Alzheimer’s Disease Susceptibility Gene Apolipoprotein E ( APOE ) and Blood Biomarkers in UK Biobank (N = 395,769)

Abstract

Background:

Alzheimer’s disease (AD) is a neurodegenerative condition where the underlying etiology is still unclear. Investigating the potential influence of apolipoprotein E (APOE), a major genetic risk factor, on common blood biomarkers could provide a greater understanding of the mechanisms of AD and dementia risk.

Objective:

Our objective was to conduct the largest (to date) single-protocol investigation of blood biomarkers in the context of APOE genotype, in UK Biobank.

Methods:

After quality control and exclusions, data on 395,769 participants of White European ancestry were available for analysis. Linear regressions were used to test potential associations between APOE genotypes and biomarkers.

Results:

Several biomarkers significantly associated with APOE ɛ4 ‘risk’ and ɛ2 ‘protective’ genotypes (versus neutral ɛ3/ɛ3). Most associations supported previous data: for example, ɛ4 genotype was associated with elevated low-density lipoprotein cholesterol (LDL) (standardized beta [b] = 0.150 standard deviations [SDs] per allele, p < 0.001) and ɛ2 with lower LDL (b = –0.456 SDs, p < 0.001). There were however instances of associations found in unexpected directions: e.g., ɛ4 and increased insulin-like growth factor (IGF-1) (b = 0.017, p < 0.001) where lower levels have been previously suggested as an AD risk factor.

Conclusion:

These findings highlight biomarker differences in non-demented people at genetic risk for dementia. The evidence herein supports previous hypotheses of involvement from cardiometabolic and neuroinflammatory pathways.

Keywords

Alzheimer’s disease APOE cholesterol dementia UK Biobank

INTRODUCTION

Alzheimer’s disease (AD) is the most common form of dementia and an important public health issue [1], hypothesized to be the result of interactions between genetic and environmental risk factors [2]. Apolipoprotein E (APOE) ɛ4 genotype is the largest common single genetic risk factor for AD and cognitive decline behind increasing age [3], with the ɛ2 allele potentially protective [1]. The exact mechanisms by which APOE genotype influences brain aging are unclear but probably due to pleiotropic pathways stemming from its core role in lipid metabolism [4].

Several studies have investigated serum biomarker differences between AD patients versus healthy individuals in order to identify potential risk factors, including low density lipoproteins (LDL) and insulin-like growth factor 1 (IGF-1) with sometimes conflicting results potentially due to methodological heterogeneity [5, 6]. Many studies investigating serum levels in AD have focused on specific biomarkers involved in amyloid-β protein precursor metabolism and phosphorylation [7]. There have been relatively few studies investigating a wide range of biomarkers in a “hypothesis-free” approach; and those studies which have done this appear to be limited by small sample size or have been cross-sectional in individuals with an extant diagnosis of dementia [8 –10]. Relatively few biomarkers have been investigated in the context of AD-susceptibility gene APOE. Gaining a greater understanding of how APOE influences biomarker serum levels could be extremely beneficial: highlighting factors significantly associated with AD genetic risk could elucidate potential pathways involved in its development, pathophysiology and ultimately treatment [11].

In this study, APOE genotype status was tested versus a range of circulating serum blood biomarkers available for approximately 396,000 participants in UK Biobank. Two separate analyses were undertaken to investigate the influence of genotypic status on biomarker levels: differences per 1) risk ɛ4, or 2) protective ɛ2 allele; each versus neutral ɛ3/ɛ3 genotype. Further analyses were undertaken to investigate the associations in males and females separately due to a priori evidence for APOE-sex differences in AD pathophysiology [12]. To our knowledge, this is the first large-scale investigation of the relationship between APOE genotype status and a wide range of biomarkers in a population cohort.

METHODS

Subjects

Over 502,000 UK residents aged 37–73 years were recruited to UK Biobank from 2006-2010. At one of 22 assessment centers across the UK, participants completed a range of phenotypic assessments and questionnaires, including genetic, urine, and blood samples [13]. We focused on participants with White British ancestry because there is evidence of different ɛ4 frequencies across ethnicities [14]. This project was completed using UK Biobank application 17689 (PI: DML).

Ethical approval

This secondary-data analysis study was conducted under generic approval from the NHS National Research Ethics Service (approval letter dated 17 June 2011, ref 11/NW/0382). Written informed consent was obtained from all participants in the study (consent for research, by UK Biobank).

Genotyping

UK Biobank participants were genotyped using Applied Biosystems UK BiLEVE Axiom array by Affymetrix and Applied Biosystems UK Biobank Axiom Array which share 95% marker content [13]. APOE status was based on two single nucleotide polymorphisms (SNPs): rs7412 and rs429358. Stringent quality control and processing were applied to the data, detailed at http://www.ukbiobank.ac.uk/scientists-3/genetic-data and http://www.ukbiobank.ac.uk/wp-content/uploads/2014/04/UKBiobank_genotyping_QC_documentation-web.pdf.

Biomarker collection and processing

Biomarker levels were analyzed in UK Biobank from serum and packed red blood cell samples obtained from all UK Biobank participants at baseline [15]. Stringent quality controls were applied to the assays used measure biomarker levels, details of biomarker quality control, instrumentation and analysis methods are available at: https://biobank.ndph.ox.ac.uk/showcase/showcase/docs/biomarker_issues.pdf, https://biobank.ndph.ox.ac.uk/showcase/showcase/docs/serum_biochemistry.pdf, http://biobank.ndph.ox.ac.uk/showcase/showcase/docs/haematology.pdf, and http://www.ukbiobank.ac.uk/wp-content/uploads/2018/11/BCM023_ukb_biomarker_panel_website_v1.0-Aug-2015-edit-2018.pdf. Processing of very low levels of estradiol and rheumatoid factor (RF) were recorded as “missing” (in the original data); these missing values were recoded conservatively as the square root of the minimum stated detectable value if individuals had data for a remaining biomarker [16] and were not coded as ‘no data returned’ or having unrecoverable aliquot problems. Estradiol and RF require attention in this context because they were the biomarkers highlighted by UK Biobank as the variables with by far the highest frequency of values below reportable levels (75–90% of results); no assays had >0.1% of results above their reportable range.

