ABSTRACT
Survival critically depends on the ability of animals to select the appropriate behavior in response to threat and safety signals from the external world. However, the synaptic and circuit mechanisms by which the brain learns to encode accurate predictors from noise remain largely ignored. Here, we show that frontal association cortex (FrA) dendrites discriminate auditory modalities through the recruitment of non-linear, NMDARs-dependent conductances. These active dendrites can further modify membrane potential dynamics by specifically integrating auditory cues and basolateral amygdala (BLA) inputs. This cooperative mechanism critically shapes the expression of safety vs. fear memories generated from sensory cues that were not explicitly paired to an aversive event (e.g., a footshock) during fear conditioning. Taken together, our data reveal a dendritic mechanism for cue discrimination in FrA, thus providing a new framework for discriminative learning and related disorders.
INTRODUCTION
Discriminative learning is an important survival strategy that depends on the repeated contingency and contiguity between sensory cues (conditioned stimuli, CS) and the events (e.g., danger, safety) that they must predict (unconditioned stimuli, US)1. Discriminative learning has been classically studied by using differential fear conditioning paradigms where two different auditory CSs are positively (CS+) and negatively (CS-) paired in time with an aversive US (e.g., foot shock). This learning protocol is supposed to assign appropriate emotional valence to the two incoming CSs1–3, thereby providing an accurate representation of the environment by increasing discriminative skills between threat and safety signals. While CS+ promotes conditioned fear responses (e.g., freezing behavior) when presented alone, CS- has been shown to serve as a learned safety predictor by reducing fear behavior and increasing positive affective response4. Previous work has thoroughly investigated how the CS+ generate fear responses5,6. However, it remains unclear how the brain learns to encode CS- and thus discriminates between threat and safety.
The medial prefrontal cortex (mPFC) appeared over the past decade as a critical region that shapes behaviors in response to both aversive and non-aversive environmental cues3,7,8. These antagonistic effects of the mPFC possibly develop through specific interaction between its different subdivisions (i.e., prelimbic (PL) and infralimbic (IL) cortices) and the basolateral complex of the amygdala (BLA)9–11. However, the mPFC does not receive direct sensory information neither from sensory cortical areas nor from the thalamus12, thereby supporting the idea that a higher-order neuronal network above the mPFC might encode opposing memories that are later preferentially selected during recall together with its downstream cortical (e.g., PL or IL mPFC) or subcortical structures (e.g., BLA). Specifically, the superficial frontal association cortex (FrA) has been shown to contribute to memory formation during associative learning13–15. This region of the lateral part of the agranular cortex (AGl)16,17 receives inputs from the BLA13,14,18 and sensory cortices12,19, and is non-reciprocally connected to the PL/IL subdivisions of the mPFC19, raising the possibility that it may function as a relay station during learning from sensory cortical areas and the BLA to the mPFC. However, whether and how FrA integrates the variety of sensory information required for discriminative learning is not understood.
The implication of FrA in auditory fear conditioning has constantly been reported. For example, the pharmacological inactivation of FrA neurons alters both the expression and extinction of learned fear13–15. Recently, fear conditioning and extinction have been shown to induce in FrA dendritic spine elimination and formation, respectively13. Importantly, this phenomenon occurs within the same dendritic branch supporting the idea that a unique FrA circuit could form distinct memory traces with opposing emotional values. Nonetheless, no previous evidence has demonstrated a contribution of the FrA in the encoding of incoming sensory cues as threat or safety predictors and, if so, how such process may be controlled by inputs from the BLA.
To address the possible role of the FrA during discriminative learning and the mechanisms behind it, we explored the dynamics of layers II/III FrA pyramidal neurons and long-range projections from the BLA during the acquisition and recall of discriminative memory traces by using two-photon (2P) calcium imaging in head-restrained mice, in vivo whole cells recordings and optogenetic conditional strategies. We found that auditory tones produced N-methyl-D-aspartate receptors (NMDARs)-dependent, dendritic plateau-like depolarizations that potentiated FrA L2/3 neurons when combined with the channelrhodopsin-2-mediated activation of BLA neurons projecting to the FrA. During conditioning, those long-range projecting BLA neurons conveyed integrated information about the CS/US association that were critical to CS- vs. CS+ discriminative learning. In conclusion, our study reveals a potent dendritic mechanism for encoding cue discrimination in FrA, and thus extends the cortical framework of discriminative learning and related disorders.
RESULTS
Auditory stimulation generates dendritic NMDARs-dependent plateau potentials
We first sought to characterize FrA L2 pyramidal cells activation by two widely used forms of auditory stimulations (pure tone vs. broad-band mixture of pure tones (white Gaussian noise)), and performed somatic whole-cell recordings in anesthetized naive animals (Fig. 1). Consistent with previous in vivo recordings of L2/3 pyramidal neurons during anesthesia20, membrane potential spontaneously fluctuated between up and down-states (Fig. 1b). For each recorded cell, postsynaptic potentials (PSP) were monitored prior, during and after the random presentation of both tones, each consisting of 27 pure (8 kHz)-tone or white noise pips (50 ms, 0.9 Hz for 30 s) (Fig. 1c-f). To reduce the variability related to spontaneous activity and detect any change in membrane potential that might be specifically induced by auditory stimulation, we computed the cumulative PSPs (cPSP) over time that were then subtracted to the linear regression calculated during the baseline period prior to auditory stimulation (cPSP change) (Fig. 1c, d). In contrast to a pure, unpaired (up) sine auditory tone (up8kHz) that failed to affect frontal pyramidal neurons activity, unpaired white Gaussian noise (upWGN) alone evoked a long-lasting subthreshold depolarization in naive animals (upWGN: 27.6 ± 4 mV, up8kHz: 1.2 ± 4 mV; n=22, p<0.001, paired t-test) that lasted for at least 30 sec after the end of the stimulation (upWGN: 32.5 ± 5 mV, up8kHz: −3.8 ± 3 mV; n=22, p<0.001, paired t-test) (Fig. 1c-f).
