The Puzzling Pulvinar

Abstract

The functional role of the pulvinar, with its widespread cortical connectivity, has remained elusive. In this issue of Neuron, Jaramillo et al. (2019) provide a computational roadmap for how the pulvinar might influence various cognitive behaviors across multiple large-scale networks.

In the context of international relations, Winston Churchill referred to the actions of Russia at the beginning of World War II as “a riddle, wrapped in a mystery, inside an enigma.” In the context of neurosscience, this same characterization might be applied to the pulvinar: the largest nucleus in the primate thalamus. The pulvinar has expanded during evolution on a scale comparable to the prefrontal cortex. This thalamic nucleus is heavily interconnected with multiple cortical areas, providing an anatomical framework for intricate cortico-pulvino-cortical path ways. The specific functions associated with these transthalamic pathways, however, have remained largely unknown. Research into the functional role of the pulvinar had a promising start in the 1970s and 1980s, with both electrophysiological and lesion studies linking the pulvinar primarily with attention-related processing (Saalmann and Kastner, 2011). Yet despite this promising start, it has proven difficult to confirm or further define the role of the pulvinar in perception and cognition. For some, investigating the pulvinar has therefore become synonymous with entering the graveyard of neuroscience, with potentially grim consequences for scientific careers. In recent years, however, pulvinar research seems to be rising from the dead. Methodological advances have allowed researchers to focus on interactions between the pulvinar and interconnected brain regions (Fiebelkom et al., 2019; Purushothaman et al., 2012; Saalmann et al., 2012; Zhou et al., 2016). As a result, the mysteries of the pulvinar are beginning to unravel. Jaramillo et al. (2019), in the present issue, show that adding the pulvinar to a cortico-cortical circuit model provides critical flexibility during various cognitive behaviors (i.e., attention, working memory, and decision making). Their cortico-pulvinar circuit model reproduces several key physiological and behavioral findings from previous investigations and thus provides a computational roadmap for our understanding of how the pulvinar influences cognition.

Unlike the sensory thalamus, which transfers information from the peripheral sensory organs to cortex, the pulvinar is considered a higher-order nucleus of the thalamus. That is, the pulvinar forms input-output loops almost exclusively with the cortex. There are two well-established types of cortico-pulvinar pathways: a transthalamic feedforward pathway that indirectly connects two cortical areas through the pulvinar and a feedback pathway that projects from a cortical area to its projection zone in the pulvinar (Sherman and Guillery, 2006). Both of these pathways primarily connect the pulvinar with cortical areas dedicated to visual processing. The specific connectivity patterns with visual cortex are indicative of at least three subdivisions in the pulvinar: the inferior pulvinar is interconnected with early visual areas, the lateral pulvinar with ventral and dorsal extrastriate areas, and the mediodorsal pulvinar with higher-order frontal and parietal areas. Although these cortico-pulvinar pathways are well established, relatively little is known about their functions.

One proposal has been that the pulvinar provides a clocking mechanism for cortico-cortical interactions, thereby promoting between-region communication. Given the general rule that any directly connected cortical regions are also indirectly connected through the pulvinar, this thalamic nucleus is well positioned to mediate both feedforward and feedback interactions between cortical regions (Saalmann and Kastner, 2011). Researchers have proposed that such pulvino-cortical coordination might be mediated through the pulvinar-directed synchronization of cortical oscillations (Saalmann and Kastner, 2011). According to this account, pulvinar-directed synchronization of cortical oscillations gates the transfer of information between cortical regions by optimally aligning high-excitability states (or phases). Thus far, two electrophysiological studies in macaques have provided evidence in support of this proposal (Fiebelkom et al., 2019; Saalmann et al., 2012). Simultaneous recordings from the ventrolateral pulvinar and visual areas V4 and TEO during a behavioral task that manipulated spatial attention revealed increased between-region oscillatory synchronization during spatial attention, with the pulvinar being the apparent source of this synchronization (Saalmann et al., 2012). These findings were recently extended to higher-order cortex by a study that reported similar results while simultaneously recording from the mediodorsal pulvinar and two cortical hubs of the network that directs spatial attention, the frontal eye fields (FEF) and the lateral intraparietal region (LIP) (Fiebelkorn et al., 2019). Together, these studies suggest that the pulvinar facilitates attentionrelated communication across a largescale network that includes both sensory and higher-order cortices, thereby coordinating cortical interactions in the temporal domain.

