Clinical and quantitative pathologic correlates of dementia with Lewy bodies

Patients.

We studied 25 cases (19 men and 6 women, age at death 78.6 ± 5.8 years) autopsied at the Massachusetts Alzheimer’s Disease Research Center (ADRC) Brain Bank, Boston, between 1993 and 1998. The neuropathologic diagnosis in all of them was diffuse LB disease or Lewy variant of AD. In addition to a diagnosis of LB-related dementia, these cases were selected with two criteria: first, that the neuropathologic diagnosis reported either no or only mild concomitant AD changes. Our purpose was to enroll cases with as pure LB disease as possible to obtain clinicopathologic correlations not confounded by the presence of tangles and neuritic plaques. Second, all these cases had been examined neurologically during the disease and had quite extensive clinical data available on the onset, symptoms, and duration of the disease. The average duration of disease was 9.4 ± 4.5 years; the average length of time individuals were followed clinically at the Memory Disorders Unit of the ADRC was 43.1 ± 40.2 months (range 1 to 123). The time between assessments for each patient was approximately 6 months. The last examination was conducted at 27.5 ± 26.5 months before the time of death.

Clinical data reviewed for each case were the following: age at onset, symptoms at onset, duration of the disease in years, presence of dementia, extrapyramidal symptoms, and visual hallucinations. Other symptoms relevant to the clinical consensus criteria for DLB such as cognitive fluctuation, systematized delusions, and history of syncope and unexplained falls were also considered. Clinical examination in all cases had been conducted according to a standardized protocol. The diagnosis of dementia was based on DSM-III-R criteria19 and supported by impairments on the Information-Memory-Concentration subscale of the Blessed Dementia Scale20 and on a set of cognitive tests administered in the clinical unit at Massachusetts General Hospital.21 The presence of at least two of the cardinal manifestations of PD (bradykinesia, rigidity, tremor at rest, and balance and gait instability) was required for extrapyramidal signs.22 Psychiatric symptoms, syncope, and falls were considered positive only if recurrent (noted on more than one examination). Because the omission of some of these symptoms in retrospectively reviewed charts may be wrongly interpreted as their absence, we were careful to consider the absence of a certain symptom when it was made explicitly clear either through the clinical history or examination. We cannot, of course, exclude the possibility that a certain symptom that was considered negative might have developed in an individual between last examination and death.

An initial attempt was made to use quantitative scores for the severity of the symptoms. However, for the majority of patients seen within 1 year of death, a “floor effect” was observed, i.e., most of the patients scored in the “worst” range or were essentially untestable. Our approach was then to compare the patterns of pathology among groups of patients who had a different symptom complex at onset (i.e., cognitive impairment, psychiatric symptoms, or parkinsonism) or who developed certain symptoms during the course of the disease (i.e., parkinsonism, visual hallucinations, delusions, recurrent falls, etc.).

Tissue processing.

All brains had been fixed in 10% buffered formalin and routinely dissected according to a standardized protocol.23 Tissue blocks were obtained from 18 areas, embedded in paraffin, and sections cut at 8 μm. Sections were stained with hematoxylin-eosin, Congo red, and the modified Bielschowky method. On this basis a pathologic diagnosis of diffuse LB disease or LB variant of AD had been made where there were classical LB within the pigmented brainstem nuclei and similar eosinophilic inclusion bodies distributed throughout limbic and neocortical regions.7

A Braak and Braak24 stage was determined for each case according to the neurofibrillary tangles present in silver stains. Thirteen cases had very few tangles only in hippocampus and perirhinal areas (Braak stages I, n = 7, or II, n = 6), and we designated them Group A because they met a low likelihood that dementia was due to AD according to the National Institute on Aging/Reagan criteria.25 Another 11 cases had enough paralimbic tangles to meet Braak stage either III (n = 7) or IV (n = 4). One case had some tangles in association cortex meeting criteria for Braak stage V. These latter 12 cases together were designated as Group B because AD-related changes had at least an intermediate chance of contributing to dementia.25 We also recorded the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) semiquantitative scores of neocortical neuritic plaques (0 = none, 1 = sparse, 3 = moderate, and 5 = frequent) and applied CERAD criteria.26 According to these criteria, Group A included two cases with no AD changes, four possible, four probable, and three definite AD; Group B included two possible, three probable, and seven definite AD cases. Thus, the degree of overlap between DLB and AD depends greatly on the definitions used for AD. The current series provides a gradient of neuropathologic material ranging from little to moderate overlap.

For quantitative neuropathologic assessment we studied areas chosen to represent brainstem (substantia nigra at the level of the red nucleus), paralimbic areas (entorhinal cortex, cingulate gyrus, insula, and hippocampus), and neocortex (medial frontal and associative occipital cortex). We included occipital cortex because of the functional studies that suggest occipital hypometabolism as a significant feature in DLB27-28 and the prominent visual hallucinations and visuospatial disorders29 often described in these patients. We performed double-label fluorescence immunohistochemistry with a polyclonal antibody against ubiquitin (DAKO, Glostrup, Denmark), according to the Consortium on DLB criteria to identify LB, and the monoclonal antibody HC3 against α-synuclein (courtesy of Dr. David Clayton, University of Illinois, Urbana, IL), which is increasingly recognized as a very specific immunomarker for LB.13-15 The secondary antibodies were cy3 anti-rabbit (Jackson ImmunoResearch, West Grove, PA) and BODIPY anti-mouse (Molecular Probes, Eugene, OR). The slides were mounted with 4′,6-diamino-2-phenylindole aqueous medium (Vectashield, Vector, CA), which stains nuclei and allows for a distinction between white and gray matter under ultraviolet illumination.

Quantification.

All round α-synuclein–positive structures were considered and counted as LB. This criterion included intracytoplasmic, intraneuritic, and extracellular LB. LB counting was performed systematically throughout the whole selected area available on each slide, ranging from 15 mm² of substantia nigra to 135 mm² of cingulate cortex. The scanning for LB was conducted under a 40× objective, first for anti–α-synuclein immunostaining and then double counted for anti-ubiquitin staining. This order allowed us to exclude globose tangles, which are ubiquitin positive but α-synuclein negative,30 especially in entorhinal cortex and hippocampus where several cases displayed some AD changes. The quantitation was performed in all cases and areas by the same investigator who was unaware of the clinical data.

LB quantitation and area measurements were recorded with a computer-based image analysis system (Bioquant Image Analysis System with Stereology Package, Nashville, TN). This program provides stereologic overlays to assist in estimating the total number of LB in cortical areas and also records the positions of each of the counted structures with x/y coordinates. These data provide an accurate visual image of the distribution of LB throughout cortical layers. The total number of LB counted in each area was divided by the area scanned to obtain a LB density, expressed in LB/mm² of tissue. Paralimbic LB burdens were calculated for each case as the mean for the four paralimbic areas, and neocortical LB burdens were calculated as the mean for the two neocortical areas. A total LB burden was calculated as the mean for nigral, paralimbic, and neocortical burdens.

Statistical analysis.

We tested whether there were significant differences in LB density across brain regions and whether the mean profile differed for Groups A and B using a mixed between (Group A versus B)-within (brain region) subjects analysis of variance. Post hoc comparisons for all possible pairwise contrasts of brain region means were adjusted with Bonferroni correction for multiple comparisons. The correlations between LB density and disease duration, Braak stages, or CERAD semiquantitation of senile plaques were examined by linear regression analysis. The association between LB accumulation in specific areas and clinical symptoms was examined in two aspects. First, we compared cases with different symptoms at onset (dementia or parkinsonism) using Student’s t or Mann-Whitney U tests according to the distribution of the data. Second, we compared LB density between groups according to the presence or absence of a certain symptom (cognitive fluctuations, parkinsonism, visual hallucinations, delusions, or falls) at any time during the disease. This analysis was performed using two different approaches. In the first, we compared groups with Student’s t or Mann-Whitney U tests and used Bonferroni correction. In the second, we performed logistic regression using LB densities in each of the seven areas and global regions as the independent variables, the presence or absence of a certain symptom as the dependent variable, and whether the case belonged to Group A or B as a covariate. Statistical analyses were performed using SAS software (Statistical Analysis Systems, Inc., Cary, NC).

Clinical symptoms.

Age at onset for the whole group was 69.3 ± 7.1 years (range 48 to 80), and duration of the disease was 9.4 ± 4.5 years (range 3 to 17). Symptoms at onset were cognitive impairment in 17 patients and parkinsonism in 7 (in 1 patient the symptoms at onset were not clear). In the first group, 12 patients started with insidious cognitive impairment and 5 started acutely after a confusional episode, and cognitive decline progressed thereafter. Note that these acute confusional episodes, often with psychotic behavior, developed in previously cognitively intact individuals and without apparent cause, except for one patient in whom confusion appeared in the context of heart surgery. Of the seven patients whose presenting symptom was parkinsonism, cognitive decline followed during the next 4 to 24 months in five, or within the 4th to 5th year in the other two. Psychiatric symptoms early in the disease were often intertwined with confusional episodes and cognitive decline and so could not be considered as a different onset. Cases that started with parkinsonism had a slightly longer duration of the disease (12.1 ± 3.5 years) than cases that started with insidious cognitive decline (8.5 ± 4.1) or with acute confusion (7.6 ± 5.6), but the differences were not statistically significant. There were no significant differences in age at onset, disease duration, or symptoms at onset between Groups A and B (table 1).

At some time during the evolution of the disease all patients developed dementia, 19 developed parkinsonian signs, and 13 had visual hallucinations. Overall, 12 patients (48%) showed the complete symptomatic triad. Other symptoms were reported more infrequently. Fluctuations in cognitive status and alertness, in a broad sense including variable cognitive performance as well as recurrent confusional, lethargic or syncopal-like episodes, were reported in 16 cases. Recurrent falls were reported in eight cases, and systematized delusions in eight.

Neuropathology.

The quantitation of α-synuclein and ubiquitin immunostaining paralleled closely and the former were the data analyzed. There was a significant effect for brain region (p < 0.0001). The density of LB showed a linear trend as follows: substantia nigra > entorhinal cortex > cingulate gyrus > insula > frontal cortex > hippocampus > occipital cortex (figure 1). This rank order was highly consistent as assessed in most of the cases, with 21 of the 25 cases having Pearson correlation coefficients across brain regions greater than 0.65. There was no significant group effect. LB density did not differ significantly between Groups A and B across brain regions, except in the entorhinal cortex where the mean value was significantly higher (p < 0.05) in Group B (see figure 1 and table 1). There were no significant differences in brain region means when comparing men and women. The following post hoc pairwise comparisons of brain regions were significant after Bonferroni correction (p = 0.002): both cingulate and entorhinal cortex were significantly different from substantia nigra and neocortical regions; substantia nigra was significantly different from neocortical regions and hippocampus; and hippocampus was significantly different from entorhinal and occipital cortex.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. LB density across brain regions (listed along x axis). The columns represent the mean (± standard deviation) for the 25 cases of dementia with Lewy bodies together, and the lines show the means for Group A (♦; Braak and Braak stages I and II, n = 13) and Group B (▪; Braak and Braak stages III to V, n = 12) separately.

There was a strong correlation between paralimbic and neocortical LB burdens (r² = 0.56, p < 0.0001, slope ± standard error [SE] = 0.38 ± 0.07), which remained even after omitting two extreme cases (r² = 0.39, p = 0.0018, slope ± SE = 0.38 ± 0.10), but neither of these correlated directly with LB density in substantia nigra (figure 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Correlations between density of Lewy bodies in substantia nigra, paralimbic areas, and neocortex. Paralimbic burden for each case is the mean for hippocampus, cingulate gyrus, insula, and entorhinal cortex. Neocortical burden for each case is the mean for frontal and occipital areas. The bands represent the 95% confidence intervals.

Although LB frequently occurred in the deep cortical layers, there was some case-to-case variability in the laminar topography of LB. In some cases synuclein-positive structures were clearly confined to deep layers, whereas in others they were more spread out. Moreover, each individual case usually displayed the same pattern in all paralimbic and neocortical areas. This pattern of “confined” versus “widespread” was only partially dependent on a higher LB density because several cases with a high LB density showed LB pathology confined to deep layers (figure 3). Laminar pattern was not related to disease duration. In general, the distribution of LB across each area was quite irregular, with clusters of LB alternating with unaffected areas. We did not find cases with paralimbic but without neocortical LB or vice versa, suggesting that LB occur in multiple areas concurrently during the evolution of the disease.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. Overviews of the distribution of Lewy bodies (LB) throughout cortical layers in entorhinal cortex in six different cases of dementia with LB. LB density increases along the cases, but duration of the disease and distribution of LB (“confined” to deep layers or “widespread”) are not consistent.

The degree of neurofibrillary tangle involvement, as assessed by Braak stages, did not correlate with LB density in any region (r² = 0.010, p = 0.63 versus paralimbic; r² = 0.001, p = 0.83 versus neocortical; and r² = 0.002, p = 0.80 versus total LB burden). The amount of cortical senile plaques, assessed according to CERAD criteria, did not correlate with neocortical (r² = 0.055, p = 0.27), paralimbic (r² = 0.002, p = 0.83), or total (r² = 0.002, p = 0.83) LB density.

Clinicopathologic correlations.

First, we examined whether patients with different symptoms at onset (cognitive decline or parkinsonism) would have different LB densities in substantia nigra and neocortex. Cases where the disease initiated with parkinsonism had higher LB density in substantia nigra, and those that started with cognitive impairment had slightly higher LB density in neocortex, but the differences were not statistically significant (table 2).

Second, we compared regional LB density (nigral, neocortical, paralimbic, or total LB densities) between groups according to the presence or absence of a certain symptom (cognitive fluctuations, parkinsonism, visual hallucinations, delusions, or falls) at any time during the disease (see table 2). Specifically, we examined whether cases with parkinsonism or recurrent falls would have higher LB burdens in substantia nigra than those without these symptoms and whether the clinical evidence of visual hallucinations, delusions, or cognitive fluctuations would be related to a higher neocortical or paralimbic LB burden. There were no significant differences in LB density in any region when comparing patients with or without cognitive fluctuations, visual hallucinations, delusions, recurrent falls, or parkinsonism. Patients with visual hallucinations had higher total LB burden than those without, but the comparison (p = 0.035) was not significant after Bonferroni correction. There were no significant relationships between the symptoms and regional LB densities in the logistic regression analysis.

Third, we analyzed whether visual hallucinations, delusions, or cognitive fluctuations would be related with more LB pathology in a specific paralimbic (entorhinal, cingulate gyrus, insula, or hippocampus) or neocortical (frontal or occipital) area. None of the symptoms were significantly associated with increased LB density in a specific area, as analyzed either by t-test comparisons or logistic regression. However, there was a trend toward the association of delusions with higher LB density in the cingulate cortex. The cases with delusions had an almost twofold higher mean LB density in that area (2.03 ± 1.48) compared with cases without delusions (1.04 ± 0.85, p = 0.08).

Finally, we assessed the correlation between disease duration and LB density. Disease duration did not correlate with LB density in any individual area. However, a total LB burden, calculated as the mean of nigral, paralimbic, and neocortical burdens, correlated with duration of the disease with r² = 0.22, p = 0.023 (slope ± SE = 0.05 ± 0.02).

LB gradient.

We found a very consistent pattern in the distribution of LB, both across brain regions and within subjects. LB density was highest in substantia nigra, followed by paralimbic areas in a consistent gradient (entorhinal cortex > cingulate gyrus > insula > hippocampus), and, finally, neocortex (frontal > occipital). This gradient was very similar in cases with or without concurrent AD and supports a different hierarchical vulnerability to LB formation across the brain. LB were distributed mainly within deep cortical layers. However, in some cases LB pathology was spread out across the whole cortical depth. We found a similar pattern of distribution of the LB pathology for each case across areas (either confined to deep layers or widespread), as if the pathogenic process were developing uniformly all throughout the cortex. The widespread pattern was not clearly dependent on a higher LB density or a longer disease duration. This specific, nonrandom distribution of the LB across brain regions and individuals may help to direct future studies to highlight the pathologic mechanisms underlying LB development and to identify phenotypes of neurons vulnerable to developing LB.

LB regional correlations.

There was a close parallel between paralimbic and neocortical LB densities, but neither of these measures were directly correlated with the amount of LB in substantia nigra. The relationship between a restricted LB pathology as in idiopathic PD and a widespread form as occurs in DLB is not clear. Both diseases are considered subtypes in the general spectrum of LB disorders. Many cases with idiopathic PD have a small number of cortical LB.31 It is tempting to hypothesize that DLB may occur as a severe or long-lasting PD, as less vulnerable regions became involved by the disease with increasing time. However, this may be an over-simplification, and some observations do not support this type of progression.

First, DLB is not a common occurrence in uncomplicated PD, nor is it a common substrate of parkinsonian dementia. Among the 30 to 40% of idiopathic PD patients who develop dementia not all have cortical LB, and these have been considered to be the likely cause of dementia in only 10% of the demented cases (no definitive cause and AD changes are more common).31 There is no clear threshold for cortical LB density that distinguishes between demented and nondemented PD. Furthermore, the neuropsychological profile of the typical dementia accompanying PD differs in some aspects from the one in DLB, suggesting a slightly different neuroanatomic substrate.29

Second, we found no correlation between cortical and subcortical LB pathology. We would have expected those cases with a higher cortical LB density to have a more severe involvement of substantia nigra (either as a higher LB count or as a lower LB density caused by severe neuronal loss). This lack of correlation might be due, in part, to different degrees of neuronal loss in cortical versus subcortical regions. The important neuronal loss in substantia nigra may lead to a floor effect when quantitating LB pathology in this area, especially in cases with a long disease duration. However, this does not seem to be a likely explanation in our group because LB density in substantia nigra was not correlated with disease duration, and the nine cases with disease duration longer than 10 years displayed a wide range of LB densities in substantia nigra.

Since cases with cortical but without brainstem LB are exceptional, it seems clear that dopaminergic neurons in substantia nigra are the primary and most vulnerable target for LB formation. However, an “extra” predisposition might be needed to also develop cortical LB. If this predisposition exists, the patient would be likely to have DLB from the beginning. The fact that widespread LB are found in patients with short duration of PD suggests the concurrent formation of LB across different brain areas.31 The weak correlation we found between LB density and disease duration, apparent only by calculating a “total brain” LB load but not analyzing any individual area, suggests that cortical LB accumulation may depend more on individual vulnerability than on disease duration.

LB do not correlate with tangles and plaques.

We did not find any correlation between cortical LB burden and the amount of neurofibrillary tangles or senile plaques, suggesting a relative independence in the formation or in the dynamic turnover of these pathologic structures. Our study included DLB cases with a wide range of Alzheimer’s changes. Alzheimer’s pathology coexists in most occasions with cortical LB, but previous reports show that these changes are different from those in pure AD. Most DLB cases had few or no cortical neurofibrillary tangles, and senile plaques are mainly of the diffuse type.2,17,18,32-35 Moreover, the distribution of Alzheimer’s changes does not show strict regional correlation with LB.1 The absence of significant cortical neuron loss in DLB,9-11 a constant feature in pure AD,36 further distinguishes the pathologic patterns of these entities. All these data support the distinction of DLB as a separate disease from AD.

Clinicopathologic correlations.

A total LB burden correlated with disease duration, but there was little evidence of a correlation between specific clinical symptoms and the amount of LB pathology in the specific regions we examined. The average duration of the disease in this group (9.4 ± 4.5 years) was longer than described in other series,1,2,10,32,37,38 mostly because the subgroup of patients who started with parkinsonism had a younger age at onset and disease duration more than 10 years. At least two of these patients would not have been included as having clinical DLB according to the recommendations of the Consortium on DLB7 because cognitive impairment developed more than 12 months after the parkinsonian features. However, many cases fulfilling neuropathologic criteria for DLB developed dementia several years after parkinsonism.37 The over-representation of men in our DLB patients is quite frequent in other series.38-39

Onset with cognitive impairment was more common than it was with parkinsonism. Overall, the clinical symptoms in this group reflect the spectrum previously described in other clinicopathologic series2,33,37-41 and support the possibility of distinguishing DLB from pure AD cases on a clinical basis. Dementia was present in all patients, parkinsonism in 75%, cognitive fluctuations in 66%, and psychiatric symptoms in 52%. Consistent with other authors,10 we did not find clinical differences between patients with pure DLB and those with concomitant AD.

We considered the clinical data in a global “yes/no” dichotomy to try to highlight associations between motor symptoms and substantia nigra or between dementia/psychiatric symptoms and cortical LB, but none of the correlations were statistically significant. Some methodologic advantages in this study include the quite large sample of pure DLB cases, the α-synuclein immunostaining that makes a clear distinction between LB and globose tangles,30 and a counting procedure that allows an unbiased and systematic counting of structures throughout large areas. In other studies the quantitation of LB has generally been based on a targeted sampling methodology, that is, the microscopic field with the highest density is identified and counted. This sampling scheme has the advantage of targeting relatively rare lesions but the disadvantage of not being “random” in a statistical sense, and thus counts are subject to upward bias, especially when the structures to count, as is the case with LB, are not distributed uniformly.

Our sampling of neocortical regions was limited. Some cognitive and psychiatric symptoms may be anatomically multifocal or diffuse-related, and clinicopathologic correlations may not be as straightforward as with parkinsonism and pathology in substantia nigra. Even in this case, it remains puzzling that a certain percentage of DLB patients, in our group and also in other series,10,38,40 do not present parkinsonian features when the substantia nigra is as affected by LB as it is in idiopathic PD cases. However, parkinsonian features are different in these entities, with DLB cases showing milder and more symmetric signs, less tremor, and minor response to l-dopa.2,42

Other studies have approached the clinicopathologic correlations in DLB by assessing the correlation between cortical LB and dementia severity, either as a functional or as a neuropsychological score, with contradictory results. Perry et al.2,43 and Gómez-Isla et al.44 did not find any correlation. Churchyard and Lees45 did not find a correlation between mental status and LD density in cingulate gyrus, but this was well correlated with the amount of Lewy neurites in hippocampus. In contrast, other groups have found a significant correlation between dementia severity and neocortical LB counts.46-49 In some of these studies the significance relies on the most severely demented cases having a higher cortical LB density. The possibility that cortical LB may be just another marker of a widespread pathologic process, including selective neuronal loss, neuritic dysfunction in crucial pathways, and neurochemical imbalances, makes it more difficult to clarify their contribution to the clinical syndrome in DLB.