The upWGN-evoked increase in cumulative potential (Fig. 1e) was similar to evoked cortical up-states which were shown to depend on the activation of NMDA receptors (NMDARs)21. In agreement, the sustained depolarization evoked by the upWGN was efficiently suppressed by both the topic application of the specific NMDAR antagonist D(-)-2-Amino-5-phosphonovaleric acid (dAP5; 1 mM) or the presence of the NMDAR open-channel blocker MK-801 (1 mM) inside the intracellular solution (iMK801; end of the stimulation; control: 27.6 ± 4 mV, n=22; +dAP5: −13.4 ± 5 mV, n=14; +iMK801: −2.2 ± 2 mV, n=5; p<0.001, anova; 30 sec later; control: 32.5 ± 5 mV, n=22; +dAP5: −8.2 ± 4 mV, n=14; +iMK801: −7.7 ± 10 mV, n=5; p<0.001, anova) (Fig. 1e and Supplementary Fig. 1). As a consequence, the difference between upWGN- and up8kHz-evoked depolarizations disappeared upon pharmacological blockade of NMDARs (dAP5 and iMK801 conditions pooled together, upWGN: −10.5 ± 3.6 mV, n=19; up8kHz: −18.2 ± 2.9 mV, n=15; p=0.125, t-test). Moreover, both upWGN and up8kHz hyperpolarized pyramidal neurons when NMDARs conductances were indifferently blocked by using dAP5 or iMK801 (Fig. 1e, f), indicating that this effect is cell-autonomous and unlikely to be due to a network wide modification of the balance between excitation and inhibition. Rather, it potentially exposed the existence of a tone non-specific feed-forward inhibitory circuit. Importantly, it indicates that up8kHz was also able to recruit NMDARs conductances, but to a lesser extent than upWGN.
The above results suggest that upWGN-evoked sustained depolarizations recorded at the soma were cell-autonomous and likely to be mediated by local dendritic Ca2+ events through the recruitment of distal NMDARs-dependent conductances that spread towards the soma21,22. To test this hypothesis, we infected mice with an AAV9-Syn-flex-GCaMP6s together with a 1:10000 dilution of AAV1-hSyn-cre (Fig. 2a) in order to obtain a sparse labeling with only a limited number of GCaMP6s-expressing dendrites. Thus, the activity of 104 isolated, non-overlapping distal dendritic branches was imaged in superficial layer 1 through a chronically-implanted cranial window in awake mice (Fig. 2b). We then segregated calcium transients based on their spatial spread along individual dendrites (Fig. 2c and Supplementary Fig. 2). Multiple local (full width at half maximum (fwhm) < 50μm) and global (fwhm ≥ 50μm) calcium transients occurred both spontaneously (i.e. during baseline prior to stimulation) and upon auditory stimulations (Fig. 2b-d). While the full width at half maximum of these local events (21.7 ± (s.d.) 13 μm, n=670 events) fell into the spatial range of NMDA spikes21,22, global events presumably resulted from backpropagating action potentials (Fig. 2d).
Interestingly, the presentation of upWGN evoked more local events (Baseline, local: 79, global: 43; upWGN, local: 338, global: 98; up8kHz, local: 253, global: 86), with a number of local dendritic events per dendrite significantly higher as compared to the presentation of up8kHz (baseline: 0.59 ± 0.13; upWGN: 3.37 ± 0.3; up8kHz: 2.49 ± 0.2; p<0.001, anova) (Fig. 2e). Nevertheless, up8kHz generated more local events as compared to baseline, which is in agreement with the effect of NMDARs block on upWGN-induced somatic cPSP change observed during anesthesia (see Fig. 1f).
Altogether, our results suggest that FrA pyramidal neurons are capable to discriminate auditory stimuli based on their spectral properties during both anesthesia and wakefulness. This occurs at the subthreshold level, with Gaussian auditory tones being more efficient in producing local, dendritic plateau potentials within the same dendritic branch as compared to pure tones.
Co-activation of segregated BLA and auditory inputs potentiated FrA L2/3 pyramidal neurons
The activation of BLA neurons instructs prefrontal circuits during learning and memory recall8,14,23. However, the optical activation of the BLA alone is not sufficient to produce learned associations24. Therefore, we hypothesized that BLA axons, along with the synaptic non-linearities evoked by auditory tones, could control L2/3 FrA pyramidal neurons through their projections into L125. To address this question, we expressed the recombinant light-gated ion channel channeIrhodopsin-2-YFP (ChR2; AAV9-CamKIIa-hChR2-eYFP) into the BLA and performed intracellular recordings in 14 L2/3 FrA neurons from 6 naive mice (Fig. 3).
First, we confirmed that BLA neurons projected to the superficial layer 1 of the ipsilateral FrA (<150μm) (Fig. 3a-c), thereby most likely contacting dendrites of L2/3 pyramidal neurons. In addition, local photostimulation of ChR2-BLA axons in acute slices produced excitatory postsynaptic current (EPSC) in FrA pyramidal neurons with short latencies (3.5 ± 0.36 ms, n=9) and low jitter (0.289 ± 0.04 ms, n=9), suggesting that a fraction of BLA neurons are monosynaptically connected to L2/3 FrA pyramidal neurons (Supplementary Fig. 3)23. However, it does not exclude the possibility that BLA inputs could recruit large-scale neuronal networks. Indeed, we found that the in vivo photostimulation of BLA neurons with an implanted optical fiber produced large, plateau-like depolarizations in all recorded neurons (peak amplitude: 6.2 ± 1.2 mV*sec; fwhm: 551 ± 80 ms; n=13) (Fig. 3d). However, this was observed only when the stimulation was delivered during down-states (Fig. 3d), suggesting that BLA inputs might facilitate the transition between down to up-states by recruiting large-scale synaptic networks. Importantly, spontaneous somatic up-states have been suggested to depend on dendritic plateaus26. Accordingly, those BLA-mediated plateaus were affected by artificial hyperpolarization that has been shown to efficiently and cell-autonomously block NMDARs in vivo21 (pre: 6.14 ± 1.6 mV*sec, hyper: 3.9 ± 1.5 mV*sec, post: 5.56 ± 1.4 mV*sec; n=4; p=0.004; anova repeated measures) (Fig. 3e, f).
Given the well-established role of the BLA during the acquisition and expression of learned association2,3, we next investigated the effect of BLA activation during auditory cue presentation on L2/3 FrA pyramidal neurons (Fig. 4). We first verified that upWGN alone was able to activate FrA pyramidal neurons in mice chronically implanted with optical fibers. Similarly to the effect of auditory stimulation in non-implanted mice (see Fig. 1), upWGN but not up8kHz evoked a long-lasting subthreshold depolarization (upWGN: 18.8 ± 6.7 mV, n=13; up8kHz: −9.7 ± 3 mV, n=8; p=0.01; Mann-Whitney t-test). Then, ChR2-expressing BLA neurons were photo-stimulated for 30 s at 0.9 Hz with 27 square light pulses (50 ms), a protocol that precisely mimicked the pattern of auditory stimuli (Fig. 4a-d). The coincident photo-activation of BLA24 during upWGN significantly increased the cumulative potential as compared to the presentation of the upWGN alone (upWGN: 18.8 ± 6.7 mV vs. upWGN+light: 39 ± 6.2 mV; n=13; p=0.008; paired t-test) (Fig. 4a, b, e). Interestingly, the photo-activation of BLA also significantly affected the cumulative potential evoked by up8kHz (up8kHz: −9.7 ± 3.3 mV vs. up8kHz+light: 16.6 ± 3.1 mV; n=8; p<0.001; paired t-test) (Fig. 4c-e) similarly to the effect of light on upWGN-induced cumulative potential (upWGN/light: 39 ± 6.2 mV, n=13; up8kHz/light: 16.6 ± 3.1 mV, n=8; p=0.138, Mann-Whitney rank sum test) (Fig. 4e).
