Skip to main content
Download PDF

Data-driven CSF biomarker profiling: imaging and clinical outcomes in a cohort at risk of Alzheimer’s disease

Alzheimer's Research & Therapy volume 16, Article number: 274 (2024) Cite this article

Abstract

Background

Cerebrospinal fluid (CSF) biomarkers of synaptic dysfunction, neuroinflammation, and glial response, complementing Alzheimer’s disease (AD) core biomarkers, have improved the pathophysiological characterization of the disease. Here, we tested the hypothesis that the co-expression of multiple CSF biomarkers will help the identification of AD-like phenotypes when biomarker positivity thresholds are not met yet.

Methods

Two hundred and seventy cognitively unimpaired adults with family history (FH) of sporadic AD (mean age = 60.6 ± 4.85 years, 64.8% women) underwent lumbar puncture, magnetic resonance imaging (n = 266) and positron emission tomography imaging (n = 239) protocols, and clinical evaluations. CSF Aβ42, Aβ40, p-tau181, p-tau217, p-tau231, NfL, neurogranin, sTREM2, YKL40, GFAP, S100, α-Synuclein, SYT1, and SNAP25 were measured. Participants were clustered based on CSF biomarker co-expression with an agglomerative algorithm. The predictive value of the classification against brain and cognitive outcomes was evaluated.

Results

Three clusters (C) were identified. Higher Aβ burden and CSF p-tau was the hallmark of C1. The other two clusters showed lower Aβ burden but higher expression of glial (C2) or synaptic markers (C3). Participants in C1 showed an AD-like clinical phenotype, comprising participants with the overall highest percentage of two parent FH and APOE-ε4 carriers, in addition to comprising more females compared to C2. C3 displayed better vascular health compared to C1. C2 were older and comprised a lower percentage of females compared to C3. C1 showed an AD-like gray matter reduction in medial temporal (notably hippocampus) and frontal regions that were not observed in Aβ42/40 + compared with Aβ42/40 − . Furthermore, Aβ42/40 − participants in C1 showed GM reduction in inferior temporal areas compared with Aβ42/40 + participants overall. C1 membership also predicted cognitive decline in executive function, but not memory, beyond Aβ + status, overall suggesting a better prognosis in Aβ42/40 + participants without C1 membership. Additionally, C1 displayed a higher rate of conversion to Aβ + (25%) over time.

Conclusions

Our results suggest that examining multiple CSF biomarkers reflecting diverse pathological pathways may complement and/or outperform AD core biomarkers and thresholding approaches to identify individuals showing a clinical and cognitive AD-like phenotype, including higher conversion to Aβ + , GM reductions and cognitive decline. The clinical utility of this approach warrants further investigation and replication in other cohorts.

Introduction

Alzheimer’s disease (AD) is biologically defined as the deposition of β-amyloid (Aβ) into plaques and hyperphosphorylated tau (p-tau) as neurofibrillary tangles [1]. The use of cerebrospinal fluid (CSF) and neuroimaging biomarkers has enabled researchers to assess neuropathological and neuronal brain dysfunction in vivo [2]. The A/T/N (amyloid/tau/neurodegeneration) system groups different biomarkers by the pathologic process and classifies individuals based on positivity thresholds [1]. Here, we present a tentative methodology that does not depend on positivity thresholds.

AD-related pathologic changes are three-to-four times higher in cognitively unimpaired (CU) middle-aged individuals with a genetic predisposition to AD [3], with early Aβ deposition detected based on proximity to parental age at symptom onset [4, 5]. Although Aβ positivity is not very prevalent in this age group, these individuals may exhibit some degree of variability within the negative range that may be clinically relevant [6]. Clinicopathologic studies suggest that subthreshold levels of pathologic changes are associated with subtle cognitive deficits in CU people [7].

New CSF biomarkers provide more detailed information about the neuropathological burden, beyond amyloidosis and tau pathology [2]. Recent work has pointed to specific reactive astrocyte biomarkers that are differentially associated with Aβ vs. tau pathology, and closely linked to neuroinflammatory proteins [8, 9]. Hence, CSF biomarkers indicating different pathological processes, such as synaptic dysfunction, neuroinflammatory response, neuronal injury or vascular dysregulation, may improve the A/T/N scheme. These pathological processes are altered early in preclinical AD [2, 10], which may surround and/or exacerbate the central amyloid-dependent pathways. Indeed, the revised criteria of the 2018 A/T/N research framework now includes biomarkers of inflammatory/immune processes in AD pathogenesis [11].

To address the heterogeneity in progression of AD pathology, a panel of CSF immunoassays (i.e., NeuroToolKit (NTK)) was developed to accelerate biomarker development in AD through the generation of robust data across cohorts [12], which has also accounted for variability in cognitive performance [10]. Age-related neurodegeneration shows heterogeneity, as evidenced by conditions like suspected non-AD pathophysiology and primary age-related tauopathy that can occur without Aβ pathology [13]. Evaluating co-expression of multiple CSF biomarkers may better capture AD heterogeneity.

