Let the neuron do the work!

Our understanding of how the brain processes and transmits information is still sketchy at best. One thing is clear however: neurons do not act in isolation. Rather, they need to join forces and emit their action potentials, or spikes, in a coordinated fashion to be effective. How exactly this is done in the brain remains elusive though. Neurons could operate either in small groups or in huge populations, their coordinated action could be very precise in time or rather sloppy. Advanced methods of data analysis can help to gain new insight into these and related questions, and it is indispensable to develop and apply new such tools that are both effective and reliable.

A common method to analyse data in order to hunt down such coordinated spike events – often dubbed correlations – is to look at spike recordings from multiple single neurons and to see what kind of statistical dependencies show up in their signals. This approach, however, has its downsides, many of them related to the technical difficulties one has to master when listening to a large number of identified neurons simultaneously. These constraints currently impose serious limits to scientific progress in this field.

Imke Reimer and her colleagues at the Bernstein Center Freiburg now present a new method in the Journal of Computational Neuroscience that evaluates the intracellular membrane potential of just one neuron, and in this way infers higher-order correlations among the neurons that send their input to this neuron using chemical synapses. More specifically, the scientists paid attention to the so-called subthreshold fluctuations. These changes within the electrical potential of a neuron are caused by input from other nerve cells, but do not cause the cell to elicit an electrical impulse itself. To achieve the generation of such a pulse, the fluctuations have to depolarise the cell enough to reach a threshold before it fires a spike.

The method of recording and analysing the correlations among several neurons using the recording of just one individual cell has two very important advantages: Very large populations of hundreds or even thousands of neurons can be studied, and “spike sorting” becomes obsolete. Spike sorting describes the labour-intensive and error-prone process of unequivocally attributing the impulses recorded from many cells within a piece of brain tissue to individual neurons. Besides rendering spike-sorting unnecessary, this new approach truly emphasizes the functional aspects of higher-order statistics, as it infers exactly those correlations that are “seen” by a neuron.

By simulating the whole procedure on a computer, Reimer and her colleagues tested the new method of using a neuron as an “antenna” that senses correlations in its input activity. Their analysis showed that this method is surprisingly sensitive to weak higher-order correlations, and that only relatively short stretches of membrane potential are required for their reliable inference. Equally important is their finding that this approach is remarkably robust and works reliably also in real-life conditions. It is for these reasons that the scientists from Freiburg are optimistic that, using this new tool, they will now be able to make a fresh attempt to address the above-mentioned questions of neuronal coordination, this time based on experiments in which the fluctuations of membrane potentials are recorded in real nerve cells.

Original article:

Imke Reimer, Benjamin Staude, Clemens Boucsein and Stefan Rotter (2013) A new method to infer higher-order spike correlations from membrane potentials. J Comput Neurosci. [Epub ahead of print]
 

Image:

Relating subthreshold membrane potential fluctuations of a neuron and properties of its input spike trains: A neuron receives action potentials which have been elicited by many presynaptic neurons (raster display at top). While this activity is not observed, intracellular recordings allow to analyse the resulting subthreshold membrane potential fluctuations of a neuron (bottom). This signal contains information about the input population, and in particular the correlated spikes there give rise to salient deflections in the membrane potential. Analysis of this signal, therefore, may reveal higher-order spike correlations in the input population (arrow). The membrane potential trace has been recorded intracellularly from a pyramidal neuron in the primary visual cortex of a rat.