Nevertheless, the light-activation of BLA neurons differently altered upWGN and up8kHz-evoked cumulative potential 30 sec after the end of the stimulation (time point 2). Indeed, we observed that the coincident photo-activation of BLA affected later on-going spontaneous up and down fluctuations following both upWGN (upWGN: 19.5 ± 7.4 mV vs. upWGN/light: 46.4 ± 11.6 mV; n=13; p=0.045; paired t-test) (Fig. 4b) and up8kHz (up8kHz: −10.8 ± 5.5 mV vs. up8kHz/light: 14.8 ± 6.6 mV; n=8; p=0.013; paired t-test) (Fig. 4d). However, this long-lasting cooperative effect was significantly higher with Gaussian tone as compared to pure-frequency tone (upWGN/light: 46.4 ± 11.6 mV, n=13; up8kHz/light: 14.8 ± 6.6 mV, n=8; p=0.039, Mann-Whitney rank sum test) (Fig. 4e). We also observed that the ectopic application of dAP5 (1mM) to the cortical surface blocked the effect of BLA activation during upWGN (light/upWGN: 39 ± 6.2 mV, n=13; light/upWGN/+dAP5: −26.6 ± 6.7 mV, n=7; p<0.001; t-test) (Fig. 4f). In fact, this resulted in a negative net effect of light (control: 20.2 ± 6.4 mV; n=13; +dAP5: −30.1 ± 7 mV; n=7) (Fig. 4g) that uncovered the activation of feed-forward inhibitory circuits. Thus, it seems that the coordinated activation BLA and auditory inputs generated a highly non-linear depolarization in FrA dendrites that is necessary to overcome evoked inhibition, and eventually induced a long-term alteration of FrA neuronal membrane potential.
Discriminative learning depends on the activation of BLA-to-FrA circuits
We next questioned the functions of BLA-to-FrA axons during fear learning. It is now well established that the BLA and its prefrontal projections actively acquire signals about safety3,7–9,27. Whether direct BLA-to-FrA axons are also necessary to encode CS-remains unknown. We addressed this question by silencing specifically BLA-to-FrA axons during conditioning but only throughout the presentation of the CS- (Fig. 5a). For that, mice were injected bilaterally with a retrograde Cav-2-CMV-Cre28 into the FrAs together with either AAV9-flex-CBA-ArchT-GFP (ArchT-expressing mice, n=11) or AAV9-CAG-flex-eGFP (control GFP-expressing mice, n=13) into both BLAs (Fig. 5a). This resulted in the expression of the light-driven inhibitory proton pump ArchT (or GFP) in a target-specific fraction of BLA neurons that project to the FrA (Fig. 5b, c).
Fear learning was then induced by using a discriminative conditioning protocol during which five auditory stimuli (each consisting of 27 pure-tone or WGN pips, 50 ms, 0.9 Hz for 30 s) were positively (CS+) or negatively (CS-) paired with the delivery of a mild electrical shock (0.6 mA) to the paws in a pseudorandom order (Supplementary Fig. 4a). Learning was tested 24h later during recall and quantified by multiplying the freezing level in each condition by the corresponding index of discrimination (learning index) (Fig. 5e). As expected, control GFP-expressing mice presented robust fear learning (43 ± 5 %; n=13). In contrast, ArchT-expressing mice showed strongly attenuated fear responses during recall (18 ± 3 %; n=11) (Fig. 5e), with a learning index that was significantly lower during recall than the one measured in GFP-expressing mice (GFP: 43 ± 5 %, n=13; ArchT: 18 ± 3 %, n=11; p<0.001, t-test) (Fig. 5f). This alteration of fear learning in ArchT-expressing mice resulted from decreased discriminative performance during recall (GFP, 0.8 ± 0.03, n=13; ArchT, 0.56 ± 0.08, n=11; p=0.003, Kolmogorov-Smirnov test; p=0.016, Mann-Whitney rank sum test) (Fig. 5g).
However, the auditory tones (8 kHz and WGN) used for CS+ and CS-during conditioning were counterbalanced across mice (protocol 1: CS+/CS- = 8kHz/WGN respectively; protocol 2: CS+/CS- = WGN/8kHz respectively) (Supplementary Fig. 4b). Because WGN appeared more effective in producing dendritic plateau potentials in FrA as compared to pure tone (Fig. 1 and Fig. 2), it remains thus possible that fear learning depends on the physical property of the sensory cue that was negatively paired to the footshock29,30. Therefore, we analyzed the impact of BLA-to-FrA inactivation on freezing behaviors for each counter-balanced protocol (Fig. 5h-j and Supplementary Fig. 4c, d). We found that the time-locked suppression of BLA-to-FrA communication during negatively-paired (np) WGN (CS-, protocol 1) significantly increased freezing responses upon subsequent npWGN presentations (GFP: 5.3 ± 1 %, n=8; ArchT: 13.7 ± 4 %, n=7;p=0.040, Mann-Whitney rank sum test), indicating that CS-acquires safety properties through the activation of a specific population of BLA neurons. Surprisingly, it also decreased freezing behaviors upon positively-paired (pp) 8kHz (CS+ protocol 1) (GFP: 55.9 ± 8 %, n=8; ArchT: 34.9 ± 3 %, n=7; p=0.037, t-test) (Fig. 5h). Consequently, the index of discrimination was strongly attenuated as compared to controls (GFP: 0.8 ± 0.03, n=8; ArchT: 0.49 ± 0.12, n=7; p=0.024, Mann-Whitney rank sum test; p=0.04, Kolmogorov-Smirnov test) (Fig. 5j). In contrast, blocking the activity of BLA-to-FrA axons during protocol 2 failed to affect discriminative performance (GFP: 0.78 ± 0.06, n=5; ArchT: 0.64 ± 0.12, n=4; p=0.335, t-test; p=0.17, Kolmogorov-Smirnov test) (Fig. 5j), with similar freezing responses between GFP- and ArchT-expressing mice upon subsequent np8kHz (CS-; GFP: 7.3 ± 2.5, n=5; ArchT: 9.6 ± 3.5, n=4; p=0.611, t-test) and ppWGN (CS+; GFP: 49.8 ± 10.2, n=5; ArchT: 32.2 ± 8, n=4; p=0.230, t-test) (Fig. 5i).
