Holographic stimulation of opposing amygdala ensembles bidirectionally modulates valence-specific behavior via mutual inhibition

Sean C Piantadosi; Zhe Charles Zhou; Carina Pizzano; Christian E Pedersen; Tammy K Nguyen; Sarah Thai; Garret D Stuber; Michael R Bruchas

doi:10.1016/j.neuron.2023.11.007

. Author manuscript; available in PMC: 2025 Feb 21.

Published in final edited form as: Neuron. 2023 Dec 11;112(4):593–610.e5. doi: 10.1016/j.neuron.2023.11.007

Holographic stimulation of opposing amygdala ensembles bidirectionally modulates valence-specific behavior via mutual inhibition

Sean C Piantadosi ^1,², Zhe Charles Zhou ^1,², Carina Pizzano ^1,², Christian E Pedersen ^1,², Tammy K Nguyen ², Sarah Thai ², Garret D Stuber ^1,^2,³, Michael R Bruchas ^1,^2,^3,^4,^*

PMCID: PMC10984369 NIHMSID: NIHMS1946419 PMID: 38086375

Summary

The basolateral amygdala (BLA) is an evolutionarily conserved brain region, well known for valence processing. Despite this central role, the relationship between activity of BLA neuronal ensembles in response to appetitive and aversive stimuli and the subsequent expression of valence-specific behavior has remained elusive. Here we leverage 2-photon calcium imaging combined with single cell holographic photostimulation through an endoscopic lens to demonstrate a direct causal role for opposing ensembles of BLA neurons in the control of oppositely valenced behavior in mice. We report that targeted photostimulation of either appetitive or aversive BLA ensembles results in mutual inhibition and shifts behavioral responses to promote consumption of an aversive tastant or reduce consumption of an appetitive tastant, respectively. Here we identify that neuronal encoding of valence in the BLA is graded and relies on the relative proportion of individual BLA neurons recruited in a stable appetitive or quinine ensemble.

eTOC

The amygdala is a critical brain locus for valence encoding and the expression of valence-specific behavior. Piantadosi et al. report that optically separable valence encoding ensembles are mutually inhibitory and can be precisely manipulated via multiphoton holographic stimulation producing bidirectional changes in consummatory behavior.

Introduction

To ensure survival, organisms must rapidly and accurately assess the safety of their environment and various stimuli. Some stimuli are innately aversive, such as bitter tastes indicating an item is unsafe to consume, while others are innately appetitive, such as sugar rich sweet tasting foods. The perceived value of these stimuli, also known as valence, dictates subsequent approach or avoidance behavior. The basolateral amygdala (BLA) has been identified as a critical node where specific information about stimulus valence is encoded^1–7. However, prior experimental and theoretical exploration of valence coding within the BLA has relied on indirect correlations of neuronal activity with behavior^3,5,8–10.

Within the BLA, spatially intermixed groups of excitatory neurons are known to respond to either appetitive or aversive stimuli (US) with innate valences^3,4,7, and this response selectivity can be imbued on previously neutral stimuli that predict the US⁵. Recent advances have been made in identifying features to subdivide this intermixed population, including genetic markers for positive (Ppp1r1b) and negative (Rspo2) valence neurons that are anatomically distributed¹¹. Valence specificity has also been identified at the circuit level, with projections from the BLA to downstream regions driving appetitive (BLA to nucleus accumbens) or aversive (BLA to central amygdala) behavior^12–15. However, neither feature has been able to fully explain valence encoding in the amygdala, as recordings made from BLA projectors identify substantial heterogeneity in response to stimuli with positive and negative valence³ and previously identified genetic markers project widely, including to regions commonly associated with the opposite valence¹¹. A recent study identified the gene Fezf2 as a potential marker that bridges the gap between genetic and projection specificity, as these neurons respond to stimuli of both positive and negative valence and specificity is determined based on projection target¹⁶. Together these data suggest that at some level, BLA neurons may be hard-wired to encode valence, and emphasize the need to investigate intra-BLA microcircuit mechanisms as potential mediators of valence ensemble specificity.

An interesting potential mechanism for coordinating valence assignment in the BLA is a potential antagonistic relationship observed between positive and negative valence ensembles^17,18. This relationship has been observed at the level of BLA projectors in vivo, with optogenetically tagged BLA projection neurons generating varying degrees of network inhibition³, and ex vivo in genetically defined valence encoding neurons, identifying mutual inhibitory connections between Rspo2 and Ppp1r1b expressing neurons¹¹. This inhibitory motif may also contribute to learned valence, as inhibitory responses to conditioned and unconditioned stimuli emerge during reward and punishment learning⁷. Existing approaches used to manipulate distinct groups of neurons in the BLA have relied on immediate early gene expression⁴, which lacks the sensitivity to detect decreases in neural activity, provides slow and imprecise labeling, and labels neurons that respond to both appetitive and aversive stimuli, which constitute a large population of BLA neurons¹⁹. To ascertain whether the expression of valence-specific behaviors results from the mutually antagonistic activity of discrete BLA ensembles, a method for manipulating the activity of these ensembles in a temporally and spatially precise manner is necessary.

Recent advances in multiphoton imaging, opsin development, and targeted photo-stimulation technology has afforded investigators the ability to conduct simultaneous recording and manipulation of neural activity at the single cell level^20–30. This convergence of technology has led to the ability to directly define causality of discrete neuronal ensembles in mediating a multitude of behaviors^{21,23,24,27,28}. However, recent investigations using this method have focused on relatively superficial cortical regions^20,23,27,29 as well as the hippocampal formation²⁸. Ensemble coding in these brain regions, while stable, is often distributed, highly stochastic, and subject to representational drift^31–33. To date, no experiments have harnessed this technology for interrogation of deep brain²⁶, evolutionarily conserved limbic structures, like the BLA, well known to encode stimulus valence. Here we determined whether valence is stably encoded at the level of distinct opposing groups of single neurons within the BLA, and whether valence-specific changes in neuronal activity are sufficient to bidirectionally control consumption or avoidance behavior.

Results

Optically separable populations of BLA neurons respond to appetitive and aversive stimuli.

To determine how individual principal neurons within the BLA encode opposing valences, we developed an optical approach to simultaneously monitor as well as modulate the activity of individual BLA neurons through a gradient refractive index lens (GRIN). To achieve this, a combination (3:1 ratio) of two adeno-associated viruses encoding either the calcium indicator GCaMP6f (AAV5-CaMKIIa-GCaMP6f) and the red-shifted excitatory opsin ChRmine (AAV8-CaMKIIa-ChRmine-mScarlet-Kv2.1-WPRE) were co-injected unilaterally into the BLA of wildtype (C57BL/6) mice (Figure 1A). A GRIN lens was implanted directly above our viral injection target. Colocalization of GCaMP6f and ChRmine was observed within the BLA underneath the GRIN lens (Figure 1B; Figure S1A), with most GCaMP-expressing cells also expressing ChRmine (3% GCaMP6f only; Fig. S1B). Post-hoc confirmation of GRIN lens placement within the BLA was conducted for each mouse (GRIN lenses spanned −1.46 to −2.18mm from Bregma; Figure S1C).

Figure 1. — (A) Surgical schematic demonstrating co-injection of viral vector to express GCaMP6f and ChRmine in the same BLA neurons and subsequent GRIN lens implantation. (B) Histological image of co-expression of GCaMP6f (green) and ChRmine (red) in the BLA just below GRIN lens. (C) Schematic depiction of head-fixed mouse under the 2-photon microscope and breakdown of trial distribution. On 70% of trials, 10% sucrose (blue) was delivered, while on 30% of trials 2mM quinine (orange) was delivered (20–25s ITI). (D) Example licking rasters aligned to tastant delivery from a single sucrose and quinine delivery session separated by trial type (top – sucrose, bottom – quinine). Lines indicate individual licks. (E) Mean number of licks on sucrose versus quinine delivery trials across mice (t(15)=8.13,p=0.0001). (F) Mean projection image from representative FOV of BLA neurons during sucrose and quinine session. Neurons are colored according to responsivity to tastants. Grey=not activated, Blue=sucrose activated, Orange=quinine activated, Purple=sucrose and quinine activated. (G) Representative continuous traces of neurons colored according to F overlayed on top of individual tastant delivery trials (Sucrose=blue, Quinine=orange). (H) Representative heatmaps of tastant-delivery aligned calcium fluorescence from individual neurons during sucrose (left) or quinine (right) delivery trials for a sucrose activated (top), sucrose and quinine activated (middle), or quinine activated (bottom) neuron. (I) Calcium fluorescence averaged across trials for fluorescence data in (H). (J) Trial-averaged fluorescence of sucrose activated neurons in response to sucrose (top) or quinine (middle) delivery. Mean response across all neurons on sucrose (blue) or quinine (orange) delivery trials (bottom). (K) Trial-averaged fluorescence of sucrose and quinine activated neurons in response to sucrose (top) or quinine (middle) delivery. Mean response across all neurons on sucrose (blue) or quinine (orange) delivery trials (bottom). (L) Trial-averaged fluorescence of quinine activated neurons in response to sucrose (top) or quinine (middle) delivery. Mean response across all neurons on sucrose (blue) or quinine (orange) delivery trials (bottom). (M) Proportions of not activated (grey), sucrose activated (blue), sucrose and quinine activated (purple), or quinine activated (orange) neurons across entire population (n=1177 neurons). (N) Mean fluorescence during post-tastant delivery period for sucrose activated, sucrose and quinine activated, and quinine activated neurons (Main effect of activated cell type [F(1,738)=32.56,p<0.0001], main effect of tastant delivery [F(2,738)=91.46,p<0.0001], and interaction between activated cell type and tastant delivery [F(2,738)=201.3,p<0.0001]. *Post-hoc* comparison of tastant delivery on sucrose activated neurons (t(738)=13.17,p<0.0001) and quinine activated neurons (t(738)=13.93,p<0.0001), no effect of tastant delivery type on sucrose and quinine activated neuron activity p > 0.05). (O) Population activity vectors from one mouse occurring during tastant consumption for sucrose (left; blue) and quinine (right; orange) trials. Each row represents the mean fluorescence for a single neuron on a single trial. (P) Schematic depicting binary linear support vector classifier (SVC) trained on single neuron mean fluorescence (Q) Accuracy of linear SVC for predicting either sucrose or quinine trials (both; black), only sucrose trials (blue), or only quinine trials (orange) across mice (n=13). Filled box-and-whisker plots indicate classifier trained on true fluorescence data. Unfilled box-and-whisker plot represents classifier trained on fluorescence traces that have been randomly shuffled in time. Whiskers represent minimum and maximum accuracy values for each individual mouse. + symbol denotes mean accuracy across mice. Both tastant (t(36)=3.35, p=0.006), sucrose only (t(36)=2.84, p=0.02), quinine only (t(36)=4.3, p=0.0004). ****p<0.0001,***p<0.001, **p<0.01, *p<0.05.

We first trained mice in a behavioral paradigm where they receive random presentations (ITI = 20–25s, total session duration = 20 min) of either un-cued 10% sucrose (sucrose) or 2mM quinine (quinine), highly appetitive and aversive tastants, respectively. To ensure consistent consumption of both tastants, trial presentation was structured such that on 70% of trials sucrose was delivered and on 30% of trials quinine was delivered (Figure 1C)³⁴. The un-cued nature of delivery combined with limited water restriction ensured that mice continued to stably sample each tastant by licking throughout the duration of an entire session (Figure 1D). During a session in which 100% of trials resulted in sucrose delivery, water restricted mice continued to lick consistently with no signs of satiety (Figure S2A-D). By contrast, when 100% of trials resulted in quinine delivery, water restricted mice ceased licking by the end of the session (Figure S2C-D) demonstrating its aversive quality. Mice licked significantly more on sucrose trials relative to quinine trials (Figure 1E). Compared to trials in which drinking water was delivered (Figure S2E), licking frequency was significantly elevated when sucrose was delivered and significantly reduced on quinine trials (Figure S2F-H). During these behavioral sessions the activity of individual excitatory neurons expressing GCaMP6f within the BLA were recorded (total=1177 neurons, n=22 mice, mean neurons per mouse=53.5) (Figure 1F). We compared individual trials of neural activity to shuffled versions of the same activity traces, maintaining identical variance, to classify neurons as either only sucrose activated, both sucrose and quinine activated, or only quinine activated (Figure 1G). Individual calcium indicator responses for classified neurons were robust and consistent to their preferred tastant across trials, and sucrose- or quinine-only activated neurons displayed negligible responses to the unpreferred tastant (Figure 1H-I). This same activity pattern was observed in the trial-averaged data across mice, with sucrose-activated neurons displaying robust increases in activity in response to sucrose, but not quinine (Figure 1J), sucrose and quinine-activated neurons exhibiting similar increases in their activity in response to both the appetitive and aversive tastant (Figure 1K), and quinine neurons showing a robust increase in activity specifically to quinine (Figure 1L). Across the entire population of BLA neurons, 22% of neurons (261/1177) were identified as sucrose and quinine activated neurons, while 16% (191/1177) and 26% (310/1177) of neurons were sucrose only and quinine only activated, respectively (Figure 1M). The mean calcium fluorescence for each neuron following tastant delivery (0 to 10s) showed strong and consistent increases in fluorescence for sucrose-activated neurons and quinine-activated neurons to their preferred trial conditions respectively, while no difference in fluorescence was detected for neurons classified as both sucrose and quinine activated (Figure 1N). Neurons identified as sucrose-only activated displayed positive correlations between mean fluorescence and licking behavior (Figure S2I), while neurons identified as quinine-only activated exhibited negative correlations between mean fluorescence and licking behavior on a trial-by-trial basis (Figure S2J). Across all classified cells, the correlation between licking and fluorescence was significantly different between sucrose- and quinine-activated only neurons (Figure S2K). While the responses correlate with licking magnitude, this fluorescence response reflects stimulus valence rather than simply a difference in licking structure, as the mean fluorescence for each sucrose- or quinine-activated neuron is significantly different when analysis was restricted to the immediate post-delivery period (1s post-delivery) where licking behavior is not different between trial types (Figure S2L-N). For each classified group, mean activity on preferred tastant trials (e.g. sucrose trials for sucrose-only activated neurons) was significantly different compared to non-preferred tastant (e.g. quinine trials for sucrose-only activated neurons) beginning 1s after tastant delivery (Figure S2O). Single neuron responses to sucrose or quinine were also stable across multiple sessions (Figure S3A-F) both in terms of amplitude of the response (Figure S3G) and the temporal dynamics of the fluorescence signal (Figure S3H-I). Finally, using the mean fluorescence response for each neuron on each tastant delivery trial (Figure 1O) we trained a binary linear support vector classifier (SVC), ideally suited for two-class problems, to predict tastant trial type (Figure 1P). Mean BLA population activity in the post-tastant epoch reliably predicted whether sucrose or quinine were delivered compared to a model trained on shuffled calcium fluorescence data, containing equal mean fluorescence and variance as the unshuffled calcium fluorescence. Classifier accuracy was also significantly better than shuffled control data for both individual tastant (sucrose or quinine) delivery trials (Figure 1Q). These results suggest that the activity of groups of individual BLA neurons stably encode the valence of a stimulus.

