The Amygdala as a Locus of Pathologic Misfolding in Neurodegenerative Diseases

Abstract

Over the course of most common neurodegenerative diseases the amygdala accumulates pathologically misfolded proteins. Misfolding of 1 protein in aged brains often is accompanied by the misfolding of other proteins, suggesting synergistic mechanisms. The multiplicity of pathogenic processes in human amygdalae has potentially important implications for the pathogenesis of Alzheimer disease, Lewy body diseases, chronic traumatic encephalopathy, primary age-related tauopathy, and hippocampal sclerosis, and for the biomarkers used to diagnose those diseases. Converging data indicate that the amygdala may represent a preferential locus for a pivotal transition from a relatively benign clinical condition to a more aggressive disease wherein multiple protein species are misfolded. Thus, understanding of amygdalar pathobiology may yield insights relevant to diagnoses and therapies; it is, however, a complex and imperfectly defined brain region. Here, we review aspects of amygdalar anatomy, connectivity, vasculature, and pathologic involvement in neurodegenerative diseases with supporting data from the University of Kentucky Alzheimer’s Disease Center autopsy cohort. Immunohistochemical staining of amygdalae for Aβ, Tau, α-synuclein, and TDP-43 highlight the often-coexisting pathologies. We suggest that the amygdala may represent an “incubator” for misfolded proteins and that it is possible that misfolded amygdalar protein species are yet to be discovered.

INTRODUCTION

Protein misfolding is a recurring theme in age-related neurodegenerative diseases: the brains of most elderly persons, with or without clinical dementia, contain >1 species of insoluble proteinaceous aggregates, often involving the amygdala (1–3). Here we review the literature on the pathobiology of the amygdala in common neurodegenerative diseases, with particular focus on protein aggregates. Excellent previously published articles have focused on related topics (4–13), including molecular mechanisms, which are not addressed in detail here. This review is oriented toward a broad readership interested in human brain pathology, and is organized as follows: (i) definition of key terms and conceptual overview; (ii) presentation of a hypothesis that has emerged from analyses of human studies, implicating pathologic synergies (as defined below) in the amygdala; (iii) review of the anatomy, connectivity, and vascularization of the amygdala; (iv) discussion of prior studies related to pathologic synergy in the amygdala, along with some relevant data from the University of Kentucky Alzheimer’s Disease Center (UK-ADC) autopsy cohort; (v) presentation of neuropathologic data on 6 representative persons’ amygdalae, including immunohistochemical stains for β-amyloid (Aβ), Tau, α-synuclein (α-SN), and TDP-43; and (vi) conclusions and possible future directions.

A primary assumption of this review article is that studies of autopsied human brains provide key contributions in the worldwide effort to advance the field of neurodegenerative disease research. At the very least, the exploration of human neuropathologies offers insights into phenomena that merit follow-up in other experimental contexts. Moreover, recent analyses of data from large autopsy series have enabled important conceptual breakthroughs. These studies (some described below) have revealed hitherto unexpected complexity and underscored that no other experimental system can fully recapitulate the unique milieu of the aged human brain. As such, research focusing on the neuropathologies of aged humans has exposed limitations of simplistic models while suggesting new opportunities for diagnostic and therapeutic strategies.

DEFINITIONS AND OVERVIEW

A “pathologic marker” refers to a microscopic structure interpretable to diagnose the presence of a disease, and possibly, as an indicator of disease severity (14). In neurodegenerative diseases, most pathologic markers are proteinaceous aggregates comprising peptide polymers and/or protein adducts (15). The proteinaceous aggregates of common neurodegenerative diseases include molecular fragments derived from the following genes (polypeptides): APP (Αβ), MAPT (Tau), SNCA (α-SN), and TARDBP (TAR-DNA binding protein 43 [TDP-43]). Over the course of neurodegenerative diseases, proteinaceous aggregates are deposited in predictable temporal and neuroanatomic patterns that may mirror the clinical features of disease. In other words, the presence and density of these pathologic markers correlate with decreased function in the affected regions (16, 17). However, the implications of each pathologic marker are context-specific and hence they are usually not pathognomonic. For example, a neurofibrillary tangle (NFT), composed of polymerized Tau protein, may be present in the brain as a result of infectious (18–20), neoplastic (21, 22), metabolic/developmental (23, 24), autoimmune (25), toxic (24, 26), or brain trauma-induced (27) conditions, in addition to the archetypal neurodegenerative diseases (such as Alzheimer disease [AD]), in which NFTs are seen (28, 29).

Nor are pathologic markers usually seen in isolation. In persons >80 years old, with or without frank dementia, it is the rule and not the exception for multiple pathologic markers to coexist in the same brain (2, 3, 30). One of the hypotheses underlying this review, which has been articulated previously, is that the pathologic aggregation of 1 protein can work synergistically to initiate or otherwise promote the aggregation of different protein species (7, 9, 31–34). This process is what we refer to with the term “pathologic synergy” (Fig. 1A). Pathologic synergy seems to be deleterious and is associated with accelerated cognitive impairment in dementing disorders (Fig. 1B, C) (12, 35–7).

FIGURE 1.

A combination of schematic illustration (A, B) and previously published data (C) help describe how pathologic synergies are associated with worsened clinical outcomes in neurodegenerative disease brains. (A) Cartoon depicts how one pathologic protein may interact with and promote the pathologic aggregation of another protein via mechanisms that are still not well understood; this is the process we are terming “pathologic synergy.” (B) The significance of pathologic synergies is that 1 misfolded protein can induce a conformational change in a separate protein species that may lead to altered (worsened) clinical course of the brain disease. (C) Data from the UK-ADC autopsy cohort provides support of this concept. The change of Mini-Mental State Exam (MMSE) scores over time, prior to death, is depicted using a statistical model of data from age-matched patients who died with Pure Alzheimer disease (AD) (n = 45) or AD with comorbid Lewy body disease, (AD + DLB, n = 20) pathologies, after having been followed longitudinally at the UK-ADC for >4 years before death. Note that the trajectory of cognitive impairment is more severe for AD + DLB patients (all with amygdala LBs) than Pure AD patients. The p value is derived from the “midpoint” parameter of the logistic model. Curves were generated using a 3-parameter logistic regression model (243) to allow between-cohort comparisons, as described in previously in detail (36).

