A novel channel of information transmission in the brain: correlations between the activities of medullary neurons encode behaviorally relevant stimuli in weakly electric fish
Final Report Abstract
To understand how the brain encodes and transmits information and generates behavior, we need to understand the interplay of activity in neuronal populations. This is complicated by the fact that neuronal responses are non-deterministic, i.e. when presented with the same stimulus repeatedly, responses will exhibit trial-to-trial variability. It has been widely observed that this trial-to-trial variability can be correlated between neurons (= “noise correlations”) a phenomenon with substantial impact on coding efficacy. Furthermore, it was shown that noise correlations can be plastic and change in magnitude based on various factors such as an animals general state, attention, motivation or even in relation to the parameters of a stimulus, such as its spatial extend. From this, theoretical studies have predicted that noise correlations could encode sensory information directly thereby forming a channel of information transmission independent of commonly known coding parameters such as firing rate or spike latency. In this project I investigated this possibility experimentally for pyramidal neurons in the medullary electrosensory lateral line lobe of weakly electric fish, a network with well characterized physiology and anatomy. I tested if noise correlations magnitude changes in relation to 2nd order stimulus statistics (i.e. the amplitude of an electrosensory envelope stimulus) as well as the underlying mechanisms. For this I used electrophysiological recordings alongside newly developed computational methods to simulate pyramidal cell noise correlations. I further established high-density silicon probes (neuropixels), which became commercially available during the project period, to record in weakly electric fish to circumnavigate methodological challenges arising in some of my experiments. My results confirm the presence of correlated variability between medullary pyramidal neurons with respect to the electrosensory envelope and the fact that this magnitude changes as a function of stimulus amplitude. While some of the data is still being analyzed and prepared for publication, noise correlations seem to encode the envelope amplitude through increases in noise correlations magnitude. This finding was surprising as it was contrary to my expectations based on preliminary data presented in the proposal. Importantly signal correlations and the population firing rate do not change significantly as a function of envelope amplitude and therefore cannot account for the seen noise correlation change. The above effects were observed over various timescales and with different types of stimuli (constant amplitude vs. time varying amplitude). Using experimental work and computational predictions I was able to show that correlation plasticity is mediated jointly by feedforward and feedback mechanisms. This is namely the differential activation of receptive field centers and surrounds under global stimulation and their geometrical interaction with the receptive fields of neurons in close vicinity. Further, I discovered a novel type of correlated variability that can be expected to be present under naturalistic stimulation. This “noise response similarity” was surprisingly found also between neurons that were recorded non-simultaneously (i.e. between pairs without shared neuronal input). Investigating the source of this type of correlated variability revealed that in naturalistic stimuli of nearly all modalities, 1st and 2nd order stimulus attributes are independent of one another and that responses to the 1st order stimulus attribute can from effectively be considered noise when computing responses with respect to 2nd order stimulus attributes. Noise response similarity had significant detrimental effects for stimulus encoding as shown by considering decoding of population activity. However, the neuronal heterogeneity of ELL and the splitting of stimulus encoding into separate On- and Off-type pathways partially mitigates the deleterious effects of this type of correlated variability. During my project, I was able to investigate most of the initially proposed questions as well as unforeseen aspects, such as the above.
Publications
- Differential receptive field organizations give rise to nearly identical neural correlations across three parallel sensory maps in weakly electric fish (2017) PLoS Comput Biol 13(9): e1005716
Hofmann V & Chacron MJ
(See online at https://doi.org/10.1371/journal.pcbi.1005716) - Population coding and correlated variability in electrosensory pathways (2018) Front Integr Neurosci 12:65
Hofmann V & Chacron MJ
(See online at https://doi.org/10.3389/fnint.2018.00056) - Novel functions of feedback in electrosensory processing. (2019) Front Integr Neurosci 13:52
Hofmann V & Chacron MJ
(See online at https://doi.org/10.3389/fnint.2019.00052) - Neural synchrony gives rise to amplitude- and duration invariant encoding consistent with perception of natural communication stimuli. (2020) Front Neurosci 14:79
Metzen MG, Hofmann V, Chacron MJ
(See online at https://doi.org/10.3389/fnins.2020.00079) - Neuronal On- and Off-type heterogeneities improve population coding of envelope signals in the presence of stimulus-induced noise. (2020) Sci Rep. 10:10194
Hofmann V & Chacron MJ
(See online at https://doi.org/10.1038/s41598-020-67258-1)