Spatial distribution of valence-encoding neurons within the BLA

Previous work suggests that, while valence encoding BLA neurons are largely intermixed throughout the BLA^2,4, spatial gradients in positive and negative valence neurons may exist across its extent^3,11. To obtain clarity on whether spatial gradients exist, we harnessed the resolution of 2-photon imaging and carefully mapped our imaging field of views and individual neuron location (n=1177) with histological registration of our lens placement (Figure S1C). We first evaluated whether there were any differences in spatial location of neurons identified as sucrose-activated or -inhibited, finding no difference in medial/lateral distribution (Figure 2A), but significantly greater proportions of sucrose-only inhibited neurons in the posterior (Figure 2B) and dorsal (Figure 2C) BLA relative to sucrose-only activated neurons. Like the sucrose responsive neurons, quinine-activated and -inhibited neurons were distributed similarly in the medial/lateral extent of the BLA (Figure 2D). Again, a significant bias toward a greater proportion of inhibited neurons located in the posterior BLA relative to activated neurons was identified for quinine neurons (Figure 2E). In contrast to sucrose responsive neurons, quinine inhibited neurons were more prevalent in the ventral BLA compared to quinine activated neurons (Figure 2F). These data suggest a consistent pattern of greater positive- and negative-valence inhibited neurons in the more posterior segment of the BLA, while these neurons are distributed in an opposite pattern in the dorsal/ventral axis, highlighting the potential importance of these negatively modulated ensembles.

Figure 2. — (A) No difference in sucrose-only activated and inhibited probability distribution function as a function of medial/lateral distance (top; Kolmogorov-Smirnov test, D=0.14, p=0.12). Scatter points represent individual neuron locations along the medial/lateral and anterior/posterior axis, light blue = inhibited neurons, dark blue = activated neurons (bottom). (B) Greater proportion of sucrose-only inhibited neurons relative to sucrose-only activated neurons in the posterior BLA relative to anterior (top; Kolmogorov-Smirnov test, D=0.22, p=0.001). Scatter points represent individual neuron locations along the anterior/posterior and dorsal/ventral axis, light blue = inhibited neurons, dark blue = activated neurons (bottom). (C) Greater proportion of sucrose-only inhibited neurons relative to sucrose-only activated neurons in the dorsal BLA relative to ventral BLA (top; Kolmogorov-Smirnov test, D=0.38, p<0.0001). Scatter points represent individual neuron locations along the dorsal/ventral and medial/lateral axis, light blue = inhibited neurons, dark blue = activated neurons (bottom). (D) No difference in quinine-only activated and inhibited probability distribution function as a function of medial/lateral distance (top; Kolmogorov-Smirnov test, D=0.13, p=0.17). Scatter points represent individual neuron locations along the medial/lateral and anterior/posterior axis, light orange = inhibited neurons, dark orange = activated neurons (bottom). (E) Greater proportion of quinine-only activated neurons in the anterior BLA while quinine-only inhibited neurons are more prevalent in posterior BLA (top; Kolmogorov-Smirnov test, D=0.26, p<0.0001). Scatter points represent individual neuron locations along the anterior/posterior and dorsal/ventral axis, light orange = inhibited neurons, dark orange = activated neurons (bottom). (F) Greater proportion of quinine-only inhibited neurons relative to quinine-only activated neurons in the ventral BLA relative to dorsal BLA (top; Kolmogorov-Smirnov test, D=0.40, p<0.0001). Scatter points represent individual neuron locations along the dorsal/ventral and medial/lateral axis, light orange = inhibited neurons, dark orange = activated neurons (bottom). (G) 3-dimensional scatterplot of tastant-activated BLA neuron locations in space (blue = sucrose only, purple = sucrose and quinine, orange = quinine only, grey = not activated). Compass length = 0.5mm. (H) No difference in location of sucrose- or quinine-only activated neurons as a function of medial/lateral location (top; Kolmogorov-Smirnov test, D=0.09, p<0.32). Scatter points represent individual neuron locations along the medial/lateral and anterior/posterior axis, colored according to panel G. (I) Greater number of quinine-only activated neurons in the anterior BLA relative to sucrose-only activated neurons (top; Kolmogorov-Smirnov test, D=0.17, p=0.005). Scatter points represent individual neuron locations along the anterior/posterior and dorsal/ventral axis, colored according to panel G. (J) Greater number of sucrose-only activated neurons in the ventral BLA relative to quinine-only activated neurons (top; Kolmogorov-Smirnov test, D=0.27, p<0.0001). Scatter points represent individual neuron locations along the anterior/posterior and dorsal/ventral axis, colored according to panel G. ****p<0.0001, ***p<0.001.

We next compared positive- and negative valence neurons (neurons selectively activated to sucrose or quinine) as a function of 3-dimensional location within the BLA (Figure 2G). No difference in the spatial distribution of sucrose- and quinine-activated neurons across the medial/lateral extent of the BLA (Figure 2H). A greater proportion of quinine-activated neurons were in the anterior BLA relative to sucrose-activated neurons (Figure 2I), while more sucrose-activated neurons were identified in the ventral BLA compared to quinine activated neurons (Figure 2J). Except for the medial/lateral axis, we identified significant gradients of positive and negative-valence neurons across the BLA. These data identify several novel gradients, specifically focusing on neurons that show reduced activity in response to appetitive and aversive stimuli (Figure 6C,F) and also lend support for previous observations that focused on the spatial location of neurons with increases in neural activity^3,11. Despite these significant gradients, considerable overlap was identified at all spatial locations establishing the intermixed nature of valence encoding ensembles⁴ and highlighting the necessity of a tool with high spatial specificity for achieving ensemble specific manipulation.

Figure 6. — (A) Heatmaps of sucrose inhibited and quinine activated neurons (n=61; left), and quinine inhibited and sucrose activated neurons (n=86; right). (B) Mean fluorescence post-tastant delivery for sucrose inhibited/quinine activated neurons and quinine inhibited/sucrose activated neurons on sucrose delivery and quinine delivery trials. Two-way repeated measures ANOVA main effect of opposing cell type [F(1,145)=5.6, p=0.019] and interaction between opposing cell type and tastant delivery [F(1,145)=168.5, p<0.0001]. The mean post-delivery fluorescence for sucrose inhibited/quinine activated neurons was significantly elevated on sucrose delivery trials relative to quinine (Sidak’s multiple comparisons test (t(145)=10.03, p<0.0001). Mean post-delivery fluorescence for quinine inhibited/sucrose activated neurons was significantly elevated on quinine trials relative to sucrose delivery trials (Sidak’s multiple comparisons test (t(145)=8.24, p<0.0001). (C) Of neurons identified as significantly inhibited in response to sucrose, 40% of those are activated in response to quinine. Of neurons identified as significantly inhibited by quinine, 55% of those are significantly activated in response to sucrose. (D) Difference map from a single FOV following sucrose delivery (blue outline) and quinine delivery (orange outline). Red indicates increased pixel intensity change from baseline, blue indicates reduced pixel intensity change from baseline. Proximal pairs of activated and inhibited neurons are identified (pairs 1–4). (E) Example traces of proximal pairs identified in D. Blue traces indicate sucrose activated and quinine inhibited neurons, orange traces indicate quinine activated and sucrose inhibited neurons. Vertical lines indicate sucrose (blue) or quinine (orange) delivery. (F) Schematic depicting quantification of cluster distance using Euclidean distance. (G; left) 2D scatterplot (ML and AP) of neuron center points from a single mouse FOV. Orange filled circles indicate quinine activated/sucrose inhibited neuron location while blue filled circles indicate sucrose activated/quinine inhibited neuron location (left). Open circles denote mean cluster location, colored according as described previously. Blue and orange shading indicates outline of antagonistic cluster location. Grey filled circles indicate non-antagonistic neuron locations. (G; right) 2D scatterplot (ML and AP) of neuron center points from a single FOV for non-opposing ensembles (e.g. sucrose activated neurons with no significant response to quinine, and quinine activated neurons with no significant response to sucrose). (H) Mean Euclidean distance between opposing ensembles and non-opposing ensembles (n=13/22 mice with antagonistic ensembles, Paired-samples t-test t(12)=2.22, p=0.046). (I) Mean projection image of FOV with location of sucrose activated/quinine inhibited neurons (blue outlines) targeted for multiphoton activation (red spirals). Orange outlines indicate quinine activated/sucrose inhibited ensemble. (J) Difference map relative to pre-multiphoton activation of sucrose activated/quinine inhibited neurons (blue outlines, red spirals). Orange outlines surround quinine activated/sucrose inhibited ensemble neurons. (K) Multiphoton activation of antagonistic sucrose ensemble neurons (n=8, representative from single mouse) produces robust activation of targeted neurons (blue trace). Antagonistic quinine ensemble neurons (n=6, orange trace) display reductions in fluorescence during photostimulation of the sucrose ensemble. Heatmaps showing mean response across stimulation trials (15 trials) for targeted (top; sucrose activated / quinine inhibited neurons) and non-targeted quinine activated / sucrose inhibited neurons (bottom). (L) As expected, photostimulation produced robust activation of neurons targeted for activation (sucrose activated/quinine inhibited ensemble; n=34 neurons, 4 mice) relative to baseline fluorescence (Two-tailed paired t-test; t(33)=8.6, p<0.0001). (M) Photostimulation of the sucrose activated/quinine inhibited ensemble produced a significant reduction in fluorescence of quinine activated / sucrose inhibited (n=28 neurons, 4 mice) relative to baseline fluorescence (Two-tailed paired t-test; t(27)=8.86, p<0.0001). (N) Difference map relative to pre-multiphoton activation of sucrose activated/quinine inhibited neurons (blue outlines, red spirals). Orange outlines surround quinine activated/sucrose inhibited ensemble neurons. (O) Multiphoton activation of antagonistic quinine ensemble neurons (n=6, representative from same mouse as in **J-M**) produces robust activation of targeted neurons (orange trace). Antagonistic sucrose ensemble neurons (n=8, orange trace) display reductions in fluorescence during photostimulation of the quinine ensemble. Heatmaps showing mean response across stimulation trials (15 trials) for targeted (top; quinine activated / sucrose inhibited neurons) and non-targeted sucrose activated / quinine inhibited neurons (bottom). (P) As expected, photostimulation produced robust activation of neurons targeted for activation (quinine activated/sucrose inhibited ensemble; n=28 neurons, 4 mice) relative to baseline fluorescence (Two-tailed paired t-test; t(27)=9.36, p<0.0001). (Q) Photostimulation of the quinine activated/sucrose inhibited ensemble produced a significant reduction in fluorescence of sucrose activated/quinine inhibited ensemble (n=34 neurons, 4 mice) relative to baseline fluorescence (Two-tailed paired t-test; t(33)=6.94, p<0.0001). ****p<0.0001, *p<0.05.

Holographic photostimulation via a GRIN lens is specific and robust at activating amygdala ensembles

Having established that intermixed but spatially separable populations of BLA neurons stably encode stimulus valence we tested whether these correlative changes in activity were sufficient to drive corresponding appetitive and aversive behavioral responding. We used a spatial light modulator (SLM)^22,26 to photostimulate predefined neurons in the field of view (Figure 3A) expressing the red-shifted excitatory opsin ChRmine²⁹ (Figure 3B). We first calibrated and determined the specificity and physiological resolution tolerance of single-neuron stimulation in the X/Y plane. To achieve this, we targeted a single neuron in our field of view expressing both GCaMP6f and ChRmine for spiral stimulation (10hz, 2s, 8mW power per ROI at the sample). We then proceeded to move the spiral stimulation target away from the cell body at a distance of 10 μm each trial (roughly 1 cell-body width, 5 trials per ROI, randomized) for 5 sequential stimulation targets up to 40 μm away from the targeted neuron (Figure 3C). We found that photoactivation of originally targeted neuron fell off dramatically as the stimulation ROI was moved laterally (Figure 3D), such that once the spiral stimulation target was > 1 cell body width (~20 μm) away there was no change in fluorescence of the original neuron (Figure 3E), indicating a very narrow window of control over the x/y cartesian plane at the single neuron level. We then conducted a subsequent experiment to determine the resolution specificity of stimulation in the Z-axis by first identifying neurons for stimulation in a recording plane and then, for different trials, stimulating at various planes above and below the recording plane. To record the effect of out of plane stimulation on the fluorescent activity of the target neuron in the recording plane, we coupled an electrically tunable lens (ETL) to the microscope imaging pathway which was used to near-simultaneously (< 5ms) compensate for the change in microscope Z-position to image the recording plane while maintaining the photostimulation z-depth location (Figure 3F). We next measured and corrected for the amount magnification caused by the GRIN lens at varying depths (Figure 3G; Figure S4A-C). Using this method, we established a clear distance-response relationship, with stimulation targeted to the recording plane (±0μm) eliciting the greatest change in fluorescence of the target neuron (Figure 3H). At the extremes of the microscope Z-position movements (−26 and +33μm in converted distance) limited activation of the target neuron (n=6 neurons) was seen relative to stimulation within the recording plane across 5 individual trials (Figure 3I). Across all distances, we found a similar relationship between distance and stimulation evoked response to what we observed in Figure 3D, with the effects of out of plane photostimulation negligible at distances of 20 μm above or below the imaging plane (Figure 3J). Finally, we found that multiphoton stimulation through a GRIN lens scaled with the number of neurons targeted for stimulation, allowing for robust activation of multiple neurons simultaneously with no change in the timing or amplitude of the photostimulated response as a function of neuron number (Figure 3K). Further, repeated stimulation (ITI 20–25s) of the same neurons produced consistently similar increases in fluorescence and did not decrease in magnitude over the course of a session (Figure 3L). Together, these results establish specificity of photostimulation through a GRIN lens within the BLA and a working paradigm for targeted, repeatable control of individual neuron activity using SLM deep in the brain of an awake, behaving mouse. These findings demonstrate holographic SLM stimulation through a GRIN lens, is capable of robust, precise, and consistent activation of a priori optically identified neurons.