The histomorphologic appearance of a pathologic marker may provide clues about its genesis. An example of a phenomenon that we interpret to indicate pathologic synergy is the neuritic amyloid plaque of AD, which contains both Αβ and Tau (Fig. 2A). More specifically, neuritic amyloid plaques are made up of extracellular Αβ deposits in close proximity with axons and dendrites that are both morphologically disfigured (38), and also packed with pathologic Tau filaments (39, 40). Because it has been demonstrated that axons from different extrinsic sources can show Tau-immunoreactive dystrophic features within a single neuritic amyloid plaque (41, 42), there are intuitive implications: apparently, a pathologic synergy existed in which some agent(s) in the extracellular plaque triggered neurochemical pathway(s) that promoted intracellular Tau misfolding. Details of this process remain incompletely understood; articles with proposed mechanisms are numerous (43–54), but a discussion of them is beyond the scope of this review. The physical nearness of misfolded Αβ and Tau within the plaque does not necessarily indicate that Αβ itself has a direct role in Tau pathology because many other potentially toxic molecules are also present in amyloid plaques (e.g. apolipoprotein E; Fig. 2B, C) (55–57). Nevertheless, even if Αβ in plaques is a proxy for other toxic agent(s), the evocative histomorphologic features of the neuritic plaque appear to support the conclusion of Barcikowska et al in regard to Tau pathology, i.e. it “may represent a nonspecific response … to different kinds of injuries, like the deposition of amyloid in Alzheimer disease” (58).

FIGURE 2.

Neuritic amyloid plaques have histomorphologic features that suggest the existence of pathologic synergy. (A) A photomicrograph of an AD brain section immunolabeled for both Αβ (brown chromogen; R1280) and Tau (black chromogen; Alz-50). The inset shows a false-colored rendering to highlight how the intracellular Tau (blue/black) pathology is entangled with the extracellular Αβ (yellow) of the plaque. Whereas the close proximity of Αβ and Tau in the neuritic amyloid plaques conveys the potential for pathologic synergies between extracellular toxic substance(s) and Tau pathology, it does not prove direct mechanistic interactions. (B, C) There are numerous complicating variables including the many biologically active molecules other than Αβ that are also found in neuritic amyloid plaques. For example, Apolipoprotein E is visualized here, stained immunohistochemically in a human amygdala (panel B is low-power, C is high-magnification from the same section). The appearance of Apolipoprotein E-immunopositive plaques (clustered near the purple arrow) in this amygdala (B) resembles the pattern demonstrated by Unger et al (139) using a Tau antibody to visualize neuritic amyloid plaques. Scale bars: A = 100 µm; B = 500 µm; C = 70 µm.

Other phenomena observed in aged brains suggest additional pathologic synergies. For example, investigators have documented the colocalization of α-SN and Tau aggregates (59–64), TDP-43 and Tau aggregates (64–67), and TDP-43 and α-SN aggregates (64, 65, 68) within the same cells, although in some brain regions, the colocalization of comorbid pathologies appears only infrequently (69–72). The development of a finite set of protein aggregates across different brain diseases indicates that diverse upstream factors can lead to common downstream reactions; these reactions may overlap with adaptive (neuroplasticity and/or inflammatory) pathways (73–76), but may ultimately evolve to become harmful (34, 77–79). The deleterious influences of any proteinaceous aggregate may include its propensity to cause other proteins to misfold. Further, since some proteinopathic processes are hypothesized to spread through the brain in a “prion-like” manner (10, 11, 80), a logical focal-point is how the prion-like agent(s) were introduced. In summary, pathologic synergies represent a potential common mechanism for the initiation and expansion of misfolded protein species in the aged human brain.

HYPOTHESIS: PATHOLOGIC SYNERGY IN THE AMYGDALA

Cross-sectional data from large autopsy cohorts indicate that neurodegenerative diseases develop in the brain over decades; these findings are increasingly confirmed via clinical biomarker studies (81–84). From these studies, evidence has also been gathered in support of the hypothesis that particularly important pathological synergies occur in the amygdala. Before more detailed discussion below, we consider this hypothesis in the context of the 2 most commonly observed neurodegenerative conditions: primary age-related tauopathy (PART) and AD (Fig. 3).

FIGURE 3.

(A–C) Cartoons depicting how pathology develops in the brain during the course of 2 of the most common known neurodegenerative conditions: primary age-related tauopathy (PART) (A) and Alzheimer disease (AD) (B, C). In the course of PART (i.e. lacking Αβ plaques) there is expanding distribution of Tau/NFT pathology, starting from the brainstem and evolving to exist also in medial temporal lobe structures including the amygdala. (B) In early AD, amyloid plaques occur first in neocortical regions and evolve in a rostral to caudal (toward the brainstem) direction. For reasons currently not well understood, the presence of Αβ amyloid plaques correlates with a significant transition such that Tau/NFT pathology occurs in the neocortex. (C) In advanced AD, both plaques and tangles are widely distributed throughout many regions of the CNS, and apparent secondary/tertiary misfolding (α-SN and TDP-43 pathologies) occur preferentially in the amygdala.

In the absence of comorbid Aβ, α-SN, or TDP-43 pathologies, Tau/NFT pathology is thought to develop during the course of aging in all humans, evolving in a brainstem-toward-cortex direction, being first seen in locus coeruleus, and later in medial temporal lobes (85, 86). This pathology underlies a disease that is now referred to as PART (87). The maximum severity of PART pathology is limited to Braak NFT Stage IV (87, 88), due to pathophysiologic factors that are not currently understood. Relatively advanced PART pathology (Braak NFT Stage III/IV) is associated with clinical subjective memory complaints (89) and mild cognitive impairment (90).

The presence of Aβ plaques constitutes a sine qua non for the expanded distribution of NFTs to merit the designation of isocortical Braak NFT stages (V/VI), the pathologic substrates for most AD-type dementia (30, 88, 91). Pathologic synergy that occurs in the amygdala between amyloid plaques and Tau/NFT may facilitate the transition to more severe clinical disease (Fig. 3). Further, in brains with severe AD pathology by consensus-based definition (92), both Aβ plaque and NFT pathologies are widely distributed, but the amygdala has a special propensity to develop apparent secondary misfolding of TDP-43 and/or α-SN (Fig. 3C).

