Skip to main content
NCBI home page
eLife logoLink to eLife
. 2020 Dec 15;9:e60812. doi: 10.7554/eLife.60812

A prefrontal-bed nucleus of the stria terminalis circuit limits fear to uncertain threat

Lucas R Glover 1,✉, Kerry M McFadden 1, Max Bjorni 2, Sawyer R Smith 1, Natalie G Rovero 2, Sarvar Oreizi-Esfahani 1, Takayuki Yoshida 1, Abagail F Postle 1, Mio Nonaka 1, Lindsay R Halladay 2, Andrew Holmes 1
Editors: Mihaela D Iordanova3, Kate M Wassum4
PMCID: PMC7899651  PMID: 33319747

Abstract

In many cases of trauma, the same environmental stimuli that become associated with aversive events are experienced on other occasions without adverse consequence. We examined neural circuits underlying partially reinforced fear (PRF), whereby mice received tone-shock pairings on half of conditioning trials. Tone-elicited freezing was lower after PRF conditioning than fully reinforced fear (FRF) conditioning, despite an equivalent number of tone-shock pairings. PRF preferentially activated medial prefrontal cortex (mPFC) and bed nucleus of the stria terminalis (BNST). Chemogenetic inhibition of BNST-projecting mPFC neurons increased PRF, not FRF, freezing. Multiplexing chemogenetics with in vivo neuronal recordings showed elevated infralimbic cortex (IL) neuronal activity during CS onset and freezing cessation; these neural correlates were abolished by chemogenetic mPFC→BNST inhibition. These data suggest that mPFC→BNST neurons limit fear to threats with a history of partial association with an aversive stimulus, with potential implications for understanding the neural basis of trauma-related disorders.

Research organism: Mouse

eLife digest

While walking home alone late one night, you hear footsteps behind you. Your heart starts to beat faster as you wonder whether someone might be following you. Being able to identify and evade threats is essential for survival. A key part of this process is learning to recognize signals that predict potential danger: the sound of footsteps behind you, for example. But many such cues are unreliable. The person behind you might simply be heading in the same general direction as you. And if you spend too much time and energy responding to such false alarms, you may struggle to complete other essential tasks.

To be useful, responses to cues that signal potential threats must thus be proportionate to the likelihood that danger is actually present. By studying threat detection in mice, Glover et al. have identified a brain circuit that helps ensure that this is the case. Two groups of mice learned to fear a tone that predicted the delivery of a mild footshock. In one group of animals, the tone was followed by a shock on every trial (it was said to be ‘fully reinforced’). But in the other group, the tone was followed by a shock on only 50% of trials (‘partially reinforced’).

After training, both groups of mice froze whenever they heard the tone – freezing being a typical fear response in rodents. But the animals trained with the partially reinforced tone showed less freezing than their counterparts in the fully reinforced group. Moreover, freezing in response to the partially reinforced tone was accompanied by activity in a specific neural pathway connecting the frontal part of the brain to an area called the bed nucleus of the stria terminalis. Inhibiting this pathway made mice respond to the partially reinforced tone as though it had been reinforced on every trial. This suggests that activity in this pathway helps dampen responses to unpredictable threat cues.

In people with anxiety disorders, cues that become associated with unpleasant events can trigger anxiety symptoms, even if the association is unreliable. The findings of Glover et al. suggest that reduced activity of circuits that constrain excessive responses to threats might contribute to anxiety disorders.

Introduction

In many cases of psychological trauma, encounters with contexts and stimuli during aversive experience(s) are interleaved with occasions when the same stimuli are experienced without consequence. Most standard rodent assays of fear (i.e., threat) memory, however, present the subject with a conditioning stimulus (CS) that on each occasion is paired with an aversive unconditioned stimulus (US) (Fanselow and Poulos, 2005). This discrepancy is pertinent to modeling traumatic memories in rodents, via back translation from human to rodent.

Theoretical accounts of associative learning predict that conditioned responses to CSs with a mixed or partial reinforcement history, which render the CS uncertain or ambiguous with regard to its expected outcome, may differ in certain respects from those that are consistently reinforced. For example, as compared to fully reinforced CSs, partially reinforced CSs can be more difficult to extinguish and produce lesser conditioned responses, due to associative strength accruing to the conditioning context or through the endowment of the CS with inhibitory (CS = no US) properties (Humphreys, 1939; Fitzgerald, 1963; Rawlins et al., 1985; Rescorla, 2007; Tsetsenis et al., 2007; Miguez et al., 2012; Harris et al., 2019).

Fear behavior that arises from partial reinforcement could involve neural circuits distinct from the well-described circuits implicated in standard (i.e., fully reinforced) fear conditioning (Pape and Pare, 2010; Bukalo et al., 2014; Tovote et al., 2015). Two brain regions that could be important for the acquisition and expression of partially reinforced fear (PRF) are the medial prefrontal cortex (mPFC, comprising, in the rodent, the prelimbic [PL], infralimbic [IL], and anterior cingulate [ACC] cortices) and the bed nucleus of the stria terminalis (BNST) (Lebow and Chen, 2016; Goode et al., 2019). The mPFC is engaged in experimental situations requiring integration of higher-order cues or disambiguation between conflicting cues to gate a level of response appropriate to the value of outcome (Sharpe and Killcross, 2018; Marek et al., 2019), while the BNST has been shown to support learning when a stimulus poorly predicts threat (Lebow and Chen, 2016; Goode et al., 2019).

These structures are also anatomically connected, with a particularly dense connection between the IL and the anterior regions of the BNST (Hurley et al., 1991; McDonald et al., 1999; Dong et al., 2001; Vertes, 2004; Radley and Sawchenko, 2011; Radley et al., 2013; Johnson et al., 2016; Glangetas et al., 2017; Tillman et al., 2018; Johnson et al., 2019). Additionally, BNST-projecting IL cells are activated by ‘unpredictable’ threat in a backward conditioning paradigm (Goode et al., 2019). Moreover, stimulation of glutamatergic mPFC inputs produces synaptic depression in the BNST (Glangetas et al., 2013). Together, these findings suggest the mPFC and BNST might form a functional circuit regulating fear to ambiguous and uncertain threats.

Here, we sought to elucidate the potential role of the mPFC and BNST and other neural circuits in PRF, using a paradigm in which a CS was paired with a footshock US on only half of the trials (McHugh et al., 2015; Glover et al., 2017). By combining immediate-early gene mapping, neuronal pathway tracing, in vivo chemogenetics, and a multiplexed approach combining in vivo chemogenetics and in vivo neuronal recordings, we demonstrate that the mPFC→BNST circuit negatively gates PRF.

Results

Lower freezing to a partially reinforced CS

The PRF conditioning procedure entailed presenting male C57BL/6J (B6) mice with three pairings of a tone CS and a footshock US, along with three interspersed presentations of the same CS without concomitant footshock (McHugh et al., 2015; Glover et al., 2017). For comparison, a fully reinforced fear (FRF) group received 3x CS+US pairings, and a CS-only control group received 6x CS presentations without the US (Figure 1A,B).

Figure 1. Lower freezing during retrieval of partially reinforced fear; effects of genetic strain.

(A) Schematic depiction of experimental procedure for assessing, in B6 mice, PRF and FRF, along with CS-only controls. (B) Schematic depiction of experimental procedure for assessing, in B6 mice, PRF and FRF retrieval in a novel context (context B) and the conditioning context (context A) (C) Lower CS-related freezing during retrieval in PRF mice than in FRF mice. Higher baseline and CS-related freezing in PRF and FRF mice relative to CS-only controls (n = 4–8 mice per group). (D) Schematic depiction of experimental procedure for assessing PRF and FRF retrieval in the B6 and S1 genetic strains. (E) Lower CS-related freezing during retrieval in PRF than in FRF in B6, not S1, mice (n = 7–8 mice per group/strain). Data are means ± SEM. *p<0.05.

Figure 1—source data 1. PRF versus FRF (Figure 1C).
Figure 1—source data 2. Strain comparison (Figure 1E).

Figure 1.

Figure 1—figure supplement 1. Freezing during conditioning.

Figure 1—figure supplement 1.

(A) Schematic depiction of experimental procedure for assessing, in B6 mice, PRF and FRF, along with CS-only controls. (B) No group differences in CS-related freezing during conditioning (n = 4–8 mice per group). (C) No difference in CS-related freezing between FRF and PRF mice during retrieval, broken down by CS presentation (n = 8 mice per group). (D) Schematic depiction of experimental procedure for assessing PRF and FRF retrieval in the B6 and S1 genetic strains. (E) No group differences in CS-related freezing during conditioning (n = 7–8 mice per group/strain). Data are means ± SEM.
Figure 1—figure supplement 2. Increased latency to feed in the NSF test after PRF.

Figure 1—figure supplement 2.

(A) Schematic depiction of experimental procedure for assessing behavior on the novelty-suppressed feeding (NSF) test after PRF or FRF conditioning, along with context-exposed controls. (B) Longer latencies to eat under high, but not low, illumination in PRF and FRF versus controls and in PRF versus FRF. n = 8 mice per group. Data are means ± SEM. *p<0.05.
Open in a new tab

Freezing increased to a similar extent over the six conditioning trials in the PRF group and over the three conditioning trials, plus the corresponding three no-trial periods, in FRF group, but did not significantly increase in the CS-only group (analysis of variance [ANOVA] group-effect: F(2,17)=5.74, p=0.0125; trial-effect: F(5,85)=13.49, p<0.0001; interaction: F(5,85)=1.64, p=0.1099). On a retrieval test conducted in a novel context (context B) the following day, the PRF and FRF groups froze more than CS-only controls during pre-CS baseline and CS presentation. Notably, however, CS-evoked freezing was lower in the PRF, relative to the FRF, group (ANOVA group-effect: F(2,17)=53.02, p<0.0001; CS-effect: F(1,17)=216.90, p=0.0001; interaction: F(2,17)=25.51, p=0.0001, followed by post-hoc tests: CS-only vs PRF p<0.0001, CS-only vs FRF p<0.0001, PRF vs FRF p=0.0008) (Figure 1C, Figure 1—figure supplement 1).

These data show that B6 mice express less freezing in the PRF, as compared to FRF, procedure despite the number of CS–US pairings being equivalent in both conditions. These differences are in line with lower freezing in the PRF procedure in a mixed C57BL/6J;CBA/J;129S6/SvEvTac genetic background (Tsetsenis et al., 2007) but, indicating a degree of strain dependency of PRF, differ from data in outbred CD-1 mice, in which freezing is equivalent between PRF and FRF groups (Glover et al., 2017).

Mice with an abnormal fear phenotype do not exhibit lower PRF

We next reasoned that an inbred strain (S1), which exhibits impaired contextual (and cued) fear discrimination, deficits in limiting fear following extinction and conditioned inhibition, and high fear expression in a different assay for PRF (Camp et al., 2012), might exhibit deficits in the current PRF assay (Camp et al., 2009; Camp et al., 2012; Figure 1D).

Across conditioning trials, there was increased freezing in the PRF and FRF groups, irrespective of strain (ANOVA group-effect: F(1,25)=0.15. p=0.7023, trial-effect: F(5,125)=38.02, p<0.0001; strain-effect: F(1,25)=7.68, p=0.0103; three-way interaction: F(5,125)=0.66, p=0.6539). On retrieval in context B, PRF B6 mice showed less CS-related freezing than their FRF counterparts, whereas freezing was equivalent in PRF and FRF S1 mice (ANOVA strain-effect: F(1,27)=8.72, p=0.0065; conditioning-type effect: F(1,27)=6.15, p=0.0197; CS-effect: F(1,27) = 524.00, p<0.0001; three-way interaction: F(1,27)=7.03, p=0.0132, followed by post-hoc tests: PRF vs FRF in B6 p=0.0038, PRF vs FRF in S1 p=0.2322) (Figure 1E, Figure 1—figure supplement 1).

The finding that S1 mice exhibit similar freezing to the PRF and FRF procedures aligns with the excessive fear shown by this strain to innocuous stimuli and following extinction (Camp et al., 2009; Camp et al., 2012) and further illustrates the strain dependency of PRF.

Increased latency to feed in the novelty-suppressed feeding test after PRF

An earlier study by Glover et al., 2017 found that following PRF conditioning, CD-1 mice had a higher latency to feed, as compared to a FRF group, in the novelty-suppressed feeding (NSF) test, an assay sensitive to anxiolytics and antidepressants (Ramaker and Dulawa, 2017).

To test whether PRF had a similar effect in B6 mice, NSF was assessed under either high or low illumination levels the day after B6 mice underwent either PRF or FRF. Under high, but not low, illumination, latencies to feed were higher in the PRF and FRF groups than unconditioned controls (ANOVA group-effect: F(1,44)=10.93, p=0.0019; illumination-effect: F(1,44)=4.11, p=0.0230; interaction: F(1,44)=2.83, p=0.0699, followed by post hoc tests: PRF vs Con p=0.0004, FRF vs Con p=0.0222) (Figure 1—figure supplement 2).

These data show that PRF conditioning increases anxiodepressive-like anxiety-like behavior under relatively aversive (high illumination) conditions of approach–avoidance conflict.

Ex vivo neuronal regional activity correlates of PRF

The complex behavioral sequelae of PRF suggest that this form of fear may have different neural substrates than FRF. We, therefore, sought to identify neural correlates of PRF by quantifying the number of c-Fos+ cells, as a proxy for neuronal activity, in forebrain regions following retrieval (for corresponding behavioral data, see Figure 1B).

There were a higher number of c-Fos+ cells in the basolateral amygdala (BLA) of FRF mice (ANOVA group-effect: F(2,17)=6.79, p=0.0068, followed by post hoc tests: FRF vs CS-only p=0.0038, FRF vs PRF p=0.0132), as compared to either PRF mice or a set of controls that had received CS-only trials during conditioning. In the paraventricular nucleus of the thalamus (PVT), another region implicated in fear (Penzo et al., 2015), c-Fos+ counts were higher in the PRF and FRF groups than controls (ANOVA group-effect: F(2,17)=4.01, p=0.0374, followed by post-hoc tests: CS-only vs PRF p=0.0281, CS-only vs FRF p=0.0145). No group differences were evident in the lateral or medial habenula, or ventral or dorsal hippocampus (Figure 2A–I, Figure 2—figure supplement 1).

