Tsuang D, Leverenz JB, Lopez OL, Hamilton RL, Bennett DA, Schneider JA, Buchman AS, Larson EB, Crane PK, Kaye JA, Kramer P, Woltjer R, Trojanowski JQ, Weintraub D, Chen-Plotkin AS, Irwin DJ, Rick J, Schellenberg GD, Watson GS, Kukull W, Nelson PT, Jicha GA, Neltner JH, Galasko D, Masliah E, Quinn JF, Chung KA, Yearout D, Mata IF, Wan JY, Edwards KL, Montine TJ, Zabetian CP. APOE ϵ4 Increases Risk for Dementia in Pure Synucleinopathies. JAMA Neurol. 2013;70(2):223-228. doi:10.1001/jamaneurol.2013.600
Author Affiliations: Veterans Affairs Puget Sound Health Care System (Drs Tsuang, Leverenz, Watson, Mata, and Zabetian, and Ms Yearout); Departments of Psychiatry and Behavioral Sciences (Drs Tsuang, Leverenz, and Watson), Neurology (Drs Leverenz, Mata, and Zabetian, and Ms Yearout), Medicine (Drs Larson and Crane), Epidemiology (Drs Kukull and Edwards, and Ms Wan), and Pathology (Dr Montine), University of Washington; Group Health Cooperative (Dr Larson), Seattle, Washington; Departments of Neurology (Dr Lopez) and Pathology (Dr Hamilton), University of Pittsburgh, Pittsburgh; Departments of Pathology and Laboratory Medicine (Drs Trojanowski, Irwin, and Schellenberg), Neurology (Drs Weintraub, Chen-Plotkin, Irwin, and Rick), and Psychiatry (Dr Weintraub), and Institute on Aging (Dr Trojanowski), University of Pennsylvania, Philadelphia; Philadelphia Veterans Affairs Medical Center (Dr Weintraub), Pennsylvania; Departments of Neurological Sciences (Drs Bennett and Buchman) and Pathology (Dr Schneider), Rush University, Chicago, Illinois; Departments of Neurology (Drs Kaye, Kramer, Quinn, and Chung) and Pathology (Dr Woltjer), Oregon Health and Science University; Portland Veterans Affairs Medical Center (Drs Quinn and Chung), Portland, Oregon; Departments of Pathology (Drs Nelson and Neltner) and Neurology (Dr Jicha), University of Kentucky, Lexington; and Departments of Neurology (Dr Galasko) and Pathology (Dr Masliah), University of California, San Diego.
Objective To test for an association between the apolipoprotein E (APOE) ϵ4 allele and dementias with synucleinopathy.
Design Genetic case-control association study.
Setting Academic research.
Patients Autopsied subjects were classified into 5 categories: dementia with high-level Alzheimer disease (AD) neuropathologic changes (NCs) but without Lewy body disease (LBD) NCs (AD group; n = 244), dementia with LBDNCs and high-level ADNCs (LBD-AD group; n = 224), dementia with LBDNCs and no or low levels of ADNCs (pure DLB [pDLB] group; n = 91), Parkinson disease dementia (PDD) with no or low levels of ADNCs (n = 81), and control group (n = 269).
Main Outcome Measure The APOE allele frequencies.
Results The APOE ϵ4 allele frequency was significantly higher in the AD (38.1%), LBD-AD (40.6%), pDLB (31.9%), and PDD (19.1%) groups compared with the control group (7.2%; overall χ24 = 185.25; P = 5.56 × 10−39), and it was higher in the pDLB group than the PDD group (P = .01). In an age-adjusted and sex-adjusted dominant model, ϵ4 was strongly associated with AD (odds ratio, 9.9; 95% CI, 6.4-15.3), LBD-AD (odds ratio, 12.6; 95% CI, 8.1-19.8), pDLB (odds ratio, 6.1; 95% CI, 3.5-10.5), and PDD (odds ratio, 3.1; 95% CI, 1.7-5.6).
Conclusions The APOE ϵ4 allele is a strong risk factor across the LBD spectrum and occurs at an increased frequency in pDLB relative to PDD. This suggests that ϵ4 increases the likelihood of presenting with dementia in the context of a pure synucleinopathy. The elevated ϵ4 frequency in the pDLB and PDD groups, in which the overall brain neuritic plaque burden was low, indicates that apoE might contribute to neurodegeneration through mechanisms unrelated to amyloid processing.
Lewy body disease (LBD) encompasses a spectrum of clinicopathologic entities that include Parkinson disease (PD), PD with dementia (PDD), and dementia with Lewy bodies (DLB). Dementia with Lewy bodies and PDD are differentiated from one another based on clinical criteria. Dementia with Lewy bodies is diagnosed when dementia occurs before or concurrently with parkinsonism, whereas in PDD, parkinsonism precedes dementia by at least 12 months.1 Lewy body disease neuropathologic changes (NCs) include classic histologic inclusions (Lewy bodies) and α-synuclein immunopositive neuronal inclusions and processes (Lewy neurites) in partially overlapping regions of the brain. However, the pathologic classification of DLB is complex because some cases show LBDNCs with no or low levels of Alzheimer disease (AD) NCs, which we refer to as pure DLB (pDLB), while many other cases show LBDNCs with coexistent high levels of ADNCs (LBD-AD). Importantly, the pathophysiologic relationship between LBD-AD, pDLB, and PDD has not been delineated, and whether these disorders share common risk factors remains unclear.
Humans are unlike other mammals in that we possess 3 common alleles of the apolipoprotein E (APOE) gene that are determined by 2 single nucleotide polymorphisms located in exon 4 at positions 3937 (T/C; rs429358) and 4075 (C/T; rs7412). The corresponding apoE isoforms (299 amino acids) differ at amino acid positions 112 (Cys for apoE2 and apoE3; Arg for apoE4) and 158 (Cys for apoE2; Arg for apoE3 and apoE4), and these isoforms have different functional and biochemical properties.2 The APOE ϵ4 allele is a well-established risk factor for both early-onset and late-onset AD.3,4 We and others have reported an association between ϵ4 and LBD-AD,5- 7 but it is unclear whether ϵ4 is a risk factor for pDLB or PDD because interpretations of existing data are limited by methodologic differences between studies. In particular, studies of DLB have often failed to assess for the presence of coexistent ADNCs, thus they have been unable to differentiate LBD-AD from pDLB.8- 10 Therefore, it is possible that all genetic risk for DLB associated with the APOE ϵ4 allele is related to its frequent comorbidity with ADNCs and is unrelated to LBDNCs. Furthermore, no studies have directly compared genetic risk factors between pDLB and PDD.
To address this gap in knowledge, we genotyped APOE in a clinically and neuropathologically well-characterized sample of control participants and subjects with AD, LBD-AD, pDLB, and PDD.
The study population comprised 640 patients with dementia and 269 control subjects who all reported their race as white. Subjects with dementia but without a clinical diagnosis of PD were enrolled in 1 of 7 AD centers (ADCs) (Oregon Health and Science University; Rush University; University of California, San Diego; University of Kentucky; University of Pennsylvania; University of Pittsburgh; and University of Washington), in the Rush Memory and Aging Project, or in the Group Health/University of Washington Alzheimer Disease Patient Registry(ADPR)/Adult Changes in Thought (ACT) study. The ADPR/ACT study is a community-based longitudinal study that enrolled individuals aged 65 years or older with dementia (ADPR11) or without dementia (ACT12) from a health-maintenance organization in the Seattle area. Expert diagnosticians at the ADCs, Rush Memory and Aging Project, and ADPR/ACT reviewed subject clinical history, physical examination findings, and neuropsychologic test results at a consensus diagnosis conference. These individuals all received a clinical diagnosis of probable AD, possible AD, or dementia (type unknown), according to the National Institute of Neurological and Communicative Disorders and Stroke and the AD and Related Disorders Association criteria.13
Subjects with PDD were enrolled in studies at the University of Washington; Oregon Health and Science University; the University of Pennsylvania; or the Veterans Affairs Northwest or Philadelphia Parkinson Disease Research, Education, and Clinical Centers. All patients satisfied UK PD Society Brain Bank clinical diagnostic criteria for PD,14 met criteria recently proposed by a Movement Disorder Society Task Force for probable dementia associated with PD,15 and had onset of parkinsonism more than 1 year before the diagnosis of dementia. Data on 57 of the PDD cases have been published elsewhere.16
Control subjects were elderly adults who were enrolled in longitudinal studies of normal aging as part of the ADPR/ACT study, the Rush Memory and Aging Project, or 6 ADCs (Oregon Health and Science University; Rush University; University of California, San Diego; University of Kentucky; University of Pittsburgh; and University of Washington). All control subjects were cognitively normal at study entry, as determined by a clinical consensus that considered both clinical history and neuropsychologic testing. Control subjects remained free of cognitive impairment at their last evaluation, as indicated by a detailed clinical assessment (Rush ADC and Rush Memory and Aging Project),17 a Cognitive Abilities Screening Instrument18 score greater than 85 (ADPR/ACT), or a Clinical Dementia Rating score less than 1 and Mini-Mental State Examination score greater than 26 (all other centers). None of the control subjects had a clinical diagnosis of a neurodegenerative disease, including PD. All control subjects were autopsied and underwent their last clinical evaluation within 3 years of death.
All study procedures were approved by the institutional review boards at each participating site.
All subjects underwent a standard histologic evaluation using hematoxylin and eosin, modified Bielschowsky, and thioflavine S methods. Immunostaining for LBDNCs was also performed on all subjects using α-synuclein immunohistochemistry with antibody LB509 (1:50 to 1:400; Zymed), as previously described.19 Cases with questionable LB509 immunoreactivity were evaluated with a second antibody to nitrated α-synuclein (syn 303, 1:1000).20 A section from each of the following 5 regions was assessed for LBDNCs: medulla, substantia nigra, amygdala, cingulate gyrus, and frontal cortex. The only exception was that 3 (substantia nigra, cingulate gyrus, and frontal cortex) rather than 5 regions were assessed for LBDNCs in cases and control subjects from Rush University Medical Center. Braak staging21 for neurofibrillary tangles and Consortium to Establish a Registry for Alzheimer Disease plaque score22 were performed using modified Bielschowsky–stained sections, tau immunohistochemistry (AT8, Endogen, 1:250, or PHF-1, 1:1000), or both.
Subjects were classified as PDD (n = 81), pDLB (n = 91), LBD-AD (n = 224), AD (n = 244), or control participants (n = 269) based on ADNCs, LBDNCs, and the aforementioned clinical criteria. The AD group was defined by the presence of high-level ADNCs (Braak stage 4, 5, or 6, and a Consortium to Establish a Registry for Alzheimer Disease plaque score of moderate or frequent) but no LBDNCs. The LBD-AD group was defined by the presence of both high-level ADNCs and limbic or neocortical stage LBDNCs.19 The PDD and pDLB groups were defined by the same neuropathologic criteria: the presence of limbic or neocortical stage LBDNCs and no or low levels of ADNCs.16,19 The control group showed no evidence of LBDNCs or high-level ADNCs. Note that all cases with ADNCs that did not meet criteria for being high level were classified as low level, and we did not include an intermediate level to maximize statistical power.
DNA was extracted from peripheral leukocytes or brain tissue using standard methods. The presence of the APOE ϵ2, ϵ3, and ϵ4 alleles was determined by genotyping rs429358 and rs7412 by TaqMan Assay (Applied Biosystems).
Age at death and age at onset of dementia were compared among the groups using analysis of variance. A Pearson χ2 test was used to evaluate group differences in sex proportion and APOE ϵ4 allele frequency. Multinomial logistic regression was used to test for an association between the ϵ4 allele and each of the 4 dementia groups under a dominant model using control subjects as the reference. Regression analyses were performed with and without adjustment for age at death and sex. A test for a trend in the odds of carrying the ϵ4 allele by disease group was performed using the score trend test, adjusting for age and sex using Mantel-Haenszel odds ratios (ORs). Analyses were performed using STATA version 11.2 and R version 2.13.0.
The demographic and clinical characteristics of the study population are shown in Table 1. Control subjects were slightly older at death than those in all 4 dementia groups, and a substantially greater proportion of subjects with PDD and pDLB were male (P < 1 × 10-4 for all). There was no significant difference in age at onset of dementia among the case groups.
The APOE genotype and allele frequencies observed in the sample are presented in Table 2. There was a highly significant overrepresentation of the ϵ4 allele in all case groups, including PDD (χ21 = 18.25; P = 1.94 × 10−5) and pDLB (χ21 = 68.61; P = 1.2 × 10−16). In a dominant genetic model adjusting for sex and age at death with control subjects as the reference, the ORs were 9.9 (95% CI, 6.4-15.3) for AD, 12.6 (95% CI, 8.1-19.8) for LBD-AD, 6.1 (95% CI, 3.5-10.5) for pDLB, and 3.1 (95% CI, 1.7-5.6) for PDD (Table 3).
Within the groups with synucleinopathy, the ϵ4 allele frequency was significantly lower in subjects with PDD (19.1%) than those with pDLB (31.9%; χ21 = 6.60; P = .01) or LBD-AD (40.6%; χ21 = 23.24; P = 1.43 × 10−6).
Inspection of the ORs suggested that for individuals with PDD and pDLB, the ϵ4 allele might convey risks intermediate between the control group and the LBD-AD and AD groups. We assessed this hypothesis using a trend test, coding control subjects as 0, PDD as 1, pDLB as 2, and AD and LBD-AD as 3. The results indicated an increased likelihood of carrying the ϵ4 allele across groups (control<PDD<pDLB<LBD-AD/AD [χ21 = 166.47; P = 4.37 × 10−38]).
We observed a strong overall association between the APOE ϵ4 allele and AD, LBD-AD, pDLB, and PDD. While the ϵ4 allele is a well-established risk factor for AD3,4 and several studies have reported an association between APOE and LBD-AD,5- 7 the finding that ϵ4 was overrepresented in pDLB and PDD was unexpected. By far, the most widely held view for the mechanism by which APOE ϵ4 influences AD risk is that apoE isoforms have direct or indirect effects on amyloid-β (Aβ) peptide metabolism. There is a large body of evidence to support this hypothesis, including experiments in mice overexpressing human apoE isoforms in which the ϵ4 allele results in lower Aβ clearance and increased deposition of insoluble Aβ in the brain than the ϵ2 and ϵ3 alleles.23,24 Thus, in humans, ϵ4 is predicted to accelerate the accumulation of neurotoxic Aβ, which ultimately leads to neuritic plaque formation and neurodegeneration. However, our observation of an elevated ϵ4 frequency in the pDLB and PDD groups in which the overall brain neuritic plaque burden is low suggests the possibility that apoE isoforms also might modulate neurodegeneration by nonamyloidogenic mechanisms. A caveat is that we were not able to account for the potential influence of ϵ4 on soluble oligomeric Aβ levels because soluble oligomeric Aβ might not be well represented by measures of histologically detectable Aβ plaques, and soluble oligomeric Aβ could adversely affect cognition. Nonetheless, evidence that Aβ-independent pathways might exist is beginning to emerge from work in model systems. For example, C-terminal–truncated fragments of apoE4 are neurotoxic in vivo and in vitro, possibly through impairment of mitochondrial function or disruption of the cytoskeleton.25,26 In the presence of lipids, apoE4 impairs neuronal plasticity in vitro.25 Also, greater microglia-mediated neurotoxicity has been observed in mice expressing human apoE4 than other apoE isoforms.27
Interpretation of previous studies of APOE in DLB has been challenging given the wide difference in methods and diagnostic criteria used between studies. Few neuropathologically verified studies have reported separate APOE frequencies in LBD-AD and pDLB,5,6,28 and some did not collect sufficient information to exclude subjects with PDD.28 Furthermore, the sample size of the pDLB group in these previous studies ranged from 6 to 18 subjects, and the ϵ4 allele frequency varied from 6% to 22%, precluding firm conclusions about the association of APOE with pDLB. On the other hand, a number of studies have reported a significant overrepresentation of APOE ϵ4 in subjects with LBD-AD, with ϵ4 allele frequencies ranging from 29% to 47%.5- 7,28,29 Finally, several studies have examined APOE in DLB without making a distinction between LBD-AD and pDLB, often because the diagnosis was solely based on clinical criteria.8- 10
Data from genomewide association studies indicate that APOE is not a susceptibility gene for PD.30- 32 While PD is clinically defined by motor symptoms, more than 50% of patients develop dementia within 10 years of diagnosis.33,34 Whether APOE acts as a modifier gene by influencing the manifestation of cognitive dysfunction in PD is still a matter of debate. There are several possible approaches to address this question, including assessing whether APOE genotypes (1) differ in frequency between patients with PDD and control subjects or cognitively intact patients with PD, (2) influence the rate of progression to dementia in PD cohorts, or (3) associate with performance on cognitive testing in cross-sectional or longitudinal studies of patients with PD. In a meta-analysis of 17 studies, Williams-Gray and colleagues35 reported a significantly higher APOE ϵ4 frequency in patients with PDD (n = 501) compared with nondemented patients with PD (n = 1145; OR, 1.74; 95% CI, 1.36-2.23). However, interpretation of the results was made difficult because there was evidence of significant heterogeneity of ORs and publication bias, and the criteria used to define PDD varied substantially across individual studies. Williams-Gray and colleagues35 also longitudinally assessed an incident cohort of 107 patients with PD for 5 years and found no effect of APOE ϵ4 on the risk for dementia or rate of cognitive decline. In contrast, a recent longitudinal study of 212 patients with PD reported that ϵ4 carriers displayed a more rapid decline in total score on the Mattis Dementia Rating Scale than noncarriers.36 In a cross-sectional study of 937 patients with PD, we observed that the ϵ4 allele is associated with lower psychometric test scores across multiple cognitive domains after adjusting for disease duration.37 Together with findings from the present study, we believe that the preponderance of the evidence indicates that APOE is a risk factor for cognitive dysfunction in PD. However, in our data set, the magnitude of the APOE ϵ4 effect was smaller for PDD than for LBD-AD and pDLB, as evidenced by the 2-fold (or more) larger ORs observed in the LBD-AD and pDLB groups (Table 3) and the results of the trend test. One explanation for these findings is that in the presence of synucleinopathy, APOE ϵ4 might increase the likelihood that dementia precedes parkinsonism, hence a clinical diagnosis of DLB rather than PDD is rendered.
This study had some limitations. Because most of the subjects died prior to the publication of the consensus clinical criteria for DLB, we were unable to fully apply these criteria retrospectively, in particular the fluctuating cognition and rapid eye movement sleep behavior disorder features.1 Although we limited the sample to white subjects, we did not have data available for ancestry informative genetic markers, thus we cannot entirely exclude the possibility of unrecognized population structure in our data set.
We have shown that APOE ϵ4 increases the risk for dementia in subjects with LBDNCs but with no or low levels of ADNCs, which provides further rationale for exploring how apoE isoforms might modulate neurodegeneration through mechanisms unrelated to Aβ metabolism. Longitudinal studies comparing cerebrospinal fluid profiles in subjects clinically diagnosed as having DLB or PDD and stratified by APOE genotype and amyloid burden by in vivo imaging might complement work in model systems to address this question in the future.
Correspondence: Cyrus P. Zabetian, MD, MS, VA Puget Sound Health Care System, GRECC S-182, 1660 S Columbian Way, Seattle, WA 98108 (firstname.lastname@example.org).
Accepted for Publication: July 13, 2012.
Published Online: November 19, 2012. doi:10.1001/jamaneurol.2013.600
Author Contributions: Drs Tsuang and Leverenz contributed equally to this work. Study concept and design: Tsuang, Leverenz, Lopez, Trojanowski, Galasko, and Zabetian. Acquisition of data: Tsuang, Leverenz, Hamilton, Bennett, Schneider, Larson, Crane, Kaye, Kramer, Woltjer, Trojanowski, Chen-Plotkin, Irwin, Rick, Schellenberg, Kukull, Nelson, Jicha, Neltner, Galasko, Masliah, Quinn, Chung, Yearout, Montine, and Zabetian. Analysis and interpretation of data: Leverenz, Buchman, Trojanowski, Weintraub, Irwin, Watson, Yearout, Mata, Wan, Edwards, and Montine. Drafting of the manuscript: Tsuang, Hamilton, Buchman, Trojanowski, Yearout, and Zabetian. Critical revision of the manuscript for important intellectual content: Tsuang, Leverenz, Lopez, Bennett, Schneider, Buchman, Larson, Crane, Kaye, Kramer, Woltjer, Trojanowski, Weintraub, Chen-Plotkin, Irwin, Rick, Schellenberg, Watson, Kukull, Nelson, Jicha, Neltner, Galasko, Masliah, Quinn, Chung, Yearout, Mata, Wan, Edwards, and Montine. Statistical analysis: Trojanowski, Mata, Wan, and Edwards. Obtained funding: Tsuang, Leverenz, Bennett, Buchman, Larson, Crane, Kaye, Schellenberg, Galasko, Montine, and Zabetian. Administrative, technical, and material support: Leverenz, Hamilton, Schneider, Larson, Kaye, Woltjer, Trojanowski, Irwin, Schellenberg, Kukull, Nelson, Jicha, Neltner, Masliah, Yearout, and Mata. Study supervision: Tsuang, Lopez, Trojanowski, Schellenberg, Edwards, Montine, and Zabetian.
Conflict of Interest Disclosures: Dr Schneider's work was funded by grants from the National Institutes of Health.
Funding/Support: This work was supported by grant 1I01BX000531 from the Department of Veterans Affairs and grants P30 AG008017, P30 AG028383, P30 AG010124, P30 AG010161, P50 NS053488, P50 AG005131, P50 NS062684, P50 AG005136, P50 AG005133, R01 NS048595, R01 NS065070, R01 AG010845, and U01 AG006781 from the National Institutes of Health.
Additional Contributions: We thank Lynne Greenup for her expert technical assistance.