Figure 3. — (A) Schematic of head-fixed 2-photon calcium imaging via 920nm laser path combined with a second 1040nm laser path for stimulation. (B) *In vivo* imaging of GCaMP6f (920nm) or the red-shifted opsin ChRmine (1080nm) expressed in BLA neurons. (C) Schematic depicting experimental setup for testing X/Y specificity of single ROI activation using a spatial light modulator (SLM). Red spirals indicate SLM targets, with each subsequent stimulation moving a fixed distance (+10, +20, +30, +40μm) laterally from the target cell. (D) Mean fluorescence across 5 stimulation trials (10hz, 2s, 6mW power) at each distance, darker colors indicate SLM spiral target being closer to the target cell. (Inset) Mean projection image of neuron recorded from (green) and spiral stimulation targets moving laterally. Order of spiral stimulation target (1–5) was randomized. (E) Mean fluorescence of single BLA neuron across 5 stimulation trials at each SLM ROI plotted as a function of distance from the target neuron. (F) Schematized representation of strategy to assess axial specificity of single-cell SLM stimulation. To stimulate (10hz, 2s, 6mW power) both above and below our target cell, the microscope Z-position was moved at fixed distances (−90 to +90μm in microscope space). At the same time, a separate imaging path with an electrotunable lens (ETL) was used to counteract this microscope movement to ensure imaging of the plane where our target cell was located. (G) Sample position and corresponding objective position for determining the degree of axial magnification caused by the GRIN lens. (H) Trial averaged fluorescence of a single BLA neuron plotted as a function of both time (left Y-axis) and stimulation distance in microscope units (X-axis). (I) Heatmaps showing stimulation aligned fluorescence traces (dotted line indicates onset of SLM stimulation) for 3 distances across 5 repeated stimulation trials (ITI=15s). (J) SLM stimulation evoked fluorescence averaged across all neurons (n=6) plotted as a function of distance corrected for GRIN lens magnification (brain distance in μm). Negative numbers indicate SLM stimulation ROIs above the target neuron, while positive numbers indicate SLM stimulation ROIs below. (K) Simultaneous activation of 6 BLA neurons through a GRIN lens using SLM stimulation (10hz, 2s, 6mW power per ROI; ITI=20–25s). Yellow box indicates first stimulation of the session, green box indicates final stimulation of the session. (L) Aligned SLM stimulation evoked fluorescence from the first stimulation (left; yellow line) and the final (right; green line) stimulation. Traces are shaded according to their color in K.

Targeted photostimulation of quinine ensemble reduces consummatory behavior to an appetitive stimulus

We next tested the central hypothesis for a direct causal link between these separable neural responses and behavior. We used a behavioral paradigm in which spiral SLM stimulation (2 s, 10 Hz activation, power per ROI 5–10 mW) occurred randomly on 50% of all tastant (35% sucrose trials and 15% quinine trials) delivery trials (Figure 4A). We first tested whether selective activation of the aversive, quinine-responsive ensemble altered licking behavior (Figure 4B). On average we targeted 13.2 quinine neurons per mouse (range 4 to 25/mouse, total=172) for SLM stimulation. We evaluated whether SLM stimulation of the quinine ensemble produced neuronal activity changes selectively in quinine activated neurons and not in sucrose activated neurons identified during baseline sessions (Figure 4C-D). On trials without holographic stimulation of the quinine ensemble, sucrose neurons maintained their preference for sucrose over quinine (Figure 4E-F; top, n = 6 neurons from representative mouse), consistent with their activity during baseline sessions (Figure 1). Critically, the activity of sucrose neurons was not affected by SLM stimulation of the quinine ensemble, with these neurons exhibiting a similar activity increase selectively to sucrose even during quinine ensemble stimulation trials (Figure 4E-F; bottom). By contrast, quinine-selective neurons displayed a typical preference for quinine delivery over sucrose on trials without SLM stimulation of the quinine ensemble (Figure 4G-H; top, n = 8 neurons from representative mouse). However, quinine ensemble neuron activity was not different on trials in which either sucrose or quinine was delivered concurrently with quinine ensemble activation (Figure 4G-H; bottom). Across all quinine neurons from all mice, the mean fluorescence in response to quinine delivery on trials without targeted photostimulation of the quinine ensemble was significantly elevated relative to sucrose delivery trials (Figure 4I). When SLM photostimulation of the quinine ensemble was paired with sucrose delivery, a significant increase in fluorescence was detected relative to the no stimulation sucrose delivery trials. Importantly, no significant difference in mean fluorescence was observed between trials in which photostimulation was paired with sucrose or quinine, establishing specific “play back” of this quinine ensemble activity pattern on sucrose delivery trials in which there would typically have been no activity in quinine neurons. No change in fluorescence was detected when quinine activated neurons were targeted for SLM stimulation in mice not expressing ChRmine (Figure S5A-D). These data indicate that SLM activation of the aversive quinine ensemble was specific in both the spatial and temporal dimensions during our sessions.

Figure 4. — (A) Schematic of trial structure for SLM stimulation sessions. 70% of trials resulted in 10% sucrose (blue) delivery, while 30% of trials resulted in 2mM quinine (orange) delivery. SLM stimulation (10hz, 2s, 6–10mW power per ROI) was randomly delivered simultaneously with tastant delivery on 50% of all trials. (B) SLM stimulation ROIs were targeted to quinine-activated (quinine-ensemble) neurons. (C) Mean projection image of FOV from a single mouse. BLA neuron contours are colored according to their activity during the sucrose and quinine delivery session. Blue contours indicate neurons that were activated in response to sucrose only, while orange contours indicate neurons activated in response to quinine only. SLM spiral stimulation targets indicate the 8 quinine activated neurons (quinine ensemble) targeted for stimulation. (D) Example traces from 3 sucrose (blue) and quinine (orange) activated neurons labeled in C. Traces are overlayed on top of tastant delivery trials (sucrose=blue, quinine=orange) as well as trials in which SLM stimulation of the quinine ensemble occurred (red shading). (E) Mean fluorescence activity of sucrose activated neurons in response to sucrose (blue) and quinine (orange) delivery on trials without stimulation (top; two-way repeated measures ANOVA, main effect of trial type F(1,10)=9.6, p=0.011, no main effect of time F(187, 1870)=1.02, p=0.43, interaction between trial type and time F(187, 1870)=3.51, p<0.0001; bar indicates bins where t≥3.68, p≤0.044 Sidak’s multiple comparison correction) and trials with SLM stimulation (bottom; two-way repeated measures ANOVA, no main effect of trial type F(1,10)=4.6, p=0.06, main effect of time F(187, 1870)=4.3, p<0.0001, interaction between trial type and time F(187, 1870)=3.54, p<0.0001; bar indicates bins where t≥3.68, p≤0.043 Sidak’s multiple comparison correction). Dotted line indicates tastant delivery. Red shading indicates onset of quinine ensemble stimulation via SLM. (F) Spatial map of mean pixel intensity during the post-sucrose delivery period subtracted from mean pixel intensity immediately preceding sucrose delivery (−5s to 0s) during trials without quinine ensemble stimulation (top) and SLM stimulation of the quinine ensemble (bottom). Blue triangles indicate location of sucrose activated neurons identified in C and D. Red shading indicates a greater difference in activity during the post-sucrose delivery period relative to baseline. Blue shading indicates reduced activity. (G) Mean fluorescence activity of quinine activated neurons in response to sucrose (blue) and quinine (orange) delivery on trials without stimulation (top; two-way repeated measures ANOVA, no main effect of trial type F(1,14)=5.64, p=0.03, main effect of time F(187, 2618)=1.67, p<0.0001, interaction between trial type and time F(187, 2618)=3.82, p<0.0001; bar indicates bins where t≥3.72, p≤0.038 Sidak’s multiple comparison correction) and trials with SLM stimulation (bottom; two-way repeated measures ANOVA, no main effect of trial type F(1,14)=2.2, p=0.16, main effect of time F(187,2618)=35.64, p<0.0001, no interaction between trial type and time F(187, 2618)=0.84, p=0.94). Dotted line indicates tastant delivery. Red shading indicates onset of quinine ensemble stimulation via SLM. (H) Spatial map of mean pixel intensity during the post-quinine delivery period subtracted from mean pixel intensity immediately preceding quinine delivery (−5s to 0s) during trials without quinine ensemble stimulation (top) and SLM stimulation of the quinine ensemble (bottom). Orange triangles indicate location of quinine activated neurons identified in C and D. Red shading indicates a greater difference in activity during the post-quinine delivery period relative to baseline. Blue shading indicates reduced activity. (I) Mean fluorescence in the post-delivery period for quinine neurons in response to sucrose and quinine delivery on no stimulation versus SLM stimulation (two-way repeated measures ANOVA main effect of stimulation [F(1,134)=42.1, p<0.0001], main effect of tastant [F(1,134)=26.96, p<0.0001], , stimulation X tastant interaction [F(1,134)=33.53, p<0.0001]. Sidak’s multiple comparison correction sucrose no stimulation vs. sucrose stimulation (t(134)=9.78, p<0.0001), sucrose no stimulation vs quinine no stimulation (t(134)=8.33, p<0.0001), all other p>0.05. (J) Lick rasters from a single mouse organized by trial type and aligned to sucrose delivery on trials without quinine ensemble SLM stimulation (top) and on trials with quinine ensemble stimulation (bottom; red shading). (K) Licking probability across all mice on sucrose delivery trials without stimulation (blue) and on trials with simultaneous quinine ensemble SLM stimulation (red) (two-way ANOVA, main effect of trial type F(1,84027)=14.3, p=0.0002), main effect of time F(110, 84027)=219.1, p<0.0001, and interaction between trial type and time F(110, 84027)=3.2, p<0.0001; bar indicates significant bins t≥3.56 and p≤0.042). (L) The mean number of licks per trial per mouse on sucrose delivery trials without quinine ensemble stimulation and on trials with quinine ensemble SLM stimulation (red shading) (two-tailed paired t-test; t(13)=4.99, p=0.0002). (M) Correlation between the number of quinine ensemble neurons stimulated via SLM and the percent change in licking on sucrose trials (No stimulation vs quinine ensemble stimulation trials) (R²=0.65, linear regression F(1,2)=22.06 p=0.0005). (N) Lick rasters from a single mouse organized by trial type and aligned to quinine delivery on trials without quinine ensemble SLM stimulation (top) and on trials with quinine ensemble stimulation (bottom; red shading). (O) Licking probability across all mice on quinine delivery trials without stimulation (blue) and on trials with simultaneous quinine ensemble SLM stimulation (red) (two-way ANOVA, main effect of trial type F(1, 20160)=11.71, p=0.0006, main effect of time F(111, 20160)=12.9, p<0.0001, no interaction between trial type and time F(111, 20160)=0.78, p=0.96). (P) The mean number of licks per trial per mouse on quinine delivery trials without quinine ensemble stimulation and on trials with quinine ensemble SLM stimulation (red shading) (two-tailed paired t-test; t(13)=0.12, p=0.90). (Q) Correlation between the number of quinine ensemble neurons stimulated via SLM and the percent change in licking on quinine trials (No stimulation vs quinine ensemble stimulation trials) (R²=0.28, linear regression F(1,12)=4.3 p=0.07). **p<0.01.

We next determined whether playing back this quinine ensemble activity during trials in which an appetitive tastant (sucrose) was delivered alters consummatory behavior. Comparing sucrose consumption during trials without quinine ensemble stimulation with trials in which the aversive quinine ensemble activation was played back via SLM stimulation (Figure 4J) we found that lick probability was significantly reduced during stimulation trials (Figure 4K). Furthermore, the mean number of licks to the appetitive sucrose was significantly reduced during trials in which the quinine ensemble was activated compared to trials without SLM stimulation (Figure 4L). No change in licking behavior was observed when SLM stimulation was targeted to quinine ensembles in mice without ChRmine expression (Figure S5E-G), further supporting the conclusion the observed reduction in licking is not an artifact of SLM stimulation itself. A significant positive correlation was found between the number of quinine neurons stimulated per mouse and the reduction in licking, indicating that activating a larger quinine ensemble produces more profound reductions in consumption of sucrose (Figure 4M). By comparison, when quinine neurons were stimulated during quinine delivery trials within the same session, no change in the probability of licking to quinine delivery (Figure 4N-O) or on the mean number of licks per trial (Figure 4P) was detected compared to trials without stimulation. No correlation between the percentage change in licking during quinine trials and the number of neurons stimulated was detected (Figure 4Q). These results suggest that the endogenous activation of these neurons likely reached a ceiling and the additional activation generated by targeted photostimulation (Figure 4I) was not sufficient to drive any further reduction in licking. These data suggest that the neural representation of an innately aversive stimulus within the BLA is sufficient to produce aversive behavioral responding.

Targeted photostimulation of sucrose ensemble increases consummatory behavior of aversive stimulus

We next investigated if activation of the appetitive, sucrose-responsive ensemble could produce the opposite behavioral effect (Figure 5A-B). Like quinine-ensemble stimulation, activation of the sucrose ensemble (10hz, 2s, 5–10mW power per ROI) was highly specific to sucrose and not quinine activated neurons which we identified during recordings from baseline sucrose and quinine delivery sessions (Figure 5C-D). Neurons that were identified as sucrose activated did not display significant activation on quinine delivery trials without sucrose-ensemble stimulation (Figure 5E; top). However, on trials in which sucrose ensemble stimulation (n=10 neurons from representative mouse) was paired with sucrose or quinine delivery, activity was not different between trial types (Figure 5E; bottom). Importantly, activity of quinine-responsive neurons was not impacted by photostimulation of the sucrose ensemble, as increases in activity on stimulation and non-stimulation trials were only observed in response to quinine and not sucrose delivery (Figure 5G-H). Across all sucrose neurons (n=115 total, mean per mouse=12.8, range 5–28 neurons) recorded from all mice and targeted for photostimulation (Figure 5I) we found that photostimulation produced a significant increase in fluorescence when paired with quinine delivery, while no significant increase in fluorescence was detected on sucrose delivery trials. No difference in mean fluorescence was detected when photostimulation was paired with sucrose or quinine. Thus, SLM stimulation of the sucrose ensemble was specific to sucrose activated neurons, indicating that the endogenous pattern of activity evoked by sucrose delivery could be replayed on trials in which there would have been no activity (quinine-delivery trials). These findings strengthen the conclusion that it is possible recapitulate the spatial and temporal specificity of the endogenous response elicited by consumption of a highly appetitive tastant.