Dementia outcomes

We validated APOE genotype’s association with dementia/AD outcomes (in UK Biobank) as a check, having previously shown associations in expected directions with brain structure [3] and cognitive abilities [17]. Dementia and AD outcomes were generated using self-report, hospital admission and death record data, with hospital and death record data utilizing International Classification of Diseases version 10 (ICD-10 codes). Individuals were designated as cases (“all-cause dementia” or “Alzheimer disease”) if they had indicated dementia or AD in either self-report, or through hospital or death records— derived by UK Biobank [18]. Those coded as missing were designated as controls (i.e., did not self-report dementia/AD, and these diagnoses were not present in hospital/death records).

Covariates

Participants self-reported their smoking history, and we collated past and current smokers into ‘ever’ (versus never). Participants self-reported medication use for cholesterol, high blood pressure, oral contraceptive, hormone replacement therapy, or insulin. We excluded those who did not know or preferred not to answer for these various items (<5%). Townsend deprivation indices were derived from postcode of residence [19]. This provides an area-based measure of socioeconomic deprivation derived from aggregated data on car ownership, household overcrowding, owner occupation, and unemployment. Higher Townsend scores equate to higher levels of area-based socioeconomic deprivation. We additionally controlled for potential population stratification using UK Biobank-derived principal components (PCs) 1–5, and genotypic array [13]. Body mass index (BMI) was derived from weight in kilograms divided by height in meters squared. Height was measured (Seca 202 stadiometer) and weight was measured to the nearest 0.1 kg (BC-418 MA body composition analyzer; Tanita Corp). All-cause cancer was derived based on self-report at baseline. Month of assessment was recorded by UK Biobank. We have previously described and derived an ‘any self-reported inflammatory condition’ variable [20], including 58 conditions described in an open-access report. Participant age, sex and assessment center were recorded by UK Biobank, and educational attainment was self-reported.

Statistical analyses

The linear regressions reported reflect average SD-changes per ɛ4 or ɛ2 allele versus neutral ɛ3/ɛ3 genotype, i.e., ɛ3/ɛ3 versus ɛ3/ɛ4 versus ɛ4/ɛ4 (dose), and ɛ2/ɛ2 versus ɛ2/ɛ3 versus ɛ3/ɛ3 (dose). Associations with APOE genotype versus dementia/AD were tested using binary logistic regressions reporting odds ratios (OR) and their 95% confidence intervals. We corrected for multiple testing using False Discovery Rate (FDR) [21], conservatively collating all tests. Biomarkers which were not normally distributed were log transformed prior to Z-score transformation and reanalyzed: the resulting effect sizes were unchanged and therefore we report the original estimates.

Three linear regression models were used to investigate potential associations with each biomarker and adjusted for potential confounders. Model 1 (‘minimally-adjusted’) adjusted for age, sex, baseline assessment center, principal components 1–5 for population stratification, and genotyping array. Model 2 (‘partially-adjusted’) also included self-reported diabetes, high blood pressure, and coronary heart disease (comprised of angina plus myocardial infarction [22]). Model 3 (‘fully adjusted’) also controlled for self-reported cholesterol, hormone replacement therapy, oral contraceptive, insulin or hypertension medication, Townsend deprivation scores, and ever versus never smoking. The cross-sectional association between APOE versus dementia outcomes were also analyzed using these models.

As additional sensitivity analyses, we adjusted for dummy-variable ‘any self-reported chronic inflammatory condition’ (n = 64,996; 17%); underweight (BMI < 18.5; n = 1,962 or 0.5%) or obesity (BMI≥30, n = 297,738 or 75.4%) versus normal to overweight (18.5 to 30 BMI; n = 95,001 or 24.1%), month of assessment, university/college degree versus not, and finally we additionally corrected any significant associations with IGF-1 for self-reported baseline cancer history (n = 33,406; 8%) [23]. SNP data was collated and quality controlled with PLINK V1.90, and analyzed with Stata V16.

RESULTS

We excluded participants with non-white British ancestry (N = 78,672; 16%), sex mismatch (self-report versus genetic), chromosomal aneuploidy, excessive heterozygosity, and genotype missing rate >10%. We excluded the minority of participants with ɛ2/ɛ4 (n = 2,556; 0.7%) genotype because this included potentially protective and risk alleles [24]. We removed outliers >5 SDs from the mean for each biomarker. This left overall N = 395,769 which varied slightly by biomarker: Table 1 shows sample size and key values per biomarker.

