Dopaminergic projections to the prefrontal cortex are critical for rapid threat avoidance learning

Abstract

To survive, animals must rapidly learn to avoid predictable threats. Such learning depends on detecting reliable cue-outcome relationships that efficiently drive behavioral adaptations. The medial prefrontal cortex (mPFC) integrates learned information about the environment to guide adaptive behaviors^{1,2,3,4,5,6,7} and is critical for threat avoidance.^{8,9,10,11,12,13,14} However, most studies focused on well-learned threat avoidance strategies, and the specific inputs that signal avoidability and drive rapid avoidance learning remain poorly understood. Dopamine (DA) inputs from the ventral tegmental area (VTA) potently modulate prefrontal function and are preferentially engaged by aversive stimuli.^{15,16,17,18,19,20,21} Pharmacological blockade, DA depletion, and microdialysis experiments implicated DA in threat avoidance^22,23,24,25 but lacked the spatiotemporal resolution required to define the timing of mPFC DA signals during avoidance learning. We used high-resolution tools to dissect the role of the VTA-mPFC DA circuit in rapid avoidance learning. Optogenetic suppression of VTA DA terminals in mPFC selectively slowed learning of a cued avoidance response without affecting cue-shock association learning, reactive escape behaviors, or expression of previously learned avoidance. Using a fluorescent DA sensor, we observed rapid, event-locked DA activity that emerged transiently during learning initiation. Increased DA encoded aversive outcomes and their predictive cues, while decreased DA encoded their omission and predicted how quickly mice learned to avoid. In yoked mice lacking control over shock omission, these dynamics were largely absent. Together, these findings demonstrate that the VTA-mPFC DA circuit is necessary for rapid acquisition of proactive avoidance behaviors and reveal transient event-related DA signals underlying this form of learning.

Keywords: dopamine; fear conditioning; prefrontal cortex; prelimbic; threat avoidance; ventral tegmental area.

Figure 1.. Optogenetic inhibition of VTA-mPFC DAergic axon terminals impairs avoidance learning but not tone-shock learning during PMA

A. Experimental timeline of PMA protocol. B. Behavioral training protocol. C. Schematic of PMA chamber (left) and frames from videos (right). D. Schematic of viral and fiberoptic targeting locations. E. Coronal sections from a representative brain showing Jaws-eGFP expression in VTA and bilateral fiber placement in PL. F. Optogenetic inhibition during PMA. 635nm laser light was presented coincidently with a 30 second 4kHz tone that co-terminated with a 2-second foot shock. G. Schematics showing a successful trial when mice preemptively avoided the shock vs. an escape trial when mice leaped to the platform after the shock began. H. Percent successful trials across days in GFP vs. Jaws mice (Mixed effects model F_time(4.26, 49)=17, P<0.0001; F_opsin(1, 12)=5.66, P=0.03; F_interaction(8, 96) = 2.197, P=0.06; n=7 GFP, n=7–8 Jaws mice). I. Percent escape trials across days in GFP vs. Jaws mice (Mixed effects model F_time(4,43, 49.86)=5.43, P=0.0007; F_opsin(1, 12)=4.78, P=0.04; F_interaction(8, 96)=1.04; P=0.4; n=7 GFP, n=7–8 Jaws mice). J. Percent time freezing during tone across days in GFP vs. Jaws mice (Mixed effects model F_time(4.32, 46.98)=8.41, P<0.0001; F_opsin(1, 12)=1.4, P=0.2; F_interaction (8, 96)=1.01; P=0.4; n=7 GFP, n=7–8 Jaws mice). K. Platform time during tone periods across days in GFP vs. Jaws mice (Mixed effects model F_time(4.2, 50) = 15, P<0.0001; F_opsin(1, 12) = 0.9, P=0.3; F_interaction(8, 96)=2.2, P=0.02; n=7 GFP, n=7–8 Jaws mice). L. Number of reward beam breaks during the tone period across days in GFP and Jaws mice (Mixed effects model F_time(1.9, 32.1)=2.5, P=0.09; F_opsin(1, 33)=5.66, P=0.9; F_interaction(2, 33)=0.2, P=0.7; n=7 GFP, n=7–8 Jaws mice). M. Number of reward beam breaks during ITI across days in GFP and Jaws mice (Mixed effects model F_time(1.9, 20.1) = 4.3, P=0.02; F_opsin(1, 12)=8.2, P=0.01; F_interaction(2, 21)=0.02, P=0.98; GFP vs Jaws Day 1 P=0.02; Day 2 P=0.2; Day 3 P=0.3, n=7 GFP, n=7–8 Jaws mice). *P<0.05, Graphs represent mean ± SEM. Triangles denote males, circles denote females. See also Figure S1.

