SUPPLEMENTARY INFORMATION

Transcription

1 Encoding of conditioned fear in central amygdala inhibitory circuits Stephane Ciocchi 1,*, Cyril Herry 1,,*, François Grenier 1, Steffen B.E. Wolff 1, Johannes J. Letzkus 1, Ioannis Vlachos 3, Ingrid Ehrlich 1,4, Rolf Sprengel 5, Karl Deisseroth 6, Michael Stadler 1, Christian Müller 1 and Andreas Lüthi 1 1 Friedrich Miescher Institute for Biomedical Research, Maulbeerstrasse 66, CH458 Basel, Switzerland. Present address: INSERM U86, Neurocentre Magendie, 146 Rue LéoSaignat, 3377 Bordeaux, France. 3 Bernstein Center for Computational Neuroscience, Freiburg, Germany. 4 Present address: Hertie Institute for Clinical Brain Research, 776 Tübingen, Germany. 5 Department of Molecular Neurobiology, Max Planck Institute for Medical Research, Jahnstrasse 9, 691 Heidelberg, Germany. 6 Department of Bioengineering, Stanford University, Stanford, CA 9435, USA. * These authors contributed equally to this work. Supplementary notes Given that both CEl off and CEl on neurons can form anatomical and functional inhibitory connections in, the question arises why output neurons are disinhibited, rather than inhibited i.e. why the CEl off pathway predominates over the CEl on pathway during sensory stimulation. A possible explanation could be that unitary synaptic connections made by CEl on and CEl off neurons onto output neurons have distinct properties in terms of their strength or their shortterm dynamics. Moreover, based on our data, we cannot exclude that CEl on and CEl off neurons might target distinct subtypes of output neurons. More detailed in vitro electrophysiological and anatomical studies will be required to address these questions. Our results on cell activity are in accord with most studies on the effects of CEA activation on fear behavior. However, the decrease in firing of brainstem projecting cells upon presentation reported by Pascoe and Kapp (1985) 1 seems at odds with our results, which can be at least partly traced to methodological differences. Their neuronal groupings were purely functional, and did not distinguish between the subnuclei in which their neurons were recorded. Furthermore, their analysis was on pooled nonnormalized data, making it hard to appreciate quantitatively the activity that can be ascribed only to brainstem projecting cells. 1. Pascoe, J.P. & Kapp, B.S. Electrophysiological characteristics of amygdaloid central nucleus neurons during Pavlovian fear conditioning in the rabbit. Behav. Brain. Res. 16, (1985). 1

