Dopamine release and dopamine-related gene expression in the amygdala are modulated by the gastrin-releasing peptide in opposite directions during stress-enhanced fear learning and extinction

Generation of the Grp^-/- mice and mapping of the GRPergic neural circuitry

We developed the Grp^-/- mice in which the majority of exon 1 of the Grp gene was deleted and replaced with the green fluorescent protein (GFP) (Fig. 1A; Supplementary Fig. 1A–C). The Grp^-/- mice of both sexes develop normally and show no gross abnormalities throughout the body, including the brain (Supplementary Fig. 1D). Staining with the anti-GFP antibody showed that the GFP is highly enriched in the lateral (LA) and basomedial (BMA), but not basal (BA), nuclei of the amygdala as well as in the ventral CA1 area of the hippocampus and mPFC (Fig. 1B–E). In addition, we observed GFP expression in the retrosplenial cortex, dorsal subiculum, auditory cortex, and the entorhinal/perirhinal cortex, similar to the endogenous Grp expression previously examined by RNA in situ hybridization [38, 45]. Therefore, expression of the GFP knocked-in in exon 1 of the Grp gene is very similar to the endogenous GRP expression.

To map the GRPergic neurons onto the amygdala-associated neural circuitry, we injected retrograde neuronal tracer rAAV2-retro-CaMKII-tdTomato (rAAV2) [46] in several brain regions in Grp^-/- mice. When rAAV2 was injected in the TE3 area, the retrograde labeling co-localized with the GRP-positive cells in the MGm/PIN area of the auditory thalamus (Fig. 1F). When rAAV2 was injected in the lateral nucleus of the amygdala (LA), the areas labeled with tdTomato were the TE3 area of the auditory cortex (Fig. 1G) and the MGm/PIN (Supplementary Fig. 2), two major regions sending projections to the LA [3, 47]. However, only the TE3 area showed co-localization of the tdTomato and GFP (Fig. 1G); the MGm/PIN area had no GRP-positive cells co-labeled with the tdTomato (Supplementary Fig. 2B). The results of the injections into TE3 and LA show that GRP-positive cells project to the LA via the indirect MGm/PIN- > TE3-LA pathway but not via the direct MGm/PIN- > LA pathway.

When rAAV2 was injected in the basal nucleus of the amygdala (BA), the co-localization of tdTomato and GFP was observed in the ventral hippocampus (vHP; Fig. 1H), the major region projecting to the BA [48].

To examine the BLA-mPFC projections, we injected rAAV2 in the mPFC. Interestingly, most of the BLA projections originated from the BA area, which by itself is GRP-negative but is located exactly between the two major GRP-positive amygdala areas, the LA and the basomedial nucleus of the amygdala (BMA). There was no co-localization between GRP-positive cells and those labeled with tdTomato in the BA area or any other areas of the amygdala projecting to the mPFC (Fig. 1I). Thus, the BLA GRPergic neurons do not project to the mPFC. The BA however receives the major dopaminergic projections from th`e ventral tegmental area (VTA) neurons, some of which are GRP-positive (see below the section about the GRPergic cells in the VTA). Quantification of these connections showed a higher density of GRP-positive cells in the TE3- > LA projections followed by the MGm/PIN - > TE3 and vHP->BLA projections (Fig. 1J). These results show that the GRP expression maps a very specific neural circuitry entering the amygdala and may play a unique role in processing of fear-related sensory information (Fig. 1K).

The Grp^-/- mice show an enhancement in cued and contextual long-term fear memory

To characterize memory of the Grp^-/- mice, we examined them in the standard protocol of fear conditioning. Post-shock freezing during training was not different between genotypes (Supplementary Fig. 3A). Grp^-/- mice showed stronger long-term memory (LTM) in both cued and contextual fear (Fig. 2A). We also analyzed Grp^-/- mice for short-term memory (STM) testing them 1 hour after training using independent groups of Grp^-/- mice. There was no significant difference between wildtype and Grp^-/- mice in both cued and contextual STM (Fig. 2B and Supplementary Fig. 3B, C). Thus, the enhancement in memory observed in Grp^-/- mice is specific to long-term, but not short-term, fear memory, demonstrating the specificity of the GRP/GRPR signaling pathway for long-term memory related to fear [38].

Next, we crossed the Grp^-/- mice (GRP KO) with GRPR KO mice and assessed GRP/GRPR double KO mice in fear conditioning. The double KO mice showed enhanced long-term contextual and cued fear memory assessed 24 h after fear conditioning (Fig. 2C and Supplementary Fig. 3D). The fact that the double KO animals have stronger fear memory than the Grp^-/- mice suggests that the GRP and GRPR may have other, perhaps secondary, ligands/receptors that they might bind, such as other members of the bombesin family [49].

To rule out possible developmental effects of removing GRP/GRPR system early in life and determine the specific effects of GPR signaling in the amygdala, we selectively ablated interneurons expressing GRPR using saporin-based cell elimination in the BLA of wildtype mice at the adult stage (Fig. 2D) [50]. Mice injected with bombesin-saporin (bombesin is a frog peptide related to the GRP with high affinity to the GRPR [51]) showed a ~ 50% reduction in GRPR-expressing cells and had significantly higher freezing compared to mice injected with control blank-saporin in cued fear conditioning. Therefore, potential developmental changes are not likely to be responsible for the enhanced fear memory phenotype in GRP or GRPR KO mice.

Enhanced amygdala activity of the Grp^-/- mice following fear conditioning

Based on our observations that GPR-related knockouts enhance fear memory, we hypothesized that the absence of GRP or GRPR may elevate neural activity in the amygdala. Expression of immediate-early genes (IEG) is activity-dependent and often used as an indicator of neural activity. Quantitative real-time PCR (qPCR) revealed that RNA expression of IEG c-Fos and Arc was increased in the BLA of Grp^-/- mice compared to their wildtype counterparts 30 min after single-paired cued fear conditioning (Supplementary Fig. 4). Activity of c-Fos and Arc was normal in naïve Grp^-/- mice. These results suggest that the neural activity in the amygdala of Grp^-/- mice is enhanced after fear conditioning, which is consistent with the enhanced memory Grp^-/- mice display in fear conditioning.

Grp^-/- mice show normal anxiety and pain sensitivity

We examined anxiety levels of the Grp^-/- mice using the elevated plus maze (EPM), open field (OF) and light-dark (LD) box (Supplementary Fig. 5A–C). The statistical analysis showed that there was no significant difference between the genotypes in all three tests. To verify that the increase in freezing displayed by the Grp^-/- mice in fear conditioning was not due to an increased sensitivity to the shock, we examined their pain sensitivity by movement, vocalization and jump (Supplementary Fig. 5D). There was no difference between genotypes in the intensity of the shock required to elicit these three behaviors. We interpret these findings to indicate that the increase in freezing observed in fear conditioning is due to differences in memory, but not in anxiety or pain sensitivity, confirming the behavioral phenotype of GRPR KO mice [38].

Grp^-/- mice exhibit increased susceptibility to stress-enhanced fear learning (SEFL)

To examine a possible role of the GRP in linking stress and fear, we examined the Grp^-/- mice in a behavioral task where mild stress precedes fear conditioning and fear extinction, using SEFL (Fig. 2E) [52]. Each genotype had two groups of mice: one group (SEFL) was exposed to 2 h of restraint stress, while another group – designated only for fear learning (FL) – was kept naïve in a home cage. Seven days later all mouse groups were trained in fear conditioning (Cued FC; Fig. 2E, F) using two tone-shock pairings. Four days after fear conditioning, all groups underwent two days of fear extinction, and recall memory was tested after additional 24 days. A repeated-measures three-way ANOVA revealed the significant main effect of the genotype on the freezing during extinction (Fig. 2G). In the recall test, a significant interaction between the effect of genotype and stress on freezing was observed in the SEFL group. Freezing during recall was higher in the KO-SEFL group compared to other groups (p = 0.0196 vs KO-FL; p = 0.0003 vs WT-SEFL; p = 0.001 vs WT-FL, Tukey post-hoc test). We calculated the percent of decrease in freezing by subtracting freezing in bin 5 from freezing in bin 1 for each extinction session and found that the extinction rate was similar between the genotype or stress groups in Extinction 1 or 2 (Supplementary Fig. 6 A, B). We also calculated the percent of increase in freezing during Recall by subtracting the freezing during the last bin of Extinction 2 from freezing during Recall and found no significant differences in the increase of freezing between the groups (Supplementary Fig. 6C).

