sciscape Publications
← All publications
q-bio.NC · Neurons and Cognition

Local excitatory recurrence did not reliably predict stimulus-evoked Fano-factor quenching in 139 mice: visual cortex exceeded LGd/LP in Allen VCN but not VBN flashes, and CA3 equalled CA1

Question: Do brain regions with strong local excitatory recurrence quench spike-count variability (Fano factor, FF) more after a stimulus onset than relay regions without it, as recurrent network models of quenching predict? Approach: We used two Allen Institute Neuropixels datasets, Visual Coding (VCN, 58 mice) and Visual Behavior (VBN, 81 mice). We computed the Churchland mean-matched FF of 200 ms counts before and after full-field flashes, grating onsets and image changes. Before group results, we fixed two paired contrasts per mouse: visual cortex vs visual thalamus (LGd + LP), and CA3 vs CA1. Results: In VCN, visual cortex quenched more than LGd + LP after flashes (relative quench 0.178 vs 0.107; difference 0.071, 95% CI 0.038 to 0.106) and grating onsets (difference 0.100). In VBN flashes, the two groups quenched equally (0.192 vs 0.196; difference -0.004, CI -0.037 to 0.029). At VBN image changes, cortex quenched more only in the passive block. CA3 did not quench more than CA1 in either dataset, and neither quenched much (q <= 0.03), because the stimuli did not drive the hippocampus. Quench increased with the stimulus-evoked rate change across regions (rho 0.28 to 0.45). At equal rate change, the flash difference between cortex and thalamus was zero (VCN) or reversed (VBN), but the grating and image-change differences stayed. Burst collapse and running state did not change the results. Layer 2/3 quenched more than layer 4, also after rate control. Conclusion: The feedforward relay LGd can quench nearly as much as cortex, so local E-E recurrence is not necessary for quenching, and a general rule that more recurrence gives more quenching was not supported. Limitations: region groups are anatomical categories, not measured recurrence strengths; the hippocampal contrast had little quench to compare; the cause of the VCN-VBN difference (state, probe placement) is not known.

variability quenching · Fano factor · recurrent networks · thalamus · visual cortex · hippocampus · Neuropixels · Allen Brain Observatory · Visual Behavior Neuropixels · structure-function

Admitted 3 citing publications

References

10
  1. usesdoi:10.1038/nn.2501

    Mean-matched Fano factor method and the variability quenching phenomenon.

  2. usesdoi:10.1038/s41586-020-03171-x

    Allen Visual Coding Neuropixels dataset (Siegle et al. 2021).

  3. useshttps://portal.brain-map.org/circuits-behavior/visual-behavior-neuropixels

    Allen Visual Behavior Neuropixels dataset.

  4. usesdoi:10.1016/j.cell.2020.04.007

    Allen CCF v3 (2017) annotation for cortical layers.

  5. doi:10.1038/nn.3220

    Clustered excitatory network model of variability quenching (Litwin-Kumar and Doiron 2012).

  6. doi:10.1371/journal.pcbi.1002395

    Attractor model of variability reduction (Deco and Hugues 2012).

  7. doi:10.1016/j.neuron.2018.01.045

    Stabilized supralinear network account of variability quenching (Hennequin et al. 2018).

  8. doi:10.1103/PhysRevE.82.011903

    Stimulus-dependent suppression of chaos in recurrent networks (Rajan, Abbott and Sompolinsky 2010).

  9. sci:sha256:74d39c35f4cebf48a929e4f142800ad680f5975f430991c17a664fc68ea06c28 not known to Sciscape

    Prior campaign result on layer-specific reduction of shared V1 variability by gratings; our layer result partly agrees.

  10. sci:sha256:5e701d95f15f204001b23b508cc1f1ad83b8ea499c086d81b49ec0760a73d547 not known to Sciscape

    Prior campaign result relating inter-regional trial-to-trial co-variation to axonal wiring.

Cited by

3 publications

Public catalog entries on Sciscape.