How do cocaine and shock activate distinct nucleus accumbens ensembles?

Tagging which neurons were active during each experience

The study asks how the nucleus accumbens (NAc), a striatal region that responds to both rewards and aversive stimuli, represents experiences of opposite valence. To answer it, the authors used TRAP2;Ai14 mice, in which the c-fos immediate early gene locus drives a tamoxifen-inducible CreER. When a neuron is active while 4-hydroxytamoxifen (4-OHT) is present, recombination permanently switches on a fluorescent reporter, tdTomato. A tdTomato-positive cell is therefore a record of activity during the tagging window, not a marker of cell type.

The two tagging protocols differed in stimulus and timing. Cocaine mice received a single intraperitoneal injection of cocaine hydrochloride (20 mg/kg) immediately followed by 4-OHT, after seven days of habituation to saline injections. Shock mice received 20 foot shocks (1.5 mA, 1 second each, 30-second intervals) and then 4-OHT. Cell identity was read out with RNAscope probes for Drd1 and A2a, with A2a used as the D2-MSN marker because cholinergic interneurons also express Drd2.

This design supports comparisons of ensemble size, anatomical distribution, and cell-type composition between a drug reward and an aversive stimulus. It does not by itself show that tagged cells cause behavior; that required the optogenetic reactivation experiments described later.

Cocaine recruits a larger, D1-enriched anterior ensemble

Cocaine administration recruited a significantly larger tdTomato-positive ensemble than saline in the NAc core (unpaired t-test, t6 = 2.4, p = 0.0500) and lateral shell (t6 = 3.7, p = 0.0095), with a trend toward more cells in the dorsomedial and ventromedial shell. The cocaine ensemble also had a distinct anteroposterior profile, with the highest proportion of activated cells in the more anterior NAc.

Cell composition was strongly biased. Among tagged cocaine-ensemble cells, 57% were D1-MSNs and 15% were D2-MSNs. The saline ensemble contained a similar share of D1-MSNs (56%) but a higher share of D2-MSNs (31%), so the cocaine-specific difference lies mainly in how few D2-MSNs were recruited. These percentages describe the composition of tagged cells, not the fraction of all D1- or D2-MSNs in the region that responded.

Foot shock recruits both MSN types across medial NAc

Foot shock produced a larger tagged ensemble than the no-shock condition in the NAc core (t5 = 2.7, p = 0.0425), dorsomedial shell (t5 = 9.1, p = 0.0003), and ventromedial shell (t5 = 2.7, p = 0.0402). Unlike the cocaine ensemble, the shock ensemble was distributed uniformly along the anteroposterior axis without clear segregation between subregions.

Its composition was also less skewed: 47% of tagged shock-ensemble cells were D1-MSNs and 32% were D2-MSNs. The paper describes this as proportional recruitment of both populations. The contrast with cocaine is therefore twofold: cocaine favors D1-MSNs and the anterior NAc, whereas shock draws on both MSN types across medial subregions.

Repeated exposure re-engages only part of each ensemble

To test whether the same stimulus re-engages previously tagged cells, mice were exposed to cocaine or shock during tagging and again two weeks later, after which endogenous c-fos was quantified. Reactivated neurons were those positive for both tdTomato and c-fos.

Re-exposure to cocaine reactivated 20.9% of cells in the NAc core and 18.3% in the shell, compared with 6.5% and 4.5% in saline controls. Repeated shock reactivated 41.1% of core cells and 39.1% of shell cells, compared with 13.5% and 5.7% in no-shock controls. Both ensembles are therefore partially, not fully, re-engaged, and the shock ensemble shows the higher reactivation rate in this assay.

Calcium imaging shows opposite D2 dynamics for the two stimuli

The tagging experiments identify which cells were active, but not whether they were excited or inhibited. The authors therefore used one-photon calcium imaging with miniaturized microscopes in freely moving D1-Cre and A2A-Cre mice expressing GCaMP6f, recording 426 D1-MSNs and 209 D2-MSNs around cocaine administration and 372 D1-MSNs and 239 D2-MSNs around foot shock.