These observations and hypotheses raise additional questions: Where exactly in the amygdala or in the peri-amygdaloid regions (e.g. anterior hippocampus, basal forebrain, entorhinal cortex, and ambient gyrus) does the pathology occur? How is the amygdala involved during the course of other diseases, such as Lewy body diseases (LBDs), TDP-43 diseases, and non-AD tauopathies? What features of the amygdala make it prone to be involved in pathologic synergies? Are Aβ, Tau, α-SN, and TDP-43 pathologies the only proteinaceous aggregates that occur in the amygdala? Before discussing these topics, we will first address a more fundamental question: what is the amygdala?

THE HUMAN AMYGDALA: ANATOMY, CONNECTIVITY, AND VASCULARIZATION

The amygdala is a centrally located brain region that plays fundamental roles in human emotion, memory, and various homeostatic responses (93–97). Situated in the rostral part of the temporal lobe and abutting the basal forebrain, the human amygdala is closely connected with olfactory structures, the hippocampal formation, basal ganglia, basal forebrain components including ventral striatum and nucleus basalis, insula, claustrum, hypothalamus, and various thalamic nuclei. The amygdala contributes to white matter tracts that stream nearby, including the anterior commissure, inferior longitudinal fasciculus, stria terminalis, and uncinate fasciculus. Excellent descriptions of the developmental, phylogenetic, neurobehavioral, and neuroimaging aspects of the amygdala are available (94–96, 98–105).

Studies of the amygdala attest to the complexity of this imperfectly defined anatomic region. Morphologic variability of the amygdala is high between primate species (106), and even among individual humans, possibly in association with their life experiences (107, 108). There also have been inconsistencies in published studies of the human amygdala. For example, some prior studies have reported that the human amygdala averages ∼1.1 cm³ in volume (96, 102), whereas other studies reported that the volume of human amygdala averages ∼1.7 cm³ (103, 109), or even larger (110). These divergent results are presumably due to substantial discrepancies in how its boundaries were defined.

Regardless of its peripheral boundaries, what we are referring to as “amygdala” comprises separate cell groups according to multiple criteria of distinction (Supplementary DataFig. S1). It has been suggested that a more appropriate term would be “amygdaloid nuclear complex” (96, 111), and, even more provocatively (because “the amygdala is neither a structural nor a functional unit”), it could be described as an “arbitrarily defined set of cell groups” (93, 111, 112). Extension of amygdala-like neurons into the basal forebrain (so-called “extended amygdala,” briefly described below) further complicates the idea of a single, well-defined anatomic structure (113).

The amygdala can be parsed into between 4 and 32 different subnuclei, as defined by various investigators (93, 95, 96, 100, 101, 114). Six regions are generally acknowledged: the cortical/transitional zone, medial nucleus, central nucleus, lateral nucleus, basal nucleus, and accessory basal nucleus. To economize on detail, 3 main regions of the amygdala, similar to de Olmos et al (94), are discussed here (Fig. 4). They are the cortical/transitional zone, the anteromedial area (including medial and central nuclei), and the basolateral complex (comprising lateral, basal, and accessory basal nuclei).

FIGURE 4.

The amygdala is an approximately 1.5-cubic-centimeter amalgamation of cell groups, with complex connections that include regions of the brain that are vulnerable to neurodegenerative disease pathologies. (A) A schematically rendered axial/horizontal oriented profile of the amygdala to depict some surrounding structures, as well as the 3 main regions of the amygdala: the basolateral complex (green), the anteriomedial area (red), and the medial cortex-like zones (yellow), roughly following de Olmos et al (94); this is not an accurate depiction because not all of these structures are in the exact same plane. (B) The medial cortex-like zone of the amygdala is the least well understood in terms of connectivity in human brains, but has strong reciprocal relationships with the hippocampal formation, entorhinal cortex, and olfactory structures. (C, D) Inputs (C) and outputs (D) of the basolateral complex and anteriomedial region are better understood from prior studies performed in rodents and nonhuman primates. Within the amygdala, information flows from the basolateral and accessory nuclei, toward the anteromedial nuclei, from which derives the main amygdalar output to the diencephalon and brainstem. (E) Panel depicts a right hemibrain to show some of the many brain regions that are directly connected with the amygdala and have been strongly implicated in the earliest phases of various neurodegenerative diseases. More information on amygdala connectivity may be found in (93, 95, 96, 100, 101, 114). Acc. Bas.n., accessory basal nucleus; ag, ambient gyrus; AnCing, anterior cingulate cortex; ARAS, ascending reticular activating system; Bas.n., basal nucleus; Bsl. Forbr., basal forebrain, including magnocellular neurons (nucleus basalis); BNST, bed nucleus of the stria terminalis; Cau, tail of the caudate; Cent.Med.n.g., centromedian nuclear group; Claus, claustrum; Co.n., cortical nucleus; Co.A.Tr., cortico-amygdaloid transition area; cs, collateral sulcus; DMThal, dorsomedial thalamus; d.m.V., dorsal motor nucleus of the vagus nerve; Ectx, Entorhinal cortex; ers, endorhinal sulcus; fg, fusiform gyrus; Ins, insula; Lat.n., lateral nucleus; LC, locus coeruleus; NbM, nucleus basalis of Meynert; PAG, periaqueductal grey matter; PFC, prefrontal cortex; phg, parahippocampal gyrus; sg, semilunar gyrus; Str., striatum/ventral extension of putamen; TEctx, transentorhinal cortex.

