A long-standing open question is how brain circuits learn what information is behaviorally relevant and should be attended. We have a new key insight to this question published in the paper “Adaptive Reinforcement Learning is causally supported by Anterior Cingulate Cortex and Striatum” (Neuron, pdf). We found that brief 0.3 s electrical stimulation at the time subjects looked at visual objects causally changed how these objects became top-down attentional target objects. The stimulation enhanced learning the relevance of visual objects in the striatum and impaired this learning in the anterior cingulate cortex. The effect was most pronounced when there were many alternate features present that subjects needed to filter out from attention, i.e. when there was higher feature uncertainty and the need for feature-specific credit assignment during learning.

The paper is an experimental tour-de-force by stimulating in two brain areas, modeling the stimulation effects on behavior (to show the effects are strongest on adaptive reinforcement learning mechanisms) and recording from neurons in the brain areas (to show which neuronal task representations are likely affected). Kudos to Louis, Kia, Charlie and Adam to succeed with such a comprehensive study and to Prof. Paul Tiesinga for help with an advanced Triple Mechanisms Adaptive Reinforcement Learning model that critically helped understanding the mechanisms underlying the observed effects.

Our new paper shows that the flexible learning of attention sets is facilitated by noradrenergic receptor action. The paper “Noradrenergic alpha-2a receptor stimulation enhances prediction error signaling and updating of attention sets in anterior cingulate cortex and striatum” is published in Nature Communications was led by PhD alumnus Ali Hassani and found that neuronal firing in the anterior cingulate cortex, striatum and lateral prefrontal cortex correlated with prediction error (PE) signals that we estimated with a reinforcement learning model. These PE signals were represented stronger when the selective alpha-2A agonist Guanfacine was given at doses that facilitated the reversal learning of feature-based attention. Guanfacine is a drug used for attention-deficit hyperactivity disorder ADHD) amongst others. Our study suggests that there is indeed a dose range for this drug that supports focusing on goal-relevant information and controlling more effectively distracting inputs. 

These findings provides insights how the noradrenergic system supports cognitive flexibility and – given the feature based reversal task used in this study –  speeds up adjusting feature-based attentional priorities. Kudos to Ali Hassani and to our collaborator Paul Tiesinga to have pulled off such an important insights. 

We published a new unity based software platform for profiling cognitive and motivational constructs in nonhuman primates and humans. The platform has multiple pre-configured tasks with some gamified features that makes them engaging to play for participants. Details are described and linked on the website: http://m-use.psy.vanderbilt.edu. The technical details are available in Watson et al. (2023) A multi-task platform for profiling cognitive and motivational constructs in humans and nonhuman primates. 1-48; bioRxiv; https://www.biorxiv.org/content/10.1101/2023.11.09.566422v1.

The platform is open source and well documented for users and for potential developers who want to extend the platform.

We have a new major finding published at Neuron. We found that spiking in different areas (ACC, Striatum, LPFC) engage in ~20ms wide correlations, and that this coordinated activity has systematic time lags that correspond to the anatomical connectivity. These ‘baseline routing states’ are amplified during beta bursts, and switch directionality (between ACC and PFC) during bursts in the theta frequency band (and with attention and during a choice).

The paper has multiple unique analyses and findings, is backed up by biophysically realistic modeling and shows that the specific neurons that contribute to a directionally leading routing state can be identified.

The implications of this paper can be huge, as it documents a neuronal ensemble mechanism that correlates in 20ms time windows during inter-areal coordination. Kudos to Kianoush Banai Boroujeni to have pulled off such a comprehensive and sophisticated novel insight !

We have new research out at PNAS about enhancing cognitive flexibility with highly selective allosteric modulation of the M1 muscarinic receptor (pdf: here)!

Muscarinic receptors are known to mediate pro-cognitive effects of acetylcholine, but it has remained unclear whether they differentially affect the cognitive subfunctions of attentional filtering, set shifting, and learning. To clarify the functional specificity of M1 mAChRs, we assessed these diverse functions using a recently developed, highly selective M1 PAM developped at the Warren Center of Neuroscience Drug Discovery by co-authors Prof. Jones and Dr. Russel. This novel M1 PAM caused domain-specific cognitive improvement of flexible learning and extradimensional set shifting, reduced perseverations and enhanced target recognition during learning without altering attentional filtering functions. These domain-specific improvements contrasted to effects of a nonselective acetylcholinesterase inhibitor that primarily enhanced attention and caused dose-limiting adverse side effects. These results demonstrate domain-specific improvements in cognitive flexibility suggesting M1 PAMs are versatile compounds for treating cognitive deficits in schizophrenia and Alzheimer’s disease.

Acetylcholinergic Drug enhances attention at different dose as cognitive flexibility
We tested how a cholinergic drug that is used to treat symptoms of dementia (donepezil, Arizept) affects cognitive abilities across multiple domains in monkeys. We found that donepezil showed stunning improvements of attentional filtering (less distraction) during visual search but at a different dose at it enhanced flexible learning of attention sets. See our paper Hassaniet al. (2023). Dose-dependent dissociation of pro-cognitive effects of donepezil on attention and cognitive flexibility in rhesus monkeys. Biological Psychiatry GOS. 3(1), p.68-77

We used focused ultrasound (FUS) sonication of the anterior cingualte and striatum to disrupt local processing during learning. FUS in ACC slowed down learning of atetntion sets – but only when the attentional demands were high and the task included the risk of loosing already attaiuned reward tokens. Under these cognitive and motivaitonally challenging conditions FUS reduce the ability to adapt and update attentional sets. See our paper Banaie Boroujeni et al. (2022) The anterior cingulate cortex causally supports flexible learning under motivationally challenging and cognitively demanding conditions. PLoS Biology. doi: https://doi.org/10.1101/2021.08.04.455080.

We tested which model mechanisms best explain how six animals learn attention sets and found a common set of most-important behavioral mechanisms that account for learning success.
When learning attention sets is easy value based reinforcement learning and working memory are powerful, but when learning problems are more complex learning is more efficient with attention and a meta-learning process that help speeding up learning when errors accumulate. (See our paper Womelsdorf at al. (2022) Learning at variable attentional load requires cooperation between working memory, meta-learning and attention-augmented reinforcement learning. Journal of Cognitive Neuroscience 34(1) 79-107.)

We now published the hardware and software design for a novel Monkey Kiosk Station that provides cognitive enrichment and the ability to assess cognition with cage-based touchscreen tasks. The paper and its appendix with the technical details are available here.

Our new paper in eLife shows that a subclass of fast spiking interneurons in prefrontal and anterior cingulate cortex gamma synchronizes when uncertainty about cues and outcomes is resolved. This finding was possible by classifying narrow spiking neurons into fast and non-fast spiking classes and correlating their firing and spike-LFP synchrony during processing of attention cues and reward outcomes in a reversal learning task. In prefrontal cortex the interneurons of the fast spiking subclass synchronized after cue onset during learning when there was uncertainty about the value of different stimuli. In anteiror cingulate cortex the same type of fast spiking interneuron class gamma synchronized after rewards were delivered, but only when the reward predictions errors were high during learning, i.e. when outcomes were uncertain. Computational modeling showed that the interneuron specific gamma synchrony could reflect a soft gating of competing excitatory inputs.
These findings are of fundamental importance for understanding cell type specific circuit functions and might be the first characterization of interneurons in the primate prefrontal cortices during higher cognitive learning performances. The paper can be downloaded here.