How the brain produces a rhythm discovered almost 100 years ago

Share this post:

In the May issue of Communications Biology features a new paper that was done in collaboration with Prof. Miles Whittington of Hull York Medical School (U.K.) and his group. In this new paper, we describe the likely cellular mechanism of the oldest known EEG (electroencephalogram) rhythm: the alpha rhythm, at around 10 cycles per second.  This work was supported in part by the IBM Research Exploratory Science Program, which sheds lights upon fundamental scientific questions underpinning the disruptive innovations and technological shifts which can help transform our company and the world.

Historical background.  Nerve cells in the brain produce electric currents that flow through the fluid and tissues that surround the cells.  If there is sufficient coherence amongst different neurons, especially neurons in the cerebral cortex, electrical potentials are thereby produced which can be measured at the surface of the head: the electroencephalogram, or EEG.  The potentials are typically in the tens of microvolt range.  The first convincing measurements of the EEG in humans were obtained by Hans Berger, in Germany, 19291.  The most striking signals that he recorded, the alpha rhythm, were best observable over the back of the head, corresponding to brain regions devoted to vision.  He and others noted that the alpha rhythm occurred when subjects were awake, but with eyes closed.  Later on, many other EEG rhythms were described, whose frequency and location depend on “brain state” (for example, whether the subject is asleep, and how deeply) and on brain pathology – most especially epileptic seizures and related phenomena.

Earlier hypothesis. Brain rhythms are presumed to occur because they serve some useful function.  But what function or functions?  In order to address this question, several approaches are necessary.  First, one must know the behavioral correlates of each rhythm.  For example, some rhythms occur during sleep; others occur prior to a volitional movement.  Second, one must know how neurons collectively generate the rhythm.  The second approach is key because knowing the mechanism allows investigators to then manipulate the rhythm, say with drugs, and to look for associations between rhythmic patterns and genes that encode for proteins – ion channels and synaptic receptors, for example – that are essential to the rhythm.  In this way, one might be able to show that degradation of a particular rhythm leads to failure in performing, say, a memory task – implying that the rhythm in question is functionally important for that task.

Up to now, a leading hypothesis has been that the alpha rhythm is generated in the thalamus, a large structure lying centrally in the brain, that provides major inputs to the cerebral cortex and receives in turn major feedback from the cortex.  An important monograph arguing this idea was that of P. Andersen and S.A. Andersson (1968) Physiological Basis of the Alpha Rhythm.

Surprisingly, and contrary to current wisdom, the new research instead indicates that the alpha rhythm is generated within visual areas of the cerebral cortex itself, by a little-studied population of neurons: small “pyramidal” cells that lie in the middle layer of the cortex (layer 4).

Background on brain oscillations. Brain oscillations are omnipresent.  The details of generation are, however, extremely variable, depending on brain region, even cortical layers, on the sleep-wake state, on sensory input, movement preparation, attention, and many other variables.  Although oscillations are presumed to have value for the behaving organism, the cellular mechanisms are usually best addressed in reduced preparations, where the experimenter can record from different identifiable cell types, at high temporal resolution, and apply drugs that block specific types of membrane channels or synaptic receptors.  Many (not all) brain oscillations are of a general pattern like this: groups of principal (excitatory) neurons discharge together, at least approximately, and then go silent for a period; the period is usually determined by the time course of synaptic inhibition.  As synaptic inhibition can have different time courses, depending on the properties of the relevant inhibitory receptors, oscillations can occur at a range of different frequencies.

The newly described alpha rhythm is different.  Synaptic inhibition is present, but it does not determine the period of the oscillation.  Instead, the period is determined by the properties of synaptic excitation between layer 4 pyramidal neurons.  As for inhibition, synaptic excitation can also occur on different time scales – something that may have importance for the design of artificial neural networks.  At present, the “neurons” in artificial neural networks operate on a single time scale, in the sense that they accept weighted inputs, then immediately generate an output.  In contrast, biological neurons process their inputs over a range of time scales.

The traces show the voltages produced by single neurons (layer 4 pyramidal cells) in the biological preparation and in the computer model: acting alone, in response to an experimental current (upper trace); as the cell participates in the alpha rhythm (middle trace); and overlay of multiple alpha periods (lower trace), to show the complex spike patterns (lower trace). [From Figure 5 of the paper. Time calibrations: 100 ms upper and middle, 10 ms lower; voltage calibration 10 mV.]