The cortical/transitional zone of the amygdala seems to be affected relatively early in the course of multiple diseases (115, 116). Anatomically it consists of a “superficial cortex-like region” (96) along the medial aspect of the amygdala as well as an ill-defined region near the entorhinal cortex, which is termed the cortical-amygdaloid transition zone. Physical connection between this region and the entorhinal cortex is extensive; according to Insausti and Amaral, the entorhinal cortex exists both anterior and posterior to the amygdala (117). The cortical/transition zone has been referred to as the “olfactory amygdala” (94), but with strong connections with the olfactory bulbs, hippocampal formation and parahippocampal cortices, it also serves memory and cognitive functions (94, 96, 100). There remains much to be learned about the medial region of the human amygdala, including the exact demarcation of borders. As stated by Yilmazer-Hanke, this area “is the most controversial amygdaloid region with regard to classification … and delineation of its sectors” (100).

Better characterized are the anteromedial subnuclei of the amygdala, sometimes referred to as the “centromedial nuclear group” (Supplementary DataFig. S1). Neurons in these subnuclei have strong connectivity with autonomic, visceral, and sensory input-related structures. Within the anteromedial region, the medial nuclei of the amygdala also receive strong olfactory input (93). The anteromedial nuclei are continuous structurally and functionally with the sublenticular region and the bed nucleus of the stria terminalis, collectively termed the “extended amygdala” (96, 118). Considered the amygdala’s main output source, the central (particularly) and medial nuclei relay integrated stimuli to autonomic effector regions in the hypothalamus, forebrain, and brainstem, and thus the anteromedial region has been described as a striatum-like “somatomotor” center (100).

By volume, the basolateral nuclear complex comprises much of what is collectively referred to as amygdala, particularly the regions of amygdala overlying the temporal/inferior horn of the lateral ventricle. The lateral nucleus is the largest subnucleus of the human amygdala and is usually asymmetric, i.e. with the right side larger than the left (96, 119). Neurons in the basolateral complex of the amygdala receive widespread input from “higher-order association” cortex and hippocampus, in addition to sensory cortex and thalamus. Notably strong bidirectional communication links the basolateral nuclear complex with the prefrontal cortex and the dorsomedial thalamus. Because of these connections, this area has been termed the amygdala’s “frontotemporal system” (93).

Synaptic connectivity within the amygdala has been described through studies of nonhuman species. Connections flow from the basolateral complex into the anteromedial nuclei, modulating the main outputs to the diencephalon and brainstem that help control bodily functions that may manifest as emotion (93, 95, 96, 100, 101, 114). These pathways incorporate multimodal stimuli, while helping to integrate the circuits of memory and emotion. Remarkably, many of the brain regions that project to and/or from the amygdala are highly prone to develop pathology in neurodegenerative diseases (Fig. 4E).

It currently is not known why specific brain areas are most vulnerable in neurodegenerative diseases. With regard to special anatomic features of the human amygdala, we highlight the brain’s internal (ependyma) and external (pia mater) limiting layers (Fig. 5). The tissues directly internal to these layers (subependymal and subpial compartments) often stain positively for misfolded proteins in neurodegenerative diseases of aging. For example, TDP-43 pathology has been observed in the subependymal and subpial regions in the periamygdaloid region of individuals with neuropsychiatric conditions (120) and subpial and subependymal Tau pathology are characteristic of aging-related tau astrogliopathy (ARTAG) (121). These observations suggest the possibilities that these compartments are susceptible to protein misfolding pathology, or alternatively, the tissue may be in direct contact with disease-stimulating agent(s). The area immediately ventromedial to the amygdala is noteworthy for the close proximity of the ependyma and the pia, separated only by a layer of gray matter (Fig. 5).

FIGURE 5.

Select features of the anatomy and histology of the human amygdala. (A) Low-power photomicrograph of medial human amygdala to visualize how the lateral ventricle (LV) and medial pia mater (Pia) are juxtaposed, separated only by a ∼2.5-mm-band of peri-amygdaloid cortical gray matter. Hematoxylin and eosin. (B) This is shown in higher power. (C, D) Inset regions from (B), depicting the ependyma (Ep) (C) and the pia mater (D). (E–G) In the brains of older persons, the regions of the superficial (subpial) medial amygdala tend to show degenerative features, including gliosis and corpora amylacea (E); corpora amylacea are indicated with green arrows. Immunohistochemistry with anti-phospho-TDP-43 antibody (red arrowheads) highlights the subpial TDP-43 pathology (F, G); the TDP-43 pathology is in close proximity to a corpus amylacea. Scale bars: A = 2 mm; B = 800 µm; C = 100 µm; D = 150 µm; E = 60 µm; F = 125 µm; G = 25 µm.

There is increasing awareness that the “vascular” and “neurodegenerative” pathogenic processes may interact (122–124); therefore, an overview of the amygdala’s blood supply is germane to this discussion. The amygdala’s blood supply is complex, deriving from multiple arterial territories (96, 125, 126), with substantial interindividual variation (125). Further, this is a region rich in vascular anastomoses (125). Primary arterial input to the amygdala derives from the middle cerebral artery (96), and/or the anterior choroidal artery (127), but the amygdala often also receives arterial branches from the internal carotid artery and/or the posterior cerebral artery (125, 126). Huther et al concluded that “… the anterior choroidal artery supplies the posteromedial part of the amygdala and … the middle cerebral artery irrigates the anterolateral part…” (125). TDP-43 pathology in the amygdala may appear in a peculiar distribution around small blood vessels, suggesting a potential link between vascular pathology and misfolded proteins (128). Di Marino et al emphasized that the amygdala’s vascular supply and drainage are aspects of human biology that merit additional study: “The vascularization of the amygdala is much less studied than the vascularization of the neighboring hippocampus…” (96). The need for further relevant studies is particularly pressing in the context of brain aging, in which small vessel pathologies tend to be widespread and are associated with cognitive impairment (123, 129–132).

NEURODEGENERATIVE DISEASE PATHOLOGIES IN THE HUMAN AMYGDALA

Whereas the amygdala harbors misfolded proteins relatively early in multiple neurodegenerative diseases, many questions remain. As summarized by McDonald and Mott: “There are discrepancies in different reports as to which amygdalar nuclei are most affected, which may be related to individual variability and/or stages of the disease” (133). The involvement of each brain region in each disease can be registered in terms of the staging systems that have been created through careful study of neurodegenerative diseases in large autopsy series. These staging systems are helpful for defining the expected distribution of pathology in specific diseases (87, 134), and for gauging disease severity. The particular stages of neurodegenerative diseases during which the amygdala is first affected are depicted in Table 1. Relatively few recent human studies have described in detail the distribution of pathologic markers within subnuclei of the amygdala, to provide a sense of expected interindividual and interdisease variability. Following those caveats, our review focuses on amygdalar pathologies in AD, in LBDs, in TDP-43 proteinopathies, and in rarer human brain conditions. We also present some primary data from the UK-ADC autopsy cohort based on methods published in prior studies.