Figure 2. PRF preferentially activates subregions of mPFC and BNST.

(A) Schematic depiction of experimental procedure for assessing ex vivo neuronal regional activity (via c-Fos immunohistochemistry) after PRF or FRF retrieval, along with CS-only controls. Representative images and c-Fos+ cell count differences for basal amygdala (B), paraventricular nucleus of the thalamus (C), infralimbic cortex (D), prelimbic cortex (E), posterior portion of the anterior cingulate cortex (F), anteroventral BNST (G), anterodorsal BNST (H), and lateral habenula (I). For corresponding behavioral data, see Figure 1B. Scale bars = 30 µm (B,D–F), 100 µm (C,I), 300 µm (G,H). n = 4–8 mice per group. Data are means ± SEM. *p<0.05.

Figure 2—source data 1. c-Fos.

Figure 2.

Figure 2—figure supplement 1. Ex vivo neuronal regional activity correlates of PRF.

Figure 2—figure supplement 1.

(A) Schematic depiction of experimental procedure for assessing ex vivo neuronal regional activity (via c-Fos immunohistochemistry) after PRF or FRF retrieval. Representative images and c-Fos+ cell counts for the dorsal dentate gyrus (B), dorsal CA3 (C), and ventral CA1/subiculum (D) regions of the hippocampus and the medial habenula (E). There were no group differences. Scale bars = 100 µm. n = 4–8 mice per group. Data are means ± SEM.
Figure 2—figure supplement 2. Connectivity between mPFC, BNST, and downstream targets.

Figure 2—figure supplement 2.

(A) Schematic depiction of viral strategy to label mPFC inputs to BNST neurons and their onward projections to the hypothalamus. (B) Representative image of synaptophysin-mCherry labeling in BNST neurons receiving mPFC input (scale bar = 1000 µm). (C) Representative image of synaptophysin-mCherry labeling in mPFC-innervated BNST neuronal projections in hypothalamus (scale bar = 500 µm). (D) Schematic depiction of viral strategy to label PVN oxytocin cell inputs to the PAG in Oxt-Cre mice. (E) Representative image of ChR2-EYFP labeling in the PVN (scale bar = 200 µm). Low (F) and high (G) magnification images of ChR2-EYFP labeling in the PAG (scale bars = 200 µm and 100 µm, respectively). (H) Schematic depiction of viral strategy to label PVN oxytocin cell inputs to the PAG in Oxt-Cre mice. (I) Representative image of synaptophysin-EYFP labeling in the PVN (scale bar = 200 µm). Low (J) and high (K) magnification images of synaptophysin-EYFP labeling in the PAG (scale bars = 200 µm and 100 µm, respectively). Note: LH = Lateral Hypothalamus; PVN = paraventricular nucleus of the hypothalamus; IPAG = lateral periaqueductal gray; vlPAG = ventrolateral periaqueductal gray; Aq = aqueduct.
Open in a new tab

In subregions of the mPFC, however, there were more c-Fos+ cells in the IL (F(2,17)=8.21, p=0.0032, followed by post-hoc tests: CS-only vs PRF p=0.0009, CS-only vs FRF p=0.0411, FRF vs PRF p=0.0420), but not the posterior ACC (F(2,17)=1.01, p=0.3862) of PRF and FRF mice, relative to CS-only controls. Counts in the PL were higher in PRF mice relative to controls and trended higher in the FRF group (F(2,17)=3.60, p=0.0499, followed by post hoc tests: CS-only vs PRF p=0.0196). The same pattern of elevated activity in the PRF group, relative to the other groups, was also evident in the BNST, though specifically in the anteroventral BNST (avBNST) (F(2,17)=19.43, p=0.0001, followed by post hoc tests: CS-only vs PRF p=0.0001, CS-only vs FRF p=0.0294, PRF vs FRF p=0.0005), not the anterodorsal BNST (adBNST) (F(2,14)=1.38, p=0.2831) (Figure 2A–I).

These findings show that retrieval of a PRF CS, despite being characterized by lower freezing than FRF, associates with a unique pattern of regional brain activation, with preferentially high activation in the IL and PL subregions of the mPFC and the avBNST.

Connectivity between mPFC, BNST, and downstream targets

Previous studies in the rat have demonstrated a direct (GABAergic) input from the mPFC to the BNST that is particularly dense between the IL and avBNST (Dong et al., 2001), but also present between the PL and avBNST (Johnson et al., 2016; Johnson et al., 2019). As our c-Fos data indicated activation of the IL, PL, and avBNST by PRF, we sought to verify an mPFC-to-BNST projection in mice.

In a combinatorial viral tracing approach to label postsynaptic targets of mPFC neurons in the BNST (Zingg et al., 2017; Sengupta and Holmes, 2019), a construct containing a Cre-containing anterograde trans-synaptic virus was infused into the mPFC and a Cre-dependent, synaptophysin-containing, mCherry-fused construct infused into the BNST (Figure 2—figure supplement 2). Indicative of monosynaptic input from the mPFC, mCherry labeling was apparent in BNST neurons, mainly in the ventral areas below the anterior commissure. In the rat, PL neurons form close appositions with GABAergic cells in the avBNST that in turn send efferents to the paraventricular nucleus of the hypothalamus (PVN), a key mediator of responses to stress and defensive behaviors (Johnson et al., 2016; Johnson et al., 2019). Indicating that a similar connection is likely present in mice, inspection of our tissue revealed mCherry/synaptophysin expression originating from mPFC-innervated BNST neurons in the PVN, as well as lateral hypothalamus.

A corollary to the existence of a disynaptic mPFC–BNST–PVN circuit in mice is whether the PVN in turn targets other fear-mediating regions in this species. To gain initial insight into this question, we infused a Cre-dependent, YFP-fused construct containing either channelrohodpsin2 (ChR2) or synaptophysin into the PVN of oxytocin-Cre mice, to label a major population of (oxytocin-positive) PVN cells. This indicated labeling in the ventrolateral periaqueductal gray (vl/PAG) (Figure 2—figure supplement 2), a region known to regulate defensive behaviors including freezing (Tovote et al., 2015).

Together these data provide evidence of input from the mPFC to the BNST in the mouse, as well as onward connections from the BNST to the PVN and in turn possibly on to the vl/PAG. Thus, PRF engagement of the mPFC and BNST can be viewed in the context of a direct connection between these regions and their downstream access to a broader fear-regulating neural circuitry.

Inhibition of mPFC→BNST neurons increases freezing to a PRF CS

To causally interrogate the contribution of the mPFC→BNST pathway to PRF, a retrogradely transported Cre-containing construct viral construct was infused into the BNST and a construct containing a Cre-dependent form of hM4Di (or mCherry control) infused into the mPFC, enabling the expression of the inhibitory DREADD in mPFC→BNST neurons to inhibit their activity, via systemic injection of clozapine N-oxide (CNO), during retrieval (Figure 3A,B).

Figure 3. Inhibition of mPFC→BNST neurons increases PRF.

(A) Schematic depiction of experimental procedure for assessing effects of chemogenetic inhibition of mPFC→BNST neurons during retrieval. (B) Cartoon of viral strategy and representative images of hM4Di–mCherry labeling in BNST neurons receiving mPFC input (scale bars = 200 µm). (C) Lower CS-related freezing during retrieval in PRF mice than in FRF mice transfected with mCherry, not hHM4Di. Data are means ± SEM. *p<0.05.

Figure 3.

Figure 3—figure supplement 1. Freezing during conditioning prior to mPFC→BNST inhibition on retrieval.

Figure 3—figure supplement 1.

(A) Schematic depiction of viral strategy to selectively inhibit BNST-projecting mPFC neurons during retrieval. (B) Freezing increased across CS trials, irrespective of virus group. (C) Trial-by-trial breakdown of freezing during each CS of retrieval indicated a non-significant trend for decreasing freezing across trials in the mCherry PRF group. n = 8–9 mice group/virus. Freezing data are means ± SEM.
Figure 3—figure supplement 2. Electrode placements and virus localization for combined chemogenetic/single-unit recordings.

Figure 3—figure supplement 2.

Estimated location of tips of the electrodes at the center of the multi-array in the IL of PRF (A) and FRF (B) mice. Estimated extent of virus, as indicated by mCherry expression, in the BNST of PRF (C) and FRF (D) mice (darker shading represents areas of greater overlap across mice).
Figure 3—figure supplement 3. CS and freezing-related IL unit activity and effects of mPFC→BNST inhibition.

Figure 3—figure supplement 3.

(A) Schematic depiction of experimental procedure for in vivo IL single-unit recordings, combined with chemogenetic inhibition of mPFC→BNST neurons during retrieval (n = 8–9 mice per group/virus). (B) Cartoon of viral strategy and representative image of hM4Di-mCherry labeling in the mPFC and electrode array placement in the IL (scale bar = 500 µm). (C) Raster plot of a representative IL unit firing in response to CS onset (CS-ON neuron). Baseline-normalized population trace of CS-ON neuronal activity during retrieval; average of all groups (D) and split by group (E). (F) Higher percentage of CS-ON units during PRF than FRF retrieval in mice transfected with mCherry, not hHM4Di (n = 17 recorded units in PRF/mCherry, n = 25 units in PRF/hM4Di, n = 20 units in FRF/mCherry, n = 17 units in FRF/hM4Di, from three mice per group/virus). Baseline-normalized population trace of freeze cessation (Freeze-OFF neurons) IL unit activity during retrieval; average of all groups (G) and split by group (H) (n = 3 CS units in PRF/mCherry, n = 6 units in PRF/hM4Di, n = 1 units in FRF/mCherry, n = 1 units in FRF/hM4Di). (I) Higher percentage of Freeze-OFF units during PRF than FRF retrieval in mice transfected with mCherry, not hHM4Di (n = 17 recorded units in PRF/mCherry, n = 25 units in PRF/hM4Di, n = 20 units in FRF/mCherry, n = 17 units in FRF/hM4Di, from three mice per group/virus, n = 4 Freeze-OFF units in PRF/mCherry, n = 6 units in PRF/hM4Di, n = 3 units in FRF/mCherry, n = 0 units in FRF/hM4Di). Baseline-normalized population trace of freeze onset (Freeze-ON neurons) IL unit activity during retrieval; average of all groups (J) and split by group (K). (L) No differences in the percentage of Freeze-OFF units during retrieval between groups (n = 17 units in PRF/mCherry, n = 25 units in PRF/hM4Di, n = 20 units in FRF/mCherry, n = 17 units in FRF/hM4Di, from three mice per group/virus, n = 4 Freeze-ON units in PRF/mCherry, n = 6 units in PRF/hM4Di, n = 4 units in FRF/mCherry, n = 2 units in FRF/hM4Di). Data are means ± SEM. *p<0.05.
Figure 3—figure supplement 4. Heat maps illustrating IL unit activity.

Figure 3—figure supplement 4.

(A) Heat plots of unit activity aligned to CS onset (left columns), freeze cessation (center columns), and freeze onset (right columns). The same data shown as peri-event histograms and % event-related activity can be found in Figure 3—figure supplement 3.
Open in a new tab

During conditioning, freezing increased over trials to a similar extent in all groups (ANOVA trial-effect: F(5,145)=23.54, p<0.0001; group effect: F(3,145)=0.91, p=0.4467; interaction: F(15,145)=0.61, p=0.08647) (Figure 3—figure supplement 1). Following CNO administration, CS-related freezing during retrieval was lower in PRF mice than in FRF mice expressing the control virus, replicating our earlier data. By contrast, there was no difference in freezing in mice expressing hM4Di (ANOVA conditioning-type effect: F(1,29)=9.35, p=0.0048; virus-group effect: F(1,29)=12.15, p=0.0016; CS: F(1,29)=1331.02, p<0.0001; three-way interaction: F(1,29)=6.58, p=0.0157, followed by post-hoc tests: mCherry PRF vs mCherry FRF p<0.0001, hM4Di PRF vs hM4Di FRF p=0.1425, mCherry PRF vs hM4Di PRF p=0.0013, mCherry PRF vs hM4Di PRF p=0.7951) (Figure 3C). Examination of the trial-by-trial freezing during retrieval indicated no significant trial-related differences in freezing, despite a trend for decreasing freezing across trials in the mCherry PRF group (ANOVA trial-effect: F(5,145)=1.83, p=0.1098; group-effect: F(3,29)=14.15, p<0.0001; trial x group interaction: F(15,145)=1.04, p=0.4213) (Figure 3—figure supplement 1).

These data show that inhibition of mPFC→BNST neurons increases freezing to a PRF CS. This finding suggests that engagement of these mPFC→BNST neurons limits the expression ofto the unreliable, PRF, though it remains possible that inhibition of these neurons also produces an increase in PRF expression, which may have been masked due to high (ceiling) levels of freezing.

IL cells signal CS onset and freezing cessation

The finding that inhibiting mPFC outputs to the BNST pathway increases freezing to a PRF CS implies that mPFC neurons likely encode some aspects of fear. To address this possibility, we devised an approach entailing chemogenetic inhibition of mPFC→BNST neurons (as described above) coupled with in vivo recordings of mPFC single-unit activity via chronically implanted electrode arrays, which we targeted at the IL (Figure 3—figure supplement 2). The average firing rate of units did not statistically differ between groups (FRF mCherry: 4.10 ± 0.64, FRF hM4Di: 3.45 ± 0.49, FRF mCherry: 2.67 ± 0.66, FRF hM4Di: 1.67 ± 0.34).

Aligning the single-unit data to the presentation of the CSs during retrieval revealed examples of IL units with activity time-locked to the onset of the CS (Figure 3—figure supplement 3). Units exhibiting activity >/<1.96 z scores from baseline in at least two 100 ms time bins within the 500 ms of CS onset were classified as CS responsive (CS-ON). Overall, CS-ON units showed a significant change in neuronal activity in response to the CS (baseline: 0.15 ± 0.35, post-CS: 1.43 ± 0.55, paired t-test: t(10)=6.51, p<0.0001) (Figure 3—figure supplement 3, and for heat maps, see Figure 3—figure supplement 4). Peak responses occurred within 200–300 s of CS onset and were highest in the mCherry FRF group (Figure 3—figure supplement 3). However, when the percentage of CS-ON units was calculated and compared across the conditioning and virus groups, this revealed a higher proportion of CS-ON units in the mCherry groups than in hM4Di groups for PRF mice (Fisher’s exact test: p=0.0122), but no differences between virus groups in the FRF mice (Fisher’s exact test: p=0.6090), and no difference between PRF and FRF groups, irrespective of virus group (Fisher’s exact test in mCherry: p=0.2510; in hM4Di: p=1.000) (Figure 3—figure supplement 3).

