Associations between polymorphisms in five inflammation-related genes and cognitive ability in older persons
Riccardo E. Marioni and Jackie F. Price, Public Health Sciences Section, Division of Community Health Sciences, The University of Edinburgh, Medical School, Teviot Place, Edinburgh EH8 9AG, Scotland, UK. E-mails: R.E.Marioni@sms.ed.ac.uk; J.Price@ed.ac.uk
Several studies have found associations between inflammatory biomarker levels and cognitive ability. This study tested the relationship between polymorphisms in genes that are associated with or encode the biomarkers and cognitive ability and estimated lifetime cognitive change. Data came from the aspirin for asymptomatic atherosclerosis trial (n = 2091, mean age = 67.2 years ). Twelve single nucleotide polymorphisms (SNPs) were genotyped from five genes (IL-1α, IL-1β, IL-6, HNF-1A and F13A1). Cognition was assessed via administration of a five-test battery of psychometric tests, which were used to derive a general intelligence factor, g. A vocabulary-based cognitive test was also administered and adjusted for in the analysis to enable an estimation of lifetime cognitive change. Age- and sex-adjusted analyses yielded one weakly significant association between the IL-1α rs2856838 SNP and a measure of mental flexibility/processing speed (P = 0.044). Adjustment for the vocabulary-based scores resulted in a single, significant association between the IL-1α rs3783546 SNP and a measure of processing speed (P = 0.048). There is little evidence to suggest an association between SNPs in the inflammation-related genes IL-1α, IL-1β, IL-6, TCF-1 and F13A1 and cognition in an elderly population of community-dwelling Scottish citizens.
There is an increasing evidence linking plasma levels of inflammatory biomarkers with cognitive ability in elderly persons (Alley et al. 2008; Dik et al. 2005; Luciano et al. 2009; Rafnsson et al. 2007; Schram et al. 2007). Most studies report small but consistent effect sizes of the order of 1% with raised marker levels associating with decreased cognitive ability. Whether these are causal relationships is unclear. In an attempt to identify causal determinants of cognitive ability and decline, studies can use genetic variants such as single nucleotide polymorphisms (SNPs) as predictor variables in their analyses. Given that genetic information is determined at conception, it is unlikely that variants will be affected by non-genetic confounder variables (Ebrahim et al. 2008).
Whilst many studies have investigated genetic variants with cognitive ability, few have focussed on SNP variants related to inflammation and inflammatory pathways. Of these studies, Marioni et al. 2010 found no relationship between C-reactive protein (CRP) polymorphisms and cognition in four Scottish cohorts. By contrast, studies by Trompet et al. (2008) and Baune et al. (2008) have found polymorphisms in the IL-1 gene to associate significantly with poorer cognition. The aim of this paper was to test the association between SNPs from five genes related to inflammatory pathways and cognitive function as part of a secondary data analysis on participants of the aspirin for asymptomatic atherosclerosis (AAA) trial. The genes under investigation included interleukin-1α (IL-1α), interleukin-1β (IL-1β), interleukin-6 (IL-6), transcription factor gene-1 (TCF-1) and coagulation factor XIII, A1 polypeptide (F13A1). These genes were selected for biological reasons. Specifically, they are all involved in inflammatory pathways and have been associated directly or indirectly with inflammatory biomarker levels, which in turn, have been associated with cognitive ability.
Interleukin-1 is composed of two distinct proteins IL-1 α and IL-1β. Both are coded at chromosome 2 and are pro-inflammatory components of the immune system. Polymorphisms from the IL-1 genes have been associated with systemic levels of inflammatory markers including CRP and IL-6 (Reiner, Wurfel, et al. 2008).
Interleukin-6 is a cytokine that can induce both pro- and anti-inflammatory responses.
It is produced by macrophages and T-cell lymphocytes and its presence helps stimulate the production of a large number of liver-based acute-phase proteins (Pradhan et al. 2001).
Transcription factor gene-1 (TCF-1) – also known as HNF1A– encodes the transcription factor hepatocyte nuclear factor (HNF)-1α (Reiner, Barber, et al. 2008). The HNF genes regulate the transcription of proteins involved with blood clotting factors, enzymes and transporters involved with glucose, cholesterol, fatty acid transport and metabolism (Armendariz & Krauss 2009). HNF-1α is expressed predominantly in the liver and has been associated with transactivation (increased gene expression triggered by endogenous cellular or viral proteins) of CRP (Toniatti et al. 1990).
The F13A1 gene is located on chromosome 6 and encodes coagulation factor XIII. After the conversion of fibrinogen into fibrin via the enzyme thrombin, factor XIII stabilises the newly formed mesh of fibrin by cross-linking the molecules (Mannila et al. 2007). This process results in the formation of clots.