Table 1

Baseline descriptive values

	Mean (SD)	Median	Min	Max
Age	56.9 (8.0)	58	39	73
Townsend deprivation score	–1.57 (2.9)	2.35	–6.26	10.8
All-cause dementia	Cases N (%)		Controls N (%)
	1,852 (0.47%)		393,917 (99.53%)
Alzheimer’s disease	Cases N (%)		Controls N (%)
	722 (0.18%)		395,047 (99.82%)
	Male		Female
Sex	181,000 (45.7%)		214,771 (54.3%)
	ɛ3/ɛ3	ɛ3/ɛ4	ɛ4/ɛ4	ɛ3/ɛ2	ɛ2/ɛ2
APOE genotype	230,094 (59.7%)	94,395 (24.5%)	9,518 (2.4%)	49,082 (12.7%)	2,557 (0.6%)
	Non-smoker		Previously smoked		Current smoker
Smoking status	216,094 (54.8%)		138,744 (35.2%)		39,580 (10%)
	No		Yes
CHD	466,843 (96.5%)		16,920 (3.5%)
High BP	381,072 (78.8%)		102,691 (21.2%)
Type 2 diabetes	376,233 (95.3%)		18,702 (4.7%)
Medication	329,322 (83.4%)		65,634 (16.6%)
Biomarkers (not standardized)	Mean	Standard deviation	Min	Max	N
Low-density lipoprotein	3.6	0.9	0.5	7.9	376,668
High-density lipoprotein	1.5	0.4	0.2	3.4	345,315
Insulin-growth like factor-1	21.4	5.5	1.4	49.6	375,135
Hemoglobin A1C	35.6	5.1	15.0	68.1	374,702
Lipoprotein A	44.1	49.5	3.8	189.0	300,249
Estradiol	71.6	173.8	8.5	1313.3	349,998
Phosphate	1.2	0.2	0.4	2.0	344,894
Rheumatoid factor	4.4	4.9	3.2	47.4	374,322
Sex hormone-binding globulin	51.4	26.4	0.4	189.3	341,418
Total Bilirubin	9.0	3.9	1.1	31.2	374,181
Direct bilirubin	1.8	0.7	1.0	6.0	320,111
Testosterone	6.5	6.0	0.4	36.7	341,975
Protein	72.4	4.0	36.3	92.5	345,085
Triphosphate	1.7	1.0	0.2	6.9	375,888
Urate	309.1	79.9	89.1	708.8	376,905
Vitamin D	49.9	20.8	10.0	154.0	360,598
C-reactive protein	2.3	3.0	0.1	24.2	373,817
Gamma-glutamyltransferase	34.9	28.1	5.0	241.7	374,958
Cystatin C	0.9	0.1	0.3	1.8	376,382
ApoB	1.0	0.2	0.4	2.0	375,576
ApoA	1.5	0.3	0.4	2.5	343,506
Creatinine	72.0	14.5	10.8	161.4	376,640
Total cholesterol	5.7	1.1	1.4	11.4	377,349
Calcium	2.4	0.1	1.2	2.8	345,223
Urea	5.4	1.3	0.8	12.3	376,453
Aspartate transaminase (AST)	25.8	7.5	3.3	78.7	374,496
Alanine transaminase (ALT)	23.0	11.6	3.0	93.7	375,656
Alkaline phosphatase	83.1	22.9	8.0	215.0	376,658
Albumin	45.2	2.6	18.9	58.2	345,558

APOE associations with dementia phenotypes

As a form of replication and to support the utility of investigating APOE genotype in the UK Biobank cohort, we tested for ɛ4 and ɛ2 allele count (versus ɛ3/ɛ3) against all-cause dementia (n = 1,852; 0.5%), and specific AD diagnosis (n = 722; 0.2%). We found that ɛ4 was associated with increased AD (fully-adjusted OR = 3.51 per allele, 95% CI = 3.14 to 3.92, p < 0.001) and all-cause dementia (OR = 2.59, 95% CI = 2.40 to 2.79, p < 0.001), whereas, ɛ2 was correspondingly associated with decreased AD (OR = 0.59, 95% CI = 0.40 to 0.85, p = 0.005) and all-cause dementia (OR = 0.78, 95% CI = 0.65 to 0.93, p = 0.007). An unadjusted chi-square test showed 64% % of people with AD had at least one ɛ4 allele versus 36% in the non-AD group.

APOE associations with biomarker values

Several significant associations (at nominal p < 0.05) were identified between APOE genotype status and biomarker values, as shown in Fig. 1, which shows fully-adjusted estimates. Increasing ɛ4 allele count associated with significant differences in several biomarker values (Supplementary Table 1). There were associations between ɛ4 allele count versus higher LDL, IGF-1, sex hormone binding globulin (SHBG), total bilirubin, triphosphate levels, ApoB, and total cholesterol. Negative associations were found between ɛ4 genotype and lower high-density lipoprotein (HDL), hemoglobin A1c (HbA1c), lipoprotein A, phosphate, C-reactive protein (CRP), gamma glutamyl transferase (GGT), vitamin D, creatinine, urate, and urea. The largest effect sizes, based on fully-adjusted model results, were seen for total cholesterol (0.13 SDs per allele in the fully adjusted model), ApoB (0.20), CRP (–0.12), and LDL (0.15), with the rest <0.1 SDs per ɛ4 allele. There was no evidence of a significant fully adjusted association between ɛ4 and estradiol, aspartate transaminase (AST), albumin, testosterone, and RF levels.

Fig. 1.

Linear regression fully-adjusted standardized betas comparing APOE ɛ4 and ɛ2 genotypes (per allele) versus neutral ɛ3/ɛ3, on biomarker values. p < 0.05, **p < 0.001.

Significant associations were seen between ɛ2 allele count versus lower LDL, HDL, IGF-1, total bilirubin, direct bilirubin, vitamin D, CRP, cystatin C (CysC), ApoA, ApoB, creatinine, and alkaline phosphatase (Fig. 1). These are in the opposite directions of effect reported for the associations to the ɛ4 allele (versus ɛ3/ɛ3) as expected. Associations between APOE ɛ2 and HbA1c, lipoprotein A, SHBG, and triphosphate levels were also identified; however, these effects were in the same direction as ɛ4. The largest overall e2 effect sizes were for LDL (–0.46 SDs per allele in the fully-adjusted model), triphosphate (0.13), ApoB (–0.61), and total cholesterol (–0.35), with the rest <0.1 SDs per ɛ2 allele.

There were instances of significant association for ɛ2 but not ɛ4 versus ɛ3/ɛ3. These were: negative associations between ɛ2 versus protein and aspartate aminotransferase levels, and positive association between ɛ2 and testosterone. There was no statistically significant association between ɛ2 count and estradiol, phosphate, GGT, urate, urea, and alanine aminotransferase levels (Supplementary Table 1). Both ɛ4 and ɛ2 were also found to associate with increased SHBG, decreased protein, and increased triphosphate (versus ɛ3/ɛ3).