To examine whether IL cells were also associated with the behavior of mice during testing, their activity was aligned to episodes of freezing and those cells displaying a reliable change relative to either the onset or cessation of freezing (i.e., resumption of movement; >/<1.96 z from baseline in at least two 100 ms time bins within the 500 ms of the event). These units, classified as Freeze-ON and Freeze-OFF, respectively, showed a significant change in baseline-normalized activity (Freeze-ON baseline: −0.81 ± 0.47, post-event: −2.12 ± 0.52, paired t-test: t(10)=4.60, p=0.0010, Freeze-OFF baseline: 0.73 ± 0.43, post-event: 1.67 ± 0.40, paired t-test: t(12)=8.54, p<0.0001) (Figure 3—figure supplement 3, and for heat maps, see Figure 3—figure supplement 4). Freeze-ON units displayed a decreased firing rate at freezing onset, which was most evident in both of the mCherry groups, while Freeze-OFF units increased firing rate at the cessation of freezing in both groups (Figure 3—figure supplement 3).

When the percentage of these cell types were compared across groups, there was a higher percentage of Freeze-OFF units in the mCherry PRF group than in the hM4Di PRF group (Fisher’s exact test: p=0.0024), whereas there was no group difference in FRF mice (Fisher’s exact test: p=1.000) and no difference between PRF and FRF groups in either the mCherry (Fisher’s exact test: p=0.4600) or hM4Di (Fisher’s exact test: p=0.0590) virus conditions (Figure 3—figure supplement 3). Conversely, there was no difference between the mCherry and hM4Di groups in the proportion of Freeze-ON units, irrespective of whether mice had undergone PRF or FRF.

Discussion

Here, we sought to provide new insight into the neural substrates regulating the fear response to an uncertain/ambiguous threat. Employing an assay of partial tone+shock reinforcement in B6 mice, we found that PRF conditioning produced a lower fear response than FRF, which was associated with preferential neuronal activation in the mPFC and BNST. We also show that the mPFC and BNST formed a monosynaptic circuit that, when chemogenetically inhibited, caused a selective increase in the expression of PRF and an attendant loss of in vivo correlates of both CS onset and freezing cessation in IL units.

The current findings align with and extend prior work implicating the mPFC and BNST in various situations in which there is ambiguity and uncertainty about a threat. For example, the mPFC is engaged in settings that require integration of higher-order cues to gate learned responses (Halladay and Blair, 2015; Halladay and Blair, 2017; Sharpe and Killcross, 2018; Marek et al., 2019), or where there is conflict between excitatory and inhibitory CS associations, for instance in fear extinction (Milad and Quirk, 2012; Bloodgood et al., 2018; Lay et al., 2020), fear discrimination (Grosso et al., 2018), threat/safety conditioning (Sangha et al., 2014; Meyer et al., 2019), and punished reward-seeking (Burgos-Robles et al., 2017; Halladay et al., 2020).

The BNST, meanwhile, supports learning in the absence of the BLA (Poulos et al., 2010; Zimmerman and Maren, 2011), and when a stimulus poorly predicts threat (Lebow and Chen, 2016; Goode et al., 2019; Bjorni et al., 2020), either because it is distal (e.g., predator odor; Fendt et al., 2003; Xu et al., 2012; Breitfeld et al., 2015; Verma et al., 2018; Goode et al., 2020), diffuse (e.g., contextual; Sullivan et al., 2004; Kalin et al., 2005; Duvarci et al., 2009; Davis et al., 2010; Luyten et al., 2012; Jennings et al., 2013; c.f. Haufler et al., 2013), or temporally ambiguous (e.g., random or sustained; Waddell et al., 2006; Walker et al., 2009; Hammack et al., 2015; Daldrup et al., 2016; Goode and Maren, 2017; Lange et al., 2017) with respect to the US.

These known functions of the mPFC and BNST make these structures well placed to mediate fear under conditions of partial reinforcement where the CS is experienced both with and without the US. As we show here, the mPFC and BNST form a discrete neural circuit, through a direct anatomical connection, that serves to limit the expression of partially reinforced fear. This observation is reminiscent of a recent study showing that IL neurons projecting to the avBNST are activated (measured by c-Fos) by a measure of unpredictable threat (backward conditioning), in which US presentation precedes the CS (Goode and Maren, 2017). In conjunction with the current data, there is convergent evidence supporting a key role for the mPFC→BNST circuit in mediating fear across various measures of threat uncertainty, ambiguity, and unpredictability.

The precise nature of this role remains to be fully clarified, however. One possibility is that when discrete cues are relatively poor predictors of danger, other environmental stimuli, such as context, modulate the expression of fear in a manner that recruits the mPFC to exert top-down control over the BNST. In support of this possibility, the mPFC is posited to subserve higher-order modulation of conditioned responding (Sharpe and Killcross, 2018); indeed, a recent study in rats found that lesions of either the PL or IL impaired one measure of such modulation known as occasion setting (Roughley and Killcross, 2019). Another possible explanation for the increase in PRF caused by mPFC→BNST inhibition is that CS-alone presentations during conditioning imbues the CS with inhibitory properties that are gated by mPFC→BNST neurons during retrieval.

Learned inhibition is a function attributed to the mPFC and in particular the IL. For example, pharmacological, optogenetic, and chemogenetic inhibition of the IL impairs the formation and/or retrieval of fear extinction memories (Laurent and Westbrook, 2009; Bukalo et al., 2015; Do-Monte et al., 2015; Kim et al., 2016; Lay et al., 2020; Bukalo et al., 2021) and the expression of learned safety acquired through explicit CS–US unpairing (Sangha et al., 2014). Conversely, presentation of a safety signal during inescapable stress decreases activity (c-Fos) in a lateral area of BNST encompassing avBNST (Christianson et al., 2011). Furthermore, lesioning this area reduces inappropriate fear to a non-reinforced CS in rats with high-trait anxiety (Duvarci et al., 2009). Indeed, a growing number of lesion and functional neuroimaging studies in non-human primates and humans have implicated the BNST in the processing of uncertain threat (see Goode et al., 2019; Miles and Maren, 2019). Together, these findings suggest that inhibitory properties of the partially reinforced CS could be signaled by the IL downstream to the BNST, thereby limiting the expression of CS-induced fear to a level appropriate to its partial reinforcement history.

It is important to note in this regard that while we targeted our virus infusions to the IL and avBNST – based on prior evidence of a dense anatomical connection between these subregions – the small size and ventral location of these areas meant that viral transfection encompassed parts of the PL and adBNST. The adBNST is engaged by challenges that produce negative affect (Centanni et al., 2019), undergoes plastic changes in response to chronic stress (Conrad et al., 2011), and, of particular relevance here, is a target of dorsal raphe (Marcinkiewcz et al., 2016), BLA (Lange et al., 2017), and central amygdala (Asok et al., 2018) neurons that sustain fear responses to predictable and unpredictable threats. As such, although the adBNST is a less densely innervated by the mPFC (Hurley et al., 1991; McDonald et al., 1999; Dong et al., 2001; Vertes, 2004; Radley and Sawchenko, 2011; Radley et al., 2013; Johnson et al., 2016; Glangetas et al., 2017; Tillman et al., 2018; Johnson et al., 2019), and current as well as prior (Goode et al., 2019) c-Fos data do not indicate adBNST activation with uncertain threat, it would be premature to exclude a contribution of this area to PRF.

The possible contribution of BNST-targeting PL neurons to PRF also should not be discounted. Precisely dissociating the roles of PL and IL inputs to BNST in PRF will be an interesting avenue for future work. While the PL has been ascribed a role in promoting FRF via its outputs to the BLA (Pape and Pare, 2010; Dias et al., 2013; Bukalo et al., 2014; Tovote et al., 2015), recent work found that optogenetically silencing PL inputs to the avBNST increased immobility and associated stress hormone responses in the rat shock-probe burying and tail suspension tests (Johnson et al., 2019; see also Radley et al., 2009). These effects suggest a role for PL inputs in attenuating negative affect and as such are broadly congruent and potentially explanatory of the current data, despite important differences in methodology.

In the current study, we found neuronal correlates of PRF, specifically within the IL. As in our prior studies of FRF in mice (Fitzgerald et al., 2014; Fitzgerald et al., 2015), a subset of IL neurons were phasically active to CS presentation during fear retrieval. Intriguingly, we also found a subset of IL neurons that displayed phasic activity during the cessation, but not onset, of freezing during fear retrieval, echoing recordings in rats that uncovered movement-related activity in IL units (Halladay and Blair, 2015; Halladay and Blair, 2017). Inhibition of BNST-projecting mPFC neurons during fear retrieval essentially abolished the CS- and Freeze-OFF-associated neuronal activity in IL neurons and, in parallel, increased freezing in the PRF group. The ability of mPFC→BNST inhibition to ablate these neuronal correlates could have arisen from the chemogenetic inhibition of BNST-projecting IL neurons, resulting in a loss-of-function in this pathway and a selective increase in PRF mice. Two caveats to this interpretation are that, firstly, inhibition-induced increases in freezing in FRF mice may have been masked by a performance ‘ceiling’ in the FRF control group and, secondly, hM4Di expression in our recording experiment was not restricted to BNST-projections within the IL, and also encompassed neurons in the PL.

With regard to the broader neural circuitry in which the mPFC→BNST circuit operates to mediate PRF, using trans-synaptic tracing, we found evidence of a disynaptic mPFC→BNST→PVN circuit in mice, as has been reported in rats (Radley and Sawchenko, 2011; Johnson et al., 2016; Johnson et al., 2019). The PVN contains a high density of cells expressing peptide hormones implicated in stress and fear, notably corticotrophin-releasing hormone (CRH) and oxytocin (Herman and Tasker, 2016; Triana-Del Río et al., 2019). Though CRH-expressing neurons in the PVN receive input from the BNST (Colmers and Bains, 2018), it is unclear whether oxytocin-producing cells do so. Interesting nonetheless, we found that mouse PVN oxytocinergic neurons strongly innervate the freezing-regulating vl/PAG, replicating earlier work in cats and naked mole rats (Holstege, 1987; Rosen et al., 2008). Given the vl/PAG also receives input from a population of avBNST neurons that is, in turn, innervated by the PL (Johnson et al., 2016; Johnson et al., 2019), these tracing results suggest that in addition to directly innervating the PVN, mPFC→avBNST neurons may also have both direct and indirect (via the PVN) access to the vl/PAG. This positions the circuit to modulate multiple behavioral and neuroendocrine responses to PRF.

In summary, the current study found that B6 mice expressed lower fear to a CS that is partially, rather than fully, reinforced with footshock. Lower PRF expression was not apparent in a mouse strain (S1) deficient in fear discrimination and learned inhibition. Furthermore, c-Fos mapping revealed PRF preferentially recruited the mPFC and BNST, and neuronal tracing showed direct neuronal projections from mPFC to the BNST, with downstream connections to stress- and fear-mediating regions. Demonstrating the causal importance of the mPFC→BNST neurons, inhibiting this pathway increased PRF and abolished neuronal correlates of CS presentation and freezing cessation in the IL. Collectively, these findings provide novel insight into the neural substrates of PRF, with potential translational relevance to anxiety and trauma- and stressor-related disorders in which threats are typically ambiguous and unpredictable.

Materials and methods

Subjects

Subjects were adult male C57BL/6J (B6), 129S1/SvImJ (S1), and B6;129S-Oxttm1.1(cre)Dolsn/J (JAX strain 024234) (Oxt-Cre) mice obtained from the Jackson Laboratory (Bar Harbor, ME, USA) and were at least 8 weeks old at the time of testing. Mice were group-housed in a temperature (22 ± 3°C) and humidity (45 ± 15%) controlled vivarium under a 12 hr light/dark cycle (lights on 0600 hr). Mice undergoing surgery for chronic implantation were single housed after surgery to prevent the implant being damaged by a cage mate. All experimental procedures were approved by the National Institute on Alcohol Abuse and Alcoholism (NIAAA) and Santa Clara University Animal Care and Use Committees (SCU AWA: D18-01042) and followed the NIH guidelines outlined in ‘Using Animals in Intramural Research’ and the local Animal Care and Use Committees.

Partially versus fully reinforced threat (standard procedure)

The threat conditioning procedures were based on previous studies with slight modifications (McHugh et al., 2015; Glover et al., 2017). For this and all other experiments, prior to testing, mice were randomly assigned to experimental groups and habituated to handling for approximately 10 min per day for 4 days. The following procedures were used in all experiments, unless stated otherwise below.

Conditioning was conducted in a 27 × 27 × 11 cm chamber with opaque metallic walls and a metal rod floor (context A). The walls of the chamber were cleaned with a 79% water/20% ethanol/1% vanilla-extract solution to provide a distinctive odor – this was repeated after each session. All conditioning procedures began with a 180 s baseline period. Conditioning for FRF entailed three presentations (60–90 s variable inter-CS interval) of a 30 s, 75 dB (50 ms rise time), white noise (CS) that co-terminated with a 2 s, 0.6 mA scrambled footshock (US). After the final pairing for all groups, there was a 120 s no-stimulus period before the mouse was returned to the home-cage. The procedure was the same for PRF, with the exception that the CS was presented, without the US, on an additional three occasions during the intervals between the CS+US pairings (order: CS+US; CS+US; CS-noUS; CS-noUS; CS+US; CS-noUS, 15–60 s inter-stimulus interval). Where a CS-only control group was included (stated below), the CS was presented on six occasions, corresponding to the order and timing of the PRF group, but without any concomitant US.

CS retrieval took place one day after conditioning in a novel context B, a 27 × 27 × 11 cm chamber with white Plexiglas walls (rear wall curved) and a solid white floor, which was housed in a different room to context A. Between each session, all surfaces of the chamber were cleaned with a 99% water/1% acetic acid solution. After a 180 s baseline period, there were six CS presentations (20–60 s inter-pairing interval). There was a 20 s no-stimulus period before the mouse was returned to the home-cage. All groups were tested in the same manner.

