A little over a month ago we took a peak at a 2004 paper from Dr. Gerald Edelman’s laboratory that updated Dr. Edelman’s group selection theory by including axon conductance delays and spike timing dependent plasticity (STDP) in a massive computer model of the cerebral cortex containing 100,000 neurons and 8.5 million synaptic connections. The first author of that paper, Dr. Eugene Izhikevich, published a new paper in 2006 (“Polychronization: Computation with Spikes” published February 2006 in Neural Computation) that focused on the remarkable properties that emerged from the addition of axon conductance delays and STDP in a highly simplified model of cerebral cortex containing 1,000 neurons. This is probably the key paper describing the author’s results and ideas surrounding what he calls “polychronization” (poly means many and chronous means time) or the spontaneous formation of neuronal groups defined as “small collectives of neurons having strong connections with matching conduction delays and exhibiting time-locked but not necessarily synchronous spiking activity” (they may fire at many different times).
The sparse network (0.1 probability of connection between any two neurons) of 1000 randomly connected spiking neurons included axon conductance delays and STDP. This relatively simple network composed of 80% excitatory and 20% inhibitory neurons displayed dynamics similar to those seen in the mammalian cerebral cortex including 4 Hertz delta oscillations, 40 Hertz gamma oscillations, and balanced excitation and inhibition. An important finding reported in this paper is that the number of coexisting polychronous groups may far exceed the number of neurons in the network. In other words, these highly dynamic and spontaneously formed groups have the potential to carry a huge amount of information.
In a recent report (“Reversing EphB2 depletion rescues cognitive functions in Alzheimer model” published January 6, 2011 in Nature) the authors hypothesized that EphB2 depletion in Alzheimer’s disease is caused by amyloid beta oligomers and that reductions in EphB2 contribute to amyloid beta induced deficits in synaptic plasticity and cognitive functions. By carrying out experiments using a wide range of techniques, from manipulating molecules and genes to testing cognitive abilities, the team was able to confirm their hypothesis and they were even able to reverse the learning and memory effects of EphB2 depletion.
Increasing EphB2 in the hippocampus reversed the NMDA receptor mediated decrease in synaptic strength and it reversed learning and memory deficits as measured by both spatial and non-spatial learning and memory tasks.
The experiments were done in mice genetically manipulated to have Alzheimer’s disease like symptoms. It’ll be important to show that similar mechanisms are at work in the human. Also, a recent article attributed changes in NMDA receptor function to tau (see blog post “Synaptic Dysfunction without Neurodegeneration“). Are these separate and unrelated mechanisms or could they be connected somehow?
They used 27 three-dimensional (3D) digital reconstructions of rat CA1pyramidal cells from four different laboratories, which are publicly available from NeuroMorpho.org. The reference (URL) to these data in the paper and in the ModelDB record is out of date. I wondered how I’d figure out which neurons were used in this study until I finally noticed that the cell identifier numbers used by NeuroMorpho.org were listed in figure 1 under the bars representing input resistance values. However, only 26 neurons were listed there. However, as I began looking at the records on NeuroMorpho.org I noticed the following. The listing 9068802 does not exist in the repository but 9068802a and 9068802b do exist (which would account for missing neuron number 27). This raises the question of the input resistance values for these two neurons. Neuron 8228804 is not in repository but 8228804a does exist. Kr1 and Kr2 do not exist but there is an NM1 and an NM2. I couldn’t find c8076 but there is a c8076e.
Note: Those interested may find the complete list of reconstructed neurons used in this study in the table at the end of this post.
This points out a huge issue that is super important for the Semantic Web. Things on the Semantic Web, like neurons, each have their unique identifier that is technically a Uniform Resource Identifier (URI) but often in practice is a Uniform Resource Locator (URL), which is a kind of URI. URIs and URLs used to identify things must remain valid across time to be useful. A published account of neuron n125 is compromised if its identifier http://neuromorpho.org/neuroMorpho/neuron_info.jsp?neuron_name=n125 (the way NeuroMorpho.org identifies their cells) is no longer valid. How do you find it? Ways to solve the problem have been proposed and even put into practice but the Semantic Web community hasn’t come to a consensus yet.
The study demonstrated that both back and forward propagation of action potentials in the dendrites can be effectively and independently modulated in individual dendritic branches of the same neuron. They found that the distribution of A-type potassium channels (KA) appears to be pivotal in gating back propagation whereas local morphology plays the lead role in regulating forward propagation. They found that spikes back propagating from the soma continuously decreased in size in the apical trunk but tended only to invade branches off the trunk in an all-or-none fashion. They also found that isolated spiking activity in a dendritic branch usually had negligible effects on the rest of the neuron.
Note: NEURON runs under most computing environments. Details on setup and trouble shooting vary by platform but are well documented at the NEURON website.
Note: Don’t forget to compile the files in the project folder. On the Macintosh computer you drag the project folder to NEURON’s mknrndll program icon.
Load the model’s fig2A.hoc file.
The NEURON application should be running and displaying some windows. One window should contain an image of the hippocampusCA1pyramidal neuron that you’ll use in the simulation (neuron 5038804). Another window titled “Neuron” shows two empty graphs and a button. The top graph will show a trace of the somatic membrane potential. The bottom graph will show the distribution of the average peak depolarization of each dendritic branch in the neuron.
Click on the “runm” button to run the model.
The result should look like Figure 1 above and the paper’s figure 2a.
Don’t forget to play around with these models. You have first rate scientific tools at your disposal to investigate one of the biggest questions we may ask about ourselves. How does our brain work?
Note: Would you like me to take you deeper into NEURON? Please let me know!
The following table identifies all 27 neurons used in the study described in the paper under review. I’ve done the best I can to identify the cells (see text above) but there could be errors. Each neuron is identified by its NeuroMorpho.org number and is linked to its record in the repository.
.”