Early visual alterations in individuals at-risk of Alzheimer’s disease: a multidisciplinary approach

Participants

This investigation was a part of a project titled: “Características cognitivas y neurofisiológicas de personas con alto riesgo para el desarrollo de demencia: una aproximación multidimensional” (COGDEM), a prospective longitudinal study targeting the identification and progression of possible biomarkers capable of detecting subjects at higher risk of developing dementia in multiple domains [25]. The total sample of the study consisted of 251 cognitively healthy subjects, all of which signed the informed consent form. The research followed the tenets of the Declaration of Helsinki, and the San Carlos Clinical Hospital Ethics Committee approved the study with the internal code 18/422-E_BS.

As inclusion criteria, participants were required to have a complete ophthalmological evaluation and characterization of APOE alleles, have valid MEG recordings, and fulfill ophthalmological inclusion criteria (see below). Exclusion criteria were defined as scoring less than 25 on the Mini-Mental State Examination (MMSE) after adjusting their score to their age and educational level, following the procedure described in Blesa et al. [26] with a Spanish population. Participants who scored lower than 10% of their normative population in the Logical Memory subscale of the Wechsler Memory Scale III [27] in both units recall and themes recall scores, were not considered further. Other exclusion criteria were having a history of neurological or psychiatric disorders or any serious medical condition or showing brain abnormalities in magnetic resonance imaging (MRI). In total, 107 subjects were subsequently excluded according to these criteria, and the final sample consisted of 144 subjects.

The resulting sample was divided into two groups: participants who had at least one first grade relative with AD (FH +) and those who had no family history of the disease (FH −). Relatives’ AD diagnosis was verified after a medical history review by a multidisciplinary diagnostic consensus panel. Only those diagnoses made under international criteria or certified by autopsy reports of the relative were accepted. In addition, in order to avoid the influence of the variable “age,” which greatly affects visual function, especially contrast sensitivity [28,29,30], the sample was divided into a subsample ranging between 40 and 60 years of age, and a subsample over 60 years of age. Finally, each subsample was stratified according to allelic characterization for the APOE ɛ4 gene. The details of each group are shown in Fig. 1.

Fig. 1figure 1

Flow diagram of study participants. The participants without family history of AD (FH −) and non-carriers of ApoE ɛ4 (ApoE ɛ4 −) in green and the participants with family history of AD (FH +) and carriers of ApoE ɛ4 (ApoE ɛ4 +) in red

Due to limitations in sample size and statistical power, as well as asymmetries between the size of the groups and the demographics of some of them, the more natural analysis for the subgroups, a 2 × 2 × 2 design, taking into account age, family history, and APOE, was not feasible. Instead, visual functioning analyses were performed for the bigger groups (FH − vs. FH +) and the more restricted subgroups (FH − 40-60ɛ4− vs. FH + 40-60ɛ4+), given the big influence of age over the variables. Additional comparisons between older groups were performed to assess if the patterns were distinct from the ones found in the younger group. Given the lack of differences among the older groups, further correlation analyses were not performed.

The analyses of retinal structure were performed only for the more restrictive subgroups because the analysis of the larger groups (FH − and FH +) has been published previously [13]. Furthermore, as changes in visual function appear between the subgroups (FH − 40-60ɛ4− vs. FH + 40-60ɛ4+), it would be interesting to complement the study by OCT analysis in these subgroups.

When testing for possible demographic confounding variables, the only significant difference was found in the age of FH − and FH + . This difference disappears in the more restricted groups with ages between 40 and 60, which constitutes an additional reason to perform the analyses in this subgroup. The detailed data is shown in Table 1.

Table 1 Participant demographicsGenotyping

The APOE genotyping was carried out at the San Carlos Clinical Hospital in Madrid. DNA was extracted from whole blood in ethylenediamine tetra-acetic acid (EDTA), using standard DNA isolation methods (DNAzol®; Molecular Research Center, Inc., Cincinnati, OH, USA). APOE haplotype was determined by analyzing single-nucleotide polymorphisms (SNPs) rs7412 and rs429358 genotypes with TaqMan assays (C____904973_10 and C___3084793_20, respectively), using an Applied Biosystems 7500 Fast Real-Time PCR machine (Applied Biosystems, Foster City, CA).

APOE ε3/ε4 and APOE ε4/ε4 subjects were considered APOE ε4 + , while APOE ε2/ε3 and APOE ε3/ε3 subjects were considered APOE ε4 − .

Ophthalmological analysis

As previously described in López-Cuenca et al. [13], all participants completed a telephone screening interview to determine the status of their visual health. Those who were confirmed to be free of any pathology were examined at the Ramon Castroviejo Institute of Ophthalmic Research clinic (Madrid, Spain). A complete ophthalmologic examination, including visual acuity, refraction, applanation tonometry (Perkins MKII tonometer), CSV-1000E test, and OCT examination, was performed, and only participants who had no ocular disease, a best-corrected visual acuity of 0.5 dec, a spherocylindrical refractive error of less than ± 5, and an intraocular pressure of less than 20 mmHg were included.

Visual acuity

As previously described in Salobrar-García et al. [31], a standard clinical Snellen eye chart (decimal scale) was employed to determine the monocular best-corrected visual acuity. Visual acuity was measured with the subject’s subjective refraction. Patients started reading each row from the top towards the bottom of the chart and the test ended when the subjects were not able to recognize at least five out of eight letters (an approximation of 56.25%, the steepest point of the psychometric acuity function).

Contrast sensitivity function

The contrast sensitivity test was performed to measure the contrast sensitivity function under the same conditions for all participants, as previously described in Salobrar-García et al. [31]. A detailed description of the procedure can be found in the Supplementary Materials.

Optical coherence tomography

Macular thickness of each layer and peripapillary retinal nerve fiber layer (pRNFL) were measured using Spectralis OCT (Heidelberg Engineering, Heidelberg, Germany) as previously described in López-Cuenca et al. [13]. A detailed description of the procedure can be found in the Supplementary Materials.

The colorimetric representation of the changes in the macular and peripapillary thickness between the study groups was done with the Excel software and the color scale function. Areas where no difference can be found are colored in white, sectors that presented thinning among the FH + 40-60ɛ4+ group and FH − 40-60ɛ4− are colored in blue tones, and those that showed thickening are colored in red tones. The color tone is provided directly by the software based on the thickness variation.

Magnetic resonance imaging acquisition

For each subject, a T1-weighted anatomical brain MRI scan was acquired at the San Carlos Clinical Hospital in Madrid, with a General Electric 1.5 T magnetic resonance scanner, using a high-resolution antenna and a homogenization Phased array Uniformity Enhancement filter (Fast Spoiled Gradient Echo sequence, TR/TE/TI = 11.2/4.2/450 ms; flip angle of 12°; 1 mm slice thickness, 256 × 256 matrix, and FOV of 25 cm).

MagnetoencephalographyData acquisition

The electrophysiological activity of each participant was recorded at the Centro de Tecnología Biomédica during the performance of a cognitive task, described at length in Serrano et al. [32]. In brief, the task consisted of a delayed match-to-sample paradigm with faces as stimuli. For the present article, only the visual-related activity generated after the presentation of the faces will be addressed. The task comprised 128 trials, each containing two face presentations, resulting in a total of 256 face-locked events. All faces were neutral, Caucasian, adult male and female faces on a gray noise background and were kept on screen for 1 s in each presentation.

A detailed description of the task, MEG specifications, preprocessing performed, and source reconstruction methods can be found in the Supplementary Materials.

M100 latency

The visual evoked field (VEF) was generated using the face visualization task aforementioned. Each participant had, at least, 100 valid phase-locked events (211.48 ± 30.53; mean ± standard deviation). There were no statistically significant differences in the number of epochs between groups (t = 0.702; p = 0.487).

For the present study, only the M100 component of the VEF was addressed. The individual latency values were defined as the point of maximal activation of the calcarine cortex, as defined by the automated anatomical labeling (AAL) atlas [33]. Figure 2A, B shows the grand-average of the activation in these areas for a subsample of 133 individuals, showing the average M100 power.

Fig. 2figure 2

M100 latency. A Grand-averaged power at the calcarine fissure. B Grand-averaged power in the brain at 100 ms

Time–frequency analysis

Evoked field amplitude in MEG is highly dependent on the participant’s position in the scan, resulting in an unreliable metric. Therefore, we analyze the level of brain activity evoked by the face presentation using a time–frequency (TF) analysis.

TF representation was calculated for a 1000-ms time window, from 500 ms before to 500 ms after the face presentation. Epochs were analyzed in the time–frequency domain using a 5-cycle Gaussian Morlet wavelet with 1 Hz steps from 2 to 30 Hz. In order to avoid edge effects, all epochs had 2 s of real data at each side as padding. Resulting data were corrected by the average basal activity before the presentation of the stimulus, resulting in a relative change representation.

The subsequent analyses were restricted to the range in which the visual response takes place. To determine this range, a grand-average TF response for 133 participants (see Fig. 3A–C) was calculated. According to this response, the analyses were performed in the range of frequencies between 4 and 10 Hz, and the range of latencies between 0 and 250 ms.

Fig. 3figure 3

Visual response related time–frequency analysis. A Grand-averaged TF representation in the sensor space. B Grand-averaged TF representation in the sensors inside the A rectangles. C Grand-averaged TF representation in the source space (calcarine cortex). Rectangles in A indicate the sensors in which the activity is best perceived. Rectangles in B and C indicate the TF range in which the activity is more prevalent, which is later used for the analysis

Statistical analysis

Chi-squared test was used to perform comparisons between groups in qualitative variables.

Fig. 4figure 4

Cluster-based permutation test for the relationships of TF activity with visual function. Blue colors indicate negative correlations, while yellow colors indicate positive correlations. The time–frequency-sensor triplets pertaining to the significant cluster are shown in solid colors. A Correlation between visual acuity and TF in the FH + group (sensors); B correlation between visual acuity and TF in the FH + 40-60ɛ4+ group (sensors); C correlation between visual acuity and TF in the FH + group (calcarine); D correlation between visual acuity and TF in the FH + 40-60ɛ4+ group (calcarine)

Wilcoxon rank sum test with continuity correction was used to perform comparisons between groups in continuous variables, and Spearman correlation coefficients were obtained to assess the relationship between ophthalmological and electrophysiological variables. These methods were chosen as they make less assumptions regarding the data distribution. Additionally, Spearman correlation is capable of assessing non-linear monotonic relationships.

The correlation differences between the groups were calculated transforming the correlations to Fisher Z scores.

To measure effect sizes, Cramer’s V and its 95% confidence interval were calculated for the comparisons between groups in qualitative variables. For the comparisons between groups in continuous variables, effect size calculations from the Mann–Whitney U test and its 95% confidence interval, which has an interpretation analogous to a correlation coefficient, were estimated [34]. For correlations, the Spearman correlation coefficient acts as the size effect statistic.

To evaluate the TF power correlation with visual acuity, contrast sensitivity, and macular thickness, two approaches were taken. First, the analyses were performed in the sensor space, using cluster-based permutation test (CBPT) for multiple comparison corrections, with a Monte Carlo procedure, implementing 10,000 randomizations and a significance threshold of 0.05, taking sensors, time, and frequency as dimensions. Second, the analyses were also performed in the source space, averaging the TF sources belonging to the calcarine cortex as defined by the AAL atlas, and using CBPT with the same parameters, taking time and frequency as dimensions. As previously noted, both types of analysis were restricted to frequencies between 4 and 10 Hz and latencies between 0 and 250 ms.

We used false discovery rate (FDR) to correct for multiple comparisons. Within each research question, we applied FDR three times, once to correct the differences between groups, and once to correct the relationship between the ophthalmological measure and the M100 latency and the TF power for each group.

The preprocessing of MEG data and the CBPT analyses were performed in MATLAB R2019b (The Mathworks, Inc., Natick, MA), using the Fieldtrip package blindly to the group each subject belonged to. All other analyses were performed using R 3.6.2.

留言 (0)

沒有登入
gif