The encoding bandwidth of the cortical gateway
Somatosensory information enters the primary sensory cortex through thalamic axons that innervate spiny stellate (SpSt) cells in cortical layer 4. We showed earlier , that the population of layer 2/3 pyramidal cells (Pyr), the main target of SpSt cells output, can encode input changes in less than a millisecond. Their large dendritic tree might significantly contribute to this ability . It is intuitively clear that the entry site for thalamic information, i.e. SpSt cells, should have an encoding bandwidth at least as wide as those of later processing stages, i.e. layer 2/3 and layer 5 Pyr cells. However, SpSt cells feature the smallest dendritic compartment of all excitatory cortical neurons, which leads to the expectation that they have a severly limited bandwidth. This discrepancy between a required wide bandwidth at the entry level and an expected severely limited bandwidth prompted us to explore whether SpSt cells employ compensatory mechanisms to retain a high bandwidth of information encoding in the absence of large dendrites.
We find that SpSt cells employ two compensatory mechanisms to obtain a superior bandwidth for information encoding. Firstly, the sub-threshold synaptic input fluctuates much slower in SpSt cells1 than in Pyr cells and that leads to an enhanced encoding (Figure 2 B, C) . This is the first report of such a mechanism employed in neural encoding. Secondly, the detrimental effects of small dendritic tree are reduced by a reduction in soma size.
With a combination of experimental and theoretical approaches, we revealed how different biophysical principles are at play to allow for ultra-fast processing of somato-sensory input during cortical processing.
Figure 1: Spiny stellate cells are the smallest excitatory cells in the cortex
Figure 2: A SpSt cells’ encoding bandwidth is only marginally smaller than those of Pyr cells. B Synaptic currents change much slower in SpSt cells. C SpSt population rate rapidly encodes input changes embedded in fluctuations. An ideal observer was faster to detect a step in the input (not shown) when the background fluctuations had a longer correlation time (blue).
1. Due to the presence of an unconventional NMDA receptor, operating already at subthreshold voltages 
 G. Eyal, et al. J. Neurosci. 34, 24 8063(2014)
 I.A. Fleidervish A.M. Binshtok, M.J. Gutnick Neuron 21 1055 (1998)
Members working within this Project:Andreas Neef