The amygdala also responds to positive affect, suggesting that the amygdala codes salience as a whole. Electrophysiological studies in monkeys indicate that amygdala neurons respond to rewarding, as well as aversive stimuli (Nishijo et al., 1988; Paton et al., 2006). The amygdala also participates in appetitive learning in which an object is given a value that is associated with the representation of a reward (review, Baxter and Murray, 2002 and O’Doherty et al., 2002). In this process, the amygdala is thought to maintain and rapidly update the value of an object, as its value changes (review, Baxter and Murray, 2002).

In primates, the amygdala also processes socially relevant stimuli (reviews, Adolphs, 2010; Amaral et al., 2003), a function especially important in species depending on a tight social order. Evidence for the amygdala’s role in social function was established by early temporal lobe resection studies documenting abnormal behavior including decreased fearfulness, hyperorality and alterations in sexual behavior in monkeys (Klüver and Bucy, 1939; Weiskrantz, 1956), which in some cases resulted in ostracization from the group, or a fall in social status (Dicks et al., 1969; Rosvold et al., 1954). More recently, selective fiber-sparing amygdala lesions have produced similar, albeit more complex results (Emery et al., 2001; Machado and Bachevalier, 2006). For instance, in dyadic relations‘, amygdalectomized monkeys are much more likely to approach other monkeys, resulting in increased grooming by their counterparts. When placed in larger social groups however, lesioned monkeys actually exhibit decreased social interactions and more avoidance behaviors compared to their conspecifics. These data suggest that amygdalectomized monkeys are actually capable of the full repertoire of social behaviors‘, but display an impaired ability to recognize the biological relevance of social stimuli‘, resulting in an inability to deploy behaviors in a contextually appropriate manner (reviews, Adolphs, 2010; Amaral et al., 2003 and Emery et al., 2001; Machado et al., 2008).

Given its role in social responses, it is not surprising that facial expressions, which convey critical social information, activate the amygdala. A subset of neurons in the monkey amygdala selectively responds to faces (Brothers and Ring, 1993; Gothard et al., 2007; Heit et al., 1988; Leonard et al., 1985; Rolls, 1984) and patients with amygdala damage are unable to recognize complex emotion in faces, particularly fear (Adolphs et al., 1994; Broks et al., 1998). While many early neuroimaging studies suggested that the human amygdala responds mainly to fearful faces (Breiter et al., 1996; Morris et al., 1996; Whalen et al., 1998), more recent data indicate that it responds to a range of facial expressions, including neutral faces (review, Adolphs, 2010 and Fitzgerald et al., 2006; Yang et al., 2002). Taken together, studies in humans and nonhuman primates support the idea that the amygdala may be more involved in detecting overall emotional salience than any one type of emotion.

Finally, amygdala dysfunction has been consistently implicated in the pathophysiology of human mood and anxiety disorders (e.g., Drevets, 2001; Phillips et al., 2003; Protopopescu et al., 2005; Schwartz et al., 2003; Stein et al., 2002). The hallmarks of these illnesses include social withdrawal, negative attentional bias (tendency to focus on negative versus positive environmental cues), and decreased ability to experience pleasure (Sadock and Sadock, 2007), suggesting that the amygdala’s ability to extract salience is biased or compromised. Evidence for the amygdala’s involvement in these disorders has been repeatedly documented in functional neuroimaging studies (review, Victor et al., 2010).

While the nomenclature and boundaries of amygdalar subregions have changed over time, today, most investigators use nomenclature initially proposed by Johnston (Johnston, 1923) and further modified by Crosby and Humphrey (Crosby and Humphrey, 1941; Humphrey, 1968) and Price and Amaral (Price et al., 1987) and DeOlmos (De Olmos, 2004; DeOlmos, 1990) (Fig. 1). The cytoarchitectonic features and differential nuclear staining that delineate these regions, are visualized by using classic stains such as cresyl violet and acetylcholinesterase (AChE), respectively.

The amygdala is divided into the ‘superficial’ amygdala and the ‘deep’ amygdala (Fig. 1). The superficial amygdala is considered ‘allocortex’, a primitive 2-3 layered cortex that is highly conserved phylogenetically (Vogt and Vogt, 1919). It includes the anterior and posterior cortical nuclei, the medial nucleus, and periamygdaloid cortex. The periamygdaloid cortex (or ‘prepiriform’ cortex) contains a molecular layer (layer I), a deep pyramidal layer (layer II), and occasional cell clusters forming a rudimentary layer III (review, Gloor, 1997a). The sulcal subdivision of the periamygdaloid cortex (PACs) surrounds the annular sulcus and is the least differentiated. Importantly, the PACs merges with the medial PL. This entire area has also been referred to as the ‘corticoamygdaloid transition zone’ (CTA) (Fig 1C, boxed area) (Rosene and Van Hoesen, 1987) other terminologies for this unique region include the ‘cortical nucleus’ (Johnston, 1923), and ‘amygdalopiriform area’ (DeOlmos, 1990). We use the ‘CTA’ designation because it focuses on the transitional nature of this zone (Johnston, 1923). Caudally, the PACs and PL merge into yet another transition region, the ‘amygdalohippocampal area (AHA) (Fig. 1D). The entire superficial amygdala is generally associated with olfactory function, however inputs from other sensory modalities project here as well (Carmichael et al., 1994; Turner et al., 1980).

