2020-2021 Doctoral Awardee: Celine Cammarata

“The role of acetylcholine in flexible cognition across age and species”

Celine Cammarata, Cornell University, Department of Human Development


Cognitive flexibility, the capacity to adjust behavior to a changing environment, is notably compromised during aging. The neuromodulator Acetylcholine (ACh) supports cognitive flexibility, and is itself reduced later in life. Through experiments across rats and humans, we explored how ACh impacts cognitive flexibility during aging, and whether a body-brain manipulation could combat age-related decline in flexibility. First, we manipulated ACh in young and aged rats performing a flexible learning task to probe the interaction of ACh and aging on cognitive flexibility, revealing that aging and ACh inhibition drive strikingly similar effects. We recorded cortical activity from frontoparietal cortices of young and aged rats during flexible learning while manipulating ACh to assess the impact of ACh on this cortical system, known to support cognitive flexibility. ACh-dependent beta oscillations in the posterior parietal cortex are important for cognitive flexibility and absent in aged rats. Finally, we drove activity in the vagus nerve, putatively stimulating cortical ACh, to reduced age effects on cognitive flexibility in humans. Young and older adults faced a flexibility challenge after a breathing manipulation that boosted vagal function, elevating older adults’ cognitive flexibility to that of young adults. Our results illuminate the role of ACh in age-related decline of cognitive flexibility, the neural mechanisms underlying this, and routes of intervention.


Navigating our dynamic environment demands that we fluidly update our behavior. However this capacity, termed cognitive flexibility, declines notably with age1–7, with consequences for other aspects of cognition and for independent living8,9.

In young adults cognitive flexibility is supported by the neurochemical Acetylcholine (ACh)10–17. A cluster of neurons in the basal forebrain synthesize ACh and release it in the cortex to modulate processing in targeted circuits, particularly classically supporting attention and memory 18–24 Among the cortical targets of ACh is the frontoparietal system 25–28. Neuroanatomical studies have pinpointed this cortical system, consisting of the prefrontal cortex and the posterior parietal cortices, as a key actor in cognitive flexibility 29–31. Lesioning either of these regions reduces cognitive flexibility 32–34  whereas neuroimaging studies repeatedly show that flexibility activates this system  35–40.

The frontoparietal cortices are functionally linked to the vagus nerve 41,42, a far-reaching peripheral nerve in the parasympathetic system43 that carries information bidirectionally between the body and brain. Stimulating the vagus can alter cortical activity 44–46 and can improve  memory and learning 47–49. This may be in part through vagal activity driving the basal forebrain cholinergic system to modulate the cortex 46,50. Thus, the vagus nerve may provide a body-brain pathway that can increase ACh activity in the cortex.

Notably, the synthesis and release of ACh decreases in later life, with this decline implicated in both pathological dementia and normal cognitive aging 51–54. Despite the parallels between ACh’s role in cognitive flexibility, it’s role in aging, and the decrease in cognitive flexibility seen with age, few studies have investigated how ACh contributes to reduced flexibility in older adults. We probed this through three studies asking: 1) how does the role of ACh in cognitive flexibility interact with aging? 2) does cholinergic modulation of cortical activity contribute to cognitive flexibility, and how does this change during aging? and 3) can a vagal manipulation – putatively elevating ACh activity – reverse age-effects on cognitive flexibility?

We explored cognitive flexibility by utilizing the phenomenon of proactive interference  (PI), i.e., the impairment of new learning that overlaps or conflict with past knowledge. While young adults – humans and rodent models alike – overcome PI readily, the ability to resolve PI is impaired by advanced age across species 55,56.

In the first two studies, we pharmacologically manipulated ACh activity in young adult and aged rats during our PI-resolution task first to test how ACh and aging interact in cognitive flexibility (study 1) and next to record population-level neural activity from frontoparietal cortices, assessing the influence of age and ACh on neural systems that support cognitive flexibility (study 2). In our third study, we explored the vagus-ACh relationship by comparing the effect of a breathing manipulation that acts on the vagus nerve on PI resolution in young and older adult.


