Functional magnetic resonance imaging (fMRI) is one technique that enables noninvasive assessment of real-time changes in cortical activation patterns that can be applied in experimental settings. Functional MRI takes advantage of the paramagnetic properties of hemoglobin that vary with the degree of oxygenation (Ogawa et al., 1990). Thus, changes in blood flow in response to changes in neural activity can be detected and used to quantify and locate the range of regions involved in response to particular stimuli. The blood-oxygenation-level-dependent (BOLD) signal provided by fMRI is complex in its origin (Arthurs and Boniface, 2002) but reflects well the temporal and spatial aspects of regional or focal cortical activation patterns in response to particular stimuli. As such, this method complements invasive electrophysiologic and/or pharmacologic approaches that study single neurons or regions.
A commonly used approach to understand the functional impact of cortical activation patterns is to correlate these with patterns of change in blood pressure, heart rate, efferent sympathetic nerve activity or a stimulus signal such as strain gauge force, during maneuvers that elicit cardiovascular stress such as handgrip exercise or lower body negative pressure (to mimic orthostatic stress). Because of difficulties in measuring peripheral physiologic signals in the MRI scanner concurrent measures that reflect autonomic nervous system variables are limited to indirect analog measurements such as heart rate or skin conductance responses (Wong et al., 2011). To bring further meaning to cortical activation patterns studies are replicated in the fMRI session as well as a physiology laboratory session. In this context, reproducibility of cortical patterns associated with identical stimuli are required, and have been established in our hands (Kimmerly et al., 2004, 2005). Moreover, the regions associated with cardiovascular adjustments to stress are more-or-less replicated across many stressors from many laboratories. Specifically, many groups have examined the cortical patterns associated with cardiovascular responses (primarily heart rate) during physical (King et al., 1999; Critchley et al., 2000, 2003; Gianaros et al., 2004, 2005; Critchley, 2005; Kimmerly et al., 2005; Wong et al., 2007b; Goswami et al., 2011), cognitive (Gianaros et al., 2004, 2005; Critchley et al., 2000, 2003), and emotional (Lane et al., 2008) stressors. Overall, the regions outlined below appear to form the core cortical regions associated with cardiovascular control.
Cortical Circuitry Associated With Cardiovascular Arousal
The inaugural study of King et al. (1999) illustrated the utility of fMRI to expose the complexity of cortical activation that occurs during straining efforts that induced large cardiovascular responses. Using short but maximal handgrip exercise, Valsalva's maneuver and a maximal inspiration task, these authors observed increased cortical activation that was localized in the insular cortex, the posterior regions of the thalamus, and the medial prefrontal cortex (MPFC). Notably, increased MPFC activity during the recovery phase of Valsalva's maneuver occurred concurrently with a decline in heart rate. This point becomes important below when HR correlations with MPFC activation or deactivation begin to expose an important role for this frontal region in cardiovagal control. Regions of brain that were less activated compared with baseline were not explored in this early study. Also, each of the handgrip, maximal inspiration and Valsalva's maneuver segments elicited significant changes in blood pressure, heart rate, and volitional effort sense (straining). Thus, the cortical patterns were observed within the context of complex stimuli from muscle, baroreceptor and top-down neural sources.
To examine the forebrain architecture associated with baroreflex-mediated sympathetic activation in the absence of volitional effort or changes in blood pressure, a model of graded lower body suction was developed to simulate orthostasis while the participant remains supine (Kimmerly et al., 2005). With lower body suction it is possible to grade the orthostatic stress, and thereby the magnitude of baroreflex unloading, so that either a change in sympathetic nerve activity occurs without a change in heart rate (−15 mmHg) or both sympathetic activation and heart rate changes occur (such −35 mmHg suction). Cortical regions demonstrating increased activity that correlated with higher HR and greater levels of lower body suction included the right superior posterior insula, frontoparietal cortex and the left cerebellum. Conversely, using the identical statistical paradigm, bilateral anterior insular cortices, the right anterior cingulate, orbitofrontal cortex, amygdala, mediodorsal nucleus of the thalamus and midbrain showed decreased neural activation. These findings were replicated in additional studies using direct unloading of baroreceptors (Kimmerly et al., 2007a, b). Such locations also covary with changes in cardiovagal baroreflex sensitivity induced by psychological stress in humans (Gianaros et al., 2011). Further, the insula cortex has long been associated with baroreflex cardiovascular control in anesthetized rodents (Saleh and Connell, 1998; Zhang et al., 1998).