The present work categorizes individuals based on multiple CSF biomarkers, including AD pathologies, neurodegeneration, synaptic dysfunction, neuroinflammation, and vascular dysfunction. Our hypothesis is that finer biomarker data granularity combined with clinical information is useful for identifying individuals with the earliest AD-related changes when positivity thresholds are not yet met [1]. Hence, the objectives include: (i) clustering participants based on similarity patterns of expression in multiple CSF biomarkers, (ii) characterizing the participants expressing these patterns using (a) demographics, clinical, and family history (FH) variables, (b) neuroimaging and longitudinal cognitive trajectories, and finally (iii) comparing the identification of AD-like profiles using this methodology vs. positivity thresholds. To this aim, we focus on a cohort of middle-aged CU adults with FH of sporadic AD. We expect to identify a group with an AD-like phenotype that displays AD hallmark features and potentially higher presence of FH and APOE-ε4 status.

Methods

Participants

The ALFA (ALzheimer’s and FAmilies) cohort was established to characterize preclinical AD and consists of 2743 CU individuals aged between 45–74 years at the time of the recruitment [14]. Most of the recruited participants were offspring of AD patients (86% had at least one parent that has suffered AD). Baseline assessments were performed between 2013 and 2014, and included sociodemographic, genetic, cognitive, and clinical data collection.

Here, we focus on participants from the nested ALFA + study, with FH of clinically diagnosed sporadic AD. These participants underwent advanced protocols of magnetic resonance imaging (MRI), Aβ positron emission tomography (PET) imaging with [18F] flutemetamol, and lumbar puncture between 2016 and 2019 in the context of ALFA + Visit 1. The present sample included 281 ALFA + participants, of which 270 (mean age [SD] = 60.6[± 4.85], range = 49.25 − 73.43 years; mean years of education [SD] = 13.46[± 3.55], range = 6 − 18 years; sex = 95 males, 175 females) were included in the final analysis.

FH of sporadic AD was considered when a self-reported medical history was presented in tandem with either (i) a clinical diagnosis; or (ii) a retrospective diagnosis consistent with AD based on reported symptoms [4].

CSF biomarkers

CSF Aβ42, Aβ40, neurofilament light (NfL), neurogranin, the soluble fragments of the triggering receptor expressed on myeloid cells 2 (sTREM2), Chitinase 3-like 1 (YKL40), glial fibrillary acidic protein (GFAP), S100 calcium-binding protein B (S100b), and α-synuclein (⍺-SYN) were measured using the NTK [12, 15]. CSF p-tau181 and p-tau217were measured with in-house Simoa-based assays [16], targeting N-terminal fragments; CSF p-tau231 was quantified using a research colorimetric Enzyme-linked immunosorbent assays (ELISA), targeting mid-region tau [16]. Synaptotagmin-1 (SYT1) and synaptosomal-associated protein-25 (SNAP-25) were measured using immunoprecipitation mass spectrometry [17]. All measurements were performed at the Clinical Neurochemistry Laboratory, Sahlgrenska University Hospital, Mölndal, Sweden. These biomarkers were chosen as the indicators of different neuronal alterations: CSF Aβ42/40to measure amyloidosis, CSF p-tau biomarkers to measure tau pathology, CSF SNAP-25, SYT1, neurogranin and ⍺-SYN as markers of synaptic dysfunction, CSF sTREM2, YKL40, GFAP and S100B as markers of glial reactivity, and NfL as a biomarker of neuronal injury [2, 15].

We considered CSF Aβ42/40 positivity when Aβ42/40 < 0.071 [15]. This threshold was derived from our cohort to detect early pathophysiological changes [18]. We used p-tau181 Elecsys® (Roche Diagnostics International Ltd., Rotkreuz, Switzerland), targeting mid-region tau, as a previously validated immunoassay to determine CSF p-tau181positivity (T + > 24 pg/mL) [15, 19].

APOE Genotyping

Total DNA was extracted from the blood cellular fraction by proteinase K digestion followed by alcohol precipitation. The variants of APOE allele were obtained from the allelic combination of the rs429358 and rs7412 polymorphism. APOE status was determined based on the APOE-ε4 genotype. Participants were classified as APOE-ε4 carriers (i.e., carriers of one ε4 allele) or non-carriers.

Imaging data acquisition and pre-processing

Flutemetamol PET imaging

PET scans were acquired using a Siemens Biograph Mct (Munich, Germany) (see details [14]). Images were preprocessed with SPM12 software (http://www.fil.ion.ucl.ac.uk/spm; Wellcome Trust Centre for Neuroimaging, UK) following the Centiloid (CL) pipeline [20, 21].

We considered a cut-off value of 12 for positivity in Aβ PET to classify participants as Aβ + (≥ 12 CL) or Aβ − (< 12 CL) [21], which has previously been shown to indicate the transition from absence of pathology to subtle pathology, with an optimal correspondence to CSF Aβ42 positivity [18, 21]. We also reported Aβ positivity when CL > 30, a cut-off that reflects the presence of established Aβ pathology [21].