Figure 2.. Prefrontal DA signals evolve with PMA learning and predict future avoidance success

A. Representative coronal section showing AAV-GRABDA_2m expression and fiber placement. B. Successful trials across time per animal (n=6 mice). C. z-scored DA fluorescence in a representative mouse. Red x marks shock trials. D. Illustration of metrics used to quantify mPFC DA dynamics and correlate with PMA performance. E. Left: mPFC DA signal during shocks across all days of PMA. Right: mean of first three shocks vs mean of last three shocks (paired t-test, P=0.03). F. Left: Linear regression of successful trials after the first three shocks with slope between first two response blocks (R²=0.26, P=0.29). Right: same but for area under the curve (AUC) of all responses (R²=0.66, P=0.048). G. Left: mPFC DA signal across successful avoids across all days of PMA. Right: mean DA during first three avoids vs mean DA during last three avoids (paired t-test, P=0.04). H. Left: Linear regression of successful trials after the first three avoids with slope between first two avoid response blocks (R²=0.77, P=0.049). Right: same but for (AUC) of all avoid responses (R²=0, P=0.97). I. Left: DA signal tone onset across all days of PMA. Right: mean DA during first three tones vs mean DA during last three tones (paired t-test, P=0.9). J. Left: Linear regression of successful trials with slope between first two tone response blocks (R²=0, P=0.89). Right: same but for AUC of all tone responses (R²=0.58, P=0.07). K. Left: DA dynamics surrounding platform entry at baseline (BL), immediately preceding the behavior (event), or after the behavior (post) averaged across all platform entries on all days. Right: Quantification of signal changes (repeated measures one-way ANOVA F(1.4, 7.4)=10, P=0.01. BL vs. event P=0.02; event vs. post P=0.007; BL vs. post P= 0.7, Tukey’s multiple comparisons tests). See also Figures S2 and S3. L. Left: Comparison of early (Day 1) and late (Day 3) platform entry DA dynamics. Middle: Timing of signal peak from Day 1 and Day 3 (Wilcoxon signed rank test, P=0.031). Right: Peak of signal from Day 1 and Day 3 (Wilcoxon signed rank test, P=0.9). M. Same as K for platform exits (repeated measures one-way ANOVA F(1.2,6)=6, P=0.04. BL vs. event P=0.03. Event vs. post P=0.4. BL vs. post P=0.04, Tukey’s multiple comparisons tests). N. Same as L for platform exits. Middle: timing of peak signal from Day 1 to Day 3 (Wilcoxon signed rank test P=0.56). Right: peak of signal from Day 1 to Day 3 (Wilcoxon signed rank test P=0.68). *P<0.05, ** P<0.01. Graphs represent mean ± SEM. Triangles denote males, circles denote females. See also Figures S2 and S3.

Figure 3.. mPFC DA activity is sensitive to the avoidability and predictability of aversive outcomes

