There is accumulating evidence that mitochondrial dysfunction is associated with the contribution of diabetes to Alzheimer disease (AD) progression. Neuronal mitochondrial proteins are found in plasma neuronal-derived exosomes (NDEs) at levels that reflect those in brain neurons. Here, we tested the performance of mitochondrial proteins in plasma NDEs to predict cognitive decline and brain injury in participants with diabetes. The study participants with type 2 diabetes mellitus (T2DM) included 41 cognitively normal control subjects, 97 individuals with mild cognitive impairment (MCI) (68 individuals with stable MCI; 29 individuals with progressive MCI), and 36 patients with AD dementia. Plasma neuroexosomal proteins were measured by ELISA kits. Spearman correlation was used to test associations between plasma neuroexosomal mitochondrial proteins and other core biomarkers of AD. Diagnostic accuracy for progressive MCI and AD was obtained for mitochondrial proteins using receiver operating characteristic curve analyses. The associations of mitochondrial proteins with the conversion from MCI to AD were assessed by Cox proportional hazard regression analysis. Plasma levels of neuroexosomal NADH ubiquinone oxidoreductase core subunit S3 (NDUFS3) and succinate dehydrogenase complex subunit B (SDHB) were significantly lower in patients with T2DM with AD dementia and progressive MCI than in cognitively normal subjects (P < 0.001 for both groups). We also found that plasma neuroexosomal NDUFS3 and SDHB levels were lower in progressive MCI subjects than in stable MCI subjects. Both plasma neuroexosomal NDUFS3 and SDHB offer diagnostic utility for AD. Low plasma neuroexosomal SDHB levels significantly predicted conversion from MCI to AD. In addition, low mitochondrial protein levels were associated with the rate of hippocampal and gray matter atrophy and reduced AD signature cortical thickness in progressive MCI over the follow-up period. These data suggest that both plasma neuroexosomal NDUFS3 and SDHB are already increased at the early clinical stage of AD, and indicate the promise of plasma neuroexosomal mitochondrial proteins as diagnostic and prognostic biomarkers for the earliest symptomatic stage of AD in participants with diabetes.
Introduction
The incidence of type 2 diabetes mellitus (T2DM) is increasing worldwide, and the disease, common in older adults, has become a significant public health problem. Cognitive decline, including mild cognitive impairment (MCI) and dementia, is increasingly recognized as an important central nervous system (CNS) complication of diabetes that places a significant self-management burden on affected individuals and families. Numerous epidemiological studies have demonstrated that T2DM is associated with accelerated age-related cognitive decline, a higher incidence of MCI, and a nearly twofold increased incident risk of Alzheimer disease (AD), even after adjusting for vascular risk factors (1,2). Furthermore, diabetes and prediabetes substantially accelerate the progression from MCI to AD (3).
Exosomes are vesicle mediators of maintenance of cellular homeostasis and of intercellular communication. Cells from the CNS use exosomes as a strategy not only to eliminate lipids, genetic material, and toxic proteins but also to mediate cell-to-cell communication as carriers of important signals and messengers (4). Exosomes cross the blood-brain barrier and are detectable in the peripheral circulation (5). The levels of plasma exosomal biomarkers reflect pathological brain changes. Jia et al. (6) reported that the levels of amyloid-β (Aβ), total tau (T-tau), and tau phosphorylated at threonine 181 (P-T181-tau) in blood neuronal-derived exosomes (NDEs) were highly correlated with their levels in the cerebrospinal fluid in AD and amnestic mild cognitive impairment (aMCI) patients. Diabetes elicits AD-like brain changes linked with cognitive impairment and neurodegeneration, such as accumulation of Aβ, elevated tau expression and phosphorylation, and mitochondrial and synaptic dysfunction (7–9). As a common denominator in AD and diabetes, mitochondrial dysfunction has also emerged as a possible mechanistic bridge between these two pathologies (10,11). Increased evidence has indicated that mitochondrial and synaptic dysfunction is an early pathological feature of AD (12). In a recent study, ultra-deep cerebrospinal fluid proteome profiling identified evident mitochondrial protein, including subunits of electron transport chain complex, such as succinate dehydrogenase complex subunit B (SDHB), reduction in AD patients (13). Some mitochondrial proteins (subunits of electron transport chain complex I and complex II, such as NADH ubiquinone oxidoreductase core subunit S3 [NDUFS3] and SDHB) also were identified in circulating exosomes from older adults with Parkinson disease (14). However, it remains unknown whether plasma neuroexosomal mitochondrial changes occur in the early clinical stage of AD among patients with T2DM and whether plasma neuroexosomal mitochondrial proteins are correlated with other core features of AD, such as cognitive decline, Aβ pathology, and structural brain changes. In this study, we investigated the capacity of plasma neuroexosomal mitochondrial proteins to detect preclinical AD before cognitive impairment in patients with T2DM and analyzed the correlation between exosomal biomarkers and changes in cognition, in AD pathogenic proteins, such as plasma neuroexosomal Aβ42, and tau proteins, and on MRI over time.
Research Design and Methods
Study Subjects
We enrolled participants with T2DM with MCI at the Endocrinology Clinic of Weihai Municipal Hospital between January 2017 and June 2017. All subjects provided written informed consent before inclusion, and the study followed the Declaration of Helsinki and was approved by the Weihai Municipal Hospital Ethics Committee. T2DM was diagnosed according to the 1999 World Health Organization criteria. The criteria for MCI included the presence of a subjective memory complaint, with a Mini-Mental State Examination (MMSE) score between 24 and 30, a Clinical Dementia Rating (CDR) of 0.5, preserved activities of daily living, and an absence of dementia (15). All subjects were included in the baseline population based on the following criteria: 1) patients with a T2DM disease duration of >3 years, 2) patients who were ≥55 years old, and 3) long-term residents of Weihai City. The exclusion criteria included cortical stroke, seizure, brain surgery, a history of traumatic brain injury, a concomitant neurologic disorder potentially affecting cognitive function (e.g., Parkinson disease), and being unable to comply with the study assessment. Subjects with MCI and depressive disorder were also excluded (16).
