Respiratory Physiology & Neurobiology 167 (2009) 72���86 Contents lists available at ScienceDirect Respiratory Physiology & Neurobiology j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / r e s p h y s i o l Cortical and subcortical central neural pathways in respiratory sensations Paul W. Davenport ��� , Andrea Vovk Department of Physiological Sciences, University of Florida, Gainesville, FL 32610, USA a r t i c l e i n f o Article history: Accepted 1 October 2008 Keywords: Central neural pathway Respiratory sensations Cerebral cortex Respiratory afferents a b s t r a c t Respiratory sensations motivate humans to behaviorally modulate their breathing and are the sensory urge component of the respiratory motivation-to-action neural system. Human and animal studies have provided evidence for the neural substrate for afferents in the respiratory tract and muscles to project to the cerebral cortex. Respiratory afferents continually transduce breathing pattern into a sensory neu- ral code. This neural code is transmitted to a subcortical gating area. Respiratory sensory information is then transmitted by respiratory modality specific convergent and divergent subcortical pathways to the cerebral cortex. There are two primary cortical pathways: (1) the discriminative pathway related to respiratory proprioception and (2) the affective pathway related to the qualitative assessment of breath- ing. Respiratory sensory information is processed by the discriminatory somatosensory-motor cortex and the affective mesocortex resulting in conscious awareness of breathing that can lead to distressing res- piratory sensations. The significance of respiratory sensory information processing is the fundamental interoceptive perception of ventilatory status. �� 2008 Elsevier B.V. All rights reserved. 1. Introduction Respiratory sensations motivate humans to behaviorally mod- ulate their breathing and are the sensory urge component of the respiratory motivation-to-action neural system (Bradley, 2000 Davenport, 2008). Conscious awareness of disrupted breathing motivates patients to seek medical care, severe sensations of res- piratory dysfunction produce significant morbidity in patients. The different types of respiratory sensations are due to different patterns of stimulation of afferent mediating respiratory sensory modalities, e.g. mechanical loads, bronchoconstriction, hyperin- flation and blood gas changes. When ventilation is obstructed, stimulated, challenged or attended to, cognitive awareness of breathing can occur. Respiratory sensations usually include the various forms of dyspnea (air hunger, chest tightness, effort of breathing) however, other respiratory sensations are common, such as an urge-to-cough, urge-to-sneeze, sense of suffocation and similar cognitive perceptions related to breathing. The effects of respiratory sensations range from a simple awareness of breathing to highly distressing fear and anxiety in both humans and animals. Respiratory sensations of sufficient magnitude can dominate cog- nitive awareness, hence there has to be a cortical and sub-cortical ��� Corresponding author at: Department of Physiological Sciences, Box 100144, HSC, University of Florida, Gainesville, FL 32610, USA. Tel.: +1 352 392 2246x3825 fax: +1 352 392 5145. E-mail address: davenportp@vetmed.ufl.edu (P.W. Davenport). neural basis for perception of breathing. It follows that appropriate manipulation of these neural processes will provide insight into the mechanisms mediating the specific forms of respiratory sensations, including dyspnea. It has been proposed that respiratory sensations are the result of neural gating into the cerebral cortex of respira- tory afferent input eliciting a somatosensory cognitive awareness of breathing and an affective response. Respiratory sensations are the result of sensory activation of subcortical and cortical neural pathways. Some of these pathways are shared across respiratory modalities (convergent) while activation of some neural areas are modality specific (divergent). Convergent neural mechanisms pro- vide generalized respiratory sensation while divergent pathways provide sensation of modality specificity. Subsequent neural pro- cessing recruits neural centers in motor pathways to initiate the respiratory compensatory behavior component of the motivation- to-action neural network. Subcortical and cortical neural pathways mediate respira- tory afferent activation of central neural structures. Respiratory sensations are, thus, the result of two subcortical and cortical processes: (1) discriminative processing���awareness of the spatial, temporal and intensity components of the respiratory input (i.e. what is sensed), and (2) affective processing���evaluative and emo- tional components of the respiratory input (i.e. how it feels). Discriminative processing involves neural pathways resulting in somatosensory cortical activation. Affective processing includes the amygdala and associated structures such as the anterior cin- gulate and insular cortex. Modality specific activation of cortical neural processing depends on a change in neural activity that 1569-9048/$ ��� see front matter �� 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.resp.2008.10.001