We next examined the effect of sucrose-ensemble SLM stimulation on consummatory licking behavior in response to either sucrose or quinine. In a similar manner to our observation when activating the quinine ensemble during quinine delivery trials (Figure 4G), photostimulation of the sucrose ensemble on sucrose delivery trials did not change licking probability (Figure 5J-K) nor the mean number of licks per trials (Figure 5L) compared to trials without SLM stimulation. No significant correlation was observed between the number of neurons activated and the mean change in licking between sucrose stimulation and non-stimulation trials (Figure 5M). By contrast, when the appetitive sucrose ensemble was activated during aversive quinine presentation, licking was significantly increased across trials compared to trials without sucrose ensemble SLM stimulation (Figure 5N). Lick probability was non-significantly increased on trials in which the quinine delivery was paired with SLM activation of the sucrose ensemble relative to no stimulation trials (Figure 5O). However, comparing the mean number of licks per mouse in response to quinine delivery on non-stimulation trials versus trials in which the sucrose ensemble was activated, mice had a significant increase in number of licks during stimulation trials (Figure 5P). Additionally, a significant positive correlation was observed between the number of activated sucrose neurons and the percentage change in mean licks, supporting the concept that stimulation of greater numbers of sucrose ensemble neurons produced larger increases in quinine licking compared to trials without stimulation (Figure 5Q). Together these data suggest that selective recruitment of an appetitive neural ensemble is capable of biasing consummatory behavior in response to an aversive tastant toward a more appetitive-like response. In summary, this all-optical deep brain approach demonstrates that separable populations of neurons within the BLA stably encode appetitive or aversive behavioral responses, and that selective photoactivation of a small neuronal population is capable of bidirectionally influencing consummatory behavior.

Antagonistic relationship between valence-encoding BLA neurons drives mutual inhibition

Previous reports suggest that antagonistic relationships exist between BLA projections known to regulate behaviors of opposing valences^3,12 and functional antagonism has been shown between genetically distinct valence encoding neurons ex vivo¹¹. This potential antagonism provides a potential mechanism by which information about stimulus valence is amplified and transmitted downstream. To test this mechanism, we first examined whether functional antagonism existed within our dataset. We found that many positive (sucrose activated) and negative (quinine activated) valence neurons displayed inverse responses to the opposite tastant (Figure 1J-L). This opposing relationship was more prominent when neurons classified as inhibited selectively to sucrose or quinine were examined (Figure S6A-D). We next identified only neurons that had significant antagonistic activity profiles (e.g. sucrose activated neurons that were quinine inhibited, and quinine activated neurons that were sucrose inhibited; Figure 6A-B), we found that 40% of sucrose inhibited neurons were also quinine activated while 55% of quinine inhibited neurons were also activated by sucrose. Interestingly, compared to the entire population of positive- and negative-valence ensembles, antagonistic ensembles did not show any significant spatial gradient in any dimension (anterior/posterior, medial/lateral, dorsal/ventral) suggesting these particular opposing ensembles have an intermixed distribution throughout the BLA at large (Figure S6E-G).

However, we did observe that within individual fields of view, cells from these antagonistic ensembles were often located proximally, such that a sucrose activated/quinine inhibited neuron was located near a quinine activated/sucrose inhibited neuron (Figure 6D-E). To quantify this phenomena, we calculated the Euclidean distance between the means of the opposing clusters (Figure 6F), and compared this value to mean distance between non-opposing valence ensembles (e.g. sucrose activated/quinine unaffected, and quinine activated/sucrose unaffected ensembles; Figure 6G). We found that the mean Euclidean distance between opposing ensembles was lower in opposing ensembles compared to non-opposing ensembles (Figure 6H). We posit that this proximity might indicate functional connections between these opposing ensembles. Given that previous work suggests that functional antagonism between genetically distinct valence-encoding neurons ex vivo, we again utilized multiphoton optogenetics to assess connectivity between these ensembles (Figure 6I). After performing initial mapping of valence ensembles as we did previously (Figure 1C), we conducted sessions in which spiral stimulation was targeted to antagonistic ensembles, starting with the sucrose activated / quinine inhibited ensemble (Figure 6J). We found that stimulation produced robust activation of the targeted sucrose activated / quinine inhibited ensemble (Figure 6K,L). However, in the non-stimulated opposing ensemble (quinine activated / sucrose inhibited) we observed a significant reduction in fluorescence, indicative of a suppression of activity (Figure 6K,M). On sessions in which quinine activated / sucrose inhibited ensembles were targeted for spiral stimulation (Figure 6N), we again observed robust and specific activation of this ensemble (Figure 6O,P). The effects of stimulation on the opposing sucrose activated / quinine inhibited ensemble, resulted in a significant suppression of fluorescence across neurons (Figure 6O,Q). For both opposing ensembles not targeted for stimulation, the fluorescence response was predominantly negative. Although quinine ensemble activation resulted in fewer significantly inhibited opposing neural responses (recording sucrose) relative to sucrose ensemble activation (recording quinine ensemble; Figure S6J). Taken together, these data indicate that opposing valence BLA neuronal ensembles engage in a functional antagonism via mutual inhibition.

Discussion

Here we report intermixed groups of neurons within the BLA that respond reliably and selectively to appetitive stimuli, aversive stimuli, or both. At the population level, this diversity of individual neural responses allows for accurate classification of whether an appetitive or aversive stimulus was delivered, and these responses were highly stable across days, indicating stable encoding of valence within the BLA (Figure 1). Owing to large numbers of neurons recorded and high spatial resolution, we observed novel gradients in inhibited neuronal ensembles, with appetitive and aversive inhibitory ensembles displaying opposite patterns of expression in the dorsal/ventral axis (Figure 2A-F). Focusing on traditional valence encoding neurons (those that increase selectively to either a positive or negative valence stimulus with no change to the other) we observed gradients that were largely consistent with previously evidence^3,11 though a high degree of spatial overlap was observed throughout the BLA (Figure 2). We next established the spatial specificity of targeted photostimulation via SLM of individual BLA neurons deep in the brain through a GRIN lens for the first time, demonstrating high spatial and temporal control of targeted neurons that was robust and repeatable (Figure 3). With spatial specificity established, we targeted BLA ensembles that responded selectively to either appetitive or aversive stimuli for photostimulation, finding that when the quinine ensemble was stimulated, licking for an appetitive tastant was reduced (Figure 4). By contrast, when the sucrose ensemble received photostimulation, licking for the aversive tastant was enhanced (Figure 5). Finally, here we identified a potential mechanism of mutual inhibition between positive and negative valence ensembles, using holographic photostimulation to conduct microcircuit connectivity mapping to isolate functional antagonism of opposing BLA neuronal ensembles (Figure 6). These findings demonstrate that the neural representation of innate stimulus valence, reflected in distinct ensembles of BLA neurons with antagonistic connectivity profiles, is causally related to the expression of valence specific behavior.

Our findings bridge a long standing gap in understanding of how individual BLA neuron activity directly contributes to valence processing¹⁷. Numerous correlational studies have identified populations of neurons of varying sizes within the BLA that respond to appetitive and aversive stimuli as well as cues that predict them^3,6,7,11. Here we identify a substantial population of neurons that respond strongly and consistently to both appetitive and aversive stimuli, suggesting that the activity of these neurons may be important for the salience of a stimulus rather than valence^19,35. Due to this substantial overlap in the activity profiles of these neurons, but also their intermixed anatomical distribution, establishing whether neurons with selective activation in response to either an appetitive or an aversive stimulus simply represent the stimulus type or if they causally contribute to valence specific behavior (approach/consumption or avoidance) has not previously been possible. Here we show that these groups of BLA neurons not only have inherent and stable encoding of valence following lick training, but that their individual activation is sufficient to bias valence specific behavior. This conclusively demonstrates a causal role in controlling the expression of valence-specific behaviors by unique BLA ensembles, providing support for prior observations that BLA neurons may encode both appetitive and conditioned behavior⁶. Future experiments will evaluate how conditioned stimuli are represented by these defined groups of BLA neurons and whether targeted photostimulation can alter the representation of these previously neutral stimuli.

The linear relationship detected between the number of BLA neurons photostimulated and the change in licking behavior (Figure 4 & 5) provides new insight into how the BLA as a whole encodes valence. This suggests that the selection of valence-specific behaviors is graded and dependent on the recruitment of stable ensembles of BLA neurons that prefer one stimulus type, as well as potentially antagonistic ensembles with the opposite response. Future experiments will evaluate whether differing concentrations of appetitive and aversive stimuli shift ensemble size, and whether the number of photostimulated neurons necessary to produce a behavioral shift depends on the perceived delta of the appetitive and aversive stimulus. If the proposed graded model of encoding valence-specific behavior exists, we expect that the more similar the appetitive and aversive stimulus is the smaller the ensemble size will become.

Prior reports have been mixed with regards to whether there is a relationship between anatomical location and whether a neuron responds selectively to appetitive or aversive stimulus^3,11. Given our large dataset and spatial specificity due to the high-resolution imaging modality, we sought to bring clarity to any spatial distribution of valence encoding neurons. Consistent with patterns of Rspo2 expression, a marker for negative valence BLA neurons¹¹, we find that more quinine-activated neurons are located in the anterior BLA than sucrose activated neurons (Figure 2I). This contrasts with what was observed while recording BLA neurons during a conditioning paradigm³, where a bias toward greater response to the sucrose paired cue than to the quinine paired cue. We also observe a substantial bias toward more sucrose responsive neurons in the ventral BLA, which is consistent with what has been shown during conditioning³. While significant spatial gradients exist in our dataset, valence encoding cells were largely intermixed and overlapping as a function of location (Figure 2). Interestingly, when we focused our analysis exclusively on neurons with antagonistic relationships, no spatial gradients were observed (Figure S6E-G). Instead, we found that these antagonistic ensembles were located more proximally relative to non-antagonistic ensembles, reminiscent of lateral inhibition motifs that have been observed throughout the brain and could be mediated via parvalbumin (PV) expressing interneurons within the BLA, which receive strong input from neighboring BLA neurons^36–38. Future experiments should seek to identify the identity of these intermediate interneurons, and how they specifically regulate valence.

An important caveat of our study is that mice were water restricted, which was necessary to generate sufficient licking to the aversive tastant quinine in order to make inferences about the link between neural activity and behavior (Figure S2A-D)^34,39. Water restriction was selected over food restriction as the latter is less well tolerated by rodents and results in increases in circulating corticosterone levels^40,41. Constant water restriction, as conducted in this study, does not significantly elevate corticosterone levels in mice⁴². While care was taken to ensure water deprivation was consistent across animals, it is possible that varying water deprivation levels may have led to differing valence assessments of the two tastants. Given that homeostatic and hunger states are capable of altering valence processing and neural responses to food-associated stimuli^43,44, including in the BLA⁴⁵, it is likely that the valence encoding ensembles (while stable Figure S3) are influenced by this state. Future experiments teasing apart the stability and formation of these ensembles as a function of homeostatic state are necessary. A second caveat is the stimuli (sucrose and quinine) used in the present study rely on taste and consumption and relies on an innate response to our select stimuli. These stimuli were chosen for several reasons. First, they are robust, well established, and have been previously used to categorize valence-encoding neurons in the amygdala^3,11. Second, we felt it was critical to select stimuli that mice would experience similarly (e.g. the same sensory modality and delivery method) with clear and obvious behavioral readouts that would allow for direct comparison of the effects of multiphoton stimulation. Previous work has used stimuli of different sensory modalities for categorizing valence responsive neurons and projections^4,7,11,12 or has presented the aversive stimulus as a no-go cue in a learning context, limiting the number of trials mice experienced the aversive stimulus³. This prior evidence has identified many critical facets of valence encoding within the amygdala and demonstrates that the presence and role of these valence encoding neurons is preserved across a range of contexts and stimulus modalities (both for learned and innate valence). However, to obtain the cleanest and consistent neural responses to manipulate with multiphoton optogenetics, we have focused on only sucrose and quinine ensembles, which share the same administration route and sensory modality and produce qualitatively similar behavioral readouts (though at differing magnitudes). While our observations may be limited to two opposing valence taste-based stimuli, we do not believe these neurons are simply sucrose- and quinine-responsive neurons. Indeed, recent work suggests that individual amygdala neurons participate in multidimensional encoding that is sensitive to many types of stimuli and behavioral state^6,10. Future experiments conducting stimuli reversal, administering differing gradients of sucrose and quinine to observe how ensembles change, and comparing with other appetitive and aversive stimuli should be conducted to assess the generalizability of our observations.

Here we have extended the depth through which targeted photostimulation of individual neurons has been achieved, which has been previously used exclusively in superficial brain structures^{20,21,23,26–29}. By accounting for optical aberration generated by the GRIN lens (Figure S4), we find that specificity of photostimulation via the SLM is qualitatively similar to what has been shown in previous experiments using cortical windows^21,23. Although the numbers of accessible neurons are lower than what can be achieved via a cortical window, consistent behavioral effects were observed through targeted photostimulation of ensembles responsive to either appetitive or aversive stimuli across mice (Figure 4, 5 and 6). Given the significant positive correlation we observed between the number of neurons activated and the percent change in licking behavior, we expect that more profound changes in behavior will be detected by maximizing the numbers of neurons targeted in future experiments. This could be achieved numerous ways, including using a larger GRIN lens, different optical hardware, more sensitive red-shifted opsins, or by recording and stimulating multiple planes within a single mouse. Here we establish that targeted photostimulation in the deep brain is possible and use this method to directly demonstrate a causal role for mutually inhibitory opposing BLA ensembles on the expression valence-specific behavior. In doing so, we identify a model of BLA function in the control of innate valence that can be applied to other valence domains.