Proactive interference resolution task – Rodent

We probed cognitive flexibility by eliciting PI in an associative learning task15,16,57. Rats learned pairs of target and distractor odors and had to indicate the target odor with a nosepoke (Fig. 1A). Rats first learned a baseline stimulus pair (Fig. 1B), then advanced to a 5-day testing phase consisting of interleaved trials of two types: PI trials in which the target is a new odor, but the distractor odor had been the target in the baseline pair; and Novel trials consisting of two new stimuli. Each test session started with 32 Baseline trials preceding 96 interleaved PI and Novel trials. Accuracy on the task was submitted to mixed-effect models, the results of which were in turn submitted to F-tests for statistical analysis.

Cholinergic manipulation

In studies 1 and 2 we manipulated ACh by systemic injections of the ACh-receptor antagonist scopolamine (SCOP), or, as a control, methylscopolamine (MSCOP) which inhibits ACh activity in the periphery but cannot pass the blood brain barrier, allowing us to isolate the effect of ACh in the brain. SCOP/MSOP, at 0.25 mg/kg body weight, were administered during the test phase 20 minutes prior to behavior. Each rat experienced both drug conditions with separate stimulus sets and drug order was counterbalanced across rats.

Neural recording and data analysis

We measured local field potentials (LFP) from the medial prefrontal cortex (PFC) and posterior parietal cortex (PPC) through stainless-steel electrodes (Fig 1C). LFP captures the oscillatory activity of neural populations; these oscillations occur at particular frequencies that research has associated with behavior and cognition58,59. We analyzed the LFP activity during the test phase, in a half-second period immediately before the rat made a response. Activity for each PI and Novel trial was normalized to the mean LFP activity for that individual’s Baseline trials in the same session. A Fast Fourier Transform was used to characterize the power in each of four characteristic frequency bands: theta (4-10Hz), beta (10-30Hz), gamma (30-70Hz), and high gamma (70-100Hz). LFP power was submitted to mixed models, the results of which were in turn submitted to F-tests for statistical analysis.

Proactive interference resolution task – Human

The human PI-resolution task largely mimicked the rodent version, with abstract visual stimuli on a computer screen. Humans first learned a baseline pair of images, then were presented with a PI-manipulation of this pair. Unlike the rat task, in the human version we compared this to a response reversal (RR) manipulation. Each baseline pair was learned in a single run and followed by a single PI or RR run (Fig 1D).

Vagus activity measurement and manipulation

The main proxy for vagal activity is heart rate variability (HRV), the difference in intervals between heart beats60, which was measured through electrocardiograms and/or photoplethysmography. We drove vagal activity through slow breathing at 0.10 Hz (“slow breathing”)61–64. As a control, participants completed two minutes of controlled breathing at 0.28 Hz (“sham breathing”), a normal pace for older adults 65. In both conditions, participants were directed when to inhale and exhale by specific tones. This accounted for the effects of consciously modulating respiration, thus isolating the impact of breathing pace. Participants completed either a slow or a sham breathing run prior to each behavioral baseline run (Fig 1D).

Fig. 1. A) Operant behavior chamber used for rat behavior. Odors are released simultaneously from the two odor ports, and the rat’s nosepoke response to indicate one odor or the other is recorded by the infrared (IR) beam. Choosing the target odor results in a water reward at the far end of the chamber, while choosing the distractor odor causes an error tone. B) Schematic of stimulus pairs, where each color/letter indicates an individual odor. Rats first learn the baseline pair (A+/B-), then learn the PI pair (C+/A-) where the distractor was a previous target, and Novel pair (D+/E-) that is entirely new. C) Locations of LFP recording electrodes in the PFC (top) and PPC (bottom). Cgl = cingulate cortex; PrL = prelimbic cortex; IL = infralimbic cortex, all subregions of the medial prefrontal cortex. MPtA = medial parietal, LPtA = lateral parietal, subregions of the posterior parietal cortex. D) Overview of the human task procedure (left) and example visual stimuli (right) with PI and RR manipulations. The sham breathing manipulation always preceded the slow breathing manipulation, but the association of the PI vs. RR task with each breathing condition was counterbalanced across participants.