The ‘deep’ nuclei include the basal, accessory basal, and lateral nuclei. Unlike other amygdaloid regions, this entire group of structures, sometimes referred to as the ‘basolateral nuclear group’, is progressive across mammalian species, reaching its relatively greatest size in primates (Stephan and Andy, 1977). These nuclei are phylogenetically recent adaptations, and their development parallels the massive expansion of the temporal and frontal cortices from which they receive their main inputs (Jerison, 1973).

The lateral nucleus is the largest nucleus, reaching its greatest extent in the human (Stephan and Andy, 1977). It is the main receiving site of the amygdala, and is the main site where sensory information from visual and polysensory association cortices of the temporal lobe project (Aggleton et al., 1980; Amaral and Insausti, 1992; Stefanacci and Amaral, 2000; Turner et al., 1980). These projections are reciprocated but also channeled into the basal and accessory basal nuclei, following a topography based on sensory modality (Pitkanen and Amaral, 1998). The basal nucleus is also the recipient of additional extrinsic projections, many of which it returns. The medial and orbital prefrontal cortex, the insula, and the entorhinal cortex densely project to the basal nucleus (Amaral and Insausti, 1992; Carmichael and Price, 1995; Ghashghaei and Barbas, 2002; Van Hoesen et al., 1981). Additional inputs from CA1 and subicular regions of the hippocampus also terminate in the basal nucleus, but interestingly, specifically target the ventral basal nucleus including the PL in the primate (Rosene and Van Hoesen, 1977; Saunders et al., 1988). In addition, the thalamus also projects to these regions (Carmichael and Price, 1996; Ghashghaei and Barbas, 2002; Mehler, 1980; Van Hoesen et al., 1981). The basal nucleus is comprised of magnocellular (Bmc), intermediate (Bi) and parvicellular (Bpc) subregions, oriented dorsally to ventrally (Amaral and Bassett, 1989; Braak and Braak, 1983). The dorsal Bmc has the largest cells and the strongest AChE staining, while the ventral Bpc has the smallest cells and the lightest AChE staining. The PL, as discussed in detail later, is closely associated with the basal nucleus in primates and has historically been considered to be part of this nucleus by some authors (Barbas and de Olmos, 1990; Jimenez-Castellanos, 1949).

The accessory basal nucleus (AB), located medial to the basal nucleus, receives many of the same intrinsic and extrinsic connections as the basal nucleus, (Freese and Amaral, 2009). It is divided into magnocellular (ABmc), parvicellular (ABpc) and sulcal divisions (ABs), which have some important differences anatomically and connectionally. One notable distinction in primates is that the ABpc receives hippocampal inputs, whereas the ABmc does not (Saunders et al., 1988). The ABmc, along with the basal nucleus, form the bulk of the amygdalostriatal pathway; the parvicellular subdivision of the accessory basal nucleus has very sparse inputs to the striatum (Fudge and Haber, 2002).

The central nucleus (CeN) is the main recipient of inputs from both the ‘superficial’ and ‘deep’ nuclei (Fudge and Tucker, 2009) and is a distinct structure composed of ‘striatal-like’ medium spiny neurons (Cassell et al., 1999; McDonald, 1982a). The central nucleus forms a continuum with the bed nucleus of the stria terminalis (BNST), a macrostructure known as the ‘extended amygdala’ (Alheid and Heimer, 1988; review, Heimer et al., 1999). Despite its location outside the amygdala (it resides caudal to the ventral striatum), the BNST is almost identical to the CeN cytoarchitecturally and histochemically. The CeN and BNST are linked by an almost continuous chain of cell islands that span the basal forebrain, hence, the designation ‘extended amygdala’. The full scope of input and output paths of the extended amygdala have been determined predominantly in rodent species (Alheid, 2003; Reynolds and Zahm, 2005), with relatively few studies in the primate.