Materials and methods
This secondary data analyses used information from 2091 participants from the AAA trial: a randomised controlled trial of aspirin for the reduction of cardiovascular events and death in people with asymptomatic atherosclerosis (Price, Stewart, Douglas, et al. 2008). Of the 3350 men and women aged over 50 years that were recruited into the trial between 1999 and 2001, 2312 underwent cognitive testing at 5-year follow-up. All subjects recruited had no history of cardiovascular disease but had a ratio of systolic blood pressure in the ankle to that in the arm (ankle brachial index) of 0.95 or less, which is indicative of atherosclerotic burden and increased risk of developing symptomatic cardiovascular disease (Ankle Brachial Index Collaboration 2008; Heald et al. 2006). In addition to symptomatic vascular disease or major illness, subjects were excluded if they had a contraindication to aspirin therapy (Price, Stewart, Deary, et al. 2008, Price, Stewart, Douglas, et al. 2008). The study population of 2091 consisted of subjects who completed three or more tests from a five-test cognitive battery and had DNA available for testing. Baseline demographic characteristics were similar to those of the 3350 original trial participants (Price, Stewart, Deary, et al. 2008). Subjects were living independently in the community and were free of symptomatic cardiovascular disease and dementia at baseline.
Genomic DNA was isolated from whole blood by standard procedure at the Wellcome Trust Clinical Research Facility Genetics Core, Western General Hospital, Edinburgh. Genotyping was carried out by KBioscience (Herts, UK) using their in-house chemistry of competitive allele specific polymerase chain reaction (KASPar). Twelve SNPs were typed from five genes: IL-1α (rs2856836, rs3783546, rs2856838), IL-1β (rs1143643, rs1143634), IL-6 (rs2069832, rs2069840, rs1800795), TCF-1 (rs1169292, rs1169301, rs2464196) and F13A1 (rs5985). The IL-1α SNPs are in high LD with three of the four polymorphisms needed in order to investigate the effect of known polymorphisms in the IL-1α gene (Knudsen et al. 2007). The IL-1β rs1143643 SNP is an eQTL while rs1143634 is a synonymous coding SNP. The three IL-6 SNPs are found in the same haplotype block and rs1800795 is believed to be of functional importance (Ng et al. 2008). The three TCF-1 SNPs have been associated with plasma fibrinogen levels (Soria et al. 2005), while rs5985 in the F13A1 gene is a non-synonymous coding SNP.
A five-test cognitive battery was administered to all participants at the trial's 5-year follow-up to assess: executive function [verbal fluency test (VFT) (Lezak 1995)]; processing speed [digit symbol test (DST) (Wechsler 1998)]; non-verbal reasoning [Raven's standard progressive matrices (RAVENS) (Raven et al. 1998)]; immediate and delayed memory [auditory verbal learning task (AVLT) (Lezak 1995)] and mental flexibility [trail making test, part B (TMT) (Spreen & Strauss 1991)]. A vocabulary-based measure of intelligence was assessed at baseline using the combined junior and senior Mill Hill vocabulary scale (MHVS) synonyms (Raven et al. 1998). The mini-mental state examination (MMSE) (Folstein et al. 1975) was included as a general mental assessment. Analyses were repeated to exclude potential dementia cases (MMSE score <24, n = 43) but this did not affect results.
Ethical approval was provided by the ethics committees of Lanarkshire, Lothian and Greater Glasgow Health Boards; all patients gave written, informed consent prior to participation.
Age- and sex-adjusted ancova was used to test for differences in late-life cognitive test scores by SNP genotypes. All cognitive variables were approximately normally distributed apart from the TMT, which was transformed using natural logarithms. Further adjustment for scores on the MHVS enabled an estimate of lifetime cognitive change. Scores on vocabulary-based tests are found to vary little over a lifetime (Crawford et al. 2001). In addition, late-life cognitive scores adjusted for such measures have been shown to correlate highly with actual lifetime cognitive change (Deary et al. 2004). Prior to the analysis, all SNPs were checked for Hardy–Weinberg equilibrium (HWE) and (for the genes where multiple SNPs had been typed) for patterns of LD. All analyses were carried out in R version 2.8.1 (Ihaka & Gentleman 1996).
The population had a mean age of 67.2 (SD 6.51) and was predominantly female (n = 1520, 72.7%). Table 1 shows descriptive information for each of the SNPs typed. All were found to be in HWE (P > 0.05) with the minor allele frequencies being similar to those of the HapMap population with European ancestry (CEU). The three TCF-1 SNPs and the IL-6 pair rs2069832 and rs1800795 were found to be in high LD (r2 > 0.80).
Table 1. Descriptive statistics for the inflammation-related SNPs
| ||rs1143634||2021||C||T||0.23||0.21||0.85||2||113306611–113307111||Coding synonymous|
| ||rs1800795||2001||G||C||0.41||0.46||0.96||7||22732920–22733420||5′ near gene|
| ||rs2464196||2017||T||C||0.29||0.3||0.33||12||119919560–119920060||Coding non-synonymous|
The P values for the age- and sex-adjusted ancova are presented in Table 2. In spite of the large number of tests conducted, given the high correlation between the individual cognitive tests, the LD structure between some of the SNPs and the independence between genes, it was not deemed necessary to correct for multiple testing. At a nominal threshold of P = 0.05, only 1 of the 72 tests was significant; the association between rs2856838 (IL-1α) and the TMT measure of mental flexibility (P = 0.044). None of the SNPs associated with the MHVS estimate of pre-morbid intelligence. The overall contribution of the SNPs towards explaining variation in the cognitive traits was small, with R2 never exceeding 0.3%.
