Highly unstable heterogeneous representations in VIP interneurons of the anterior cingulate cortex

Abstract

A hallmark of the anterior cingulate cortex (ACC) is its functional heterogeneity. Functional and imaging studies revealed its importance in the encoding of anxiety-related and social stimuli, but it is unknown how microcircuits within the ACC encode these distinct stimuli. One type of inhibitory interneuron, which is positive for vasoactive intestinal peptide (VIP), is known to modulate the activity of pyramidal cells in local microcircuits, but it is unknown whether VIP cells in the ACC (VIP^ACC) are engaged by particular contexts or stimuli. Additionally, recent studies demonstrated that neuronal representations in other cortical areas can change over time at the level of the individual neuron. However, it is not known whether stimulus representations in the ACC remain stable over time. Using in vivo Ca²⁺ imaging and miniscopes in freely behaving mice to monitor neuronal activity with cellular resolution, we identified individual VIP^ACC that preferentially activated to distinct stimuli across diverse tasks. Importantly, although the population-level activity of the VIP^ACC remained stable across trials, the stimulus-selectivity of individual interneurons changed rapidly. These findings demonstrate marked functional heterogeneity and instability within interneuron populations in the ACC. This work contributes to our understanding of how the cortex encodes information across diverse contexts and provides insight into the complexity of neural processes involved in anxiety and social behavior.

Fig. 1. VIP^ACC encode diverse contexts in the EZM.

A Experimental timeline. B Left: EZM. Pink zone: open arms. White zone: closed arms. Right: Time spent in each EZM arm type (%). Closed vs. Open. N = 19. ****p < 0.0001. Trial duration = 10 min. C Representative image of VIP^ACC expressing GCaMP6f (white) in vivo while the mouse navigates the EZM. Pink circles: open-selective, green circles: closed-selective. Neutral cells are not circled. Scale bar = 50 μm. D ROC curves for 3 representative neurons demonstrating how well cells encode behavioral states. An open-selective (pink, auROC = 0.75), a closed-selective (green, auROC = 0.71), and a neutral (gray, auROC = 0.51) VIP^ACC. E Representative Ca²⁺ transients from selective cells: closed-selective (top), open-selective (bottom). Shaded areas indicate the location of the mouse: open (pink) or closed (green) arms. Scale bar = 25 s and 2 SD. F 17% of VIP^ACC were classified as open-selective, 10% as closed-selective, and 73% as neutral. G–I Average Ca²⁺ activity of selective VIP^ACC in the EZM. N = 19 mice, n = 1171 cells. G Closed-selective. Closed vs. Open. ****p < 0.0001. N =17 mice, n = 110 cells. H Open-selective. Open vs. Closed. ****p < 0.0001. N = 16 mice, n = 203 cells. I Neutral cells. Closed vs. Open. p = 0.7100. N = 19 mice, n = 858 cells. J Movement of a mouse from an open to a closed arm (left) or a closed to an open arm (right) of the EZM. K–N Heatmaps (top) show activity of individual selective cells during behavioral transitions. The activity of each cell is normalized and presented on a scale from 0 (dark blue, least active) to 1 (yellow, most active). Traces (bottom) show normalized average activity of selective cells during these transitions. Shaded areas (bottom) indicate the location of the mouse: open (pink) or closed (green) arms. K, L Activity of closed-selective cells from 10 s prior to 10 s after entering either a closed (K) or an open (L) arm. Activity of open-selective cells from 10 s prior to 10 s after entering either an open (M) or closed (N) arm. O–R Average Ca²⁺ activity of selective-cell. O Closed-selective entering closed arm. −5 vs. +5 s. ****p < 0.0001. N = 17 mice. P Closed-selective entering open arm. −5 vs. +5 s. ****p < 0.0001. N = 17 mice. Q Open-selective entering open arm. −5 vs. +5 s. ****p < 0.0001. N = 16 mice. R Open-selective entering closed arm. −5 vs. +5 s. ****p < 0.0001. N = 16 mice. S Schematic for logistic regression (LR) analysis. Red and green shading represent data from open and closed arms in the EZM, which were both used for training and testing. In total, 70% of frames were used to train the LR and 30% to test. T LR performance when trained with either Ca²⁺ data from selective cells (blue) or neutral cells (black) in the EZM. Shaded regions in the curves are the SD: Selective cells: light blue. Neutral cells: gray. All statistics performed with Paired t test. For panels G-I and O-R, Ca²⁺ activity refers to the area under the curve (integral df/f). Each ROC curve and trace is from one representative neuron. Each replicate in (B), (G–I), and (O–R) represents one mouse. ROC receiver operating characteristic, TP true positive rate, FP false positive rate, C closed arm, Clo closed arm, O open arm, Ope open arm, Closed-sel closed selective, Open-sel open-selective, df/f Deltaf/f, LR logistic regression.