MRI

Participants underwent an anatomical 3D T1-weighted Fast Field Echo sequence performed on a Philips 3 T Ingenia CX scanner (Best, Netherlands) with the following parameters: voxel size = 0.75 mm3 isotropic, field of view = 240 × 240 × 180 mm3, flip angle = 8°, repetition time = 9.9 ms, echo time = 4.6. ms, and inversion time = 900 ms in sagittal acquisition.

Voxel-based morphometry (VBM) preprocessing analysis

The VBM preprocessing was conducted using SPM12. Segmented gray matter (GM) masks, extracted from tissue probability maps and normalized to Montreal Neurological Institute (MNI) space using Dartel [22], were used to generate a group-level mask with GM probability > 0.3 (see Additional File 1 / Voxel-based morphometry preprocessing analysis).

Clinical variables

Family history variables

We created a dichotomous variable to classify the participants based on FH of AD dementia load with one or two parents affected.

Vascular health

History of comorbid conditions was collected via structured interviews considering hypertension, arrhythmia, dyslipidemia, thyroid, obesity, and diabetes mellitus. Systemic vascular health was calculated from the number of reported vascular risk factors (i.e., hypertension, diabetes mellitus, dyslipidemia, heart failure, ischemic heart disease, and atrial fibrillation). We then created a factor with 2 levels: 0 = no reported comorbid conditions and 1 = one or more comorbid conditions.

Cognitive assessments

Participants completed several cognitive tasks during ALFA + study Visits 1 and 2. Here, we considered the following cognitive assessments: the Trail Making Test [23] Part A [TMT-A] and the TMT Part B [TMT-B], where the difference TMT-B − TMT-A served as the measure of executive function after accounting for the motor and visual search component from Part A. We also used the Total Delayed Free Recall component of Memory Binding Test [MBT-TDFR] [24] as a measure of memory.

Statistical analysis

For pairwise t-tests and linear mixed effects models of demographic and clinical distinctions between clusters, reported results are uncorrected at a two-tailed significance threshold of p < 0.05. For assessment of Aβ conversion across time, results are uncorrected at a one-tailed significance threshold of p < 0.05.

Clustering analysis

Participants were clustered based on their CSF biomarker expression using an unsupervised bottom-up agglomerative hierarchical clustering analysis (HCA), with ‘complete’ linkage and Euclidean distance. In brief, raw biomarker distributions were inspected, missing values were imputed, and the participant (observations) by biomarkers (features) matrix was z-score normalized across features and observations, respectively (see Additional File 1 / Clustering analysis).

Cluster selection and sensitivity analysis

To assure the validity of the clusters obtained, we combined two internal validity indices to evaluate the quality of clustering results. We chose the optimal number of clusters as the cluster count displaying the smallest difference between these two metrics. For details, see Additional File 1—Cluster sensitivity analysis.

Between-cluster comparisons

Clinical and demographic variables

We performed pairwise between-cluster comparisons in demographic and clinical outcomes of interests to explore basic cluster-level differences using t-test or chi-squared test, when appropriate.

VBM analyses

We performed whole brain exploratory analysis to compare GM volumes between defined clusters. We used ANOVA to identify clusters with lower GM volumes and performed pairwise comparisons between clusters while adjusting for any other identified clusters. All analyses were total intracranial volume (TIV)-corrected, without adjusting for age or sex to focus on the clusters' demographic characteristics.

Given that the study sample is not expected to show advanced atrophy/neurodegeneration, we used a threshold of p uncorrected (puncorr) 0.005.

Longitudinal analyses

Cognitive health trajectories

Longitudinal assessments were available for 248 participants (mean follow-up time [SD]: 3.47[± 0.58] years). We ran linear mixed effects models to assess cluster-level differences on change in cognitive performance. We adjusted for age, sex, years of education, APOE-ε4 status, and FH, and included ‘participant’ as a random effect, allowing for individual variability at the level of intercept. Additionally, we included the interaction term Time*Cluster to investigate whether a specific cluster displayed disproportionate declines in any outcome over time.

Aβ-PET positivity conversion

We investigated the utility of clustering to identify individuals who convert to Aβ + over time, in accordance with the A/T/N framework considering a low threshold for Aβ + to capture the earliest changes. Of all participants with longitudinal Aβ PET scans available, we charted how many participants in each cluster converted to Aβ + in PET between the two visits. Fisher’s exact test was used to assess the significance of conversion between clusters among Aβ − .

Results

Clustering

Three clusters were identified using HCA.

Cluster selection

The optimal number of clusters was determined based on the smallest difference between the Dunn and Davies-Bouldin internal validity indices (Additional File 2, left panel). Additionally, permutation testing revealed that the mean within-cluster variances between the three identified clusters was significantly lower than those given by chance (p < 0.001; Additional File 2, right panel).

