Anterior Basolateral Amygdala Neurons Comprise A Remote Fear Memory Engram Part 2

Remote fear memory recall induces asub-region-specific aBLA Fos ensemble

Analysis of immunostained sections revealed that memoryrecall testing activated a significantly larger Fos + populationof aBLA neurons in fear-conditioned (128 ± 5) than context-conditioned (83 ± 3) mice [unpaired t-test, t(8)=8.027, P < 0.0001](Table 1). 

Second, immunostaining can enhance the connection between memory and emotion. Immune cells in the brain are mainly distributed in areas controlled by memory and emotion. Therefore, appropriate immunostaining can increase the connection between memory and emotion and enhance the long-term effectiveness and stability of memory.

Examination of rostral-caudal aBLA sub-region countdensities (Figures 3A–C) by two-way ANOVA revealed asignificant sub-region and treatment interaction [F(2,24)=5.402,P=0.0116] (Figure 3D). 

In the remote fear memory recall group,aBLA Fos + count density was greater in the middle (254 ± 17) thanthe rostral (104 ± 11, P < 0.0001) or caudal (158 ± 26, P=0.0004)sub-region and was greater in the caudal than rostral sub-region(P=0.0490). 

In the context memory recall group, Fos + countdensity was comparable in the middle (134 ± 24) and caudal(135 ± 11) sub-regions, yet both were greater than in the rostral(51 ± 10) sub-region (middle, P=0.0019; caudal, P=0.0018).Between-group analyses revealed Fos + count density was greateronly in the middle aBLA sub-region of the fear memory recallgroup (P < 0.0001).

Analysis of Fos + count density within quadrants by threeway ANOVA (Figure 3E) revealed quadrant [F(3,32)=6.450,P=0.0015] and treatment [F(1,32)=78.63, P < 0.0001] effectswithin the middle aBLA sub-region, with greater Fos + countdensity in the fear compared to context group in quadrant 1(289 ± 28 vs. 165 ± 12, P=0.0031) and 2 (361 ± 22 vs.167 ± 30, P < 0.0001). 

Furthermore, within-group analyses showed that fear-recall induced greater Fos + count density in quadrant1 (289 ± 28) than quadrant 3 (180 ± 37, P=0.{{20}}168), whilequadrant 2 (361 ± 222) was greater than quadrants 3 (P < 0.0001)and 4 (197 ± 20, P < 0.0001). A significant quadrant effect wasalso observed in the caudal aBLA sub-region [F(3,32)=16.21,P < 0.0001]. 

The fear-recall group had greater Fos + count densityin quadrants 2 (284 ± 59, P < 0.0001) and 4 (190 ± 32, P=0.0423)compared to quadrant 3 (70 ± 6) while the context-recall grouphad greater Fos + count density in quadrant 2 (309 ± 35) thanquadrant 1 (132 ± 6, P=P < 0.0001), 3 (57 ± 19, P < 0.0001)or 4 (148 ± 7, P < 0.0001). 

Comparison of Fos + counts densitywithin a treatment group and quadrants across aBLA rostrocaudal sub-regions (Figure 3E and Supplementary Table 2) yieldedfindings similar to those of tdTomato where quadrant 2 of themiddle (361 ± 22, P < 0.0001) and caudal (284 ± 59, P=0.0015)aBLA sub-regions each had greater Fos + count density than did therostral sub-region (109 ± 30).

We also evaluated the relationship between freezing behaviorand the size of the Fos + neuronal ensemble in the aBLA as a whole, inits middle sub-region, and quadrant 2 of the middle and caudalsub-regions. Again, no significant correlations were observed ineither the fear or context group (Figure 3F), suggesting thatfear behavior does not reflect the size of the aBLA sub-regionor quadrant-specific for ensemble activated during remote fearmemory recall.

Data in Figure 3 illustrate that remote fear memory recallrecruited a larger aBLA Fos ensemble than context memory recall.Like the fear memory formation ensemble, the highest activationdensity following fear memory recall was localized to the middlesub-region and in quadrant 2 of both the middle and caudal subregions, but again fear behavior did not correlate with the size ofthese ensemble populations.

