pubmed.ncbi.nlm.nih.gov

Controlling Visually Guided Behavior by Holographic Recalling of Cortical Ensembles - PubMed

  • ️Tue Jan 01 2019

Controlling Visually Guided Behavior by Holographic Recalling of Cortical Ensembles

Luis Carrillo-Reid et al. Cell. 2019.

Abstract

Neurons in cortical circuits are often coactivated as ensembles, yet it is unclear whether ensembles play a functional role in behavior. Some ensemble neurons have pattern completion properties, triggering the entire ensemble when activated. Using two-photon holographic optogenetics in mouse primary visual cortex, we tested whether recalling ensembles by activating pattern completion neurons alters behavioral performance in a visual task. Disruption of behaviorally relevant ensembles by activation of non-selective neurons decreased performance, whereas activation of only two pattern completion neurons from behaviorally relevant ensembles improved performance, by reliably recalling the whole ensemble. Also, inappropriate behavioral choices were evoked by the mistaken activation of behaviorally relevant ensembles. Finally, in absence of visual stimuli, optogenetic activation of two pattern completion neurons could trigger behaviorally relevant ensembles and correct behavioral responses. Our results demonstrate a causal role of neuronal ensembles in a visually guided behavior and suggest that ensembles implement internal representations of perceptual states.

Keywords: Learned behavior manipulation; Neuronal ensembles; Pattern Completion; Two-photon imaging; Two-photon optogenetics.

Copyright © 2019 Elsevier Inc. All rights reserved.

PubMed Disclaimer

Conflict of interest statement

DECLARATION OF INTERESTS

R.Y. is listed as an inventor for the patent (USA Patent 9846313) “Devices, apparatus and method for providing photostimulation and imaging of structures.” The other authors declare no competing financial interests.

Figures

Figure 1.
Figure 1.. Visually Guided Go/No-Go Task

(A) Experimental design: simultaneous two-photon calcium imaging and two-photon holographic optogenetic manipulation of targeted neurons in visually guided Go/No-Go task. (B) Performance assessment. (C) Improvement in performance as a function of training session (n = 9 mice). (D) Performance increased in expert mice because of increased hits and reduced false choices (**p < 0.005; n = 9 mice; Wilcoxon matched-pairs signed rank test). (E) Licking behavior of a representative mouse showing that enhancement of behavioral performance was reflected as shorter licking delays. Colored bars: visual stimuli (Go: green; No-Go: blue; expert: day 10). Dark markers represent licks. (F) Tuning of behavioral performance by lowering the contrast of visual stimuli in expert animals (**p < 0.005; n = 7 mice; Mann-Whitney test).

Figure 2.
Figure 2.. Identification of Neuronal Ensembles and Pattern Completion Neurons

(A) Population vectors representing neuronal activity at different time points. (B) Cartoon of population vectors in a multidimensional space. Each dot represents one population vector and clusters of population vectors define a neuronal ensemble. The normalized inner product compares population vectors by the cosine of the angle between any pair of vectors in a multidimensional space. (C) Graphical representation of CRF models. Circles represent neurons. Visual stimulus is represented by an added node (square). Shaded nodes (x) represent observed data. White nodes (y) represent neurons from the graphical model. Edges indicate the mutual probabilistic dependencies between neurons. Node potentials indicate whether a neuron is active or inactive. Edge potentials represent states of adjacent neurons. (D) Identification of pattern completion neurons defined by predictability values computed as the area under the curve (AUC) from the ROC curve and node strengths (top right neurons). Red indicates neurons with pattern completion capability. Dotted lines indicate cutoff values from random models. (E) Neurons with pattern completion capability that co-express GCaMP6s and C1V1 were simultaneously photo-stimulated using an SLM.