Table 2. Age- and sex-adjusted P values for the ancova models of cognitive ability by SNP genotypes
|ln(TMT)||97 (76, 128)*||0.14||0.94||0.044||0.26||0.74||0.53|
For the models of estimated lifetime cognitive change, the age-, sex- and MHVS-adjusted ancovaP values are presented in Table 3. The general pattern of results is similar to the age- and sex-adjusted models with a single association reaching significance at the P = 0.05 threshold. This was again for an IL-1α SNP, rs3783546, but this time with the DST measure of processing speed (P = 0.048). After performing a Bonferroni correction for multiple testing, there were no statistically significant findings in either the baseline model or the model of estimated lifetime cognitive change.
Table 3. Age-, sex- and MHVS-adjusted P values for the ancova models of estimated cognitive change by SNP genotypes
This study tested the association between SNP variants in the IL-1α, IL-1β, IL-6, TCF-1 and F13A1 genes and cognitive ability in a sample of over 2000 healthily ageing adults living in central Scotland. No strong significant results were found between the SNPs and either late-life cognitive ability or estimated lifetime cognitive change. Weak associations were found between the IL-1α rs2856838 SNP and a measure of mental flexibility and between IL-1α rs3783546 and estimated decline taken from a measure of processing speed. However, these disappeared after a correction for multiple testing.
Whilst coverage of the genes investigated was by no means comprehensive, the findings provided no suggestion of an association with cognitive function. By contrast, there are empirical evidence and plausible mechanistic hypotheses to relate inflammatory biomarker levels with cognitive function (Alley et al. 2008; van den Biggelaar et al. 2007; Dik et al. 2005; Holmes et al. 2003; Jordanova et al. 2007; Rafnsson et al. 2007; Trompet et al. 2008; Wright et al. 2006; Yaffe et al. 2003).
The results of the current study imply that it is unlikely for IL-1 and IL-6 to be causally associated with cognitive ability. By contrast, previous studies in this area did find some evidence for associations between SNPs in these genes and childhood cognition (Harding et al. 2005) and late-life cognitive ability (Baune et al. 2008; Trompet et al. 2008). The null findings for the TCF-1 and F13A1 genes are not surprising. Despite their associations with plasma CRP and fibrinogen levels (Mannila et al. 2006, 2007; Soria et al. 2005), which are consistent predictors of late-life cognitive ability and decline (Alley et al. 2008; Dik et al. 2005; Luciano et al. 2009; Rafnsson et al. 2007; Schram et al. 2007), they are not directly involved in the coding of these proteins. However, a limitation of this study was the lack of any biomarker–SNP associations.
The strengths of this study included the use of a large cognitive battery that included a vocabulary-based measure allowed for the calculation of estimated lifetime cognitive change. Given that very few studies contain serial cognitive assessments across a lifespan, having an estimate of lifetime change is extremely useful. Whilst study size was a strength compared with those reported in the literature (Baune et al. 2008) the inability of the SNPs to explain a sufficiently large percentage of cognitive variation indicates either that the study was underpowered or that the null findings were accurate. For example, assuming a minor allele frequency (MAF) of 0.23, which was the lowest observed, the study had over 90% power for a two-sided significance test (α = 0.05) to detect 1% of the cognitive variance but just over 60% power to detect 0.5% of the cognitive variance. By contrast, the maximum cognitive variance explained by the SNPs was below 0.3%. Other limitations included the lack of control for potential confounding factors other than age and sex. However, as no strong positive associations were found this would have made very little difference to the overall conclusions. Finally, over 1000 persons were excluded from the analysis because of a lack of cognitive test results (less than three of the five tests completed). Whilst there were few differences between the baseline characteristics of these groups (Price, Stewart, Deary, et al. 2008), some bias may have been introduced as people with the worst cognitive decline may have been less likely to attend follow-up. This may have reduced the magnitude of the findings slightly but it is unlikely to have had a major effect because the incidence of frank dementia is likely to be small.
In conclusion, we have shown 12 SNP variants in the IL-1α, IL-1β, IL-6, TCF-1 and F13A1 genes not to associate with late-life cognitive ability or estimated lifetime cognitive change in an elderly Scottish population. Although this does not add evidence to a causal inflammation-cognition hypothesis, replication is required to confirm these findings. Nonetheless, we recommend that future studies focus on variants that encode directly the inflammatory proteins that have been consistently associated with cognition, such as C-reactive protein, interleukin-1, interleukin-6, tumour necrosis factor-alpha and fibrinogen.
This research was supported by the Wellcome Trust, Chest Heart and Stroke Scotland, the British Heart Foundation, the Chief Scientist Office, Scotland, and the Medical Research Council (funding for R.M. and the MRC Centre for Cognitive Ageing and Cognitive Epidemiology).