Sex-specific analyses

There were instances of male/female sex versus genotype interactions for several biomarkers (Supplementary Table 2). Out of 33 biomarkers, 16 showed some interaction (p < 0.05): LDL, HDL, HbA1c, estradiol, RF, SHBG, Testosterone, protein, triphosphate, urate, ApoB, ApoA, total cholesterol, AST, and alanine transaminase (ALT). We provide complete sex-specific Z-score associations in Supplementary Table 3. Most of the significant associations in the collated analyses remained so in individual sexes. Certain associations were only significant in males: ɛ4 versus estradiol and calcium, and ɛ2 versus IGF-1, SHBG, testosterone, and cystatin c.

Sensitivity analyses

All nominally significant associations survived correction for FDR. Results were unchanged in terms of effect size and p-value when individuals with incident dementia/AD were removed from analyses. When we added presence of any self-reported inflammatory condition, month of assessment, degree versus not, underweight or obesity (versus normal to overweight) to the final model, no results were meaningfully changed (Supplementary Table 4). Results were unchanged when we re-analyzed all outcome variables with inverse-rank normalization to avoid potential false positives caused by outlying values (prior to removing values >5 SD from the mean), and when we used the maximum 40 PCs (versus 5 reported). IGF-1 results were unchanged when we additionally controlled for all-cause cancer. As a check, we re-ran all final-model tests as 0 versus 1 allele, and 0 versus 2 allele contrasts (rather than a 0/1/2 dose effect); results were consistently indicative of dose effects in the same direction.

DISCUSSION

The APOE ɛ4 allele is known to associate in UK Biobank with worse non-demented cognitive abilities [17], cerebrovascular health [3], and here clinically-ascertained AD/dementia risk (mostly; a small minority of cases were baseline self-report). In this study, we found several significant associations per ɛ4 allele on circulating biomarker levels (versus neutral ɛ3/ɛ3). In many instances these were supported by corresponding associations between the putatively protective ɛ2 genotype in the opposite direction as would be expected. Gaining a greater understanding of the biomarker profiles of individuals at genetic risk of AD may be useful in the future for earlier detection of AD and could potentially highlight pathways as therapeutic targets [11]. In some instances, the directions of effect for ɛ4 (‘risk’) and ɛ2 (‘protective’) alleles conflicted with what would be expected based on levels reported elsewhere in people with prevalent AD. For example: lower levels of IGF-1 have previously associated with increased risk of AD and cognitive decline [6, 25]. By contrast here we showed association between ‘risk’ ɛ4 and increased IGF-1, and between ‘protective’ ɛ2 genotype and lower IGF-1. This is surprising as IGF-1 stimulates neurogenesis and promotes cell survival [26]. Previous reports could reflect some degree of reverse causality or bias in cross-sectional AD patient sample studies showing lower IGF-1 levels, i.e., where disease onset affects biomarker health.

Interpretation

Vitamin D

Previous studies have reported APOE ɛ4 association with higher vitamin D serum concentration [27]; however, here ɛ4 associated with decreased, and ɛ2 with increased vitamin D. In ɛ4 homozygotes, a higher vitamin D concentration has been associated with higher memory function, suggesting higher vitamin D levels could be protective for people at risk for AD [28].

Lipids

Investigations into the effect of differing cholesterol levels on cognitive decline and AD have produced conflicting results with both low and high cholesterol being associated [8 , 29–31]. The lack of consensus could be partially due to the smaller sample sizes previously used. The three APOE alleles encode for different ApoE protein isoforms with altered lipid interactions in serum; the ɛ3 encoded protein isoform is associated with “normal” plasma lipid levels [32]. APOE ɛ4 and its associated higher LDL have been previously associated with early onset of AD [33], while ɛ2 has protective effects on the concentrations of cholesterol, lipids, and phospholipids [34]. Here we reinforce those observations, particularly in the context of respective dose ɛ2/ɛ4 protective versus deleterious associations [5 , 35].

It has been hypothesized that lipid metabolism is important in the pathophysiology of AD [11]. The significant associations between APOE genotype status and ApoA/ApoB support this. ApoA and ApoB proteins are major surface proteins of HDL and LDL, respectively [36]; previously ApoA has been associated with lower risk of cardiovascular disease [37], whereas, ApoB is reported to be proatherogenic [36]. We identified lower ApoA levels in ɛ4 carriers and higher levels in ɛ2 carriers; ApoA has reportedly neuroprotective effects by inhibiting amyloid-β plaque aggregation [36, 38]. Our findings are supported by previous associations reported between decreased serum ApoA and increased AD risk [38]. With ApoB, we found ɛ4 was associated with elevated levels of ApoB and ɛ2 with lower levels of ApoB compared to ɛ3. This is consistent with the direction of effect reported in most [10, 33] but not all studies [36].

It is hypothesized APOE allele-encoded protein isoforms have different affinities for lipoproteins [39]. It has been difficult to define the exact effects of APOE genotype on lipoprotein A from previous studies’ generally small sample sizes. Our data showed that both APOE ɛ2 and ɛ4 were significantly associated with decreased levels of lipoprotein A (versus ɛ3) [39]. Elevated levels of lipoprotein A are considered to have a causal relationship with myocardial infarction, which has previously been associated with increased dementia risk [39]. The influence of lower lipoprotein A on the pathophysiology of AD remains unclear.

APOE genotype status may modify the effect of sex hormones on dementia symptoms [40]. We did not find any evidence for a significant association between APOE genotype and estradiol in the whole sample, although in males ɛ4 associated with lower levels. Estradiol has suggested overall neuroprotective effects; however, evidence is conflicting, and this relationship is not fully understood [40]. We found an association between APOE ɛ2 and higher testosterone levels, specific to males. There is conflicting evidence regarding the effect of testosterone on AD risk [41, 42]; some cross-sectional studies report lower levels of testosterone in AD patients versus healthy controls [42], consistent with our findings as ɛ2 is associated with decreased AD risk [1].

Inflammation

The underlying pathophysiology of AD has been suggested to be at least partially influenced by neuroinflammation [43]. Serum CRP levels are a marker for inflammation but the evidence for association with AD risk is conflicting. Interactions between APOE ɛ4 and elevated CRP have been reported to associate with early onset of AD [44]. However, consistent with our findings ɛ4 has been associated with lower CRP levels, and ɛ2 with higher CRP levels [10, 45]. Further research is required to elucidate the effects of CRP levels on the pathophysiology of AD, particularly given it is a marker of acute rather than chronic inflammation [43]. Another inflammatory marker associated with increased risk of AD is raised GGT [46]. We identified lower levels of GGT in ɛ4 carriers: this may reflect bias in cross-sectional studies of GGT and AD.