To study the forebrain regions and patterns involved with exercise, the cortical response to graded moderate intensity handgrip (HG) exercise was assessed (Wong et al., 2007a, b; Goswami et al., 2011). When performed at <40% of maximal contractile force for brief (e.g., <30 sec) in young adults, HG elicits an intensity-dependent tachycardia (in most individuals) that is apparent within the first 1–2 sec of the handgrip onset, growing to about 10–15 bpm increase over the 30-sec contraction duration (Fig. 1). Pharmacologic blockade evidence suggests that PNS withdrawal mechanistically controls the bulk of this rapid HR response (Hollander and Bouman, 1975; Fagraeus and Linnarsson, 1976; Mitchell et al., 1989). It follows that regions of the brain that change their activity in association with these rapid HR changes may reflect sites that modulate PNS. With HG, the motor cortex, bilateral insula, thalamus, cerebellum, and basal ganglia regions are all increased in their activation (Wong et al., 2007b). However, the ventral medial prefrontal cortex (vMPFC) was the only region to correlate strongly and inversely with heart rate changes with a time course and magnitude of change that reflected variations in exercise intensities. This patterned response was not affected by one's handedness or sex although females tend to produce smaller HR and cortical responses for the same relative workload (Wong et al., 2007a). Similar deactivation patterns within the medial prefrontal/genual ACC region that correlated inversely with HR were noted in the LBNP study above. Thus, handgrip maneuvers that appear to emphasize parasympathetic withdrawal elicit increased cortical responses in the posterior inferior bilateral insula activation and deactivation within the MPFC. This pattern has been observed in repeated studies (Wong et al., 2007a, b; Goswami et al., 2011). Furthermore, this strong relationship between MPFC and HR is consistent with other reports examining the brain-heart relationship during cognitive and emotional stressors (Critchley et al., 2000, 2003, 2004; Critchley, 2004, 2005; Gianaros et al., 2004, 2005; Lane et al., 2008).
Figure 1. Time course of changes in heart rate (HR; top panel), mean arterial pressure (MAP; middle panel) during a 30-sec of handgrip exercise performed at 40% of maximal voluntary contraction (MVC) strength.
Download figure to PowerPoint
Thus, the evidence from lower body suction and handgrip exercise models indicates that changes in HR are consistently associated with reduced activation within the vMPFC and subgenual ACC complex. These observations are consistent with extensive functional anatomical studies in lower animals which indicate the extensive projections from this region to brain stem cardiovascular centers (Verberne and Owens, 1998). While pharmacologic evidence discussed above suggests that PNS withdrawal mediates this rapid HR response, there may be a similar time course of sympathetic activation of visceral organs (Momen et al., 2005; Frances et al., 2008). However, conditions under which sympathetic activation may be responsible for the rapid HR change have not been studied. Of note, moderate intensity handgrip exercise also elicits increased activation within the supplementary motor area, basal ganglia, hippocampus, motor cortex, thalamus, and cerebellum. Of these, the cerebellum vermis (Bradley et al., 1991) and hippocampus (Ansakorpi et al., 2004; Castle et al., 2005) have been associated with autonomic cardiovascular modification. Otherwise, these regions have received minimal attention with regard to human cardiovascular control.
The above studies were performed under conditions of short-term and moderate-intensity handgrip to minimize concurrent blood pressure and sympathetic responses that occur with exercise. Nonetheless, these studies are still complicated by the complex neural patterns that occur at the onset of exercise. In particular, concurrent cardiovascular arousal, ventilatory activation, and motor control are coordinated at the exercise onset and are enveloped in the concept of “central command” (Krogh and Lindhard, 1913). The cortical activation patterns associated with central command in general have been studied in a series of complex studies by Williamson and colleagues using both SPECT and fMRI imaging (Williamson et al., 1996, 1999, 2002, 2003). These studies collectively demonstrate involvement of the insula in particular as an important region associated with volitional and effortful muscular exercise. These data raise the important question of the purposefulness of cortical activation during cardiovascular arousal. Particularly, to what extent does the insula–amygdala–medial prefrontal/anterior cingulate complex impact heart rate alone versus a broader coordination of physiologic responses required to initiate and sustain, effortful exercise?