Altogether, our data confirm the sophisticated nature of differential conditioning protocols1,29, during which auditory tones that were not explicitly paired to the footshock might actively participate in discriminative learning (i.e. threat vs. safety encoding) through the interaction between the BLA and the FrA during conditioning. Importantly, it reveals that the BLA-to-FrA circuit encodes WGN as a safety predictor more efficiently than pure frequency tones.
BLA-to-FrA axons are not activated by auditory stimulation but are recruited during fear conditioning
The results above suggest a possible cooperative, Hebbian-like frontal mechanism that could theoretically integrate into safety memory traces any complex sensory inputs that are contiguous to the activation of BLA-to-FrA axons14,24,25,31. Consequently, this mechanism could only occur if BLA-to-FrA axons are activated during conditioning upon CS-presentation. To test this hypothesis, we injected a virus expressing the genetically-encoded calcium indicator GCaMP6f into the right BLA and imaged axonal Ca2+ responses in superficial L1 of the right FrA of awake head-restrained mice during fear conditioning (Fig. 6a-c). We choose to use exclusively the protocol 1 (Supplementary Fig. 4b) as BLA-to-FrA axons appeared to specifically encode safety only when WGN is used as CS-during conditioning (Fig. 5h-j).
We conditioned awake mice (n=7) under the 2-photon microscope, during which GCaMP6f calcium transients (ΔF/F0) provided a direct measure of the activation of BLA neurons projecting to the FrA (Fig. 6c-f). Again, learning was tested 24h later and quantified by assessing the learning index (Fig. 6d). Mice were classified as learners (learning+) when the learning index was higher than 20% during recall (Fig. 6d). We then compared the activity of individual boutons between mice that learned (learning+, n=4 mice) and those that failed to learn (learning-, n=3 mice). While the activity of 299 individual BLA boutons in FrA was relatively low at rest, it increased significantly upon successive pairings (Fig. 6e, f). However, this occurred only for mice that learned (learning+, baseline: 206 events; pairings: 310 events; n=4; +95 ± 51 %; learning-, baseline: 160 events; pairings: 127 events; n=4; −30 ± 10 %; χ2=18.6, p<0.001). Interestingly, it also never occurred before the end of the first US presentation. Indeed, the number of axonal transients observed during the first CS+ (that is, before the delivery of the first footshock) was not different from baseline (learning+, baseline: 206 events; CS+1/No US: 220 events; n=4; learning-, baseline: 160 events; CS+1/No US: 127 events; n=3; χ2=3.03, p=0.08) (Fig. 6g). In contrast, the activity of boutons measured during both the first CS- (learning+, baseline: 206 events; CS-1/US: 342 events; n=4; learning-, baseline: 160 events; CS-1/US: 129 events; n=3; χ2=28.9, p<0.001) and the second CS+ (learning+, baseline: 206 events; CS+2/US: 357 events; n=4; learning-, baseline: 160 events; CS+2/US: 126 events; n=3; χ2=24.3, p<0.001) were significantly increased as compared to the baseline period (Fig. 6g, h). Hence, neither the tone alone nor the footshock appeared to have an effect on the activity of BLA-to-FrA axons (Fig. 6i). Instead, our data support the idea that BLA axons projecting to the FrA conveyed information about the learning, i.e. the CS+/US association itself rather than about the nature or the valence of the auditory tones14.
Fear learning occludes auditory tone-specific dendritic plateau potentials
Finally, we tested the impact of fear learning on auditory-evoked dendritic nonlinearities. We observed that plateau potentials evoked by WGN were strongly and specifically affected in conditioned animals (Fig. 7). Indeed, as compared to naive mice, WGN failed to activate FrA pyramidal neurons in conditioned mice (npWGN: 5.4 ± 9 mV, n=8; upWGN: 27.6 ± 4 mV, n=22; p=0.008) (Fig. 7a-d). Thus, the difference between upWGN and up8 kHz-induced non-linear depolarizations observed in naive mice (14 naive mice; upWGN: 27.6 ± 4 mV; up8kHz: 1.1 ± 4 mV; n=22; p<0.001; paired t-test) disappeared after fear conditioning (5 conditioned mice; npWGN: 5.4 ± 9 mV; pp8kHz: 0.5 ± 5 mV; n=8; p=0.698; paired t-test) (Fig. 7d). Accordingly, the averaged number of local dendritic events per dendrite observed upon npWGN presentation in awake mice significantly decreased after fear learning (upWGN: 3.37 ± 0.3, n=9 naive mice; npWGN: 1.88 ± 0.3, n=5 mice; p=0.003) (Fig. 7e), indicating that npWGN no longer generated local dendritic activation during both anesthesia and wakefulness. We next plotted the tone-evoked cPSP as a function of the behavioral performance (Fig. 7f). As opposed to pp8kHz-cPSPs that were not different between behavioral performance (r2<0.001), npWGN induced stronger plateau depolarizations in low freezing mice indicating that WGN-dendritic plateaus are negatively correlated with behavioral performance (npWGN: r2=0.55; pp8kHz: r2<0.001) (Fig. 7f). Importantly, it indicates that learning occluded multiple input-specific synapses depending on the strength of learning13, possibly through postsynaptic plasticity mechanisms32,33.
DISCUSSION
The present study describes the role of the BLA-to-FrA circuit in the integration of sensory information, and how such process participates in the acquisition of specific memory traces during differential conditioning. We identified segregated auditory and BLA inputs streams whose interaction at the dendritic level of FrA pyramidal neurons might support the creation of safety representation and further influences the representation of cue predicting threat. Although our results depend on the specific association between pure and Gaussian auditory tones during conditioning, they bring new conceptual perspectives to central questions regarding how frontal circuits contribute to learning, thus expanding beyond the BLA-mPFC interactions classically described in fear learning studies3.
