Rethinking the role of taste processing in insular cortex and forebrain circuits

Abstract

Over the years, many approaches towards studying the taste-responsive area of insular cortex have focused on how basic taste information is represented, and how lesions or silencing of this area impact taste-focused behaviors. Here, we review and highlight recent studies that imply that insular cortex does not contain a “primary” taste cortex in the traditional sense. Rather, taste is employed in concert with other internal and external sensory modalities by highly interconnected regions of insular cortex to guide ingestive decision-making, especially in context of estimating risk and reward. In rodent models, this may best be seen in context of foraging behaviors, which require flexibility and are dependent on learning and memory processes.

Introduction

Commensal rodents such as rats and mice possess an omnivorous diet, and are well adapted to live alongside humans. The taste abilities of these commonly-used laboratory species are a fairly good match for those of humans, both in terms of sensing and distinguishing among basic taste qualities, as well as in hedonically oriented consumption decisions (i.e. they avidly consume sweet-tasting foods while rejecting those tasting bitter). While homeostatic controls of ingestion often function via similar mechanisms and pathways in vertebrates, behavioral strategies for obtaining and consuming food are often very different. Recent research into the function of forebrain neural pathways for taste processing, including its cortical representation, increasingly indicate that taste circuits are best understood when considered in context of an animal’s natural feeding behavior [1–4]. Rodents, including rats and mice, spend a great proportion of their time searching for food in the wild. House mice (mus musculus), a short-lived yet highly adaptable species, are generally opportunistic feeders, eating small but frequent meals [5]. Foraging in this and other rodent species is an activity naturally counterbalanced with environmental dangers such as predation or toxicity of unfamiliar foods [6]. Successful foraging in a natural environment therefore requires flexibility in decision-making strategies [7]; for example, with increased experience, mice are able to change their navigation tactics to reach a distal odor source [8].

Other forms of feeding decision-making involve the sense of taste. When mice are faced with a decision concerning the nutrition/reward value of an available food source, they can correct for prior disadvantageous choices and display behavior analogous to regret [9]. Sampling foods in nature can be fraught with danger, as many potential food sources have varying levels of toxicity [10]. Rodents utilize a pair of strategies to deal with this problem: neophobia, and conditioned taste aversion (CTA). The former involves consuming only a small amount of a novel food upon first exposure, even one that is highly palatable. Once the food has proven safe to consume, intake increases to expected levels in subsequent exposures. By contrast, CTA occurs when the novel taste is paired with visceral malaise. Here, animals will display aversion to subsequent presentations of the taste.

Within the field of gustatory research, brain regions including cortex are often examined first and foremost for evidence of how primary tastes are represented, with the implicit idea that taste-evoked neural activity underlies some form of perceptual discrimination. Taste-responsive neurons found in the insular cortex (IC) are described as collectively comprising a “gustatory cortex”, analogous to other primary sensory cortices. However, the insular cortex is known to be an area of broad sensory and interoceptive representation, with evidence for a role in a large and diverse number of processes, including sensory processing, autonomic control, learning, anxiety and emotion [11]. In this review, we argue against the idea of a primary cortical area for taste perception or discrimination, highlighting evidence that gustatory processing is polymodal and non-topographical. We then discuss how recent research using circuit-specific techniques reveals that gustatory activity in cortex and its connected forebrain areas underlies a variety of behaviors that are all related to feeding strategies in rodents, consistent with the relative importance of feeding and foraging to rodents, including mice.

The taste representation in rodent insular cortex IC is multimodal and non-topographic

In rodents, IC is located on the lateral surface of the brain, spanning almost 4 mm from anterior to posterior in mice (Allen Brain Atlas; atlas.brain-map.org). Concomitant with its role as a complex, multimodal hub, it has a rich and diverse pattern of connectivity with other cortical and subcortical areas [12, 13]. This includes reciprocal connectivity with the gustatory thalamus (GT) and amygdala, especially its basolateral (BLA) and central (CeA) subnuclei (Figure 1). In a new, comprehensive viral tracing study quantitative differences in connectivity to major brain regions were found to exist along the anterior-posterior axis of IC [14]. While posterior and medial IC possess more connectivity with sensory-related cortical and subcortical areas, the anterior IC preferentially receives higher-order associative and motor input, and drives motivated behavior via output to the striatum.

Figure 1.

