Abstract
INTRODUCTION
There is considerable evidence from observational studies and meta-analyses that type 2 diabetes increases the risk of dementia [1–4], and that this increased risk applies to both Alzheimer’s disease and vascular dementia [2, 3]. Several pathophysiological processes that are common to type 2 diabetes and Alzheimer’s disease suggest that type 2 diabetes may promote Alzheimer-associated neurodegeneration [5]. Measurements of plasma amyloid-β (Aβ) peptides have been explored as potential biomarkers for Alzheimer’s disease given that cerebral Aβ accumulation is a pathological hallmark of the condition [6–8]. A recent review concluded that plasma Aβ40 and Aβ42, the major circulating Aβ peptides, had limited diagnostic utility but that the ratio Aβ42:Aβ40 may have value [6]. In two recent large studies, longitudinal measures of these Aβ peptides were considered to have modest but clinically useful prognostic significance as a simple non-invasive screening test [7, 9].
Despite the putative association between type 2 diabetes and Alzheimer’s disease, few studies have examined plasma Aβ in type 2 diabetes. Increased plasma Aβ40 and Aβ42 levels have been reported in hyperglycemic individuals [10] and increased Aβ42 levels were found in people treated with insulin [11]. Several variables, some of which are associated with type 2 diabetes, are known to influence plasma Aβ levels including age, renal function, obesity, and APOE ɛ4 genotype [7, 12]. The aim of the present study was to investigate plasma Aβ concentrations in type 2 diabetes using a matched case-control study designed to avoid confounding from age, sex, and APOE ɛ4 status.
MATERIALS AND METHODS
Participants
The present matched case-control study included 194 patients with type 2 diabetes and 194 controls without diabetes. The patients with type 2 diabetes were participants in the longitudinal, community-based Fremantle Diabetes Study Phase II (FDS2) [13]. Subjects were selected because they had long-standing diabetes (≥15 years duration), were currently aged 50 years or over, and were judged to be cognitively normal (Mini-Mental State Examination (MMSE) score≥27). The control sample were healthy individuals from the Perth arm of the Australian Imaging, Biomarkers and Lifestyle (AIBL) Study [9, 14] and were matched 1:1 by age, sex, and APOE ɛ4 status with each FDS2 patient. The FDS2 protocol was approved by the Fremantle Hospital Human Research Ethics Committee and the AIBL study by the institutional ethics committees of Hollywood Private Hospital and Edith Cowan University [14]. Written informed consent was obtained from all study participants.
Clinical assessments
The FDS2 and AIBL participants underwent clinical assessments that included common demographic, clinical, biochemical, and cognitive data (body mass index (BMI), fasting glucose and insulin concentrations, creatinine and the Mini–Mental State Examination (MMSE)) [9, 13]. The glomerular filtration rate was estimated (eGFR) using the Chronic Kidney Disease Epidemiology Collaboration (CKDEPI) equation [15], and explored as both a continuous variable and by eGFR categories defined by internationally agreed guidelines [16]. Fasting blood samples for all participants were collected using an established AIBL protocol specific for analysis of plasma Aβ [10], and plasma Aβ40 and Aβ42 concentrations were measured via a multiplex microsphere-based immunoassay using Luminex xMAP technology (Innogenetics NV, Ghent, Belgium) in the same laboratory [9]. APOE genotypes were determined by TaqMan® SNP genotyping assays (Life Technologies Australia Pty Ltd, Victoria, Australia).
The FDS2 participants underwent a comprehensive assessment that included additional measures including assessment of cardiovascular risk factors, common comorbidities, microvascular complications, and macrovascular disease using published methods [13, 17]. Microalbuminuria was assessed using the urinary albumin:creatinine ratio on a first morning urine sample, peripheral neuropathy as a score of >2/8 on the Michigan Neuropathy Screening Instrument clinical portion [13], and retinopathy was assessed by retinal photography using a non-mydriatic camera. Peripheral arterial disease (PAD) was defined as an ankle:brachial index≤0.90 or a diabetes-related lower extremity amputation.
Statistical analyses
Statistical analyses were performed using the computer package SPSS for Windows (version 22; SPSS Inc., Chicago, IL, USA). Descriptive statistics of the data are given as proportions, means±SD, geometric mean (SD range) or median [interquartile range] for variables that deviate from a normal or log-normal distribution. A logarithm transformation [log (maximum score+1-score)] was used to normalize the distribution of MMSE score. The frequency distributions of Aβ peptides (Aβ40 and Aβ42) were investigated using histograms. Spearman’s rank order correlation was used to investigate relationships between variables. Paired comparisons were by McNemar’s test for proportions, paired t-test for normally distributed variables and by Wilcoxon signed-rank test for non-normally distributed variables, while two-way independent comparisons were by Fisher’s exact test, Student’s t-test, and the Mann-Whitney U-test, respectively. Multiple logistic regression was used in the participants with diabetes to investigate associations with membership of sub-groups defined by Aβ40 levels. All clinically plausible variables with p < 0.20 from these comparisons were considered for entry into multiple models. A two-tailed significance level of p < 0.05 was used for all analyses.
RESULTS
Sample characteristics and plasma Aβ distributions
The participants with and without type 2 diabetes were closely matched on MMSE scores as well as the matching criteria (age, sex, and APOE ɛ4 status; see Table 1). Diabetes was associated with significantly higher BMI, fasting glucose, and insulin and significantly worse renal function. Diabetes was also associated with significantly lower median plasma Aβ40 and Aβ42 peptide concentrations. However, the distributions of plasma Aβ40 and Aβ42 were different between diabetes and controls, being normally distributed in the group without diabetes and bimodal in the group with diabetes (Fig. 1). This difference was marked for Aβ40 and with a local minimum between the peaks for Aβ40 at 80 pg/mL in type 2 diabetes. This cut-point defined two distinct subgroups comprising 39.2% and 60.8% of the diabetic group who had Aβ40 levels in the low and high peaks, respectively. There was no difference in plasma Aβ40 between the groups without diabetes and individuals with diabetes and a plasma Aβ40 concentration in the higher peak (median [IQR] Aβ40: 149.8 [134.3–165.3] versus 142.9 [131.1–164.1] pg/mL, p = 0.081). The bimodal distribution of Aβ42 was less distinct than Aβ40, but members of the subgroup with lower Aβ40 (≤80 pg/mL) also had significantly lower plasma Aβ42 concentrations (median [IQR] Aβ42: 13.7 [8.6–16.2] versus 35.9 [29.5–42.6] pg/mL, p < 0.001). There was a significant linear correlation between Aβ40 and Aβ42 in type 2 diabetes (rs = 0.84, p < 0.001; see Fig. 2) that was stronger than seen in the group without diabetes (rs = 0.40, p < 0.001).
Associations with the bi-modal Aβ distribution in type 2 diabetes
To investigate possible explanations for the bimodal distribution of Aβ peptides in type 2 diabetes, we used the FDS2 dataset and compared a wide range of variables in the two subgroups defined by the Aβ40≤80 pg/mL cut-point (see Table 2). In bivariate analysis, there were significant differences between those belonging to the low and high Aβ40 subgroups in age, diabetes duration, MMSE and eGFR. Using multiple logistic regression and including plausible variables with p < 0.20 in bivariate analyses, membership of the low Aβ40 subgroup was independently associated with eGFR (odds ratio (95% CI) per ml/min/1.73 m2: 1.017 (1.002–1.033), p = 0.026) and (negatively) with diabetes duration (per 5 years’ increase: 0.703 (0.514–0.962), p = 0.028). With eGFR categories substituted for eGFR as a continuous variable in the model, membership of the low Aβ40 subgroup was strongly associated with an eGFR≥90 ml/min/1.73 m2 (2.398 (1.199–4.795), p = 0.013).
Plasma Aβ42:Aβ40 ratio
The ratio of Aβ42:Aβ40 in plasma was normally distributed and significantly higher in the subjects with diabetes compared with those without diabetes (Table 1), and there was no difference between the subgroups with diabetes defined by the plasma Aβ40≤80 pg/mL cut-point. Multiple linear regression analysis was used in the combined sample to further investigate the association between type 2 diabetes and the plasma Aβ42:Aβ40 ratio with all available variables (age, sex, BMI, ln(glucose), APOE ɛ4 status, and eGFR) entered into the model. After adjustment, type 2 diabetes remained independently associated with the plasma Aβ42:Aβ40 ratio (β= 0.19 (95% CI 0.10–0.29), p < 0.001).
DISCUSSION
In this matched case-control study, we found important differences in plasma Aβ40 and Aβ42 concentrations by type 2 diabetes status. Plasma Aβ40 and Aβ42 were lower than normal and had a bimodal distribution in type 2 diabetes whereas both peptides were normally distributed in the controls. We investigated possible causes of the bimodal distribution of Aβ40 because it had such a distinct bimodal pattern, but consider that a common mechanism may explain the phenomenon in both peptides given that they were strongly correlated. While the diabetic participants had worse renal function than the controls, those with low plasma Aβ40 concentrations were more likely to have normal or high eGFR, consistent with increased clearance from the circulation. Peripheral Aβ clearance is known to involve renal mechanisms [18] and renal failure is associated with high plasma Aβ concentrations [7, 19–21]. Glomerular hyperfiltration is an early manifestation of diabetic nephropathy that is common in type 2 diabetes where initially normal or high eGFR is followed by a rapid decline in eGFR [22, 23] and glomerular hyperfiltration is associated with increased clearance of several plasma proteins [24]. That Aβ40 is more likely to be of peripheral origin than Aβ42 [18] could account for the larger effect of glomerular hyperfiltration on Aβ40 than Aβ42. We conclude that type 2 diabetes is associated with lower plasma concentrations of Aβ peptides probably as a consequence of increased clearance from the peripheral circulation due to diabetes-associated glomerular hyperfiltration.
We explored associations of clinically relevant variables and the plasma Aβ42:Aβ40 ratio given the reported associations between type 2 diabetes and Alzheimer’s disease [2, 3] and suggestions that the index may have value as a biomarker for Alzheimer’s disease [6]. We found that type 2 diabetes was independently associated with higher plasma Aβ42:Aβ40 ratios, consistent with a lower risk of Alzheimer’sdisease [6]. Whether this indicates that this biomarker has less validity in the setting of type 2 diabetes is unknown. By the same token, lower plasma Aβ40 and Aβ42 concentrations might be expected to be associated with a lower risk of Alzheimer’s disease if, as has been suggested, the increased excretion of Aβ peptides acts as a peripheral sink that helps reduce the cerebral amyloid burden [18, 25]. Longitudinal studies will be required to clarify these issues.
These findings have a number of important implications. The use of plasma Aβ peptides as biomarkers for Alzheimer’s disease has proven to be problematic because of both analytic issues and a range of factors found to influence circulating concentrations [6]. The present study suggests that there are additional layers of complexity related to type 2 diabetes and the presence of diabetic nephropathy, a progressive condition but one which can have a period of increased function as its earliest manifestation. The biphasic distribution of both Aβ peptides in type 2 diabetes may reflect the dynamic influence of the development of diabetic nephropathy. Reduced plasma Aβ concentrations may occur early in the course because of glomerular hyperfiltration to be followed later by increased levels when glomerular filtration declines [20]. Such a biphasic pattern of Aβ clearance could have significance for the risk of Alzheimer’s disease in patients with type 2 diabetes if peripheral clearance influences cerebral amyloid accumulation [18, 26]. This could help explain why some studies have found a lack of association between type 2 diabetes and CSF Aβ levels or with cerebral amyloidosis assessed at autopsy or with PiB positron emission tomography scans [27, 28] despite clinical associations between diabetes and Alzheimer’s disease. This could also help explain why an association was seen between insulin resistance and cerebral amyloid uptake in middle aged study participants but not in those with diabetes [29]. Glomerular hyperfiltration complicates several other renal diseases and has been reported in obesity and sleep apnea [23], suggesting that the present findings may have wider clinical relevance. Future studies of plasma Aβ as a biomarker for Alzheimer’s disease need to include moredetailed assessments of renal function over time.
The strengths of our study are the close matching of the cases and controls for important variables including age, sex, APOE ɛ4 genotype, and level of cognition. All samples were collected and analyzed using identical methods in the same laboratory. Other strengths include the community nature of the source of type 2 diabetes participants and their standardized comprehensive assessment. The main limitations relate to the cross-sectional nature of the study and a consequent inability to infer causality. In addition, the sources of cases and controls differed, the former being identified from many sources in the community and the latter having been recruited fromadvertisements.
In summary, type 2 diabetes is associated with substantial differences in concentrations of circulating plasma Aβ peptides compared to those in matched control subjects. These findings have clinical and scientific implications relevant to the study of amyloid-related biomarkers for Alzheimer’s disease.
CONFLICT OF INTEREST
The authors have no conflicts of interest to report.
Footnotes
ACKNOWLEDGMENTS
We are grateful to FDS staff for help with collecting and recording clinical information for the patients with diabetes. We thank the AIBL staff for supplying control data and for performing the amyloid-β assays and APOE genotyping. The Fremantle Diabetes Study was supported by National Health and Medical Research Council of Australia grants (513781 and 1042231). TMED is supported by a National Health and Medical Research Council of Australia Practitioner Fellowship (1058260).