Stimulus presentation was controlled by the Med Associates VideoFreeze system (Med Associates, Burlington, VT, USA). Freezing, scored manually every 5 s (as no visible movement except that required for breathing), was measured as an index of fear (Blanchard and Blanchard, 1972) and converted to a percentage ([number of freezing observations/total number of observations] x 100).

Novelty-suppressed feeding

Mice (assigned to FRF, PRF, and a control group exposed to the conditioning context for 1 min) underwent conditioning and were then food-deprived for 24 hr. Subsequently, they were assessed using the NSF test for anxiodepressive-like behavior, as previously described (Glover et al., 2017). The test apparatus was a novel, 50 cm3 white Plexiglas box with the floor covered by fresh cage substrate. A single pellet of regular home-cage food chow was placed within a plastic weigh-boat in the center of the box. Separate groups of mice underwent the test under 180 lux (low) and 1350 lux (high) illumination. The mouse was placed in a corner of the box, facing the center, and the latency to begin eating the chow was measured from a video-recording. The test ended when eating started or when 600 s had elapsed.

Behavior in mouse strain with a persistent and generalized fear phenotype

B6 and S1 mice underwent conditioning and retrieval as described under ‘Partially versus fully reinforced threat (standard procedure)’.

Regional patterns of fear-related c-Fos activity

Mice (assigned into FRF, PRF, and a CS-only group) underwent conditioning and retrieval as described under ‘Partially versus fully reinforced threat (standard procedure)’. Ninety minutes after retrieval, mice were deeply anesthetized with sodium pentobarbital and transcardially perfused with ice-cold phosphate-buffered saline (PBS, pH 7.4) followed by ice-cold 4% paraformaldehyde (PFA). Brains were removed, and 50 µm coronal sections were cut on a vibratome (Leica VT1000 S, Leica Biosystems Inc, Buffalo Grove, IL, USA) and stored free floating in 0.1 M phosphate buffer (PB) at 4°C for <1 week.

Sections were incubated successively with 10% normal goat serum and 1% bovine serum albumin in PBS-TritonX (0.3%) for 2 hr, a mixture of rabbit anti-c-Fos (9F6) (cat# 2250S, 1:1000, Cell Signaling Technology, Danvers, MA, USA) and a mouse monoclonal anti-NeuN antibody (MAB377, Millipore, 1:1000) in a dilution of 1% normal goat serum and 0.1% bovine serum albumin in PBS-TritonX (0.3%) for two nights on a platform rocker at 4°C. Sections were then rinsed 3× for 10 min in PBS and incubated in anti-rabbit Alexa 488 secondary antibody (cat# A-11034, 1:500, Invitrogen, Eugene, OR, USA) and Alexa Fluor 555 anti-mouse antibody (cat# A-21422, 1:500, Invitrogen) in a dilution of 1% normal goat serum and 0.1% bovine serum albumin in PBS-TritonX (0.3%) at room temperature on a platform rocker for 2 hr. Sections were rinsed in PBS 2× for 10 min and then counterstained with Hoechst 33342 (5 µg/mL, cat# H1399, Thermo Fisher Scientific, Waltham, MA, USA) in PBS. Sections were rinsed 3× for 10 min in PBS before each series. After rinsed once in 0.1 M PB for 10 min, serial sections were mounted onto slides, air-dried, coverslipped with aqueous mounting media (10 mM Tris–HCl [pH 8.0] [5 mL], DABCO [cat# D27802-25G, Sigma–Aldrich] [1.42 g], and glycerol [cat# 5516, Sigma–Aldrich] [50 mL]), then sealed with clear nail polish.

Images of all three channels (c-Fos, NeuN, Hoechst) for all sections were acquired using an Olympus VS120 Virtual Slide Microscope system (Olympus, Center Valley, PA, USA, VS_ASW software) with a 20× objective (U Plan S Apo; 20×, NA 0.75). The NeuN channel, in the autofocus mode, was used as a focus reference, in the autofocus mode. For image analysis, the FIJI (https://imagej.net/Fiji) (Schindelin et al., 2012) with VSI reader plugin (BIOP, Zurich, Switzerland, https://c4science.ch/w/bioimaging_and_optics_platform_biop/image-processing/imagej_tools/ijab-biop_vsireader/) was used. A contour of each brain area (region of interest, ROI) was manually drawn on the Hoechst channel with reference to a mouse brain atlas (Paxinos and Franklin, 2001) on the thumbnail image that covers the whole coronal sections and the full resolution image of the ROI was extracted for all channels.

Counts were made in the following brain regions: PL, IL, ventromedial BNST, dorsolateral BNST, BLA, lateral and medial habenula, and ventral and dorsal hippocampus (for cartoons depicting region definitions and example images, see Figure 2, Figure 2—figure supplement 1). For each brain region, cell counts were conducted (blind to test group) in two to four sections from each hemisphere, for a total of six data points per region per mouse. It was unnecessary to correct for double counting because sections were non-consecutive. The ROIs were transferred to the c-Fos channel, and the mean number of c-Fos positive cells per 0.25 mm2 within the ROI was quantified in a semi-automated manner using a custom-written macro.

Trans-synaptic tracing of mPFC→BNST neuronal outputs to hypothalamus

Mice were placed in a stereotaxic alignment system (Kopf Instruments, Tujunga, CA, USA) and kept under isoflurane anesthesia. AAV1-hSyn Cre-WPRE-hGH (titer: 3.5 × 1013 GC/mL, plasmid# 55637, obtained from Addgene, Cambridge, MA, USA, and packaged by Vigene Biosciences, Rockville, MD, USA) was unliterally infused (0.15 µL) into the PFC, and AAV5-Ef1a-DIO-eYFP (titer: 2.1 × 1012, obtained from the UNC Vector Core) was unilaterally infused (0.15 µL) into the BNST of the same hemisphere. The coordinates for BNST infusions were Medial/Lateral = ±0.80, Dorsal/Ventral = −4.15, Anterior/Posterior = +0.03. The coordinates for mPFC infusions were Medial/Lateral = ±1.30 (20° angle), Dorsal/Ventral = −3.00, Anterior/Posterior = +2.00.

Four weeks later, mice were terminally anesthetized with sodium pentobarbital (50–60 mg/kg). Brains were removed and initially suspended in 4% PFA overnight and then at 4°C in 0.1 M PB for 1–2 days. The general histological procedures were also the same as described under ‘Regional patterns of fear-related c-Fos activity’, with the exception that sections were successively immunostained with rabbit anti-DsRed (1:200 dilution, cat# 632496, Takara Bio, Mountain View, CA, USA) and Alexa Fluor 555 Goat Anti-Rabbit (1:500 dilution, cat# A-21428, Thermo Fisher Scientific), PBS (9 mL), 10% Triton X-100 (0.3% final) (300 µL), and blocking buffer (as above; 1 mL), and incubated on a platform rocker for 2 hr (20°C). The sections were mounted onto slides and allowed to dry.

Far-Red nuclear staining dye (four to five drops of NucRed Dead 647 ReadyProbe Reagent, Thermo Fisher Scientific, and 0.1 M phosphate buffer [2.5 mL]) was pipetted onto the mounted sections. After 15–20 min, the excess solution was suctioned using a benchtop aspirator. Once sufficiently dried, the slides were coverslipped using the same aqueous mounting media. Fluorescent images were taken with a Zeiss (LSM 700, Carl Zeiss Microscopy, Thornwood, NY, USA) confocal microscope under a Plan-Apochromat 10x/0.8 M27 objective.

Output tracing of oxytocin PVN cells

Oxt-Cre mice were placed in a stereotaxic alignment system (Kopf Instruments) under isoflurane anesthesia. Either a viral vector-containing ChR2, fused to GFP (AAV2-EF1a-DIO-hChR2(E123T/T159C)-EYFP, titer: 6.10 × 1012 vp/mL, Addgene plasmid#35509, obtained from the UNC Vector Core, Chapel Hill, NC, USA) or a vector-containing synaptophysin, fused to GFP (AAV8.2-hEF1a-DIO-synaptophysin-EYFP, titer: 2.1 × 1013 vg/mL, generously provided by Dr. R. Neve, Massachusetts General Hospital, Belmont, MA, USA) was bilaterally infused (0.15 µL) into the PVN (coordinates ML = ±1.50 mm (15° angle), DV = −4.87 mm, and AP = +0.75 mm, relative to bregma).

Five weeks later, mice were deeply anesthetized with sodium pentobarbital and transcardially perfused with PBS followed by 4% PFA. Coronal (50 µm thick) sections were prepared by vibratome (VT1000S; Leica). The general histological procedures were also the same as described under ‘Regional patterns of fear-related c-Fos activity’, with the exception that sections were successively immunostained with chicken anti-GFP (1:5000 dilution, cat# ab13970, Abcam, Cambridge, UK) and anti-chicken Alexa 488 secondary antibody (1:500 dilution, cat# ab150169, Abcam). Images were taken with a fluorescence microscopy (VS120; U Plan S Apo; 20×, NA 0.75; Olympus).

Effects of in vivo chemogenetic mPFC→BNST pathway inhibition

Mice were placed in a stereotaxic alignment system (Kopf Instruments) and kept under isoflurane anesthesia. rAAV2-retro-Ef1a-Cre (titer: 1.0 × 1013 gc/mL, obtained from the Salk Institute, La Jolla, CA, USA) was bilaterally infused targeting the BNST (0.15 µL/hemisphere). Additionally, either AAV8-hSyn-DIO-hM4D(Gi)-mCherry-WPRE (titer: 2.25 × 1013 gc/mL, obtained from the Massachusetts General Hospital Gene Delivery Technology Core, Cambridge, MA, USA) or AAV8.2-hEF1-DIO-mCherry-WPRE (titer: 2.13 × 1013 vg/mL, obtained from the Massachusetts General Hospital Gene Delivery Technology Core) was bilaterally infused targeting the IL (0.15 µL/hemisphere). Each infusion was done over 10 min using a Hamilton syringe and 33-gauge needle. The needle was left in place for a further 5 min to ensure diffusion. The coordinates for mPFC and BNST infusions were as described above.

Four weeks after surgery, mice underwent conditioning and retrieval testing as described under ‘Partially versus fully reinforced threat (standard procedure)’. Thirty minutes prior to the retrieval test, CNO (0.01 mg/mL/kg) was injected intraperitoneally.

After the completion of testing, mice were terminally anesthetized with sodium pentobarbital (50–60 mg/kg). Brains were removed and suspended in 4% PFA overnight and then at 4°C in 0.1 M PB for 1–2 days. Coronal sections (50 μm thick) were cut with a vibratome (Leica VT1000 S, Leica Biosystems Inc) and coverslipped with Vectashield HardSet mounting medium with DAPI (Vector Laboratories, Inc, Burlingame, CA, USA). Sections were imaged using an Olympus BX41 microscope (Olympus America Inc, Center Valley, PA, USA). Mice without viral (i.e., mCherry) expression in the region of interest were removed from the analysis.

In vivo mPFC single-unit recordings during chemogenetic mPFC→BNST inhibition

Mice were placed in a stereotaxic alignment system (Kopf Instruments) and kept under isoflurane anesthesia. rAAV2-retro-Ef1a-Cre (titer: 1.0 × 1013 GC/mL, obtained from the Salk Institute) was bilaterally infused targeting the BNST (0.25 µL/hemisphere). In addition, either AAV8-hSyn-DIO-hM4D(Gi)-mCherry-WPRE (titer: 2.25 × 1013 GC/mL, obtained from the Massachusetts General Hospital Gene Delivery Technology Core) or AAV8.2-hEF1-DIO-mCherry-WPRE (titer: 2.13 × 1013 vg/mL, obtained from the Massachusetts General Hospital Gene Delivery Technology Core) was bilaterally infused targeting the IL (0.25 µL/hemisphere). Infusions were done over 10 min using a Hamilton syringe and 33-gauge needle. The needle was left in place for a further 5 min to ensure diffusion. The coordinates for mPFC and BNST infusions were as described above. During the same surgery, a microelectrode array (two rows of eight electrodes with 35 µm electrode spacing and 200 µm row spacing [Innovative Neurophysiology, Durham, NC, USA]) was unilaterally (hemisphere counterbalanced) targeting the IL (array center: ML = ±0.30 mm, DV = −2.70 mm, AP = +1.75 mm) and affixed to the skull with dental cement.

Five weeks after surgery, mice were habituated to the recording tethers for 30 min in their home-cage for two consecutive days prior to behavioral testing. Mice underwent conditioning and retrieval testing using the standard procedure described above, with the exception that retrieval was conducted in a 30 cm diameter clear acrylic cylinder with an open top to accommodate the cable connecting the head-stage. Thirty minutes prior to the retrieval test, CNO (0.01 mg/mL/kg) was intraperitoneally injected. Electrophysiological and behavioral recordings were acquired using SpikeGadgets main control unit and Trodes software (SpikeGadgets, San Francisco, CA, USA). Unit recordings were carried out using 16-channel digitizing head-stages, sampled at 30 kHz. Behavioral videos were scored offline by an experimenter blind to conditions.

Single units were sorted manually using Offline Sorter v3.0 (Plexon Inc, Dallas, TX, USA) and analyzed using NeuroExplorer, version 5 (Nex Technologies, Colorado Springs, CO, USA) as previously described (Halladay and Blair, 2015; Halladay et al., 2020). Unit data were aligned to CS and freezing events. Freezing was manually scored (by an experimenter blind to experimental group) and resultant time stamps aligned with the neuronal data. Freezing onset was defined as a transition from movement to no visible movement except that required for breathing. Freezing cessation was defined as a transition from freezing to movement. To determine whether units were responsive to the CS, data during a 500 ms window following the start of the CS for each unit were binned in 100 ms bins and normalized to a 1 s baseline defined as the 10 bins immediately prior to the start of the CS. Units with at least two bins of the same sign in the 500 ms following the start of the CS with a value of >1.96 (p<0.05) were considered significantly different from baseline and classified as CS responsive. Unit responsiveness to freezing onset and freezing cessation were analyzed similarly, with the exception that the baseline was shifted from −2 to −1 (rather than −1 to 0) s prior to start of an onset or cessation event to ensure that event and the baseline were temporally separate.

