Cortical, thalamic, and amygdaloid projections of the rat anterior and posterior insular cortices were examined using the anterograde transport of biocytin. Granular and dysgranular posterior insular areas between bregma and 2 mm anterior to bregma projected to the gustatory thalamic nucleus. Granular cortex projected to the subjacent dysgranular cortex which in turn projected to the agranular (all layers) and granular cortices (layers I and VI). Both granular and dysgranular posterior areas projected heavily to the dysgranular anterior insular cortex. Agranular posterior insular cortex projected to medial mediodorsal nucleus, agranular anterior insular and infralimbic cortices as well as granular and dysgranular posterior insula. No projections to the amygdala were observed from posterior granular cortex, although dysgranular cortex projected to the lateral central nucleus, dorsolateral lateral nucleus, and posterior basolateral nucleus. Agranular projections were similar, although they included medial and lateral central nucleus and the ventral lateral nucleus. Dysgranular anterior insular cortex projected to lateral agranular frontal cortex and granular and dysgranular posterior insular regions. Agranular anterior insular cortex projected to the dysgranular anterior and prelimbic cortices. Anterior insuloamygdaloid projections targeted the rostral lateral and anterior basolateral nuclei with sparse projections to the rostral central nucleus. The data suggest that the anterior insula is an interface between the posterior insular cortex and motor cortex and is connected with motor-related amygdala regions. Amygdaloid projections from the posterior insular cortex appear to be organized in a feedforward parallel fashion targeting all levels of the intraamygdaloid connections linking the lateral, basolateral, and central nuclei.

The central component of the extended amygdala (CEA) comprises the central amygdaloid nucleus (Ce), the dorsal substantia innominata (SI), and the bed nucleus of the stria terminalis (BNST). Anatomical studies have suggested the presence of an intrinsic system of GABAergic neurons that not only connects homologous subareas of the Ce, SI, and BNST but that also acts as an interface between sensory afferents and brain stem-projecting neurons. CEA outputs, with a few exceptions, arise from separate populations of neurons, but all, including GABAergic neurons themselves, are heavily innervated by GABAergic terminals. GABAergic neurons may serve to integrate output activity of the CEA, though GABAergic neurons form a heterogeneous population whose differential intrinsic connections appear related to their peptide content. Afferents from the dysgranular insular cortex and lateral parabrachial complex preferentially innervate GABAergic neurons, suggesting these neurons may also integrate afferent activity. Afferents from the basolateral amygdala (BL) appear to innervate both output neurons and intrinsic GABAergic neurons. Evidence will be presented to show that BL afferents form synaptic complexes with cortical, GABAergic, and TH-immunoreactive terminal boutons on GABAergic dendritic spines. These complexes may be a key element in control of CEA output activity.

The projections of the cerebral cortex to the extended amygdala were studied in the rat using anterograde and retrograde tract-tracing techniques. Most cortical areas with strong projections to the extended amygdala preferentially targeted either the medial extended amygdala (including the medial amygdalar nucleus, ventromedial substantia innominata, and the medial part of the bed nucleus the stria terminalis) or the central extended amygdala (including the central amygdalar nucleus, dorsolateral substantia innominata, and the lateral part of the bed nucleus of the stria terminalis). Some cortical areas, however, had equal projections to both medial and central portions. The main areas projecting preferentially to the medial extended amygdala were the ventral subiculum, infralimbic cortex, ventral agranular insular area, and the rostral part of the ventrolateral entorhinal area. The main areas projecting preferentially to the central extended amygdala were the prefrontal cortex, viscerosensory and somatosensory portions of the insular cortex, and the amygdalopiriform transitional area. It is suggested that these cortical inputs may be important for cognitive, mnemonic, and affective aspects of emotional and motivated behavior.

It is well established that the basolateral amygdala is critically involved in the association between an unconditioned stimulus (US), such as a foot shock, and a conditioned stimulus (CS), such as a light, during classic fear conditioning. However, little is known about how the US (pain) inputs are relayed to the basolateral amygdala. The present studies were designed to define potential US pathways to the amygdala using lesion methods. Electrolytic lesions before or after training were placed in caudal granular/dysgranular insular cortex (IC) alone or in conjunction with the posterior intralaminar nuclei of the thalamus (PoT/PIL), and the effects on fear conditioning were examined. Pretraining lesions of both IC and PoT/PIL, but not lesions of IC alone, blocked the acquisition of fear-potentiated startle. However, post-training combined lesions of IC and PoT/PIL did not prevent expression of conditioned fear. Given that previous studies have shown that lesions of PoT/PIL alone had no effect on acquisition of conditioned fear, these results suggest that two parallel cortical (insula-amygdala) and subcortical (PoT/PIL-amygdala) pathways are involved in relaying shock information to the basolateral amygdala during fear conditioning.

Although there is a close correspondence between fear and anxiety, and the study of fear in animals has been extremely valuable for understanding the neural basis of anxiety, it is also clear that a richer animal model of human anxiety disorders would include measures of both stimulus-specific fear and something less stimulus specific, more akin to anxiety. Patients with posttraumatic stress syndrome seem to show normal fear reactions but abnormal anxiety measured with the acoustic startle reflex. Studies in rats, also using the startle reflex, indicate that highly processed explicit cue information (lights, tones) activates the central nucleus of the amygdala, which projects to and modulates the acoustic startle pathway in the brain stem. Less explicit information, such as that produced by exposure to a threatening environment or by intraventricular administration of corticotropin-releasing hormone, may activate another part of the extended amygdala, the bed nucleus of the stria terminalis, which also projects to the startle pathway. Because this information may be less specific and of long duration, activation of the bed nucleus of the stria terminalis may mediate anxiety, whereas activation of the central nucleus of the amygdala may mediate stimulus-specific fear.