Tumor Necrosis Factor α and Interleukin 10 Promoter Region Polymorphisms and Risk of Late-Onset Alzheimer Disease FREE

Alzheimer disease (AD) is the most common form of dementia in individuals older than 65 years and presents a substantial public health problem. The ultimate cause of dementia is the loss of neurons and the synaptic connections between them. However, the insults that produce this loss remain to be determined. The innate immune response and resulting neuroinflammation may be important in mediating neuronal damage in AD. This response includes the activation of microglia, astrocytes, and the complement system, as well as increased cytokine expression and acute-phase protein response.^{1 - 2} The harmful effect of the inflammatory response is supported by epidemiological evidence of a potential protective effect of anti-inflammatory agents for AD.^{3 - 4} In addition, evidence suggests that the risk of AD is affected by genetic variation in inflammatory modulators, such as interleukin 1α (IL-1α), IL-1β, IL-6, tumor necrosis factor α (TNF-α), α₂-macroglobulin, and α₁-antichymotrypsin.⁵ Nucleotide variations in genes encoding these molecules could affect biological activity by regulating transcription, translation, and secretion. Therefore, factors modulating the degree of inflammatory damage might influence the risk of AD.

In the inflammatory response, TNF-α and IL-10 have opposing roles. While TNF-α is generally proinflammatory, IL-10 limits and terminates inflammatory reactions.⁶ Interleukin 10 is a potent suppressor of TNF-α, IL-1α, IL-1β, and IL-6, all of which have been investigated for their potential association with AD.⁷ Furthermore, in vitro studies^{8 - 9} have shown that IL-10 suppresses amyloid β peptide– or lipopolysaccharide-induced inflammatory cytokines in rat and murine microglia. Several polymorphisms in TNF-α were found to be associated with AD when siblings from high-risk families were evaluated.¹⁰ The effect of IL-10 polymorphisms on AD risk also has been reported.^{11 - 14} These study results vary significantly, and the variation might be attributed to small sample sizes and differences in study populations. In this study, we genotyped 4 single nucleotide polymorphisms (SNPs) located in the promoter regions of TNF-α and IL-10 and analyzed their associations, as individual SNPs and as haplotypes, with AD in a community-based case-control series.

The subjects evaluated in this analysis were part of a large case-control study (Genetic Differences in Alzheimer Disease Cases and Controls) whose population base was a local health maintenance organization (Group Health Cooperative of Puget Sound). This is a stable and geographically defined organization whose members are representative of the surrounding Seattle area with respect to sex, race/ethnicity, and age distribution but have a slightly higher proportion of persons with education beyond high school.¹⁵ Newly recognized (incident) cases (n = 390) were enrolled in the University of Washington/Group Health Cooperative of Puget Sound Alzheimer's Disease Patient Registry¹⁶ between January 1987 and December 1996 and met (1) the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition criteria for dementia¹⁷ and (2) the National Institute of Neurological and Communication Disorders and Stroke–Alzheimer's Disease and Related Disorders Association criteria for probable or definite AD.¹⁸ Control subjects (n = 379) were selected by simple random sampling of the Group Health Cooperative of Puget Sound population and were matched on sex and age to the case series. To qualify for study inclusion, controls were required to achieve a score of 28 of 30 on the Mini-Mental State Examination (27 if >80 years old) and to have no previously diagnosed dementia or a disease-causing dementia. All subjects received annual follow-up visits consisting of cognitive tests, behavioral assessment, and medical history updates and gave written informed consent to participate.

Following standard procedures, DNA was prepared from blood samples that were previously collected and stored at −70°C. A Bs1-I–based mismatched polymerase chain reaction–restriction fragment length polymorphism assay was used for genotyping of IL-10 −592, IL-10 −1082, TNF-α −308, and TNF-α −863 SNPs as previously described.^{19 - 20} Genotyping was completed for 660 of 769 initially enrolled subjects (298 cases and 362 controls). The remaining 109 subjects did not have genotyping results for any of the 3 markers because (1) they or their families declined giving a blood sample, (2) they discontinued study participation before a sample was collected, (3) the study ended before a sample was taken, or (4) their samples failed the genotyping reaction.