Study 1: Cholinergic disruption mimics the effect of natural aging on cognitive flexibility

In young (n = 12, 10 months) and aged rats (n = 11, 21 months), across drugs, accuracy was lower for the PI trials than for the Baseline (Fig. 2, t(576) = 15.451, p  < 0.01) or Novel trials (t(576) = 13.278, p < 0.01), indicating interference. As seen previously, in young rats SCOP specifically reduced accuracy in the PI trials (t(311) = 0.111, p < 0.001) compared to MSCOP (Fig. 2A-B).

Older rats were selectively impaired on the PI trials compared to young rats (t(81.6) = 2.496, p = 0.015) on MSCOP. Strikingly, among aged rats SCOP had no impact on performance (F(1,266.590) = 3.561, p = 0.060). Together these results indicate that while age and ACh-inhibition alone each reduce cognitive flexibility, together these have no further effect (Fig. 2C-D).

Fig. 2. * = p < 0.05 for all panels. A) Young rat accuracy (# trials with correct response / total # trials completed) for each trial type and drug condition, averaging over all testing days. B) Same as A, expanded to show learning over the testing period. Significance tests refer to pairwise comparisons of accuracy between the PI trials and Novel trials specifically. C-D) Same as A-B for aged rats.

Study 2: Acetylcholine supports cognitive flexibility through parietal beta oscillations that are compromised in aging

We recorded LFP in a subset of the rats from study 1. We recorded 2,416 trials of PPC activity from 7 young rats. In the PPC, under MSCOP we observed elevated power in the beta band prior to correct responses (Fig. 3A F(1,4809.714) = 4.841, p = 0.028), in the PI condition only (t (4809.72) = 2.577, p = 0.010). Beta power was suppressed under SCOP (Fig. 3C-D, F(1,4800.637) = 32.521, p < 0.001), indicating dependence on ACh. Furthermore, logistic regressions showed that the probability of a correct response increased with increasing beta power for PI trials but not Novel trials, further supporting that this beta activity may contribute to rats making a correct response in the face of PI. In the young rat PFC, where we recorded 2,339 trials from 8 rats, this beta signature failed to reach significance (Fig. 3B).

In either cortical region, aged rats failed to show the beta signature associated with PI-resolution in young rats (PPC = 1,080 from 4 rats and PFC = 1,245 trials from 5 rats, Fig 4). Indeed, LFP activity was largely dysregulated in aged rats under both drug conditions, particularly under SCOP.

Study 3: Taking twelve slow breaths enhances cardiovascular and cognitive plasticity

Both young (n = 51, mean 21 years) and older adults (n = 24, mean 67 years) modulated their breathing rate in accordance with our instructions (Fig. 5A), and in both groups the slow breathing selectively increased HRV ((F(2,104) = 17.02, p < 0.001, Fig. 5B).

Fig. 3. Spectrograms showing oscillatory power by frequency (Y axis) and time (X axis), where the 0 timepoint corresponds to the rats making a response. Heatmap indicated the difference in power between trials that resulted in a correct response and trials that resulted in an incorrect response. A) LFP power in posterior parietal cortex of young rats. Left side = MSCOP, right side = SCOP. Top row = PI trials, bottom row = Novel trials. Box highlights elevated beta oscillatory power in PI trials where the rat responded correctly compared to those with an incorrect response. B) Same as A for young rat prefrontal cortex.
Fig. 4. Spectrograms showing oscillatory power by frequency (Y axis) and time (X axis) for aged rats, where the 0 timepoint corresponds to the rats making a response. Heatmap indicated the difference in power between trials that resulted in a correct response and trials that resulted in an incorrect response. A) LFP power in posterior parietal cortex of aged rats. Left side = MSCOP, right side = SCOP. Top row = PI trials, bottom row = Novel trials. B) Same as A for aged rat prefrontal cortex.

In both PI and RR, aging was associated with poorer cognitive flexibility relative to young adults (Fig. 5C, t(61) = 2.842, p = 0.006).  As expected, there was also a significant interaction between task and group (F(1,443) = 9.087, p = 0.003), with a larger accuracy cost in older adults in adapting to PI (t(91) = 2.501, p = 0.014, Fig. 5C). Preceding the task runs with slow breathing increased accuracy across age groups and tasks (F(1,443) = 72.007, p < 0.001), while an interaction of breathing condition and age (F(1,443) = 10.200, p = 0.002) revealed a sufficiently greater impact on the older adults that abolished the age difference in PI resolution (t(95) = 0.541, p = 0.590).