The amygdala contains several other regions that are not easily grouped including, the anterior amygdaloid area (AAA), the amygdalo-striatal area (AStr), and the intercalated cell islands (IC). The ICs are a unique group of cells that have recently received attention because they inhibit amygdala output (Amano et al., 2010; Likhtik et al., 2008). The ICs are small heterogeneous cell clusters embedded in fiber tracts located between major amygdaloid nuclei. They are found in the rat and the primate, including the human. The morphology and chemical profile of the ICs indicate that they contain GABAergic, inhibitory neurons (Pitkanen and Amaral, 1994). Pertinent to this review, similarities between the IC and the PL have been reported (Amaral and Bassett, 1989; Braak and Braak, 1983; Krettek and Price, 1978; Millhouse, 1986), raising the idea that the PL may be either continuous with some ICs or developmentally linked to them. However, the low levels of GABA expression in the PL stand in contrast to the ICs, and are an important discontinuity, as we shall see.

A lack of consensus over its nomenclature has made the PL an elusive subregion to study (see Table 1 for a summary of alternative terminologies). Importantly, the PL varies significantly from species to species, both in its anatomical position and its overall prominence within the amygdala. Such species variations and multiple nomenclatures may account for why researchers have largely ignored the overall structure and function of the PL (see Species variations of the PL below).

The PL was first identified by Crosby and Humphrey in the human amygdala (Crosby and Humphrey, 1941). They proposed that this region was a distinct subdivision of the basal nucleus because of its higher cell density compared to the overlying Bpc and described the PL region as the most “superficial” (ventromedial) portion of the basal nucleus, referring to it as the “pars medialis superficialis of the basal nucleus”. The term “pars medialis superficialis” was maintained in the Macaque (Lauer, 1945). In some studies, the PL was identified as a subdivision of the basal nucleus (Barbas and de Olmos, 1990; Jimenez-Castellanos, 1949) while in other studies the PL was recognized as separate nucleus, and identified with several different names, including the ‘amygdaleum profundum ventrale’ (Brockhaus, 1938), the ‘granular nucleus’ (Braak and Braak, 1983), and ultimately, the ‘paralaminar nucleus’ (Amaral and Price, 1984; De Olmos, 2004; Price, 1981; Price et al., 1987).

Golgi and lipofuscin staining studies, in addition to Nissl staining, have been used as a justification for identifying the PL as separate nucleus from the basal nucleus (Braak and Braak, 1983; Tosevski et al., 2002). Golgi staining, which involves silver chromate impregnation, stains a limited sample of cells but allows visualization of the cell soma, its axon, and dendritic arborization in their entirety. Lipofuscin staining identifies a by-product of lysosome degradation and can be used to identify different classes of neurons. In contrast, Nissl staining detects broad cellular populations through the use of alkaline dyes that preferentially bind negatively charged RNA and proteins. All of these techniques have shown that the PL and the adjacent parvicellular basal nucleus have a much more complex array of cell types compared to the remaining subdivisions of the basal nucleus. Using these techniques in human samples, Braak and Braak found a clear structural distinction between the PL (their ‘granular’ nucleus), the basal nucleus, and the ‘transitory’ zone between them. This finding was echoed by Tosevski, who concluded that the degree of cellular heterogeneity and fascicular dendritic arrangements in the human PL warranted its classification as its own nuclear region (Tosevski et al., 2002).

The PL is progressive, increasing in relative size from lower species such as rats and cats and tree shrews to higher species such as Macaques and humans (Stephan and Andy, 1977). In lower species such as the rat and cat, the PL appears as a thin band of cells and makes up only a small proportion of the amygdala. In the rat, the PL is so small that it only appears in a few rostral sections, which may explain why many rat atlases and studies do not identify it (Fig. 1 E, F). It also may explain why some authors identify this region as intercalated islands. For example, Millhouse identified the region encompassing the PL as the ‘largest of the intercalated islands’ (Millhouse, 1986). More recently, the islands of cells interposed between the lateral nucleus and the external capsule has been termed the ‘lateral paracapsular island‘ (Fuxe et al., 2003; Marowsky et al., 2005). In contrast to its small appearance in rodents, the PL in humans and non-human primates is quite extensive as it extends around the basal, and to a lesser extent around the lateral nuclei as a large sheet of cells. The increasing relative size of the PL across phylogeny is suggestive of a greater functional role in higher species.

In humans, the inferior GE is situated ventral to the developing amygdala and persists throughout most of development (Ulfig et al., 2000) (Fig. 3). The GE is a source of neuronal, astrocytic and oligodendrocytic cell types (Birling and Price, 1998; Culican et al., 1990; Schmechel and Rakic, 1979; Ulfig et al., 1998). As a proliferative site, the human GE contains a number of densely packed cells that stain for markers found in proliferating and migrating cells such as PCNA (Ulfig et al., 2000), and the polysialylated form of neural cell adhesion marker (PSA-NCAM), as has been noted in the human (Ulfig and Chan, 2004). Immunoreactivity for PSA-NCAM is particularly apparent in the 5^th gestational month and gradually diminishes after the 6-7^th gestational months in humans (Ulfig and Chan, 2004; Ulfig et al., 1998).