In summary, the regions that consistently are associated with cardiovascular arousal include the dorsal anterior cingulate, medial prefrontal cortex/subgenual anterior cingulate, and insula cortex. Another site commonly observed is the amygdala. Sympathetic activation, studied in our hands only in the context of baroreflex unloading, is most consistently associated with the superior, posterior insula cortex (right side) and dorsal ACC. In our experience, HR responses always correlate negatively with the medial prefrontal cortex. From these temporal and spatial patterns, the collective evidence suggests that these regions reflect portions of a network that processes and supports the complex features associated with reflex cardiovascular control. Nonetheless, evidence supporting the concept of an integrated network is required.
Attempts to interpret the role of particular cortical regions based on BOLD responses must be considered carefully due to the limitations of this methodological approach. These challenges include the following: (1) uncertainty regarding the physiological basis of the BOLD response, (2) need for short but repeated stimuli to enhance signal-to-noise, (3) sensitivity to artifacts generated by movement, or tissue/air interfaces and global changes in brain blood volume (such as might occur with changes in blood pressure or ventilation patterns), and, (4) ambiguity in interpretation of the BOLD signal. Generally, our study designs aim to deal with such limitations by optimizing the levels of lower body negative pressure (LBNP) or handgrip to elicit important cardiovascular reflex adjustments with minimal movement or minimal changes in blood pressure and ventilation. Such practices are important to consider further. As mentioned, BOLD imaging offers advantages of relatively high temporal and spatial resolution, an advantage that makes this technique sensitive to changes in heart rate, blood pressure, and ventilation. This sensitivity to pulsatile or rhythmic physiologic events can interfere with detection of other signals that are unrelated to cardiovascular or ventilatory control. Thus, some neuroimaging specialists treat BOLD oscillations due to these autonomic variables as nuisance outcomes to be removed from the overall signal (Iacovella and Hasson, 2011). For example, increases in blood pressure can, in fact, enhance BOLD signal detection (Wang et al., 2006; Qiao et al., 2007). This effect likely includes somatosensory inputs from baroreceptors, an issue presented in detail below. In contrast, many other laboratories, such as those reflected in this review, consider these autonomic outcomes to be a theoretically meaningful component of the BOLD signal. Further, the sensitivity of cerebrovasculature to changes in blood oxygen and/or carbon dioxide create a concern for ventilatory patterns and their integration with task-specific brain activation patterns that may be specific to different brain regions (Hall et al., 2011).
The issue of interpretational ambiguity in BOLD responses (i.e., problem no. 4 above), requires special attention as well. Specifically, although relationships between cortical activation/deactivation patterns are exposed by correlations with heart rate or the stimulus time course, these correlations cannot convey causality or directional relationships. Two interpretational challenges arise from this latter limitation. First, by itself, the BOLD signal does not indicate whether the regions of activation form an integral network or discreet regions that are processing information directly or indirectly related to the stimulus. Second, a single activation pattern cannot inform the viewer regarding the inhibitory or excitatory nature of the brain's response or whether various regions are (a) “listening” to afferent sensory signals arising in the brain that reflect a cardiovascular change or (b) “talking” in the sense that they are directing an efferent motor response to adjust the cardiovascular status. To illustrate this point, the rapid increase in heart rate and blood pressure with handgrip exercise, or the elevated HR during lower body suction, will stimulate mechanosensor or baroreceptor afferents in the cardiac chambers, aortic arch and carotid sinus' whose neural pathways link into the nucleus tractus solitaries in the brain stem, with possible subsequent pathways that project through the thalamus to the insula, amygdala and other forebrain areas (Cechetto and Saper, 1990). Furthermore, sensory signals from muscle spindles arise during muscular work and these may cause a change in cortical activation; the cortical irradiation of such muscle sensory signals in humans is not reported. Finally, even mild handgrip exercise requires cognitive engagement during volitional effort, an aspect of the task that will carry its own and perhaps variable cortical activation pattern. The problem then arises as to whether the activation patterns observed in the cortex during handgrip or other cognitive tasks, are causing (or at least modulating) the heart rate or blood pressure change or, rather, are representing that change in sensory input as a cortical activation that does not influence cardiovascular control. Stated more specifically, the reduced activation of the vMPFC during HG may reflect a sensory response to afferent neural signals arising from the heart or skeletal muscle that signal a change in cardiac function or muscle tension within the MPFC region. Thus, studies are being conducted to address sensory representation within cortical autonomic regions and the concurrent cardiovascular outcomes.