The relationship between dementia and direct involvement of the hippocampus and amygdala in Parkinson's disease

Dementia is a frequent and disabling feature of advanced Parkinson's disease (PD). At the onset of symptomatic PD, even before treatment, subtle cognitive deficits consistent with dysfunction of the frontostriatal loop are common,^1-2 and 10 to 20% of patients eventually become severely demented.³ The cause of dementia in PD is probably manifold but likely includes direct cortical involvement as evidenced by the presence of Lewy bodies (LBs) and Lewy neurites (LNs), degeneration of subcortical nuclei causing cortical transmitter deficiencies, and coincidental conditions such as Alzheimer's disease (AD). Involvement of the entorhinal cortex (ERC) may be important in the development of dementia in many patients.⁴ No pathologic cause of cognitive decline could be identified in 55% of patients with PD and late dementia in one series (which did not examine the hippocampus and amygdala).⁵ Because both the amygdala and the hippocampus are involved in intellectual function in humans,^6-8 involvement of either might contribute to cognitive impairment in PD. In PD, the amygdala is atrophic,⁹ and LBs and LNs are present in most amygdala nuclei,¹⁰ but the relationship of these changes to cognitive impairment is unknown. LBs and LNs have been observed in the CA2-3 fields in patients classified as having diffuse Lewy body disease (DLBD),^11,12 and the density of LNs has been related to that of cortical LBs in DLBD.^11,12 Hippocampal LNs and LBs are present in demented PD patients¹² but were not present in PD patients in a series in which cognitive function was not documented.¹¹ Neuronal density is preserved in the CA1, CA2, and CA3 fields in PD as well as in senile dementia of the LB type(SDLT).¹³ Thus, although the CA1 and CA4 fields are spared in PD,^11,12 it is not clear whether involvement of the CA2-3 fields (as evidenced by neuronal loss or local LNs and/or LBs) contributes to dementia in PD.

To determine if involvement of either the hippocampus or amygdala might contribute to dementia in PD, we examined the density of neurons, LBs, LNs, neurofibrillary tangles (NFTs), and neuritic plaques in the amygdala and the CA1-4 fields of the hippocampus in demented and nondemented PD patients without AD or other causes of dementia. Cortical LB involvement was assessed in the anterior cingulate gyrus, which is characteristically involved in DLBD.

Methods. Twenty-seven patients fulfilling the UK Parkinson's Disease Society Brain Bank criteria for PD¹⁴ and 11 controls were studied. Selection into the study required detailed prospective documentation of initial presentation, course of the disease, mental state, and cognitive function. Cognitive function was rated with prospectively recorded annual Folstein Mini-Mental State Examination (MMSE) scores,¹⁵ and patients were rated functionally as normal or mildly, moderately, or severely demented according to DSM-III criteria¹⁶ within 6 months of death. Other causes of dementia identified by prior independent pathologic examination, including pathologic changes sufficient to satisfy the various criteria for AD,^17-19 excluded selection. Cortical LBs or low densities of plaques and NFTs insufficient to diagnose AD did not influence selection. Thus, this study was intentionally confined to those with the clinicopathologic characteristics of PD alone. The case notes of the controls were insufficient to assess cognition, but none had cerebral infarction, LBs, or AD at autopsy.

Ten serial 16-µm-thick coronal sections were taken through the amygdala at the level of the internal pallidum (figure 1), through the hippocampus at the level of the cornu Ammonis, and through the anterior cingulate gyrus at the level of the anterior frontal lobe. Successive sections were stained with hematoxylin-eosin or Luxol fast blue-cresyl violet or modified Bielschowsky silver impregnation or were immunostained with polyclonal anti-ubiquitin (Dako 1:150) or polyclonal anti-glial fibrillary acid protein (Dako 1:400) or monoclonal -A4 (Dako 1:100) or monoclonal anti-tau (Dako 1:1,500). The CA1-4 hippocampal fields²⁰ and the central (Ce), cortical (Co), accessory cortical (AcCo), and lateral(La) nuclei of the amygdala¹⁰ were demarcated on the slide coverslip with a felt pen, their areas determined with an eyepiece graticule, and the densities of neurons, LBs, NFTs, and neuritic plaques calculated. The area of the anterior cingulate gyrus was determined with a Colomorph Image Analyser (Perceptive Instruments). Gliosis and LN density were estimated qualitatively (0 to 5). The density of cortical LB involvement was estimated from a single anti-ubiquitin-stained section of the anterior cingulate gyrus as well as by the Kosaka criteria for DLBD¹² from an adjacent hematoxylin-eosin section. Sections were examined with the examiner blind to the diagnosis and cognitive state. Densities of each object in patients and controls were compared using the paired t-test. Cognitive function was related to the density of neurons and each pathologic feature using Kruskal-Wallis (MMSE) and one-way ANOVA (DSM-III).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Amygdala nuclei examined according to reference 10. Ce = central, Co = cortical, Ac = accessory cortical, La = lateral, ot = optic tract, GPi = internal pallidum.

Results. Clinical data. All patients presented with asymmetric motor deficits of PD (14 tremor, 5 akinesia, 8 tremor and akinesia) without overt cognitive dysfunction. Dementia, confusion, and hallucinations, when present, occurred late (onset of overt cognitive impairment: 9.3 ± 1.2 years after initial motor symptoms). None of the patients fulfilled the clinical criteria for DLBD,²¹ LB variant of AD,²² or SDLT.²³ At death, 17 were demented according to DSM-III criteria (3 mildly: MMSE 18 to 24; 4 moderately: MMSE 13 to 17; and 10 severely: MMSE <11). The mildly and moderately demented groups were combined because of their low numbers. The other 10 patients had MMSE scores >25 and were not demented according to DSM-III criteria. Controls and each PD group were matched for age. PD groups were matched for duration of disease and treatment, Hoehn and Yahr score, and daily dose of L-dopa (table 1). Confusion and hallucinations were precipitated by drugs in two nondemented patients but resolved after withdrawal of the causative medication. The incidence of abnormal behavior (social withdrawal, agitation, aggression, and wandering), confusion, and hallucinations increased with the degree of dementia. These neuropsychiatric complications usually became chronic in the severely demented patients, three of whom were drug free at death because antiparkinsonian therapy worsened their mental state. Depression was uncommon.

Pathologic data. LBs, LNs, densities of plaques, and NFTs sufficient to satisfy pathologic criteria of AD^17-19 or nigral neuronal loss were not detected in any control. Severe neuronal loss and LBs occurred in the substantia nigra and locus coeruleus in all patients (data not shown). No nondemented PD patient fulfilled the Kosaka criteria for DLBD.²¹ Two moderately and three severely demented patients had five or more cortical LBs in a single hematoxylin and eosin-stained section through the anterior cingulate gyrus and could therefore be classified as having DLBD.²¹

Neuronal densities were unaffected by PD, regardless of the degree of dementia, in all hippocampal fields and amygdala nuclei (tables 2 and 3). Densities of LBs and LNs were greatest in the CA2-3 hippocampal fields (figure 2) and the AcCo, Co, and Ce nuclei of the amygdala. In the CA2 field, LBs and LNs were denser in the severely demented patients than in the nondemented or less cognitively impaired patients (see table 2). Neither LBs nor LNs were present in the CA2-3 field of one patient with mild to moderate dementia and one patient with severe dementia. Thus, CA2 field LNs and LBs were detected in 88% of demented PD patients. Low levels of CA2-3 LNs or LBs were present in five nondemented patients (50%), but no hippocampal pathology could be detected in the remaining cognitively intact patients. In the amygdala, densities of LNs and LBs were the same in all groups (seetable 3), and LBs and LNs were detected in all cases apart from one mildly demented patient. NFTs and neuritic plaques occurred at low densities in all patients and were not increased in either the amygdala or hippocampus in comparison with controls (see tables 2 and 3). No patient or control had congophilic angiopathy or more than very occasional senile plaques in either the amygdala or hippocampus. Significant atrophy was present only in the CA2 fields of PD patients with mild to moderate and severe dementia (table 4). Gliosis was generally very mild (rated 0 to 1), even in the presence of high densities of LBs and LNs, and was increased only in the Ce nuclei in PD patients with mild to moderate dementia. Nonetheless, marked gliosis (rated 3) was present throughout the hippocampus and amygdala in two patients satisfying the criteria for DLBD²¹ and with very high densities of hippocampal and amygdala LBs and LNs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Anti-ubiquitin immunostained sections through the hippocampus in a severely demented patient (magnification×100, before 29% reduction). (A) Frequent Lewy neurities (arrowheads) in CA2 field. (B) Intraneuronal Lewy body (asterisk) in CA4 field.