Data on demographic characteristics and vascular risk factors were collected. The patients underwent neuropsychological testing, including the MMSE, Montreal Cognitive Assessment (MoCA), and CDR, the 17-item version of the Hamilton Depression Rating Scale (HAM-D), and the Activities of Daily Living (ADL) scale. Brain MRI was conducted at baseline and during the follow-up period.
All patients with T2DM with MCI were followed up for >24 months (mean 40.6 months). The point of dementia conversion was determined by two neurologists based on criteria modified from the DSM-IV (17). The subjects with dementia were further assessed with brain MRI. Dementia of the Alzheimer type was diagnosed based on the National Institute of Neurological Disorders and Stroke (NINDS)–Alzheimer's Disease and Related Disorders Association (ADRDA) criteria for the clinical diagnosis of probable AD (18). The diagnosis of vascular dementia was based on the criteria of the NINDS–Association Internationale pour la Recherche et l'Enseignement en Neurosciences (AIREN) (19). AD patients were included in the current study. At the time of the last available assessment, MCI patients were classified as having progressive MCI (pMCI) if they progressed to developing AD at any time during follow-up or as having stable MCI (sMCI) (no progression to AD dementia during follow-up) if their diagnosis remained MCI (20). We excluded subjects who were diagnosed with MCI at baseline but reverted to cognitively normal (CN) during follow-up.
We also recruited age-, sex-, and education-matched patients with type 2 diabetes who were CN and AD patients. The criteria for CN subjects included no history of neurologic or psychiatric disorders, an MMSE score ranging between 27 and 30, and a CDR score of 0. AD patients fulfilled the NINDS-ADRDA criteria and had MMSE scores between 20 and 26 and a CDR score of 0.5 or 1.0.
Cognitive Assessment
Global cognition was assessed by MMSE and MoCA scores. MMSE and MoCA scores were selected at eight time points: baseline and at 6, 12, 18, 24, 30, 36, 42, and 48 months.
Brain MRI Data
MRI examinations were performed with a Magnetom Trio whole-body 3.0-T MR scanner (Siemens, Erlangen, Germany) with a 12-channel head-matrix coil and identical technical parameters at baseline and at the point when MCI patients converted to dementia during follow-up. All subjects underwent T1- and T2-weighted diffusion-weighted imaging scans and fluid-attenuated inversion recovery (FLAIR) sequence imaging. Sagittal three-dimensional T1-weighted structural images were acquired with the following parameters: repetition time/echo time = 2,300/2.98 ms; time inversion = 900 ms; field of view = 256 × 256 mm; flip angle = 9; section thickness = 1 mm; and 192 sagittal slices. The major parameters of FLAIR images were as follows: repetition time/echo time = 9,000/96 ms; field of view = 240 × 240 mm; matrix size = 512 × 512; and section thickness = 5 mm.
Normalized volumes of target brain compartments (hippocampal volumes, cortical thickness, and total gray matter volume) were quantified with the FreeSurfer image analysis suite (version 5.3.0), which is documented and freely available for download online at https://surfer.nmr.mgh.harvard.edu/. AD signature cortical thickness was defined by averaging cortical thickness of the entorhinal cortex, inferior temporal lobes, middle temporal lobes, and fusiform gyrus, with a lower value indicating more severe AD pathology (21). White matter hyperintensity (WMH) was defined as the presence of hyperintensity in the white matter area on FLAIR images. Periventricular hyperintensity (PVH) and deep WMH (DWMH) volumes were quantitatively analyzed by a neurologist using 3D Slicer semiautomated freeware (https://www.slicer.org).
Collection and Confirmation of NDEs From the Blood
A fasting blood sample of all participants was drawn at baseline between 6 and 7 a.m. and stored in a polypropylene tube containing EDTA. The blood samples were centrifuged at 4,000g for 10 min to obtain the plasma. Plasma (0.5 mL) was used to enrich exosomes. Specific NDEs were isolated according to our published protocol (22). In brief, ExoQuick exosome precipitation solution (EXOQ, EXOQ20A-1; System Biosciences, Palo Alto, CA), was used to collect total exosomes from plasma. NDEs were then isolated by coimmunoprecipitation using a rabbit anti-L1 cell adhesion molecule (L1CAM) antibody (13-1719-82; eBiosciences, San Diego, CA) and labeled with biotin by the EZ-Link Sulfo-NHS-Biotin system (53117; Thermo Fisher Scientific, Waltham, MA).
Western blotting (WB) and transmission electron microscopy (TEM) were performed to confirm the success of exosomal collection according to our previous protocols (22). The degree of purity was verified by WB with the positive exosomal marker TSG101 (Abcam, Cambridge, MA), HSP70 (Abcam), and negative exosomal marker calnexin (Proteintech, Rosemont, IL). L1CAM-positive plasma NDEs were characterized based on shape and size using TEM and nanoparticle tracking analysis.
Quantification of NDEs and ELISA
L1CAM-positive exosomal proteins were measured by ELISA kits for human NDUFS3 (abx381746; Abbexa Ltd., Cambridge, U.K.), human SDHB (abx383076, Abbexa Ltd.), and synaptosomal-associated protein 25 (SNAP-25) (ELH-SNAP25-1; RayBiotech, Norcross, GA). The amount of CD81 protein was measured by ELISA kits (ELH-CD81-1, RayBiotech) to normalize the relative values for each sample (23). The plasma neuroexosomal Aβ42, T-tau, and P-T181-tau levels were measured by ELISA kits according to our published protocol (19).