Acute cocaine generated significantly more excitation in D1-MSNs than in D2-MSNs (chi-square test, p < 0.0001): it primarily excited D1-MSNs while inhibiting most D2-MSNs. Peak z-scores were also higher in D1-MSNs after cocaine (Mann-Whitney U = 39907, p = 0.0338). Foot shock, by contrast, induced excitatory responses in both MSN types. A support vector machine decoder could predict cocaine exposure from D1-Cre and A2A-Cre population activity, and likewise predicted foot shock delivery, indicating that the activity patterns carry stimulus-specific information.

The paper notes that many neurons responded to both stimuli but in opposite directions. That observation supports flexible valence encoding by overlapping populations, but it is a correlation between activity and stimulus, not proof that either population causes the behavioral response.

Reactivation dissociates cocaine preference from shock effects

Optogenetic reactivation tested causality. In a cocaine conditioned place preference paradigm, stimulating the cocaine-tagged ensemble did not significantly change the difference in time spent between stimulation and non-stimulation compartments relative to saline-ensemble stimulation. However, preference for the stimulation side correlated positively with cocaine ensemble size in the NAc core (r2 = 0.7920, F1,5 = 19.04, p = 0.0073) and medial shell (r2 = 0.5744, F1,5 = 6.7, p = 0.0484), with no such correlation in the saline group. In a two-choice operant task, cocaine-ensemble stimulation reinforced lever pressing.

For the shock ensemble, the picture was weaker. Stimulation did not produce clear avoidance in place preference or operant paradigms; it modestly decreased overall lever pressing, which the authors interpret as general suppression of motivation rather than directed aversion. In a fear conditioning and extinction experiment, mice previously exposed to shock or no shock showed similar conditioned freezing, and both groups declined across seven extinction days despite optical stimulation of the shock ensemble. Freezing on the last extinction day did not correlate with shock ensemble size.

The authors conclude that reactivation of the cocaine ensemble is sufficient to promote preference and reinforcement, whereas reactivation of the shock ensemble alone is not sufficient to maintain or reinstate fear. They also caution that ensemble identity and recruitment size cannot be separated causally by this design, since larger cocaine ensembles correlated with stronger preference.

Overlap in cells does not mean overlap in function

The study's central claim is that rewarding and aversive experiences recruit partially overlapping NAc ensembles with distinct cellular composition, activity dynamics, anatomical distribution, and behavioral impact. The cocaine ensemble is D1-biased and anterior; the shock ensemble is mixed and medially distributed; cocaine excites D1-MSNs while inhibiting most D2-MSNs, whereas shock excites both.

The authors offer mechanistic interpretations rather than demonstrated causes. They suggest the D1 bias fits dopamine-dependent potentiation during cocaine exposure, and that balanced D1/D2 recruitment by shock may reflect afferent input from regions such as the basolateral amygdala, prefrontal cortex, or ventral tegmental area. They also raise the possibility that the shock-tagged ensemble captures not only the aversive experience but the transition to a post-aversive relief state, because 4-OHT was given immediately after the shock session. That would help explain D1-MSN recruitment, overlap with the cocaine ensemble, and the modest behavioral effect of shock-ensemble reactivation.

These are hypotheses consistent with the data, not results the experiments isolate. The behavioral dissociation is the firmer finding: only reactivation of the cocaine-responsive ensemble changed preference and reinforcement in these paradigms.

Where the conclusion stops

The evidence comes from mice, with group sizes of three to four animals in the tagging comparisons and specific stimulus parameters: 20 mg/kg cocaine, 20 foot shocks at 1.5 mA for tagging, and 10 shocks at 0.7 mA during imaging. The imaging and tagging experiments use different shock protocols, so the activity dynamics and the tagged composition should not be treated as the same measurement.

The authors also note that investigators were not blinded during behavioral testing, though they were blinded during data analysis, and that the independent contributions of ensemble identity and ensemble size remain unresolved. The shock-ensemble result in particular should be read as a failure to produce strong avoidance under these reactivation conditions, not as evidence that the NAc plays no role in aversion.

The paper's most useful priority is the contrast it establishes: in the NAc, a cocaine-tagged D1-enriched ensemble is sufficient to drive preference and reinforcement, while a shock-tagged mixed ensemble is not sufficient to drive avoidance or sustain fear. Readers comparing valence representations should therefore weigh cell composition, anatomical position, and reactivation paradigm together, rather than treating overlap between tagged populations as evidence of shared function.

Sources used in this answer