2 Statistical analysis Fig. 1c. Left: Blue laser light (473 nm) was used for bilateral stimulation of ChR expressing neurons (1s periods of light stimulation with interstimulation intervals of 36s). Time bins for freezing determination were 1 s for light stimulation and 36 s for interstimulation intervals. Right: The summary data shows significant lightinduced freezing in AAVChRAtdimer infected animals compared to shamoperated controls. There were significant main effects for infection: F (1,5) = , P <.1; and light stimulation: F (1,5) = , P <.1, two way ANOVA; significant interaction infection x light stimulation: F (1,5) = 5.53, P <.1. Infected animals spent 15.9 ± 6.% of time freezing without light, and 63.1 ± 4.% with light (n = 5, P <.5, posthoc Bonferroni ttest). Freezing in shamoperated controls: without light, 7.6 ± 1.8%; with light: 4.8 ±.4% (n = 1, P >.5 vs. no light, P <.5 vs. infected animals with light). Fig. 1e. Freezing Values for inactivation of the different subnuclei: CEl: 5.9 ± 5.%, n = 8, F (3,) = 6.65, P <.1, one way ANOVA, P <.5, CEl vs. Ctrl and, posthoc Bonferroni ttest. :.9 ± 3.4%, n = 8, P >.5 vs. Ctrl. CEA: 6.8 ± 4.5%, n = 5, P >.5 vs. Ctrl. Control: 6. ± 9.6%, n = 5. Fig. 1f. Freezing Values for inactivation of the different subnuclei: Control: 59.4 ± 8.6%, n = 5. CEl: 4.4 ± 3.7%, n = 14, F (3,31) = 1.719, P <.1, one way ANOVA, P <.1, posthoc Bonferroni ttest. CEA: 7.4 ± 6.4%, n = 7, P <.1. : 54.3 ± 5.5%, n = 9, P >.5. Fig. 1g. Freezing Values for inactivation of the different subnuclei: Control: 65.3 ± 8.%, n = 7. : 39. ± 6.%, n = 9, F (3,) = 5.345, P <.1, one way ANOVA, P <.5, posthoc Bonferroni ttest. CEA: 3.1 ± 1.%, n = 7, P <.1. CEl: 57.7 ± 7.6%, n = 8. Fig. a. CEl on neurons: n = 5 units, 1 mice; zscore, habituation:.79 ±.8; post fear conditioning:.45 ±.39, P <.1, blocks of 4 s. Fig. b. CEl off neurons: n = 41 units, 1 mice; zscore, habituation:.14 ±.31; post fear conditioning: 1.3 ±.3, P <.1, blocks of 4 s. Fig. 3c. Fear conditioninginduced changes in evoked firing in units: n = 15 units, 6 mice; average zscore, 1 ms, habituation:.15 ±.3; post fear conditioning: 3.13 ±.77, P <.1. Fig. 4a. Fear conditioninginduced changes in tonic activity: neurons: before: 9.88 ±.99 Hz; after: 5.98 ±.3 Hz, n = 15, P <.1. CEl on neurons: before: 3.99 ±.7 Hz; after: 3. ±.54 Hz, n = 5, P <.5. CEl off neurons: before: 4.65 ± 1.1 Hz; after: 6.11 ± 1.41 Hz, n = 41, P <.5. Fig. 4d. Correlations between tonic activity and generalization: neurons: freezing ratios for discriminating trial blocks: 1.83 ±.11, n = 4 trial blocks; generalizing trial blocks:.9 ±.11, n = 8, P <.1 vs. discriminating trial blocks; Z score ratios of phasic responses for discriminating trial blocks: 1.79 ±.49, n = 4 trial blocks; generalizing trial blocks:.81 ±.16, n = 8, P <.5 vs. discriminating trial blocks. CEl off neurons: freezing ratios for discriminating trial blocks: 3.48 ±.66, n =

3 19 trial blocks; generalizing trial blocks:.7 ±.9, n = 3, P <.1 vs. discriminating trial blocks; Zscore ratios of phasic responses for discriminating trial blocks: 4.15 ± 1.46, n = 19 trial blocks; generalizing trial blocks: 1.6 ±.35, n = 3, P <.5 vs. discriminating trial blocks. CEl on neurons: freezing ratios for discriminating trial blocks: 3.31 ±.63, n = 1 trial blocks; generalizing trial blocks:.8 ±.11, n = 19, P <.1 vs. discriminating trial blocks; Zscore ratios of phasic responses for discriminating trial blocks:.97 ±., n = 1 trial blocks; generalizing trial blocks: 1.56 ±.3, n = 19, P =. vs. discriminating trial blocks. 3

4 Light on a b 6 Counts Supplementary Figure 1 ChRmediated optical control of neuronal firing in vivo. a, Highmagnification picture illustrating AAVinfected neurons. Scale bar is 5 µm. b, Example raster plot and PSTH illustrating light induced sustained firing of an amygdala neuron recorded in an anaesthetised mouse. Similar lightinduced responses were observed in several neurons (n = 5) in AAVChRAtdimer infected animals. 4

5 a Flinching prob. b Flinching prob. c Freezing (%) 1..5 No drug Muscimol CEA CEl.5 No drug Muscimol US intensity (ma) US intensity (ma) no *** no CEA CEl *** Vocalisation prob. Vocalisation prob No drug Muscimol CEA US intensity (ma) CEl No drug Muscimol US intensity (ma) Supplementary Figure Targeted inactivation of CEA or CEl does not affect US sensitivity and does not lead to permanent damage. a, Inactivation of CEA with muscimol does not affect US sensitivity as measured by the US threshold inducing flinching behavior (left, n = 5, P >.5 vs. control) or vocalization (right, n = 5, P >.5 vs. control). b, Likewise, inactivation of CEl with muscimol does not affect US sensitivity (left: flinching behavior, n = 8, P >.5 vs. control; right: vocalizations, n = 8, P >.5 vs. control). c, Reconditioning of mice 4 hrs after inactivation of CEA or CEl. After washout of muscimol, mice in which CEA or CEl had been inactivated showed normal acquisition and retrieval of conditioned fear responses. Twenty four hours after muscimol application, mice (n = 7 in each group) were reconditioned and tested for fear memory retrieval again 4 hrs later (CEA, 65.3 ± 8.% during exposure vs ± 6.6% before exposure; CEl, 57.7 ± 7.6% during exposure vs. 7.1 ± 4.9% before exposure). **P <