A. Schematic showing GRABDA recording in prelimbic (PL) cortex. B. Yoked experimental design: animals were placed in a fear conditioning (FC) arena and exposed to tones. Presence of a co-terminal shock was dependent on the matched trial outcome of a subject from PMA. C. Freezing during the tone across days (n=5 mice). D. Heatmap of GRABDA fluorescence (zscored) from representative mouse. E. Left: DA levels during shocks across time. Right: Comparison of DA during first vs. last shock block (paired t-test, P=0.3, n=5 mice). F. Left: DA levels during shock omission across time. Right: Comparison of DA during first vs. last shock omission block (paired t-test P=0.2, n=5 mice). G. Left: DA levels during tone onset across time. Right: Comparison of DA during first vs. last tone block (paired t-test P=0.03, n=5 mice). H. Linear regression of successful trials after the first shock block with (left) area under the curve (AUC) of shock responses (R²=0.81, P=0.03, n=5 mice) and (right) slope between first two shock response blocks (R²=0.55, P=0.13, n=5 mice). I. Linear regression of successful trials after the first avoid block with (left) area under the curve (AUC) of avoid responses (R²=0.05, P=0.71, n=5 mice) and (right) slope between first two avoid blocks (R²=0, P=0.97, n=5 mice). J. Linear regression of successful trials with (left) area under the curve (AUC) of tone responses (R²=0, P=0.94, n=5 mice) and (right) slope between first two tone response blocks (R²=0.09, P=0.62, n=5 mice). K. AAV and fiber placement for GRABDA fiber photometry in fear conditioning. L. Experimental design for fear conditioning recordings. M. Freezing during tone (n=6 mice). N. Heatmap from representative mouse showing GRABDA fluorescence during fear conditioning. O. Same as N for fear memory retrieval. P. Left: DA levels during shock during across the fear conditioning session. Right: Comparison of first to last shock response (paired t-test P=0.01, n=6 mice). Q. Left: DA levels during end of tone (shock omission) across the fear memory retrieval session. Right: Comparison of first and last tone response (paired t-test P=0.03, n=6 mice). R. Left: DA levels during tone onset across the fear conditioning session. Right: Comparison of first and last tone response (paired t-test P=0.3, n=6 mice). S. Left: DA levels during tone onset across the fear memory retrieval session. Right: Comparison of first and last tone response (paired t-test P=0.008, n=6 mice). T. Cartoon summary of changes in mPFC DA dynamics across learning in different aversive learning paradigms. * P<0.05, ** P<0.01, ns P>0.05. Graphs represent mean ± SEM. Triangles denote males, circles denote females. See also Figures S2, S3, S4.

Figure 4.. Optogenetic inhibition of VTA-mPFC DA terminals during PMA without motivational conflict and following learning.

A. Schematic of viral and fiber optic targeting locations. B. Protocol for optogenetic inhibition during PMA training. On training day, 635nm laser light was presented coincidently with a 30 second 4kHz tone that co-terminated with a 2-second footshock. On retrieval day, mice were presented with 10 tones without shocks or laser. C. Left: Fraction of successful trials across days in control vs. Jaws mice (F_time(2.85, 25.72)=5.15, P=0.006; F_opsin(1, 9)=13.6, P=0.005; F_interaction(4, 36)=1.92, P=0.12; Two-way ANOVA, n=5 Jaws, n=6 control mice). Middle: Fraction of escape trials across days in control vs. Jaws mice (F_time (1.9, 17.12)=5.52, P=0.01; F_opsin(1, 9)=5.76, P=0.03; F_interaction (4, 36)=1.6; P=0.18. Two-way ANOVA, n=5 Jaws, n=6 control mice). Right: Fraction time freezing during tone across days in control vs. Jaws mice (F_time(4, 36)=11.51, P<0.0001; F_opsin(1, 8)=0.12 P=0.72; F_interaction(4, 36)=0.98; P=0.42. Two-way ANOVA, n=5 Jaws, n=6 control mice). D. Fraction time on platform and fraction of time freezing during tones on retrieval day (platform: P=0.0003, freezing: P=0.7, two-tailed t-test, n=6 control, n=5 Jaws mice). E. Protocol for optogenetic inhibition during PMA retrieval. Mice are trained in PMA without laser inhibition. The next day, mice are presented with 10 randomly interleaved tones, half of which are paired with constant 635 nm light. F. Successful trials during PMA training (n=3 mice). G. Percent time on platform and percent time freezing during the tone on interleaved light-on and light-off trials (n=3 mice; Platform Wilcoxon signed-rank test P=0.75; Freezing Wilcoxon signed-rank test P=0.5). *P<0.05, **P<0.01. Graphs represent mean ± SEM. Triangles denote males, circles denote females.