Cluster description

Each cluster (C) describes a comprehensive combination of the differential weights of each biomarker. The three clusters (C1, C2, and C3) included a total of 104, 55, and 111 participants, respectively (see Fig. 1A for dendrogram). Higher Aβ burden (lower CSF Aβ42/40 ratio) was the hallmark of C1 only, while the other two clusters showed lower Aβ burden but higher expression of biomarkers related to neuroinflammatory (C2) and synaptic (C3) factors (see Fig. 1B for radial plot of cluster identification).

Fig. 1
figure 1

Dendrogram of clustering and radial plot by biomarker group. Blue, yellow, and orange colors represent the AD-like co-expression (C1), Neuroinflammation (C2), and Synaptic (C3) clusters, Upper panel (A): Dendrogram of clustering result from HCA. The number of participants within each cluster is indicated in parentheses. The height of the dendrogram associated with each branch represents the dissimilarity, as Euclidean distance, between the observations being joined at that particular point. Middle panel (B): Radial plot of the mean z-score normalizations. Plotted range of z-score values can be found in a small schematic in the legend to the right. Literature-supported biomarker class is indicated along the circumference of the plot [19]. Lower panel (C): Heat plot of the pairwise cluster comparisons and cluster profiles. All three cluster comparisons are listed vertically and variables are listed horizontally. Plotted colors represent signed p-values to preserve the direction of the effect. “Sex” is coded as 0 for males and 1 for females. “FH” denotes family history (one or two parents with history of AD dementia), “Vasc-H” denotes vascular health, with higher values associated with poorer health/higher vascular burden. The symbol (▴) denotes the reference group. *p uncorr < 0.05, ** p uncorr < 0.01, *** p uncorr < 0.001. Descriptive boxes below the heat plot list cluster profiles from pairwise comparisons

Full size image

C1 displayed the lowest CSF Aβ42/40 ratio and highest expression of CSF p-tau levels for the three protein phosphorylation sites (p-tau181, p-tau217 and p-tau231). C2 displayed the highest expression of neuroinflammation markers S100b, YKL40, sTREM2, and GFAP. C3 displayed the highest expression of synaptic markers, ⍺-synuclein, neurogranin, SNAP-25, and SYT1.

Between-cluster comparisons

Demographic and clinical variables

C1 showed a higher proportion of APOE-ε4 carriers than both C2, X2(1, N = 159) = 13.28, p < 0.001, and C3, X2(1, N = 215) = 14.14, p < 0.001; a greater number of participants with two parents diagnosed with sporadic AD than both C2 (Fisher’s exact test, p = 0.028) and C3 (Fisher’s exact test, p = 0.008); and displayed higher CL values than both C2 (t(144) = 3.7, p < 0.001) and C3 (t(189) = − 4.47, p < 0.001) (Fig. 1C for heat plot).

Additionally, C1 showed a higher percentage of females when compared to C2, X2(1, N = 159) = 7.9, p = 0.004). C1 also displayed greater vascular health burden when compared to C3 (t(212) = − 2.43, p = 0.016). C1, therefore, showed an AD-like clinical phenotype.

C2 comprised slightly older individuals compared to C3 (t(164) = − 2.57, p = 0.011), whereas C3 comprised more females than C2 (X2(1, N = 166) = 15.91, p < 0.001) (Table 1).

Table 1 Descriptive table of demographics, CSF biomarkers, and clinical variables divided by cluster
Full size table

VBM analysis

For GM volume analyses, we adopted a data visualization approach to cluster threshold selection by merging results across all comparisons of interest (= 21) and plotting each contrast of interest at puncorr = 0.005, with no cluster threshold applied. The resulting cluster sizes (k) were plotted, and the threshold defined as the first point of break in continuity of the histogram. This method resulted in a threshold selection of k > 275 voxels (Additional File 3). We did not observe any GM volume differences for Family-Wise Error corrected p-values, as expected, due to the sample characteristics. The results are reported for puncorr at both the voxel and cluster level for completeness.

HCA cluster comparison

Results from the ANOVA showed lower GM volumes only in C1 in comparison to C2 and C3, which is compatible with an AD-like atrophy pattern. The differences were observed in the left amygdala, left hippocampus along with portions of left parahippocampal gyrus, right parahippocampal gyrus extending to the right hippocampus, right inferior and middle frontal gyrus, left anterior cingulate gyrus with an extension to the right side, and left superior and middle temporal pole (Fig. 2, Panel A). There were not significant regions showing less GM in C2 and C3 compared with C1 and C3, and C1 and C2, respectively.

Fig. 2
figure 2

Voxel-wise gray matter differences between the clusters (Panels A-C), Aβ42/40 + vs. Aβ42/40 − (Panel D), and T + vs. T − (Panel E). Voxel-wise results are shown with a cluster-size threshold of 275 voxels with p uncorrected < .005 on the left side and with the T-maps throughout the right side of the figure. CSF Aβ42/40 positivity was considered when Aβ42/40 < 0.071. CSF p-tau181 positivity was considered when p-tau181 > 24 pg/mL. C1 = AD-like co-expression, C2 = Neuroinflammation, and C3 = Synaptic cluster. R: Right, L: Left

Full size image

Posthoc pairwise analyses revealed that C1 had lower GM volumes than C2 (Fig. 2, Panel B), but not than C3 (Fig. 2, Panel C), in the left parahippocampal gyrus, left amygdala, left middle and superior temporal gyrus, right middle frontal gyrus, left anterior cingulate gyrus (with bilateral extension), right parahippocampal gyrus, left superior temporal gyrus and left middle cingulate gyrus.