On completion of testing, mice were anesthetized with 2% isoflurane and a current stimulator (S48 Square Pulse Stimulator, Grass Technologies, West Warwick, RI, USA) that delivered 2 s of 40 µA DC current through each electrode to make a small marking lesion. The next day, mice were overdosed via an intraperitoneal injection of 150 mg/kg Euthasol (Henry Schein, Melville, NY, USA) and perfused intracardially with PBS followed by 4% PFA. Brains were left in 4% PFA overnight, then transferred to a 30% sucrose PBS solution for cryoprotection. Coronal sections (50 µm thick) were cut on a cryostat (Leica Biosystems Inc) and mounted onto slides. Tissue was stained with DAPI (Sigma–Aldrich) and imaged using a Keyence BZ-X800 fluorescence microscope (Keyence Corporation of America, Itasca, IL, USA). Mice without viral (i.e., mCherry) expression in the mPFC or correct electrode placement in the IL were removed from the analysis.

Statistical analysis

Differences in freezing and c-Fos counts were analyzed using ANOVA followed by Dunn’s post hoc tests. Differences in z scored single-unit values were analyzed using paired t-tests. Differences in the percentage of recorded units responsive to the CS onset, freezing onset, and freezing cessation were analyzed using non-parametric Fisher’s exact tests. The threshold for statistical significance was set at p<0.05; significance values are shown up to p<0.0001.

Acknowledgements

We are very grateful to Dr. L Ostroff, Dr. N Justice, Dr. S Maren, and Dr. TL Kash for valuable discussions. Research supported by the NIAAA Intramural Research Program.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Lucas R Glover, Email: lucasglover@email.gwu.edu.

Mihaela D Iordanova, Concordia University, Canada.

Kate M Wassum, University of California, Los Angeles, United States.

Funding Information

This paper was supported by the following grant:

  • National Institute on Alcohol Abuse and Alcoholism NIAAA-IRP to Andrew Holmes.

Additional information

Competing interests

No competing interests declared.

Author contributions

Formal analysis, Investigation, Methodology, Writing - review and editing.

Investigation.

Investigation.

Investigation, Writing - review and editing.

Investigation.

Investigation, Writing - review and editing.

Formal analysis, Investigation, Methodology, Writing - review and editing.

Formal analysis, Investigation, Methodology, Writing - review and editing.

Formal analysis, Investigation, Methodology, Writing - review and editing.

Formal analysis, Supervision, Investigation, Methodology, Writing - original draft, Writing - review and editing.

Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology.

Ethics

Animal experimentation: All experimental procedures were approved by the NIAAA (protocol # LBGN-AH-01) and Santa Clara University (SCU AWA: D18-01042) Animal Care and Use Committees and followed the NIH guidelines outlined in 'Using Animals in Intramural Research' and the local Animal Care and Use Committees.

Additional files

Transparent reporting form

Data availability

Some of the data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided for Figure 1. Further data has been uploaded to Dryad (https://doi.org/10.5061/dryad.j9kd51cbn).

The following dataset was generated:

Glover LR, McFadden KM, Bjorni M, Smith SR, Rovero NG, Oreizi-Esfahani S, Yoshida T, Postle AF, Nonaka M, Halladay LR, Holmes A. 2021. mCherry and Gi DREADD electrophysiology for freezing levels during retrieval. Dryad Digital Repository.