Schematic representation of connectivity between major forebrain gustatory areas in regards to taste processing and feeding. The parabrachial nucleus (PBN) in the pons, which receives gustatory and visceral input directly from the medulla, is also shown as it is the source (in rodents) of divergent taste projections to the gustatory thalamus (GT) and central nucleus of the amygdala (CeA), as well as a relatively small projection to insular cortex (IC). In turn, PBN neurons are modulated by descending feedback from CeA and IC, but not GT. IC receives major inputs from GT and the basolateral amygdala (BLA), and projects heavily back to those areas. CeA itself does not project to IC, but communicates with BLA. BLA-IC reciprocal connectivity is directly implicated in the formation and expression of conditioned taste aversions (CTA). Other studies suggest that cortico-amygdalar outputs drive motivated, feeding-related behaviors, likely through striatal projections.

According to the Allen Brain Atlas the gustatory representation in mouse IC is centered in the medial IC, but extends into both posterior and anterior IC. Results from a recent electrophysiological “mapping” study of IC indicate that each of four primary tastes (sweet, salt, acid and bitter) had about equal representation among cortical ensembles, which a majority of neurons responded to more than one stimulus [15]. No spatial clustering based on taste response was apparent either along 1.2 mm of the AP axis of IC, or within its dorsal-to-ventral subdivisions. This lack of somatotopy based on response has also been found in two-photon imaging studies with taste stimuli, food stimuli, or multimodal stimuli [2, 16–18]. Older studies noted that taste-responsive cortical cells often possessed responses to somatosensory or visceral stimuli [19]; such convergence can actually be considered a feature rather than exception of the taste system in general, especially centrally [20, 21]. Electrophysiological studies in rats show that individual IC neurons can respond to tastes, retronasal odors, or both modalities; this multimodality plays a key role in palatability and preference learning [22, 23]. In addition, a subset of taste-responsive cortical neurons can respond to somatosensory, odor, auditory or visual cues, even in untrained rats [24]. In the posterior IC, two-photon imaging shows individual neurons preferentially respond to aversive stimuli including tail shock, bitter taste stimuli, anxiety, thirst, or visceral malaise [25]. Over a third of sensory-responsive cells were polymodal, with the largest cross response (50%) in these to bitter taste and pain. Collectively, this prevalence of polymodality, and lack of spatial or topographic organization such as is found in visual or somatosensory cortex, argues that insular cortex does not contain a dedicated “gustatory cortex”. Rather, taste inputs are one type of sensory channel that converges with others in a type of association cortex.

Activity in several types of sensory cortices is characterized by “expansive sparseness”: even though there are far greater numbers of cortical neurons as compared to downstream regions, only a small fraction respond to sensory stimuli [26, 27]. For example, in piriform (olfactory) cortex less than 15% of neurons respond to a given odorant [28, 29]. A recent imaging study of central insular cortex in awake mice found that 24.1% of neurons responded to gustatory stimulation [16]. In posterior insular cortex, 49% of neurons responded to a high concentration of quinine, but only 10% responded to sucrose [25]. In terms of neural processing, sparseness is theorized to yield computational benefits [26], although more work is needed in insular cortex to determine the extent and contribution of this type of coding to gustatory processing.

Taste processing in the IC is related to feeding decisions

Early studies showed that decerebrate rats possess relatively normal taste reactivity reflexes in response to basic taste stimuli [30]. Similarly, the balance of evidence over years of studies combining lesions of forebrain taste areas (including cortex) with behavior found only minor-to-negligible effects on basic taste preferences or licking rates [31, 32]. When such effects have been found, they are invariably related to a deficit in learning [33]. On the other hand, electrophysiological studies with behaving rats demonstrated that gustatory responses can be modulated by anticipation of a stimulus, including expectation of taste that is signaled by an external cue [34]. Following cue-taste association learning, the proportion of taste-responsive neurons with cross-modal activity increases [24]. Moreover, silencing of medial IC disrupts learned cue-food associations in mice [35]. When mice are trained to associate contextual cues with delivery of a palatable food, they will overeat in the presence of these cues even when sated; inhibition of IC eliminates this effect [36]. A recent study extends the role of the IC in mediating food approach behaviors to that of a specific cortico-amygdalar circuit [37]. Selective inhibition of IC neurons that project to the CeA prevented the production of conditioned approach or avoidance responses to cues predicting an aversive or rewarding taste stimuli. This specific circuit appears to underlie flexible anticipatory behaviors to particular tastes, further supporting a role for IC in guiding food choice during feeding and foraging behaviors.