In addition to serving a functional role as a thalamic timekeeper for cortical large-scale networks, there is evidence that the pulvinar supports—and may even be necessary for—basic sensory processing (Purushothaman et al., 2012; Zhou et al., 2016). Studies pairing pulvinar lesions with electrophysiological recordings in visual cortex indicate that pulvino-cortical inputs are required for normal sensory-cortical function. Deactivating the pulvinar greatly diminishes sensory responses in visual cortex (Purushothaman et al., 2012; Zhou et al., 2016), suggesting that the role of the pulvinar is not limited to gating between-region interactions (Fiebelkorn et al., 2019; Saalmann et al., 2012; Zhou et al., 2016). Rather, the pulvinar also provides critical support for local cortical computations during visual-sensory processing (Purushothaman et al., 2012; Zhou et al., 2016). This evidence corroborates classical proposals, which suggested that sensory information could not get routed between cortical areas without a functioning pulvinar (Sherman and Guillery, 2006). It should be noted, however, that chronic pulvinar lesions in humans often result in subtle or temporary behavioral deficits, indicating that the pulvinar has more of a modulatory impact on cortical functions (Halassa and Kastner, 2017; Saalmann and Kastner, 2011).

The functional role of the pulvinar is not exclusively mediated through its influence on cortex. Komura et al. (2013), for example, reported that the firing rate of single neurons in the mediodorsal pulvinar explicitly represents decision confidence. Here, macaques had to decide whether to make a decision that would result in a larger reward or, if their confidence in that decision was low, to opt out of the task and receive a smaller reward. The firing rate in pulvinar neurons was predictive of the decision to opt out. These results thus indicate that the functional role of the pulvinar is not limited to supporting computations that occur in other brain regions; some behaviorally relevant computations instead occur within the pulvinar itself.

The diversity of cognitive functions associated with the pulvinar reflects the diversity of cognitive functions associated with the brain regions that are interconnected with the pulvinar. For example, the mediodorsal pulvinar in macaques is interconnected with LIP. This region of parietal cortex, which was described above as a cortical hub of the attention network, is also involved in decision making. The mediodorsal pulvinar has likewise been linked to both attention-related processing and decision making (Fiebelkorn et al., 2019; Komura et al., 2013). Whatever the specific functional role of the pulvinar, it seems to perform that role both across multiple, large-scale networks and in the context of multiple cognitive behaviors. Murray et al. (2017) previously demonstrated that a two-module cortical circuit (i.e., without the pulvinar) could support many of the cognitive behaviors that have thus far been associated with the pulvinar. So what precisely do such cortico-centric views of cognitive function miss by excluding the pulvinar (and other subcortical structures)?

Jaramillo et al. (2019), in the present issue, investigated this question by combining a two-module cortical circuit with known cortico-pulvinar pathways (Figure 1): first, a transthalamic feedforward pathway (Figure 1A), indirectly connecting two cortical areas that are directly connected at the cortical level (referred to as cortical area 1 and 2). Their manipulations of this pathway suggest that the pulvinar provides cortical computations with critical flexibility. The total connectivity between cortical areas depends on contributions from both direct connectivity between the cortical areas and indirect connectivity through the pulvinar. The authors changed the total connectivity by adjusting a single variable, pulvinar excitability, proposing that this variable might be influenced by behavioral state (e.g., attended versus unattended). Changes in this variable had a dramatic effect on network computations. For example, the authors demonstrated that changing pulvinar excitability could resolve a conflict between stimulus-driven signals (such as luminance) and goal-directed signals (such as expected reward). Increasing pulvinar excitability increased the strength of feedforward signals, biasing responses toward stimulus-driven signals (or the choice within cortical area 1), while decreasing pulvinar excitability decreased the relative strength of feedforward signals, biasing responses to ward goal-directed signals (or the choice within cortical area 2). As a second example, the authors demonstrated that increasing pulvinar excitability could lead to spatially selective, persistent activity during a simple memory task. However, a particularly large excitability value in the pulvinar was also associated with the increased likelihood that a distractor would override such persistent activity. For both of these examples, manipulating pulvinar excitability greatly influenced computational outcomes and therefore behavioral outcomes.

Figure 1. Jaramillo et al. (2019) Modeled Two Cortico-Pulvinar Pathways.

(A) The first pathway is a transthalamic feedforward pathway that indirectly connects two cortical areas that are also directly connected. Here, we illustrate this first pathway in the context of higher-order cortical regions (i.e., LIP and FEF), which are interconnected with mdPul. (B) The second pathway is a feedback pathway, with one cortical area projecting both directly to pulvinar and indirectly to the pulvinar through the TRN. The pulvinar then projects back to the same cortical area. Here, we illustrate this second pathway in the context of extrastriate cortex (i.e., TEO), which is interconnected with vlPul. LIP, lateral intraparietal area; FEF, frontal eye fields; TRN, thalamic reticular nucleus; mdPul, mediodorsal pulvinar; vlPul, ventrolateral pulvinar; iPul, inferior pulvinar.