Accumulating evidence from anatomical and functional studies has demontrated that despite its pivotal role in the acquisition and expression of associations between sensory stimuli and the emotional valence of these stimuli34,35, the mPFC is not directly involved in sensory processing12,19,36. In contrast, due to its anatomical connections with distributed cortical and subcortical regions12,19, the FrA might serve as a hub that coordinates incoming sensory information before reaching the mPFC. Here, we provide the first demonstration, to our knowledge, that auditory sensory stimulation produces dendritic non-linear NMDARs-dependent plateau potentials in FrA L2/3 pyramidal neurons during both anesthesia and wakefulness. More specifically, we observed that the occurrence of these events within the same dendritic branch was higher upon broadband complex noise as compared to pure frequency auditory stimulation (Fig. 1 and Fig. 2). Although we cannot exclude that WGN tones are structured and further abstracted along the entire auditory system37, the simplest explanation suggests a wide and heterogeneous distribution of frequency-tuned spines throughout the same FrA dendrite, similar to what has been detailed in the auditory cortex of anesthetized mice38. As a consequence, the multiple frequencies composing WGN would promote the activation of a dense pattern of neighboring spines that might in turn facilitate the generation and propagation towards the soma of local non-linear events26. However, multiple calcium transients occurring simultaneously in multiple dendritic branches are necessary to affect somatic voltage22. These data support the idea that, in our study too, WGN-induced long-lasting depolarization recorded at the soma (Fig. 1) built on multiple local calcium events across multiple dendritic branches. In contrast, pure-frequency tone appeared unable to activate enough branches simultaneously (Fig. 2), thereby making the alteration of somatic voltage less probable (Fig. 1).
The FrA and the BLA are anatomically interconnected13,14,18. Yet, the functional properties of these connections remain unknown. Our results confirmed that the BLA projected heavily to the superficial layer of the FrA, thereby most likely contacting dendrites of L2/3 pyramidal neurons (Fig. 3). The photo-stimulation of ChR2-expressing BLA neurons produced plateau-like depolarizations that were strongly affected by hyperpolarization indicating that they also possibly emerge from dendritic NMDARs-mediated conductances21. These BLA-mediated depolarizations were rather weak (Fig. 3), but might summate during rhythmic activation to create favorable conditions for the integration of coincident sensory-driven inputs21,26,39,40. Alternatively, it is possible that the modest activation of BLA synapses in FrA apical dendrites facilitates or gates the propagation towards the soma of tone-evoked dendritic events41. In agreement, we observed that the temporal coincident activation of BLA-to-FrA inputs increased both WGN and 8 kHz-evoked depolarization during anesthesia (Fig. 4). Nevertheless, it is striking that only WGN, but not pure-frequency tone, potentiated FrA pyramidal neurons when combined with the photo-stimulation of BLA-to-FrA inputs (Fig. 4). Even though this effect was measured no longer than 30 sec after the end of the auditory stimulation, it prompts the speculation that long-range BLA projections in superficial layer of FrA produce additional non-linear dendritic depolarization and gain control over coincident WGN-related inputs. Indeed, compelling experimental evidence has demonstrated that non-linear interactions between compartmentalized streams of neural activity induce long-lasting change in synaptic strength and intrinsic excitability21,25,39,41–43 that could have affected permanently the dynamics of FrA membrane potential.
The BLA presumably transfers to the FrA information that is relevant to fear learning13,14. Here, we showed that BLA neurons projecting to the FrA participated in the acquisition of safety memory traces (Fig. 5e-g). These data are consistent with recent studies showing an increase of theta synchronization between mPFC and BLA during safety and CS discrimination that possibly inhibits fear response and anxiety-related behaviors7,8. However, this occurred only when WGN, but not pure-frequency tone, is used as CS-during conditioning (protocol 1) (Fig. 5h-j), thereby revealing for the first time the unique nature of Gaussian tones during learning. In addition, blocking transiently the activity of BLA projecting neurons during CS-surprisingly altered the expression of fear responses (Fig. 5). This is unlikely to be the consequence of a non-sufficient activation of BLA. Instead, given the low number of BLA neurons expressing ArchT (Fig. 5c) and their time-locked inhibition during CS- (Fig. 5a), we propose alternatively that npWGN when combined with the activation of BLA projecting axons might actively influence the representation within the FrA of sensory cues predicting threat (i.e. pp8kHz)1. Importantly, this hypothesis is supported by the late modification of membrane potential fluctuations observed after the activation of BLA with WGN presentation, which occurred at the time when pp8kHz is presented during conditioning (Fig. 4e).
BLA neurons send projections to multiple cortical and subcortical areas that have been shown to project also to the FrA. Thus it remains possible that the information is transmitted from the BLA to the FrA through indirect pathway14,44. Here, we demonstrated that the expression of the genetically encoded calcium indicator GCaMP6 in BLA neurons permitted to monitor optically the activity of target-specific BLA axons during learning (Fig. 6). Using this strategy in awake mice, we revealed for the first time that BLA-to-FrA axons were progressively recruited upon successive conditioning trials, thereby ruling out an indirect activation of the FrA circuit. First, we showed that BLA-to-FrA axonal activity was never affected upon the presentation of auditory cues alone (Fig. 6i). Our data contrast with previous work showing an increase of local field potential and unit activity in the BLA upon auditory stimulation45, and suggest instead the existence of a subpopulation of BLA neurons projecting specifically to the FrA that might play specific role during emotional learning. Indeed, those axons were only activated after the first CS+/US pairing, and thus seem to transmit integrated information about the association itself14. Nevertheless, the activation of axons was independent of the nature of the CS presented (Fig. 6g-i). It appears thus unlikely that BLA-to-FrA axons conveyed the emotional valence of this association. In addition, the activation of BLA alone, while necessary, is not sufficient to trigger learning24,31. What might be then the function of its projections during learning? Previous studies highlighted the critical function of the BLA in attention for learning34,35. In that context, BLA neurons might signal to frontal circuits any new association independently of its valence, which might be subsequently assigned by the mPFC depending on the nature of the incoming stimuli to further supervise the activity of BLA3,46.
Collectively, our results raise the possibility that the learned association formed in the BLA could be combined during learning into the FrA with npWGN-evoked non-linearities to activate FrA prefrontal neurons. This would eventually lead to the recruitment of neurons into specific cue memory traces. In agreement, we found that differential fear conditioning significantly decreased the number of npWGN, but not pp8kHz-evoked local dendritic transients (Fig. 7c-e). Dendritic plateau potentials have been shown to regulate synaptic strength and synaptic plasticity21,22,47–49 which might subsequently facilitate the stabilization or pruning of synaptic inputs during learning50,51. In support of this view, the level of fear learning has been shown to correlate with the percentage of spine elimination in FrA13 which possibly explains the negative relation we observed during anesthesia between npWGN-evoked subthreshold depolarizations and the strength of learning (Fig. 7f).