P.W. Davenport, A. Vovk / Respiratory Physiology & Neurobiology 167 (2009) 72���86 73 gates-in modality specific information to the higher brain centers (Boutros et al., 1991, 1995 Boutros and Belger, 1999 Arnfred et al., 2001 Grunwald et al., 2003 Kisley et al., 2004). This activation leads to cognitive awareness of the modality. The significance of gating-in and gating-out sensory modalities is the need to attend or ignore essential physiological functions. Changing the status of the respiratory system alters the associated sensory inputs, respiratory information can then be gated-in, eliciting a cognitive awareness of breathing which can be distressing, i.e. dyspnea. The transition between breathing that does not reach consciousness to cogni- tive awareness of breathing suggests neural information processing similar to non-respiratory sensory modalities. Sensory gating has been demonstrated with inspiratory loads (Davenport et al., 2007 Chan and Davenport, in press) and auditory, visual and somatosen- sory systems (Boutros et al., 1991, 1995 Boutros and Belger, 1999 Arnfred et al., 2001 Grunwald et al., 2003 Kisley et al., 2004). The neural gating element is likely the thalamus since it has been implicated in gating sensory information from auditory, visual and somatosensory modalities. The significance of respiratory sensory information processing is the fundamental interoceptive percep- tion of ventilatory status. Self-monitoring of ventilatory state is critical to respiratory disease management and morbidity. In order to investigate respiratory sensations, several things must be considered: the modality mediating the sensation, threshold, mag- nitude of stimulation, neural processing mechanisms, and motor outcomes/compensations. In the following sections, the human neural areas activated by different respiratory sensation eliciting modalities are summarized. The following section presents human sensory electrophysiological studies that determined the tempo- ral relationship between respiratory mechanical stimuli and brain neural activity. Electrophysiological and anatomical animal stud- ies of brain areas mediating cortical and subcortical respiratory sensory neural pathways are then summarized. These human and animal results are used to develop a model of respiratory sensory gating and an integrated respiratory neural network mediating res- piratory cortical and subcortical pathways. 2. Cerebral cortical processing of respiratory information���human brain imaging Brain imaging tools, such as functional MRI and PET, have pro- vided valuable insights into the cortical and sub-cortical neural mechanisms that may be involved in processing respiratory sensory information in humans. A drawback of brain imaging tools is that they cannot distinguish between the structures that are involved in discriminative and affective processing, and motor behavioral responses. Thus, the brain activation maps generated represents the entire neural network activated by a modality specific respi- ratory perturbation, which could include sensory, cognitive and motor areas as well as areas activated by non-respiratory stimuli inherent of the experimental procedure. The respiratory methods and modalities used in past brain imag- ing studies, such as hypercapnia, inspiratory and expiratory loads, and restricted tidal volume below resting levels (low VT), pro- duce different respiratory sensory experiences (Binks et al., 2002 Lansing et al., 2000) and behavioral responses. This is a reflection of the cortical and sub-cortical neural mechanisms activated. The neural mechanisms recruited depends upon the type or combina- tion of afferents activated (i.e. mechanoreceptors, chemoreceptors and nociceptors in the chest wall, lung, airways and blood vessels), the magnitude of the respiratory stimulus (i.e. supra-threshold, intensity), the affective nature of the respiratory stimulus (i.e. unpleasant, neutral), and the quality of the respiratory sensa- tion (i.e. chest tightness, urge to breathe). Thus, each respiratory modality will generate a unique brain activation map, but may share similar activated neural substrates with other respiratory modalities. Comparing activation maps generated from different respiratory modalities with a focus on the convergent and divergent patterns of activation provides a unique insight into the mecha- nisms by which respiratory sensory information is processed in humans. Recent brain imaging studies in respiratory sensation have focused on the function of a specific group of limbic structures in affective processing (Banzett et al., 2000 Evans et al., 2002 Peiffer et al., 2001, 2008 von Leupoldt et al., 2008). These structures are highly integrated with sensory, motor and cognitive areas. They are constantly sharing and comparing incoming and outgoing neu- ral information. Respiratory sensory processing involves a dynamic modulation of each structure, by each structure, within the neural network. Therefore, it is important to consider the entire activated neural network in respiratory sensation studies. In this section we review the respiratory modality specific and shared areas of activation, and report the patterns of activation of cortical and sub- cortical structures above the medulla in humans. These structures are the neural substrates for respiratory sensory processing. 