STAR Methods

RESOURCE AVAILABILITY

Lead contact

Further information and requests should be directed to the lead contact, Dr. Michael Bruchas (mbruchas@uw.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

All data reported in this paper will be shared by the lead contact upon reasonable request.
All original code has been deposited on Github and is publicly available as of the date of publication. DOIs are listed in the key resources table.
Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
AAV5-CAMKIIa-GCaMP6f-WPRE-SV40	Penn Vector Core	N/A
AAV8-CAMKIIa-ChRmine-mScarlet-Kv2.1-WPRE	Stanford Vector Core	GVVC-AAV-194

Chemicals, peptides, and recombinant proteins
VECTASHIELD Hardset Antifade Mounting Medium with DAPI	Vector Laboratories	CAT#H-1800
Sucrose, 99%	Fisher Scientific	AAA155830C
Quinine hydrocholoride dihydrate	Sigma-Aldrich	Q1125

Experimental models: Organisms/strains
C57BL/6J	Jackson Laboratory	000664

Software and algorithms
FIJI/ImageJ	NIH	https://fiji.sc/
Prairie View	Bruker	N/A
MATLAB	Mathworks	https://www.mathworks.com/products.html
Python 3.0	Python Software Foundation	https://www.python.org/
SIMA	⁴⁶	http://www.losonczylab.org/sima/1.3.2/
NAPE Calcium imaging pipeline	Github/Zenodo	https://doi.org/10.5281/zenodo.10076964
PRISM 8	GraphPad	https://www.graphpad.com/
Illustrator CS8	Adobe	https://www.adobe.com/products/illustrator.html

Other
ProView GRIN lens (0.6mm x 7.3mm)	Inscopix	1050-004608
Ultima 2P Plus w/ NeuraLight 3D Spatial Light Modulator	Bruker	https://www.bruker.com/en/products-and-solutions/fluorescence-microscopy/multiphoton-microscopes/ultima-2pplus.html
Insight X3 Tunable Ultrafast Laser	Spectra-Physics	https://www.spectra-physics.com/en/f/insight-x3-tunable-laser
Spirit High Power Femtosecond Lasers	Spectra-Physics	https://www.spectra-physics.com/en/f/spirit-femtosecond-laser

Open in a new tab

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animals

All procedures were carried out in accordance with the guidelines for the care and use of laboratory animals from the NIH and with approval from the University of Washington Institutional Animal Care and Use Committee (IACUC). Male and female C57BL/6 mice were used for all experiments. Mice in all cohorts were approximately 3–5 months old at the time of surgery. Mice were housed on a reverse light/dark cycle (lights on: 9:00PM, lights off 9:00AM).

METHOD DETAILS

Stereotactic surgery

For all surgeries, mice were anesthetized using 5% isoflurane mixed with oxygen and maintained at 1–2% isoflurane maintenance for the duration of the surgery. Mice were secured on a small-animal stereotactic instrument (Kopf Instruments, Tujunga, CA USA). Body temperature was maintained throughout surgery at 37°C using an infrared heating pad (Kent Scientific, Torrington, CT USA). Ophthalmic ointment was immediately applied to both eyes once mice were anaesthetized to prevent drying. Three rounds of betadine and alternating 70% ethanol wipes were applied to the scalp prior to an incision being made down the midline suture. A single skull screw was placed just in front of the lambdoid suture at a depth that did not penetrate the skull. After skull screw placement, a 1mm drill-bit was used to create a craniotomy above the BLA (coordinates A/P: −1.45, M/L: −3.22, D/V: −4.55). The D/V coordinate was measured relative to skull surface above the BLA, while A/P and M/L were measured from an interpolated bregma. 700nl of a 1:3 mixture of opsin (AAV8-CAMKIIa-ChRmine-mScarlet-Kv2.1-WPRE; Stanford Viral Vector Core, Titer 2.3×10¹³) and GCaMP6f (AAV5-CAMKIIa-GCaMP6f-WPRE-SV40, Penn Viral Vector Core, Titer 1.31×10¹³) was injected using a fixed-needle Hamilton Neuros syringe (Hamilton Company, Reno, NV USA) and an infusion pump (World Precision Instruments, Sarasota, FL USA) at a rate of 100 nl/min. Immediately after viral injection, a GRIN lens (Inscopix, ProView 7.3mm length, 0.6mm diameter or 9mm length, 1mm diameter) was lowered directly above the viral injection target at the same coordinates. The GRIN lens was lowered steadily until the D/V approached −3.55mm from skull surface, at which point it was lowered slowly at a rate of 100 μm/min until reaching final depth of −4.55mm. The skull was then dried and a layer of all purpose krazy-glue (Krazy Glue, Columbus, OH USA) was applied to the skull. After this application, a layer of black dental acrylic (Lang Dental, Wheeling, IL USA) was added to the skull and sides of the GRIN lens and built up into a secure base. A custom machined steel head-ring was then placed on top of the acrylic base such that the GRIN lens was in the center of the ring. The ring was then secured in place with another layer of dental acrylic. Animals were injected with slow-release Carprofen (5 mg/kg) prior to surgery to act as a post-surgery analgesic. Animals were monitored daily for 7 days following surgery.

Water restriction protocol

Mice recovered from surgery for a minimum of 4 weeks before beginning water restriction. Mice were water-restricted by giving them access to water only once per day in order to properly motivate consumption of sucrose, quinine, or water in behavioral assays. Starting 1 week prior to behavioral testing, the mice were given 1.0 mL of water in a 2×2 cm paper cup in their home cage each day at approximately 3–4pm. The mice promptly consumed the water while monitored by the researcher to assure they consumed most of the water. Animals were also weighed routinely to ensure they were at 85–90% of their body weight while water-restricted. No health issues related to dehydration occurred throughout the water-restriction period. After the behavioral testing concluded, water-restriction was discontinued and the mice were given ad-libitum access to water in their home cages again.

Head-fixed sucrose and quinine behavioral paradigm

After recovery from surgery, the mice were water restricted (see Water Restriction Protocol) and trained to lick 10% sucrose solution while immobilized and head-fixed. Head-fixation was accomplished by scruffing the mice, putting their back end into a 70 mL conical tube, and then placing the tube with the mouse into a custom headstage⁴⁷. All mice were habituated to the head-fixation and lick training with 10% sucrose for 4 days. After habituation, an Arduino controlled liquid delivery system was used to randomly deliver either 10% sucrose (70% of trials , pseudorandom) or 2mM quinine (30% of trials, pseudorandom) via distinct solenoids, tubes, and lick spouts. The licks spouts were positioned in a triangular formation such that liquid was accessed from the same point regardless of trial type. The intertrial interval was 20–25 seconds. Discrete lick events were detected and recorded using a contact lickometer circuit. TTLs reflecting each event (tastant delivery and licks) were sent directly to the data acquisition system on the two-photon microscope to ensure accurate tracking of both behavior and imaging data. The behavioral testing was conducted in the dark.

Sucrose-only and quinine-only head-fixed behavioral paradigm

After undergoing lick training and the sucrose and quinine behavioral paradigm listed above, a subset of mice underwent a sucrose-only paradigm for 1 day. The mice were head-fixed as described above. Testing was performed with a 20 minute session with 100% of trials delivering 10% sucrose. Delivery of sucrose and recording of licks was the same as described in the sucrose and quinine behavioral paradigm above.

Following the sucrose-only paradigm listed above, the same mice underwent a quinine-only paradigm for 1 day. The mice were head-fixed as described above. The test consisted of a 20 minute session with 100% of the trials delivering 2mM quinine. Delivery of quinine and recording of licks was the same as described in the sucrose and quinine behavioral paradigm above.

Sucrose, water, and quinine head-fixed behavioral paradigm

To compare licking of the appetitive tastant sucrose and aversive tastant quinine to regular drinking water, a 3 day paradigm was established. On the first day, the lick-trained mice first underwent a head-fixed sucrose and water test, consisting of a 30 minute session where they were presented with a droplet of either 10% sucrose or water. The percentage of sucrose and water trials were 70% and 30%, respectively, which were presented pseudorandomly. On the second day, the mice were offered a droplet of either 2mM quinine solution or water during the 30 minute session. The percentage of water and quinine trials were 70% and 30%, respectively, which were presented pseudorandomly. Conditions of delivery of the solutions and recordings of the licks were the same as described above.

Two-photon imaging

4 to 6 weeks after surgery mice were habituated to head-fixation for at least 4 consecutive days. Head-fixation was achieved by securing the circular head-ring into a metal clamp attached to a custom head-stage while the body of the mouse was lightly restrained in a 70ml conical tube. During habituation, mice were placed underneath the two-photon microscope for 5 minutes and given access to random presentations of 10% sucrose to familiarize mice with the procedure. Following habituation, combined two-photon imaging and behavior sessions were conducted. GCaMP6f imaging was acquired via an Ultima 2P Plus with the Neuralight 3D Spatial Light Modulator (SLM) (Bruker Fluorescence Microscopy, Middleton WI, USA) two-photon microscope using Prairie View Software (Bruker Fluorescence Microscopy, USA). Individual frames were acquired at 30Hz using a galvo-resonant scanner and a 4 times temporal average was performed on-line for an effective frame rate of 7.5Hz with a resolution of 512px x 512px. We used a long working distance 20x air objective designed for infrared wavelengths (Olympus, LCPLN20XIR, 0.45 numerical aperture (NA), 8.3mm working distance) combined with an Insight X3 Tunable Ultrafast Laser at 920nm (Spectra-Physics, Santa Clara, CA USA). The microscope was also equipped with an electrically tunable lens (ETL) in the resonant scan path. The ETL provides dynamic positioning of the imaging focal plane and is decoupled from the photostimulation laser path, enabling simultaneous imaging and photoactivation at different Z-planes. The same imaging FOV was identified by carefully measuring the distance from the top of the GRIN lens to the field of view from day to day.

Holographic photostimulation during behavior

For targeted photostimulation, the same microscope and acquisition system (Bruker) was used with a second laser path consisting of a 1040nm high power femtosecond pulsed laser (Spirit One, Spectra-Physics), spatial light modulator (512×512 pixel density) to generate multi-point stimulation montages, and a pair of galvanometer mirrors for generating spiral patterns of the montage²⁴. Neurons were selected for targeted photostimulation based on their selectivity for either sucrose or quinine (see Tastant related activity classification) from a prior session. Photostimulation sessions occurred 24–72 hours after initial sucrose and quinine imaging session, depending on the length of time necessary to process, segment, and analyze all data. During the photostimulation session, the correct FOV was identified and a 128 frame average image was generated in order to clearly highlight all neurons. Using the selectivity map (e.g. Figure 1F) as a reference, spiral stimulation targets (15–20μm diameter) were manually placed on top of GCaMP6f-expressing neurons selective for a particular tastant (sucrose or quinine). Laser power was adjusted based on the number of spiral stimulation targets (6–10mW of stimulation per target) for each individual animal and session.

A TTL was randomly delivered via Arduino to the Bruker system to elicit holographic photostimulation (20hz, 2s, 5ms pulse width) of select ensembles on 50% of tastant delivery trials (either sucrose or quinine delivery). Tastant delivery and stimulation occurred simultaneously for all experiments. With the exception of randomly delivered photostimulation, all other aspects of the behavioral paradigm remained the same. Due to the manual nature of spiral stimulation placement, experimenters were not blind while photostimulation sessions were conducted.

Measurment of lateral and axial resolution during holographic photostimulation

To measure the physiological resolution of 2-photon holographic photostimulation, two sets of experiments were performed for lateral and axial dimensions. For estimating lateral resolution, an isolated target cell was initially identified. A stimulation protocol with the same temporal and spatial parameters as our main experiments was set up, with the exception that the SLM mask targeted three ROIs and five stimulation trials were collected. One of these ROIs was positioned over the target cell and the other two were positioned outside of the GRIN lens FOV. The reason for this motif was to properly engage the SLM, as single ROI stimulation would not utilize the SLM. We then repeated the stimulation protocol several times with the SLM mask shifted laterally by the size of a cell (20μm) for each iteration. The data were then analyzed to examine response magnitude as a function of stimulation ROI distance from the target cell.

For measuring axial resolution, an isolated target cell was again identified. The same protocol as outlined for the lateral resolution experiments were performed up until shifting the ROI mask. Instead, in order to move the stimulation ROIs in the Z-dimension, the microscope Z position was offset by a predetermined distance. This step would also offset the imaging plane; however we aimed to record target cell activity in response to out of plane stimulation. Accordingly, we compensated for this imaging plane displacement by offsetting the microscope’s ETL in the opposite direction of the microscope Z movement. We then repeated this stimulation protocol several times with the microscope Z positioned at randomized distances and corresponding ETL offsets. The data were then analyzed to examine response magnitude as a function of stimulation ROI distance from the target cell.

Holographic photostimulation for connectivity mapping

The same microscope and photostimulation system as above was used for mapping of connectivity between ensembles. Using the selectivity map from previous sucrose and quinine delivery sessions (e.g. Figure 1F) as a reference, spiral stimulation targets (15–20μm diameter) were manually placed on top of GCaMP6f-expressing neurons selective for a particular tastant (sucrose or quinine). Laser power was adjusted based on the number of spiral stimulation targets (6–10mW of stimulation per target) for each individual animal and session.

A TTL was randomly delivered to the Bruker system via Arduino to elicit holographic photostimulation (20hz, 2s, 5ms pulse width) of select ensembles with a truncated exponential distribution with a mean of 30s and a maximum of 90s. No tastants (sucrose or quinine) were delivered during these trials. These trials occurred subsequent to Photostimulation during behavior. Sessions ended when an ensemble was targeted for photostimulation 15 times. Due to potential plastic changes, sucrose and quinine ensembles were targeted for photostimulation on separate days.

Histological registration of virus and GRIN lens targeting

Mice were transcardially perfused with 10% formalin and phosphate buffered saline (1X PBS). Immediately after perfusion, heads (with implants intact) were placed into 10% formalin for 24h for post-fixation after which brains were removed and transferred to a 30% sucrose (in 1X PBS) solution. Brains were frozen and 35um sections were cut on a cryostat in 1:6 series. 1 series was mounted, counterstained with DAPI mounting media (Fluoroshield, Sigma-Aldrich) and imaged at 20x resolution on epifluorescence (Leica Microsystems, Wetzlar, Germany) to visualize virus and implant sites.

GRIN lens placement in millimeters was determined by identifying the centerpoint of the bottom of the lens and identifying this on a stereotactic atlas. For precise localization in the medial/lateral and anterior/posterior axis, lens location was identified on the Scalable Brain Atlas (https://scalablebrainatlas.incf.org/mouse/ABA_v3). For dorsal/ventral lens location, the bottom of the lens was identified on The Mouse Brain in Stereotaxic coordinates⁴⁸. For histological and subsequent analyses, the bottom of the lens was used as the imaging plane as it is difficult to assess the precise imaging field of view in tissue. All coordinates (in mm) are relative to bregma. If GRIN lens implants were determined to be outside the BLA (Figure S1C), mice were excluded from the study.

