Michael Hausser – Dendritic computation

Michael Häusser

I am especially excited about this one. The work of Michael Häusser and his lab is quite influential in my PhD. Let’s see how this goes …

It turned out I wasn’t quite able to blog as he talked. One reason was the fact that the talk was really interesting and my brain couldn’t really follow it and write at the same time. Second reason is my loud laptop fan, which was driving everybody else crazy.

So here you have the post-mortem abridged version, full of my subjective views and links to what I am generally thinking of doing during the course of my PhD.

Computation in the single neuron

According to Michael Häusser they are focusing on computation in single neurons. Mapping the computation on different structural parts of single neurons (spines, spine cluster, branchlet, dendritic regions) – what are the computational compartments in neurons.

Let’s take a step back. Our brain takes care of pretty much everything we think and do. To do so, it is constantly computing. How is it doing all this computation? It is widely believed this is mostly achieved by the activities of neurons. We have about 100 billions of them inside our brain. The huge computational power comes from the huge number of those neurons and mostly their connections. So this network of neurons can do all sort of clever computations, as demonstrated by a number of connectionist models. Most of those models simply use neurons as a passive point, which sums up all the inputs from the neurons projecting to it (afferent neurons) and sends the output to the neurons to which it projects (efferent, normally not the same as the afferent ones). Now we know there is more to it.

What the lab of Michael Häusser (and partly my PhD as well) is trying to show is the computation of a single neuron. Throwing away the connections and network effects, you can still find some (be it limited) computations performed. The question is not really weather this happens or not, but how much does it happen, is it robust, is it relevant or can it be neglected, which types of neurons do what (there are many different types) and so on.

Let’s see what they have got going so far …

The work of Tiago Branco et al.

This talk was focusing practically inclusively on the work of post-doc in the lab, namely Tiago Branco [1]:

Using a clever technology called two-photon glutamate uncaging, they find that single dendrites of cortical pyramidal neurons exhibit sensitivity to the sequence of synaptic activation.

This means that they perform several stimulations (mimicking the real synaptic input) on a single neuron on different locations in a short time period and are observing the voltage response in the soma of that neuron. They find that the response is different, if you change the sequence of the stimulation locations. This is most obvious when your sequence stimulation is located on the same dendrite section. To obtain the greatest difference in the response, you can then look at the sequence of stimulus going towards  the soma vs. the sequence of stimulus going away from the soma. Former sequence elicits a greater response then the later.

They do this in the real neurons (in vivo) as well as in simulations. They check this for various neuron types and find it everywhere but in the purkinje cells. They also illustrates this effect is present due to the NMDA receptors. Also they show the difference in this effect in the proximal dendrites (close to the soma) vs. the distal dendrites (away from the soma). The difference suggests that for the inputs in the distal dendrites, the number of inputs is more predominant factor and thus suggests the rate coding, while for proximal distance the timings are more important, thus suggesting the time coding. Check the coding strategies in the brain section.

Where do I fit in?

Honestly the talk gave me a bit of a scare. I was aware of their interests – as mentioned, they are a bit of an inspiration for my own investigations as well. Also I knew and read the Branco paper [1] well before the talk. Still I was surprised to see they are quite further then this already. Besides the quite elegant and technically challenging in vivo experiments, they have worked a lot on the simulations and did quite some things I was planning to look into.

In my case, I will 95% surely not be doing any experimental work during my PhD, so I am focusing on the simulations. I am collaborating with the lab of Clemens Boucsein in Freiburg (more specifically with Philipp and Mihael). There they have a neat experimental setup called Dynamic Photo Stimulation [2]. The benefit of this process is, that you can elicit true synaptic inputs to the patched neuron (as opposed to the mimicked ones of Branco). The setback is that you do not have the information about where on the dendritic this synaptic inputs are elicited.

Now it’s time to figure out a systematic way of assessing the computational capabilities of single neurons given the tools we have. I think it will be hard fun.

[1] Branco T, Clark BA, Häusser M. Dendritic discrimination of temporal input sequences in cortical neurons. Science 2010 Sep;329(5999):1671-1675.

[2] Boucsein C, Nawrot M, Rotter S, Aertsen A, Heck D. Controlling Synaptic Input Patterns In Vitro by Dynamic Photo Stimulation. J Neurophysiol 2005 Oct;94(4):2948-2958.