TABLE 1.

Amygdala and Selected Anatomical Regions in Neurodegenerative Disease Pathology Staging Systems

Anatomic Region Reference	Classification Schemes [possible number of stages in each system]
	Braak NFT Stages [0–VI] (136)	Thal Aβ Phases [0–5] (241)	AGD [0–3] (242)	Braak PD Stages [0–6] (151)	Amygdala-Predominant LBD*(166)	TDP-43 in AD [0–6] (188)	TDP-43 in Aging [0–3] (195)
	Below are the stages each anatomic region is first affected in each staging scheme
Amygdala	I–II	2–3	1†	4	Early/Strong	1	1
Locus coeruleus	0–i	5	n/a	2	0/mild	n/a	n/a
Entorhinal cortex	I	2	2	5	n/a	2	2
Hippocampus	II	2	2	4	n/a	2–3	2
Neocortex-affected early	V	1	3	6	0/mild	6	3
Neocortex-affected late	VI	n/a	n/a	6	0	n/a	3

AGD, argyrophilic grain disease; PD, Parkinson disease; LBD, Lewy body disease; n/a, not applicable.

^*Amygdala-predominant LBD is not a stage-based scheme, and is currently considered “low-likelihood of dementia with Lewy bodies” (159).

^†Gyrus ambiens is immediately rostro-dorsal to the amygdala’s cortical/transition region with indistinct boundaries.

In terms of AD-type pathology, the amygdala is first affected by amyloid plaques in Thal Αβ stage 2 (135), and by NFTs in Braak NFT stage I–II (“a few isolated NFTs”) (136, 137). According to the seminal NFT staging paper of Braak and Braak, “…the corticomedial complex of the amygdala reveals the presence of many neuritic plaques, while NFT and neuropil threads predominate in the basolateral nuclei.” (136). Scott et al correspondingly highlighted atrophy in the basal nuclei in AD (138). Unger et al described a distinctive pattern of neuritic amyloid plaques early in the disease in the amygdala cortical transition zone (139), and Tsuchiya and Kosaka reported that the density of NFTs and cell loss was greatest in the corticomedial group in AD (140). Overall, it is generally agreed that neither plaques nor tangles are seen first in the amygdala but in late AD the amygdala harbors extensive Αβ amyloid and Tau pathology (141–148).

Additional insights can be gained if one considers the importance of pathologic synergies. In Figure 6, data from the UK-ADC autopsy cohort show that neocortical regions (dorsal prefrontal, inferior parietal, superior/mid-temporal neocortex, and occipital visual neocortex) are affected early in AD by neuritic amyloid plaques but not by NFTs. In sharp contrast, many hippocampal and entorhinal NFTs are present in the early stages of AD, whereas numerous neuritic amyloid plaques are seen later in those medial temporal lobe areas. The pathologic stages are oriented toward indicating where any pathology is recognized (Table 1). A more quantitative approach provides complementary information, in this case providing support for the hypothesis of early convergence of substantial densities of both neuritic amyloid plaques and NFTs in the amygdala (Fig. 6). This highlights a key transition-point of early AD at which some aspect of amyloid plaque biology correlates with (promotes?) a change from relatively benign PART to the more malignant, and possibly auto-propagating, widespread tauopathy of AD. Given the early concentration of numerous NFTs and neuritic amyloid plaques in the amygdala and its connectivity linking brainstem, allocortical, and neocortical structures, the amygdala is a credible anatomical location for this critical transition to occur.

FIGURE 6.

The amygdala (large red arrows) is distinguished by having relatively high numbers of both neuritic amyloid plaques and neurofibrillary tangles (NFTs), relatively early in the clinical course of Alzheimer disease. (A, B) Neuritic amyloid plaque counts (A) and NFT counts (B) from 7 different brain areas among 603 research subjects. Cases were stratified by final MMSE scores, which provide a gauge of global cognitive status. As expected, pathologic marker counts were higher in persons that experienced more severe antemortem cognitive impairment. Research subjects who had come to autopsy from the UK ADC cohort were the basis for the study, using the same methodology as described in detail in (244). The number of cases in each group of final MMSE scores: MMSE 26–30 (n = 47); MMSE 21–25 (n = 184); MMSE 16–20 (n = 63); MMSE 11–15 (n = 70); MMSE 6–10 (n = 53); and MMSE 0–5 (n = 186). Most of these individuals were last evaluated within 1 year of death. Information related to the UK-ADC cohort and neuropathology protocols, including lesion counting methods, were presented in detail elsewhere (2, 218, 244).

The amygdala also is the anatomical location where presumed secondary misfolding of α-SN and TDP-43 most often is observed in AD, particularly in advanced AD, when the amygdala appears to be a veritable “incubator” of protein misfolding. It has previously been shown that medial temporal lobe α-SN and TDP-43 pathologies with an epicenter in the amygdala are more likely to occur in advanced AD (Braak NFT stage VI) than in cases lacking, or with less severe, AD pathology (7, 65, 149, 150). Data from the UK-ADC autopsy series are shown in Figure 7. In this sample (n = 172), Braak NFT stage is associated with the combination of TDP-43 and LB pathologies (${\mathit{\chi }}_9^2$ p = 0.004; Fig. 7). Further examining the a priori hypothesis that subjects with both TDP-43 and LB are more likely to have advanced NFT pathology, we see that 61.1% of TDP-43-positive/LB-positive subjects were Braak NFT stage VI (61.1%) in comparison to 27.9% who were Braak NFT stage VI among those that lack either or both of TDP-43 and LB pathologies (p = 0.009). These data also underscore that a large percentage of brains harbor comorbid pathologies. Relatedly, both α-SN and TDP-43 pathologies occur in early-onset, autosomal dominant AD cases (see below), thereby providing added support for the hypothesis that these pathologies are downstream/secondary effects in brains with advanced AD pathology. However, LB pathology is also frequently seen in persons lacking advanced AD.