Figure 3.
Figure 3.. Reliable Activation of Neuronal Ensembles by Go Stimulus

(A) PCA of population vectors from a representative mouse evoked by visual stimuli show that co-active groups of neurons responding to the “Go” signal (green) define a cluster of vectors that differs from those activated by the “No-Go” signal (No-Go: blue). Each dot represents a population vector (n = 463 population vectors). (B) Sorted similarity map from a representative mouse showing lack of overlap between population vectors from Go and No-Go ensembles. (C) Cosine similarity between population vectors related to Go and No-Go visual stimuli from different mice. Population vectors from Go and No-Go ensembles are different (p < 0.001; n = 7 mice). (D) Top: time course of ensembles identified with SVD (green: Go; blue: No-Go). Shown in the middle are raster plot of neurons belonging to No-Go ensembles (blue) and Go ensemble (green). Note variability in individual responses. Shown at the bottom is a histogram of activity from all recorded neurons. No-Go trials have reduced network activity. (“x” denotes missed trials for Go stimuli; “o” denotes false choice trials for No-Go stimuli). (E) Spatial maps of same data showing subsets of neurons belong to Go and No-Go ensembles. Scale bar, 50 μm. (F) Calcium transients from neurons belonging to Go ensemble. Shown on the right are calcium transients of Go ensemble cells aligned to the Go visual stimuli (black line represents the mean value from trials shown on the left). Scale bars, 10%; 1 s. (G) Mean value of calcium transients from neurons belonging to the Go ensemble from the representative mouse depicted in the figure (onset is defined by first time derivative > 2.5 SD of noise level; mean onset ± SEM: 461 ± 121 ms; black line represents the mean value from all neurons). Scale bars, 10%; 1 s. (H) Reliability of Go ensembles is higher than that of No-Go ensembles (p < 0.005; n = 7 mice). (I) Number of co-active neurons in different mice is reduced during No-Go stimuli (p < 0.001; n = 7 mice). (J) Higher fluorescence responses to visual stimuli from neurons that belong to the Go ensemble compared with neurons that belong to No-Go ensembles in expert mice (p < 0.005; n = 7 mice). (K) Number of neurons from Go and No-Go ensembles is similar (p > 0.05; n = 7 mice). Mann-Whitney test.

Figure 4.
Figure 4.. Holographic Two-Photon Optogenetics of Cortical Neurons

(A) Calcium transients from neurons co-expressing C1V1 and GCaMP6s. Red shadows show reliably responsive neurons when one or multiple cells were targeted using an SLM. Note complete lack of cross activation. Scale bars, 10 s and 50% change in fluorescence. (B) Spatial map of targeted neurons co-expressing C1V1 and GCaMP6s (red dots). Scale bar, 50 μm. (C) Changes in fluorescence evoked in targeted neurons as a function of the number of simultaneously photo-stimulated neurons showing that there are not significant changes in calcium transients as a function of photo-stimulated neurons (p > 0.1; ANOVA test; n = 14 targeted neurons). (D) Overall spontaneous activity of non-targeted neurons as a function of the number of simultaneously photo-stimulated neurons showing that spontaneous events in non-targeted neurons are not affected by photo-stimulation of targeted neurons (p > 0.1; ANOVA test; n = 16 non-targeted neurons).

Figure 5.
Figure 5.. Unspecific Neuronal Activation Degrades Ensemble Identity and Visual Performance

(A) Experimental protocol. Go ensemble (green); No-Go ensemble (blue). During the Disrupt condition, unspecific sets of neurons (red) (including some neurons from No-Go ensemble) are simultaneously photo-stimulated during Go stimulus presentation. (B) Disruption of Go ensemble identity by stimulation of Disrupt neurons. PCA of population vectors evoked by “Go” stimulus alone and with concomitant photo-activation of disrupt neurons, which generates a different cortical response (red). Each dot represents a population vector. (C) Similarity maps of population vectors. Go (green line) versus Disrupt (red line). (D) Similarity between population vectors showing that Go and Disrupt ensembles are significantly different (p < 0.005; n = 6 mice). (E) Top: Neuronal ensemble analysis shows an artificial neuronal ensemble (red) evoked by targeted activation of Disrupt neurons. Shown in the middle are raster plots of neurons belonging to Go and Disrupt ensembles. Shown on the bottom is a histogram of network activity from all the neurons (“x” denotes miss trials for Go visual stimuli). (F) Spatial map of neurons from Go (green) and Disrupt ensembles (red). Optogenetic targeting included neurons belonging to No-Go ensemble (blue). Scale bar, 50 μm. (G) Calcium transients from neurons belonging to the Go ensemble during visual stimuli and Disrupt stimuli (left). Responses of Go ensemble neurons evoked by visual stimuli decreased (right) (p < 0.005; n = 6 mice). (H) Disrupt ensemble is composed mainly of optogenetically targeted neurons (p < 0.005; n = 6 mice). (I) Reliability of Go ensemble during disruption is significantly decreased during disrupt protocol (p < 0.005; n = 6 mice). Green dotted line: Go ensemble reliability in control. (J) Cross-correlation of Go ensemble neurons is significantly reduced by Disrupt protocol (p < 0.005; n = 6 mice). Mann-Whitney test. (K) Behavioral performance is significantly decreased during Disrupt protocol (p < 0.05; n = 6 mice). (L) Licking onset is significantly increased by Disrupt protocol (p < 0.05; n = 6 mice). Wilcoxon matched-pairs signed rank test.

