The frontoparietal control system: a central role in mental health

Abstract

Recent findings suggest the existence of a frontoparietal control system consisting of flexible hubs that regulate distributed systems (e.g., visual, limbic, motor) according to current task goals. A growing number of studies are reporting alterations of this control system across a striking range of mental diseases. We suggest this may reflect a critical role for the control system in promoting and maintaining mental health. Specifically, we propose that this system implements feedback control to regulate symptoms as they arise (e.g., excessive anxiety reduced via regulation of amygdala), such that an intact control system is protective against a variety of mental illnesses. Consistent with this possibility, recent results indicate that several major mental illnesses involve altered brain-wide connectivity of the control system, likely altering its ability to regulate symptoms. These results suggest that this "immune system of the mind" may be an especially important target for future basic and clinical research.

Keywords: brain networks; cognitive control; executive functions; prefrontal cortex; psychiatric disease.

Figure 1. The control system and flexible hubs

A) Clustering applied to resting-state functional connectivity MRI identified large-scale neural systems (Thomas Yeo and others 2011). Components of the control system are co-active in a wide variety of task domains (i.e., the system is domain-general) (Duncan 2010; Fedorenko and others 2013), are sensitive to a variety of cognitive control demands (Niendam and others 2012), and this system is split here into three sub-systems. The frontoparietal sub-system is labeled in bold due to its centrality to adaptive task control (Dosenbach and others 2007; Cole, Reynolds, and others 2013), though all sub-systems are highly inter-connected and functionally related (Cole and Schneider 2007; Vincent and others 2008). B) Recent evidence suggests the core control system has highly global functional connectivity (Cole and others 2010; Power and others 2011) that updates systematically across tasks (Cole, Reynolds, and others 2013). Further, the control system inhibits the default-mode system when it is irrelevant to task performance (Shulman and others 1997; A.C. Chen and others 2013), and control system inhibition of the default-mode system is impaired in mental illness (Anticevic, Cole, and others 2012). C) A schematic of how temperature is regulated by a thermostat (a controller in a feedback control loop). D) Several biologically realistic computational models suggest a two-step process of cognitive control (Braver and Cohen 1999; O'Reilly and Frank 2006). The first is reward prediction by the basal ganglia selecting a goal representation via the control system. The second step involves goal maintenance with searching for sub-goals to accomplish the goal (matching the current state to the maintained goal state representation). This is similar to feedback control in other contexts (e.g., controlling temperature with a thermostat).

Figure 2. Control system capacity interacts with dysfunctions to regulate symptoms

Like the body’s immune system is protective against symptoms of bodily disease, the control system is postulated to be protective against symptoms of mental disease – likely via the flexible hub mechanisms described above. Theoretical probability distributions are shown to indicate any given individual’s likelihood of control system capacity (top) and the severity of a harmful dysfunction in any given mental process (left). The likely levels of experienced symptoms are indicated at different combinations of control system capacity and dysfunctionality. Treatment for each mental disease is postulated to be specific to that disease when harmful dysfunctions are reduced (left), but may be common across diseases when control system capacity is enhanced (top) due to the domain generality of the control system (Chein and Schneider 2005; Duncan 2010).

Figure 3. The control system is disrupted across mental diseases

A) ‘Primary control’ disorders are defined as involving neural dysfunction of the control system itself, such that control system capacity is more likely to be compromised in all individuals with such disorders. This predicts that primary control disorders are more difficult to treat due to the diseases’ disruption of natural health-promoting processes. B) In contrast, ‘secondary control’ disorders are characterized as those that are exacerbated by low control system capacity (potentially by chance), but whose root neural dysfunction does not directly affect control system capacity. It may be possible, however, that if/when symptoms arise the control system capacity is in turn compromised (as a secondary downstream effect). C) Resting-state functional connectivity (inter-region temporal correlations during rest) disruptions have been found with a key control system region – lateral prefrontal cortex (LPFC) – across a variety of mental diseases. LPFC's global connectivity (temporal correlations across all other regions) was altered in major depression (Zhang and others 2011), obsessive-compulsive disorder (Anticevic, Hu, and others 2013), and schizophrenia (Cole and others 2011). These alterations, as well as altered connectivity with a subcortical hub (mediodorsal thalamus) in bipolar disorder (Anticevic, Cole, and others 2013), are consistent with the flexible hub theory.

Figure 4. Direct and indirect feedback control

An anxiety disorder (e.g., social phobia) is illustrated as an example of two ways that the control system could regulate symptoms of anxiety. A) Direct feedback control is illustrated by direct monitoring and inhibition of a hyperactive amygdala via brain connectivity. Note that direct feedback control can also involve brain connectivity over multiple synapses (e.g., control system to orbitofrontal cortex to amygdala). B) Indirect feedback control is illustrated by direct monitoring and indirect inhibition of a hyperactive amygdala via changes to the environment. These changes are implemented via the motor system. Examples include 1) deep breathing, which changes sympathetic vs. parasympathetic balance and indirectly influences amygdala, and 2) moving to a room that is comforting, thereby removing any environmental stimuli contributing to the hyper-active amygdala. A combination of direct and indirect feedback control strategies is likely the most effective, though implementing multiple control strategies would require a highly effective control system.

Figure 5. Rapid instructed task learning, flexible hubs, and psychotherapy

A) Psychotherapy is illustrated as augmenting the control system’s search for ways to regulate symptoms. The therapist utilizes listening (with questions) to detect the nature of symptoms (and previous attempts to regulate them), which informs instructed strategies that are ultimately implemented by the control system (via rapid instructed task learning (Cole, Laurent, and others 2013)). A specific example of this therapeutic process may be exposure-response prevention with cognitive-behavioral therapy elements designed for obsessive-compulsive disorder (Abramowitz and Arch 2014). B) Potential neural population mechanisms within LPFC for rapid instructed task learning (Cole, Laurent, and others 2013). Verbal instructions activate primitive components (bottom) and their relations are built via local connectivity to build task procedures on the fly. Substantia nigra (SN), ventral tegmental area (VTA), and other basal ganglia (BG) (O'Reilly and Frank 2006) help update LPFC with new instructions (see Figure 1D). Task 1 is a typical cognitive laboratory task, involving an arbitrary stimulus-response association. Task 2 is an adaptive real-world strategy used in psychotherapy to reduce anxiety. Both tasks are illustrated using the same mechanisms. Note that the mechanisms are simplified for illustration (e.g., representations overlap across neural populations in LPFC) (Cole, Laurent, and others 2013; Rigotti and others 2013).