Taken together, our data revealed for the first time the instructive properties of Gaussian tone during learning, especially when not explicitly paired to the footshock, in encoding safe vs. threat predictors. Given that complex tones and WGN are abundant in the environment and communication of most mammals, they might be well-suited to facilitate safety prediction. This is of crucial importance as many anxiety-related disorders such as post-traumatic stress disorder are associated with a loss of cue discrimination that may result in fear generalization to harmless signals2.
AUTHOR CONTRIBUTION
MA, EA, VK, NC performed the experiments. YH and CM provided technical assistance. MA and FG conceived the studies and analyzed the data with the help of NC and EA. FG supervised the research and wrote the manuscript with the help from MA, EA and NC.
DECLARATION OF INTERESTS
The authors declare no competing financial interests.
METHODS
All experiments were performed in accordance with the Guide for the Care and Use of Laboratory Animals (National Research Council Committee (2011): Guide for the Care and Use of Laboratory Animals, 8th ed. Washington, DC: The National Academic Press.) and the European Communities Council Directive of September 22th 2010 (2010/63/EU, 74). Experimental protocols were approved by the institutional ethical committee guidelines for animal research (N°50DIR_15-A) and by the French Ministry of Research (N°02169.01). We used male C57Bl6/J 6-weeks old mice from Charles River that were housed with littermates (3-4 mice per cage) in a 12-h light-dark cycle. Cages were enriched and food and water were provided ad libitum.
Surgery and virus injection
Mice were anesthetized with an intraperitoneal (i.p.) injection of a mix containing medetomidine (sededorm, 0.27 mg kg-1), midazolam (5 mg kg-1) and fentanyl (0.05 mg kg-1) in sterile NaCl 0.9% (MMF-mix). Analgesia was achieved by local application of 100 μl of lidocaine (lurocaine, 1%) and subcutaneous (s.c.) injection of buprenorphine (buprécare, 0.05 mg kg-1). 40 μl of dexamethasone (dexadreson, 0.1mg ml-1) was administrated intramuscularly (i.m.) in the quadriceps to prevent inflammation potentially caused by the friction of the drilling. A heating-pad was positioned underneath the animal to keep the body temperature at 37°C. Eye dehydration was prevented by topical application of ophthalmic gel. The skin above the skull was disinfected with modified ethanol 70% and betadine before an incision was made. Stereotaxic injections were done as previously described21. Briefly, the bregma and lambda were aligned (x and z) and a hole for injection was made using a pneumatic dental drill (BienAir Medical Technologies, AP-S001). The injections were targeted either to the layer 2/3 of the FrA (from bregma: AP, +2.8 mm; DV, −0.2-0.3 mm; ML ±1.0 mm) or to the BLA (from bregma: AP, −1.3 mm; DV, −4.5 to 4.8 mm; ML, ±2.9 mm), or to both at the same time. 200 nl of virus were injected at a maximum rate of 60 nl/min, using a glass pipette (Wiretrol, Drummond) attached to an oil hydraulic manipulator (MO-10, Narishige).
The following viruses were used depending on the experiments. AAV-ChR2 (AAV9.CamKIIa.hChR2(H134R).eYFP.WPRW.SV40, Penn Vector Core) was unilaterally injected in the right BLA, whereas AAV-ArchT-Flex (AAV9.CBA.flex.Arch-GFP.WPRE.SV40, Penn Vector Core) and CAV2-Cre (Cav2.CMV.Cre, IGMM BioCampus Montpellier) were bilaterally injected into the BLA and FrA, respectively. Control experiments were performed using an AAV containing the DNA construct for GFP (AAV9.CAG.flex.eGFP.WPRE.bGH). For axonal calcium imaging, AAV-GCaMP6f (AAV1.Syn.GCaMP6f.WPRE.SV40, Penn Vector Core) was injected to the right BLA. For dendritic calcium imaging, AAV-GCaMP6s (AAV9.Syn.Flex.GCaMP6s.WPRE.SV40, Penn Vector Core) and a 1:10000 dilution of AAV-Cre (AAV1.hSyn.Cre.WPRE.hGH, Penn Vector Core) were injected together into the right FrA. After injections, the viruses were allowed to diffuse for at least 10 min before the pipette was withdrawn. Mice were then either prepared for cranial window implantation or waked-up by a sub-cutaneous injection of a mixture containing atipamezole (revertor, 2.5 mg kg-1), flumazenil (0.5 mg kg-1), and buprenorphine (buprécare, 0.1 mg kg-1) in sterile NaCl 0.9% (AFB-mix).
The cranial windows were made as previously described21. Briefly, after skull’s exposure a ~5 mm plastic chamber was attached on the area of interest and a 3 mm craniotomy was made on the right hemisphere above FrA and M2, with a pneumatic dental drill, leaving the dura intact. The craniotomy was covered with sterile saline (0.9% NaCl) and sealed with a 3 mm glass cover slip after viral injection (for imaging experiments). The chamber, the cover slip and a custom-made stainless steel head stage were well attached to the skull using dental acrylic and dental cement (Jet Repair Acrylic, Lang Dental Manufacturing).
To evaluate the viral expression profiles in BLA and FrA, fixed brain slices were imaged post-hoc using a wide-field epifluorescence microscope (Nikon, Eclipse N-iU). Illumination was set such that the full dynamic range of the 16-bit images was utilized. A two-dimensional graph of the intensities of pixel was plot using Fiji Software. 16-bit images’ brightness was processed and masks were registered to the corresponding coronal plates (ranging from −1.94 to −2.70 mm) of the mouse brain atlas using Illustrator (Adobe), at various distances anterior (FrA) or posterior (BLA) to the bregma.
Fear conditioning and quantification of learning
At least 5 days before starting behavioral experiments, mice went through handling with the same experimenter that performed the experiments in order to decrease stress. For consistency across experiments, mice were then habituated to auditory tones during 3 successive days. During habituation, mice were placed on the conditioning compartment (context A, consisting of a squared box with a grid floor that allows the delivery of a foot shock and with home cage litter under; cleaned between individuals with 70% ethanol). Two conditional auditory stimuli (CS) (8 kHz pure tone; and white Gaussian noise (WGN); each composed of 27 pips, 50 ms in duration, 0.9 Hz for 30 s) were presented 4 times with a 80 dB sound pressure level and variable inter stimulus interval (ISI). The freezing time during each CS presentation was measured and the mice returned to their home cage. Mice were fear conditioned 24 hours after the last habituation phase by using a classical differential protocol. Briefly, mice were exposed to context A and 5 auditory tones (CS+) were paired with the unconditional stimulus (US, 1s foot-shock, 0.6 mA). The onset of US coincided with the CS+ offset. 5 CS-presentations were intermingled with CS+ presentations with a variable (10-60 s) ISI. CS were counterbalanced with WGN and 8 kHz pure tones being used as CS+ and CS-, respectively. Recall tests were carried out 24, 48 and 72 hours after the conditioning phase by measuring the freezing time during the presentation of 2 CS+ and 2 CS- in a new context (context B, consisting of a cylindrical white compartment with home cage litter on the floor; cleaned between individuals with septanios MD 2%).