Two-photon imaging analysis and data processing pipeline

Individual .tif files from a single recording session were collected at 7.5Hz and combined into HDF5 format using custom code (https://github.com/zhounapeuw/NAPE_imaging_analysis). HDF5 files were then imported into the Sequential IMaging Analysis (SIMA) Python package⁴⁶ as previously described for motion correction in the x-y plane⁴⁷. Following motion correction, the motion corrected HDF5 file as well as the mean projection image across all motion corrected frames were imported into Fiji is just ImageJ (Fiji, version 1.52)⁴⁹ for selecting neuronal regions of interest (ROIs). ROIs were manually drawn by a trained observer using the polygon selection tool. Care was taken to capture only somatic ROIs and to minimize any potential overlap between multiple soma. ROIs were then imported into SIMA for signal extraction. These aforementioned preprocessing steps were facilitated by the NAPECA analysis pipeline that wraps around the SIMA package (https://github.com/zhounapeuw/NAPE_imaging_analysis). Across sessions the same ROI file was used with minor adjustments depending on slight translation or rotation of the field of view (FOV) across sessions. Tracking across sessions was highly reliable (Figs. 4,5,6). For photostimulation sessions, frames in which stimulation was occurring and the imaging shutter was closed were removed from analysis.

Continuous SIMA extracted fluorescence signals for each neuron were then normalized to the mean and standard deviation from the entire session (Z-score). This normalized fluorescence was used for all down-stream analyses.

QUANTIFICATION AND STATISTICAL ANALYSIS

Tastant related activity classification

In order to perform unbiased classification of an individual neuron’s responsiveness (activated, inhibited, or unaffected) to a tastant delivery (e.g. sucrose or quinine delivery) we adapted a previously used strategy^50,51. For each individual cell, raw calcium traces 3 seconds prior to tastant delivery onset and 10 seconds after tastant delivery (188 total samples at 7.5 Hz, 133ms per frame) were shuffled in time for each sample (200x) removing any temporal information that was previously in each trace but maintaining the variance within each trial. This shuffle was then performed 1000 times per cell to obtain a null distribution of tastant delivery associated calcium activity. A cell was considered responsive to a particular tastant onset if its average behavioral event Z-normalized calcium fluorescence amplitude between 0s before delivery to 5s after delivery exceeded a 1 standard deviation threshold from the null distribution.

To categories neurons as sucrose-only, quinine-only, or sucrose and quinine responsive, logical indexing was used. Thus, a sucrose-only activated neuron had significantly increased activity in response to sucrose without a significant increase to quinine. Likewise, a quinine-only activated neuron had significant increase in activity in response to quinine without a significant increase to sucrose. Sucrose and quinine activated neurons had significant increases in activity in response to both sucrose and quinine delivery. For inhibited neurons, the same indexing rule applied.

Population decoding of tastant type

To investigate whether delivery of sucrose or quinine could be uniquely identified only by the simultaneous activity of BLA neurons, a linear support vector classifier (SVC) model (cost = 0.8) was created. The linear SVC model was trained and tested on the data for each animal individually. The class of the tastant condition and the magnitude of the reward response (0–5 s from onset of tastant delivery) of every neuron for a given trial was used as input for the model. 75% of the trials were used to train the linear SVC and 25% were used for cross-validation. The accuracy of the model during cross-validation for each animal is reported in the Results and Figure 1. The linear SVC model was created in Python using the Scikit-learn library.

Anatomical distribution of tastant-selective neurons

For plotting 2- and 3-dimensional representations of putative neuronal contours, lens location coordinate values were combined with individual cell masks within a given FOV. The center point of each contour was then identified for each neuron. These center point values were then plotted as scatterplots (scatter for 2D or scatter3 for 3D in MATLAB) for each putative neuron for all mice. Two-dimensional representations (e.g. medial/lateral and anterior/posterior) were used to more easily visualize neuron identity as a function of spatial location. To compare the probability distribution functions of cell identity/classification (e.g. sucrose activated vs. sucrose inhibited, or sucrose activated vs. quinine activated) as a function of spatial location, we used the two-sample Kolmogorov-Smirnov test.

Analysis of ensemble proximity

To assess the proximity of valence-encoding ensembles, we calculated Euclidean distance measures between mean ensemble locations within a single FOV across mice. For each neuronal contour, the center point was determined. We then calculated the mean of all center points in a given ensemble (e.g. sucrose activated/quinine inhibited). Using the ensemble mean, we calculated the Euclidean norm/distance distance in 2-dimensional space (medial/lateral and anterior/posterior) using the MATLAB function norm(). Euclidean distances between opposing and non-opposing ensembles were then compared with a Paired two-sample t-test.

Statistical analysis of fluorescence data

Analysis of continuous fluorescence traces comparing activity of neurons in response to sucrose or quinine were carried out using two-way repeated measures ANOVA. When a significant tastant X time interaction was detected, post-hoc comparisons of individual time bins were conducted using Sidak’s multiple comparison correction. Unless otherwise noted in the main text, mean post-delivery fluorescence was calculated for each individual neuron (trial-averaged fluorescence) using a window from 0 to 5s post-tastant delivery. To compare mean fluorescence of individual neurons, again a two-way repeated measures ANOVA was used with Sidak’s multiple comparison correction if a significant interaction was observed.

To probe for stability of valence encoding, Pearson’s correlations were used. For assessing the stability of fluorescence amplitude, a difference score (Sucrose – quinine) was conducted using the trial-averaged fluorescence responses for each individual neuron. Thus, for each session a single value (Δ Fluorescence) was computed for each neuron for each session. A positive value indicates a sucrose-selective neuron, while a negative value indicates a quinine-selective neuron. These difference scores were then correlated between the two recording sessions. To assess the consistency of the temporal dynamics of the calcium signal across sessions in response to sucrose or quinine, the area under the curve (AUC) of the trial-averaged fluorescence was obtained for each individual neuron using the trapz() function in MATLAB. These AUC values were then correlated across sessions. GraphPad Prism 8.0 (GraphPad, La Jolla, CA USA) was used for all statistics and correlations. Unless otherwise noted, alpha was set to 0.05 and plotted data represent mean ±standard error of measurement (SEM).

Statistical analysis of licking behavior

Mean number of licks on sucrose or quinine trials (and on photostimulation trials versus no stimulation trials) was analyzed using a paired two-tailed independent samples t-tests. For comparing the probability of licking in response to sucrose or quinine delivery two-way repeated measures analysis of variance (ANOVA) was used. Main effects and interactions are reported and in the case of significant interactions, post-hoc comparisons of significant time bins were made using Sidak’s multiple comparison correction. Two-way repeated measures ANOVAs were also used for analyzing the effect of targeted photostimulation on licking probability in response to either sucrose or quinine. GraphPad Prism 8.0 (GraphPad, La Jolla, CA USA) was used for all statistics and correlations. Unless otherwise noted, alpha was set to 0.05 and plotted data represent mean ±standard error of measurement (SEM).

Supplementary Material

NIHMS1946419-supplement-2.pdf^{(8.5MB, pdf)}

Download video file^{(31MB, mp4)}

HIGHLIGHTS.

Valence-specific ensembles within the BLA are stable and optically separable.
Holographic stimulation of BLA neurons through a GRIN lens is specific and robust.
Multiphoton stimulation of valence-specific ensembles bidirectionally alters behavior.
Opposing ensembles exhibit mutual inhibitory connectivity in vivo.

Acknowledgements:

We thank Dr. Azra Suko for lab management and organization and Drs. David Ottenheimer and Adam Gordon-Fennell as well as Madelyn Hjort (all members of the Stuber lab) for assistance and input related to head-fixed behavior. We also thank the entire Bruchas laboratory as well as the Center for Neuroscience of Addiction, Pain, and Emotion at the University of Washington for resources, alongside critical feedback.

Funding:

This work was supported by the National Institute of Mental Health (MRB, R01MH112355, R01MH111520), the National Institute of Drug Abuse (GDS – R37DA032750–10, CEP - F31DA051124), and the National Institute of General Medical Sciences (SCP – T32GM086270–12, PI Dr. Tonya M. Palermo). We also acknowledge the generous support from the Murdock Foundation for resources to fund the SLM, and the NAPE Center NIDA funded P30DA048736. MRB and SCP also received support from the Weill Neurohub.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Declaration of Interests: The authors have no competing interests to declare.

Supplemental Video 1. Stimulation of quinine ensemble during task, related to Figure 4.

In vivo two-photon calcium imaging of BLA neurons during a session in which quinine ensemble neurons were targeted for holographic photostimulation. Video depicts selected quinine ensemble neurons (red) targeted for photostimulation (10hz, 2s duration, 5ms pulse width) during the presentation of either sucrose or quinine. Video acquired at 7.5hz and played back at 13x speed. Initial targeted photostimulation is slowed down to 3x speed.