Following fear conditioning, the stressed group was separated into two subgroups, resilient and susceptible, based on their freezing performance during one minute of post-shock freezing (Fig. 2F), as this measure is used as an index of stress susceptibility in the SEFL protocol [52]. Animals that froze above the mean percent freezing in the stressed group were classified as stress-susceptible (SS), while those that fell below the mean were classified as stress-resilient (SR). The ratio of the number of mice in the susceptible group to the resilient group was higher in the Grp^-/- mice compared to wildtype mice. There was also a significant difference in freezing during extinction and recall between the resilient and susceptible subgroups in wildtype mice (Fig. 2H). However, there was no significant difference between the knockout stress-resilient group and knockout stress-susceptible group in extinction and recall (Fig. 2I). A comparison between Genotype and Susceptibility factors revealed that the KO mice, regardless of being stress-resilient or stress-susceptible, showed higher freezing compared to the WT mice (statistical differences are in Genotype but not in the interaction of Genotype X Susceptibility; Supplementary Fig. 7). The results suggest that the deletion of the Grp gene confers susceptibility to stress combined with conditioned fear.

Downregulation of the dopamine-signaling pathway in the amygdala of the Grp^-/- mice following SEFL recall

We used qPCR to examine transcription of several genes known to be induced following SEFL recall [52], PTSD or stress [10, 53]. We found that several genes involved in the dopamine signaling were downregulated in the basolateral amygdala of the Grp^-/- mice (KO-SEFL) compared to control mice following SEFL memory recall (Fig. 3A and Supplementary Fig. 8). One of these genes encodes for the tyrosine hydroxylase (TH). There was a significant interaction between the effect of genotype and stress on a decrease in expression of the Th mRNA as shown by two-way ANOVA (Fig. 3A and Table S1). To verify that the observed Th mRNA was trafficked from the VTA cell nuclei to the BLA synapses to be translated into TH protein there, we examined TH protein expression. Western analysis showed similar levels of the TH protein between Grp^-/- and wildtype mice in naïve conditions in the hippocampus (Supplementary Fig. 9A). Thus, Th mRNA and TH protein expression in the BLA and hippocampus confirm previous evidence that the Th mRNA is trafficked and locally translated at synapses [54, 55].

Other dopamine-related genes found in our qPCR analysis include nuclear receptor related 1 protein (NURR1) and dopamine receptor D1 (DRD1). The Drd1 mRNA was also decreased in the Grp^-/- mice that received fear extinction but no stress (FL group). A two-way ANOVA followed by post-hoc test also revealed a significant acute stress-dependent decrease of Nurr1 mRNA expression in the Grp^-/- mice and GRP knockout-dependent decrease of Drd1 mRNA expression (Fig. 3A and Table S1). qPCR analysis revealed no difference in mRNA expression of the Th, Nurr1, Drd1 and Drd2 genes in the VTA and BLA of naïve wildtype and Grp^-/- mice (Supplementary Fig. 9B, C). These results suggest the possibility that the changes in expression of dopamine-related genes found after recall were induced in the Grp^-/- mice by long-term memory retrieval during SEFL.

Co-expression of the GRP and tyrosine hydroxylase in the ventral tegmental area

To investigate the relationship between the GRPergic and dopaminergic neurons, we performed co-immunostaining for the TH and Grp-promoter-driven GFP on brain sections of the Grp^-/- mice. TH immunostaining showed expression in the fibers in the BLA and in cell bodies in the VTA, confirming previous work showing that the TH is expressed in the VTA, which sends projections to the amygdala (Fig. 3B, C) [56]. In the VTA, we found that the GFP and TH overlap. Combined, our results suggest that a certain proportion of the VTA neurons, are positive for both dopamine and GRP, and that TH- and GRP-positive fibers are present in the BLA (Fig. 3D).

Enhanced fear is supported by differences in dopamine signaling

To determine whether changes in dopaminergic signaling might contribute to the modulation of fear memories, we recorded dopamine responses during SEFL training using the fluorescent dopamine sensor dLight1.2 (Fig. 4A, B). We observed that both the Grp^-/- and wildtype mice showed increased dLight responses to the tone and shock during fear learning (Main effect of event: F(2,49) = 19.47, p < 0.0001; Fig. 4C), that carried forward to the tone during extinction (Fig. 4D). The dLight responses to the shock were greater in both the Grp^-/- and wildtype mice (Grp^-/-: p = 0.009; WT: p = 0.016). However, responses to the tone were only greater in the Grp^-/- mice when compared to baseline (p = 0.03). Additionally, dLight responses to the shock were greater in the Grp^-/- mice during fear learning than those observed in wildtype mice (Group x Event interaction: F(2, 48) = 4.60, p < 0.05; Grp^-/- vs. WT Shock: p = 0.0007; Fig. 4E), and subsequently greater to the tone during the first extinction session (Main effect of Group: F(1, 16) = 6.054, p = 0.026; Fig. 4F). Moreover, the magnitude of the dLight response to the tone during extinction was directly related to the magnitude of the dLight response to the shock during learning, but not to that of the tone during fear learning (Fig. 4G). Analysis of the behavior during photometry recordings showed that post-shock freezing was not different between genotypes, and that the ratio of the susceptible group to the resilient group was higher in the Grp^-/- mice compared to wildtype mice (Supplementary Fig. 10A, B). Freezing during extinction was not different between genotypes, most likely because of changes in mobility due to the implanted optic fibers (Supplementary Fig. 10C). When we included calculations of within-subject changes, the increased signal observed in the Grp^-/- mice during training and extinction was observed as well (Supplementary Fig. 10D, E). Thus, the observed differences in dopamine responses during SEFL are driven by genotype (GRP KO vs. WT) differences. From these results, we infer that dopamine signaling is important for regulating the initial response to a noxious stimulus, and that amplified responses to the cue and stimulus during training may be critical for determining subsequent susceptibility to learned fear cues.

Occluded presynaptic VTA-BLA connectivity in the Grp^-/- mice

To examine functional connectivity between the VTA and BLA, we injected the VTA of the Grp^-/- and WT mice with the AAV-hSyn-ChR2-EGFP (Fig. 5A, B) and examined the properties of spontaneous EPSPs (sEPSPs) and IPSPs (sIPSPs) in BLA neurons during optogenetic stimulation of VTA-BLA projecting axons. Light stimulation increased the frequency of sEPSPs in the WT mice (Fig. 5C, D). Notably, the frequency of sEPSPs observed in Grp^-/- mice at baseline conditions (without light stimulation) matched that of the WT mice under light stimulation and did not significantly change under light stimulation (Fig. 5C, D). We found similar results with the number of sEPSPs (Supplementary Fig. 11A). In contrast, sEPSP amplitude did not significantly change in WT or Grp^-/- mice under light stimulation, nor sEPSP amplitude was different between the groups (Fig. 5E, F). Similarly, the median sEPSP frequency in WT or Grp^-/- mice remained unchanged with light stimulation, although the Grp^-/- mice showed significantly lower median sEPSP frequency than the WT mice at baseline conditions (Supplementary Fig. 11B). Light stimulation did not significantly change sIPSP frequency in the WT mice, although the Grp^-/- mice showed a significantly heightened frequency of sIPSPs at baseline compared to WT neurons at baseline, which did not significantly change with light stimulation (Fig. 5G, H). Light stimulation did not significantly affect the amplitude of sIPSPs in WT or Grp^-/- mice, nor was sIPSP amplitude significantly different between the groups (Fig. 5I, J). Similar results were found with the analysis of the number of sIPSPs (Supplementary Fig. 11C) and median sIPSP frequency (Supplementary Fig. 11D). In addition to the optogenetic studies, we found a 4 mV more depolarized resting membrane potential and increased spike activity in the Grp^-/- compared to WT mice (Supplementary Fig. 11E, F). Our data, showing strong dopaminergic VTA-BLA projections (Fig. 3B, D), confirm previous research that indicates that the majority ( > 95%) of the VTA-BLA projections are dopaminergic [56–58]. Therefore, our results, showing occluded presynaptic VTA-BLA connectivity in the Grp^-/- mice, demonstrate a possible role of dysregulated dopaminergic input from the VTA on BLA function in the absence of the GRP.

RNA-sequencing confirms and extends the qPCR data on the number of genes in the dopamine-signaling pathway with decreased expression in the Grp^-/- mice following SEFL recall

To characterize the effects of the Grp gene deletion on genome-wide patterns of mRNA expression, we generated RNA-seq datasets following SEFL recall (we used the same time point of SEFL memory recall as in our qPCR analysis above, but we used a separate group of animals) and performed differential gene expression analysis using DESeq2 (Fig. 6). Consistent with our qPCR results, we found that the Th has lower abundance (in transcripts per million (TPM)) in the KO-SEFL compared to all the other samples (Fig. 6A) in our RNA-seq datasets. We also found Nurr1 RNA abundance is lower in SEFL relative to FL in both the wildtype and knockout mice, validating our qPCR results (Fig. 6A). These differences are significant only for the knockout mice likely due to the fact that Nurr1 abundance in WT-FL group is highly variable. We also find that Drd1 abundance is significantly lower in KO-SEFL relative to WT-SEFL consistent with our qPCR analysis (Fig. 6A). Some additional dopamine-related genes showed differences in RNA-seq that we did not see in our qPCR analysis: we find that Drd2 and Grik2 abundances are significantly lower in KO-SEFL relative to WT-SEFL (Supplementary Fig. 12A). Interestingly, we also find abundance is significantly lower in KO-SEFL relative to WT-SEFL for the Ppm1f gene, which encodes protein phosphatase, Mg²⁺/Mn²⁺ dependent 1 F (PPM1F). The Ppm1f gene was found to be regulated by stress in mice as well as associated with anxiety and PTSD in humans [53, 59]. While the average abundances of Ppm1f in the GRP KO mice across both conditions are lower than in the wildtype control mice, these differences are small and not significant due to high biological variability in mRNA abundances across individual samples in the wildtype mice.

GRP removal primes dopaminergic system for stress-induced plasticity

We find that in the wildtype mice, SEFL (WT-SEFL) led to no significant changes in gene-expression patterns genome-wide relative to non-stressed fear-conditioned mice that underwent fear extinction (WT-FL; Supplementary Fig. 12B). Similarly, under non-stress conditions, the combination of Grp gene deletion and fear conditioning (KO-FL) failed to elicit large changes in gene-expression patterns relative to the wildtype control mice that underwent fear conditioning (WT-FL). However, the combination of stress and fear conditioning (SEFL treatment) drives significant changes in gene-expression patterns in KO-SEFL relative to both WT-SEFL wildtype and KO-FL. Together this indicates that neither the stress alone nor the Grp gene deletion by itself has large effects, but that these two conditions can (and perhaps must) act synergistically. More specifically, when comparing KO-SEFL to KO-FL mice (Fig. 6B), we find a modest number of genes with altered expression levels (452, q-value < 0.05). Similarly, when comparing KO-SEFL to WT-SEFL mice, we find only 336 genes significantly altered (q-value < 0.05) (Fig. 6B). Functional characterization of these differentially expressed genes suggests that genes involved in regulation of synaptic transmission are typically downregulated, while genes involved in microtubule and cilium assembly are upregulated (Fig. 6C). Genes belonging to the sperm assembly categories are microtubule-associated genes with more specific annotations. Overall, these results suggest a specific and consistent functional effect of the Grp gene deletion on synaptic regulation and microtubule-associated genes. Furthermore, many of the genes that were significantly affected were associated with dopamine signaling and recapitulated effects seen in our behavioral and qPCR analyses. Taken together, these results suggest that the GRP may link stress to fear through the modulation of the dopamine signaling.

The GRP regulates dopamine signaling in opposite directions in fear learning and extinction

We have identified and characterized the GRP as a molecular link regulating dopaminergic control of fear learning and extinction. When this molecular link is disrupted, dopamine is enhanced in the amygdala when the tone-CS or shock-US are presented during fear learning and when tone-CS is presented during extinction on day 1, while dopamine-related gene transcription is decreased following CS memory recall of fear extinction (Supplementary Fig. 13). Combined optogenetics and ex vivo electrophysiology suggest disrupted VTA-BLA connectivity in Grp^-/- mice, demonstrating an importance of the GRP for the input from the VTA to BLA. The dependence of BLA function on the VTA inputs as well as changes in dopamine signaling at the early and late stages of fear learning and extinction provide an initial insight into how the GRP may regulate dopamine function and confer susceptibility to exaggerated fear in animals and possibly PTSD in humans. Also, our study confirms recently emerging work, showing that the GRP links stress and fear processing [60, 61].

The GRP may serve as a functional biomarker of learned fear processing

Several studies show the importance of dopamine neurons in the VTA during fear extinction [22, 24]. Importantly, the Grp mRNA is a selective marker for dopaminergic subpopulations in the VTA both in mice and humans [39, 62, 63]. Moreover, our results support recent work that dopamine may play a more general role in salience signaling [64] and suggest that the GRP-dopamine link is important for the initial encoding of fear memory as well as for the long-term memory recall in fear extinction processes. Thus, if deficiencies in GRP support aberrant dopaminergic signaling and enhanced salience at the time a fear memory is acquired, it may serve as an important marker for the long-term susceptibility to protracted fear responses and mental illness.