References

  1. Asok A, Draper A, Hoffman AF, Schulkin J, Lupica CR, Rosen JB. Optogenetic silencing of a corticotropin-releasing factor pathway from the central amygdala to the bed nucleus of the stria terminalis disrupts sustained fear. Molecular Psychiatry. 2018;23:914–922. doi: 10.1038/mp.2017.79. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bjorni M, Rovero NG, Yang ER, Holmes A, Halladay LR. Phasic signaling in the bed nucleus of the stria terminalis during fear learning predicts within- and across-session cued fear expression. Learning & Memory. 2020;27:83–90. doi: 10.1101/lm.050807.119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Blanchard DC, Blanchard RJ. Innate and conditioned reactions to threat in rats with amygdaloid lesions. Journal of Comparative and Physiological Psychology. 1972;81:281–290. doi: 10.1037/h0033521. [DOI] [PubMed] [Google Scholar]
  4. Bloodgood DW, Sugam JA, Holmes A, Kash TL. Fear extinction requires infralimbic cortex projections to the basolateral amygdala. Translational Psychiatry. 2018;8:60. doi: 10.1038/s41398-018-0106-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Breitfeld T, Bruning JE, Inagaki H, Takeuchi Y, Kiyokawa Y, Fendt M. Temporary inactivation of the anterior part of the bed nucleus of the stria terminalis blocks alarm pheromone-induced defensive behavior in rats. Frontiers in Neuroscience. 2015;9:321. doi: 10.3389/fnins.2015.00321. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Bukalo O, Pinard CR, Holmes A. Mechanisms to medicines: elucidating neural and molecular substrates of fear extinction to identify novel treatments for anxiety disorders. British Journal of Pharmacology. 2014;171:4690–4718. doi: 10.1111/bph.12779. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Bukalo O, Pinard CR, Silverstein S, Brehm C, Hartley ND, Whittle N, Colacicco G, Busch E, Patel S, Singewald N, Holmes A. Prefrontal inputs to the amygdala instruct fear extinction memory formation. Science Advances. 2015;1:e1500251. doi: 10.1126/sciadv.1500251. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Bukalo O, Nonaka M, Weinholtz CA, Mendez A, Taylor WW, Holmes A. Effects of optogenetic photoexcitation of infralimbic cortex inputs to the basolateral amygdala on conditioned fear and extinction. Behavioural Brain Research. 2021;396:112913. doi: 10.1016/j.bbr.2020.112913. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Burgos-Robles A, Kimchi EY, Izadmehr EM, Porzenheim MJ, Ramos-Guasp WA, Nieh EH, Felix-Ortiz AC, Namburi P, Leppla CA, Presbrey KN, Anandalingam KK, Pagan-Rivera PA, Anahtar M, Beyeler A, Tye KM. Amygdala inputs to prefrontal cortex guide behavior amid conflicting cues of reward and punishment. Nature Neuroscience. 2017;20:824–835. doi: 10.1038/nn.4553. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Camp M, Norcross M, Whittle N, Feyder M, D’Hanis W, Yilmazer-Hanke D, Singewald N, Holmes A. Impaired Pavlovian fear extinction is a common phenotype across genetic lineages of the 129 inbred mouse strain. Genes, Brain and Behavior. 2009;8:744–752. doi: 10.1111/j.1601-183X.2009.00519.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Camp MC, Macpherson KP, Lederle L, Graybeal C, Gaburro S, Debrouse LM, Ihne JL, Bravo JA, O'Connor RM, Ciocchi S, Wellman CL, Lüthi A, Cryan JF, Singewald N, Holmes A. Genetic strain differences in learned fear inhibition associated with variation in neuroendocrine, autonomic, and amygdala dendritic phenotypes. Neuropsychopharmacology. 2012;37:1534–1547. doi: 10.1038/npp.2011.340. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Centanni SW, Morris BD, Luchsinger JR, Bedse G, Fetterly TL, Patel S, Winder DG. Endocannabinoid control of the insular-bed nucleus of the stria terminalis circuit regulates negative affective behavior associated with alcohol abstinence. Neuropsychopharmacology. 2019;44:526–537. doi: 10.1038/s41386-018-0257-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Christianson JP, Jennings JH, Ragole T, Flyer JG, Benison AM, Barth DS, Watkins LR, Maier SF. Safety signals mitigate the consequences of uncontrollable stress via a circuit involving the sensory insular cortex and bed nucleus of the stria terminalis. Biological Psychiatry. 2011;70:458–464. doi: 10.1016/j.biopsych.2011.04.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Colmers PLW, Bains JS. Presynaptic mGluRs control the duration of Endocannabinoid-Mediated DSI. The Journal of Neuroscience. 2018;38:10444–10453. doi: 10.1523/JNEUROSCI.1097-18.2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Conrad KL, Louderback KM, Gessner CP, Winder DG. Stress-induced alterations in anxiety-like behavior and adaptations in plasticity in the bed nucleus of the stria terminalis. Physiology & Behavior. 2011;104:248–256. doi: 10.1016/j.physbeh.2011.03.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Daldrup T, Lesting J, Meuth P, Seidenbecher T, Pape HC. Neuronal correlates of sustained fear in the anterolateral part of the bed nucleus of stria terminalis. Neurobiology of Learning and Memory. 2016;131:137–146. doi: 10.1016/j.nlm.2016.03.020. [DOI] [PubMed] [Google Scholar]
  17. Davis M, Walker DL, Miles L, Grillon C. Phasic vs sustained fear in rats and humans: role of the extended amygdala in fear vs anxiety. Neuropsychopharmacology. 2010;35:105–135. doi: 10.1038/npp.2009.109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Dias BG, Banerjee SB, Goodman JV, Ressler KJ. Towards new approaches to disorders of fear and anxiety. Current Opinion in Neurobiology. 2013;23:346–352. doi: 10.1016/j.conb.2013.01.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Do-Monte FH, Manzano-Nieves G, Quiñones-Laracuente K, Ramos-Medina L, Quirk GJ. Revisiting the role of infralimbic cortex in fear extinction with optogenetics. Journal of Neuroscience. 2015;35:3607–3615. doi: 10.1523/JNEUROSCI.3137-14.2015. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Dong HW, Petrovich GD, Swanson LW. Topography of projections from amygdala to bed nuclei of the stria terminalis. Brain Research Reviews. 2001;38:192–246. doi: 10.1016/S0165-0173(01)00079-0. [DOI] [PubMed] [Google Scholar]
  21. Duvarci S, Bauer EP, Paré D. The bed nucleus of the stria terminalis mediates inter-individual variations in anxiety and fear. Journal of Neuroscience. 2009;29:10357–10361. doi: 10.1523/JNEUROSCI.2119-09.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Fanselow MS, Poulos AM. The neuroscience of mammalian associative learning. Annual Review of Psychology. 2005;56:207–234. doi: 10.1146/annurev.psych.56.091103.070213. [DOI] [PubMed] [Google Scholar]
  23. Fendt M, Endres T, Apfelbach R. Temporary inactivation of the bed nucleus of the stria terminalis but not of the amygdala blocks freezing induced by trimethylthiazoline, a component of fox feces. The Journal of Neuroscience. 2003;23:23–28. doi: 10.1523/JNEUROSCI.23-01-00023.2003. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Fitzgerald RD. Effects of partial reinforcement with acid on the classically conditioned salivary response in dogs. Journal of Comparative and Physiological Psychology. 1963;56:1056–1060. doi: 10.1037/h0048204. [DOI] [PubMed] [Google Scholar]
  25. Fitzgerald PJ, Whittle N, Flynn SM, Graybeal C, Pinard CR, Gunduz-Cinar O, Kravitz AV, Singewald N, Holmes A. Prefrontal single-unit firing associated with deficient extinction in mice. Neurobiology of Learning and Memory. 2014;113:69–81. doi: 10.1016/j.nlm.2013.11.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Fitzgerald PJ, Pinard CR, Camp MC, Feyder M, Sah A, Bergstrom HC, Graybeal C, Liu Y, Schlüter OM, Grant SG, Singewald N, Xu W, Holmes A. Durable fear memories require PSD-95. Molecular Psychiatry. 2015;20:901–912. doi: 10.1038/mp.2014.161. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Glangetas C, Girard D, Groc L, Marsicano G, Chaouloff F, Georges F. Stress switches cannabinoid type-1 (CB1) receptor-dependent plasticity from LTD to LTP in the bed nucleus of the stria terminalis. Journal of Neuroscience. 2013;33:19657–19663. doi: 10.1523/JNEUROSCI.3175-13.2013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Glangetas C, Massi L, Fois GR, Jalabert M, Girard D, Diana M, Yonehara K, Roska B, Xu C, Lüthi A, Caille S, Georges F. NMDA-receptor-dependent plasticity in the bed nucleus of the stria terminalis triggers long-term anxiolysis. Nature Communications. 2017;8:14456. doi: 10.1038/ncomms14456. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Glover LR, Schoenfeld TJ, Karlsson RM, Bannerman DM, Cameron HA. Ongoing neurogenesis in the adult dentate gyrus mediates behavioral responses to ambiguous threat cues. PLOS Biology. 2017;15:e2001154. doi: 10.1371/journal.pbio.2001154. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Goode TD, Ressler RL, Acca GM, Miles OW, Maren S. Bed nucleus of the stria terminalis regulates fear to unpredictable threat signals. eLife. 2019;8:e46525. doi: 10.7554/eLife.46525. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Goode TD, Acca GM, Maren S. Threat imminence dictates the role of the bed nucleus of the stria terminalis in contextual fear. Neurobiology of Learning and Memory. 2020;167:107116. doi: 10.1016/j.nlm.2019.107116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Goode TD, Maren S. Role of the bed nucleus of the stria terminalis in aversive learning and memory. Learning & Memory. 2017;24:480–491. doi: 10.1101/lm.044206.116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Grosso A, Santoni G, Manassero E, Renna A, Sacchetti B. A neuronal basis for fear discrimination in the lateral amygdala. Nature Communications. 2018;9:1214. doi: 10.1038/s41467-018-03682-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Halladay LR, Kocharian A, Piantadosi PT, Authement ME, Lieberman AG, Spitz NA, Coden K, Glover LR, Costa VD, Alvarez VA, Holmes A. Prefrontal regulation of punished ethanol Self-administration. Biological Psychiatry. 2020;87:967–978. doi: 10.1016/j.biopsych.2019.10.030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Halladay LR, Blair HT. Distinct ensembles of medial prefrontal cortex neurons are activated by threatening stimuli that elicit excitation vs. inhibition of movement. Journal of Neurophysiology. 2015;114:793–807. doi: 10.1152/jn.00656.2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Halladay LR, Blair HT. Prefrontal infralimbic cortex mediates competition between excitation and inhibition of body movements during pavlovian fear conditioning. Journal of Neuroscience Research. 2017;95:853–862. doi: 10.1002/jnr.23736. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Hammack SE, Todd TP, Kocho-Schellenberg M, Bouton ME. Role of the bed nucleus of the stria terminalis in the acquisition of contextual fear at long or short context-shock intervals. Behavioral Neuroscience. 2015;129:673–678. doi: 10.1037/bne0000088. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Harris JA, Kwok DWS, Gottlieb DA. The partial reinforcement extinction effect depends on learning about nonreinforced trials rather than reinforcement rate. Journal of Experimental Psychology: Animal Learning and Cognition. 2019;45:485–501. doi: 10.1037/xan0000220. [DOI] [PubMed] [Google Scholar]
  39. Haufler D, Nagy FZ, Pare D. Neuronal correlates of fear conditioning in the bed nucleus of the stria terminalis. Learning & Memory. 2013;20:633–641. doi: 10.1101/lm.031799.113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Herman JP, Tasker JG. Paraventricular hypothalamic mechanisms of chronic stress adaptation. Frontiers in Endocrinology. 2016;7:137. doi: 10.3389/fendo.2016.00137. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Holstege G. Some anatomical observations on the projections from the hypothalamus to brainstem and spinal cord: an HRP and autoradiographic tracing study in the cat. The Journal of Comparative Neurology. 1987;260:98–126. doi: 10.1002/cne.902600109. [DOI] [PubMed] [Google Scholar]
  42. Humphreys LG. The effect of random alternation of reinforcement on the acquisition and extinction of conditioned eyelid reactions. Journal of Experimental Psychology. 1939;25:141–158. doi: 10.1037/h0058138. [DOI] [Google Scholar]
  43. Hurley KM, Herbert H, Moga MM, Saper CB. Efferent projections of the infralimbic cortex of the rat. The Journal of Comparative Neurology. 1991;308:249–276. doi: 10.1002/cne.903080210. [DOI] [PubMed] [Google Scholar]
  44. Jennings JH, Sparta DR, Stamatakis AM, Ung RL, Pleil KE, Kash TL, Stuber GD. Distinct extended amygdala circuits for divergent motivational states. Nature. 2013;496:224–228. doi: 10.1038/nature12041. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Johnson SB, Emmons EB, Anderson RM, Glanz RM, Romig-Martin SA, Narayanan NS, LaLumiere RT, Radley JJ. A basal forebrain site coordinates the modulation of endocrine and behavioral stress responses via divergent neural pathways. The Journal of Neuroscience. 2016;36:8687–8699. doi: 10.1523/JNEUROSCI.1185-16.2016. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Johnson SB, Emmons EB, Lingg RT, Anderson RM, Romig-Martin SA, LaLumiere RT, Narayanan NS, Viau V, Radley JJ. Prefrontal-Bed nucleus circuit modulation of a passive coping response set. The Journal of Neuroscience. 2019;39:1405–1419. doi: 10.1523/JNEUROSCI.1421-18.2018. [DOI] [PMC free article] [PubMed] [Google Scholar]
  47. Kalin NH, Shelton SE, Fox AS, Oakes TR, Davidson RJ. Brain regions associated with the expression and contextual regulation of anxiety in primates. Biological Psychiatry. 2005;58:796–804. doi: 10.1016/j.biopsych.2005.05.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Kim H-S, Cho H-Y, Augustine GJ, Han J-H. Selective control of fear expression by optogenetic manipulation of infralimbic cortex after extinction. Neuropsychopharmacology. 2016;41:1261–1273. doi: 10.1038/npp.2015.276. [DOI] [PMC free article] [PubMed] [Google Scholar]
  49. Lange MD, Daldrup T, Remmers F, Szkudlarek HJ, Lesting J, Guggenhuber S, Ruehle S, Jüngling K, Seidenbecher T, Lutz B, Pape HC. Cannabinoid CB1 receptors in distinct circuits of the extended amygdala determine fear responsiveness to unpredictable threat. Molecular Psychiatry. 2017;22:1422–1430. doi: 10.1038/mp.2016.156. [DOI] [PubMed] [Google Scholar]
  50. Laurent V, Westbrook RF. Inactivation of the infralimbic but not the prelimbic cortex impairs consolidation and retrieval of fear extinction. Learning & Memory. 2009;16:520–529. doi: 10.1101/lm.1474609. [DOI] [PubMed] [Google Scholar]
  51. Lay BP, Pitaru AA, Boulianne N, Esber GR, Iordanova MD. Different methods of fear reduction are supported by distinct cortical substrates. eLife. 2020;9:e55294. doi: 10.7554/eLife.55294. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Lebow MA, Chen A. Overshadowed by the amygdala: the bed nucleus of the stria terminalis emerges as key to psychiatric disorders. Molecular Psychiatry. 2016;21:450–463. doi: 10.1038/mp.2016.1. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Luyten L, Casteels C, Vansteenwegen D, van Kuyck K, Koole M, Van Laere K, Nuttin B. Micro-positron emission tomography imaging of rat brain metabolism during expression of contextual conditioning. Journal of Neuroscience. 2012;32:254–263. doi: 10.1523/JNEUROSCI.3701-11.2012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Marcinkiewcz CA, Mazzone CM, D'Agostino G, Halladay LR, Hardaway JA, DiBerto JF, Navarro M, Burnham N, Cristiano C, Dorrier CE, Tipton GJ, Ramakrishnan C, Kozicz T, Deisseroth K, Thiele TE, McElligott ZA, Holmes A, Heisler LK, Kash TL. Serotonin engages an anxiety and fear-promoting circuit in the extended amygdala. Nature. 2016;537:97–101. doi: 10.1038/nature19318. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Marek R, Sun Y, Sah P. Neural circuits for a top-down control of fear and extinction. Psychopharmacology. 2019;236:313–320. doi: 10.1007/s00213-018-5033-2. [DOI] [PubMed] [Google Scholar]
  56. McDonald AJ, Shammah-Lagnado SJ, Shi C, Davis M. Cortical afferents to the extended amygdala. Annals of the New York Academy of Sciences. 1999;877:309–338. doi: 10.1111/j.1749-6632.1999.tb09275.x. [DOI] [PubMed] [Google Scholar]
  57. McHugh SB, Barkus C, Lima J, Glover LR, Sharp T, Bannerman DM. SERT and uncertainty: serotonin transporter expression influences information processing biases for ambiguous aversive cues in mice. Genes, Brain and Behavior. 2015;14:330–336. doi: 10.1111/gbb.12215. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Meyer HC, Odriozola P, Cohodes EM, Mandell JD, Li A, Yang R, Hall BS, Haberman JT, Zacharek SJ, Liston C, Lee FS, Gee DG. Ventral Hippocampus interacts with prelimbic cortex during inhibition of threat response via learned safety in both mice and humans. PNAS. 2019;116:26970–26979. doi: 10.1073/pnas.1910481116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  59. Miguez G, Witnauer JE, Miller RR. The role of contextual associations in producing the partial reinforcement acquisition deficit. Journal of Experimental Psychology: Animal Behavior Processes. 2012;38:40–51. doi: 10.1037/a0024410. [DOI] [PubMed] [Google Scholar]
  60. Milad MR, Quirk GJ. Fear extinction as a model for translational neuroscience: ten years of progress. Annual Review of Psychology. 2012;63:129–151. doi: 10.1146/annurev.psych.121208.131631. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Miles OW, Maren S. Role of the bed nucleus of the stria terminalis in PTSD: insights from preclinical models. Frontiers in Behavioral Neuroscience. 2019;13:68. doi: 10.3389/fnbeh.2019.00068. [DOI] [PMC free article] [PubMed] [Google Scholar]
  62. Pape HC, Pare D. Plastic synaptic networks of the amygdala for the acquisition, expression, and extinction of conditioned fear. Physiological Reviews. 2010;90:419–463. doi: 10.1152/physrev.00037.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  63. Paxinos KBJ, Franklin G. The Mouse Brain in Stereotaxic Coordinates. Second Edition. London: Academic Press; 2001. [Google Scholar]
  64. Penzo MA, Robert V, Tucciarone J, De Bundel D, Wang M, Van Aelst L, Darvas M, Parada LF, Palmiter RD, He M, Huang ZJ, Li B. The paraventricular thalamus controls a central amygdala fear circuit. Nature. 2015;519:455–459. doi: 10.1038/nature13978. [DOI] [PMC free article] [PubMed] [Google Scholar]
  65. Poulos AM, Ponnusamy R, Dong HW, Fanselow MS. Compensation in the neural circuitry of fear conditioning awakens learning circuits in the bed nuclei of the stria terminalis. PNAS. 2010;107:14881–14886. doi: 10.1073/pnas.1005754107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  66. Radley JJ, Gosselink KL, Sawchenko PE. A discrete GABAergic relay mediates medial prefrontal cortical inhibition of the neuroendocrine stress response. Journal of Neuroscience. 2009;29:7330–7340. doi: 10.1523/JNEUROSCI.5924-08.2009. [DOI] [PMC free article] [PubMed] [Google Scholar]
  67. Radley JJ, Anderson RM, Hamilton BA, Alcock JA, Romig-Martin SA. Chronic stress-induced alterations of dendritic spine subtypes predict functional decrements in an hypothalamo-pituitary-adrenal-inhibitory prefrontal circuit. Journal of Neuroscience. 2013;33:14379–14391. doi: 10.1523/JNEUROSCI.0287-13.2013. [DOI] [PMC free article] [PubMed] [Google Scholar]
  68. Radley JJ, Sawchenko PE. A common substrate for prefrontal and hippocampal inhibition of the neuroendocrine stress response. Journal of Neuroscience. 2011;31:9683–9695. doi: 10.1523/JNEUROSCI.6040-10.2011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  69. Ramaker MJ, Dulawa SC. Identifying fast-onset antidepressants using rodent models. Molecular Psychiatry. 2017;22:656–665. doi: 10.1038/mp.2017.36. [DOI] [PubMed] [Google Scholar]
  70. Rawlins JN, Feldon J, Ursin H, Gray JA. Resistance to extinction after schedules of partial delay or partial reinforcement in rats with hippocampal lesions. Experimental Brain Research. 1985;59:273–281. doi: 10.1007/BF00230907. [DOI] [PubMed] [Google Scholar]
  71. Rescorla RA. Spontaneous recovery after reversal and partial reinforcement. Learning & Behavior. 2007;35:191–200. doi: 10.3758/BF03206425. [DOI] [PubMed] [Google Scholar]
  72. Rosen GJ, de Vries GJ, Goldman SL, Goldman BD, Forger NG. Distribution of oxytocin in the brain of a eusocial rodent. Neuroscience. 2008;155:809–817. doi: 10.1016/j.neuroscience.2008.05.039. [DOI] [PMC free article] [PubMed] [Google Scholar]
  73. Roughley S, Killcross S. Loss of hierarchical control by occasion setters following lesions of the prelimbic and infralimbic medial prefrontal cortex in rats. Brain Sciences. 2019;9:48. doi: 10.3390/brainsci9030048. [DOI] [PMC free article] [PubMed] [Google Scholar]
  74. Sangha S, Robinson PD, Greba Q, Davies DA, Howland JG. Alterations in reward, fear and safety cue discrimination after inactivation of the rat prelimbic and infralimbic cortices. Neuropsychopharmacology. 2014;39:2405–2413. doi: 10.1038/npp.2014.89. [DOI] [PMC free article] [PubMed] [Google Scholar]
  75. Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, Preibisch S, Rueden C, Saalfeld S, Schmid B, Tinevez JY, White DJ, Hartenstein V, Eliceiri K, Tomancak P, Cardona A. Fiji: an open-source platform for biological-image analysis. Nature Methods. 2012;9:676–682. doi: 10.1038/nmeth.2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
  76. Sengupta A, Holmes A. A discrete dorsal raphe to basal amygdala 5-HT circuit calibrates aversive memory. Neuron. 2019;103:489–505. doi: 10.1016/j.neuron.2019.05.029. [DOI] [PMC free article] [PubMed] [Google Scholar]
  77. Sharpe MJ, Killcross S. Modulation of attention and action in the medial prefrontal cortex of rats. Psychological Review. 2018;125:822–843. doi: 10.1037/rev0000118. [DOI] [PubMed] [Google Scholar]
  78. Sullivan GM, Apergis J, Bush DE, Johnson LR, Hou M, Ledoux JE. Lesions in the bed nucleus of the stria terminalis disrupt corticosterone and freezing responses elicited by a contextual but not by a specific cue-conditioned fear stimulus. Neuroscience. 2004;128:7–14. doi: 10.1016/j.neuroscience.2004.06.015. [DOI] [PubMed] [Google Scholar]
  79. Tillman RM, Stockbridge MD, Nacewicz BM, Torrisi S, Fox AS, Smith JF, Shackman AJ. Intrinsic functional connectivity of the central extended amygdala. Human Brain Mapping. 2018;39:1291–1312. doi: 10.1002/hbm.23917. [DOI] [PMC free article] [PubMed] [Google Scholar]
  80. Tovote P, Fadok JP, Lüthi A. Neuronal circuits for fear and anxiety. Nature Reviews Neuroscience. 2015;16:317–331. doi: 10.1038/nrn3945. [DOI] [PubMed] [Google Scholar]
  81. Triana-Del Río R, van den Burg E, Stoop R, Hegoburu C. Acute and long-lasting effects of oxytocin in cortico-limbic circuits: consequences for fear recall and extinction. Psychopharmacology. 2019;236:339–354. doi: 10.1007/s00213-018-5030-5. [DOI] [PubMed] [Google Scholar]
  82. Tsetsenis T, Ma XH, Lo Iacono L, Beck SG, Gross C. Suppression of conditioning to ambiguous cues by pharmacogenetic inhibition of the dentate gyrus. Nature Neuroscience. 2007;10:896–902. doi: 10.1038/nn1919. [DOI] [PMC free article] [PubMed] [Google Scholar]
  83. Verma D, Tasan R, Sperk G, Pape HC. Neuropeptide Y2 receptors in anteroventral BNST control remote fear memory depending on extinction training. Neurobiology of Learning and Memory. 2018;149:144–153. doi: 10.1016/j.nlm.2018.01.011. [DOI] [PubMed] [Google Scholar]
  84. Vertes RP. Differential projections of the infralimbic and prelimbic cortex in the rat. Synapse. 2004;51:32–58. doi: 10.1002/syn.10279. [DOI] [PubMed] [Google Scholar]
  85. Waddell J, Morris RW, Bouton ME. Effects of bed nucleus of the stria terminalis lesions on conditioned anxiety: aversive conditioning with long-duration conditional stimuli and reinstatement of extinguished fear. Behavioral Neuroscience. 2006;120:324–336. doi: 10.1037/0735-7044.120.2.324. [DOI] [PubMed] [Google Scholar]
  86. Walker DL, Miles LA, Davis M. Selective participation of the bed nucleus of the stria terminalis and CRF in sustained anxiety-like versus phasic fear-like responses. Progress in Neuro-Psychopharmacology and Biological Psychiatry. 2009;33:1291–1308. doi: 10.1016/j.pnpbp.2009.06.022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  87. Xu HY, Liu YJ, Xu MY, Zhang YH, Zhang JX, Wu YJ. Inactivation of the bed nucleus of the stria terminalis suppresses the innate fear responses of rats induced by the odor of cat urine. Neuroscience. 2012;221:21–27. doi: 10.1016/j.neuroscience.2012.06.056. [DOI] [PubMed] [Google Scholar]
  88. Zimmerman JM, Maren S. The bed nucleus of the stria terminalis is required for the expression of contextual but not auditory freezing in rats with basolateral amygdala lesions. Neurobiology of Learning and Memory. 2011;95:199–205. doi: 10.1016/j.nlm.2010.11.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  89. Zingg B, Chou XL, Zhang ZG, Mesik L, Liang F, Tao HW, Zhang LI. AAV-Mediated anterograde transsynaptic tagging: mapping corticocollicular Input-Defined neural pathways for defense behaviors. Neuron. 2017;93:33–47. doi: 10.1016/j.neuron.2016.11.045. [DOI] [PMC free article] [PubMed] [Google Scholar]

Decision letter

Editor: Mihaela D Iordanova1
Reviewed by: Mihaela D Iordanova2, Michael A McDannald3, Rebecca Shansky4

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

The paper by Glover and colleagues elegantly combines behavioral, immunohistochemistry, chemogenetic and recording techniques to elucidate the neural mechanisms of partial compared to continuous reinforcement. The neural regulation of partial reinforcement in the fear domain studied in this paper links well with extensive work done in the field of extinction of fear. Therefore, the discoveries reported in this paper will be sure to guide future research directions in the field.