6 Freezing (%) 1 CEACtrl CEA CEl 5 BPY Musc ** ns *** *** ns ns *** ns *** *** *** ** Supplementary Figure 3 Differential role of CEl and in fear expression. Animals were fear conditioned in the absence of BPY or muscimolbpy and tested 4 hrs later. BPY or muscimolbpy was applied before animals were retested on the same day. Bar graphs illustrate freezing levels before and during ex posure in the presence and absence of BPY or muscimol. Consistent with the results obtained in naïve unconditioned animals, we found that inactivation of CEl in conditioned animals resulted in increased baseline (pre) freezing levels. However, when animals with CEl inactivation were exposed to the, they exhibited a further increase in freezing levels. This could be explained by several possibilities. First, it is possible that we did not inactivate the entire CEl, and that conditioned freezing was therefore not completely occluded by unconditioned freezing. Alternatively, our findings are also consistent with the notion that conditioned freezing might be elicited through other pathways (e.g. via disinhibition of output neurons by ITCs or by direct excitatory inputs from BLA or thalamus). Data were analyzed by one way ANOVA followed by posthoc Bonferroni ttest (vs. the no, notreatment group). CEACtrl, n = 7; CEA, n = 7;, n = 9; CEl, n = 8. Error bars indicate mean ± s.e.m. ns: not significant, **P <.1, ***P <

7 a c e Zscore Zscore 1 CEl 1.46 mm CEl on b Freezing (%) d Zscore Zscore no 1 *** CEl on CEl off f CEl off g Zscore i Zscore CEl on CEl off Time (ms) Time (ms) h j Zscore Zscore ms Time (ms) 35 ms Time (ms) Supplementary Figure 4 Characterization of fear conditioninginduced plasticity of evoked firing of CEl units. a, Coronal section of the am ygdala showing the location of the recordings sites of CEl on (green) and CEl off (blue) units. Numbers indicate the anteroposterior coordinates caudal to bregma. b, Summary graph illustrating changes in freezing behavior during fear conditioning. Comparing the first with the last reveals an increase in freezing behavior during conditioning (1 st : 6.3 ± 4.9% of the time spent freezing; 5 th : 6. ± 5.1%). c,d, Increased freezing levels were associated with significant changes in evoked activity in CEl on neurons (n = 15 neurons from 14 mice; note that zscored 1 responses do not reach significance) and CEl off neurons (n = 3 neurons from 17 mice). Note that not all units recorded before and after fear conditioning could be reliably identified during conditioning. e,f, Discrimination between and. Averaged and evoked responses of all CEl on units (n = 3) and CEl off units (n = 3) recorded from a mouse exhibiting behavioral discrimination. g,h, Analysis of response latencies of CEl on units. Normalized population peristimulus time histograms illustrating evoked responses of CEl on neurons (n = 5) before (left panel; onset latency of excitation: 115 ms) and after fear conditioning (right panel; onset latency: 115 ms). i,j, Analysis of response latencies of CEl off units. Normalized population peristimulus time histograms illustrating evoked responses of CEl off neurons (n = 41) before (left panel) and after fear conditioning (right panel; onset latency of inhibition; 354 ms). ***P <

8 a b CEl CEl Injection site c CEl HSVGFP d CEl Injection site HSVGFP Supplementary Figure 5 Retrograde tracing of CEl to projections. a, HSVGFP injection site in revealed by coinjection of fluorescent latex beads. Anteroposterior coordinates caudal to bregma: 1.34 mm. b, HSVGFP injection to reveals retrograde labeling of CEl neurons detected with antigfp immunostaining. c, HSVGFP injection site in CEl revealed by coinjection of fluorescent latex beads. Anteroposterior coordinates caudal to bregma: 1. mm. d, HSVGFP injection to CEl reveals local labeling in CEl as detected with antigfp immunostaining. Very few labeled neurons were observed in, indicating that the CEl to projection is unidirectional. Scale bars: 13 µm. 8