Alkaline phosphatase may have some involvement in the inflammatory/AD process [47]. We identified lower levels of alkaline phosphatase in ɛ4 carriers and elevated levels in ɛ2 carriers. It has been suggested that alkaline phosphatase could potentially be used as a therapy to reduce neuroinflammation in AD [47]; our findings may support this but more in depth investigations are required.

CysC, typically a marker of kidney dysfunction, is also involved in modulation of inflammatory responses and reported to have neuroprotective effects in AD as it co-localizes with amyloid-β and inhibits oligomerization [48]. We found that ɛ4 associated with lower CysC and ɛ2 with higher CysC levels compared to ɛ3. Lower baseline CysC has been reported to precede AD onset in an 11-year longitudinal study of non-demented elderly men at baseline; in one small study (N = 82) [49] which suggested the finding may be due to attrition bias because higher CysC is a risk factor for cardiovascular disease and earlier mortality. Our findings potentially lend support to low CysC serum levels as an AD risk factor.

Other biomarkers

We report suggestive association between ɛ4 and lower phosphate levels, potentially contradicting prior research showing association between (age-dependent) higher serum phosphorous and incident dementia [50]. We found significant associations between ɛ4 and higher total bilirubin and urea, and lower direct bilirubin, urate, creatinine, calcium, and alanine aminotransferase. In the whole sample analysis, we found evidence of associations between ɛ2 and higher direct bilirubin, creatinine and aspartate aminotransferase, and lower total bilirubin and albumin. Both ɛ4 and ɛ2 associated in same direction with HbA1c, SHBG, protein, and triphosphate. This is unexpected: ɛ4 and ɛ2 tend to show opposing effects regarding AD. We found proportionally moderate evidence of interaction between sex, genotype, and biomarker; this is a significant area of research and warrants further study along with investigation age- and multimorbidity-related interactions [12].

Limitations and future research

A limitation of UK Biobank is that participants are overall likely to have fewer health conditions, be better educated, of older age, female, and living in less socio-economically deprived areas than the general population [13]. This study was conducted in individuals of White European ancestry only and so these results may not be generalizable to a more mixed population. Findings may not be truly representative of the effects of APOE genotype in the wider UK population [13]. There may be conflicting biases at play in at least some of the results. For example, confounding effects may exist where ɛ4 carriers are of poorer health, affecting their lifestyle and in turn influencing biomarker values, or selection bias, where the ɛ4 carriers here are of relatively good health relative to carriers in the general population.

Although this study reports significant associations between serum biomarkers and APOE genotype, which could influence the risk of AD, these alleles are unlikely to be entirely responsible [1]. Associations may to some extent reflect early prodromal AD; future study may investigate biomarkers in AD-by-proxy (i.e., family history) cases. Some biomarker levels may only be pathogenic in combination with other biomarkers [8]. Some of the biomarkers may be influenced by environmental factors such as seasonality, although this should to some extent average out; sensitivity analyses showed no evidence of confounding. The effect sizes reported here are in many cases small and the results require replication; the clinical applicability of these findings remains unclear. It is not necessarily possible to identify the exact biological pathways involved in the pathophysiology of AD from these analyses. Further work is required to investigate the underlying pathways to identify processes which could be modified or targeted to decrease the risk of AD, including longitudinal biomarker data [12, 35]. Future research, e.g., using Mendelian randomization may investigate whether pharmaceutically altering serum biomarker levels or implementing lifestyle changes to manage these biomarkers may be beneficial to individuals at greater risk of developing AD.

Conclusion

The exact influence of APOE on AD and dementia pathophysiology is unclear. Through this study we have identified associations between the high-risk AD susceptibility gene locus APOE and a range of serum blood biomarkers in UK Biobank. These associations highlight potential pathways involved in the development of cognitive impairment and have potential to lead to earlier detection of AD risk through the analysis of biomarkers and APOE genotype.

Footnotes

ACKNOWLEDGMENTS

UK Biobank was established by the Wellcome Trust medical charity, Medical Research Council, Department of Health, Scottish Government and the Northwest Regional Development Agency. It has also had funding from the Welsh Assembly Government and the British Heart Foundation. The funders had no role in study design, data collection or management, analyses or interpretation of the data, nor preparation, review or approval of the manuscript. DML is supported by The Neurosciences Foundation, and American Psychological Foundation. Smith is partially funded by the Lister institute. Cavanagh is funded by the Sackler Trust, Wellcome Trust, Medical Research Council, and holds a Wellcome Trust strategic award, an industrial-academic collaboration with Janssen & Jannsen, GlaxoSmithKline, and Lundbeck. Pell has received funding from the Medical Research Council and Chief Scientist Office

This research has been conducted using the UK Biobank resource; we are grateful to UK Biobank participants.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Van Cauwenberghe

, Van Broeckhoven

, Sleegers

(2016) The genetic landscape of Alzheimer disease: Clinical implications and perspectives. Genet Med 18, 421–430.

Gatz

, Reynolds

, Fratiglioni

, Johansson

, Mortimer

, Berg

, Fiske

, Pedersen

(2006) Role of genes and environments for explaining Alzheimer disease. Arch Gen Psychiatry 63, 168.

Lyall

, Cox

, Lyall

, Celis-Morales

, Cullen

, Mackay

, Ward

, Strawbridge

, McIntosh

, Sattar

, Smith

, Cavanagh

, Deary

, Pell

(2019) Association between APOE e4 and white matter hyperintensity volume, but not total brain volume or white matter integrity. Brain Imaging Behav doi: 10.1007/s11682-019-00069-9.

Holtzman

, Herz

, Bu

(2012) Apolipoprotein E and apolipoprotein E receptors: Normal biology and roles in Alzheimer disease. Cold Spring Harb Perspect Med 2, a006312.

Moriarty

, Varvel

, Gordts

PLSM

, McConnell

, Tsimikas

(2017) Lipoprotein(a) mass levels increase significantly according to APOE Genotype: An analysis of 431 239 patients. Arterioscler Thromb Vasc Biol 37, 580–588.

Kimoto

, Kasanuki

, Kumagai

, Shibata

, Ichimiya

, Arai

(2016) Serum insulin-like growth factor-I and amyloid beta protein in Alzheimer’s disease: Relationship with cognitive function. Psychogeriatrics 16, 247–254.

Olsson

, Lautner

, Andreasson

, Öhrfelt

, Portelius

, Bjerke

, Hölttä

, Rosén

, Olsson

, Strobel

, Wu

, Dakin

, Petzold

, Blennow

, Zetterberg

(2016) CSF and blood biomarkers for the diagnosis of Alzheimer’s disease: A systematic review and meta-analysis. Lancet Neurol 15, 673–684.

Szalkai

, Grolmusz

(2017) Identifying combinatorial biomarkers by association rule mining in the CAMD Alzheimer’s database. Arch Gerontol Geriatr 73, 300–307.

Tynkkynen

, Chouraki

, van der Lee

, Hernesniemi

, Yang

, Li

, Beiser

, Larson

, Sääksjärvi

, Shipley

, Singh-Manoux

, Gerszten

, Wang

, Havulinna

, Würtz

, Fischer

, Demirkan

, Ikram

, Amin

, Lehtimäki

, Kähönen

, Perola

, Metspalu

, Kangas

, Soininen

, Ala-Korpela

, Vasan

, Kivimäki

, van Duijn

, Seshadri

, Salomaa

(2018) Association of branched-chain amino acids and other circulating metabolites with risk of incident dementia and Alzheimer’s disease: A prospective study in eight cohorts. Alzheimers Dement 14, 723–733.

10.

Soares

, Potter

, Pickering

, Kuhn

, Immermann

, Shera

, Ferm

, Dean

, Simon

, Swenson

, Siuciak

, Kaplow

, Thambisetty

, Zagouras

, Koroshetz

, Wan

, Trojanowski

, Shaw

, Biomarkers Consortium Alzheimer’s Disease Plasma Proteomics Project (2012) Plasma biomarkers associated with the apolipoprotein E genotype and Alzheimer disease. Arch Neurol 69, 1310–1317.

11.

Yamazaki

, Painter

, Bu

, Kanekiyo

(2016) Apolipoprotein E as a therapeutic target in Alzheimer’s disease: A review of research and clinical evidence. CNS Drugs 30, 773–789.

12.

Riedel

, Thompson

, Brinton

(2016) Age, APOE and sex: Triad of risk of Alzheimer’s disease. J Steroid Biochem Mol Biol 160, 134–147.

13.

Bycroft

, Freeman

, Petkova

, Band

, Elliott

, Sharp

, Motyer

, Vukcevic

, Delaneau

, O’Connell

, Cortes

, Welsh

, Young

, Effingham

, McVean

, Leslie

, Allen

, Donnelly

, Marchini

(2018) The UK Biobank resource with deep phenotyping and genomic data. Nature 562, 203–209.

14.

Eisenberg

DTA

, Kuzawa

, Hayes

(2010) Worldwide allele frequencies of the human apolipoprotein E gene: Climate, local adaptations, and evolutionary history. Am J Phys Anthropol 143, 100–111.

15.

Welsh

, Celis-Morales

, Brown

, Mackay

, Lewsey

, Mark

, Gray

, Ferguson

, Anderson

, Lyall

, Cleland

, Jhund

, Gill

JMR

, Pell

, Sattar

, Welsh

(2019) Comon of conventional lipoprotein tests and apolipoproteins in the prediction of cardiovascular disease. Circulation 13, 542–552 .

16.

Sinnott-Armstrong

, Tanigawa

, Amar

, Mars

, Aguirre

, Venkataraman

, Wainberg

, Ollila

, Pirruccello

, Qian

, Shcherbina

, FinnGen , Rodriguez

, Assimes

, Agarwala

, Tibshirani

, Hastie

, Ripatti

, Pritchard

, Daly

, Rivas

(2019) Genetics of 38 blood and urine biomarkers in the UK Biobank.660506;. bioRxiv doi: https://doi.org/10.1101/660506.

17.

Lyall

, Celis-Morales

, Lyall

, Graham

, Mackay

, Strawbridge

, Ward

, Gill

JMR

, Sattar

, Cavanagh

, Smith

, Pell

(2019) Assessing for interaction between ɛ4, sex, and lifestyle on cognitive abilities. Neurology 92, e2691–e2698.

18.

Wilkinson

, Schnier

, Bush

, Rannikmäe

, Henshall

, Lerpiniere

, Allen

, Flaig

, Russ

, Bathgate

, Pal

, O’Brien

, Sudlow

CLM

, Dementias Platform UK and UK Biobank (2019) Identifying dementia outcomes in UK Biobank: A validation study of primary care, hospital admissions and mortality data. Eur J Epidemiol 34, 557–565.

19.

Townsend

, Phillimore

, Beattie

(1988), Health and Deprivation: Inequality and the North. Routledge, London.

20.

Lyall

, Cullen

, Lyall

, Leighton

, Siebert

, Smith

, Cavanagh

(2019) The associations between self-reported depression, self-reported chronic inflammatory conditions and cognitive abilities in UK Biobank. Eur Psychiatry 60, 63–70.

21.

Benjamini

, Hochberg

(1995) Controlling the false discovery rate: A practical and powerful approach to multiple testing. J R Stat Soc B 57, 289–300.

22.

Lyall

, Celis-Morales

, Anderson

, Gill

JMR

, Mackay

, McIntosh

, Smith

, Deary

, Sattar

, Pell

(2017) Associations between single andmultiple cardiometabolic diseases and cognitive abilities in 474 129 UK Biobank participants. Eur Heart J 38, 577–583.