Decision letter after peer review:

Thank you for submitting your article "A prefrontal-bed nucleus of the stria terminalis circuit limits fear to uncertain threat" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Mihaela D Iordanova as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Kate Wassum as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Michael A McDannald (Reviewer #2); Rebecca Shansky (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

Summary:

In a series of studies, Glover and colleagues examined the role of the prefrontal cortex and its input to the bed nucleus of stria terminalis pathway in fear to uncertain threats. The authors first demonstrate that partially reinforced cues support lower fear levels than fully reinforced cues, this effect is context-specific and is dependent on mouse strain. Comprehensive c-fos mapping is used to show that differential activation of the prefrontal cortex (prelimbic [PL] and infralimbic regions [IL]) and the bed nucleus of the stria terminalis (BNST) in threat uncertainty. DREADD inhibition of BNST-projecting mPFC neurons increased freezing to the partially reinforced cue. Electrophysiological examination revealed that IL neurons are preferentially activated in the PRF group at CS-ON and the number of these neurons is greatly reduced when the mPFC-BNST pathway is inactive. All reviewers agreed that the data presented in the manuscript are exciting and of great importance in that it uncovered a novel role for the mPFC-BNST pathway in threat uncertainty. The multifaceted approach is also a major strength.

Through the review process a number of essential revisions were identified. These are specified below. None of the revision require additional data collection, but some do require additional analyses.

Essential revisions:

1) Some conceptual issues. These can be addressed in the text.

a) The role of context as a modulator. The claim that the context modulates the PRF effect is not strong so it would be best to modify this claim. Firstly, the order of testing is not counterbalanced (mice tested in B before tested in A). Because of the order and because the difference in freezing between PRF and FRF during the baseline (i.e. context) in context B in Figure 1D does not seem to be very reliable (we don't see it in Figure 1B), it is hard to know what role the context really plays. That is, this masks potential differences in fear to A and B, which could account for the lack of PRF effect in Context A (through summation). What happens if the baseline for each mouse is subtracted from its individual CS freezing score? Is the PRF effect still lacking in context A? It is important to check this to see if the contextual effect is robust.

b) The DREADDs pathway study. The mPFC-BNST hM4Di data show that freezing in the PRF group is increased, reaching the levels of the FRF group. Could the lack of effect on FRF be due to ceiling levels? The trial by trial data may speak to this provided a decline in fear is seen across trials in at least one of the groups. Further, Figure 3H is suggestive of a difference between the FRF hM4Di and the mCherry groups. This matters because it may shed light on potential ceiling effect in the mPFC-BNST hM4Di behavioural data in the FRF groups. Further, an exciting difference would be between the mCherry groups. Was this tested?

2) Statistics.

a) Only overall ANOVAs are reported for the Fos+ analyses for each brain region, but the figures show asterisks suggestive of pairwise comparison tests. These should be reported in the text for CON, PRE and FRE for each brain area. This is also the case for the CS test in the mPFC-BNST pathway study.

b) Figure 3H is suggestive of a difference between the FRF hM4Di and the mCherry groups. Were the data analyzed together? An ANOVA seems appropriate here. Unsure why a Fisher's exact test was chosen and how it was conducted – for the 2x2 factorial or for individual cells of the table.

c) Direct comparisons between the PRF and FRF in the recording study are necessary.

d) Individual data points should be included in all the figures where possible.

3) Additional neural analyses.

a) It would be more helpful (beyond the rasters provided in Figure 3F) to see a summary heat plot of all recorded units, in all conditions, with activity aligned to cue onset/offset as well as freezing onset/offset. This will give the reader a much better idea of the patterns of activity observed in all neurons in all conditions. An example would be a heat plot like that in Figure 5C of Beyeler et al., 2018 Organization of valence-encoding and projection-defined neurons in the BLA, Cell Reports. If single-unit activity was suppressed by DREADD inhibition a heat plot would nicely illustrate this.

b) What group/condition was used for population firing in Figures 3G and 3I. Were these the PRF/mCherry neurons? Or was this all PRF neurons combined? (moot point for 3J if that is the case). This should not only be specified but the population should be plotted for all groups/conditions. This would allow the reader to see if cue responses to PRF and FRF were equivalent in the mCherry conditions – a result that is interesting no matter the outcome. This approach would provide less info for the data in 3J, because no PRF/hM4Di neurons were observed, but plotting the observed populations would still be helpful. The insets for 3G and 3I were not informative and should be removed as these neurons were selected because they were responsive.

c) Please provide waveform and firing rate properties for the neurons recorded. Did they differ between the groups? This is borne out of the between-subjects design of the recording experiments as different neuron types could have been sampled in each condition.

d) Full histology for electrode placements for all mice from which neurons were obtained should be provided. It would also be helpful to include schematics showing viral spread for all subjects (e.g., overlaid traces).

4) Materials and methods. Specific information about the number of mice and neurons recorded from in each group/condition need to be reported. Ranges are provided in the Figure 3 caption (n=17-25 units/group, from 3 mice per group/virus). But considering that the main statistics are based on the % of numbers, specific details need to be provided.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "A prefrontal-bed nucleus of the stria terminalis circuit limits fear to uncertain threat" for consideration by eLife. Your article has been reviewed by two peer reviewers, including Mihaela D Iordanova as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Kate Wassum as the Senior Editor.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

Summary:

This is a revised version of the paper by Glover et al. We would like to reiterate the interest and excitement in the paper, which uncovers key brain areas and pathways that are involved in partial reinforcement in fear. While we appreciate that the authors have addressed the majority of the comments, the requests for additional analyses of the recording data were not well addressed well. The single-unit results cannot be meaningfully interpreted without additional analyses and visualizations as requested in the previous decision letter. Therefore, these concerns remain and we ask the authors to address them.

Essential revisions:

1) Please provide a summary heat plot of all recorded units, in all conditions, with activity aligned to cue onset/offset as well as freezing onset/offset. An example would be a heat plot like that in Figure 5C of Beyeler et al., 2018 Organization of valence-encoding and projection-defined neurons in the BLA, Cell Reports. This will give the reader a much better idea of the patterns of activity observed in all neurons in all conditions. If single-unit activity was suppressed by DREADD inhibition a heat plot would nicely illustrate this.

2) Population line graphs must be separately plotted for all groups/conditions when it is possible. This is necessary for the reader to see if cue responses to PRF and FRF were equivalent in the mCherry conditions – a result that is interesting no matter the outcome.

3) One other concern that remains is the possible ceiling levels in the DREADDs experiment. Ceiling in Figure 3—figure supplement 1 panel C and Figure 3—figure supplement 2 in panel C – Does the Figure 3—figure supplement 2C show the trial data for Figure 3—figure supplement 1C? This suggests ceiling across the board for all groups except PRF mCherry. This possibility needs to be reflected in the text as the possibility that the mPFC-BNST pathway could regulate fear in both cases remains.

4) Please report exact p values in all instances.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your article "A prefrontal-bed nucleus of the stria terminalis circuit limits fear to uncertain threat" for consideration by eLife. Your article has been reviewed by two peer reviewers, including Mihaela D Iordanova as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Kate Wassum as the Senior Editor.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, we are asking editors to accept without delay manuscripts, like yours, that they judge can stand as eLife papers without additional data, even if they feel that they would make the manuscript stronger. Thus the revisions requested below only address clarity and presentation.

Summary:

In a series of studies, Glover and colleagues examined the role of the prefrontal cortex and its input to the bed nucleus of stria terminalis pathway in fear to uncertain threats. The authors first demonstrate that partially reinforced cues support lower fear levels than fully reinforced cues. Comprehensive c-fos mapping is used to show that differential activation of the prefrontal cortex (prelimbic [PL] and infralimbic regions [IL]) and the bed nucleus of the stria terminalis (BNST) in threat uncertainty. DREADD inhibition of BNST-projecting mPFC neurons increased freezing to the partially reinforced cue. Electrophysiological examination show that mPFC neurons activated during PRF and FRF are reduced when the mPFC-BNST pathway is inactive. This is a second revision of the manuscript.

Essential revisions

We thank the authors for addressing all of the comments in the previous decision letter which pertained primarily to the analyses of the recording data. The reviewers have looked closely at the electrophysiology data and have noted that those data are rather preliminary in nature. Specifically, there are two main problems.

1) The data appeared underpowered: n=3 CS units in PRF/mCherry, n=6 units in PRF/hM4Di, n=1 units in FRF/mCherry, n=1 units in FRF/hM4Di.

2) The recording data only partially support the hypotheses of the paper. In some instances, the proportion of neurons do not support the hypotheses. For example, neither the heat plot nor the stats support the statement “Thus, IL cells were preferentially activated to the CS after PRF”. Further, the magnitude of the firing reveal a different pattern from the proportion. The former show

a) mPFC neurons show greatest CS onset firing to FRF

b) CS onset firing to FRF is diminished by hM4Di

c) mPFC neurons show equivalent firing increases to freezing cessation

d) Freezing cessation firing to FRF appears to be diminished by hM4Di

Other example of statements that are not supported by that data include, and therefore need to be revised:

"Thus, IL cells were preferentially activated to the CS after PRF and this activation was abolished by inhibition of the mPFC→BNST pathway."

"Thus, IL cells not only respond to presentation of the CS after PRF, but also to the ongoing behavior of the mouse and specifically during the transition from freezing to movement. As with CS responses, these neuronal correlates of behavior are absent when mPFC projections to the BNST are inhibited."

"it is tempting to conclude that this activity is a neuronal representation or even a causal driver of the lesser expression of PRF".

Ideally, the authors would have a full recording dataset. We recognize this may not feasible at this stage. So we give the authors two options: (1) To revise the language in the manuscript so that it can more accurately reflect the recording data, (2) To remove the behavioral electrophysiology data altogether.

eLife. 2020 Dec 15;9:e60812. doi: 10.7554/eLife.60812.sa2

Author response

Essential revisions:

1) Some conceptual issues. These can be addressed in the text.

a) The role of context as a modulator. The claim that the context modulates the PRF effect is not strong so it would be best to modify this claim. Firstly, the order of testing is not counterbalanced (mice tested in B before tested in A). Because of the order and because the difference in freezing between PRF and FRF during the baseline (i.e. context) in context B in Figure 1D does not seem to be very reliable (we don't see it in Figure 1B), it is hard to know what role the context really plays. That is, this masks potential differences in fear to A and B, which could account for the lack of PRF effect in Context A (through summation). What happens if the baseline for each mouse is subtracted from its individual CS freezing score? Is the PRF effect still lacking in context A? It is important to check this to see if the contextual effect is robust.

On reflection, we share the concerns regarding the open questions posed by these data and the experimental design used to obtain them – clearly follow up experiments are needed. Given their preliminary nature, we have removed them from the manuscript.

b) The DREADDs pathway study. The mPFC-BNST hM4Di data show that freezing in the PRF group is increased, reaching the levels of the FRF group. Could the lack of effect on FRF be due to ceiling levels? The trial by trial data may speak to this provided a decline in fear is seen across trials in at least one of the groups. Further, Figure 3H is suggestive of a difference between the FRF hM4Di and the mCherry groups. This matters because it may shed light on potential ceiling effect in the mPFC-BNST hM4Di behavioural data in the FRF groups. Further, an exciting difference would be between the mCherry groups. Was this tested?

We have looked at the trial by trial data for CS retrieval and there is no group difference (2x2 ANOVA P>.05) that would speak to the question of a potential ceiling effect in the FRF groups, as we now state as follows: “Examination of the trial by trial freezing during retrieval indicated no significant group differences, despite a trend for decreasing freezing across trials in the mCherry PRF group (ANOVA conditioning-type effect: P>.05; virus: P>.05; interaction: P>.05)…” The trend for an across-trial reduction in the FRF mCherry group that would be interesting to follow-up, for example with formal multi-trial extinction procedures.

2) Statistics.

a) Only overall ANOVAs are reported for the Fos+ analyses for each brain region, but the figures show asterisks suggestive of pairwise comparison tests. These should be reported in the text for CON, PRE and FRE for each brain area. This is also the case for the CS test in the mPFC-BNST pathway study.