Within the amygdala-associated neural circuitry, the GRP is released from excitatory neurons and binds to the GRP receptor expressed exclusively by GABAergic interneurons stimulating GABA release and inhibiting principal neurons [38, 45, 65–67]. The lack of the GRPR leads to an increase in long-term potentiation in the BLA, enhanced fear memory and deficits in fear extinction [38, 45]. However, within the VTA, less is known about the functional role of the GRP, besides its almost exclusive co-localization with markers of dopaminergic neurons [39, 40]. Nonetheless, one interesting possibility given the role of GRP in the BLA and the observed increase in excitatory potentials that we observed here is that GRP could be transmitted by VTA neurons that co-transmit dopamine and glutamate [63, 68], some of which also participate in salience signaling [69].

In addition to assessing the GRP-dopamine link, the Grp^-/- mice allowed for mapping the learned fear circuitry of the tone-CS-specific projections entering the amygdala from the auditory thalamus and auditory cortex. Combining visualization of the GRP-positive cells by the GFP knocked-in into the Grp gene locus in the Grp^-/- mice together with rAAV2-retro-based tracing [46], we were able to examine which of these two auditory pathways contain the GRP. The results demonstrate that the GRP labels only one of the two pathways from the auditory thalamus to the amygdala: the GRP labels the multisynaptic indirect thalamo-cortico-amygdala pathway but not the direct thalamo-amygdala pathway [3, 47]. Overall, because the GRP is selectively present in the neural circuitry regulating fear learning and extinction – the BLA, auditory thalamus, auditory cortex, VTA, ventral hippocampus and mPFC [38, 45, 70, 71] – it is a unique functional biomarker of learned fear processing.

The Grp^-/- mice may serve as a genetic model of PTSD-like symptoms

Our data suggest that the Grp^-/- mice can be used as a genetic model of PTSD symptoms. Naïve Grp^-/- mice displayed normal anxiety and pain sensitivity (Supplementary Fig. 3 and 5), consistent with normal expression of immediate-early genes (IEG; c-Fos and Arc; Supplementary Fig. 4) and dopamine-related genes (Drd1, Drd2, Th and Nurr1; Supplementary Fig. 9) in the BLA and VTA of naïve Grp^-/- mice. These behavioral and gene expression data suggest that delayed extinction observed in SEFL in Grp^-/- mice is likely to reflect changes in stress-enhanced learning and memory processes rather than in anxiety in naïve state. It appears that both learning and extinction of fear are affected by the GRP-dopamine interactions. Importantly, fear extinction is strongly dependent on initial fear learning – so, it is expected to see an effect in both processes, when a manipulation is done before fear testing (fear learning and fear extinction) is conducted (unless there is a specific experimental design that affects only extinction, for example). Moreover, examining individuals with different levels of initial fear learning and their subsequent differences in extinction is also highly translationally relevant, as these are precisely the differences that recapitulate pathological fear responses. Overall, our transgenic mouse model supports the idea that preexisting changes in genes and neural circuits may predict and cause deficits in PTSD in humans [72, 73].

Delayed extinction in the Grp^-/- mice appears to be long term, as evident in the enhanced freezing during the recall test that is performed three-four weeks following the initial stages of SEFL: acute stress, fear learning and extinction. Our current data provide genetic evidence that a decrease in the GRP signaling leads to deficiency in fear extinction, whereas an increase in the GRP signaling improves fear extinction [38, 45, 74–77]. While the SEFL paradigm clearly showed fear learning and extinction deficits in the Grp^-/- mice vs. their wildtype counterparts as well as it showed deficits in the susceptible group compared to the resilient group of wildtype mice over two days of fear extinction and during the Recall session (Fig. 2H), we did not find that stress induces fear extinction deficits in wildtype mice as a whole group, as we reported earlier [52]. It is possible that because we kept both the knockout and wildtype mice together, these two groups influenced each other, as mixed-group housing of knockout and wildtype mice was reported previously to affect behavior [78].