For optogenetic experiments using archeorhodopsin (ArchT) or GFP controls, mice were subjected to the same behavioral protocol described above. Optogenetic inhibition of BLA-to-FrA projections upon CS-presentation was achieved during the conditioning phase by synchronizing each pip (50 ms) composing the CS-with a 50 ms-laser pulse. For the experiments in which the conditioning phase was taken place under the 2 photon microscope, the context consisted of the microscope shading box in which the mice were head-restrained in a custom tube containing a shocking grid at the bottom. CS and US presentations were triggered by a MATLAB routine, associated to a pulse-stimulator (Master-8, A.M.P.I) capable of triggering the foot shock. For somatic and dendritic calcium imaging experiments, behavior was assessed at least 6 hours after imaging sessions. For whole-cell recordings experiments, mice were anesthetized and prepare for patch recordings immediately after behavior.
For each behavioral session, the total time duration (sec) of freezing episodes upon CS+ and CS-presentation was quantified automatically using a fire-wire CCD-camera connected to an automated freezing detection software (AnyMaze, Ugo Basile, Italy), and expressed as % of freezing. Learning index was further quantified for each CS by multiplying the % of freezing in each condition by the corresponding index of discrimination by using the following equation:
Learning index <20% during recall was considered as a failure of conditioning.
In vivo whole cell recordings
Isoflurane (4% with ~0.5 l min-1 O2) combined with an i.p. injection of urethane (1.5 g kg-1, in lactated ringer solution containing in [mM] 102 NaCl, 28 Na L Lactate, 4 KCl, 1.5 CaCl2) was used to induce anesthesia and prolonged by supplementary urethane (0.15 g kg-1) if necessary. To prevent risks of inflammation, brain swelling and salivary excretions, 40 μl of dexamethasone (dexadreson, 0.1 mg ml-1, i.m.) and glycopyrrolate (Robinul-V, 0.01 mg kg-1, s.c.) were injected before the surgery. Adequate anesthesia (absence of toe pinch and corneal reflexes, and vibrissae movements) was constantly checked and body temperature was maintained at 37°C using a heating-pad positioned underneath the animal. Ophthalmic gel was applied to prevent eye dehydration. Analgesia was provided as described for viral injection (with lidocaine and buprenorphine). After disinfection of the skin (with modified ethanol 70% and betadine), the skull was exposed and a ~3mm plastic chamber was attached to it above the prefrontal cortex using a combination of super glue (Loctite) and dental acrylic and dental cement (Jet Repair Acrylic, Lang Dental Manufacturing). A small ~1 x 1 mm craniotomy centered above the FrA (+2.8 mm from bregma, ±1.0 mm midline) was made using a pneumatic dental drill, leaving the dura intact.
Whole-cell patch-clamp recordings of L2/3 pyramidal neurons were obtained as previously described 21 Briefly, high-positive pressure (200–300 mbar) was applied to the pipette (5–8 MΩ) to prevent tip occlusion, when passing the pia. Immediately after, the positive pressure was reduced to prevent cortical damage. The pipette resistance was monitored in the conventional voltage clamp configuration during the descendent pathway through the cortex (until −200 μm from the surface) of 1 μm steps. When the pipette resistance abruptly increased, the 3–5 GΩ seal was obtained by decreasing the positive pressure. After break-in, Vm was measured, and dialysis was allowed to occur for at least 5 min before launching the recording protocols. Current-clamp recordings were made using a potassium-based internal solution (in mM: 135 potassium gluconate, 4 KCl, 10 HEPES, 10 Na2-phosphocreatine, 4 Mg-ATP, 0.3 Na-GTP, and 25 μM, pH adjusted to 7.25 with KOH, 285 mOsM), and acquired using a Multiclamp 700B Amplifier (Molecular Devices). Spontaneous activity was recorded prior, during and after the presentation of auditory stimulation. Spiking pattern of patched cells was analyzed to identify pyramidal neurons. dAP5 (1 mM, Tocris) was topically applied to the dura mater, before whole cell recordings. Offline analysis was performed using custom routines written in Sigmaplot (Systat), IGOR Pro (WaveMetrics) and Matlab (Mathworks).
In vivo optogenetics
After virus injection for ChR2 or ArchT expression, mice were subsequently implanted with fiber optic cannula for optogenetics (CFML22U, Thorlabs) in the BLA. The optic fibers were previously cleaved with a fiber optic scribe (S90R, Thorlabs) at 4.5mm for BLA. The cannula were guided and stereotaxically inserted inside the brain with the help of a cannula holder (XCL, Thorlabs) through the same burr hole used for the viral injections (BLA coordinates from bregma: AP, −1.3mm; DV, −4.5 mm; ML, ±2.9mm) and secured in place with a mix of super glue (Loctite) and dental acrylic and dental cement (Jet Repair Acrylic, Lang Dental Manufacturing). Anesthesia was reversed using AFB-mix for mice assigned to behavioral experiments. For in vivo photostimulation of ChR2-expressing BLA neurons, the fiber optic cannula and the optogenetic patch cable (M83L01, Thorlabs) were connected through a ceramic split mating sleeve (ADAL1, Thorlabs). The patch cable was then coupled to a blue DPSS laser (SDL-473-050MFL, Shanghai Dream Lasers Technology) which was triggered by a pulse-stimulator (Master-9, A.M.P.I), able to synchronize 50 ms laser pulses with 50 ms sound pips composing the CS. For inhibition of BLA-to-FrA projections during learning, in vivo bilateral optic stimulation of ArchT-expressing neurons was achieved by coupling the optic fibers implanted in BLA to a multimode fiber optic coupler (FCMH2-FCL, Thorlabs), with a ceramic split mating sleeve, and subsequently connected to a yellow DPSS laser (SDL-LH-1500, Shanghai Dream Lasers Technology).