Remote fear memory recall induces arobust sub-region-specific aBLA forreactivation ensemble

Next, we quantified neurons double-labeled with tdTomatoand Fos immunoreactivity (i.e., those both TRAPed during fearconditioning and reactivated during remote memory recall). 

Double-labeled counts in the aBLA were greater in the fear (35 ± 2)than context (17 ± 2) group [unpaired t-test, t(8)=6.505,P=0.0002]. Moreover, the reactivated population was a largerproportion of the population TRAPed during fear conditioning(30 ± 2%) than context conditioning (20 ± 3%) [unpaired t-test,t(8)=3.346, P=0.0101] (Table 1). Analysis of count density by two-way ANOVA yielded a significant interaction between the aBLA subregion and treatment [F(2,24)=5.477, P=0.0110] (Figures 4A–D). 

In the fear group, reactivation density was greater in the middle(P < {{0}}.0001) and caudal (P=0.0001) than in the rostral (13 ± 4) subregion. In the context group, a similar pattern was observed in themiddle (68 ± 5 vs. 27 ± 4, P < 0.0001) and caudal (53 ± 10 vs.27 ± 10, P=0.0098) sub-regions.

Comparing count density across quadrants of aBLA subregions and treatment (Figure 4E) by three-way ANOVA showedmain effects for both the middle [quadrants: F(3,32)=28.98,P < 0.0001; treatment: F(1,32)=49.10, P < 0.0001] and caudal[quadrants: F(3,32)=18.10 P < 0.0001; treatment: F(1,32)=10.94, P=0.0023] sub-regions. 

For the middle sub-region, remote fearmemory recall induced greater neuronal reactivation compared tocontext in quadrant 1 (94 ± 7 vs. 31 ± 6, P < 0.0001), 2 (83 ± 15 vs.38 ± 10, P=0.0072), and 4 (63 ± 9 vs. 20 ± 3, P=0.0129). Withinthe fear group, greater reactivation was also observed in quadrants 1(P < 0.0001) and 2 (P=0.0009) compared to 3. Within the caudalaBLA, greater reactivation (P=0.0100) was observed in quadrant2 of the fear group (133 ± 35) compared to the context group(72 ± 10). 

Moreover, quadrant 2 reactivation in the fear group wassignificantly greater than in quadrant 1 (34 ± 7, P < 0.0001), 3(23 ± 9, P < 0.0001), and 4 (39 ± 7, P=0.0002). Similarly, quadrant2 reactivation in the context group was greater than quadrant 3(4 ± 2, P=0.0023) and 4 (16 ± 4, P=0.0292). 

Comparisonof within-group and within-quadrant reactivated count densitybetween aBLA rostrocaudal sub-regions again yielded findingssimilar to tdTomato + and Fos + counts as quadrant 2 of the middle(83 ± 15, P < 0.0001) and caudal (133 ± 35, P < 0.0001) subregions showed greater reactivation than did the rostral sub-region(16 ± 6) (Figure 4E and Supplementary Table 2).

We next evaluated the relationship between freezing behaviorand the size of the reactivated neuronal ensemble in the aBLA asa whole, in its middle sub-region, and quadrant 2 of the middleand caudal sub-regions. 

No significant correlations were observedin either the fear or context group (Figure 4F), suggesting that fearbehavior does not reflect the size of the fos ensemble that was bothactivated during fear memory formation and reactivated duringrecall.

Figure 4 findings indicate that remote fear memory recallrecruited a larger aBLA Fo's reactivation ensemble than contextmemory recall. Like the fear memory formation and recallensembles, the highest reactivation density was again localized tothe middle sub-region and in quadrant 2 of both the middle andcaudal sub-regions, but again fear behavior did not correlate withthe size of these ensemble populations.

We used TRAP2 transgenic mice to investigate aBLA neuronalFos ensembles activated during contextual fear learning and duringremote fear memory recall. We found that larger Fos ensembleswere activated by fear learning and fear memory recall comparedto context-only controls. The population of neurons activated bothduring conditioning and again during recall (reactivated) was alsolarger in the fear group. We also observed that ensembles weredifferentially distributed in aBLA sub-regions and sub-regionalquadrants, but that topographical count distributions were notcorrelated with fear behavior.