FIGURE 7.

In brains with Braak NFT Stage VI, there is a relatively high likelihood of having both TDP-43 and α-Synuclein (Lewy bodies, or LB) pathologies simultaneously in the medial temporal lobe (amygdala). We hypothesize that these data may indicate secondary (downstream of AD pathology) misfolding in advanced AD. Patients who had come to autopsy from UK ADC cohorts, and had hippocampal TDP-43 pathology and amygdala α-Synuclein neuropathologic assessment available (n = 172) were the basis for the study. For the statistical comparison of Braak NFT stage versus the combination of TDP-43 and LB pathologies, a 9 degree of freedom chi-square test with 10,000 Monte Carlo simulations conferred a p value of 0.004. In addition, a 2-sample test for proportions with a continuity correction was used to test the a priori hypothesis that the proportion of subjects with Braak stage VI NFTs was different between those with both TDP-43 and LB pathologies and individuals that lacked either or both TDP-43 and LB pathologies. The 2-sided test confers a p value of 0.009 suggesting a higher proportion of Stage VI in TDP-positive/LB-positive (61.1%) versus those that lacked TDP and/or LB pathology (27.9%). The number of cases in each group: TDP-negative/LB-negative (n = 93); TDP-negative/LB-positive (n = 17); TDP-positive/LB-negative (n = 44); TDP-positive/LB-positive (n = 18).

Prevalent LBDs include Parkinson disease (PD) and dementia with Lewy bodies (DLB). In a study of cases with a broad spectrum of PD severity, Braak et al reported that α-SN pathology develops in a predictable pattern with the dorsal motor nucleus of the vagus nerve affected first, followed by an expansion of the pathology through more brain regions in clinical PD (151, 152) (Table 1). Data from other high-quality autopsy series also support the hypothesis of a caudal-to-rostral directional spread of α-SN pathology in PD (153, 154). In the Braak PD staging scheme, the amygdala is affected in Stage 4 (out of 6) (137), and Braak et al also reported that the earliest areas affected in the amygdala are the central nucleus and the accessory cortical nucleus (155); Harding et al confirmed many early LBs in the cortical nucleus (156).

Yet there is appreciable interindividual variability in the anatomic distribution of α-SN pathology among LBD cases in comparison to AD cases (157–161). It may be that no single staging scheme could encompass all the persons who are affected by LBD while being predictive of the severity of each patient’s various clinical signs and symptoms. These findings partly indicate demographic differences in large autopsy cohorts. The reflections of Jellinger are also applicable: “It should be emphasized that interpretation of abnormal accumulations may be influenced by the conditions of staining, the antibodies used, of fixation” (153).

While cohort- and laboratory-specific technical factors are relevant, many clues implicate another potential source of variation in the anatomic distribution and clinical manifestations of α-SN pathology: pathologic synergies. For example, the presence of AD/Tau pathology seems to be associated with altered distribution of α-SN pathology. In a study of familial, early-onset AD brains, the amygdala showed α-SN pathology in >90% of patients with PSEN1 mutations and >70% of patients with PSEN2 mutations, whereas the dorsal motor nucleus of the vagus nerve usually was not affected (162). Further, studies from different centers have reported multiple brains where the amygdala was the area most strongly affected with α-SN pathology (160, 163–171), and still other cases where the olfactory bulb was the only brain area affected (172–174). A key observation is that α-SN aggregates tend to be present in diseases that also have Tau pathology (175). Further, misfolded Tau and α-SN were shown to colocalize in the same nerve cells, and those double-immunolabeled cells tended to be located in the limbic regions, particularly the amygdala (59, 176) and olfactory bulb (60). These findings were summarized by Schmidt et al, who studied Tau and α-SN inclusions in human brains: “…these 2 lesions frequently occurred together in the same neurons of the amygdala. These findings are in contrast in other sites that accumulate Lewy bodies and NFTs, but rarely both lesions in the same neuron” (176).

In some ways similar to α-SN pathology, TDP-43 pathology demonstrates tropism for the amygdala. TDP-43 pathology illustrates that the disease in which a pathological marker was discovered is not necessarily the most prevalent condition in which that misfolded protein may be exerting a deleterious impact on human brains. The phenomenon of pathologic misfolding of TDP-43 was discovered in diseases along the spectrum that include frontotemporal dementia (FTD), and its pathologic substrate frontotemporal lobar degeneration (FTLD), with or without motor neuron disease (177). However, it is now known that TDP-43 pathology is most likely to be present in individuals lacking either clinical FTD or motor neuron disease symptoms (37, 178, 179). Recent data indicate that the brains of 20%–50% of individuals in advanced old age contain TDP-43 pathology (34, 130, 179–185), in sharp contrast to FTLD, which has a lifetime incidence of ∼1:700 (186, 187). Thus, FTLD and motoneuron disease combined are more than a 100-fold less common than non-FTLD aging-associated TDP-43 pathology. This is important because the public health impact of TDP-43 pathology is far broader than was originally thought, with strong relevance to amygdalar pathology in elderly persons.

Analyses of data from large autopsy series have showed that the amygdala constitutes a primary anatomical location where TDP-43 pathology is observed in aged brains, with or without comorbid AD pathology (120, 149, 188–195). A significant proportion of subjects have been found to have TDP-43 pathology in the amygdala but essentially nowhere else in the brain: 18% (166/946) of the cohort in James et al (195), and 16% (31/193) in Josephs et al (188). As with α-SN pathology, TDP-43 pathology in the amygdala is a common, although not universal, feature of PSEN1 mutant early-onset AD cases (196).