Pairwise comparisons showed that C3 had lower GM volumes than C1 in the right caudate region (Additional File 4, Panel A). Further, participants in C3 had lower GM volumes than those in C2 in the left supramarginal gyrus with an extension to left angular and middle temporal gyrus (Additional File 4, Panel B).

Comparisons between HCA cluster division and A/T/N system

Our previous analyses showed an AD-like clinical phenotype and GM reductions in C1. Here, we compared the utility of identifying clusters based on our data-driven approach to that of a division based on the A/T/N system (i.e. A and T) [1]. We first ran VBM analyses to compare whole-brain GM volume differences between CSF Aβ42/40 + (n = 77) vs. Aβ42/40 − (n = 189) and CSF T + (n = 26) vs. T − (n = 240) participants in our sample irrespective of cluster assignment. Second, considering the clustering division, we performed the following comparisons: (i) Aβ42/40 − participants in the C1 (n = 45) vs. Aβ42/40 + participants in all clusters (n = 77), (ii) T − participants in the C1 (n = 92) vs. T + participants in all clusters (n = 26), (iii) Aβ42/40 + participants in the C1 (n = 56) vs. Aβ42/40 + participants in the other two clusters (n = 21), (iv) T + participants in the C1 (n = 9) vs. T + participants in the other two clusters (n = 17). All comparisons were performed using t-tests, were controlled for the TIV, and reported with a threshold of puncorr 0.005.

In the whole sample, CSF Aβ42/40 + and Aβ42/40 − participants did not show any GM volume differences (Fig. 2, Panel D). When considering the cluster division, instead, Aβ42/40 − participants in C1 showed lower GM volumes in the left fusiform gyrus and left inferior temporal gyrus as compared to Aβ42/40 + participants in the whole sample (Fig. 3, Panel A). Further, Aβ42/40 + participants in the C1 had lower GM volumes than Aβ42/40 + participants in the C2 and C3 clusters in the left middle and anterior cingulate gyrus (extending to the right side), right calcarine sulcus (extending to the left side), bilateral middle occipital gyrus, left middle temporal gyrus, right superior frontal gyrus, right precuneus and cuneus (Fig. 3, Panel B).

Fig. 3
figure 3

Multi-planar views showing the voxel-wise results comparing the A/T/N positivity threshold approach and clustering approach. Voxel-wise results are shown with a cluster-size threshold of 275 voxels with p uncorrected < .005 on the left side and with the T-maps throughout the right side of the figure. CSF Aβ42/40 positivity was considered when Aβ42/40 < 0.071. CSF p-tau181 positivity was considered when p-tau181 > 24 pg/mL. C1 = AD-like co-expression, C2 = Neuroinflammation, and C3 = Synaptic cluster. R: Right, L: Left

Full size image

Regarding CSF p-tau, T + participants did not show lower GM than T − participants in any brain region. Considering the cluster division, we observed that T − participants in the C1 had lower GM volumes in the bilateral insular cortex, and right middle and superior temporal pole extending to the left side (Fig. 3, Panel C). Further, T + participants in C1 had lower GM than the T + in C2 and C3 in the left medial orbitofrontal cortex, right middle cingulate gyrus extending to the left side, right postcentral and precentral gyrus, bilateral middle temporal gyrus, left supramarginal gyrus, and right superior temporal and occipital gyrus (Fig. 3, Panel D).

Descriptive statistics for gray matter volume differences are reported in Additional File 5 for all performed comparisons. Results from the performed comparisons are displayed with statistics and MNI coordinates for peak regions in Additional File 6.

Finally, to test the complementarity of our approach compared to the positivity threshold approach, we repeated our whole brain exploratory VBM analyses adjusting by CSF Aβ42/40 status. The results remained largely unchanged as illustrated in the Additional File 7.

Cognitive assessments

Linear mixed effect model results showed that Cluster*Time was significant for the TMT (B-A) task. Age-related change over time on task-switching was moderated by cluster affiliation [F(2, 364.96) = 6.707, p = 0.001] such that C1 significantly increased in reaction time performance compared to C2 and C3, despite displaying initially quicker reaction time at baseline. Conversely, longitudinal change in performance on the MBT-TDFR—a task regarded as sensitive to predicting future progression to AD-dementia [25]— did not display a cluster moderation (Fig. 4). The marginal R2 of the model was 0.19 and the conditional R2 was 0.47. Additionally, C1 membership was still associated with greater decline even after adjusting for Aβ-positivity status (Table 2). Model robustness was further confirmed by model refitting after removal of standardized residual errors higher than 3, the results withstanding this procedure.