References

1.Morrison SE, and Salzman CD (2011). Representations of appetitive and aversive information in the primate orbitofrontal cortex. Ann N Y Acad Sci 1239, 59–70. 10.1111/j.1749-6632.2011.06255.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Paton JJ, Belova MA, Morrison SE, and Salzman CD (2006). The primate amygdala represents the positive and negative value of visual stimuli during learning. Nature 439, 865–870. 10.1038/nature04490. [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Beyeler A, Chang C-J, Silvestre M, Lévêque C, Namburi P, Wildes CP, and Tye KM (2018). Organization of Valence-Encoding and Projection-Defined Neurons in the Basolateral Amygdala. Cell Reports 22, 905–918. 10.1016/j.celrep.2017.12.097. [DOI] [PMC free article] [PubMed] [Google Scholar]
4.Gore F, Schwartz EC, Brangers BC, Aladi S, Stujenske JM, Likhtik E, Russo MJ, Gordon JA, Salzman CD, and Axel R (2015). Neural representations of unconditioned stimuli in basolateral amygdala mediate innate and learned responses. Cell 162, 134–145. 10.1016/j.cell.2015.06.027. [DOI] [PMC free article] [PubMed] [Google Scholar]
5.Grewe BF, Grundemann J, Kitch LJ, Lecoq JA, Parker JG, Marshall JD, Larkin MC, Jercog PE, Grenier F, Li JZ, et al. (2017). Neural ensemble dynamics underlying a long-term associative memory. Nature 10.1038/nature21682. [DOI] [PMC free article] [PubMed]
6.Kyriazi P, Headley DB, and Pare D (2018). Multi-dimensional Coding by Basolateral Amygdala Neurons. Neuron 99, 1315–1328.e5. 10.1016/j.neuron.2018.07.036. [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Zhang X, and Li B (2018). Population coding of valence in the basolateral amygdala. Nat Commun 9, 5195. 10.1038/s41467-018-07679-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Borst A, and Theunissen FE (1999). Information theory and neural coding. Nat Neurosci 2, 947–957. 10.1038/14731. [DOI] [PubMed] [Google Scholar]
9.Kumar A, Rotter S, and Aertsen A (2010). Spiking activity propagation in neuronal networks: reconciling different perspectives on neural coding. Nature reviews. Neuroscience 11, 615–627. 10.1038/nrn2886. [DOI] [PubMed] [Google Scholar]
10.Gründemann J, Bitterman Y, Lu T, Krabbe S, Grewe BF, Schnitzer MJ, and Lüthi A (2019). Amygdala ensembles encode behavioral states. Science 364, eaav8736. 10.1126/science.aav8736. [DOI] [PubMed] [Google Scholar]
11.Kim J, Pignatelli M, Xu S, Itohara S, and Tonegawa S (2016). Antagonistic negative and positive neurons of the basolateral amygdala. Nat Neurosci 19, 1636–1646. 10.1038/nn.4414. [DOI] [PMC free article] [PubMed] [Google Scholar]
12.Namburi P, Beyeler A, Yorozu S, Calhoon GG, Halbert SA, Wichmann R, Holden SS, Mertens KL, Anahtar M, Felix-Ortiz AC, et al. (2015). A circuit mechanism for differentiating positive and negative associations. Nature 520, 675–678. 10.1038/nature14366. [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Beyeler A, Namburi P, Glober GF, Simonnet C, Calhoon GG, Conyers GF, Luck R, Wildes CP, and Tye KM (2016). Divergent Routing of Positive and Negative Information from the Amygdala during Memory Retrieval. Neuron 90, 348–361. 10.1016/j.neuron.2016.03.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Stuber GD, Sparta DR, Stamatakis AM, van Leeuwen WA, Hardjoprajitno JE, Cho S, Tye KM, Kempadoo KA, Zhang F, Deisseroth K, et al. (2011). Excitatory transmission from the amygdala to nucleus accumbens facilitates reward seeking. Nature 475, 377–380. 10.1038/nature10194. [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Britt JP, Benaliouad F, McDevitt RA, Stuber GD, Wise RA, and Bonci A (2012). Synaptic and Behavioral Profile of Multiple Glutamatergic Inputs to the Nucleus Accumbens. Neuron 76, 790–803. 10.1016/j.neuron.2012.09.040. [DOI] [PMC free article] [PubMed] [Google Scholar]
16.Zhang X, Guan W, Yang T, Furlan A, Xiao X, Yu K, An X, Galbavy W, Ramakrishnan C, Deisseroth K, et al. (2021). Genetically identified amygdala–striatal circuits for valence-specific behaviors. Nat Neurosci 24, 1586–1600. 10.1038/s41593-021-00927-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Tye KM (2018). Neural Circuit Motifs in Valence Processing. Neuron 100, 436–452. 10.1016/j.neuron.2018.10.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
18.Pignatelli M, and Beyeler A (2019). Valence coding in amygdala circuits. Current Opinion in Behavioral Sciences 26, 97–106. 10.1016/j.cobeha.2018.10.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
19.Shabel SJ, and Janak PH (2009). Substantial similarity in amygdala neuronal activity during conditioned appetitive and aversive emotional arousal. PNAS 106, 15031–15036. 10.1073/pnas.0905580106. [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Jennings JH, Kim CK, Marshel JH, Raffiee M, Ye L, Quirin S, Pak S, Ramakrishnan C, and Deisseroth K (2019). Interacting neural ensembles in orbitofrontal cortex for social and feeding behaviour. Nature 565, 645–649. 10.1038/s41586-018-0866-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
21.Daie K, Svoboda K, and Druckmann S (2021). Targeted photostimulation uncovers circuit motifs supporting short-term memory. Nat Neurosci 24, 259–265. 10.1038/s41593-020-00776-3. [DOI] [PubMed] [Google Scholar]
22.Carrillo-Reid L, Yang W, Kang Miller J-E, Peterka DS, and Yuste R (2017). Imaging and Optically Manipulating Neuronal Ensembles. Annu Rev Biophys 46, 271–293. 10.1146/annurev-biophys-070816-033647. [DOI] [PubMed] [Google Scholar]
23.Carrillo-Reid L, Han S, Yang W, Akrouh A, and Yuste R (2019). Controlling Visually Guided Behavior by Holographic Recalling of Cortical Ensembles. Cell 178, 447–457.e5. 10.1016/j.cell.2019.05.045. [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Yang W, Carrillo-Reid L, Bando Y, Peterka DS, and Yuste R (2018). Simultaneous two-photon imaging and two-photon optogenetics of cortical circuits in three dimensions. eLife 7, e32671. 10.7554/eLife.32671. [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Sridharan S, Gajowa MA, Ogando MB, Jagadisan UK, Abdeladim L, Sadahiro M, Bounds HA, Hendricks WD, Turney TS, Tayler I, et al. (2022). High-performance microbial opsins for spatially and temporally precise perturbations of large neuronal networks. Neuron 110, 1139–1155.e6. 10.1016/j.neuron.2022.01.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Adesnik H, and Abdeladim L (2021). Probing neural codes with two-photon holographic optogenetics. Nat Neurosci 24, 1356–1366. 10.1038/s41593-021-00902-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
27.Gill JV, Lerman GM, Zhao H, Stetler BJ, Rinberg D, and Shoham S (2020). Precise Holographic Manipulation of Olfactory Circuits Reveals Coding Features Determining Perceptual Detection. Neuron 108, 382–393.e5. 10.1016/j.neuron.2020.07.034. [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Robinson NTM, Descamps LAL, Russell LE, Buchholz MO, Bicknell BA, Antonov GK, Lau JYN, Nutbrown R, Schmidt-Hieber C, and Häusser M (2020). Targeted Activation of Hippocampal Place Cells Drives Memory-Guided Spatial Behavior. Cell 183, 1586–1599.e10. 10.1016/j.cell.2020.09.061. [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Marshel JH, Kim YS, Machado TA, Quirin S, Benson B, Kadmon J, Raja C, Chibukhchyan A, Ramakrishnan C, Inoue M, et al. (2019). Cortical layer– specific critical dynamics triggering perception. Science 365, eaaw5202. 10.1126/science.aaw5202. [DOI] [PMC free article] [PubMed] [Google Scholar]
30.Russell LE, Dalgleish HWP, Nutbrown R, Gauld OM, Herrmann D, Fişek M, Packer AM, and Häusser M (2022). All-optical interrogation of neural circuits in behaving mice. Nat Protoc, 1–42. 10.1038/s41596-022-00691-w. [DOI] [PMC free article] [PubMed]
31.Pérez-Ortega J, Alejandre-García T, and Yuste R (2021). Long-term stability of cortical ensembles. eLife 10, e64449. 10.7554/eLife.64449. [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Ziv Y, Burns LD, Cocker ED, Hamel EO, Ghosh KK, Kitch LJ, El Gamal A, and Schnitzer MJ (2013). Long-term dynamics of CA1 hippocampal place codes. Nat Neurosci 16, 264–266. 10.1038/nn.3329. [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Rule ME, O’Leary T, and Harvey CD (2019). Causes and consequences of representational drift. Current Opinion in Neurobiology 58, 141–147. 10.1016/j.conb.2019.08.005. [DOI] [PMC free article] [PubMed] [Google Scholar]
34.K Namboodiri VM, Hobbs T, Trujillo-Pisanty I, Simon RC, Gray MM, and Stuber GD (2021). Relative salience signaling within a thalamo-orbitofrontal circuit governs learning rate. Current Biology 31, 5176–5191.e5. 10.1016/j.cub.2021.09.037. [DOI] [PMC free article] [PubMed] [Google Scholar]
35.Belova MA, Paton JJ, Morrison SE, and Salzman CD (2007). Expectation Modulates Neural Responses to Pleasant and Aversive Stimuli in Primate Amygdala. Neuron 55, 970–984. 10.1016/j.neuron.2007.08.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
36.Smith Y, Paré J-F, and Paré D (2000). Differential innervation of parvalbumin-immunoreactive interneurons of the basolateral amygdaloid complex by cortical and intrinsic inputs. Journal of Comparative Neurology 416, 496–508. . [DOI] [PubMed] [Google Scholar]
37.Krabbe S, Gründemann J, and Lüthi A (2018). Amygdala Inhibitory Circuits Regulate Associative Fear Conditioning. Biological Psychiatry 83, 800–809. 10.1016/j.biopsych.2017.10.006. [DOI] [PubMed] [Google Scholar]
38.Fu X, Teboul E, Weiss GL, Antonoudiou P, Borkar CD, Fadok JP, Maguire J, and Tasker JG (2022). Gq neuromodulation of BLA parvalbumin interneurons induces burst firing and mediates fear-associated network and behavioral state transition in mice. Nat Commun 13, 1290. 10.1038/s41467-022-28928-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Gordon-Fennell A, Barbakh JM, Utley MT, Singh S, Bazzino P, Gowrishankar R, Bruchas MR, Roitman MF, and Stuber GD (2023). An open-source platform for head-fixed operant and consummatory behavior. eLife 12, e86183. 10.7554/eLife.86183. [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Tucci V, Hardy A, and Nolan PM (2006). A comparison of physiological and behavioural parameters in C57BL/6J mice undergoing food or water restriction regimes. Behavioural Brain Research 173, 22–29. 10.1016/j.bbr.2006.05.031. [DOI] [PubMed] [Google Scholar]
41.Heiderstadt KM, McLaughlin RM, Wright DC, Walker SE, and Gomez-Sanchez CE (2000). The effect of chronic food and water restriction on open-field behaviour and serum corticosterone levels in rats. Lab Anim 34, 20–28. 10.1258/002367700780578028. [DOI] [PubMed] [Google Scholar]
42.Vasilev D, Havel D, Liebscher S, Slesiona-Kuenzel S, Logothetis NK, Schenke-Layland K, and Totah NK (2021). Three Water Restriction Schedules Used in Rodent Behavioral Tasks Transiently Impair Growth and Differentially Evoke a Stress Hormone Response without Causing Dehydration. eNeuro 8. 10.1523/ENEURO.0424-21.2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Burgess CR, Ramesh RN, Sugden AU, Levandowski KM, Minnig MA, Fenselau H, Lowell BB, and Andermann ML (2016). Hunger-Dependent Enhancement of Food Cue Responses in Mouse Postrhinal Cortex and Lateral Amygdala. Neuron 91, 1154–1169. 10.1016/j.neuron.2016.07.032. [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Livneh Y, Ramesh RN, Burgess CR, Levandowski KM, Madara JC, Fenselau H, Goldey GJ, Diaz VE, Jikomes N, Resch JM, et al. (2017). Homeostatic circuits selectively gate food cue responses in insular cortex. Nature 546, 611–616. 10.1038/nature22375. [DOI] [PMC free article] [PubMed] [Google Scholar]
45.Calhoon GG, Sutton AK, Chang C-J, Libster AM, Glober GF, Lévêque CL, Murphy GD, Namburi P, Leppla CA, Siciliano CA, et al. (2018). Acute Food Deprivation Rapidly Modifies Valence-Coding Microcircuits in the Amygdala 285189. 10.1101/285189. [DOI]
46.Kaifosh P, Zaremba JD, Danielson NB, and Losonczy A (2014). SIMA: Python software for analysis of dynamic fluorescence imaging data. Front Neuroinform 8, 80. 10.3389/fninf.2014.00080. [DOI] [PMC free article] [PubMed] [Google Scholar]
47.Namboodiri VMK, Otis JM, van Heeswijk K, Voets ES, Alghorazi RA, Rodriguez-Romaguera J, Mihalas S, and Stuber GD (2019). Single-cell activity tracking reveals that orbitofrontal neurons acquire and maintain a long-term memory to guide behavioral adaptation. Nature Neuroscience 22, 1110–1121. 10.1038/s41593-019-0408-1. [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Franklin KBJ, and Paxinos G (2013). Paxinos and Franklin’s The mouse brain in stereotaxic coordinates Fourth edition. (Academic Press, an imprint of Elsevier; ). [Google Scholar]
49.Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, Preibisch S, Rueden C, Saalfeld S, Schmid B, et al. (2012). Fiji: an open-source platform for biological-image analysis. Nat Methods 9, 676–682. 10.1038/nmeth.2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
50.Jimenez JC, Su K, Goldberg AR, Luna VM, Biane JS, Ordek G, Zhou P, Ong SK, Wright MA, Zweifel L, et al. (2018). Anxiety Cells in a Hippocampal-Hypothalamic Circuit. Neuron 97, 670–683 e6. 10.1016/j.neuron.2018.01.016. [DOI] [PMC free article] [PubMed] [Google Scholar]
51.Piantadosi SC, Manning EE, Chamberlain BL, Hyde J, LaPalombara Z, Bannon NM, Pierson JL, Namboodiri VM, and Ahmari SE (2022). Hyperactivity of indirect pathway-projecting spiny projection neurons drives compulsive behavior 2022.02.17.480966. 10.1101/2022.02.17.480966. [DOI] [PMC free article] [PubMed]

Associated Data

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

Supplementary Materials

NIHMS1946419-supplement-2.pdf^{(8.5MB, pdf)}

Download video file^{(31MB, mp4)}

Data Availability Statement

All data reported in this paper will be shared by the lead contact upon reasonable request.
All original code has been deposited on Github and is publicly available as of the date of publication. DOIs are listed in the key resources table.
Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
AAV5-CAMKIIa-GCaMP6f-WPRE-SV40	Penn Vector Core	N/A
AAV8-CAMKIIa-ChRmine-mScarlet-Kv2.1-WPRE	Stanford Vector Core	GVVC-AAV-194

Chemicals, peptides, and recombinant proteins
VECTASHIELD Hardset Antifade Mounting Medium with DAPI	Vector Laboratories	CAT#H-1800
Sucrose, 99%	Fisher Scientific	AAA155830C
Quinine hydrocholoride dihydrate	Sigma-Aldrich	Q1125

Experimental models: Organisms/strains
C57BL/6J	Jackson Laboratory	000664

Software and algorithms
FIJI/ImageJ	NIH	https://fiji.sc/
Prairie View	Bruker	N/A
MATLAB	Mathworks	https://www.mathworks.com/products.html
Python 3.0	Python Software Foundation	https://www.python.org/
SIMA	⁴⁶	http://www.losonczylab.org/sima/1.3.2/
NAPE Calcium imaging pipeline	Github/Zenodo	https://doi.org/10.5281/zenodo.10076964
PRISM 8	GraphPad	https://www.graphpad.com/
Illustrator CS8	Adobe	https://www.adobe.com/products/illustrator.html

Other
ProView GRIN lens (0.6mm x 7.3mm)	Inscopix	1050-004608
Ultima 2P Plus w/ NeuraLight 3D Spatial Light Modulator	Bruker	https://www.bruker.com/en/products-and-solutions/fluorescence-microscopy/multiphoton-microscopes/ultima-2pplus.html
Insight X3 Tunable Ultrafast Laser	Spectra-Physics	https://www.spectra-physics.com/en/f/insight-x3-tunable-laser
Spirit High Power Femtosecond Lasers	Spectra-Physics	https://www.spectra-physics.com/en/f/spirit-femtosecond-laser

Open in a new tab

[R1] 1.Morrison SE, and Salzman CD (2011). Representations of appetitive and aversive information in the primate orbitofrontal cortex. Ann N Y Acad Sci 1239, 59–70. 10.1111/j.1749-6632.2011.06255.x. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R2] 2.Paton JJ, Belova MA, Morrison SE, and Salzman CD (2006). The primate amygdala represents the positive and negative value of visual stimuli during learning. Nature 439, 865–870. 10.1038/nature04490. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R3] 3.Beyeler A, Chang C-J, Silvestre M, Lévêque C, Namburi P, Wildes CP, and Tye KM (2018). Organization of Valence-Encoding and Projection-Defined Neurons in the Basolateral Amygdala. Cell Reports 22, 905–918. 10.1016/j.celrep.2017.12.097. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R4] 4.Gore F, Schwartz EC, Brangers BC, Aladi S, Stujenske JM, Likhtik E, Russo MJ, Gordon JA, Salzman CD, and Axel R (2015). Neural representations of unconditioned stimuli in basolateral amygdala mediate innate and learned responses. Cell 162, 134–145. 10.1016/j.cell.2015.06.027. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R5] 5.Grewe BF, Grundemann J, Kitch LJ, Lecoq JA, Parker JG, Marshall JD, Larkin MC, Jercog PE, Grenier F, Li JZ, et al. (2017). Neural ensemble dynamics underlying a long-term associative memory. Nature 10.1038/nature21682. [DOI] [PMC free article] [PubMed]

[R6] 6.Kyriazi P, Headley DB, and Pare D (2018). Multi-dimensional Coding by Basolateral Amygdala Neurons. Neuron 99, 1315–1328.e5. 10.1016/j.neuron.2018.07.036. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R7] 7.Zhang X, and Li B (2018). Population coding of valence in the basolateral amygdala. Nat Commun 9, 5195. 10.1038/s41467-018-07679-9. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R8] 8.Borst A, and Theunissen FE (1999). Information theory and neural coding. Nat Neurosci 2, 947–957. 10.1038/14731. [DOI] [PubMed] [Google Scholar]

[R9] 9.Kumar A, Rotter S, and Aertsen A (2010). Spiking activity propagation in neuronal networks: reconciling different perspectives on neural coding. Nature reviews. Neuroscience 11, 615–627. 10.1038/nrn2886. [DOI] [PubMed] [Google Scholar]

[R10] 10.Gründemann J, Bitterman Y, Lu T, Krabbe S, Grewe BF, Schnitzer MJ, and Lüthi A (2019). Amygdala ensembles encode behavioral states. Science 364, eaav8736. 10.1126/science.aav8736. [DOI] [PubMed] [Google Scholar]