Recent work provides clear evidence of the importance of the VTA in fear extinction, with the VTA sending dopamine projections to other brain areas critically involved in fear extinction [79, 80]. While recent work has focused on the dopamine-related neural circuitry of fear extinction, we know very little about the molecular pathways supporting the role of dopamine in this process. Our work here begins to address this, by describing how the GRP may regulate dopamine function in the VTA-BLA circuitry. Combined optogenetics and ex vivo electrophysiology show that in the absence of the GRP, the BLA is overexcited, and the VTA-BLA connectivity is disrupted and cannot be activated under conditions we tested. Altogether these results support the notion that the GRP may have a specific role in modulating the dopamine control of fear extinction.

Dopamine release is enhanced during fear learning and early phases of extinction, but dopamine-related gene transcription is down following extinction memory recall in the Grp^-/- mice

The increased release of dopamine observed during shock presentation during initial fear conditioning and extinction on day 1 suggests that excessive release of dopamine may have led to a down regulation of dopamine-related genes to compensate for increased availability of dopamine in the Grp^-/- mice. Compensatory changes in gene expression of dopamine receptors have been reported when dopamine availability is altered [81, 82]; however, more research is needed to explore this mechanism in the VTA-BLA circuit of the Grp^-/- mice. Intriguingly, dopamine-related gene transcription was normal in naïve Grp^-/- mice, but following Recall, gene transcription of Th, Nurr1, Drd1 and Drd2 is decreased. While more experiments are required to better understand when and how the lack of the GRP may modify BLA synaptic activity and gene expression, our results demonstrate the importance of stress and fear experiences in these processes.

A decrease in the Th mRNA levels in the BLA of the Grp^-/- mice following SEFL recall suggests that the Th mRNA, a critical enzyme involved in dopamine synthesis, was located in synapses of the VTA presynaptic neurons projecting to the BLA, confirming previous work [54, 55, 83]. In support of our finding of the Drd1 mRNA, previous work showed that the D1 receptors expressed in the BLA [21, 23, 32, 34, 84–87]. Another gene we found codes for the NURR1, a transcription factor and an IEG located in the cell nucleus; thus, the changes we see in the NURR1 are in the BLA neurons. NURR1 is known to modulate dopamine signaling [88–92]. Overall, these RNA transcription data suggest that GRP’s influence on dopamine may influence the acquisition and extinction of stress-enhanced fear memories and indicate for the first time that therapies targeting GRP as a potential dopamine-oriented drug in psychotherapy approaches to maximizing consolidation of successful fear extinction and safety learning [93]. Interestingly, an additional support for the idea that the Grp^-/- mice may be a genetic model for PTSD is provided by the fact that we found the Ppm1f gene to be significantly decreased in the Grp^-/- mice following SEFL. The Ppm1f gene is regulated by stress in mice and is associated with anxiety, depression and PTSD in humans [53, 59].

A limitation in the interpretation of our data is that the Grp^-/- mice are generated using the classical gene knockout method, which means the gene is removed from every cell of the body and during early embryo development. To mitigate some of these concerns, we used the toxin-based removal of the GRPR-positive interneurons in the amygdala of the adult mice. While we were able to replicate the GRP removal effects on fear conditioning, this doesn’t fully address this concern. Thus, future works is required to better understand where and when the loss of GRP/GRPR leads to changes in the VTA-BLA synaptic connectivity and gene expression in the amygdala, and how to link these changes to the behavior.

It is important to emphasize that the molecular and behavioral changes we observe in the Grp^-/- mice suggest that the GRP may link stress and fear. An acute restraint stress elicits the release of both corticotropin-releasing factor (CRF) and GRP in the amygdala [94]. The glucocorticoid receptor gene knockout in the GRPergic neural circuitry by Inoue et al. [60] and work by Goto et al. [61] also demonstrate that the GRP is critically involved in integrating stress and fear processing [95]. Since our paper focuses on the GRP-dopamine connection, we did not study other neurotransmitters in the Grp^-/- mice. The GRP seems to regulate other molecular pathways as well, as seen in our RNA-seq data (Fig. 6B, C). Also, our electrophysiological studies suggest that glutamatergic synaptic circuits are affected downstream of the VTA-BLA projections in the Grp^-/- mice and can add to the effects described here. Overall, our results suggest that enhancing dopamine function long-term by increasing transcription of dopamine-related genes can improve extinction of undesired fear, which may have clinical relevance for PTSD. Therefore, GRP- and dopamine-transcription-based approaches can be used together with exposure-based cognitive behavior therapy in extinction-impaired individuals.