Statistical Analysis
SPSS 22.0 and MedCalc 19 statistical software were used for the statistical analysis. Tests for the homogeneity of variances were performed. The Kolmogorov-Smirnov test was also performed to ascertain the normality of the distribution of continuous variables. The statistical significance of differences between means for groups conforming to a normal distribution was determined with the Student unpaired t test or one-way ANOVA with the Bonferroni post hoc test. The variables with a nonnormal distribution were compared using the nonparametric Mann-Whitney U test or the Kruskal-Wallis test. Categorical variables were compared using the χ2 test. Spearman correlation was used to test associations between plasma neuroexosomal mitochondrial proteins and other core biomarkers. The relationship between mitochondrial proteins and brain structure (adjusted for age, sex, and the levels of fasting blood glucose) and cognition (adjusted for age, sex, education, and the levels of fasting blood glucose) was analyzed by multiple linear regression models. Diagnostic accuracy (the area under the curve [AUC]) for pMCI and AD was obtained for mitochondrial proteins using receiver operating characteristic (ROC) curve analyses. The associations of mitochondrial proteins with the conversion from MCI to AD were assessed by calculating hazard ratios (HRs) with 95% CIs using Cox proportional hazard regression analysis with adjustment for age and sex. All of the tests were two tailed, and the threshold for statistical significance was P < 0.05.
Data and Resource Availability
The data sets generated during and/or analyzed during the current study are not publicly available due to under construction but are available from the corresponding author upon reasonable request. No applicable resources were generated or analyzed during the current study.
Results
Baseline Demographic and Biomarker Characteristics of the Study Participants
The purity of plasma NDEs was also validated with WB by two exosomal positive markers TSG101 and HSP70, and one negative exosomal marker, calnexin (Supplementary Fig. 1A). TEM images show plasma NDEs that were successfully collected (Supplementary Fig. 1B). In Supplementary Fig. 1C, the size of the NDEs was directly determined by nanoparticle tracking analysis.
The demographic features and biomarker characteristics of CN participants, MCI patients, and AD dementia patients are shown in Table 1. The participants with T2DM in the study included 41 CN control subjects, 97 individuals with MCI (68 individuals with sMCI and 29 with pMCI), and 36 patients with AD dementia. There were no significant differences in age, sex, educational level, HbA1c, or vascular risk factors among the groups. Compared with MCI (pMCI and sMCI) and CN participants, patients with AD dementia had lower MMSE and MoCA scores. For participants with baseline MRI measurements, total hippocampal volume and AD signature cortical thickness were lower in patients with T2DM with pMCI (5.99 ± 0.79 cm3 and 2.47 ± 0.13 mm) than in sMCI patients (6.39 ± 0.71 cm3 and 2.55 ± 0.19 mm). Total hippocampal volume, AD signature cortical thickness, and total gray matter volume in AD patients (5.11 ± 0.84 cm3, 2.17 ± 0.33 mm, and 548.9 ± 24.8 cm3, respectively) were lower than those in MCI patients (pMCI: 5.99 ± 0.79 cm3, 2.47 ± 0.13 mm, and 607.5 ± 23.3 cm3, respectively, and sMCI: 6.39 ± 0.71 cm3, 2.55 ± 0.19 mm, and 609.8 ± 22.4 cm3, respectively) and CN participants (7.24 ± 0.73 cm3, 2.59 ± 0.18 mm, and 612.3 ± 27.5 cm3, respectively). The PVH in MCI (pMCI: 9,028 ± 6,726 mm3 and sMCI: 8,376 ± 6,026 mm3) and AD patients (1,1781 ± 8,921 mm3) was higher than that in CN participants (6,119 ± 4,464 mm3). The plasma neuroexosomal concentrations of Aβ42, T-tau, and P-T181-tau in AD patients (4.15 ± 0.60, 207.2 ± 36.5, and 89.4 ± 22.0 pg/mL, respectively) were higher than those in MCI patients (pMCI: 3.78 ± 0.75, 184.7 ± 25.3, and 66.8 ± 16.5 pg/mL, respectively, and sMCI: 3.38 ± 0.84, 167.7 ± 30.0, and 57.3 ± 13.2 pg/mL, respectively) and CN participants (3.17 ± 0.77, 163.5 ± 34.1, and 55.9 ± 10.2 pg/mL, respectively). Furthermore, the plasma neuroexosomal concentrations of Aβ42, T-tau, and P-T181-tau in pMCI patients were higher than those in sMCI patients and CN participants. Compared with MCI (pMCI: 530.6 ± 131.5 pg/mL and sMCI: 593.0 ± 130.4 pg/mL) and CN participants (627.6 ± 152.0 pg/mL), patients with AD dementia (461.4 ± 116.9 pg/mL) had lower exosomal concentrations of SNAP-25. The exosomal concentrations of SNAP-25 in pMCI patients were lower than those in sMCI and CN participants.
Demographic and clinical characteristics of the participants at baseline
. | CN (n = 41) . | sMCI (n = 68) . | pMCI (n = 29) . | AD (n = 36) . |
---|---|---|---|---|
Age, years | 69.8 ± 7.1 | 71.0 ± 8.0 | 72.6 ± 7.7 | 72.0 ± 7.1 |
Education, years | 9.1 ± 4.9 | 8.1 ± 5.3 | 8.8 ± 4.5 | 8.0 ± 4.9 |
Female sex | 22 (53.6) | 41 (60.3) | 16 (55.2) | 20 (55.6) |
Duration of type 2 diabetes, years | 9.2 ± 3.6 | 9.6 ± 3.3 | 10.8 ± 4.2 | 11.0 ± 4.1a |
Fasting blood glucose, mmol/L | 7.58 ± 1.86 | 7.75 ± 2.09 | 8.23 ± 2.05 | 8.40 ± 2.56 |
HbA1c, % | 7.9 ± 1.4 | 7.7 ± 2.0 | 7.9 ± 2.2 | 8.6 ± 2.7 |
HbA1c, mmol/mol | 62.5 ± 16.1 | 60.7 ± 22.2 | 62.5 ± 24.3 | 70.0 ± 39.2 |
BMI, kg/m2 | 24.9 ± 2.4 | 25.1 ± 3.1 | 25.0 ± 2.9 | 25.3 ± 3.0 |
Hypertension | 10 (24.4) | 12 (17.6) | 6 (20.7) | 7 (19.4) |
Hyperlipidemia | 20 (48.8) | 31 (45.6) | 14 (48.3) | 17 (47.2) |
Current smoker | 4 (9.8) | 5 (7.4) | 2 (6.9) | 3 (8.3) |
Current drinker | 7 (17.1) | 11 (16.1) | 4 (13.8) | 5 (13.9) |
MMSE | 28.8 ± 1.0 | 25.4 ± 1.5a | 25.3 ± 1.5a | 23.2 ± 1.7a,b,c |
MoCA | 26.8 ± 1.7 | 21.6 ± 1.9a | 21.5 ± 1.6a | 18.2 ± 3.4a,b,c |
Total hippocampal volume, cm3 | 7.24 ± 0.73 | 6.39 ± 0.71a | 5.99 ± 0.79a,b | 5.11 ± 0.84a,b,c |
AD signature cortical thickness, mm | 2.59 ± 0.18 | 2.55 ± 0.19 | 2.47 ± 0.13a,b | 2.17 ± 0.33a,b,c |
Total gray matter volume, cm3 | 612.3 ± 27.5 | 609.8 ± 22.4 | 607.5 ± 23.3 | 548.9 ± 24.8a,b,c |
PVH, mm3 | 6,119 ± 4,464 | 8,376 ± 6,026a | 9,028 ± 6,726a | 11,781 ± 8,921a,b |
DWMH, mm3 | 2,871 ± 4,756 | 3,836 ± 4,808 | 4,367 ± 5,179 | 6,103 ± 6,789 |
NDEs Aβ42, pg/mL | 3.17 ± 0.77 | 3.38 ± 0.84 | 3.78 ± 0.75a,b | 4.15 ± 0.60a,b,c |
NDEs T-tau, pg/mL | 163.5 ± 34.1 | 167.7 ± 30.0 | 184.7 ± 25.3a,b | 207.2 ± 36.5a,b,c |
NDEs P-T181-tau, pg/mL | 55.9 ± 10.2 | 57.3 ± 13.2 | 66.8 ± 16.5a,b | 89.4 ± 22.0a,b,c |
NDEs SNAP-25, pg/mL | 627.6 ± 152.0 | 593.0 ± 130.4 | 530.6 ± 131.5a,b | 461.4 ± 116.9a,b,c |
. | CN (n = 41) . | sMCI (n = 68) . | pMCI (n = 29) . | AD (n = 36) . |
---|---|---|---|---|
Age, years | 69.8 ± 7.1 | 71.0 ± 8.0 | 72.6 ± 7.7 | 72.0 ± 7.1 |
Education, years | 9.1 ± 4.9 | 8.1 ± 5.3 | 8.8 ± 4.5 | 8.0 ± 4.9 |
Female sex | 22 (53.6) | 41 (60.3) | 16 (55.2) | 20 (55.6) |
Duration of type 2 diabetes, years | 9.2 ± 3.6 | 9.6 ± 3.3 | 10.8 ± 4.2 | 11.0 ± 4.1a |
Fasting blood glucose, mmol/L | 7.58 ± 1.86 | 7.75 ± 2.09 | 8.23 ± 2.05 | 8.40 ± 2.56 |
HbA1c, % | 7.9 ± 1.4 | 7.7 ± 2.0 | 7.9 ± 2.2 | 8.6 ± 2.7 |
HbA1c, mmol/mol | 62.5 ± 16.1 | 60.7 ± 22.2 | 62.5 ± 24.3 | 70.0 ± 39.2 |
BMI, kg/m2 | 24.9 ± 2.4 | 25.1 ± 3.1 | 25.0 ± 2.9 | 25.3 ± 3.0 |
Hypertension | 10 (24.4) | 12 (17.6) | 6 (20.7) | 7 (19.4) |
Hyperlipidemia | 20 (48.8) | 31 (45.6) | 14 (48.3) | 17 (47.2) |
Current smoker | 4 (9.8) | 5 (7.4) | 2 (6.9) | 3 (8.3) |
Current drinker | 7 (17.1) | 11 (16.1) | 4 (13.8) | 5 (13.9) |
MMSE | 28.8 ± 1.0 | 25.4 ± 1.5a | 25.3 ± 1.5a | 23.2 ± 1.7a,b,c |
MoCA | 26.8 ± 1.7 | 21.6 ± 1.9a | 21.5 ± 1.6a | 18.2 ± 3.4a,b,c |
Total hippocampal volume, cm3 | 7.24 ± 0.73 | 6.39 ± 0.71a | 5.99 ± 0.79a,b | 5.11 ± 0.84a,b,c |
AD signature cortical thickness, mm | 2.59 ± 0.18 | 2.55 ± 0.19 | 2.47 ± 0.13a,b | 2.17 ± 0.33a,b,c |
Total gray matter volume, cm3 | 612.3 ± 27.5 | 609.8 ± 22.4 | 607.5 ± 23.3 | 548.9 ± 24.8a,b,c |
PVH, mm3 | 6,119 ± 4,464 | 8,376 ± 6,026a | 9,028 ± 6,726a | 11,781 ± 8,921a,b |
DWMH, mm3 | 2,871 ± 4,756 | 3,836 ± 4,808 | 4,367 ± 5,179 | 6,103 ± 6,789 |
NDEs Aβ42, pg/mL | 3.17 ± 0.77 | 3.38 ± 0.84 | 3.78 ± 0.75a,b | 4.15 ± 0.60a,b,c |
NDEs T-tau, pg/mL | 163.5 ± 34.1 | 167.7 ± 30.0 | 184.7 ± 25.3a,b | 207.2 ± 36.5a,b,c |
NDEs P-T181-tau, pg/mL | 55.9 ± 10.2 | 57.3 ± 13.2 | 66.8 ± 16.5a,b | 89.4 ± 22.0a,b,c |
NDEs SNAP-25, pg/mL | 627.6 ± 152.0 | 593.0 ± 130.4 | 530.6 ± 131.5a,b | 461.4 ± 116.9a,b,c |
Data are presented as mean ± SD. AD signature cortical thickness: cortical thickness in AD signature regions calculated as the average of cortical thickness in entorhinal, inferior temporal, middle temporal, and fusiform regions.
Significant at P < 0.05 vs. CN.
Significant at P < 0.05 vs. sMCI.
Significant at P < 0.05 vs. pMCI.
We assessed the use of medications by all of the enrolled patients. The results indicate that there were no significant differences in medication use between the groups (Supplementary Table 1).
Plasma Neuroexosomal Mitochondrial Protein Levels in Different Diagnostic Groups
Plasma neuroexosomal NDUFS3 (Fig. 1A) and SDHB (Fig. 1B) levels were significantly lower in patients with T2DM with AD dementia (232.7 ± 63.4 and 1,360.7 ± 328.5 pg/mL) and pMCI (274.4 ± 78.6 and 1,536.7 ± 342.8 pg/mL) than in CN subjects (333.9 ± 96.7 and 2,050.4 ± 628.9 pg/mL; P < 0.001 for both groups). Lower neuroexosomal NDUFS3 (Fig. 1A) and SDHB (Fig. 1B) levels were found in AD dementia (232.7 ± 63.4 and 1,360.7 ± 328.5 pg/mL) than in sMCI (319.9 ± 109.8 pg/mL, P < 0.001; 1,824.7 ± 606.4 pg/mL, P < 0.001) and pMCI (274.4 ± 78.6 pg/mL, P < 0.05; 1,536.7 ± 342.8 pg/mL, P < 0.05). We also found that plasma neuroexosomal NDUFS3 (Fig. 1A) and SDHB (Fig. 1B) levels were lower in pMCI (274.4 ± 78.6 and 1,536.7 ± 342.8 pg/mL) than in sMCI (319.9 ± 109.8 pg/mL, P < 0.05; 1,824.7 ± 606.4 pg/mL, P < 0.05) subjects.
Associations Between Plasma Neuroexosomal Mitochondrial Proteins and Aβ42
There were no significant associations between plasma neuroexosomal NDUFS3 and SDHB and Aβ42 in CN (r = −0.195, P = 0.222; r = −0.259, P = 0.102) or sMCI (r = −0.103, P = 0.405; r = −0.225, P = 0.065) subjects (Fig. 2A and B). NDUFS3 and SDHB were negatively correlated with Aβ42 in pMCI (NDUFS3: r = −0.462, P = 0.012; SDHB: r = −0.622, P < 0.001) and AD (r = −0.527, P = 0.001; r = −0.449, P = 0.006) patients (Fig. 2A and B).
Plasma NDEs NDUFS3 and SDHB levels in different diagnostic groups. Scatter plots show plasma NDEs NDUFS3 (A) and SDHB (B) levels in subjects with T2DM with CN, sMCI, pMCI, and AD. ***P < 0.001, **P < 0.01, *P < 0.05.
Plasma NDEs NDUFS3 and SDHB levels in different diagnostic groups. Scatter plots show plasma NDEs NDUFS3 (A) and SDHB (B) levels in subjects with T2DM with CN, sMCI, pMCI, and AD. ***P < 0.001, **P < 0.01, *P < 0.05.
Associations Between Plasma Neuroexosomal Mitochondrial Proteins and Tau Biomarkers
There were no significant associations between plasma neuroexosomal NDUFS3 and T-tau in CN subjects (r = −0.280, P = 0.076) (Fig. 3A). NDUFS3 was negatively correlated with P-181T-tau in CN subjects (r = −0.321, P = 0.041) (Fig. 3C). There were no significant associations between plasma neuroexosomal SDHB and T-tau (r = −0.289, P = 0.067) or P-181T-tau (r = −0.277, P = 0.079) in CN subjects (Fig. 3B and D). NDUFS3 and SDHB were not associated with T-tau in sMCI patients (Fig. 3A and B). There were significant associations of plasma neuroexosomal NDUFS3 and SDHB with P-181T-tau in sMCI subjects (r = −0.271, P = 0.026; r = −0.276, P = 0.022) (Fig. 3C and D). NDUFS3 and SDHB were negatively correlated with plasma neuroexosomal T-tau (r = −0.475, P = 0.009; r = −0.448, P = 0.015) (Fig. 3A and B) and P-181T-tau (r = −0.579, P = 0.001; r = −0.448, P = 0.015) (Fig. 3C and D) in pMCI patients. There were also significant associations of plasma neuroexosomal NDUFS3 and SDHB with T-tau (r = −0.492, P = 0.002; r = −0.583, P < 0.001) (Fig. 3A and B) and P-181T-tau (r = −0.589, P < 0.001; r = −0.699, P < 0.001) in AD patients (Fig. 3C and D).
Levels of mitochondrial proteins in relation to Aβ42 in plasma NDEs. Correlations between NDUFS3 (A) and SDHB (B) levels and Aβ42 in different diagnostic groups.
Levels of mitochondrial proteins in relation to Aβ42 in plasma NDEs. Correlations between NDUFS3 (A) and SDHB (B) levels and Aβ42 in different diagnostic groups.
Associations Between Plasma Neuroexosomal Mitochondrial Proteins and SNAP-25
There were no significant associations between plasma neuroexosomal NDUFS3 and SDHB and SNAP-25 in CN (r = 0.126, P = 0.433; r = 0.218, P = 0.170) (Fig. 4A and B) and sMCI (r = 0.205, P = 0.094; r = 0.214, P = 0.080) subjects (Fig. 4A and B). Plasma neuroexosomal NDUFS3 and SDHB were both positively correlated with SNAP-25 in pMCI (r = 0.614, P < 0.001; r = 0.633, P < 0.001) (Fig. 4A and B) and AD (r = 0.547, P = 0.001; r = 0.623, P < 0.001) patients (Fig. 4A and B).
Levels of mitochondrial proteins in relation to tau biomarkers in plasma NDEs. Correlations between NDUFS3 (A) and SDHB (B) levels and T-tau in different diagnostic groups. Correlations between NDUFS3 (C) and SDHB (D) levels and P-181-tau in different diagnostic groups.
Levels of mitochondrial proteins in relation to tau biomarkers in plasma NDEs. Correlations between NDUFS3 (A) and SDHB (B) levels and T-tau in different diagnostic groups. Correlations between NDUFS3 (C) and SDHB (D) levels and P-181-tau in different diagnostic groups.
Plasma Neuroexosomal Mitochondrial Proteins in Relation to Brain Structure and WMH
The associations of mitochondrial proteins with baseline brain structure and WMH are shown in Table 2. Both NDUFS3 and SDHB were correlated with total hippocampal volumes in pMCI (β = 0.374, P = 0.011; β = 0.390, P = 0.012) and AD (β = 0.556, P = 0.002; β = 0.461, P = 0.012) patients. There were also significant associations of NDUFS3 and SDHB with AD signature cortical thickness in pMCI (β = 0.487, P = 0.008; β = 0.478, P = 0.014) and AD (β = 0.457, P = 0.012; β = 0.436, P = 0.015) patients. NDUFS3 was correlated with total gray matter volume only in the AD group (β = 0.420, P = 0.023). Except for weak associations of NDUFS3 and SDHB with baseline PVH volumes in the AD group, neither baseline NDUFS3 nor SDHB levels were correlated with WMH (PVH and DWMH) volumes in the other groups.
Analysis of association between mitochondrial proteins and baseline brain structure and WMH
. | Total hippocampal volumes . | AD signature cortical thickness . | Total gray matter volume . | PVH . | DWMH . | |||||
---|---|---|---|---|---|---|---|---|---|---|
β . | P . | β . | P . | β . | P . | β . | P . | β . | P . | |
CN | ||||||||||
NDUFS3 | −0.197 | 0.210 | −0.075 | 0.657 | 0.021 | 0.901 | −0.220 | 0.196 | −0.086 | 0.617 |
SDHB | −0.184 | 0.243 | −0.189 | 0.362 | −0.091 | 0.588 | −0.227 | 0.181 | −0.155 | 0.368 |
sMCI | ||||||||||
NDUFS3 | 0.112 | 0.279 | 0.206 | 0.097 | 0.063 | 0.616 | −0.176 | 0.159 | −0.147 | 0.240 |
SDHB | 0.178 | 0.086 | 0.166 | 0.183 | −0.015 | 0.905 | −0.143 | 0.256 | −0.129 | 0.304 |
pMCI | ||||||||||
NDUFS3 | 0.374 | 0.011 | 0.487 | 0.008 | 0.374 | 0.065 | −0.131 | 0.513 | −0.153 | 0.460 |
SDHB | 0.390 | 0.012 | 0.478 | 0.014 | 0.380 | 0.073 | −0.122 | 0.560 | −0.261 | 0.223 |
AD | ||||||||||
NDUFS3 | 0.556 | 0.002 | 0.457 | 0.012 | 0.420 | 0.023 | −0.364 | 0.051 | −0.307 | 0.088 |
SDHB | 0.461 | 0.012 | 0.436 | 0.015 | 0.317 | 0.085 | −0.334 | 0.069 | −0.253 | 0.154 |
. | Total hippocampal volumes . | AD signature cortical thickness . | Total gray matter volume . | PVH . | DWMH . | |||||
---|---|---|---|---|---|---|---|---|---|---|
β . | P . | β . | P . | β . | P . | β . | P . | β . | P . | |
CN | ||||||||||
NDUFS3 | −0.197 | 0.210 | −0.075 | 0.657 | 0.021 | 0.901 | −0.220 | 0.196 | −0.086 | 0.617 |
SDHB | −0.184 | 0.243 | −0.189 | 0.362 | −0.091 | 0.588 | −0.227 | 0.181 | −0.155 | 0.368 |
sMCI | ||||||||||
NDUFS3 | 0.112 | 0.279 | 0.206 | 0.097 | 0.063 | 0.616 | −0.176 | 0.159 | −0.147 | 0.240 |
SDHB | 0.178 | 0.086 | 0.166 | 0.183 | −0.015 | 0.905 | −0.143 | 0.256 | −0.129 | 0.304 |
pMCI | ||||||||||
NDUFS3 | 0.374 | 0.011 | 0.487 | 0.008 | 0.374 | 0.065 | −0.131 | 0.513 | −0.153 | 0.460 |
SDHB | 0.390 | 0.012 | 0.478 | 0.014 | 0.380 | 0.073 | −0.122 | 0.560 | −0.261 | 0.223 |
AD | ||||||||||
NDUFS3 | 0.556 | 0.002 | 0.457 | 0.012 | 0.420 | 0.023 | −0.364 | 0.051 | −0.307 | 0.088 |
SDHB | 0.461 | 0.012 | 0.436 | 0.015 | 0.317 | 0.085 | −0.334 | 0.069 | −0.253 | 0.154 |
AD signature cortical thickness: cortical thickness in AD signature regions calculated as the average of cortical thickness in entorhinal, inferior temporal, middle temporal, and fusiform regions.
During the follow-up period, 29 MCI patients progressed to developing AD and underwent further brain MRI at the point of dementia conversion. Analysis of the correlation between mitochondrial proteins and changes in brain structure or WMH was performed. Lower baseline NDUFS3 and SDHB levels were correlated with volumetric loss in the total hippocampus (β = −0.537, P = 0.005; β = −0.578, P = 0.004) and total gray matter (β = −0.511, P = 0.009; β = −0.484, P = 0.019), and reduced AD signature cortical thickness (β = −0.427, P = 0.031; β = −0.442, P = 0.033). There were no significant associations between baseline NDUFS3 or SDHB levels and increased PVH or DWMH volumes (data not listed).
Plasma Neuroexosomal Mitochondrial Proteins in Relation to Cognition and Future Cognitive Changes
There were no significant associations of plasma neuroexosomal NDUFS3 and SDHB with baseline MoCA scores in CN (β = 0.067, P = 0.694; β = 0.139, P = 0.414) and sMCI (β = 0.133, P = 0.279; β = 0.203, P = 0.094) subjects. SDHB levels were correlated with baseline MoCA scores in pMCI (β = 0.451, P = 0.035) patients. There were no significant associations between NDUFS3 and baseline MoCA scores in pMCI (β = 0.303, P = 0.134) patients. Both NDUFS3 and SDHB were associated with baseline MoCA scores in AD patients (β = 0.431, P = 0.022; β = 0.422, P = 0.019).
Low plasma neuroexosomal NDUFS3 and SDHB levels were correlated with a more rapid decrease in MoCA scores in pMCI during the clinical follow-up period (β = 0.425, P = 0.039; β = 0.500, P = 0.023).
Diagnostic Power of Plasma Neuroexosomal Mitochondrial Proteins for pMCI and AD
The results obtained from the ROC curve analyses of the pMCI patients and CN subjects revealed that biomarkers in plasma NDEs had lower diagnostic value for patients with pMCI (Table 3). Compared with Aβ42, T-tau, and SNAP-25, NDUFS3 and SDHB had almost the same range of diagnostic accuracy for AD (NDUFS3 vs. Aβ42, P = 0.594; NDUFS3 vs. T-tau, P = 0.862; NDUFS3 vs. SNAP-25, P = 0.724; SDHB vs. Aβ42, P = 0.975; SDHB vs. T-tau, P = 0.468; SDHB vs. SNAP-25, P = 0.340) (Fig. 5). Compared with P-181T-tau, SDHB also had almost the same range of diagnostic accuracy for AD (P = 0.114). However, P-181T-tau provided higher diagnostic accuracy than NDUFS3 for AD (P = 0.029) (Fig. 5).
Levels of mitochondrial proteins in relation to SNAP-25 in plasma NDEs. Correlations between NDUFS3 (A) and SDHB (B) levels and SNAP-25 in different diagnostic groups.
Levels of mitochondrial proteins in relation to SNAP-25 in plasma NDEs. Correlations between NDUFS3 (A) and SDHB (B) levels and SNAP-25 in different diagnostic groups.
AUC of biomarkers in plasma NDEs
. | Aβ42 . | T-tau . | P-181T-tau . | SNAP-25 . | NDUFS3 . | SDHB . |
---|---|---|---|---|---|---|
pMCI | 0.707 (0.585–0.829) | 0.697 (0.554–0.803) | 0.714 (0.592–0.836) | 0.678 (0.544–0.802) | 0.680 (0.555–0.805) | 0.746 (0.633–0.859) |
P | 0.003 | 0.011 | 0.002 | 0.012 | 0.011 | <0.001 |
AD | 0831 (0.742–0.920) | 0.790 (0.691–0.889) | 0.907 (0.839–0.975) | 0.779 (0.678–0.880) | 0.801 (0.705–0.896) | 0.833 (0.745–0.920) |
P | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 |
. | Aβ42 . | T-tau . | P-181T-tau . | SNAP-25 . | NDUFS3 . | SDHB . |
---|---|---|---|---|---|---|
pMCI | 0.707 (0.585–0.829) | 0.697 (0.554–0.803) | 0.714 (0.592–0.836) | 0.678 (0.544–0.802) | 0.680 (0.555–0.805) | 0.746 (0.633–0.859) |
P | 0.003 | 0.011 | 0.002 | 0.012 | 0.011 | <0.001 |
AD | 0831 (0.742–0.920) | 0.790 (0.691–0.889) | 0.907 (0.839–0.975) | 0.779 (0.678–0.880) | 0.801 (0.705–0.896) | 0.833 (0.745–0.920) |
P | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 |
Plasma Neuroexosomal Mitochondrial Proteins Predict Conversion From MCI to AD
We analyzed whether plasma neuroexosomal mitochondrial proteins predicted conversion from MCI to AD. Cox proportional hazard regression analysis was performed for NDUFS3 and SDHB as continuous variables after adjusting for age and sex. Only SDHB significantly predicted conversion from MCI to AD. The HR was then calculated for SDHB as a dichotomous variable using the median values of SDHB as a cutoff (adjusting for age and sex). Subjects with low SDHB (HR 0.387, P = 0.018), corresponding to subjects whose SDHB values were ≤1,645.8 pg/mL, progressed much more rapidly to AD than subjects with higher values (>1,645.8 pg/mL, corresponding to the higher median values of SDHB) (Fig. 6).
ROC curve analyses were performed to test the plasma neuroexosomal mitochondrial proteins in relation to clinical diagnoses for AD.
ROC curve analyses were performed to test the plasma neuroexosomal mitochondrial proteins in relation to clinical diagnoses for AD.
Baseline SDHB in plasma NDEs as predictors of conversion from MCI to AD. Survival from AD as a function of SDHB in plasma NDEs measures (dichotomized at the median values) are shown. Analyses were adjusted for age and sex. Cutoff value was 1,645.8 pg/mL for SDHB.
Baseline SDHB in plasma NDEs as predictors of conversion from MCI to AD. Survival from AD as a function of SDHB in plasma NDEs measures (dichotomized at the median values) are shown. Analyses were adjusted for age and sex. Cutoff value was 1,645.8 pg/mL for SDHB.
Discussion
In the present longitudinal study of diabetes, we investigated the associations of plasma neuroexosomal mitochondrial proteins (NDUFS3 and SDHB) with other key biomarkers across the AD spectrum. Additional evidence in the results showed that mitochondrial proteins reflect the AD pathophysiologic process and are able to distinguish between diagnostic groups. Finally, low mitochondrial protein levels were predictive of clinical conversion from MCI to AD and brain structure injury.
Mitochondrial dysfunction is a key feature of both diabetes and neurodegenerative diseases, including AD. Mitochondrial dysfunction and oxidative stress have been extensively reported in patients with diabetes and AD as well as in rodent models of all of these conditions. For example, a large decrease in the NDUFS3 protein subunit of complex I decreased the level of the mRNA and impaired the catalytic activity of the complex in diabetic rats compared with controls (24). In addition, the activity of succinate dehydrogenase (complex II), a key marker of mitochondrial content, was reduced by 10–30% in diabetic versus control mice (25). In rat primary cortical neurons, high glucose concentrations cause decreased mitochondrial respiration, protein expression of peroxisome proliferator–activated receptor γ coactivator 1-α (PGC1α) and complex I of the electron transport chain, and insulin resistance (26). A recent study found that impairment of the respiratory chain has also been observed in mitochondria isolated from sucrose-treated wild-type and AD transgenic mice (27). Some studies have found that mitochondrial complexes, including subunits of the electron transport chain (including NDUFS3 and SDHB) and ATP synthase, were altered in AD patients (13,28,29). Furthermore, mitochondrial dysfunction in the nervous system is considered a key pathophysiologic feature of early-stage AD (30,31) and could be a promising biomarker for early AD by measuring mitochondrial function (32). In the current study we show that, comparing CN and sMCI participants with diabetes, the protein cargo of NDEs in pMCI and AD patients was characterized by lower levels of the mitochondrial components NDUFS3 (complex I) and SDHB (complex II). Plasma neuroexosomal SDHB offers predictive value for future disease progression in MCI subjects. This finding suggests that neuroexosomal mitochondrial proteins (NDUFS3 and SDHB) are an early pathophysiological indicator of AD-related mitochondrial dysfunction.
In the diabetic brain, both Aβ and tau can cause mitochondrial alterations leading to neuronal energy deficits, synaptic disturbances, and neurodegeneration (11). In our study, there were significant correlations of the levels of plasma neuroexosomal mitochondrial proteins (NDUFS3 and SDHB) with Aβ42, tau, and SNAP-25 (markers indicating synaptic damage) in pMCI and AD patients. Hyperglycemia exacerbated mitochondrial defects, synaptic injury, and cognitive dysfunction in the brains of transgenic AD mice overexpressing Aβ, as shown by decreased mitochondrial respiratory complex I enzyme activity and a greatly decreased mitochondrial respiratory rate (33). Aβ has toxic effects on mitochondrial respiration, synthesis of ATP, and the activities of various enzymes related to energy production, including the I and II enzyme complexes in the mitochondrial respiratory chain (34). Overexpression and hyperphosphorylation of tau appear to impair mitochondrial axonal transportation, mitochondrial dynamics and function, and finally, neuronal health (35). Moreover, mitochondrial dysfunction is also involved in promoting tau pathology in AD (36,37). Mitochondria are essential for synaptic function by providing energy and regulating intrasynaptic metabolic homeostasis (38,39). Several years of research have shown that synaptic pathology and mitochondrial oxidative damage, caused by Aβ and phosphorylated tau, are early events in AD progression (40).
A recent multicenter study showed that the levels of Aβ42 and tau biomarkers in NDEs were highly correlated with their levels in cerebrospinal fluid and confirmed that plasma neuroexosomal Aβ42, T-tau, and P-T181-tau have the same capacity as those in cerebrospinal fluid for the diagnosis of AD (6). In the current study, our results showed that mitochondrial proteins (NDUFS3 and SDHB) offer diagnostic sensitivity for AD that is comparable to that of Aβ42, T-tau, and P-T181-tau in plasma NDEs. In addition, we found associations between low neuroexosomal mitochondrial protein levels and volumetric loss in the total hippocampus and total gray matter as well as reduced AD signature cortical thickness in pMCI. These findings therefore suggest that plasma neuroexosomal mitochondrial proteins might be potential diagnostic biomarkers for early-stage AD.
Limitations of the Study
Some limitations of our study should be considered when interpreting the results. First, the results were drawn from a small-scale hospital-based study, and future investigations are necessary to replicate and validate our findings in a large population of patients. Second, we did not have additional information, such as cerebrospinal fluid data and pathological evidence or inflammatory biomolecules, to confirm the results. Third, the functional study of neuroexosomal mitochondrial proteins, such as NDUSFS3 and SDHB, has not been investigated in the diabetic condition with AD. Fourth, L1CAM, as a commonly used specific marker for isolations of the NDEs, might not be the best candidate because it is also present in other tissues and organs, and controversially, its association with exosomes has been challenged (41). Fifth, our results are limited to a cohort of elderly individuals with type 2 diabetes, therefore, they cannot be generalized to the general population.
Conclusions
In summary, we identified significantly decreased plasma neuroexosomal mitochondrial proteins in the predemential stages of AD in participants with diabetes, and lower concentrations were correlated with higher hippocampal and gray matter atrophy, reduced AD signature cortical thickness, and a reduced rate of cognitive decline in some stages of AD. These findings support the use of neuroexosomal mitochondrial proteins as potential diagnostic biomarkers in AD.
H.C. and R.Y. contributed equally to this work.
This article contains supplementary material online at https://doi.org/10.2337/figshare.19337807.
Article Information
Funding. This work was supported by the Development Plan of Medical Sciences of Shandong Province (2019 WS225).
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. H.C. and R.Y. drafted the manuscript. H.C., R.Y., C.S., B.L., T.W., and S.Z. performed experiments. T.S., H.S., and J.Z. analyzed and interpreted data. M.L. performed neuroimaging analyses. Y.Y., Z.L., and J.Z. revised the manuscript. The final version of the manuscript was approved by all authors. J.Z. conceived and designed the study. J.Z. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.