Fig. 2. VIP^ACC encode diverse stimuli in a social task.

A Behavioral paradigm. Left: Sociability (Day 1). Right: Social Novelty (Days 2 and 3). Littermate zone (purple). Empty cup zone (pink). Novel mouse zone (orange). Neutral zone (white). B Sociability. Interaction Time (%). Cup vs. Littermate. **p = 0.0051. N = 13. Trial duration = 10 min. C Social Novelty. Interaction Time (%). Littermate vs. Novel mouse. *p = 0.0121. N = 6. Trial duration = 10 min. D Images of VIP^ACC expressing GCaMP6f (white) in vivo during Sociability (left) or Social Novelty (right). Pink circles: cup-selective, purple circles: littermate-selective, orange circles: novel mouse-selective. Neutral cells are not circled. Scale bar = 50 μm. E ROC curves for representative VIP cells: littermate-selective (purple, auROC = 0.65), novel-mouse-selective (orange, auROC = 0.64), cup-selective (pink, auROC = 0.73), and neutral (gray, auROC = 0.51). F Ca²⁺ transients of representative cells: littermate-selective (top), cup-selective (middle), novel mouse-selective (bottom). Shaded areas indicate location of mouse: littermate (purple), cup (pink), novel (orange), and neutral (white) zones. Scale bars = 25 s and 1 SD. G In Sociability, 14% of VIP^ACC were classified as cup-selective, 10% as littermate-selective, 1% as selective for both cup and littermate, and 75% as neutral. In Social Novelty, VIP^ACC were classified as littermate-selective (10% Day 2, 17% Day 3), novel-mouse-selective (19% Day 2, 13% Day 3), selective for both littermate and novel (<1% Day 2 and 3), and neutral (71% Day 2, 69% Day 3). N = 13 mice. For Sociability, n = 662 cells, for Social Novelty, n = 354 cells on Day 2 and 203 cells on Day 3. H–K Average Ca²⁺ activity of selective cells per mouse. H Littermate-selective cell during Sociability. Littermate vs. cup. *p = 0.0179. N = 11 mice. n = 68 cells. I Empty cup-selective during Sociability. Cup vs. littermate. ***p = 0.0006. N = 12 mice. n = 59 cells. J Littermate-selective during Social Novelty. Littermate vs. novel mouse. **p = 0.0085. N = 11 mice. n = 59 cells. K Novel-mouse-selective during Social Novelty. Littermate vs. novel mouse. ****p < 0.0001. N = 11 mice. n = 63 cells. All ROC curves, traces, and images are representative. Each replicate in (B) and (C) and (H–K) represents one mouse. All statistics performed with Paired t test. Lit littermate, Nov novel mouse, Lit-sel littermate-selective cell, Cup-sel cup-selective cell, Social nov social novelty, df/f Deltaf/f, ROC receiver operating characteristic, TP true positive, FP false positive.

Fig. 3. Highly unstable VIP^ACC representations over multiple trials during an anxiogenic task.