Fig. 4
figure 4

Plot and results for the TMT (B-A) task and comparison to MBT-TDFR. Spaghetti plot of the interaction Cluster*Time on performance for the Trail Making Test (TMT) B-A (i.e., adjusted for baseline processing speed; left plot) and MBT-TDFR (right plot) tasks. TMT (B-A) performance is represented in seconds; delayed recall of MBT-TDFR is represented as accuracy. Colored dots represent each participant’s performance divided by cluster affiliation, with colored line linkage denoting single participant performance over time. Colored ribbons represent the standard error around the mean. The interaction for TMT (B-A) is significant; the interaction for MBT-TDFR is not significant (see the legend for significance level of interaction for the TMT task). *puncorr < 0.05, ** puncorr < 0.01, *** puncorr < 0.001

Full size image
Table 2 Significant results for the linear mixed effects models
Full size table

Conversion to Aβ-PET positivity

Of those individuals with longitudinal Aβ-PET scans available (n = 139), we charted how many participants in each cluster converted to Aβ + in PET (Table 3) from Visit 1 to Visit 2 (mean follow-up time [SD]: 3.51[± 0.53] years). Considering only those individuals who were Aβ − at baseline, or “at risk” for converting, results from the Fisher’s exact test showed that C1 had significantly higher conversion to Aβ + when compared to both C2 (p = 0.021) and C3 (p = 0.027).

Table 3 Aβ + status and conversion rate at Visit 1 and Visit 2 per cluster
Full size table

Discussion

In a cohort of CU adults with FH of sporadic AD and low pathological burden, our results suggest that classifying individuals based on the co-expression of AD-related CSF biomarkers is useful to identify participants with an AD-like clinical, cognitive and brain structure phenotype, including Aβ − participants. Thus, this classification showed a potential to outperform and/or complement Aβ + thresholding approaches to identify participants with AD-like GM reductions, as well as conversion to Aβ + PET status, and cognitive decline over time. Moreover, these findings suggest a more favorable prognosis in Aβ + adults who do not exhibit the AD-like co-expression pattern.

Through rigorous quality assessment via various evaluation metrics, we identified three discrete clusters with features corresponding to a co-expression of AD-like (C1), Neuroinflammation (C2), and Synaptic markers (C3). Specifically, C1 conformed to the expected biological profile of AD, expressing reduced CSF Aβ42/40 ratio and increased CSF p-tau levels across all three phosphorylation sites studied. The Neuroinflammation and Synaptic groups expressed higher Aβ42/40 ratios along with elevated biomarker levels corresponding to neuroinflammatory responses and synaptic dysfunction, respectively. Thus, the observed patterns suggest a clear distinction within our risk-enriched cohort, with one pattern demonstrating co-expression involving Aβ while the other two do not. Further exploration of Aβ positivity thresholds unveiled a more nuanced perspective, as elaborated below.

The most prominent demographic, vascular, and cognitive differences were found for C1 (AD-like co-expression); this cluster contained significantly more individuals with two AD dementia-affected parents, APOE-ε4 status, and greater Aβ-PET burden than other clusters. It is well-documented that the APOE-ε4 allele is a strong genetic risk factor for AD development, accelerating symptomatic onset by around 10 years [26]. Additionally, research has shown that CU adults with a FH of sporadic AD start to display pathological changes during midlife [27], with FH linked to increased Aβ-load [4]. Crucially, our clustering method identified individuals with reduced GM in key AD-related regions and aided in predicting conversion to Aβ + status beyond binary classification according to the A/T scheme [28].

Prior data-driven approaches to cluster discriminability of AD pathology has either exclusively focused on phenotypes of GM atrophy [29] and/or biomarker inclusion has been limited to Aβ and tau assays [30]. Recent clustering approaches have attempted to extend classification beyond GM atrophy [29] and Aβ and tau assays [30] to other pathological dimensions [31] and utilize multidimensional data to differentiate between fast and slow mild cognitive impairment (MCI) decliners at risk of AD progression [32]. CSF biomarkers beyond Aβ and tau pathology may also facilitate nuanced classification of AD as there are several pathological pathways that are relevant to cognitive decline [33]. Indeed, 24.1% of all participants classified as Aβ42/40 − based on established CSF positivity thresholds, were included in the AD-like co-expression cluster. Variability in subthreshold Aβ deposition may still be clinically relevant [6] and associated with subtle cognitive deficits in CU individuals [7]. For example, neuroinflammatory processes driven by microglial reactivity [34, 35] and GFAP and YKL-40 levels [8] can influence the effect of Aβ and tau on neurodegeneration. Indeed, our clustering approach better captured GM volume reduction in the AD-like group beyond both “A” and “T” of the A/T/N system. This is important as GM loss predicts cognitive decline [36]. Potential differential observations that cut across CSF Aβ42/40-status dichotomization are particularly significant as CSF Aβ-proteinopathy is the earliest biomarker of the AD pathological cascade model [37].

