CO2 reactivity is associated with individual differences in appetitive extinction memory

Abstract

Pavlovian conditioning can underly the maladaptive behaviors seen in psychiatric disorders such as anxiety and addiction. In both the lab and the clinic, these responses can be attenuated through extinction learning, but often return with the passage of time, stress, or a change in context. Extinction to fear and reward cues are both subject to these return of behavior phenomena and have overlap in neurocircuitry, yet it is unknown whether they share any common predictors. The orexin system has been implicated in both fear and appetitive extinction and can be activated through a CO₂ challenge. We previously found that behavioral CO₂ reactivity predicts fear extinction and orexin activation. Here, we sought to extend our previous findings to determine whether CO₂ reactivity might also predict extinction memory for appetitive light-food conditioning. We find that the same subcomponent of behavioral CO₂ reactivity that predicted fear extinction also predicts appetitive extinction, but in the opposite direction. We show evidence that this subcomponent remains stable across two CO₂ challenges, suggesting it may be a stable trait of both behavioral CO₂ reactivity and appetitive extinction phenotype. Our findings further the possibility for CO₂ reactivity to be used as a transdiagnostic screening tool to determine whether an individual would be a good candidate for exposure therapy.

Keywords: Appetitive conditioning; CO(2); Extinction; Individual differences; Orexin.

Figure 1.

Experimental timeline. Male rats (n = 48) underwent a CO₂ reactivity assessment. They then underwent appetitive conditioning with 16 trials of light-food pairings per day for 4 days. The next day, they received an extinction session with 18 trials of the light alone. The day after, they received a long-term memory (LTM) test with 4 trials of the light alone. They then underwent a second CO₂ reactivity assessment prior to euthanasia.

Figure 2.

CO₂ concentration measured in the chamber during each of the two-minute-long induction, 25% hold, and flush out phases of the CO₂ reactivity assessment. Mean (± SEM) as measured by a device located inside the chamber. Note that the concentration reaches a peak of approximately 25%.

Figure 3.

Distribution of CO₂ reactivity behaviors. Boxplots of the behaviors analyzed (ambulation, rearing, labored breathing, grooming) during each phase (induction, 25% hold, flush out 1, flush out 2) of the two CO₂ reactivity assessments. The quantity of each of the four behaviors observed during the four phases yielded 16 behavioral subcomponents which were entered as possible predictors of long-term appetitive extinction memory in our model. Note the degree of variability of each subcomponent that remains consistent between CO₂ sessions.

Figure 4.

Conditioned food seeking behavior across conditioning (A) and extinction and long-term memory (LTM) tests (B). Each trial was calculated as the difference between the second half of the 10-s CS (CS2) and the 5-s before the CS (preCS) and shown in blocks as two trials averaged (± SEM). Food cup responses increase with conditioning and decrease with extinction. (C) Data points of food cup responses for each subject from the last two-trial block of extinction and the first two-trial block of LTM. Note the high degree of variability between subjects. (D) The difference between the two values shown in (C) was used to calculate adjusted LTM for each subject and used as the outcome variable in our model. Note the high degree of variability between subjects.

Figure 5.

Relationship of the two strongest predictors from CO₂ session 1 with adjusted LTM of food cup responses. We used the best subset approach to linear regression to determine which subcomponents of CO₂ reactivity predict the most variance in long term appetitive extinction memory. This approach enabled us to choose a model that has the fewest predictors, the highest predictive power, and the highest degree of cross-sample replicability. The model suggests that the best predictive effect of CO₂ reactivity is from ambulation during the first half of the flush out, rearing during the second half of the flush out phase, and grooming during the induction phase with a combined R² = 0.23 for the train sample and 0.077 for the cross-validation sample. The relationship of ambulation during flush out 1 (A) and rearing with flush out 2 (B) with adjusted LTM of food cup responses are shown.

Figure 6.

Relationship of ambulation during flush out 1 between CO₂ reactivity sessions 1 & 2. To determine if any subcomponents of behavioral CO₂ reactivity were stable across sessions, we computed a correlation matrix (see Supplemental Figure 1). Of the significant correlations, the only one that was also identified as a predictor of adjusted LTM of food seeking behavior was ambulation during flush out 1. Thus, this predictor may also be a stable trait of behavioral CO₂ reactivity.