Over a series of studies, we explored how neuromodulation by ACh contributes to age-related loss of cognitive flexibility and probe a possible mechanism to combat this through the vagus nerve.

Building on past findings that  SCOP impairs PI resolution in young rats and humans16,17, we compared the effect of this ACh-blockade in young and aged rats. In young rats under MSCOP, our control condition, this resolved over the testing phase such that by the last days there was no difference in accuracy between PI and Novel trials. However, both young rats on SCOP and aged rats in either drug condition failed to overcome PI over the testing days. There was no global effect of either age or SCOP on overall accuracy, rather both manipulations specifically impaired cognitive flexibility while leaving Novel learning and Baseline performance intact.

Furthermore, SCOP had no impact among the aged rats; that is, while these animals showed inefficient cognitive flexibility under MSCOP, blocking ACh receptors did not further reduce their performance. This suggests that aged rats already represent a state where ACh is not contributing to cognitive flexibility. The non-additive effects of ACh receptor antagonism and of age hint that these may capture the same underlying effect, i.e., that age-related decline in ACh is what drives the observed loss of cognitive flexibility.

Fig. 5. A) Respiratory rate by breathing condition, averaging over young and older adults. B) Root mean square of successive difference (RMSSD), a measure of heart rate variability, by breathing condition. Results are averaged over young and older adults. Baseline corresponds to a 30-second period when participants attended to their breathing but did not intentionally modulate breathing pace. C) Normalized accuracy by task, breathing condition, age group and quartile bin within task run. Each run consisted of 112 trials. For analysis we separated runs into quartile bins of 28 trials each. To account for practice and for global effects of age, we normalized accuracy in the PI and RR runs by subtracting the accuracy in the last quartile bin of the associated baseline run for that participant. Thus, normalized accuracy represents the cost associated with cognitive flexibility in transitioning from the baseline condition to the PI or RR condition, where a normalized accuracy of 0.0 means the participant did no worse or better on the PI/RR run than on the preceding baseline.

Recording population activity in the frontoparietal cortex of a subset of these rats revealed an ACh-sensitive beta oscillation pattern in the PPC corresponding with flexibility in young rats. Not only was this beta burst absent in the Novel trials, indicating a selective role in PI resolution, but it was suppressed by SCOP, suggesting that ACh orchestrates this activity to cognitive flexibility.

Beta oscillations have previously been associated with cognitive and behavioral inhibition 66–68 and filtering relevant information 69,70. In the present task, inhibition-related beta activity may assist in suppressing the previously learned response, permitting rats to overcome PI and learn to choose the new target.

Aged rats had dysregulated population activity and failed to show this ACh-dependent beta signature. Aging is also associated with reduced inhibition and reduced filtering of relevant information 71 72, further suggesting that beta oscillations may provide an inhibitory filtering mechanism whose absence in aged rats leaves them vulnerable to PI.

Among human participants, we saw that aging selectively challenged the integration old and new information (PI), but not remapping old information (RR). As expected a 2-minute slow breathing intervention elevated vagal activity, as measured by HRV. We demonstrated that after this breathing intervention, cognitive flexibility was increased and the age difference in flexibility was abolished. Although observed in both cognitive flexibility tasks, this effect was especially pronounced for PI resolution.

We suggest that this may occur through breathing activating the vagus nerve, in turn driving ACh activity in the brain, however further work must be done to clarify the mechanisms. Regardless, our finding that such a brief, simple manipulation was able to drive older adults’ cognitive flexibility to resemble that of individuals 30-40 years their junior sets an exciting precedent for possible future intervention.

Impact Statement

With the average age increasing around the globe a growing segment of the population is vulnerable to cognitive decline, but current treatments are often ineffective. By understanding the neural mechanisms of decline, as we have probed in studies one and two, researchers will be better able to predict decline and help individuals maintain a higher level of cognition and, consequently, independence and comfort throughout their later years. We took a first step toward this in our third study, where we tapped in a connection between the body and brain to boost cognitive flexibility through a simple breathing manipulation. Such a technique is cost-free and easy to disseminate, potentially helping older adults alleviate challenges to cognition and helping combat the global risk of cognitive decline.