Furthermore, 26.6% participants classified as Aβ42/40 + based on established CSF positivity thresholds were not included in C1 (AD-like co-expression cluster), while they present with AD pathological change based on the current biological definition [1] and with AD based on the revised criteria of the A/T/N research framework [11]. Our results suggest that, in the short term, this group may exhibit a more favorable prognosis within the AD pathway. Indeed, for Aβ42/40 − individuals in C1 compared to Aβ42/40+ individuals in other clusters, we observed greater GM atrophy in the left inferior temporal gyrus, which in prior work has predicted decline from cognitively normal to AD status [38]. Furthermore, it is an early site for tau deposition and accelerated cortical thinning [39]. Tau and cortical amyloid deposition in this area has also been linked to decline in daily activities functioning among individuals with MCI and AD [40]. Additionally, reduced GM was observed in the left fusiform, a core AD signature region [41].

For Aβ42/40 + individuals in C1 compared to Aβ42/40 + individuals in other clusters, we observed greater GM reductions in the right calcarine. The calcarine cortex has shown to display prominent age-related atrophy [42] but has little additional atrophy in AD [43]. Lastly, Aβ42/40+ C1 individuals displayed lower GM volume in the left anterior-mid cingulate cortex with an extension to the contralateral side. Importantly, GM loss in the anterior cingulate cortex has been suggested to predict progression to AD [44].

T − individuals in C1 (AD-like co-expression cluster) displayed lower GM volume in the superior and inferior temporal pole compared to T + individuals in other clusters. Interestingly, previous research has indicated this region as the core of brain damage in language variant of early onset AD [45]. The fact that our biomarker clustering approach identified individuals with GM reductions in this region merits further consideration.

The left medial orbitofrontal cortex and right middle cingulate gyrus, extending to the left, displayed notable GM reductions in T + C1 individuals compared to T + individuals in other clusters. Recent work has shown the medial prefrontal cortex to be highly interconnected with subcortical regions, like the hippocampus, and displays altered functional connectivity in AD, even potentially preceding aberrant structural changes [46]. Future research may consider early monitoring of this region along the AD continuum.

Our findings revealed reduced GM in the bilateral hippocampus in participants within the C1 (AD-like co-expression cluster), a hallmark of AD [47]. This cluster also showed increased expression of all three CSF p-tau biomarkers, potentially indicating early-stage AD even before noticeable Aβ pathology [16]. Longitudinal cognition changes indicated that the AD-like co-expression cluster had a significant increase in completion time for TMT B–A, an executive control indicator [48], despite better initial performance. Moreover, C1 affiliation predicted longitudinal cognitive decline above and beyond Aβ42/40status, consistent with prior findings linking TMT B–A performance to Aβ level [49]. Differences were not observed for MBT-TDFR despite prior linkage to AD-related biomarkers [50]. The fact that the AD-like co-expression cluster expressed reduced GM hippocampal volume and cognitive decline in a non-memory domain is particularly interesting given recent work that similarly utilized a biomarker clustering approach that predicted memory decline, but not hippocampal atrophy, in a cluster of individuals with notably higher expression of tauopathy [51]. Overall, our results support early monitoring of executive function in asymptomatic middle-aged adults. Furthermore, participants in C1 (AD-like co-expression cluster) had higher conversion rates to Aβ-PET positivity than other clusters, which is expected given that AD-related pathologic changes are more probable in individuals with genetic predisposition to AD [3]. Notably, four of the eleven converters in C1 did not have APOE-ε4 or a family history of AD, suggesting genetic factors alone do not account for the increased conversion rates. When we compared the C2 (Neuroinflammation cluster) to C3 (Synaptic cluster), C3 showed reduced GM volume in the left supramarginal gyrus and left angular and middle temporal gyrus. Recent work from the ALFA cohort has shown that increases in CSF Neurogranin were linked to reduced cortical thickness in these AD signature regions; additionally, synaptic biomarkers were also associated with higher levels of CSF p-tau and NfL [17]. The fact that C3 i.) displayed lower GM volume outside of AD-signature regions, ii.) displayed striking homogeneity across synaptic biomarker class, iii.) was distinguishable from AD-like pathology in terms of relatively lower p-tau levels, and iv.) presented with a significantly higher number of females, warrants further investigation.

Limitations

The current study is not without limitations. First, the sample size is relatively small, which potentially limits the stability and extendibility of findings at a broader scale. Second, the choice of both clustering method and distance parameter could influence cluster identification. However, we hoped to enhance the reliability of our cluster division by choosing a common distance metric and utilizing sensitivity indices and a variance permutation test for increased robustness. Finally, with the inclusion of longitudinal data from CSF biomarkers, brain, and cognitive measures, the temporal ordering of pathological events can be better discerned.

Conclusions

Using a data-driven HCA approach, we identified subsets of CU older adults with an AD-like clinical, cognitive, and brain structure phenotype. A cluster emerged with AD-like pathology characteristics that went beyond Aβ- or tau-thresholding of the A/T/N system. Our study demonstrates the value of unsupervised data-driven methods in uncovering important patterns with clinical implications.

