Correlated variability modifies working memory fidelity in primate prefrontal neuronal ensembles

TY - JOUR

T1 - Correlated variability modifies working memory fidelity in primate prefrontal neuronal ensembles

AU - Leavitt, Matthew

AU - Pieper, Florian

AU - Sachs, Adam J.

AU - Martinez-Trujillo, Julio C.

PY - 2017/3/8

Y1 - 2017/3/8

N2 - Neurons in the primate lateral prefrontal cortex (LPFC) encodeworking memory (WM) representations via sustained firing, aphenomenon hypothesized to arise from recurrent dynamics withinensembles of interconnected neurons. Here, we tested this hypothesisby using microelectrode arrays to examine spike count correlations(rsc) in LPFC neuronal ensembles during a spatial WM task.We found a pattern of pairwise rsc during WM maintenance indicativeof stronger coupling between similarly tuned neurons and increasedinhibition between dissimilarly tuned neurons. We thenused a linear decoder to quantify the effects of the high-dimensionalrsc structure on information coding in the neuronal ensembles.We found that the rsc structure could facilitate or impaircoding, depending on the size of the ensemble and tuning propertiesof its constituent neurons. A simple optimization proceduredemonstrated that near-maximum decoding performance could beachieved using a relatively small number of neurons. These WMoptimizedsubensembles were more signal correlation (rsignal)-diverse and anatomically dispersed than predicted by the statisticsof the full recorded population of neurons, and they often containedneurons that were poorly WM-selective, yet enhanced codingfidelity by shaping the ensemble’s rsc structure. We observed apattern of rsc between LPFC neurons indicative of recurrent dynamicsas a mechanism forWM-related activity and that the rsc structurecan increase the fidelity ofWMrepresentations. Thus,WMcoding inLPFC neuronal ensembles arises from a complex synergy betweensingle neuron coding properties and multidimensional, ensemblelevelphenomena.

AB - Neurons in the primate lateral prefrontal cortex (LPFC) encodeworking memory (WM) representations via sustained firing, aphenomenon hypothesized to arise from recurrent dynamics withinensembles of interconnected neurons. Here, we tested this hypothesisby using microelectrode arrays to examine spike count correlations(rsc) in LPFC neuronal ensembles during a spatial WM task.We found a pattern of pairwise rsc during WM maintenance indicativeof stronger coupling between similarly tuned neurons and increasedinhibition between dissimilarly tuned neurons. We thenused a linear decoder to quantify the effects of the high-dimensionalrsc structure on information coding in the neuronal ensembles.We found that the rsc structure could facilitate or impaircoding, depending on the size of the ensemble and tuning propertiesof its constituent neurons. A simple optimization proceduredemonstrated that near-maximum decoding performance could beachieved using a relatively small number of neurons. These WMoptimizedsubensembles were more signal correlation (rsignal)-diverse and anatomically dispersed than predicted by the statisticsof the full recorded population of neurons, and they often containedneurons that were poorly WM-selective, yet enhanced codingfidelity by shaping the ensemble’s rsc structure. We observed apattern of rsc between LPFC neurons indicative of recurrent dynamicsas a mechanism forWM-related activity and that the rsc structurecan increase the fidelity ofWMrepresentations. Thus,WMcoding inLPFC neuronal ensembles arises from a complex synergy betweensingle neuron coding properties and multidimensional, ensemblelevelphenomena.

U2 - https://doi.org/10.1073/pnas.1619949114

DO - https://doi.org/10.1073/pnas.1619949114

M3 - SCORING: Journal article

C2 - 28275096

VL - 114

SP - E2494 - E2503

JO - P NATL ACAD SCI USA

JF - P NATL ACAD SCI USA

SN - 0027-8424

IS - 12

ER -