Need for an experimental model of alpha. Prof. Whittington succeeded in producing a reduced preparation of rat visual cortex that generated an alpha rhythm having properties similar to the human alpha rhythm, but that was – at the same time – experimentally tractable in terms of ability to record fluctuating membrane potentials of many types of neurons; and having the ability to allow experimental blockade of multiple types of ion channels and synaptic receptors.  He did this by applying a drug to the tissue that excited it, producing fast so-called gamma oscillations, this corresponding to visual stimulation in situ (the latter also induces gamma oscillations).  He then blocked one of the two main types of synaptic excitation with a drug, the fast type, and he blocked a specific ion channel.  These manipulations correspond to shutting off visual input in situ.  The result was a stable alpha rhythm, that lasted many minutes.

Need for a computational model.  Understanding the collective behavior of neurons requires computational modeling.  For the proper study of brain oscillations, the necessary models include details of the intrinsic properties of the different types of neurons themselves, as well as the properties of the synaptic interconnections – taking into account that there is a variety of different sorts of connection, operating over a range of time scales.  This part of the work was done at IBM Research.  A good model allows one to consider whether known properties of the cells and the connections can account for collective behavior.  It should also make non-obvious predictions.  In the present case, the model predicted the shape of potentials in the dendrites of the pyramidal cells – parts of the cells which receive most of the synaptic inputs, but which also have their own special membrane properties.  In order for the model to be useful in this way, several requirements were necessary.  First, the individual model neurons had to incorporate sufficient biological detail.  This was accomplished by having each model neuron contain a cell body, portions of the axon, and a dendritic arbor.  In addition, each compartment of a model neuron contained biophysical machinery that was capable of generating the different sorts of action potentials that real neurons can produce, as well as other mechanisms that determine the detailed shape and interval between action potentials.  Specifically, it was necessary to incorporate the ability to produce fast action potentials dependent on Na+ currents and also slower ones that depend on Ca2+ currents.  Second, it was also critical to interconnect the neurons with sufficient biological accuracy.  That meant, in particular, that it was necessary to include a multiplicity of synaptic receptor types, each with its own kinetics.  Only by including both of these features could the model predict the time-course of the dendritic electrical potentials, as well as the oscillation period2.

Basic mechanisms. The experimental and computational models indicate that the alpha rhythm is produced by a combination of interesting features: first, the ability of dendrites to act as more than passive recipients of synaptic inputs; second, the fact that there is a prominent type of synaptic excitation between layer 4 pyramidal neurons that involves “NMDA” receptors; these produce a slow, and voltage-dependent, type of excitation.  NMDA receptors have long been known to be key in development of the nervous system, in learning and memory, and in epilepsy.  It is novel to find them at the center of an experimental alpha rhythm.

Interestingly, NMDA receptors are accessible with drugs that can be given in vivo, such as ketamine, and have also been proposed to be important in schizophrenia3.

Functional significance. The new work suggests a function for the alpha rhythm, a long-standing mystery.  That is, that the alpha rhythm disconnects visual input to the cortex from most of the pyramidal (principal excitatory) neurons.  It does this not by inhibiting those cortical neurons that receive said visual input; but rather by causing the recipient neurons to oscillate at a different frequency than the other neurons, thus producing a functional disconnection.  The ability of brain regions to connect with each other via oscillations is an idea with a long history and is associated with important figures in Neuroscience such as Walter Freeman and Wolf Singer4, 5.  The ability of brain regions to disconnect via disparate oscillations has a more recent history and has been discussed previously by, for example, Miles Whittington and Nancy Kopell of Boston University.  Both abilities are likely to reflect a fundamental feature of a large brain, which of course lacks a central clock that coordinates all neuronal operations.  There are vast opportunities for further investigation of these concepts through study of oscillations in additional brain regions6.

One message of this research is that the brain remains full of mysteries.  Phenomena that are well-known, even taken for granted, may involve mechanisms quite unexpected. The practical consequences are impossible to predict a priori.

Inventing What’s Next.

Stay up to date with the latest announcements, research, and events from IBM Research through our newsletter.


More Science stories

We’ve moved! The IBM Research blog has a new home

In an effort better integrate the IBM Research blog with the IBM Research web experience, we have migrated to a new landing page:

Continue reading

Pushing the boundaries of human-AI interaction at IUI 2021

At the 2021 virtual edition of the ACM International Conference on Intelligent User Interfaces (IUI), researchers at IBM will present five full papers, two workshop papers, and two demos.

Continue reading

From HPC Consortium’s success to National Strategic Computing Reserve

Founded in March 2020 just as the pandemic’s wave was starting to wash over the world, the Consortium has brought together 43 members with supercomputing resources. Private and public enterprises, academia, government and technology companies, many of whom are typically rivals. “It is simply unprecedented,” said Dario Gil, Senior Vice President and Director of IBM Research, one of the founding organizations. “The outcomes we’ve achieved, the lessons we’ve learned, and the next steps we have to pursue are all the result of the collective efforts of these Consortium’s community.” The next step? Creating the National Strategic Computing Reserve to help the world be better prepared for future global emergencies.

Continue reading