A–M Animals explored the EZM twice with a 24 h ITI. Data were registered to identify the same neurons in Trial 1 and 2 (24 h later). N = 5 mice, n = 211 cells. Trial duration = approximately 10 min. A Schematic showing behavioral pipeline for multi-trial 24 h ITI EZM. Pink: open arms, white: closed arms. B Images of VIP^ACC expressing GCaMP6f (white) in vivo during Trial 1 (left) and Trial 2 (right). The same cells were identified in both trials. Second and fourth images are zoomed versions of the yellow boxed regions. Pink circles: open-selective, green circles: closed-selective, gray circles: neutral cells. Scale bar = 50 μm in first and third images, 15 μm in second and fourth images. C In Trials 1 and 2, VIP^ACC were classified as closed-selective (green, 14% in Trial 1 and 2), open-selective (pink, 24%, 11%), or neutral (gray, 62%, 75%). Percentages of functional subtypes did not significantly change across trials. Two-way Repeated measures ANOVA. Selectivity: F(_2,12) = 65.75. ****p < 0.0001. Trial x Selectivity: F(_2,12) = 2.290. p = 0.1437. D % of time in open arm. Trial 1 vs. 2. p = 0.7452. E Schematic for logistic regression (LR) analysis. Red or green shading represents data from open or closed arms in the EZM, which were both used for training and testing. 70% of frames were used to train the LR and 30% to test. LR was trained and tested with either data from Trial 1 or 2. F LR trained with either Trial 1 (left) or Trial 2 (right) data and tested on Trial 2 (blue) or Trial 1 (red) data, respectively. Shaded regions are the SD. G Using selectivity classifications from Trial 1, Average Ca²⁺ activity was quantified for these registered cells in Trial 1 or 2. Top: Analysis schematic. Bottom: Ca²⁺ activity of Trial 1 selective cells in Trial 1 (left) and 2 (right). Green: closed-selective, pink: open-selective. Trial 1, closed-selective: Closed vs. Open. **p = 0.0034. Trial 1, open-selective: Closed vs. Open. *p = 0.0112. Trial 2, closed-selective: Closed vs. Open. p = 0.6567. Trial 2, open-selective: Closed vs. Open. p = 0.3996. H % of cells identified as selective in Trial 1 that were stable, switched selectivity, or were neutral in Trial 2. Repeated measures ANOVA. F_{(1.145,4.582)} = 86.27. ***p = 0.0003. Stable vs Switched. p = 0.2360. Stable vs Neutral. **p = 0.0011. Switched vs Neutral. **p = 0.0020. I–M Average Ca²⁺ activity for all selective or neutral cells as classified in Trial 1 (I, J) or 2 (K, L). I Trial 1 selective cells. Trial 1 vs. 2. p = 0.8634. J Trial 1 neutral cells. Trial 1 vs. 2. p = 0. 3816. K Trial 2 selective cells. Trial 1 vs. 2. p = 0. 4642. L Trial 2 neutral cells. Trial 1 vs. 2. p = 0.2573. M Average Ca²⁺ activity for all VIP^ACC in Trials 1 and 2. p = 0.2751. N–W Animals explored the EZM twice with a 1 h ITI. Data were registered to identify the same neurons in Trials 1 and 2. N = 6 mice, n = 260 cells. Trial duration = 7 min. N Schematic showing behavioral pipeline for 1 h ITI EZM. Pink: open arms, white: closed arms. The same cells were identified in both trials. O In Trials 1 and 2, VIP^ACC were classified as closed-selective (green, 9% in Trials 1 and 2), open-selective (pink, 8%, 15%), or neutral (gray, 83%, 76%). Percentages of functional subtypes did not significantly change across trials. Two-way Repeated measures ANOVA. Selectivity: F(_2,15) = 273.0. ****p < 0.0001. Trial x Selectivity: F(_2,15) = 2.361. p = 0.1284. P % of time in open arm. Trial 1 vs. 2. p = 0.1681. Q Using selectivity classifications from Trial 1, Average Ca²⁺ activity was quantified for these selective cells in Trial 1 or 2. Top: Analysis schematic. Bottom: Average Ca²⁺ activity of Trial 1 selective cells in Trial 1 (left) and 2 (right). Green: closed-selective, pink: open-selective. Trial 1, closed-selective: Closed vs. Open. **p = 0.0022. Trial 1, open-selective: Closed vs. Open. **p = 0.0072. Trial 2, closed-selective: Closed vs. Open. p = 0.2980. Trial 2, open-selective: Closed vs. Open. p = 0.2326. R % of cells identified as selective in Trial 1 that were stable, switched selectivity, or were neutral in Trial 2. Repeated measures ANOVA. F_{(1.998,9.989)} = 7.345. *p = 0.0109. Stable vs Switched. p > 0.9999. Stable vs Neutral. *p = 0.0491. Switched vs Neutral. *p = 0.00448. Average Ca²⁺ activity for all selective or neutral cells as classified in Trial 1 (S, T) or 2 (U, V). S Trial 1 selective cells. Trial 1 vs. 2. p = 0.8809. T Trial 1 neutral cells. Trial 1 vs. 2. p = 0.8212. U Trial 2 selective cells. Trial 1 vs. 2. p = 0. 9117. V Trial 2 neutral cells. Trial 1 vs. 2. p = 0.7664. W Average Ca²⁺ activity for all VIP^ACC in Trials 1 and 2. p = 0.7664. Each replicate represents one mouse. All statistics performed with Paired t test unless otherwise noted. ITI inter-trial interval, Tr trial, LR logistic regression, C closed, O open, df/f Deltaf/f, auROC area under the receiver operating characteristic (ROC) curve, h hours, Sel selective, Neu neutral.

Fig. 4. Highly unstable VIP^ACC representations over multiple trials during a sociability task.

