Local memory allocation recruits memory ensembles across brain regions

Abstract

Memories are thought to be stored in ensembles of neurons across multiple brain regions. However, whether and how these ensembles are coordinated at the time of learning remains largely unknown. Here, we combined CREB-mediated memory allocation with transsynaptic retrograde tracing to demonstrate that the allocation of aversive memories to a group of neurons in one brain region directly affects the allocation of interconnected neurons in upstream brain regions in a behavioral- and brain region-specific manner in mice. Our analysis suggests that this cross-regional recruitment of presynaptic neurons is initiated by downstream memory neurons through a retrograde mechanism. Together with statistical modeling, our results indicate that in addition to the anterograde flow of information between brain regions, the establishment of interconnected, brain-wide memory traces relies on a retrograde mechanism that coordinates memory ensembles at the time of learning.

Keywords: CREB; auditory fear conditioning; conditioned taste aversion; cross-regional recruitment; memory allocation; memory coordination; memory ensemble; rabies; retrograde mechanism.

Figure 1.. Memory allocation in BLA during CTA specifically recruits presynaptic neurons in IC.

(A) The CRANE system. RG-CREB and TVA viruses target Rabies-mCherry virus to memory allocated neurons. (B) Experimental design. RG-CREB (orange) or RG-CFP (Control) were injected to BLA before CTA. RG-CREB and Rabies-mCherry (magenta) in BLA (left) and in presynaptic IC neurons expressing c-Fos following CTA retrieval (middle). Right: magnified yellow square, arrow indicates co-expression with c-Fos (cyan). (C) Both groups learn CTA with no differences during habituation (‘Hab’), acquisition (‘CTA’) and retrieval (2w-RM ANOVA). (D) Higher density (mm⁻²) of rabies c-Fos-positive neurons in the gustatory IC (‘Gustatory Ctx.’) of the CREB group, but not the Control group. (E) Presynaptic IC neurons projecting to BLA RG-CREB neurons are 2.5 times more likely to co-express c-Fos in comparison to presynaptic IC neurons projecting to BLA RG-CFP neurons. Top: Rabies-mCherry (magenta) and c-Fos (cyan) in IC neurons projecting to RG-CREB BLA neurons or BLA CFP neurons. Middle: Higher activation of rabies-positive neurons in the CREB group. Bottom-Left: Higher c-Fos expression in IC rabies-positive neurons in the CREB group (t-test,). Bottom-Right: c-Fos expression in rabies-positive IC neurons, normalized to chance levels per mouse. CREB group showed increase over chance levels (Fisher’s exact test). (F) Task-specific c-Fos expression in rabies-positive IC neurons following AFC. No difference in c-Fos expression between the CREB, Control (t-test) and also from chance levels (Fisher’s exact test). Scale bars, 100μm (B, 20μm inset). **P<0.01, ****P<0.0001. Mean ± SEM.

Figure 2.. Recruitment of presynaptic memory ensembles is bidirectional and task-specific between cortical and sub-cortical regions.

(A) Experimental design. RG-CREB (orange) or control RG-CFP were injected in the IC. Rabies-mCherry (magenta) and c-Fos (cyan) in presynaptic BLA neurons projecting to RG-CREB/Control neurons in the IC. Middle-right: magnified yellow square, arrow indicates co-expression with c-Fos. Right: same CTA level for CREB/Control (n=7–8). (B) Rabies-mCherry BLA co-localized with c-Fos. CREB group shows more co-localization between rabies and c-Fos positive cells in the BLA (left, white arrows). Presynaptic BLA neurons projecting to RG-CREB neurons in the IC are 3 times more likely to co-express c-Fos in comparison to Control (middle, n=5–6, t-test) and to chance (right, Fisher’s exact test). (C) AFC experimental design. (D) Same freezing levels following AFC for CREB/Control. (E) More activated rabies-positive neurons (expressing c-Fos, white arrowheads) in anterior auditory cortex of the CREB group (n=7–8; t-test), and also above chance levels (Fisher’s exact test). (F) Neuroanatomic-based analyses of activated rabies-positive neurons in subregions of the auditory cortex showed more activation (c-Fos) in the temporal association area (TEa) of CREB group (t-test) and to chance levels (Fisher’s exact test). (G) In the CREB group, more activated rabies-positive neurons in the anterior TEa in comparison to posterior TEa and primary auditory cortex (AUDp). Scale bars, 100μm (20μm inset). T-test for (B,D-F). One-way (G) or 2w-RM ANOVA (A), ****P<0.0001, ***P<0.001, **P<0.01, *P<0.05, Mean± SEM.

Figure 3.. Expression of CREB in the BLA and IC improves CTA memory by enhancing connection probability between BLA and IC memory neurons.

(A) In the CREB group, vCREB was injected in the IC (green) in addition to RG-CREB (orange) in the BLA (Control = no vCREB in IC). Three weeks later, Rabies-mCherry (magenta) was infused into the BLA. Middle-right: IC neurons co-expressing Rabies-mCherry and vCREB in the IC. Right: yellow square magnified. (B) Within the CREB group, BLA-projecting vCREB IC neurons are more likely to be recruited into memory ensembles compared to IC vCREB neurons that do not project to the BLA. Yellow square shows co-expression of vCREB, Rabies-mCherry, and c-Fos (cyan). IC vCREB rabies-positive neurons (EGFP/mCherry/c-Fos-positive) more activated than IC vCREB neurons that do not project to the BLA (EGFP/c-Fos positive, mCherry-negative). (C) Simultaneous CREB expression in BLA (RG-CREB) and IC (vCREB) enhanced memory in comparison to the Control group (EGFP in IC; n=7–8). (D) Increased connectivity between vCREB and RG-CREB neurons across brain regions. Comparison of BLA-projecting IC vCREB neurons (rabies/EGFP-positive). More rabies-positive vCREB IC neurons (mm^–2) in the CREB group, in comparison to Control group (t-test) and to chance levels (Fisher’s exact test, ****P<0.0001). Scale bars, 100 μm (20 μm insets A,B). T-test, *P<0.05, Mean± SEM.

Figure 4.. Retrograde model underlying cross-regional coordination of memory allocation.

(A) Statistical model for cross-regional coordination of memory allocation. (B) Comparison of four models’ predictions (black dashed line) to experimental results (dashed and dotted red lines indicate median, Q1, and Q3). Predicted activation of IC neurons in the Retrograde model (2.22%) fits experimental findings from RG-CREB-projecting IC memory ensembles (3%). Predictions from the Random co-activation model (chance activation of connected neurons in both regions), the Anterograde model (IC memory neurons choose BLA memory neurons) and the Mixed model (IC memory neuron formed by monosynaptic bidirectional connectivity with BLA memory neurons) underestimate observed experimental findings. (C) A new model for memory ensemble recruitment across brain regions.