The second pulvino-cortical pathway modeled by Jaramillo et al. (2019) is a feedback pathway, projecting from a cortical area to the pulvinar through the reticular nucleus of the thalamus (TRN)— a sheet of inhibitory neurons surrounding the thalamus (Figure 1B). Here, their cortico-TRN-pulvinar circuit was able to reproduce, for example, the results of previous studies examining the neural basis of decision making in the pulvinar (Komura et al., 2013). Jaramillo et al. (2019) assumed that single neurons in the pulvinar integrate input from two excitatory populations in the same cortical area, with input from each cortical population engaging the pulvinar both directly and indirectly through the TRN. Their model revealed that, as the weighting of firing rates changed across the two cortical populations, the amplitude of the firing rate in an interconnected pulvinar neuron indicated the decision made during a given trial. That is, the pulvinar neuron signaled whether there was (1) a “correct decision” in the decision-making task, leading to a larger reward; (2) an error in that task, leading to no reward; or (3) a choice to opt out of the decision-making task, leading to a smaller reward (i.e., based on low confidence in the decision-making task). Jaramillo et al. (2019) thus demonstrate how behaviorally relevant computations might occur in the pulvinar itself (Komura et al., 2013).

In summary, Jaramillo et al. (2019) impressively show that a simple corticopulvinar circuit model can reproduce previously reported behavioral and physiological findings from investigations into multiple cognitive functions. The authors conclude that the pulvinar “provides crucial contextual modulation to cortical computations.” That is, adding the pulvinar to a two-module cortical circuit provides another layer of cortical control, which can both gate effective communication between cortical areas and alter cortical computations based on behavioral state. Based on their work, we begin to see how diverse cognitive behaviors that match the diversity of functions in brain regions interconnected with the pulvinar can be supported through specific cortico-pulvinar circuits.

As typical for groundbreaking studies, the one by Jaramillo et al. (2019) raises many intriguing ideas for future research. The present model highlighted the gating properties of the pulvinar by demonstrating that changes in “pulvinar excitability” re-weight the information that is represented by a cortical circuit. However, the mechanisms through which pulvinar excitability changes remain largely unknown. The authors suggest that regions interconnected with the pulvinar but not included in the present model, such as cholinergic inputs from the brainstem, might provide the tuner for pulvinar excitability. That is, these cholinergic inputs might help to regulate arousal and attention through their influence on the pulvinar.

A deeper understanding of pulvinar excitability might also be gained from further analysis of the connectivity patterns that create temporal signatures of interactions between the pulvinar and other brain regions. Attention-related pulvino-cortical interactions seem to be largely characterized by oscillatory activity in the alpha (Fiebelkorn et al., 2019; Saalmann et al., 2012; Zhou et al., 2016) and gamma frequency ranges (Zhou et al., 2016). Jaramillo et al. (2019) began to address these known temporal dynamics of pulvino-cortical interactions by incorporating laminar structure into their cortical modules, with deep layers generating alpha oscillations and superficial layers generating gamma oscillations. Following a simulated pulvinar lesion, their laminar model reproduced a couple of key findings from the previous literature (Zhou et al., 2016): an increase in local cortical alpha activity and a decrease in cortico-cortical gamma-band coherence. However, their results do not seem to explain the facilitative properties of pulvinar-driven alpha-band activity, both in cortex and in the pulvinar itself. Whereas increased alpha-band activity in sensory cortex is typically associated with the suppression of sensory processing (Fiebelkorn et al., 2019; Zhou et al., 2016), pulvinar-driven increases in alpha-band activity (e.g., pulvinar-driven increases in cortico-cortical alpha-band coherence) have instead been associated with enhanced sensory processing (Fiebelkorn et al., 2019; Saalmann et al., 2012). Future research will need to focus on deciphering the differences in the circuitry that underlies region-specific alpha oscillations, accounting for source-dependent differences in function. For example, pairing recordings from the pulvinar with laminar recordings from cortex may reveal layerspecific variations in functional connectivity during different behavioral states. Such multi-contact recordings (i.e., array recordings) in the pulvinar itself might also provide critical insight into its largely unknown micro-circuitry. For example, to what extent do neurons in the pulvinar, either within or between subdivisions (e.g., the mediodorsal and ventrolateral pulvinar), interact with each other? Future studies will thus need to refine and extend the computational roadmap laid out by Jaramillo et al. (2019), ultimately revealing the mysteries of one of the brain’s most puzzling structures.