23.

Weroha

, Haluska

(2012) The insulin-like growth factor system in cancer. Endocrinol Metab Clin North Am 41, 335–350.

24.

Wisdom

, Callahan

, Hawkins

(2011) The effects of apolipoprotein E on non-impaired cognitive functioning: A meta-analysis. Neurobiol Aging 32, 63–74.

25.

Vidal

, Hanon

, Funalot

, Brunel

, Viollet

, Rigaud

, Seux

, le-Bouc

, Epelbaum

, Duron

(2016) Low serum insulin-like growth factor-I predicts cognitive decline in Alzheimer’s disease. J Alzheimers Dis 52, 641–649.

26.

Lee

, Son

(2009) Adult hippocampal neurogenesis and related neurotrophic factors. BMB Rep 42, 239–44.

27.

Egert

, Rimbach

, Huebbe

(2012) ApoE genotype: From geographic distribution to function and responsiveness to dietary factors. Proc Nutr Soc 71, 410–424.

28.

Maddock

, Cavadino

, Power

, Hyppönen

(2015) 25-hydroxyvitamin D, APOE ɛ4 genotype and cognitive function: Findings from the 1958 British birth cohort. Eur J Clin Nutr 69, 505–508.

29.

Wagner

, Helmer

, Tzourio

, Berr

, Proust-Lima

, Samieri

(2018) Evaluation of the concurrent trajectories of cardiometabolic risk factors in the 14 years before dementia. JAMA Psychiatry 75, 1033.

30.

, Shofer

, Kukull

, Peskind

, Tsuang

, Breitner

JCS

, McCormick

, Bowen

, Teri

, Schellenberg

, Larson

(2005) Serum cholesterol and risk of Alzheimer disease. Neurology 65, 1045–1050.

31.

Reitz

, Tang

, Schupf

, Manly

, Mayeux

, Luchsinger

(2010) Association of higher levels of high-density lipoprotein cholesterol in elderly individuals and lower risk of late-onset Alzheimer disease. Arch Neurol 67, 1491–1497.

32.

Lozano

, Naghavi

, Foreman

, Lim

, Shibuya

, Aboyans

, Abraham

, Adair

, Aggarwal

, Ahn

, Alvarado

, Anderson

, Andrews

, Atkinson

, Baddour

, Barker-Collo

, Bartels

, Bell

, Benjamin

, Bennett

, Bhalla

, Bikbov

, Bin Abdulhak

, Birbeck

, Blyth

, Bolliger

, Boufous

, Bucello

, Burch

, Burney

, Carapetis

, Chen

, Chou

, Chugh

, Coffeng

, Colan

, Colquhoun

, Colson

, Condon

, Connor

, Cooper

, Corriere

, Cortinovis

, de Vaccaro

, Couser

, Cowie

, Criqui

, Cross

, Dabhadkar

, Dahodwala

, De Leo

, Degenhardt

, Delossantos

, Denenberg

, Des Jarlais

, Dharmaratne

, Dorsey

, Driscoll

, Duber

, Ebel

, Erwin

, Espindola

, Ezzati

, Feigin

, Flaxman

, Forouzanfar

, Fowkes

, Franklin

, Fransen

, Freeman

, Gabriel

, Gakidou

, Gaspari

, Gillum

, Gonzalez-Medina

, Halasa

, Haring

, Harrison

, Havmoeller

, Hay

, Hoen

, Hotez

, Hoy

, Jacobsen

, James

, Jasrasaria

, Jayaraman

, Johns

, Karthikeyan

, Kassebaum

, Keren

, Khoo

, Knowlton

, Kobusingye

, Koranteng

, Krishnamurthi

, Lipnick

, Lipshultz

, Ohno

, Mabweijano

, MacIntyre

, Mallinger

, March

, Marks

, Matsumori

, Matzopoulos

, Mayosi

, McAnulty

, McDermott

, McGrath

, Mensah

, Merriman

, Michaud

, Miller

, Mock

, Mocumbi

, Mokdad

, Moran

, Mulholland

, Nair

, Naldi

, Narayan

, Nasseri

, Norman

, O’Donnell

, Omer

, Ortblad

, Osborne

, Ozgediz

, Pahari

, Pandian

, Rivero

, Padilla

, Perez-Ruiz

, Perico

, Phillips

, Pierce

, Pope

3rd , Porrini

, Pourmalek

, Raju

, Ranganathan

, Rehm

, Rein

, Remuzzi

, Rivara

, Roberts

, De León

, Rosenfeld

, Rushton

, Sacco

, Salomon

, Sampson

, Sanman

, Schwebel

, Segui-Gomez

, Shepard

, Singh

, Singleton

, Sliwa

, Smith

, Steer

, Taylor

, Thomas

, Tleyjeh

, Towbin

, Truelsen

, Undurraga

, Venketasubramanian

, Vijayakumar

, Vos

, Wagner

, Wang

, Watt

, Weinstock

, Weintraub

, Wilkinson

, Woolf

, Wulf

, Yeh

, Yip

, Zabetian

, Zheng

, Lopez

, Murray

, AlMazroa

, Memish

(2012) Global and regional mortality from 235 causes of death for 20 age groups in 1990 and 2010: A systematic analysis for the Global Burden of Disease Study 2010. Lancet 380, 2095–2128.

33.

Wingo

, Cutler

, Wingo

, Le

, Rabinovici

, Miller

, Lah

, Levey

(2019) Association of early-onset alzheimer disease with elevated low-density lipoprotein cholesterol levels and rare genetic coding variants of APOB. JAMA Neurol 76, 809.

34.

Karjalainen

J-P

, Mononen

, Hutri-Kähönen

, Lehtimäki

, Juonala

, Ala-Korpela

, Kähönen

, Raitakari

, Lehtimäki

(2019) The effect of apolipoprotein E polymorphism on serum metabolome - a population-based 10-year follow-up study. Sci Rep 9, 458.

35.

Whelan

, Mattsson

, Nagle

, Vijayaraghavan

, Hyde

, Janelidze

, Stomrud

, Lee

, Fitz

, Samad

, Ramaswamy

, Margolin

, Malarstig

, Hansson

(2019) Multiplex proteomics identifies novel CSF and plasma biomarkers of early Alzheimer’s disease. Acta Neuropathol Commun 7, 169.

36.

Tynkkynen

, Hernesniemi

, Laatikainen

, Havulinna

, Sundvall

, Leiviskä

, Salo

, Salomaa

(2016) Apolipoproteins and HDL cholesterol do not associate with the risk of future dementia and Alzheimer’s disease: The National Finnish population study (FINRISK). Age (Dordr) 38, 465–473.

37.

Emerging Risk Factors Collaboration TERF , Di Angelantonio

, Gao

, Pennells

, Kaptoge

, Caslake

, Thompson

, Butterworth

, Sarwar

, Wormser

, Saleheen

, Ballantyne

, Psaty

, Sundström

, Ridker

, Nagel

, Gillum

, Ford

, Ducimetiere

, Kiechl

, Koenig

, Dullaart

RPF

, Assmann

, D’Agostino

, Dagenais

, Cooper

, Kromhout

, Onat

, Tipping

, Gómez-de-la-Cámara

, Rosengren

, Sutherland

, Gallacher

, Fowkes

FGR

, Casiglia

, Hofman

, Salomaa

, Barrett-Connor

, Clarke

, Brunner

, Jukema

, Simons

, Sandhu

, Wareham

, Khaw

K-T

, Kauhanen

, Salonen

, Howard

, Nordestgaard

, Wood

, Thompson

, Boekholdt

, Sattar

, Packard

, Gudnason

, Danesh

(2012) Lipid-related markers and cardiovascular disease prediction. JAMA 307, 2499–2506.

38.

Slot

RER

, Van Harten

, Kester

, Jongbloed

, Bouwman

, Teunissen

, Scheltens

, Veerhuis

, van der Flier

(2017) Apolipoprotein A1 in cerebrospinal fluid and plasma and progression to Alzheimer’s disease in non-demented elderly. J Alzheimers Dis 56, 687–697.

39.

Kritharides

, Nordestgaard

, Tybjærg-Hansen

, Kamstrup

, Afzal

(2017) Effect of APOE ɛ genotype on lipoprotein(a) and the associated risk of myocardial infarction and aortic valve stenosis. J Clin Endocrinol Metab 102, 3390–3399.

40.

Xing

, Jia

, Ji

, Tian

(2013) Estrogen associated gene polymorphisms and their interactions in the progress of Alzheimer’s disease. Prog Neurobiol 111, 53–74.

41.

Chu

, Tam

, Wong

RLC

, Yik

, Song

, Cheung

BMY

, Morley

, Lam

KSL

(2010) Bioavailable testosterone predicts a lower risk of Alzheimer’s disease in older men. J Alzheimers Dis 21, 1335–1345.

42.

Panizzon

, Hauger

, Xian

, Vuoksimaa

, Spoon

, Mendoza

, Jacobson

, Vasilopoulos

, Rana

, McKenzie

, McCaffery

, Lyons

, Kremen

, Franz

(2014) Interaction of APOE genotype and testosterone on episodic memory in middle-aged men. Neurobiol Aging 35, e1–e8.

43.

Heneka

, Carson

, El Khoury

, Landreth

, Brosseron

, Feinstein

, Jacobs

, Wyss-Coray

, Vitorica

, Ransohoff

, Herrup

, Frautschy

, Finsen

, Brown

, Verkhratsky

, Yamanaka

, Koistinaho

, Latz

, Halle

, Petzold

, Town

, Morgan

, Shinohara

, Perry

, Holmes

, Bazan

, Brooks

, Hunot

, Joseph

, Deigendesch

, Garaschuk

, Boddeke

, Dinarello

, Breitner

, Cole

, Golenbock

, Kummer

(2015) Neuroinflammation in Alzheimer’s disease. Lancet Neurol 14, 388–405.

44.

Tao

, Ang

TFA

, DeCarli

, Auerbach

, Devine

, Stein

, Zhang

, Massaro

, Au

, Qiu

(2018) Association of chronic low-grade inflammation with risk of Alzheimer disease in ApoE4 Carriers. JAMA Netw Open 1, e183597.

45.

Martiskainen

, Takalo

, Solomon

, Stančáková

, Marttinen

, Natunen

, Haapasalo

, Herukka

S-K

, Kuusisto

, Soininen

, Kivipelto

, Laakso

, Hiltunen

(2018) Decreased plasma C-reactive protein levels in APOE ɛ4 allele carriers. Ann Clin Transl Neurol 5, 1229–1240.

46.

Yavuz

, Yavuz

, Halil

, Cankurtaran

, Ulger

, Cankurtaran

, Aytemir

, Ariogul

(2008) Serum elevated gamma glutamyltransferase levels may be a marker for oxidative stress in Alzheimer’s disease. Int Psychogeriatrics 20, 815–823.

47.

Pike

, Kramer

, Blaauboer

, Seinen

, Brands

(2015) An alkaline phosphatase transport mechanism in the pathogenesis of Alzheimer’s disease and neurodegeneration. Chem Biol Interact 226, 30–39.

48.

Kaur

, Levy

(2012) Cystatin C in Alzheimer’s disease. Front Mol Neurosci 5, 79.

49.

Sundelöf

, Ärnlöv

, Ingelsson

, Sundström

, Basu

, Zethelius

, Larsson

, Irizarry

, Giedraitis

, Rönnemaa

, Degerman-Gunnarsson

, Hyman

, Basun

, Kilander

, Lannfelt

(2008) Serum cystatin C and the risk of Alzheimer disease in elderly men. Neurology 71, 1072.

50.

, Xie

, Bowe

, Xian

, Al-Aly

(2017) Serum phosphorus levels and risk of incident Dementia. PLoS One 12, e0171377.