Reports of Fos expression induced in the BLA by fear memory testing are inconsistent (Holahan and White, 2004; DoMonte et al., 2016; Liu et al., 2022). Even after habituationtraining, fear conditioning with applied electrical shocks andcontext conditioning without shocks have both been reportedto increase Fos expression compared to home cage residence (Campeau et al., 1991; Pezzone et al., 1992; Milanovic et al.,1998; Radulovic et al., 1998; Rosen et al., 1998; Day et al.,2001; Holahan and White, 2004; Cho et al., 2017), and yet Fosexpression after remote fear memory recall has been reported eitherto increase (Silva et al., 2019; Liu et al., 2022) or not increase(Do-Monte et al., 2015; Cho et al., 2017) compared to contextrecall. 

With context conditioningfollowing habituation training, again as performed in the presentstudy, Fos expressing neurons primarily represent those thatremain responsive to contextual cues, some of which may undergopotentiating plasticity to encode context-specific memory. There ispresently no universally accepted experimental design that entirelycontrols for non-memory-related Fos activation. Quantifying Fosexpression in home cage resident controls, though potentiallyaiding interpretation by allowing subtraction of "background" Fosactivity, does not control specifically for the direct effects of appliedshocks. Moreover, neurons directly responsive to delivered shocksmay not be entirely distinct from those that encode fear memory. 

Our useof adult TRAP2 x Ai14 offspring, however, appears unlikely tosolely explain quantitative differences in Fos ensemble sizes duringrecall as c-fos transcription and translation are intact in these mice(DeNardo et al., 2019; Roy et al., 2022).

Fos protein is typically quantified by immunohistochemistry(IHC). Prior IHC studies show that Fos induced in the BLA byfear-learning peaks after ∼90 min, and remains elevated for up to5 h (Chowdhury and Caroni, 2019). Despite the latter, Fos IHCresults represent a "snapshot" of Fos activity as compared to resultsobtained with the TRAP2 transgenic system, which appears tocapture neurons in which Fos transcription was induced over ∼6 or more hours surrounding 4-OHT administration(DeNardo et al., 2019). 

Therefore, tdTomato expression, as an indexof Fos induction, potentially reflects cumulative expression broughton during a more prolonged period. This raises the possibilitythat the tdTomato + TRAPed ensemble includes neurons Fosactivated by stimuli unrelated to fear/context conditioning. Asnoted, our use of habituation training was employed to strengthenthe fear-conditioning specific signal relative to context cue- or novelenvironment exploration-related "noise." 

With this experimentaldesign, differential tdTomato expression may be highly relevantto understanding fear memory processes as Fos expressionduring hours following fear conditioning is thought necessaryfor memory consolidation (Chowdhury and Caroni, 2018). The TRAP2 system, therefore, might label Fos-expressing neuronsthat comprise a single ensemble activated specifically during fearlearning or additional ensembles recruited in the hours thereafter,possibly including memory consolidation ensembles. Together,these considerations could explain the larger fos ensemble(tdTomato +) we detected during the formation of fear memory rather thancontext memory.

We utilized IHC to capture the Fos ensemble (Fos +) activatedduring remote fear memory recall, which was larger than thatcaptured following remote context recall. This finding is consistentwith other reports of BLA Fos expression during remote fearmemory recall (Silva et al., 2019; Liu et al., 2022) and parallelsprevious aBLA-specific reports describing increased c-fos mRNAexpression during recent fear memory formation/recall (Kim et al.,2016; Zhang et al., 2020).

Notably, fear memory formation and remote memoryrecall differentially activated aBLA neurons as only about one-third underwent reactivation. 

Non-overlapping aBLA neuronalpopulations was not unexpected as previous studies indicatethat a portion of BLA neurons are responsive to the sensationof the unconditioned stimulus (i.e., shocks) (Corder et al., 2019)present during fear memory formation, while another portionmay be responsive specifically to conditioned stimuli (contextcues) during memory recall (Beyeler et al., 2018). 

These variableactivation patterns within aBLA ensembles during distinct fear memory tasks (i.e., formation vs. recall) may be explained by the recruitment of distinct neural circuits resulting from plasticityinitiated during fear memory formation and subsequent networkplasticity resulting in engram migration before remote memoryrecall testing (Grewe et al., 2017; DeNardo et al., 2019). Futurein vivo electrophysiological studies are required to investigateresponses during fear memory formation and recall, comparing BLA fos ensemble (tdTomato +) neurons and non-fosensemble neurons.

Reactivation of fear-learning-activated BLA Fos ensembleneurons is documented in two previous studies where remote fearmemory recall occurred 14 (Kitamura et al., 2017) and 28 (Leeet al., 2022) days after conditioning. While our findings at 21 daysare similar, our experiments also revealed findings not previouslydescribed.

As noted, one strength of our experimental design is the incorporation of a positive-control (i.e., context cue exposure only)group. This contrasts with prior remote fear-memory investigationsthat utilized a negative-control (i.e., home-cage) group. The latter,although permitting evaluation of basal BLA for activity, canlimit interpretation by precluding comparison of the reactivationensemble size between context- and fear-conditioned groups (Leeet al., 2022). Our design enables this comparison and revealssignificantly greater reactivation during fear-memory recall thanduring re-exploration of the context-conditioned environment.

Additionally, a key finding of our experiments was the identification of aBLA sub-regional localization of the reactivated(engram) ensemble. Specifically, we found higher ensembledensities in dorsal zones (quadrants 1 and 2) of the middle aBLAsub-region and in the dorsomedial zone (quadrant 2) of the caudalsub-region. 

Prior fear memory studies have identified input andoutput projection neurons near the dorsomedial zone of BLAthat may contribute both to memory and valence processing(Kim et al., 2016; Beyeler et al., 2018). Implicated are reciprocalconnections between dorsomedial BLA and the prelimbic area of the medial prefrontal cortex (PL/PFC) (McGarry and Carter, 2017)and the CA1 region of ventral hippocampus (vCA1) (Jimenezet al., 2020; Kim and Cho, 2020). These connections appear torepresent synaptic substrates driving regional and sub-regionalFos induction amongst our reactivated ensembles. 

Another aBLAoutput worth noting is the capsular part of the central amygdala(Kim et al., 2016, 2017), which is responsive to noxious inputs andimplicated in anxiety and fear behaviors (Bourgeois et al., 2001).

Unclear is the extent of functional heterogeneity withinidentified aBLA Fos ensembles. While prior studies that focusedon vCA1 inputs to BLA (Kim and Cho, 2020; Lee et al.,2022) identified LTP-like synaptic potentiation amongst apparentlearning ensemble neurons, it is difficult to precisely comparethe location of these potentiated neurons relative to ouraBLA Fos ensemble neurons. Assessing the extent to whichplasticity amongst previously identified BLA neurons implicatedin shorter-term contextual fear memory (Kim and Cho, 2020; Leeet al., 2022) is present among sub-regional aBLA Fos ensembleneurons of the present study and assessing their contribution toremote fear memory-related behaviors will require more detailedcharacterization of their neurochemical phenotypes and anatomicalconnectivity.

Here, correlation analysis failed to reveal a relationship betweenthe size of any aBLA Fos ensemble and fear-related behavior(i.e., postural freezing), suggesting either that Fos ensemble sizealone is not the primary driver of fear behavior or that the Fosensemble is itself a representation of the memory engram and nota neuronal population directly contributing to behaviors incitedby memory recall.

Within the entire aBLA (see Table 1), larger Fos ensembles(i.e., TRAPed, Fos +, and reactivated neurons) were observed inthe fear group compared to the context group. A similar patternwas observed in the middle aBLA sub-region (see Figures 2D,4D). These size differences may represent a threshold levelof aBLA populational recruitment required for eliciting fearrelated behaviors or it could indicate that fear behaviors requirethe recruitment of additional circuits. Notably, ensemble sizedifferences between context and fear groups disappeared forsome aBLA quadrants which may suggest that fear behaviorsrequire more anatomically dispersed recruitment of aBLA neurons.Furthermore, studies show that positive valence posterior BLA(pBLA) and negative valence aBLA neurons are mutually inhibitorythrough the activation of BLA GABAergic interneurons (Kim et al.,2016; Zhang et al., 2020). 

Therefore, fear behavior might reflect the excitation of negative valence aBLA neurons that not only drivefear-specific outputs broadly throughout the limbic system but alsoindirect inhibition of positive valence neurons of the BLA. Thelatter possibility must be reconciled with evidence that individualprincipal neurons in the aBLA and pBLA respond both to negativeand positive valence stimuli (Beyeler et al., 2018), indicating thatvalence and related memory encoding in BLA may not be fullysegregated.

It should be stressed that the number of neurons expressinga detectable level of Fos protein is not a measure of ensemblefunction. What is known about the BLA fear-formation ensembleis it receives monosynaptic input from corresponding ensemblesin the PL/PFC and vCA1 region of the hippocampus (Kim and Cho,2020). At remote time points, these inputs can both drive fearlike behaviors in a non-fear-associated context (Lee et al., 2022) andrestrain fear-like behaviors in a fear-associated context (Kitamuraet al., 2017). However, since PL/PFC fear ensemble inputs do notexclusively synapse on BLA fear ensemble neurons (Lee et al.,2022), it is unclear to what extent differential fear behaviors reflectBLA fear ensemble activity. Therefore, future studies will not onlyneed to delineate if the aBLA fear-formation ensemble can influence fear-like behavior but also the capacity offear-memory activated/reactivated ensembles to do so.

Here we demonstrated that less than half of Fos ensembleneurons activated during memory formation are reactivated duringremote memory recall, and yet fear ensembles are larger thantheir context counterparts specifically in the middle sub-regionof aBLA and its dorsomedial zone more caudally. Collectively,findings suggest that the remote contextual fear memory engramincludes ensemble neurons of the aBLA with a common populationactivated during fear learning and reactivated during fear memoryrecall. The latter population may represent a critical sub-regionalaBLA substrate through which learned fear is stored for recall ata remote time point. Maladaptive plasticity amongst these andother functionally coupled neuron populations may be key tofear-associated psychiatric disorders.

Beyeler, A., Chang, C. J., Silvestre, M., Leveque, C., Namburi, P., Wildes, C. P.,et al. (2018). Organization of valence-encoding and projection-defined neurons in thebasolateral amygdala. Cell Rep. 22, 905–918. doi: 10.1016/j.celrep.2017.12.097

Bourgeais, L., Gauriau, C., and Bernard, J. F. (2001). Projections from thenociceptive area of the central nucleus of the amygdala to the forebrain: A PHAL study in the rat. Eur. J. Neurosci. 14, 229–255. doi:10.1046/j.0953-816x.2001.01640.x

Campeau, S., Hayward, M. D., Hope, B. T., Rosen, J. B., Nestler, E. J., and Davis, M.(1991). Induction of the c-fos proto-oncogene in rat amygdala during unconditionedand conditioned fear. Brain Res. 565, 349–352. doi: 10.1016/0006-8993(91)91669-r

Cho, J. H., Rendall, S. D., and Gray, J. M. (2017). Brain-wide maps of Fos expressionduring fear learning and recall. Learn. Mem. 24, 169–181. doi: 10.1101/lm.044446.116

Chowdhury, A., and Caroni, P. (2018). Time units for learning involvingmaintenance of system-wide cFos expression in neuronal assemblies. Nat. Commun.9:4122.

Chowdhury, A., and Caroni, P. (2019). Author correction: Time units for learninginvolving maintenance of system-wide cfos expression in neuronal assemblies. Nat.Commun. 10:3083. doi: 10.1038/s41467-019-11208-7

Corder, G., Ahanonu, B., Grewe, B. F., Wang, D., Schnitzer, M. J., and Scherrer,G. (2019). An amygdalar neural ensemble that encodes the unpleasantness of pain.Science 363, 276–281. doi 10.1126/science.aap8586

Day, H. E., Badiani, A., Uslaner, J. M., Oates, M. M., Vittoz, N. M., Robinson,T. E., et al. (2001). Environmental novelty differentially affects c-fos mRNA expressioninduced by amphetamine or cocaine in subregions of the bed nucleus of the striaterminalis and amygdala. J. Neurosci. 21, 732–740.

For more information:1950477648nn@gmail.com