Neuronal density in each hippocampal field and amygdala nucleus was unrelated to cognitive function (p > 0.05; MMSE: Kruskal-Wallis; DSM-III: one-way ANOVA). Increasing LN densities in the CA2 field were related to worsening dementia as assessed by both the MMSE and DSM-III (p< 0.05; MMSE: Kruskal-Wallis; DSM-III: one-way ANOVA). No relationship between LBs in the CA2 field and dementia was identified (p > 0.05; MMSE: Kruskal-Wallis; DSM-III: one-way ANOVA), even though their density was greater in the severely demented group (see table 2). Levels of CA3 LBs and LNs increased with the degree of cognitive impairment, but no relationship between either pathologic abnormality and the severity of dementia was established (p > 0.05; MMSE: Kruskal-Wallis; DSM-III: one-way ANOVA). Densities of LBs and LNs in the CA1 and CA4 fields and each amygdala nucleus as well as of NFTs and neuritic plaques throughout the amygdala and hippocampus were unrelated to the degree of dementia(p > 0.05; MMSE: Kruskal-Wallis; DSM-III: one-way ANOVA).

LB densities (LB/mm²) in the anterior cingulate gyrus increased with worsening cognitive impairment (controls: 0 ± 0; nondemented: 0.2± 0.04; mildly to moderately demented: 1.1 ± 1.0; severe dementia: 1.5 ± 0.5), but this trend was not significant. Anterior cingulate LB densities in the demented PD patients ranged from 0 to 5.6/mm², which may have prevented detection of a relationship between cognitive function and cortical involvement. The highest cortical LB densities invariably occurred in those patients satisfying the criteria for DLBD.²¹ Anterior cingulate LB densities could not be related to cognitive function (p > 0.05; MMSE: Kruskal-Wallis; DSM-III: one-way ANOVA) or to levels of CA2 LNs (r= 0.31).

Discussion. The pattern of involvement of the hippocampus and amygdala conformed to that previously described; densities of LBs and LNs were greatest in the CA2-3 fields¹² and the AcCo, Co, and Ce nuclei of the amygdala.¹⁰ However, only LN density in the CA2 field of the hippocampus was related to the severity of dementia in this study. This finding is consistent with the presence of CA2-3 LNs in demented PD¹² and DLBD¹¹ patients noted previously. The density of LNs is proportional to that of cortical LBs in DLBD.^11,12 In contrast, we found that the density of anterior cingulate LBs could not be related to either hippocampal LNs or to cognitive impairment. Moreover, DLBD occurred in only a minority of demented patients, including many with prominent CA2-field LNs. Thus, the presence of hippocampal LNs was not invariably associated with DLBD, and marked involvement of the CA2-3 fields was associated with dementia in the absence of severe and widespread anterior cingulate LBs. With regard to other possible causes of dementia, formation of NFTs and neuropil threads in the pre-layer of the ERC has been reported in demented PD patients,⁴ and the total cortical LB load has been related to the severity of dementia in PD.²⁴ Whether total cortical LB load (in multiple cortices) or damage to the ERC contributed to cognitive impairment in these cases was not examined. The low densities of NFTs and neuritic plaques observed in the hippocampus and the amygdala are consistent with previous findings in the nondemented elderly,²⁵ and none had coincidental AD or multiple cerebral infarction on pathologic examination.

The CA2-3 fields receive major inputs from the hippocampal dentate gyrus, ERC, septal nuclei, and supramammillary region of the hypothalamus and then project onto the CA1 field, which in turn projects to the subiculum.⁷ Because the ERC, septal nuclei, and hypothalamus are all markedly affected in PD,^26-28 additional involvement of the CA2-3 fields would result in disruption of all the major inputs to the CA1 field. This possibility is reinforced by the observation that loss of synapses in the CA2-3 fields has been strongly correlated with dementia in AD.²⁹ Thus, we hypothesize that the pathologic process associated with LN formation in the CA2-3 fields disrupts hippocampal function, and hence cognition, by interfering with inputs to the CA1 field. Additional coincidental involvement of the ERC,⁴ which we did not examine in this study, and extensive cortical LBs²⁴ are also likely to contribute to dementia in PD. The source of hippocampal LNs is unknown. The preservation of hippocampal neurons suggests that they are not intrinsic fibers, and whether they arise from the diseased hypothalamus or septum^7,26-28 has not been determined.