Figure 6.
Figure 6.. Activation of Pattern Completion Neurons Reliably Enhances Go Ensemble and Task Performance

(A) Experimental protocol. Go stimulus neurons (green) in low contrast conditions and simultaneously targeted pattern completion neurons (red) that belonged to the Go ensemble, with same low contrast stimulus (Recall condition). (B) PCA of population vectors evoked by the low contrast Go stimulus (green) and concomitant low contrast Go visual stimulation and activation of pattern completion neurons (red). Each dot represents a population vector. (C) Similarity maps of population vectors representing the Go ensemble in low contrast condition alone and with simultaneous holographic photo-stimulation. (D) Recall condition increases Go ensemble reliability (p < 0.005; n = 6 mice). (E) Raster plot from neurons belonging to the Go ensemble shows change in overall activity evoked by the simultaneous activation of two neurons with pattern completion capability (neurons 5 and 21; red bars). Note that the reliability of individual neuronal responses is increased (“x” denotes missed trials for the Go stimuli; “o” denotes false choice trials for the No-Go stimuli). (F) Spatial map of layer 2/3 neurons highlighting neurons belonging to the Go ensemble (green). SLM photo-stimulated pattern completion neurons in red. Scale bar, 50 μm. (G) Number of active neurons evoked by visual stimuli in low contrast and during concomitant photo-stimulation of pattern completion neurons (p.c. cells) and non-pattern completion neurons (non-p.c. cells) in hit and miss trials. Red shadow indicates photo-stimulation epochs (Hit trials: Go low versus Go p.c. cells: p < 0.005; Go low versus Go non-p.c. cells: p > 0.05; Miss trials: Go low versus Go p.c. cells: p < 0.0001; Go low versus Go non-p.c. cells: p > 0.05; n = 6 mice). (H) Behavioral response to low contrast Go-Signal is significantly enhanced by the targeted activation of pattern completion neurons (p < 0.05; n = 6 mice). (I) Increased behavioral performance is due to increased hits and reduced false choices (p < 0.05; n = 6 mice). (J) Reliability of Go and No-Go ensembles during visual stimulation and SLM photostimulation of pattern completion neurons and non-pattern completion neurons belonging to the Go ensemble. Left: reliability of recalled Go ensemble is significantly increased from Go ensemble in low contrast conditions (p < 0.005; n = 6 mice). Right: the reliability of No-Go ensemble remains unaltered (p > 0.05; n = 6 mice). Green and blue dotted lines represent the mean values from Go and No-Go ensemble reliability in control conditions respectively. (K) Cross-correlation of neurons belonging to the Go ensemble increased by SLM targeting of neurons with pattern completion capability but not by the targeting of non-pattern completion neurons (p < 0.005; n = 6 mice). (L) The mean value of the licking onset was not significantly reduced under Recall conditions (p > 0.05; n = 6 mice). Mann-Whitney test (D, G, J, and K). Wilcoxon matched-pairs signed rank test (H, I, and L).

Figure 7.
Figure 7.. Behavior Induced by Recalling Go Ensemble in Absence of Visual Stimuli

(A) Experimental conditions. Simultaneous optogenetic stimulation of pattern completion neurons in the absence of visual stimuli (animals viewed a gray screen) and behavioral cues. (B) Behavioral performance evoked by recalling the Go ensemble by optogenetic stimulation in the absence of visual stimuli was significantly higher than performance in partially recalled trials (p < 0.005; n = 6 mice). (C) Licking onset from successfully driven optogenetic behavioral events was not significantly different from licking onset in visual evoked behavior (p > 0.05; n = 6 mice). (D) Paired cross-correlation of Go ensemble neurons was enhanced during successful recall, compared with partial recall trials (p < 0.005; n = 6 mice). (E) Raster plot of most representative neurons from Go ensemble during holographic stimulation of two pattern completion neurons. Vertical red lines indicate photo-stimulation. Horizontal lines highlight targeted neurons. Red marker shows successful recalling of Go ensemble and licking behavior. Black marker shows an example of partial recall. (F) Spatial map of E showing stimulated and recalled neurons during successful licking trial. (G) Number of recalled neurons after optogenetic activation in the absence of visual stimuli and behavioral cues was not significantly different from active neurons evoked by visual stimuli (dotted line; p > 0.05; n = 6 mice). Mann-Whitney test. See also Figure S1.

Similar articles

Cited by

References

    1. Abeles M (1991). Corticonics (Cambridge, England: Cambridge University Press; ).
    1. Afraz SR, Kiani R, and Esteky H (2006). Microstimulation of inferotemporal cortex influences face categorization. Nature 442, 692–695. - PubMed
    1. Bartlett JR, and Doty RW (1980). An exploration of the ability of macaques to detect microstimulation of striate cortex. Acta Neurobiol. Exp. (Warsz.) 40, 713–727. - PubMed
    1. Brecht M, Schneider M, Sakmann B, and Margrie TW (2004). Whisker movements evoked by stimulation of single pyramidal cells in rat motor cortex. Nature 427, 704–710. - PubMed
    1. Brown SL, Joseph J, and Stopfer M (2005). Encoding a temporally structured stimulus with a temporally structured neural representation. Nat. Neurosci 8, 1568–1576. - PubMed

Publication types

MeSH terms

Substances