There also is evidence for TDP-43 pathology, again like α-SN pathology, to be synergistic with Tau/NFT pathology. Non-AD tauopathy diseases, including argyrophilic grain disease (197), anti-IgLON5 tauopathy (198), dentate granule NFTs (66), corticobasal degeneration, and progressive supra-nuclear palsy (64, 199–202) have been reported to also have comorbid TDP-43 pathology, often both in the amygdala. Another intriguing [Tau + TDP-43] brain disease is chronic traumatic encephalopathy (CTE), which develops following repetitive trauma-induced brain injury (27, 71, 203). In the earliest stages of CTE, clusters of Tau and TDP-43 pathologic aggregates are observed in the depths of cerebral sulci and/or around blood vessels (27, 71, 204). Persons with this pathology may be normal functioning or they may manifest subtle symptoms that are relatively stable (205). An unknown, but possibly large, number of participants in contact sports have some version of this condition (206, 207). After years of clinical dormancy, even following decades during which traumatic injury had ceased, CTE may transform clinically into a progressive dementing disorder (208). In these cases, Tau and TDP-43 pathologies coexist in the amygdala and medial temporal lobe and spread through the cortex and brainstem (27, 67, 71, 208). This is yet another example where the amygdala region may be part of a pivotal transition wherein pathologic synergies contribute to what eventually becomes an auto-propagating, dementia-inducing neurodegenerative disease.

Although rare diseases (i.e. lifetime risk <1%) are not a focus of the current review, there are 2 themes that have emerged with direct relevance to this discussion. First, unusual diseases share with common ones the tendency for pathologic markers to be conspicuous in the amygdala. For example, in the parkinsonism-dementia complex of Guam, both Tau and α-SN aggregates are often colocalized in the amygdala (209–212). Second, there are many proteins other than Αβ, Tau, α-synuclein, or TDP-43 that misfold in the human brain, corresponding with rare neurodegenerative diseases: proteinaceous aggregates composed of misfolded fused in sarcoma (FUS), Rho-guanine nucleotide exchange factor (RGNEF), optineurin, and other proteins have been demonstrated within pathologic aggregates that correlate with neurological diseases (213–217). Thus, the amygdala appears to be a key nexus of neurodegenerative disease pathology, and proteins other than Αβ, Tau, α-SN, or TDP-43 may be prone to misfolding in the human brain.

SIX SELECTED CASES: Α β, TAU, α-SYNUCLEIN, AND TDP-43 in ADJACENT SECTIONS OF AMYGDALA

To convey how the common neurodegenerative disease-related pathologic markers are distributed in human amygdalae in a small sample, we show digitally rendered low-magnification photomicrographs from 6 human brains (Fig. 8). These individuals were followed longitudinally in a community-based cohort from normal clinical status at baseline examination until their eventual autopsy (by which time 4 individuals had documented cognitive impairment). The autopsies showed pathologic features commonly experienced in this autopsy series (2, 218) (Table 2). Immunohistochemical stains were used as described previously (66), along with digital pathologic methods that enable visualization of pathologic markers at low magnification (219). The consecutive adjacent sections of amygdala were all stained in the same order: Αβ (NAB228, gift from Dr. Eddie Lee; 1:10,000 dilution; positive staining in Table 2 indicates parenchymal plaque pathology), phospho-Tau (PHF-1, gift from Dr. Peter Davies; 1:500 dilution), α-synuclein (LB509, gift from Dr. Virginia Lee; 1:500 dilution), and phospho-TDP-43 (1D3 clone, EMD Millipore, Billerica, MA; 1:500 dilution). The next adjacent section was stained with the modified Klüver–Barrera technique to delineate the subfields of the amygdala and peri-amygdaloid regions. Labeling of the amygdalar anatomy was performed based on microscopic review of patient sections stained with a modified Klüver–Barrera protocol, incorporating 0.1% Luxol fast blue, Eosin Y, and a 0.5% aqueous solution of Cresyl violet. Anatomic references (220–226) were used in labeling the Klüver–Barrera-stained sections; the terminology of Crosby and Humphrey was applied for the amygdalar subnuclei (220). Note that each immunostain has distinct properties; for example, α-SN has the highest “background,” which is true normal staining.

FIGURE 8.

Histopathology of 6 amygdalae from the UK-ADC autopsy cohort. See Table 2 for demographic and pathologic information for these cases. The cases were selected to convey non-end-stage pathologies and relatively early clinical disease. Shown are 5 adjacent 8-µm-thick sections for each case. A Klüver–Barrera (modified) stained section (upper left) with labels is provided for anatomic orientation, followed by adjacent sections stained immunohistochemically for Aβ, phospho-Tau, α-Synuclein, and phospho-TDP-43. The immunopositive inclusions are highlighted using automated digital pathology (150, 219). Also shown in the lower right panel for each case is a “highlight” from one of the stained sections; the portion of the figure that is depicted as a highlight is indicated by a black arrow. Case 1 shows phospho-TDP-43 pathology, and phospho-Tau pathology that includes ARTAG (highlighted), particularly in the area medial to the lateral ventricle. Case 2 has large ectatic blood vessels between the lateral ventricle and the medial region of the amygdala, and near those blood vessels is scattered phospho-TDP-43 pathology (highlighted). Case 3 represents a relatively common clinical-pathologic scenario in the UK-ADC cohort: clinical dementia and a diagnosis of “Probable AD,” followed by an autopsy that revealed the presence of all 4 pathologic markers, including extensive phospho-TDP-43 pathology in the basal and lateral amygdalar subnuclei (highlighted). Case 4 also has both AD and α-synuclein pathologies. The highlighted region shows the pial surface with some scattered α-Synuclein aggregates (purple arrows). Case 5 is an interesting case of preclinical AD + LB, both within the amygdala and in the nearby cortex. The phospho-Tau antibody highlights abundant NFTs and neuritic plaques; the combination of dense plaques and tangles is rare in other brain areas. Case 6 is another case with preclinical or early disease (final MMSE score of 28), showing mild degree of LBD and phospho-TDP-43. The highlight depicts subpial Aβ deposits. Cases with comorbid Aβ, Tau, α-Synuclein, and phospho-TDP-43 pathologies are common among cognitively impaired subjects in this cohort, but are quite unusual among cognitively intact individuals (2) (Fig. 7). Abbreviations: ac – anterior commissure, ag – ambient gyrus, Acc. Bas.n. – Accessory basal nucleus, Bas.n. – basal nucleus, Bsl.Forbr. – basal forebrain, including magnocellular neurons (nucleus basalis), BV, blood vessel, Cent.Med.n.g. – Centromedian nuclear group, Claus – claustrum, Co.n. – Cortical nucleus, Co.A.Tr. – cortico-amygdaloid transition area, cs – collateral sulcus, Ectx - Entorhinal cortex, ers – endorhinal sulcus, fg – fusiform gyrus, Icm – intercalated cell mass, Ins – insula, Lat.n. – Lateral nucleus, lv – lateral ventricle, phg – parahippocampal gyrus, sas – semiannular sulcus, sg – semilunar gyrus, Sub – subiculum, Str. – striatum/ ventral extension of putamen, TEctx – transentorhinal cortex, TPC – temporopolar cortex (area TG, perirhinal cortex, Brodmann area 35-36).

TABLE 2.

Clinical and Pathological Data on 6 Cases From the University of Kentucky ADC Autopsy Cohort

Case #	Age at Death (Years)	Final Clinical Diagnosis	Final Overall Diagnosis	Braak NFT Stage	Medial Temporal Lobe Pathology
					Aβ+	TDP-43 Rating (0–2)	α-Synuclein Severity Rating (0–4)
1	87	MCI	CARTS + ARTAG + PART	III	N	2	0
2	78	AD	CARTS + PART	II	N	2	0
3	88	AD	AD + DLB + TDP-43	IV	Y	2	3
4	96	AD + DLB	AD + DLB	V	Y	0	4
5	87	Normal	Preclinical AD (mixed)	V	Y	1	2–3
6	96	Normal	Preclinical AD (mixed)	V	Y	1	1

AD, Alzheimer disease; ARTAG, aging-related tau astrogliopathy CARTS, cerebral age-related TDP-43 and sclerosis; DLB, dementia with Lewy bodies; N, no; MCI, mild cognitive impairment; PART, primary age-related tauopathy; Y, yes. Amygdalae were stained immunohistochemically for Aβ, Tau, TDP-43, and α-synuclein and are illustrated in Figure 8.

FIGURE 8.

Continued

FIGURE 8.

Continued

The stained brain sections showcase common patterns of pathologic markers seen in the UK-ADC cohort, and were selected to include pathologies in early and/or preclinical stages. From these photographs one can discern the following: (i) pathology often exists around the periphery of the amygdalae, near the meninges and/or lateral ventricle; (ii) peri-amygdaloid grey matter, including the entorhinal cortex, frequently shows pathologies; (iii) cortical and transitional regions also seem vulnerable to the accumulation of pathologic markers; and (iv) a constant in all cases among aged individuals is the presence of phospho-Tau pathology, which can be seen in gray or white matter, and within neurons (usually AD and/or PART) and/or astrocytes (usually ARTAG). As may be expected, there is not an easily deducible “one-to-one” overlap in the pathologies and the pathologic synergies (if they exist) may be specific for certain diseases and/or may be “seeded” within small subnuclei. A more systematic exploration of the anatomic associations between different pathologic markers requires far more cases and is beyond the scope of the current article.

ADDITIONAL CONSIDERATIONS AND CONCLUSIONS

In this review, we postulated that there is a synergistic relationship between misfolded protein species in some aged human brains and that these mechanisms may be extraordinarily strong within the amygdala. It is intuitively unrealistic to hypothesize that every neurodegenerative disease case is affected by pathologic synergies in the amygdala. And an alternative (arguably the null) hypothesis is that there is no pathologic synergy between misfolded proteins in neurodegenerative diseases, including in the amygdala. The underlying conditions in some aged brains may favor misfolding of multiple proteins, perhaps a “phenotype of neurodegeneration” (7). The misfolding of those proteins could be independent of each other. For example, the data in Figure 1C indicating more severe impairment for AD + DLB than pure AD may merely reflect the additive effect of separate underlying diseases. For there to be no pathologic synergies, many phenomena remain to be explained. For example, it is challenging to develop a hypothesis to explain the features of a neuritic amyloid plaque which does not include pathologic synergy (Fig. 2), and why, in the absence of pathologic synergy, there are α-SN and/or TDP-43 pathologies in middle-aged persons with APP mutations (227–229). However, it is useful to remind ourselves that biologic mechanisms in the brain may be neither all synergistic nor all independent of each other, but a combination of influences is quite possible.

Another concern is that much of the data gathered on pathologic synergies in neurodegeneration were from cross-sectional (postmortem) studies and therefore lack mechanistic insights. It remains unproven that any misfolded proteins in human brains are deleterious for the cells they are in or near. Whether or not a specific protein’s misfolding has a direct “toxic” impact, an assumption of this review is that the protein misfolding in the aged human brain heralds a deleterious process. This assumption is based on extensive prior studies from multiple institutions that indicate a strong correlation between the density and/or distribution of pathologic markers and clinical dysfunction (16, 17). If a pathologic marker were only a proxy for the presence of a “toxic oligomer,” or some other molecular adduct or agent(s), that fact could be important information for investigators seeking therapeutic strategies. However, if the correlation is reliable between a readily identified pathologic marker and something directly toxic that is less easily identified in situ, then the identifiable marker could still offer a robust indicator of the harmful process(es) that drive symptomatology.

We conclude by speculating that the amygdala may provide an anatomic setting for the pursuit of entirely new diagnostic and therapeutic targets. Given that the known pathologic protein species for common neurodegenerative diseases were characterized relatively recently—Αβ: 1984 (230), Tau: 1986–1987 (231–236), α-SN: 1997–1998 (237, 238), and TDP-43: 2006 (177, 239)—are we certain that there are no additional misfolded proteins that are directly relevant to prevalent brain diseases? And, if there are as yet unidentified proteins that could function as new pathologic markers and potential therapeutic targets, isn’t the amygdala the logical place to find them? Despite the great amount of scientific progress during the past century, the final sentences of Alois Alzheimer’s seminal case study, published in 1907, still ring true in translation: “There are without any doubt many more psychic illnesses than listed in our textbooks. In some of these instances a later histological examination will subsequently reveal peculiarities of the specific case. Then, we will gradually arrive at a stage, when we will be able to separate out individual disease from the large illness categories of our textbooks; to delineate them clinically more accurately.” (240).