During foraging in the natural environment, rodents are aware of external threats such as predators, and will adjust their behavior accordingly to minimize those threats [6, 38, 39]. This includes the ability to switch rapidly from food-searching to defensive or escape behaviors. Optogenetic activation of the posterior IC in mice has been shown to elicit precisely these kinds of behaviors [25]. Cortical neurons that respond during bouts of fear behavior (freezing) exist in the same part of IC as neurons that respond to sweet or bitter taste, and as mentioned above, a combination of these stimuli may activate the same cells. The IC-to-CeA circuit appears to be critical for this behavioral flexibility during foraging in response to threat; when activated, a mouse will immediately cease feeding or drinking, and demonstrate fear behavior [25, 37].

Other recent evidence suggests that parts of the IC play a role in homeostatic control of feeding or drinking relative to environmental cues. IC silencing had no effect on normal home cage or ad lib feeding, but impaired an association between a visual cue and palatable liquid food in mice [2]. Two-photon imaging revealed that a large number of IC neurons responded to the food cue, or during feeding (licking), and a subset responded to both cue and feeding. These responses were significantly modulated by the hunger of the animal, with neurons much more likely to respond to the food cue in a food-deprived state. These hunger-related effects could be mimicked by activating AgRP neurons in the hypothalamus, which likely influence IC via an indirect pathway through the basolateral amygdala (BLA). Moreover, there is evidence that the cue-stimulus response in IC is also modulated by thirst [18].

Taste learning involves specific insular-to-subcortical circuits

Earlier lesion studies indicated a role for both IC and amygdala, among other brain areas, in conditioned taste aversion learning [40]. Lesion studies utilizing high-resolution mapping have designated a zone in posterior IC (that includes the posterior-most gustatory area) critical for normal CTA expression [30, 41]. Recent papers have focused on how a key IC-BLA circuit functions in CTA. Following CTA, two-photon imaging in mouse IC revealed that both the number of responsive BLA-projection neurons, as well as response magnitude, increased to the conditioned stimulus (Saccharin or NaCl), suggesting a recruitment of this pathway following learning [42]. Activating the IC-BLA circuit during presentation of a novel taste, without the unconditioned stimulus (i.e., visceral malaise), still resulted in a CTA [43]. This circuit was not involved in neophobia, nor was it involved in non-feeding related forms of learning, such as trace fear conditioning.

Despite a robust input from gustatory thalamus (GT) into taste-responsive areas within IC, it is interesting that the thalamo-cortical and cortico-thalamic circuitry is seemingly not involved in either cued taste/food learning or CTA [32]. Traditionally, the GT was considered to primarily function as a cortical relay, without much in the way of processing. Input from GT contributes to the responses of cortical neurons that are activated specifically by tastes but is not required for cued taste responses [44]. On the other hand, activity in the GT itself is modulated by expectation, likely dependent on descending input from IC [45]. Rats with pharmacological inhibition of GT show impaired neophobia; As the IC is also implicated in neophobia, the necessity of an intact thalamocortical circuit for a normal response is implied [46]. Other feeding-related behaviors may involve this pathway, such as anticipatory contrast, where an animal must make a relative judgement (based on taste) between foods that vary in nutritional or reward value [47]. Although thalamic and amgydalar circuits appear to differ in function with regard to feeding, it is important to note that the GT and BLA terminal fields in IC overlap, and inputs of either type converge onto some cortical neurons [48], strongly implying the necessity of both circuits for normal feeding behaviors.

Conclusion

Recent research into the function of forebrain taste processing makes it increasingly clear that taste-activated neurons in cortex, including those that are part of cortico-amygdalar circuits, play a role not only in sensory discrimination per se, but in feeding- and foraging-related processes that are especially important to the rodent species serving as models. These processes include the implementation of flexible food-seeking strategies that depend on both external and interoceptive cues. Adding credence to this idea is that IC is organized as a multi-modal processing area without topographic taste representation. In addition to its connectivity with amygdaloid nuclei, IC is interconnected with many other brain regions; the role of these other circuits in feeding behaviors still await discovery and clarification.