A–K Multi-trial Sociability task. Animals explored an arena with an empty cup and a cup containing a novel mouse over 3 trials with 10 min ITIs. Data were registered to identify the same neurons in Trials 1–3. N = 5 mice, n = 203 cells. Trial duration = 5 min. A Schematic showing behavioral pipeline for multi-trial Sociability. Pink: empty cup zone, orange: novel mouse zone, white: neutral zone. B Images of VIP^ACC expressing GCaMP6f (white) in vivo during Trials 1–3 (left, middle, and right, respectively). The same cells were identified in all trials. Second, fourth, and sixth images are zoomed versions of the yellow boxed regions. Pink circles: cup-selective, orange circles: mouse-selective, gray circles: neutral cells. Scale bar = 85 μm in first, third, and fifth images, 30 μm in second, fourth, and sixth images. C In each Trial, VIP^ACC were classified as mouse-selective (orange, 17% in Trial 1, 11% in Trial 2, and 9% in Trial 3), cup-selective (pink, 10%, 15%, 11%), or neutral (gray, 73%, 74%, 80%). % of functional subtypes did not significantly change across trials. Two-way Repeated measures ANOVA. Selectivity: F (_2,12) = 374.7. ****p < 0.0001. Trial x Selectivity: F (_4,24) = 1.516. p = 0.2291. D % of time interacting with mouse cup. Repeated measures ANOVA F _{(1.497,5.987)} = 0.2268. p = 0.7434. E Using cells classified as selective in Trial 1, average Ca²⁺ activity was quantified in each trial. Repeated measures ANOVA. F _{(1.532,6.130)} = 3.493. p = 0.1027. % of cells identified as mouse-selective in Trial 1 that were stable, switched selectivity, or were neutral in Trials 2 (F) or 3 (G). F Trial 1 vs. Trial 2. Repeated measures One-way ANOVA. F _{(1.682,6.724)} = 7.609. *p = 0.0211. Stable vs. Switched, p = 0.4758. Stable vs. Neutral, *p = 0.0365. Switched vs. Neutral, p = 0.1668. G Trial 1 vs. Trial 3. Repeated measures ANOVA. F _{(1.588,6.353)} = 106.9. ****p < 0.0001. Stable vs. Switched, p = 0.9607. Stable vs. Neutral, **p = 0.0011. Switched vs. Neutral, ***p = 0.0003. H Survival curve showing cells identified as mouse-selective in Trial 1 and the % of those cells that remained mouse-selective in Trials 2 and 3. Magenta: real data, gray: shuffled control data with SD. I–K For cells that were classified as mouse-selective in Trial 1, average Ca²⁺ activity during interactions with the mouse and empty cups was quantified in each trial. I Trial 1. Mouse vs. Cup. ***p = 0.0003. J Trial 2. Mouse vs. Cup. p = 0.1421. K Trial 3. Mouse vs. Cup. p = 0.8520.

Fig. 5. Highly unstable VIP^ACC representations over multiple trials during an object interaction task.

A–K Multi-trial Object Interaction task. Animals explored an arena with a novel object over 3 trials with 10 min ITIs. Data were registered to identify the same neurons in Trials 1–3. N = 5 mice, n = 162 cells. Trial duration = 5 min. A Schematic showing behavioral pipeline for multi-trial Object task. Blue: novel object zone, white: neutral zone. B Images of VIP^ACC expressing GCaMP6f (white) in vivo during Trials 1–3 (left, middle, and right, respectively). The same cells were identified in all trials. Second, fourth, and sixth images are zoomed versions of the yellow boxed regions. Blue circles: object-selective, gray circles: neutral cells. Scale bar = 30 μm in first, third, and fifth images, 45 μm in second, fourth, and sixth images. C In each Trial, VIP^ACC were classified as cup-selective (blue, 28% in Trial 1, 27% in Trial 2, and 21% in Trial 3) or neutral (72%, 73%, 79%). Percentages of functional subtypes did not change across trials. Two-way Repeated measures ANOVA. Selectivity: F(_1,8) = 18.77. **p = 0.0025. Trial x Selectivity: F(_2,16) = 1.261. p = 0.3101. D % time interacting with novel object. Repeated measures ANOVA. F_{(1.642,6.567)} = 0.6484. p = 0.5244. E Using cells classified as selective in Trial 1, average Ca²⁺ activity was quantified in each trial. Repeated measures ANOVA. F_{(1.364,5.454)} = 0.3859. p = 0.6227. % of cells identified as object-selective in Trial 1 that were stable, switched selectivity, or were neutral in Trials 2 (F) or 3 (G). F Trials 1 and 2. Stable vs. Neutral. p = 0.3243. G Trials 1 and 3. Stable vs. Neutral. *p = 0.0248. H Survival curve showing cells identified as object-selective in Trial 1 and the % of those cells that remained object-selective in Trials 2 and 3. Magenta: real data, gray: shuffled control data with SD. I–K For cells that were classified as object-selective in Trial 1, average Ca²⁺ activity during interactions with the object or exploration of the empty zone was quantified in each trial. I Trial 1. Object vs. Empty. *p = 0.0274. J Trial 2. Object vs. Empty. p = 0.1291. K Trial 3. Object vs. Empty. p = 0.4231. All images are representative. Each replicate represents one mouse. All statistics performed with Paired t test unless otherwise noted. ITI inter-trial interval, Tr Trial, df/f Deltaf/f, Sel selective, O object, E empty.