2.1. Telencephalon 2.1.1. Frontal lobe Discrete areas within the frontal lobes are activated with inspira- tory (left inferior frontal gyrus) (Macey et al., 2006) and expiratory (right lateral frontal) loads (Macey et al., 2003) as well as areas that are common to both loads (frontal operculum and basal forebrain) (Peiffer et al., 2001). Hypercapnic activations are restricted to the middle frontal gyrus (Brannan et al., 2001 Liotti et al., 2001) but shares activations in the prefrontal cortex (Corfield et al., 1995a Macey et al., 2006). Low VT activates areas that are divergent from loads and hypercapnia, specifically the superior frontal gyrus and the pars opercularis (Evans et al., 2002), as well as convergent areas such as the middle and inferior frontal gyri and the frontal oper- culum (Banzett et al., 2000 Evans et al., 2002). The overlapping activation map between respiratory loads and low VT may be the result of stimulating common lung and chest wall mechnorecep- tors activated by changes in thoracic volume. The frontal lobe has long been associated with ���executive function���, and is implicated retaining memories associated with emotions derived from limbic input (Miller and Cohen, 2001). Thus, the frontal lobe may part of generating the perceptual experience to the respiratory modality. 2.1.2. Temporal lobe Respiratory loads and hypercapnia activate the fusiform gyrus (Corfield et al., 1995b Gozal et al., 1995, 1996 Brannan et al., 2001 Liotti et al., 2001 Peiffer et al., 2001 Macey et al., 2006), an area involved with face, word and numerical recognition (Davidoff and Landis, 1990). Inspiratory loads activate the left (Macey et al., 2006) and right (Peiffer et al., 2008) superior temporal gyrus, whereas hypercapnia activates the right middle temporal gyrus and the bilateral anterior temporal poles (Brannan et al., 2001 Liotti et al., 2001). The aforementioned areas are involved in audi- tion, speech, vision and word meaning while reading (Squire et al., 2004 Vandenberghe, 2007 Winer and Lee, 2007). Low VT does not activate the temporal lobe, but this may be the result of image subtraction between the experiment (low VT + air hunger) vs. con- trol (low VT + no air hunger) conditions (Banzett et al., 2000 Evans et al., 2002). Thus, activations in the temporal lobe may be the result of testing conditions (i.e. deciding on a breathlessness score) rather than respiratory afferent sensory processing and perception generation.

74 P.W. Davenport, A. Vovk / Respiratory Physiology & Neurobiology 167 (2009) 72���86 2.1.3. Occipital lobe Activations in the left cuneus with low VT (Evans et al., 2002) and in the lingual gyrus with both hypercapnia and low VT are reported (Brannan et al., 2001 Liotti et al., 2001 Evans et al., 2002). Respiratory loads, however, do not activate the occipital lobe. This difference may be related to the task of rating breathlessness. In all but two load studies, subjects did not rate their respiratory dis- comfort during the scanning session (Peiffer et al., 2001). Given that the occipital lobe processes visual information, and that the lingual gyrus in particular is related to word and numerical recognition (Borowsky et al., 2007), activation of this area with the hypercap- nic and low VT studies may be a result of rating breathlessness on visual scales. 2.1.4. Parietal lobe Respiratory loads, hypercapnia and low VT activate discrete regions within the posterior parietal lobe. Inspiratory loads activate the supramarginal gyrus (Macey et al., 2006), an area postulated to be involved with word meaning and cognition (Capek et al., 2004). Inspiratory and expiratory loads and hypercapnia activate the inferior parietal lobule (Gozal et al., 1995, 1996 Brannan et al., 2001 Liotti et al., 2001) which processes body-in-space information (Bonda et al., 1995). Low VT activates the adjacent intraparietal sulcus (Evans et al., 2002) which may be due to press- ing buttons on the rating box given that this area is activated with finger movements (Binkofski et al., 1998). Thus, activations in these parietal areas may not be directly involved with sensory processing of respiratory information but rather to the task of rat- ing. 2.1.5. Sensorimotor cortex Respiratory loads and low VT show similar activation maps within the sensorimotor cortex (Gozal et al., 1995, 1996 Fink et al., 1996 Banzett et al., 2000 Peiffer et al., 2001 Evans et al., 2002 Macey et al., 2003, 2006 von Leupoldt et al., 2008)). Both respi- ratory modalities activate the precentral gyrus (Gozal et al., 1995, 1996 Banzett et al., 2000 Peiffer et al., 2001 Evans et al., 2002 Macey et al., 2006). Association motor cortices, particularly the bilateral presupplementary (Evans et al., 2002) and the left medial supplementary motor areas (Banzett et al., 2000) are also activated with low VT. Respiratory loads activate the latter area (Fink et al., 1996 Peiffer et al., 2001 von Leupoldt et al., 2008)) in addition to the right lateral premotor cortex (Peiffer et al., 2008). Activation in these motor areas are likely involved with the ���top-down��� gen- eration of motor-behavioral responses rather than adding to the ���bottom-up��� sensory experience. The somatosensory cortex is activated bilaterally with inspira- tory loads (Fink et al., 1996 Macey et al., 2006) and low VT (Evans et al., 2002), with specific activations in the right secondary sen- sory cortex during inspiratory and expiratory loads only (Peiffer et al., 2001). Interestingly, hypercapnia does not activate the sensori- motor cortex (Corfield et al., 1995a,b Brannan et al., 2001 Liotti et al., 2001 Harper et al., 2005), which may suggest a divergence of the sensorimotor pathway between respiratory afferents activated by load and low VT (i.e. lung and chest wall mechanoreceptors) from those activated by hypercapnia (i.e. carotid and medullary chemoreceptors). It is noteworthy to mention that subjects can- not accurately localize or quantify the magnitude of hypercapnic perturbations as they can with inspiratory loads (unpublished observation), suggesting that discriminative processing may be governed by the somatosensory cortex. 2.1.6. Insular and cingulate cortices Activation in these areas has received the most attention in recent respiratory sensation studies because they play an impor- tant role in the unpleasantness of dyspnea (Peiffer et al., 2008 von Leupoldt and Dahme, 2005). Activations in the anterior and posterior insular cortex (Peiffer et al., 2001 Macey et al., 2003, 2006), and the anterior cingulate cortex are reported with loads (Macey et al., 2003, 2006), hypercapnia (Corfield et al., 1995a,b Brannan et al., 2001 Liotti et al., 2001 Harper et al., 2005) and low VT (Banzett et al., 2000 Evans et al., 2002), suggesting that the unpleasant sensations produced by respiratory challenges are processed in the same areas despite the fact that different respiratory modalities were implemented. Activations in the pos- terior and bilateral medial cingulate cortex with inspiratory loads (Macey et al., 2006) and hypercapnia (Corfield et al., 1995a,b) are reported. The insular and cingulate cortices purportedly ascribe emotional and cognitive attributes to sensory information (Evans et al., 2002 von Leupoldt and Dahme, 2005), thus they proba- bly play an important role in the motivation-to-action behavioral responses to respiratory discomfort. The insular and cingulate cor- tices are activated with other respiratory perceptual experiences not described here, such as the urge-to-cough (Mazzone et al., 2007) as well as non-respiratory experiences, such as pain (Treede et al., 1999), suggesting these structures may be part of a com- mon neural pathway for several respiratory and non-respiratory modalities. 2.1.7. Limbic system The structures comprising the limbic system contain rich inter- connections between the cerebral cortex, thalamus and brainstem and are important in integrating sensory discriminative experi- ences into affective perceptual responses (Papez, 1995). Respiratory loads, hypercapnia and low VT show both overlapping and dis- tinct areas of activity. Bilateral and unilateral activations in the left caudate during hypercapnia (Brannan et al., 2001 Harper et al., 2005) and low VT (Evans et al., 2002) may be related to the task of rating as this area functions in comprehension and articu- lation of words (Packard and Knowlton, 2002). Inspiratory loads produce bilateral activations in the caudate (Gozal et al., 1995) as well as activations specific to the right head of the caudate (Peiffer et al., 2008). The putamen, which contains rich intercon- nections with the thalamus, supplementary motor area and the sensorimotor cortex (Szabo, 1967), are activated with all respira- tory modalities (Gozal et al., 1995 Banzett et al., 2000 Brannan et al., 2001 Evans et al., 2002). Bilateral and unilateral activations in the right amygdala are reported with hypercapnia (Brannan et al., 2001) and low VT (Evans et al., 2002), however its activity dur- ing respiratory load perturbations remains equivocal. Inspiratory loads activate the amygdala (von Leupoldt et al., 2008) whereas expiratory loads decrease its activity (Gozal et al., 1996 Macey et al., 2003), possibly indicating that expiratory loads generate less affective responses than inspiratory loads, low VT and hypercap- nia. Alternatively, the intensity of the expiratory loads may not have been great enough to evoke an unpleasant sensation. The sublenticular portion of the internal capsule is a major route for neural connections between the thalamus and the somatosensory, visual and auditory cortices, and between the frontal lobe and the pons, thalamus and brainstem (Adams et al., 1997). The sublentic- ular as well as the hippocampus are activated during hypercapnic (Corfield et al., 1995a,b Adams et al., 1997 Liotti et al., 2001 Harper et al., 2005) and load perturbations (Gozal et al., 1995, 1996 Peiffer et al., 2001 Macey et al., 2003), but not with low VT. In addition, hypercapnia bilaterally activates the parahippocampal gyrus (Corfield et al., 1995a,b Brannan et al., 2001 Harper et al., 2005). These areas have been implicated in spatial learning and memory formation (Bird and Burgess, 2008), however, their pre- cise role in respiratory sensory information processing remains unclear.