We now indicate pairwise differences (performed via Newman Keuls post hoc tests, as indicated in the Statistical analysis section), for example as follows: “Notably, however, CS-evoked freezing was lower in the PRF, relative to FRF, group (ANOVA group-effect: F(2,17)=53.02, P<.01; CS: F(1,17)=216.90, P<.01; interaction: F(2,17)=25.51, P<.01, followed by post hoc tests: CS vs PRF P<.05, CS vs FRF P<.05, PRF vs FRF P<.05).”

b) Figure 3H is suggestive of a difference between the FRF hM4Di and the mCherry groups. Were the data analyzed together? An ANOVA seems appropriate here. Unsure why a Fisher's exact test was chosen and how it was conducted – for the 2x2 factorial or for individual cells of the table.

We performed Fisher’s exact test to compare the percentage data in Figure 3H and 3J, rather than ANOVA, because these data are categorical and nonparametric – i.e., the percentage of cells with event-related activity in each of the 4 groups (meaning there is no variability). We do, however, now report an overall Fisher’s exact when all 4 groups are considered, as follows: “This revealed a higher proportion of CS-ON units in the mCherry than hM4Di groups for PRF mice (Fisher’s exact test: P<.05) but no differences for FRF mice (Fisher’s exact test: P>.05)…” and “When the percentage of these cell-types were compared across groups (Fisher’s exact test: P<.05), there was a higher percentage of Freeze-OFF units in the mCherry PRF group than in the hM4Di PRF group (Fisher’s exact test: P<.01), whereas there was no group difference in FRF mice (Fisher’s exact test: P>.05)…”

c) Direct comparisons between the PRF and FRF in the recording study are necessary.

We do, however, now report the Fisher’s exact comparing the PRF and FFR groups, as follows: “…and no difference between PRF and FRF groups, irrespective or virus group, despite a trend for (Fisher’s exact test in mCherry: P>.05; in hM4Di: P>.05) (Figure 3H)…” and “…and no difference between PRF and FRF groups in either the mCherry (Fisher’s exact test: P>.05) or hM4Di (Fisher’s exact test: P>.05) virus conditions (Figure 3J).”

d) Individual data points should be included in all the figures where possible.

As requested, individual data points are now included for the main figures, where possible.

3) Additional neural analyses.

a) It would be more helpful (beyond the rasters provided in Figure 3F) to see a summary heat plot of all recorded units, in all conditions, with activity aligned to cue onset/offset as well as freezing onset/offset. This will give the reader a much better idea of the patterns of activity observed in all neurons in all conditions. An example would be a heat plot like that in Figure 5C of Beyeler et al., 2018 Organization of valence-encoding and projection-defined neurons in the BLA, Cell Reports. If single-unit activity was suppressed by DREADD inhibition a heat plot would nicely illustrate this.

While we agree that heat maps can provide a nice visual illustration of the temporal dynamics of unit activity, especially when driven by a temporally precise artificial stimulus as in the Beyeler et al. example, but we feel that showing these for all neurons is somewhat redundant to the example raster plot (Figure 3F) and peri-event histograms we already now show for all groups (Figure 3G,I), as requested (point #4 above).

b) What group/condition was used for population firing in Figures 3G and 3I. Were these the PRF/mCherry neurons? Or was this all PRF neurons combined? (moot point for 3J if that is the case). This should not only be specified but the population should be plotted for all groups/conditions. This would allow the reader to see if cue responses to PRF and FRF were equivalent in the mCherry conditions – a result that is interesting no matter the outcome. This approach would provide less info for the data in 3J, because no PRF/hM4Di neurons were observed, but plotting the observed populations would still be helpful. The insets for 3G and 3I were not informative and should be removed as these neurons were selected because they were responsive.

The data depicted in Figures 3G and 3I were all mice (PRF and FRF, mCherry and hM4Di) combined – the purpose to show the whole recorded population to first illustrate that IL neurons signal CS onset and freezing offset, and then subsequently show how this signaling varies between groups (Figures 3H and 3J). For this reason and because there are no differences between the percentage of phasic neurons in the mCherry groups (see point #5 above), we feel retain showing all groups combined and note this in the figure and figure legends. We maintain that even though they show event-responsive cells, the insets provided valuable information because they confirm that the event related activity is statistically robust (which does not necessarily follow from simply selecting cells that are event-responsive). Nonetheless, we have removed these and now report the values in the text, as follows: “Overall, CS-ON units showed a significant change in neuronal activity in response to the CS (baseline: 0.15±0.35, post-CS: 1.43±0.55, paired t-test: t(10)=6.51, P<.01) (Figure 3G).” And “These units, classified as Freeze-ON and Freeze-OFF, respectively, showed a significant change in baseline-normalized activity (Freeze-ON baseline: -0.81±0.47, post-event: -2.12±0.52, paired t-test: t(10)=4.60, P<.01, Freeze-OFF baseline: 0.73±0.43, post-event: 1.67±0.40, paired t-test: t(12)=8.54, P<.01) (Figure 3I).”

c) Please provide waveform and firing rate properties for the neurons recorded. Did they differ between the groups? This is borne out of the between-subjects design of the recording experiments as different neuron types could have been sampled in each condition.

The firing rates were given in the original manuscript, as follows “The average firing rate units did not differ between groups (FRF mCherry: 4.10±0.64, FRF hM4Di: 3.45±0.49, FRF mCherry: 2.67±0.66, FRF hM4Di: 1.67±0.34).”

While we agree that examining waveform in an effort to parse the molecular identity of the recorded cells would be of potential interest, but we do not feel confident that our data allow us to confidently explore this question given the small fraction of cells that would be putatively identifiable interneurons; i.e., it would be unlikely that we have sufficient numbers of interneurons recorded to make any strong inferences about their patterns of activity in this task. And though we recognize it is done in some studies, we did not perform waveforms analysis to categorize neurons as principal versus interneurons because we are not confident categorizing firing rate and waveform per se provides a reliable indicator of cortical cell type and would instead want to apply the current standard for categorization – optogenetic phototagging in an, e.g., parvalbumin-positive, interneuron Cre-driver line.

d) Full histology for electrode placements for all mice from which neurons were obtained should be provided. It would also be helpful to include schematics showing viral spread for all subjects (e.g., overlaid traces).

As requested, we now show schematics for electrode placements and viral spread.

4) Materials and methods. Specific information about the number of mice and neurons recorded from in each group/condition need to be reported. Ranges are provided in the Figure 3 caption (n=17-25 units/group, from 3 mice per group/virus). But considering that the main statistics are based on the % of numbers, specific details need to be provided.

The number of units and mice per group is now specified, as follows: “(n=17 units in PRF/mCherry, n=25 units in PRF/hM4Di, n=20 units in FRF/mCherry, n=17 units in FRF/hM4Di, from 3 mice per group/virus).”

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Essential revisions:

1) Please provide a summary heat plot of all recorded units, in all conditions, with activity aligned to cue onset/offset as well as freezing onset/offset. An example would be a heat plot like that in Figure 5C of Beyeler et al., 2018 Organization of valence-encoding and projection-defined neurons in the BLA, Cell Reports. This will give the reader a much better idea of the patterns of activity observed in all neurons in all conditions. If single-unit activity was suppressed by DREADD inhibition a heat plot would nicely illustrate this.

We now provide a summary heat plot of all recorded units, in all conditions, with activity aligned to cue onset as well as freezing onset/offset in Figure 3—figure supplement 3. These provide a further visual demonstration of how DREADD inhibition suppresses CS and Freeze-OFF related activity, selectively in the PRF group – thank you for the useful suggestion.

2) Population line graphs must be separately plotted for all groups/conditions when it is possible. This is necessary for the reader to see if cue responses to PRF and FRF were equivalent in the mCherry conditions – a result that is interesting no matter the outcome.

As requested, line graphs are now separately plotted for all groups/conditions in Figure 3 and Figure 3—figure supplement 7. We present these along with the original all-group averages and discuss the graphs as follows: “Peak responses occurred withing 200-300 seconds of CS onset and were highest in the mCherry FRF group (Figure 3H). However, when the percentage of CS-ON units was calculated and compared across the conditioning and virus groups (Fisher’s exact test: P<.05), this revealed a higher proportion of CS-ON units in the mCherry than hM4Di groups for PRF mice (Fisher’s exact test: P=.0122), but no differences between virus groups in the FRF mice (Fisher’s exact test: P=.6090), and no difference between PRF and FRF groups, irrespective or virus group (Fisher’s exact test in mCherry: P=.2510; in hM4Di: P=1.000) (Figure 3I),” and “Freeze-ON units displayed a decreased firing rate at freezing onset, which was most evident in both of the mCherry groups, while Freeze-OFF units increased firing rate at the cessation of freezing in both groups. (Figure 3K, Figure 3—figure supplement 3).”

3) One other concern that remains is the possible ceiling levels in the DREADDs experiment. Ceiling in Figure 3—figure supplement 1 panel C and Figure 3—figure supplement 2 in panel C – Does Figure 3—figure supplement 2C show the trial data for Figure 3—figure supplement 1C? This suggests ceiling across the board for all groups except PRF mCherry. This possibility needs to be reflected in the text as the possibility that the mPFC-BNST pathway could regulate fear in both cases remains.

That is correct, Figure 3—figure supplement 2C shows the trial data for Figure 3—figure supplement 1C. In the text, we now refer to the figure, as follows: “Examination of the trial-by-trial freezing during retrieval indicated no significant trial-related differences in freezing, despite a trend for decreasing freezing across trials in the mCherry PRF group (ANOVA trial-effect: F(5,145)=1.83, P=.1098; group-effect: F(3,29)=14.15, P<.0001; trial x group interaction: F(15,145)=1.04, P=.4213) (Figure 3—figure supplement 1).” We also now discuss the possibility of a ceiling effect, as follows: “These data show that inhibition of mPFC→BNST neurons increases freezing to a PRF CS. This finding suggests engagement of these mPFC→BNST neurons limits the expression of PRF, though it remains possible that inhibition of these neurons also produces an increase in PRF expression which may have been masked due to high (“ceiling”) levels of freezing.”

4) Please report exact p values in all instances.

All p values are now reported, up to a threshold of P=.0001, as stated under “Statistical analysis,” as follows: “The threshold for statistical significance was set at P<0.05; significance values are shown up to P<.0001.”

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Essential revisions

We thank the authors for addressing all of the comments in the previous decision letter which pertained primarily to the analyses of the recording data. The reviewers have looked closely at the electrophysiology data and have noted that those data are rather preliminary in nature. Specifically, there are two main problems.

1) The data appeared underpowered: n=3 CS units in PRF/mCherry, n=6 units in PRF/hM4Di, n=1 units in FRF/mCherry, n=1 units in FRF/hM4Di.

2) The recording data only partially support the hypotheses of the paper. In some instances, the proportion of neurons do not support the hypotheses. For example, neither the heat plot nor the stats support the statement “Thus, IL cells were preferentially activated to the CS after PRF”. Further, the magnitude of the firing reveal a different pattern from the proportion. The former show

a) mPFC neurons show greatest CS onset firing to FRF

b) CS onset firing to FRF is diminished by hM4Di

c) mPFC neurons show equivalent firing increases to freezing cessation

d) Freezing cessation firing to FRF appears to be diminished by hM4Di

Other example of statements that are not supported by that data include, and therefore need to be revised:

"Thus, IL cells were preferentially activated to the CS after PRF and this activation was abolished by inhibition of the mPFC→BNST pathway."

"Thus, IL cells not only respond to presentation of the CS after PRF, but also to the ongoing behavior of the mouse and specifically during the transition from freezing to movement. As with CS responses, these neuronal correlates of behavior are absent when mPFC projections to the BNST are inhibited."

"it is tempting to conclude that this activity is a neuronal representation or even a causal driver of the lesser expression of PRF".

Ideally, the authors would have a full recording dataset. We recognize this may not feasible at this stage. So we give the authors two options: (1) To revise the language in the manuscript so that it can more accurately reflect the recording data, (2) To remove the behavioral electrophysiology data altogether.

In view of the reviewers’ concerns about the number of recorded units showing event-related activity, we have moved in vivo unit recording data to the supplemental materials and revised the language in the manuscript by:

Removing the following statements: “Thus, IL cells were preferentially activated to the CS after PRF and this activation was abolished by inhibition of the mPFC→BNST pathway,” “Thus, IL cells not only respond to presentation of the CS after PRF, but also to the ongoing behavior of the mouse and specifically during the transition from freezing to movement. As with CS responses, these neuronal correlates of behavior are absent when mPFC projections to the BNST are inhibited” and “it is tempting to conclude that this activity is a neuronal representation or even a causal driver of the lesser expression of PRF.”

Changing the title of the corresponding Results section from “IL cells preferentially signal CS-onset and freezing cessation to a PRF CS” to “IL cells signal CS-onset and freezing cessation.”

The relevant statement in the Abstract to “Multiplexing chemogenetics with in vivo neuronal recordings showed elevated infralimbic cortex (IL) neuronal activity during CS-onset and freezing-cessation; these neural correlates were abolished by chemogenetic mPFC→BNST inhibition.”

And made other changes in the Discussion to avoid making claims about the specify of the unit correlates to PRF; for example “Intriguingly, we also found a subset of IL neurons that displayed phasic activity during the cessation, but not onset, of freezing during [add “fear”] retrieval, echoing recordings in rat…”

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Data Citations

    1. Glover LR, McFadden KM, Bjorni M, Smith SR, Rovero NG, Oreizi-Esfahani S, Yoshida T, Postle AF, Nonaka M, Halladay LR, Holmes A. 2021. mCherry and Gi DREADD electrophysiology for freezing levels during retrieval. Dryad Digital Repository. [DOI]

    Supplementary Materials

    Figure 1—source data 1. PRF versus FRF (Figure 1C).
    elife-60812-fig1-data1.xls (2KB, xls)
    Figure 1—source data 2. Strain comparison (Figure 1E).
    elife-60812-fig1-data2.xls (3KB, xls)
    Figure 2—source data 1. c-Fos.
    elife-60812-fig2-data1.xls (5.6KB, xls)
    Transparent reporting form
    elife-60812-transrepform.docx (66.7KB, docx)

    Data Availability Statement

    Some of the data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided for Figure 1. Further data has been uploaded to Dryad (https://doi.org/10.5061/dryad.j9kd51cbn).

    The following dataset was generated:

    Glover LR, McFadden KM, Bjorni M, Smith SR, Rovero NG, Oreizi-Esfahani S, Yoshida T, Postle AF, Nonaka M, Halladay LR, Holmes A. 2021. mCherry and Gi DREADD electrophysiology for freezing levels during retrieval. Dryad Digital Repository.

    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES