Study population

The study included 96 participants divided into three clinical groups based on their cognitive status, comprising 33 patients with amnesic MCI, 14 patients with AD and 49 HCs. AD, HC and MCI criteria used to recruit subjects in this study are described in our earlier work [4],[13]. Briefly, cognitive assessment was performed according to routine procedures utilized at the Fundaciè ACE Memory Clinic (Barcelona, Spain), as described elsewhere [14]. MCI subjects fulfilled the Petersen’s diagnostic criteria [15], including subjective memory complaints, normal general cognition, preserved performance in activities of daily living, absence of dementia and a measurable impairment in memory function, with or without deficit in other cognitive domains [16]. All MCI subjects had a Clinical Dementia Rating (CDR) of 0.5. On the basis of the Cut-off scores of a Brief Neuropsychological Battery (NBACE, a subtest of the Wechsler Memory Scale III) [14], impaired delayed verbal recall for which recognition testing does not improve performance classifies patients with amnesic MCI as having an “encoding/storage” pattern of memory loss. The diagnosis of AD was made according to National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association [17],[18] criteria, where AD is defined as a CDR of 1 point or more and a Mini Mental State Examination (MMSE) score below 24. HCs were cognitively normal when evaluated at the Fundaciè ACE Memory Clinic, had MMSE scores of at least 26 (taking into consideration the MMSE cutoff of <25 in the Spanish population [19]) and had a normal neuroimaging MRI profile.

Blood sampling and biochemical determinations

Blood samples from each participant were drawn on the morning after an overnight fast and were collected in polypropylene vials with ethylenediaminetetraacetic acid and a protease inhibitor cocktail (Complete Mini; Roche, Madrid, Spain). The samples were immediately cooled to 4 °C until processing, which occurred within 24 hours after collection. Blood samples were centrifuged, and both the plasma and the cellular pellet (CP) were divided into aliquots and stored in polypropylene tubes at -80°C until analyzed. The material was not thawed or refrozen at any time. All samples were analyzed in triplicate in the same run for each of the three blood fractions using two specific sandwich ELISA kits, ABtest 40 and ABtest 42 (Araclon Biotech, Zaragoza, Spain), as described elsewhere [4]. Before analysis, plasma and blood cell samples were pretreated using dilution in a formulated saline buffer with 1% blocking polymer according to the supplier’s instructions. We carried out three counts for both the Aβ40 and Aβ42 peptides in each blood sample. One count was performed using the undiluted plasma sample, another using the plasma sample diluted 1:3 with the aforementioned formulated buffer and a third using the CP that remained after plasma collection. The peptide amount in the undiluted plasma sample corresponded to the directly accessible (DA) peptide. The 1:3 dilution of the plasma was chosen because it provided the maximum peptide recovery from the sample (that is, the total in plasma (TP)). Thus, this count included the DA peptide and the peptide that was recovered from the plasma matrix. Additionally, the peptide associated with the CP was measured in a 1:5 dilution of the pellet that remained after plasma collection. The sum of these three amounts is described as the total Aβ pool in blood for either Aβ(1 - 40) or Aβ(1 - 42).

Brain imaging and magnetic resonance imaging analysis

All MRI scans were performed with a 1.5-T MRI scanner (Magnetom Symphony; Siemens Medical Solutions, Erlangen, Germany) at the Department of Diagnostic Imaging, Corachan Clinic, Barcelona. The protocol for the acquisition of the MRI data was identical for all patients and consisted of three-dimensional T1-weighted sagittal Magnetization-prepared rapid acquisition with gradient echo, two-dimensional (2D) axial T2-weighted turbo spin echo, 2D axial fluid-attenuated inversion recovery, 2D axial T2*-weighted gradient echo and 2D axial diffusion-weighted imaging. Brain images were also visually inspected by experienced clinicians who were blinded to the participants’ demographic, anthropometric and clinical data.

Cortical reconstruction and volumetric segmentation were performed with the Freesurfer image analysis suite, which is documented and freely available for download online. The technical details of these procedures are also described in prior publications (see [20],[21] and references therein).

Briefly, this processing method includes motion correction and averaging of multiple volumetric T1-weighted images (when more than one image is available), removal of nonbrain tissue performed by using a hybrid watershed/surface deformation procedure, an automated Talairach transformation, segmentation of the subcortical white matter and deep gray matter volumetric structures (including hippocampus, amygdala, caudate, putamen and ventricles), intensity normalization, tessellation of the gray-matter-white-matter boundary, automated topology correction and surface deformation following intensity gradients to optimally place the gray-white and gray-cerebrospinal fluid borders at the location where the greatest shift in intensity defines the transition to the other tissue class.

Once the cortical models are complete, a number of deformable procedures can be performed for further data processing and analysis, including surface inflation, registration to a spherical atlas by utilizing individual cortical folding patterns to match cortical geometry across subjects, parcellation of the cerebral cortex into units based on gyral and sulcal structures and creation of a variety of surface based data including maps of curvature and sulcal depth. With this method, both intensity and continuity information from the entire three-dimensional MR volume in segmentation and deformation procedures to produce representations of cortical thickness, which are calculated as the closest distance from the gray-matter-white-matter boundary to the gray-matter-CSF boundary at each vertex on the tessellated surface. The maps are created using spatial intensity gradients across tissue classes and are therefore not reliant on absolute signal intensity only. The maps produced are not restricted to the voxel resolution of the original data; thus, they are capable of detecting submillimetric differences between groups. It should be noted that the procedures described here for the measurement of cortical thickness have been validated against histological analysis [22] and manual measurements [23],[24]. Freesurfer morphometric procedures have been proven to show good test-retest reliability across scanner manufacturers and across field strengths [20],[21].

Alzheimer’s Disease Neuroimaging Initiative data

Aβ(1 - 40) plasma concentrations and Aβ(1 - 42) plasma levels of the ADNI project subjects were measured using module A of the INNO-BIA plasma Aβ forms immunoassay kit (Innogenetics, Ghent, Belgium; for research use-only reagents) on the Luminex 100 immunoassay platform and IS v.2.3 software (Luminex, Austin, TX, USA) with a fully automated sample preparation approach [25],[26]. MRI scan processing was performed with the Freesurfer image analysis software suite as explained above.

Statistical analysis

All the statistical procedures were carried out using R software [27]. Because of the shape of the empirical amyloid data distributions, the amyloid fractions were assumed following a lognormal distribution. Logarithms of Aβ concentrations were used in calculations. Because logarithms must be used only for dimensionless quantities, the Aβ fractions were transformed into dimensionless numbers by dividing them by a constant. Thus, for any fraction, a new quantity A β^ln was calculated as $A{\beta }^{\text{ln}}=\mathit{ln}\left(\frac{A{\beta }_i}{\left(\mathit{A\; \beta }\right)}\right)$, where i runs over individuals and 〈Aβ〉 is the median value for the fraction. This new quantity characterized the amyloid fraction distributions in plasma for the present study.

Pearson partial correlations were calculated between the MRI measurements for 45 ROIs given in the Freesurfer subcortical atlas and different Aβ^ln values with age, gender and ApoE, education and creatinine levels [28]-[30] as covariates. The same procedure was applied to ADNI data. Partial correlations were calculated with ApoE, age, gender and education and creatinine levels as covariates.

Pearson partial correlations adjusted for the covariates age, gender and ApoE, education and creatinine levels were also calculated between Aβ^ln values and cortical volumes and cortical thickness average for 68 ROIs in the Freesurfer cortical atlas. There was no correction for multiple comparisons in any case.

The whole procedure was repeated for both data sets, AB128 and ADNI, taking age, gender and ApoE, education and creatinine levels and phenotype as covariates. Partial correlations were calculated between the 45 Freesurfer subcortical atlas ROI volumes and Aβ^ln values and also between the 68 Freesurfer cortical atlas ROI volumes and thickness average values and Aβ^ln values. No correction for multiple comparisons was done.

An additional analysis was performed to determine the association between Aβ^ln CP Aβ(1 - 40) values and more correlated brain ROIs was conducted. The subjects were divided into phenotype groups (AD, MCI and HC), and each group was analyzed separately. Linear regressions were performed between Aβ^ln CP Aβ(1 - 40) and left hippocampal volume in each case.

Finally, Pearson partial correlations adjusted for the covariates age, gender and ApoE, education and creatinine levels were also calculated between Aβ^ln values for each amyloid fraction and MMSE score.