[R11] 11.Kim J, Pignatelli M, Xu S, Itohara S, and Tonegawa S (2016). Antagonistic negative and positive neurons of the basolateral amygdala. Nat Neurosci 19, 1636–1646. 10.1038/nn.4414. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R12] 12.Namburi P, Beyeler A, Yorozu S, Calhoon GG, Halbert SA, Wichmann R, Holden SS, Mertens KL, Anahtar M, Felix-Ortiz AC, et al. (2015). A circuit mechanism for differentiating positive and negative associations. Nature 520, 675–678. 10.1038/nature14366. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R13] 13.Beyeler A, Namburi P, Glober GF, Simonnet C, Calhoon GG, Conyers GF, Luck R, Wildes CP, and Tye KM (2016). Divergent Routing of Positive and Negative Information from the Amygdala during Memory Retrieval. Neuron 90, 348–361. 10.1016/j.neuron.2016.03.004. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R14] 14.Stuber GD, Sparta DR, Stamatakis AM, van Leeuwen WA, Hardjoprajitno JE, Cho S, Tye KM, Kempadoo KA, Zhang F, Deisseroth K, et al. (2011). Excitatory transmission from the amygdala to nucleus accumbens facilitates reward seeking. Nature 475, 377–380. 10.1038/nature10194. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R15] 15.Britt JP, Benaliouad F, McDevitt RA, Stuber GD, Wise RA, and Bonci A (2012). Synaptic and Behavioral Profile of Multiple Glutamatergic Inputs to the Nucleus Accumbens. Neuron 76, 790–803. 10.1016/j.neuron.2012.09.040. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R16] 16.Zhang X, Guan W, Yang T, Furlan A, Xiao X, Yu K, An X, Galbavy W, Ramakrishnan C, Deisseroth K, et al. (2021). Genetically identified amygdala–striatal circuits for valence-specific behaviors. Nat Neurosci 24, 1586–1600. 10.1038/s41593-021-00927-0. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R17] 17.Tye KM (2018). Neural Circuit Motifs in Valence Processing. Neuron 100, 436–452. 10.1016/j.neuron.2018.10.001. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R18] 18.Pignatelli M, and Beyeler A (2019). Valence coding in amygdala circuits. Current Opinion in Behavioral Sciences 26, 97–106. 10.1016/j.cobeha.2018.10.010. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R19] 19.Shabel SJ, and Janak PH (2009). Substantial similarity in amygdala neuronal activity during conditioned appetitive and aversive emotional arousal. PNAS 106, 15031–15036. 10.1073/pnas.0905580106. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R20] 20.Jennings JH, Kim CK, Marshel JH, Raffiee M, Ye L, Quirin S, Pak S, Ramakrishnan C, and Deisseroth K (2019). Interacting neural ensembles in orbitofrontal cortex for social and feeding behaviour. Nature 565, 645–649. 10.1038/s41586-018-0866-8. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R21] 21.Daie K, Svoboda K, and Druckmann S (2021). Targeted photostimulation uncovers circuit motifs supporting short-term memory. Nat Neurosci 24, 259–265. 10.1038/s41593-020-00776-3. [DOI] [PubMed] [Google Scholar]

[R22] 22.Carrillo-Reid L, Yang W, Kang Miller J-E, Peterka DS, and Yuste R (2017). Imaging and Optically Manipulating Neuronal Ensembles. Annu Rev Biophys 46, 271–293. 10.1146/annurev-biophys-070816-033647. [DOI] [PubMed] [Google Scholar]

[R23] 23.Carrillo-Reid L, Han S, Yang W, Akrouh A, and Yuste R (2019). Controlling Visually Guided Behavior by Holographic Recalling of Cortical Ensembles. Cell 178, 447–457.e5. 10.1016/j.cell.2019.05.045. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R24] 24.Yang W, Carrillo-Reid L, Bando Y, Peterka DS, and Yuste R (2018). Simultaneous two-photon imaging and two-photon optogenetics of cortical circuits in three dimensions. eLife 7, e32671. 10.7554/eLife.32671. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R25] 25.Sridharan S, Gajowa MA, Ogando MB, Jagadisan UK, Abdeladim L, Sadahiro M, Bounds HA, Hendricks WD, Turney TS, Tayler I, et al. (2022). High-performance microbial opsins for spatially and temporally precise perturbations of large neuronal networks. Neuron 110, 1139–1155.e6. 10.1016/j.neuron.2022.01.008. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R26] 26.Adesnik H, and Abdeladim L (2021). Probing neural codes with two-photon holographic optogenetics. Nat Neurosci 24, 1356–1366. 10.1038/s41593-021-00902-9. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R27] 27.Gill JV, Lerman GM, Zhao H, Stetler BJ, Rinberg D, and Shoham S (2020). Precise Holographic Manipulation of Olfactory Circuits Reveals Coding Features Determining Perceptual Detection. Neuron 108, 382–393.e5. 10.1016/j.neuron.2020.07.034. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R28] 28.Robinson NTM, Descamps LAL, Russell LE, Buchholz MO, Bicknell BA, Antonov GK, Lau JYN, Nutbrown R, Schmidt-Hieber C, and Häusser M (2020). Targeted Activation of Hippocampal Place Cells Drives Memory-Guided Spatial Behavior. Cell 183, 1586–1599.e10. 10.1016/j.cell.2020.09.061. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R29] 29.Marshel JH, Kim YS, Machado TA, Quirin S, Benson B, Kadmon J, Raja C, Chibukhchyan A, Ramakrishnan C, Inoue M, et al. (2019). Cortical layer– specific critical dynamics triggering perception. Science 365, eaaw5202. 10.1126/science.aaw5202. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R30] 30.Russell LE, Dalgleish HWP, Nutbrown R, Gauld OM, Herrmann D, Fişek M, Packer AM, and Häusser M (2022). All-optical interrogation of neural circuits in behaving mice. Nat Protoc, 1–42. 10.1038/s41596-022-00691-w. [DOI] [PMC free article] [PubMed]

[R31] 31.Pérez-Ortega J, Alejandre-García T, and Yuste R (2021). Long-term stability of cortical ensembles. eLife 10, e64449. 10.7554/eLife.64449. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R32] 32.Ziv Y, Burns LD, Cocker ED, Hamel EO, Ghosh KK, Kitch LJ, El Gamal A, and Schnitzer MJ (2013). Long-term dynamics of CA1 hippocampal place codes. Nat Neurosci 16, 264–266. 10.1038/nn.3329. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R33] 33.Rule ME, O’Leary T, and Harvey CD (2019). Causes and consequences of representational drift. Current Opinion in Neurobiology 58, 141–147. 10.1016/j.conb.2019.08.005. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R34] 34.K Namboodiri VM, Hobbs T, Trujillo-Pisanty I, Simon RC, Gray MM, and Stuber GD (2021). Relative salience signaling within a thalamo-orbitofrontal circuit governs learning rate. Current Biology 31, 5176–5191.e5. 10.1016/j.cub.2021.09.037. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R35] 35.Belova MA, Paton JJ, Morrison SE, and Salzman CD (2007). Expectation Modulates Neural Responses to Pleasant and Aversive Stimuli in Primate Amygdala. Neuron 55, 970–984. 10.1016/j.neuron.2007.08.004. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R36] 36.Smith Y, Paré J-F, and Paré D (2000). Differential innervation of parvalbumin-immunoreactive interneurons of the basolateral amygdaloid complex by cortical and intrinsic inputs. Journal of Comparative Neurology 416, 496–508. . [DOI] [PubMed] [Google Scholar]

[R37] 37.Krabbe S, Gründemann J, and Lüthi A (2018). Amygdala Inhibitory Circuits Regulate Associative Fear Conditioning. Biological Psychiatry 83, 800–809. 10.1016/j.biopsych.2017.10.006. [DOI] [PubMed] [Google Scholar]

[R38] 38.Fu X, Teboul E, Weiss GL, Antonoudiou P, Borkar CD, Fadok JP, Maguire J, and Tasker JG (2022). Gq neuromodulation of BLA parvalbumin interneurons induces burst firing and mediates fear-associated network and behavioral state transition in mice. Nat Commun 13, 1290. 10.1038/s41467-022-28928-y. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R39] 39.Gordon-Fennell A, Barbakh JM, Utley MT, Singh S, Bazzino P, Gowrishankar R, Bruchas MR, Roitman MF, and Stuber GD (2023). An open-source platform for head-fixed operant and consummatory behavior. eLife 12, e86183. 10.7554/eLife.86183. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R40] 40.Tucci V, Hardy A, and Nolan PM (2006). A comparison of physiological and behavioural parameters in C57BL/6J mice undergoing food or water restriction regimes. Behavioural Brain Research 173, 22–29. 10.1016/j.bbr.2006.05.031. [DOI] [PubMed] [Google Scholar]

[R41] 41.Heiderstadt KM, McLaughlin RM, Wright DC, Walker SE, and Gomez-Sanchez CE (2000). The effect of chronic food and water restriction on open-field behaviour and serum corticosterone levels in rats. Lab Anim 34, 20–28. 10.1258/002367700780578028. [DOI] [PubMed] [Google Scholar]

[R42] 42.Vasilev D, Havel D, Liebscher S, Slesiona-Kuenzel S, Logothetis NK, Schenke-Layland K, and Totah NK (2021). Three Water Restriction Schedules Used in Rodent Behavioral Tasks Transiently Impair Growth and Differentially Evoke a Stress Hormone Response without Causing Dehydration. eNeuro 8. 10.1523/ENEURO.0424-21.2021. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R43] 43.Burgess CR, Ramesh RN, Sugden AU, Levandowski KM, Minnig MA, Fenselau H, Lowell BB, and Andermann ML (2016). Hunger-Dependent Enhancement of Food Cue Responses in Mouse Postrhinal Cortex and Lateral Amygdala. Neuron 91, 1154–1169. 10.1016/j.neuron.2016.07.032. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R44] 44.Livneh Y, Ramesh RN, Burgess CR, Levandowski KM, Madara JC, Fenselau H, Goldey GJ, Diaz VE, Jikomes N, Resch JM, et al. (2017). Homeostatic circuits selectively gate food cue responses in insular cortex. Nature 546, 611–616. 10.1038/nature22375. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R45] 45.Calhoon GG, Sutton AK, Chang C-J, Libster AM, Glober GF, Lévêque CL, Murphy GD, Namburi P, Leppla CA, Siciliano CA, et al. (2018). Acute Food Deprivation Rapidly Modifies Valence-Coding Microcircuits in the Amygdala 285189. 10.1101/285189. [DOI]

[R46] 46.Kaifosh P, Zaremba JD, Danielson NB, and Losonczy A (2014). SIMA: Python software for analysis of dynamic fluorescence imaging data. Front Neuroinform 8, 80. 10.3389/fninf.2014.00080. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R47] 47.Namboodiri VMK, Otis JM, van Heeswijk K, Voets ES, Alghorazi RA, Rodriguez-Romaguera J, Mihalas S, and Stuber GD (2019). Single-cell activity tracking reveals that orbitofrontal neurons acquire and maintain a long-term memory to guide behavioral adaptation. Nature Neuroscience 22, 1110–1121. 10.1038/s41593-019-0408-1. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R48] 48.Franklin KBJ, and Paxinos G (2013). Paxinos and Franklin’s The mouse brain in stereotaxic coordinates Fourth edition. (Academic Press, an imprint of Elsevier; ). [Google Scholar]

[R49] 49.Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, Preibisch S, Rueden C, Saalfeld S, Schmid B, et al. (2012). Fiji: an open-source platform for biological-image analysis. Nat Methods 9, 676–682. 10.1038/nmeth.2019. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R50] 50.Jimenez JC, Su K, Goldberg AR, Luna VM, Biane JS, Ordek G, Zhou P, Ong SK, Wright MA, Zweifel L, et al. (2018). Anxiety Cells in a Hippocampal-Hypothalamic Circuit. Neuron 97, 670–683 e6. 10.1016/j.neuron.2018.01.016. [DOI] [PMC free article] [PubMed] [Google Scholar]

[R51] 51.Piantadosi SC, Manning EE, Chamberlain BL, Hyde J, LaPalombara Z, Bannon NM, Pierson JL, Namboodiri VM, and Ahmari SE (2022). Hyperactivity of indirect pathway-projecting spiny projection neurons drives compulsive behavior 2022.02.17.480966. 10.1101/2022.02.17.480966. [DOI] [PMC free article] [PubMed]

PERMALINK

Holographic stimulation of opposing amygdala ensembles bidirectionally modulates valence-specific behavior via mutual inhibition

Sean C Piantadosi

Zhe Charles Zhou

Carina Pizzano

Christian E Pedersen

Tammy K Nguyen

Sarah Thai

Garret D Stuber

Michael R Bruchas

Summary

eTOC

Introduction

Results

Optically separable populations of BLA neurons respond to appetitive and aversive stimuli.

Figure 1. Optically separable populations of BLA neurons respond to appetitive and aversive stimuli.

Spatial distribution of valence-encoding neurons within the BLA

Figure 2. Spatial mapping of valence encoding BLA neurons.

Figure 6. Proximal BLA positive and negative valence neuronal ensembles display functional antagonism and mutual inhibition.

Holographic photostimulation via a GRIN lens is specific and robust at activating amygdala ensembles

Figure 3. Deep brain holographic stimulation of individual BLA neurons is specific and robust.

Targeted photostimulation of quinine ensemble reduces consummatory behavior to an appetitive stimulus

Figure 4. Selective simulation of quinine ensemble reduces licking of appetitive tastant.

Targeted photostimulation of sucrose ensemble increases consummatory behavior of aversive stimulus

Figure 5. Selective stimulation of sucrose ensemble increases licking to aversive tastant.

Antagonistic relationship between valence-encoding BLA neurons drives mutual inhibition

Discussion

STAR Methods

RESOURCE AVAILABILITY

Lead contact

Materials availability

Data and code availability

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Animals

METHOD DETAILS

Stereotactic surgery

Water restriction protocol

Head-fixed sucrose and quinine behavioral paradigm

Sucrose-only and quinine-only head-fixed behavioral paradigm

Sucrose, water, and quinine head-fixed behavioral paradigm

Two-photon imaging

Holographic photostimulation during behavior

Measurment of lateral and axial resolution during holographic photostimulation

Holographic photostimulation for connectivity mapping

Histological registration of virus and GRIN lens targeting

Two-photon imaging analysis and data processing pipeline

QUANTIFICATION AND STATISTICAL ANALYSIS

Tastant related activity classification

Population decoding of tastant type

Anatomical distribution of tastant-selective neurons

Analysis of ensemble proximity

Statistical analysis of fluorescence data

Statistical analysis of licking behavior

Supplementary Material

HIGHLIGHTS.

Acknowledgements:

Funding:

Footnotes

References

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases