Gaussian Process-based prediction of memory performance and biomarker status in ageing and Alzheimer's disease-A systematic model evaluation

Neuroimaging markers based on Magnetic Resonance Imaging (MRI) combined with various other measures (such as genetic covariates, biomarkers, vascular risk factors, neuropsychological tests etc.) might provide useful predictions of clinical outcomes during the progression towards Alzheimer's disease (AD). The use of multiple features in predictive frameworks for clinical outcomes has become increasingly prevalent in AD research. However, many studies do not focus on systematically and accurately evaluating combinations of multiple input features. Hence, the aim of the present work is to explore and assess optimal combinations of various features for MR-based prediction of (1) cognitive status and (2) biomarker positivity with a multi-kernel learning Gaussian process framework. The explored features and parameters included (A) combinations of brain tissues, modulation, smoothing, and image resolution; (B) incorporating demographics & clinical covariates; (C) the impact of the size of the training data set; (D) the influence of dimensionality reduction and the choice of kernel types. The approach was tested in a large German cohort including 959 subjects from the multicentric longitudinal study of cognitive impairment and dementia (DELCODE). Our evaluation suggests the best prediction of memory performance was obtained for a combination of neuroimaging markers, demographics, genetic information (ApoE4) and CSF biomarkers explaining 57% of outcome variance in out-of-sample predictions. The highest performance for Aβ42/40 status classification was achieved for a combination of demographics, ApoE4, and a memory score while usage of structural MRI further improved the classification of individual patient's pTau status.

Dietary patterns are related to cognitive functioning in elderly enriched with individuals at increased risk for Alzheimer's disease

Purpose

To investigate cross-sectional associations between dietary patterns and cognitive functioning in elderly free of dementia.

Methods

Data of 389 participants from the German DELCODE study (52% female, 69 ± 6 years, mean Mini Mental State Score 29 ± 1) were included. The sample was enriched with elderly at increased risk for Alzheimer’s disease (AD) by including participants with subjective cognitive decline, mild cognitive impairment (MCI) and siblings of AD patients. Mediterranean and MIND diets were derived from 148 Food Frequency Questionnaire items, and data-driven patterns by principal component analysis (PCA) of 39 food groups. Associations between dietary patterns and five cognitive domain scores were analyzed with linear regression analyses adjusted for demographics (model 1), and additionally for energy intake, BMI, other lifestyle variables and APOe4-status (model 2). For PCA-derived dietary components, final model 3 included all other dietary components.

Results

In fully adjusted models, adherence to Mediterranean and MIND diet was associated with better memory. The ‘alcoholic beverages’ PCA component was positively associated with most cognitive domains. Exclusion of MCI subjects (n = 60) revealed that Mediterranean and MIND diet were also related to language functions; associations with the alcoholic beverages component were attenuated, but most remained significant.

Conclusion

In line with data from elderly population samples, Mediterranean and MIND diet and some data-derived dietary patterns were related to memory and language function. Longitudinal data are needed to draw conclusions on the putative effect of nutrition on the rate of cognitive decline, and on the potential of dietary interventions in groups at increased risk for AD.

Electronic supplementary material

The online version of this article (10.1007/s00394-020-02257-6) contains supplementary material, which is available to authorized users.

Resting-state global functional connectivity as a biomarker of cognitive reserve in mild cognitive impairment

Cognitive reserve (CR) shows protective effects in Alzheimer's disease (AD) and reduces the risk of dementia. Despite the clinical significance of CR, a clinically useful diagnostic biomarker of brain changes underlying CR in AD is not available yet. Our aim was to develop a fully-automated approach applied to fMRI to produce a biomarker associated with CR in subjects at increased risk of AD. We computed resting-state global functional connectivity (GFC), i.e. the average connectivity strength, for each voxel within the cognitive control network, which may sustain CR due to its central role in higher cognitive function. In a training sample including 43 mild cognitive impairment (MCI) subjects and 24 healthy controls (HC), we found that MCI subjects with high CR (> median of years of education, CR+) showed increased frequency of high GFC values compared to MCI-CR- and HC. A summary index capturing such a surplus frequency of high GFC was computed (called GFC reserve (GFC-R) index). GFC-R discriminated MCI-CR+ vs. MCI-CR-, with the area under the ROC = 0.84. Cross-validation in an independently recruited test sample of 23 MCI subjects showed that higher levels of the GFC-R index predicted higher years of education and an alternative questionnaire-based proxy of CR, controlled for memory performance, gray matter of the cognitive control network, white matter hyperintensities, age, and gender. In conclusion, the GFC-R index that captures GFC changes within the cognitive control network provides a biomarker candidate of functional brain changes of CR in patients at increased risk of AD.

Effects of aging on default mode network activity in resting state fMRI: does the method of analysis matter?

Functional MRI (fMRI) of default mode network (DMN) brain activity during resting state is gaining attention as a potential non-invasive biomarker to diagnose incipient Alzheimer's disease. The aim of this study was to identify effects of normal aging on the DMN using different methods of fMRI processing and evaluation.

METHODS

fMRI was acquired in 17 young and 21 old healthy subjects and the data were analyzed with (a) volumes of interest (VOI)-based signal time course and (b) independent component analyses (ICA). In the first approach, the strength of DMN region inter-connectivity (as expressed with correlation coefficients) was of primary interest, the second method provided a measure of the magnitude of DMN co-activation.

RESULTS

The older subjects exhibited significantly lower DMN activity in the posterior cingulate (PCC, t-test P<.001) as well as a tendency to lower activity in all other DMN regions in comparison to the younger subjects. We found no significant effect of age on DMN inter-connectivity.

CONCLUSION

Effects of normal aging such as loss of PCC co-activity could be detected by ICA, but not by signal time course correlation analyses of DMN inter-connectivity. This either indicates lower sensitivity of inter-connectivity measures to detect subtle DMN changes or indicate that ICA and time course analyses determine different properties of DMN co-activation. Our results, therefore, provide fundamental knowledge for a potential future use of functional MRI as biomarker for neurodegenerative dementias where diminished DMN activity needs to be reliably differentiated from that observed in health aging.

CSF total and phosphorylated tau protein, regional glucose metabolism and dementia severity in Alzheimer's disease

BACKGROUND AND PURPOSE

We investigated associations between severity of cognitive impairment, cerebrospinal fluid (CSF) concentrations of total-tau (t-tau) protein and tau phosphorylated at threonin 181 (p-tau(181)) and regional glucose metabolism measured with 18F-fluorodeoxyglucose-positron emission tomography (18F-FDG-PET) in patients with probable Alzheimer's disease (AD).

METHODS

In 38 patients (mean age 66.5 +/- 8.0 years) with AD, Mini-Mental State Examination (MMSE) scores were evaluated and CSF levels of t-tau and p-tau(181) measured. All patients underwent an 18F-FDG-PET scan. Image analysis including correlation analysis and principal component analysis (PCA) were performed using SPM5 and VINCI.

RESULTS

Dementia severity (MMSE 21.2 +/- 4.9) correlated well with metabolic impairment especially in left hemisphere association areas that are typically affected in patients with AD (e.g. inferior parietal lobule, r = 0.512; medial temporal gyrus, r = 0.478; inferior temporal gyrus, r = 0.488; precuneus, r = 0.468; PCA: r = 0.639, F = 7.751; all P < 0.001). There were no associations between t-tau and p-tau(181) with dementia severity and only weak correlations between t-tau and cerebral glucose metabolism (superior parietal gyrus, r = -0.325, P < 0.05; precentral gyrus r = -0.418, P < 0.01; amygdala r = -0.362, P < 0.05). No correlations were found between p-tau(181) and regional hypometabolism in FDG-PET.

CONCLUSION

MMSE and CSF t-tau represent different aspects of disease severity. Whilst MMSE is closely related to impaired cerebral glucose metabolism, CSF t-tau is less closely related and appears to be less well suited for assessment of disease progression.

Multicenter assessment of CSF-phosphorylated tau for the prediction of conversion of MCI

BACKGROUND

The measurement of hyperphosphorylated tau (p-tau) in CSF has been proposed as a biomarker candidate for the prediction of Alzheimer disease (AD) in patients with mild cognitive impairment (MCI). However, a standard quantitative criterion of p-tau has not been evaluated.

OBJECTIVE

To assess in a multicenter study the predictive accuracy of an a priori defined criterion of tau phosphorylated at threonine 231 (p-tau(231)) for the prediction of conversion from MCI to AD during a short-term observation interval.

METHODS

The study included 43 MCI converters, 45 stable MCI (average follow-up interval = 1.5 years), and 57 healthy controls (at baseline only). Subjects were recruited at four international expert sites in a retrospective study design. Cox regression models stratified according to center were used to predict conversion status. Bootstrapped 95% CIs of classification accuracy were computed.

RESULTS

Levels of p-tau(231) were a significant predictor of conversion (B = 0.026, p = 0.001), independent of age, gender, Mini-Mental State Examination, and ApoE genotype. For an a priori-defined cutoff point (27.32 pg/mL), sensitivity ranged between 66.7 and 100% and specificity between 66.7 and 77.8% among centers. The bootstrapped mean percentage of correctly classified cases was 79.95% (95% CI = 79.9 to 80.00%). Post hoc defined cutoff values yielded a mean bootstrapped classification accuracy of 80.45% (95% CI = 80.24 to 80.76%).

CONCLUSIONS

An a priori defined cutoff value of p-tau(231) yields relatively stable results across centers, suggesting a good feasibility of a standard criterion of p-tau(231) for the prediction of Alzheimer disease.

Value of CSF beta-amyloid1-42 and tau as predictors of Alzheimer's disease in patients with mild cognitive impairment

Subjects with mild cognitive impairment (MCI) are at a high risk of developing clinical Alzheimer's disease (AD). We asked to what extent the core biomarker candidates cerebro-spinal fluid (CSF) beta-amyloid(1-42) (Abeta(1-42)) and CSF tau protein concentrations predict conversion from MCI to AD. We studied 52 patients with MCI, 93 AD patients, and 10 healthy controls (HC). The MCI group was composed of 29 patients who had converted to AD during follow-up, and of 23 patients who showed no cognitive decline. CSF Abeta(1-42) and tau protein levels were assessed at baseline in all subjects, using enzyme-linked immunosorbent assays. For assessment of sensitivity and specificity, we used independently established reference values for CSF Abeta(1-42) and CSF tau. The levels of CSF tau were increased, whereas levels of Abeta(1-42) were decreased in MCI subjects. Abeta(1-42) predicted AD in converted MCI with a sensitivity of 59% and a specificity of 100% compared to HC. Tau yielded a greater sensitivity of 83% and a specificity of 90%. In a multiple Cox regression analysis within the MCI group, low baseline levels of Abeta(1-42), but not other predictor variables (tau protein, gender, age, apolipoprotein E epsilon4 carrier status, Mini Mental Status Examination score, observation time, antidementia therapy), correlated with conversion status (P<0.05). Our findings support the notion that CSF tau and Abeta(1-42) may be useful biomarkers in the early identification of AD in MCI subjects.

Increased frequency of a new polymorphism in the cell division cycle 2 (cdc2) gene in patients with Alzheimer's disease and frontotemporal dementia

Recent studies show linkage between Alzheimer's disease (AD) and two loci on chromosome 10. The cell division cycle 2 (cdc2) gene is located close to one of the chromosome 10 markers, and is a candidate gene for AD since it is involved in the pathogenesis of AD. We sequenced coding exons and flanking intronic sequences and the promoter region on the cdc2 gene and found three new single nucleotide polymorphisms (SNPs). We analyzed 272 Caucasian AD cases, 160 controls and 70 cases with frontotemporal dementia (FTD) for these SNPs. Homozygosity for one of the SNPs (Ex6+7I/D) was more frequent in both AD and FTD cases than in controls. In the combined tauopathy (AD and FTD) group the odds ratio (OR) was 1.77 (95% CI 1.19-2.63) for the Ex6+7II genotype. Our findings suggest that the Ex6+7I allele is associated with tauopathies, both AD and FTD.

CSF tau protein phosphorylated at threonine 231 correlates with cognitive decline in MCI subjects

In this longitudinal study of 77 patients with mild cognitive impairment (MCI), the authors analyzed whether levels of tau protein phosphorylated at threonine 231 (p-tau(231)) in CSF correlate with progression of cognitive decline. High CSF p-tau(231) levels at baseline, but not total tau protein levels, correlated with cognitive decline and conversion from MCI to AD. Independently, old age and APOE-epsilon 4 carrier status were predictive as well. Our data indicate that an increased p-tau(231) level is a potential risk factor for cognitive decline in patients with MCI.

A new rapid landmark-based regional MRI segmentation method of the brain

BACKGROUND

Neurodegenerative and cerebrovascular diseases show a distinct distribution of regional atrophy and subcortical lesions.

OBJECTIVE

To develop an easily applicable landmark-based method for segmentation of the brain into the four cerebral lobes from MRI images.

METHOD

The segmentation method relies on a combination of anatomical landmarks and geometrical definitions. It is applied on the surface reconstruction of the MRI volume. The internal borders between the lobes are defined on the axial slices of the brain. The reliability of this method was determined from MRI scans of 10 subjects. To illustrate the use of the method, it was applied to MRI scans of an independent group of 10 healthy elderly subjects and 10 patients with vascular dementia to determine the regional distribution of white matter hyperintensities (WMH).

RESULTS

The intra-rater relative error (and intra-class correlation coefficient) of the lobe segmentation ranged from 1.6% to 6.9% (from 0.91 to 0.99). The inter-rater relative error (and intra-class correlation coefficient) ranged from 1.4% to 5.2% (from 0.96 to 0.99). Density of WMH was significantly higher in all four lobes in VD patients compared to controls (p<0.05). Within each group, WMH density was significantly higher in frontal and parietal than in temporal and occipital lobes (p<0.05).

CONCLUSION

This landmark based method can accommodate age and disease-related changes in brain morphology. It may be particularly useful for the study of neurodegenerative and cerebrovascular disease and for the validation of template-based automated techniques.

Reduced levels of amyloid beta-peptide antibody in Alzheimer disease

OBJECTIVE

To investigate whether it was possible to detect the presence and different levels of naturally occurring anti-beta-amyloid (Abeta) antibodies in the CSF of patients with AD and age-matched controls by employing a sensitive ELISA.

BACKGROUND

Immunization with preaggregated amyloid beta-peptide (Abeta(1-42)) and administration of antibodies against Abeta into amyloid precursor protein APP(V717F)- transgenic mice (an animal model of AD) have recently been reported to dramatically reduce amyloid plaque deposition, neuritic dystrophy, and astrogliosis, most likely by enhancing Abeta clearance from brain.

METHODS

A sensitive ELISA was performed to detect levels of naturally occurring anti-Abeta antibodies in the CSF of patients with AD and age-matched controls. Additionally, an immunoprecipitation assay was performed to confirm that naturally occurring anti-Abeta antibodies also exist in the human blood.

RESULT

- Naturally occurring antibodies directed against Abeta were found in the CSF and plasma of patients with AD and healthy control subjects. Moreover, CSF anti-Abeta antibody titers are significantly lower in patients with AD compared with healthy control subjects.

CONCLUSION

Naturally occurring antibodies directed against Abeta exist in human CSF and plasma. The CSF anti-Abeta antibody titers may be helpful in better understanding the effects of future immunologic therapies for AD.

Large-scale, multicenter study of cerebrospinal fluid tau protein phosphorylated at serine 199 for the antemortem diagnosis of Alzheimer's disease

We surveyed a total of 570 cerebrospinal fluid (CSF) samples from a variety of diseases, including Alzheimer's disease (AD; n = 236), non-AD-demented and nondemented diseases (n = 239), and normal controls (n = 95) to quantitate levels of tau protein phosphorylated at serine 199 (CSF/phospho-tau199) by a recently established sandwich ELISA. The CSF/phospho-tau199 levels in the AD group were significantly elevated compared to those in all the other non-AD groups. Receiver operating characteristics curves showed that the diagnostic sensitivity and specificity for the AD group vs all the other non-AD groups using the CSF/phospho-tau199 were 85.2% and 85.0%, respectively. Furthermore, there was a significant positive correlation between CSF/phospho-tau199 and CSF/total-tau levels in the AD group. Elevated CSF/phospho-tau199 in the AD group was noted irrespective of age, gender, dementia severity, and number of apolipoprotein E4 alleles. Thus, we suggest that CSF/phospho-tau199 may be a novel and logical biomarker in supporting antemortem diagnosis of AD.

Detection of tau phosphorylated at threonine 231 in cerebrospinal fluid of Alzheimer's disease patients

A new sandwich ELISA is described which shows specificity for tau phosphorylated at threonine 231 and preferentially reacts with Alzheimer's disease (AD) brain extracts relative to other dementias. This assay was used to analyze 58 antemortem cerebrospinal fluid samples. Twenty-three of 27 AD samples (85% sensitivity) yielded signals greater than the cutoff, while only one of 31 non-AD samples (97% specificity) were greater. This indicates that detection of phosphotau in cerebrospinal fluid with this sandwich ELISA could prove useful in the diagnosis of AD.