Apolipoprotein E (APOE) ε4 genotypes for 316 cases and 368 controls had been previously determined by polymerase chain reaction.^{21 - 23} The APOE genotypes were unavailable for 85 subjects for the same 4 reasons just enumerated.

The Genetic Data Analysis software package²⁴ was used to evaluate Hardy-Weinberg equilibrium for each of the 4 SNPs and linkage disequilibrium (LD) between the 2 polymorphisms of each loci. PHASE v 1.0 (http://www.stat.washington.edu/stephens/software.html) was used to infer TNF-α haplotypes from the available TNF-α −308 and TNF-α −863 genotype data.²⁵ Crude odds ratios were calculated and stratified by the presence of any APOE ε4 allele to assess effect modification (ie, interaction between APOE and TNF-α). Adjusted logistic regression analysis was used to evaluate the risk of AD when considering intake age, education, and the presence of any APOE ε4 allele. The test for trend was used to evaluate the risk across the genotype categories of each polymorphism and the TNF-α haplotype. STATA version 7.0 (StataCorp LC, College Station, Tex) was used for all statistical analyses except Hardy-Weinberg equilibrium, LD, and haplotype estimation. To avoid population stratification concerns, the data analyses were restricted to white subjects, reducing the sample size to 265 cases and 347 controls.

Relevant characteristics of the cases and controls are given in Table 1. As expected, the cases and controls differ significantly by education, initial Mini-Mental State Examination score, and APOE genotype. The mean ± SD Mini-Mental State Examination scores were 21.0 ± 6.9 for cases and 28.2 ± 1.6 for controls. Slightly more than 58% (58.2%) of cases and 26.5% of controls had at least 1 copy of the APOE ε4 allele.

For each polymorphism, allele and genotype frequencies were determined by gene counting and are given in Table 2 and Table 3. The TNF-a −863 minor allele (A) was present in 16.9% of controls and in 12.6% of cases (Table 2). Our stratified analysis did not detect an interaction effect between TNF-α and APOE ε4; therefore, we considered APOE ε4 as a potential confounder and not as an effect-modifying variable. The test for trend (adjusted for age, education, and ε4 status) was significant when comparing the 3 genotypes (odds ratios [C/C, reference], 0.66 for C/A and 0.58 for A/A; P = .04), suggesting a dose-response (ie, additive) allele affect.

We found no notable differences in allele or genotype frequencies for the TNF-α −308 variant, nor did we detect a difference in age at onset of dementia symptoms in carriers vs noncarriers of the A allele (data not shown), as reported in a previous study.²⁶ However, although nonsignificant, the test for trend demonstrated a trend toward increased risk of AD with each additional −308 A allele (odds ratios [G/G, reference], 1.10 for G/A and 2.12 for A/A; P = .22). The IL-10 −1082 and IL-10 −592 allele and genotype frequencies were not significantly different between cases and controls (Table 3) and did not warrant additional analysis.

Because of LD in the TNF-α region (P = .04, Genetic Data Analysis exact test for LD), we used PHASE v 1.0 to reconstruct TNF-α promoter region haplotypes as defined by the SNPs at positions −863 and −308. The −863 A allele was exclusively associated with the −308 G allele, resulting in 3 observed haplotypes (A-G, C-G, and C-A) and 6 corresponding genotypes. Haplotype frequencies are given in Table 4. Based on knowledge of TNF-α protein production, we ordered the haplotypes from apparent least transcriptional activity (−863 A/A and −308 G/G) to highest transcriptional activity (−863 C/C and −308 A/A). After adjusting for age, education, and the presence of any APOE ε4 allele, the test for trend showed significantly increasing odds of AD with increasing transcriptional activity (P = .02). We had small sample sizes in some subgroups, and our confidence intervals for individual odds ratios reflect this limited power.

We studied functional variants in the promoter regions of 2 cytokine genes (TNF-α and IL-10) to assess their relationship with late-onset AD. Our results suggest an association between TNF-α promoter region polymorphisms and the risk of AD. Individuals with 1 or 2 copies of the TNF-α −863 A allele were found to have reduced odds of developing AD. We also observed a trend indicating that the minor allele, TNF-α −308 A, was associated with an increased risk of AD. Last, our results reflect a relationship between the risk of AD and the TNF-α promoter region haplotypes that are associated with higher transcriptional activity.

The TNF-α promoter region polymorphisms have been implicated as AD susceptibility sites in previous studies.^{10 ,26 - 28} The results from our community-based case-control series add to this evidence suggesting that genetic modulation of TNF-α production affects AD susceptibility. However, it is possible that the nucleotide substitution in these 2 sites might represent a marker in LD with other genetic elements that are affecting the risk of AD. TNF-α is located within a putative AD-associated region at chromosome 6p21.3.²⁹ This region also contains the major histocompatibility complex and HLA loci. Genetic variation in these loci might contribute to the modulation of microglial activity and inflammatory responses. Numerous TNF-α polymorphisms were found to be in LD with HLA class I and II alleles,^{30 - 31} and the association of HLA alleles with AD has been demonstrated.³²

The TNF-α −863 A allele has been linked with 31% lower transcriptional activity in reporter gene studies and with significantly lower serum TNF-α levels.^{30 ,33} In our data, this allele appears to have a protective effect that is independent of APOE genotype status. Inconsistent results have been reported on how the TNF-α −308 G/A substitution affects TNF-α production,³⁰ although in approximately half of instances the A allele was associated with higher transcriptional activity.³⁴ In addition, this allele has been associated with an earlier age at onset of AD.^{10 ,26} In our study sample, TNF-α −308 A/A was associated with a higher risk of AD compared with the G/G reference group (odds ratio, 2.12), but this association was not statistically significant (P=.22). In view of the strong LD between the polymorphisms in this region, we believe that studying the role of isolated SNPs may not be informative. Indeed, a more powerful haplotype analysis (PHASE v 1.0) revealed a significant trend of an increased risk of AD for individuals possessing the TNF-α −863 C and TNF-α −308 A alleles (C-A haplotype) (P=.02).

Our results indicate that TNF-α promoter region variation, or possibly polymorphisms in nearby genes, affects the risk of AD. The data support that therapeutic strategies designed to reduce TNF-α protein production or activity might be a valuable treatment for AD. Further insight into the mechanism underlying this relationship may suggest new strategies for reducing the intensity of the inflammatory response and attenuating the progression of AD.

Correspondence: Lee-Way Jin, MD, PhD, M.I.N.D. Institute and Department of Pathology, University of California–Davis, 2805 50th St, Sacramento, CA 95817 (lee-way.jin@ucdmc.ucdavis.edu).

Accepted for Publication: February 10, 2006.

Author Contributions:Study concept and design: Lin, Larson, Schellenberg, and Jin. Acquisition of data: Lin, Larson, Maezawa, Tseng, Schellenberg, Hansen, and Kukull. Analysis and interpretation of data: Ramos, Lin, Edwards, Hansen, Kukull, and Jin. Drafting of the manuscript: Ramos, Tseng, and Jin. Critical revision of the manuscript for important intellectual content: Ramos, Lin, Larson, Maezawa, Edwards, Schellenberg, Hansen, Kukull, and Jin. Statistical analysis: Ramos, Edwards, and Kukull. Obtained funding: Larson, Schellenberg, Jin, Kukull. and Hansen. Administrative, technical, and material support: Lin, Larson, Maezawa, Tseng, Schellenberg, Hansen, and Jin. Study supervision: Lin, Schellenberg, Hansen, Kukull, and Jin.

Funding/Support: This study was supported by grants AI49213, AI33484, CA18029, and CA15704 from the US Public Health Service; by grants AG07584 and AG06781 from the National Institute on Aging; and by an M.I.N.D. Institute start-up fund (Dr Jin).