Neuronal densities in the CA2-3 fields were preserved in the presence of LBs, and gliosis was rarely prominent, even in the presence of high densities of LBs and LNs, in contrast to the substantia nigra, in which gross neuronal loss is commonly accompanied by gliosis. However, because the CA2 field was atrophic in the demented PD patients, neuronal loss may have occurred undetected because of the methods used. Similarly, neuronal loss could not be detected in the hypothalamus in PD, even though LBs were present.²⁷ Thus, despite evidence that LBs are a marker of cell death,¹⁴ their presence may not invariably indicate gross neuronal destruction. The ERC, which forms the major interface between the cortex and hippocampus,⁷ was not examined in this study, but densities of NFTs and other changes of AD in the ERC increase with the degree of dementia in PD.⁴ Thus, the preservation of neurons in the CA1 and CA3 fields suggests that involvement of the ERC does not lead to severe anterograde neuronal degeneration in the hippocampus.

The function of the amygdala in humans is poorly understood. Bilateral amygdala lesions that spare the hippocampus do not grossly diminish memory or IQ^6,8 but impair activation and reactivation of remembered emotionally significant events,⁸ suggesting a specific mnemonic function. Bilateral amygdala lesions reduce emotional expression^6,8 and impair recognition and processing of emotional facial expression in others.³⁰ Similarly, emotional facial expression and recognition of facial emotions in others is diminished in PD in the absence of dementia.³¹ Thus, amygdala degeneration may contribute to the abnormal expression and recognition of facial emotions in PD without causing dementia per se.

The patients in this study all presented with classical PD as confirmed by severe nigral neuronal loss and LBs. Dementia, when present, occurred late; that is, all had PD complicated by dementia and did not conform to the LB variant of AD²² or SDLT,²³ both of which are characterized by early and predominant dementia, cortical LBs, and changes of AD, often with less severe nigral involvement than in PD. Only a minority of the demented patients satisfied the pathologic criteria for DLBD.²¹ McKeith et al.³² recently established consensus guidelines for a new and as yet untested clinicopathologic entity, dementia with Lewy bodies (DLBs), in an attempt to resolve the ongoing nosologic controversy as to the relationship between cortical LBs, PD, and AD. According to the authors, the diagnosis of DLB requires progressive mental impairment and at least two of the following: fluctuating cognitive function, recurrent visual hallucinations, and motor parkinsonism.³² The global dementia of DLB typically occurs over "a period of months, but more commonly over several years" (p. 1115), motor parkinsonism is typically mild, and tremor is less common than akinesia and rigidity (p. 1116), and parkinsonism of greater than 12 months' duration before the onset of dementia is better classified as PD with dementia.³² Thus, as stated by the authors, the existence of DLB supposes that it is a separate and discernible clinicopathologic entity with regard to both the evolution and time course of dementia and the pattern and severity of motor parkinsonism. The patients examined in this study do not satisfy the criteria for DLB because their motor parkinsonism was marked from the outset, tremor was frequent, and overt cognitive impairment occurred late (9.3 ± 1.2 years after presentation). Nonetheless, this study is relevant to the debate about these putative entities. In this series, episodic confusion, hallucinations, agitation, and severe personality changes were increasingly frequent with worsening dementia but occurred even with minimal cognitive impairment. In a recent autopsy study of patients with parkinsonism and dementia at death and dense cortical and nigral LBs without AD, the various initial symptoms included parkinsonism alone, pure dementia, and both dementia and parkinsonism³³; that is, the neuropsychiatric complications of dementia in PD overlap those of DLB,³² DLBD,²¹ LB variant of AD,²² and SDLT,²³ whereas CA2-field LNs and extensive cortical LBs appear to correlate with the severity of dementia.²⁴ Seemingly, the only major differences between these entities and PD with dementia are the severity of parkinsonism and whether frank cognitive impairment occurs early or late in the course of the disease. Thus, the most parsimonious explanation of dementia in parkinsonism and LB disease is that it exists as part of a spectrum that embraces classical PD with minor cognitive impairment and minimal cortical involvement and also a severe dementia, with or without antecedent parkinsonism, in which there is severe cortical and hippocampal involvement as evidenced by LBs and LNs. Whether DLB exists as a clinically useful entity that can be readily separated from PD complicated by late dementia remains uncertain, but our study and those of others³³ suggest that DLB and PD with dementia overlap clinically and pathologically.

Acknowledgments

Professors Heikko and Eva Braak, Zentrum der Morphologie, Frankfurt a. Main, generously allowed A.C. to visit their laboratory to view their superb studies of the hippocampus and amygdala and unselfishly shared their vast expertise in this complex field. Dr. Susan Daniels performed the initial diagnostic histologic examination, Miss Hardev Sangha provided invaluable technical assistance in the laboratory, and Mrs. Susan Stoneham kindly assisted with the photography.