In a recent study led by Ian Schmitt in the lab (http://www.nature.com/nature/journal/vaop/ncurrent/full/nature22073.html?WT.feed_name=subjects_neuroscience), where we ran a number of experiments aimed at understanding the circuit and computational principles underlying holding a task rule in mind (what probably is similar to keeping a phone number in one’s head for a few moments en route to dialing that number; also related to what we think working memory is; also related to executive control) and found something quite unexpected. While we did observe what has been known about the circuit substrates for executive function, the prefrontal cortex (PFC) was important for maintaining an active representation of the task rule in the absence of any sensory stimulation. This maintained representation was also dependent on local synaptic connections within the PFC (again, not super surprising). What was surprising though was:
Surprise number 1: PFC representations were dependent on inputs from the mediodorsal thalamus
Surprise number 2: While PFC representations could easily tell two task rules apart, thalamic representations did not
Surprise number 3: Thalamic inputs did not drive PFC spikes
How does this all fit together in a coherent framework? We hypothesized that this type of thalamic input (which is different from that in sensory systems) leads to changes in functional connectivity between PFC neurons rather than their overall excitability (probably through mechanisms that target compartment-specific neural excitability). This was consistent with many subsequent findings, including direct demonstration of this functional connectivity enhancement and the fact that enhancing thalamic excitability augments PFC representations and task performance… very very different than what happens when a sensory thalamic circuit is broadly excited and the animal is expected to perceive a sensory stimulus with specific features.
What was particularly nice is that this paper appeared with two others; one from Josh Gordon/Christoph Kellendonk’s labs and another from Karel Svoboda’s lab. All papers indicate that the thalamus can enhance local recurrence in its target cortical circuit. There are some interesting differences seen across different frontal thalamo-cortical loops (particularly comparing prefrontal to motor), in relation to the degree of task-variable selectivity in thalamus and possibly in the impact of thalamic input on cortical circuit in relation to those particular variables. Regardless of these differences, one principle can be taken to the bank; the thalamus is not only a source of categorical information to cortex but rather is important for building cortical representations that are dependent on local recurrence.
Here’s a couple of press releases that highlight these studies together:
NYU eureka alert: https://www.eurekalert.org/pub_releases/2017-05/nlmc-nrf050217.php
We are super excited about this and will be pursuing this direction of thalamic control of functional cortical connectivity in a cognitive context. So many exciting collaborations are now possible including those with human neuroimaging as well as AI (afterall, this is a novel computational architecture that seems to be at work, and one that differs from a corticocentric worldview of cognitive computations).
Related to all of the above, we wrote a very short but to the point “appetizer” review on this topic: http://www.sciencedirect.com/science/article/pii/S0959438816302525
More to come in terms of science and synthesis, so stay tuned…