9 a CEl on b CEl on Frequency (Hz) c CEl off d Counts Frequency (Hz) Counts CEl off Supplementary Figure 6 CEl on and CEl off units functionally inhibit output neurons. a, Raster plot illustrating evoked responses of an identified CEl on unit. b, Crosscorrelation analysis reveals a short latency inhibitory interaction between the CEl on unit taken as a reference and a neuron. Dotted lines indicate 95% confidence intervals. Dashed vertical line indicates time of reference spikes from CEl on unit. Because the presynaptic neuron exhibited bursting activity, inhibition of the postsynaptic neuron appears to start before the presynaptic spike. c, Raster plot illustrating evoked responses of an identified CEl off unit. d, Crosscorrelation analysis reveals a short latency inhibitory interaction between the CEl off unit taken as a reference and a neuron. Dotted lines indicate 95% confidence intervals. Dashed vertical line indicates time of reference spikes from CEl off unit. Probabilites for finding inhibitory crosscorrelations between identified CEl units and output neurons were: CEl on pairs: 1/; CEl off pairs: /. These data were obtained from a single mouse. Because of the low n, additional multisite unit recordings or other approaches (paired recordings in acute slices) will be necessary in order to obtain a quantitative understanding of the connectivity between defined subtypes of CEl neurons and ouput neurons. 9

10 a d Zscore 3 1 CEl 1.46 mm Time (ms) b Zscore e Zscore ms Time (ms) 45 ms c Zscore Supplementary Figure 7 Characterization of fear conditioninginduced plasticity of evoked firing of units. a, Coronal section of the amygdala showing the location of the recordings sites of units. Numbers indicate the anteroposterior coordinates caudal to bregma. b, evoked firing of units changes during fear conditioning. Increased freezing levels were associated with significant changes in evoked activity in neurons (n = 7 neurons from 4 mice). Note that not all units recorded before and after fear conditioning could be reliably identified during conditioning. c, Discrimination between and. Averaged and evoked response of a unit recorded from a mouse exhibiting behavioral discrimination. d,e, Analysis of response latencies of units. Normalized population peristimulus time histograms illustrating evoked responses of neurons (n = 15) before (left panel; onset latency of excitation: 115 ms) and after fear conditioning (right panel; onset latencies: 115 ms and 455 ms). 1

11 TEST a BLA Thal. CELon CELoff inhibitory excitatory b HABITUATION / CELoff FEAR CONDITIONING DISCRIMINATION Freezing GENERALIZATION Supplementary Figure 8 Summary of main results. a, Functional connectivity of the CEA microcircuit. The CEl contains two distinct neuronal populations changing responses with fear conditioning: CElon neurons acquire an excitatory response, whereas neurons are inhibited by the. CEloff neurons express PKCδ; REF 1). These two populations are reciprocally connected, and in addition project to. Firing of neurons triggers freezing. In addition to these local inhibitory interactions within CEA, both CElon and neurons also receive shortlatency excitatory input, presumable from thalamus and/or BLA. Based on our data, we cannot exclude that CElon and CEloff neurons might target distinct subtypes of output neurons. b, Changes in activity of CEloff and neurons with discriminative fear conditioning. During habituation, both populations are spontaneously active but showed no response. After conditioning, CEloff neurons acquired an inhibitory response, whereas neurons increased their activity in response to the. Mice were split into a group with discriminating fear responses and one group displaying stimulus generalization. In discriminating animals, baseline firing rates where very similar to habituation (dashed grey line). In contrast, generalization was associated with changes in baseline firing of opposite polarity for the two neuronal populations: CEloff neurons increased tonic firing, whereas cells displayed decreased tonic activity. The tonic changes were of opposite direction to the phasic response in both neuronal populations, and increased the signaltonoise ratio of both and, leading to reduced neuronal discrimination between the two stimuli. 11

12 a b c Freezing (%) Freezing (%) 1 5 r =.43 p< Phasic activity (Zscore) r =.81 p> Tonic activity (Hz) r =.119 p< Phasic activity (Zscore) r =.7 p> Tonic activity (Hz) r =.98 p< Phasic activity (Zscore) r =.39 p> Tonic activity (Hz) Freezing (%) 1 5 r =.17 p> r =.1 p> r =.33 p> Tonic activity (Zscore) Tonic activity (Zscore) Tonic activity (Zscore) Supplementary Figure 9 Correlations of phasic and tonic activity with postconditioning freezing levels: raw data. a, Phasic zscored evoked responses of neurons (left), CEl off neurons (middle), and CEl on neurons (right) are highly correlated with behavioral freezing. Correlations were obtained by plotting the averaged zscored responses for all neurons of a given subtype for each animal over blocks of s. Plots include both and responses. P values indicate significance levels of Pearson s correlation coefficients. Dashed lines indicate 95% confidence intervals. b, Absolute levels of tonic activity do not correlate with freezing behavior. Correlations were obtained as described above. c, Zscored tonic activity does not correlate with freezing behavior. Correlations were obtained as described above. 1