In vitro whole-cell recordings
Mice were anesthetized with a mixture of ketamine/xylazine (100mg/kg and 10mg/kg respectively) and cardiac-perfused with ice-cold, oxygenated (95% O2, 5% CO2) cutting solution (NMDG) containing (in mM): 93 NMDG, 93 HCl, 2.5 KCl, 1.2 NaH2PO4, 30 NaHCO3, 25 Glucose, 10 MgSO4, 0.5 CaCl2, 5 Sodium Ascorbate, 3 Sodium Pyruvate, 2 Thiourea and 12mM N-Acetyl-L-cysteine (pH 7.3-7.4, with osmolarity of 300-310 mOsm). Brains were rapidly removed and placed in ice-cold and oxygenated NMDG cutting solution (described above). Coronal slices (300 μm) were prepared using a Vibratome (VT1200S, Leica Microsystems, USA) and transferred to an incubation chamber held at 32°C and containing the same NMDG cutting solution. After this incubation (9-11 min), the slices were maintained at room temperature in oxygenated modified ACSF containing (mM): 92 NaCl, 2.5 KCl, 1.2 NaH2PO4, 30 NaHCO3, 20 HEPES, 25 Glucose, 2 MgSO4, 2 CaCl2, 5 Sodium Ascorbate, 3 Sodium Pyruvate, 2 Thiourea and 12mM N-Acetyl-L-cysteine (pH 7.3-7.4, with osmolarity of 300-310 mOsm) until recording.
Whole-cell recordings of layer 2/3 FrA principal neurons were performed on coronal slices (from bregma: +2.58 mm to +3.08 mm) at 30-32°C in a superfusing chamber. Patch electrodes (3-5 MΩ) were pulled from borosilicate glass tubing and filled with a K-gluconate-based intracellular solution (in mM: 140 K-gluconate, 5 QX314-Cl, 10 HEPES, 10 phosphocreatine, 4 Mg-ATP and 0.3 Na-GTP (pH adjusted to 7.25 with KOH, 295 mOsm). BLA-to-FrA monosynaptic EPSCs were elicited by 1-50 ms light stimulations delivered by an ultrahigh power 460 nm LED (Prizmatix Ltd, Israel). Data were recorded with a Multiclamp700B (Molecular Devices, USA), filtered at 2 kHz and digitized at 10 kHz. Data were acquired and analysed with pClamp10.2 (Molecular Devices).
2-photon laser-scanning microscope (2PSLM)-based calcium imaging
Head-fixed awake mice were placed and trained under the microscope every day for at least 7 days prior to the experiment, and then imaged 21 to 35 days after virus injection using an in vivo non-descanned FemtoSmart 2PLSM (Femtonics, Budapest, Hungary) equipped with a ×16 objective (0.8 NA, Nikon). The MES Software (MES v.4.6; Femtonics, Budapest, Hungary) was used to control the microscope, the acquisition parameters, and the TTL-driven synchronization between the acquisition and auditory/footshock stimuli. The GCaMPs were excited using a Ti:sapphire laser operating at λ=910 nm (Mai Tai DeepSee, Spectra-Physics) with an average excitation power at the focal point lower than 50 mW. Time-series images were acquired within a field-of-view of 300 x 300 μm (256 lines, 1ms/line) for axons; for dendrite: 200 x 60 μm (64 lines, 0.5ms/line). Each imaging session consisted of 30 s of baseline recording followed by 8 gaussian and 8 pure (8kHz)-tone auditory stimuli delivered with pseudo-random delays. We imaged on average 3500 frames (~900 s) per session, and no visible photo-bleaching was observed. Images were then analyzed as previously described 21 using custom routines written in Fiji and Matlab (Mathworks). We registered images over time and corrected XY motion artifacts within a single imaging session by using cross-correlation based on rigid body translation (Stack aligner, Image J, NIH, USA). Motion corrections were then assessed by computing pair-wise 2D correlation coefficient (Image correlation, Image J, NIH, USA), and frames were discarded from the analysis if lower than 0.7. Similar rigid body translation was used to align inter-sessions images with the session 4 (first session post learning) selected as a reference template. Regions of interest (ROIs) for pyramidal neurons and putative axonal boutons were selected and drawn manually. All pixels within each ROI were first averaged providing a single time-series of raw fluorescence. To limit the effect of fluorescence drift over time, the baseline fluorescence (F0) was calculated as the mean of the lower 50% of previous 3 s fluorescence values. Change in fluorescence (ΔFt/F0) was defined as (Ft-F0)/F0, were Ft is the fluorescence intensity at time t (time of the first pixel in each frame). Calcium events were then detected using a template-based method with a custom library of calcium transients. Templates were created by extracting and averaging segments of data that were visually identified as corresponding to a transient. Calcium transients whose peak amplitude reached a 3 X background standard deviation threshold were further considered for analysis. Each detected event was inspected visually and analysis was restricted to detected events rather than on raw fluorescence. For extracting spatial profiles of dendritic calcium events, small ROIs of 2 X 2 pixels are generated along the dendrite by using custom routine in Fiji. The spread of Ca2+ events was then quantified by calculating the full-width at half-max (fwhm, expressed as % of total dendritic length) of the normalized gaussian fit at the time when the averaged ΔF/F0 was maximal.
Data availability and Statistics
All data generated or analyzed during this study are included in the manuscript, and provided in the supplementary statistical table. Data are presented as the median ± interquartile range or mean ± sem (except where stated differently). All statistics were performed using Matlab (Mathworks) and Sigmaplot (Systat) with an α significant level set at 0.05. Normality of all value distributions and the equality of variance between different distributions were first assessed by the Shapiro-Wilk and Levene median tests, respectively. Standard parametric tests were only used when data passed the normality and equal variance tests. Non-parametric tests were used otherwise. Only two-sided tests were used. When applicable, pair-wise multiple post-hoc comparisons were done by using the Holm-Sidak method. Randomization and blinding methods were not used. No statistical methods were used to estimate sample size, but β-power values were calculated for parametric tests.
ACKNOWLEDGEMENTS
We thank Deforges, E. Normand, and B. Darracq (Imetronic) for their technical expertise and support, and A. Holtmaat, P. Fossat, and S. Valerio (AquiNeuro) for their critical reading of our manuscript, and all the members of the Gambino laboratory for technical assistance and helpful discussions. We thank K. Deisseroth and the Standford University, E. Boyden and the MIT, E.J. Kremmer and the IGMM BioCampus Montpellier, L.L. Looger and D. Kim of the GENIE project, and K. Svoboda at the Janelia Farm Research Campus (HHMI) for distributing viral vectors.
This work was supported by the following grants (to FG): FP7 Marie-Curie Career Integration grant 631044, ANR JCJC grant 14-CE13-0012-01, University of Bordeaux and Initiative of Excellence (IdEx) senior chair 2014, Fondation NRJ/Institut de France grant 2015, Laboratory of Excellence (LabEx) Brain grant 2015, and the European Research Council grant ERC-StG-2015-677878.