During the initial phases of amygdalar development in both monkey and human, columns of neuroblasts from the GE migrate along radial glial fibers to form a primordial amygdala (Kordower et al., 1992; Muller and O’Rahilly, 1990; Ulfig et al., 2003) (Fig. 3B). In the monkey, the amygdala grows steadily in size throughout the second trimester, with its nuclei becoming more differentiated, medially to laterally (Kordower et al., 1992). Cell proliferation, as detected by the thymidine analogue, bromodeoxyuracil continues until the end of gestation. Late in fetal development, the GE recedes from a massive neurogenetic zone to a thin band of cells that borders the ventral portion of the maturing amygdala. In humans, the PL begins to form during the 8^th and 9^th gestational month, growing at the site of the remnant GE (Ulfig et al., 2003). Consequently, in postnatal monkeys and humans the PL resides in the same location as the former GE.

With regards to its major afferent inputs, the primate PL receives converging projections from the lateral nucleus of the amygdala and from the hippocampus. Polysensory information is channeled via in the ventral subdivision of the lateral nucleus, the main intra-amygydaloid input to the PL (Pitkanen and Amaral, 1998). Hippocampal inputs arising from the rostral CA1 and subiculum are also an important input (deCampo et al., 2009; Rosene and Van Hoesen, 1977; Saunders et al., 1988). These latter projections terminate in a relatively discrete pattern over the PL and its contiguous transition zones, in contrast to the situation in rodents, where the hippocampus broadly targets several amygdaloid nuclei including the basal, lateral, and central nucleus (Canteras and Swanson, 1992).

In humans and monkeys, immature-appearing cells in the PL contain protein markers typical of immature neurons such as class III beta-tubulin (TUJ1), doublecortin (DCX), and polysialyated neural cell adhesion molecule (PSA-NCAM) (Bernier et al., 2002; Bernier and Parent, 1998; deCampo et al., 2009; Draberova et al., 1998; Fudge, 2004; Meyer et al., 2002; Seki and Arai, 1993; Yachnis et al., 2000; Zhang et al., 2009). In one study examining neural proliferation using bromodeoxyuracil (BrdU) labeling in the primate, some of these cells were noted to have recently divided (Bernier et al., 2000). While the idea of neurogenesis outside the subgranular zone of the hippocampus or subventricular zone is controversial, some evidence supports this idea (reviews, Gould, 2007 and Rakic, 2002). As noted above, much of the primate PL does indeed occupy the region directly apposed to the temporal horn of the lateral ventricle (Fig. 1C, D), and in fact, occupies the position of the former ganglionic eminence (Fig. 3). This anatomic position makes it plausible that proliferating cells arising in the caudal subventricular zone might migrate into the amygdala. An alternative possibility is that immature neurons in the adult PL arise much earlier, perhaps during fetal life, and persist in an immature form post-natally, similar to Cajal-Retzius cells (Marin-Padilla, 1998; Meyer et al., 1999). This latter idea was advanced early on by Johnston (Johnston, 1923), and then by Sanides, who noted an abundance of immature-appearing neurons in the human amygdala (Sanides and Sas, 1970).

In our own material DCX, class III beta-tubulin, and PSA-NCAM, are co-localized in immature-appearing neurons in adult animals (deCampo et al., 2009) (Fig. 5A-D), although they decline in senescence (unpublished observations). Immature-appearing neurons also strongly express bcl-2 protein (Fudge, 2004) and its activated transcription factor, phosphorylated cyclic AMP response element binding protein (pCREB) (Fig. 6). Bcl-2 is significant in young neurons because it influences their rate of neuronal differentiation, including the timing at which neuron-specific markers are expressed and the rate of neurite extension (Middleton et al., 1998; Suzuki and Tsutomi, 1998; Zhang et al., 1996).

The PL, with its immature neuronal component, is located at the crossroads of circuitry involved in contextual conditioning as mentioned earlier. Its main afferent inputs appear relatively specific, carrying information about temporal and spatial information from the past (via hippocampus) (Fig. 5F, G) that converges with polymodal sensory information occurring in the present (via the ventral lateral nucleus). This arrangement suggests that excitatory signals involved in the process of contextual learning or memory formation may shape the fate of immature neurons. In regions such as the dentate gyrus of the hippocampus, activity-dependent processes such as differentiation, axon and dendrite remodeling, and migration are mediated through activation of intracellular signaling pathways in immature neurons (reviews, Balu and Lucki, 2009; Kempermann et al., 2004).