13 Δ Phasic activity (Zscore) 1 5 r =.53 p< Δ Tonic activity (Hz) CEloff r =.17 p< Δ Tonic activity (Hz) 4 CElon r =.4 p> Δ Tonic activity (Hz) Supplementary Figure 1 Changes in tonic activity offset phasic responses. Direct correlation between fear conditioninginduced changes in tonic activity and in evoked phasic activity (zscored) in neurons (left, n = 15), CEl off neurons (middle, n = 41), and CEl on neurons (right, n = 5). For clarity, data were averaged and binned (bins contain equal number of data points). Linear correlations were performed using the nonbinned raw data. P values indicate significance levels of Pearson s correlation coefficients. Dashed lines indicate 95% confidence intervals. 13

14 Freezing (%) 1 5 r =.17 p> r =.1 p> r =.33 p> Tonic activity (Zscore) Tonic activity (Zscore) Tonic activity (Zscore) Supplementary Figure 11 Zscored tonic activity does not correlate with freezing behavior. Zscored tonic activity at test (4 hrs after fear conditioning) of neurons (left), CEl off neurons (middle), and CEl on neurons (right) is poorly correlated with behavioral freezing. Correlations were obtained by plotting the averaged zscored tonic activity levels for all neurons of a given subtype for each animal over blocks of s. For clarity, data were averaged and binned (bins contain equal number of data points). Linear correlations were performed using the nonbinned raw data. P values indicate significance levels of Pearson s correlation coefficients. Dashed lines indicate 95% confidence intervals. 14

15 Frequency (%) Hab. Test CEl off Hab. Test CEl on Hab. Test.5..5 Supplementary Figure 1 Changes in tonic activity during stimulation. Normalized population peristimulus ( ) time histograms of neurons (left), CEl off neurons (middle), and CEl on neurons (right). Fear conditioning was associated with decreased tonic activity in units (n = 15) and increased tonic activity in CEl off units (n = 41). 15

16 Tonic activity (Hz) (gen.) CEl off (gen.) CEl on (disc.) * *** *** * * * * ** ** pre pre pre Hab. Test Hab. Test Hab. Test Supplementary Figure 13 Changes in tonic activity across behavioral sessions. Plots show averaged absolute firing frequencies of, CEl off and CEl on units. Averages were obtained from behavioral trials (blocks of two ) in which changes in tonic activity were observed (e.g. trial blocks in which animals exhibited generalization for recordings, n = 8 trial blocks; CEl off, n = 3 trial blocks; or discrimination CEl on, n = 1 trial blocks). Spontaneous activity levels before onset ( pre ) were measured when animals were first placed in context B. 16

17 a b Absolute values CEloff CElon J3 DB J3 DB J3DB Channels w/o units J3 DB Supplementary Figure 14 Isolation of unit recordings. a, Left: Superimposed waveforms recorded from four different units. Right: Spikes originating from individual units were sorted using 3Dprincipal component analysis. b, Quantitative J3 and Davies Bouldin validity index (DB) statistics calculated for, CEl off and CEl on neurons. Controls values were obtained using two clusters defined from the centered cloud of points from channels in which no units could be detected. High values for the J3 and low values for the DB are indicative of good single unit isolation. 17

18 a b c Probability of freezing Probability of freezing Δ freezing probability Supplementary Figure 15 Time course of freezing behavior during stimulation. a, Freezing probability starts increasing early on during presentation. Freezing probability in relation to presentation in fear conditioned mice. Pooled data from mice with 1 presentations on the day following fear conditioning. Sampling period for freezing was 1 ms. The consisted of a series of 7 pips of 5 ms presented at.9 Hz. Vertical red lines indicate the occurrence of the individual pips composing the. The whole is displayed. b, Freezing probability to the first few pips is displayed, showing that freezing probability increases around 1.3 s after the onset of the. c, Freezing increase following onset can be fitted with an exponential curve rising to a maximum. Baseline freezing was subtracted from total freezing to yield freezing increase linked to presentation. The zero time point was set at 1.3 s following onset, when the freezing starts increasing. 18

19 19