Data availability

Data that support the findings of this study are available on reasonable request from the ALFA Study Investigators.

Abbreviations

Aβ:

Amyloid-beta

AD:

Alzheimer’s disease

ALFA:

ALzheimer’s and FAmilies

A/T/N:

Amyloid/Tau/Neurodegeneration

CL:

Centiloid

CU:

Cognitively unimpaired

FH:

Family history

GFAP:

Glial fibrillary acidic protein

GM:

Gray matter

HCA:

Hierarchical clustering analysis

MBT:

Memory Binding Test

MCI:

Mild cognitive impairment

MRI:

Magnetic resonance imaging

NfL:

Neurofilament light

PET:

Positron emission tomography

p-tau:

Phosphorylated tau

sTREM2:

Soluble fragments of the triggering receptor expressed on myeloid cells 2

S100b:

S100 calcium-binding protein B

TMT:

Trail Making Test

VBM:

Voxel-based morphometry

YKL40:

Chitinase 3-like 1

⍺-SYN:

α-Synuclein

References

  1. Jack CR, Bennett DA, Blennow K, Carrillo MC, Dunn B, Haeberlein SB, et al. NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimers Dement. 2018;14(4):535–62.

    Article  PubMed  Google Scholar 

  2. Milà-Alomà M, Suárez-Calvet M, Molinuevo JL. Latest advances in cerebrospinal fluid and blood biomarkers of Alzheimer’s disease. Ther Adv Neurol Disord. 2019;18(12):1756286419888819.

    Article  Google Scholar 

  3. Jansen WJ, Ossenkoppele R, Knol DL, Tijms BM, Scheltens P, Verhey FRJ, et al. Prevalence of Cerebral Amyloid Pathology in Persons Without Dementia. JAMA. 2015;313(19):1924–38.

    Article  PubMed  PubMed Central  Google Scholar 

  4. Arenaza-Urquijo EM, Salvadó G, Operto G, Minguillón C, Sánchez-Benavides G, Crous-Bou M, et al. Association of years to parent’s sporadic onset and risk factors with neural integrity and Alzheimer biomarkers. Neurology. 2020;95(15):e2065–74.

    Article  PubMed  PubMed Central  Google Scholar 

  5. Villeneuve S, Vogel JW, Gonneaud J, Pichet Binette A, Rosa-Neto P, Gauthier S, et al. Proximity to Parental Symptom Onset and Amyloid-β Burden in Sporadic Alzheimer Disease. JAMA Neurol. 2018;75(5):608–19.

    Article  PubMed  PubMed Central  Google Scholar 

  6. Bischof GN, Jacobs HIL. Subthreshold amyloid and its biological and clinical meaning: Long way ahead. Neurology. 2019;93(2):72–9.

    Article  PubMed  Google Scholar 

  7. Monsell SE, Mock C, Hassenstab J, Roe CM, Cairns NJ, Morris JC, et al. Neuropsychological changes in asymptomatic persons with Alzheimer disease neuropathology. Neurology. 2014;83(5):434–40.

    Article  PubMed  PubMed Central  Google Scholar 

  8. Ferrari-Souza JP, Ferreira PCL, Bellaver B, Tissot C, Wang YT, Leffa DT, et al. Astrocyte biomarker signatures of amyloid-β and tau pathologies in Alzheimer’s disease. Mol Psychiatry. 2022;27(11):4781–9.

    Article  PubMed  PubMed Central  Google Scholar 

  9. Pelkmans W, Shekari M, Brugulat-Serrat A, Sánchez-Benavides G, Minguillón C, Fauria K, et al. Astrocyte biomarkers GFAP and YKL-40 mediate early Alzheimer’s disease progression. Alzheimers Dement. 2023.

  10. Van Hulle C, Jonaitis EM, Betthauser TJ, Batrla R, Wild N, Kollmorgen G, et al. An examination of a novel multipanel of CSF biomarkers in the Alzheimer’s disease clinical and pathological continuum. Alzheimers Dement. 2021;17(3):431–45.

    Article  PubMed  Google Scholar 

  11. Revised Criteria for Diagnosis and Staging of Alzheimer’s | AAIC [Internet]. Revised Criteria for Diagnosis and Staging of Alzheimer’s | AAIC. [cited 2024 Mar 20]. Available from: https://aaic.alz.org/diagnostic-criteria.asp.

  12. Johnson SC, Suárez-Calvet M, Suridjan I, Minguillón C, Gispert JD, Jonaitis E, et al. Identifying clinically useful biomarkers in neurodegenerative disease through a collaborative approach: the NeuroToolKit. Alzheimers Res Ther. 2023;15(1):25.

    Article  PubMed  PubMed Central  Google Scholar 

  13. Jack CR. PART and SNAP. Acta Neuropathol. 2014;128(6):773–6.

    Article  PubMed  PubMed Central  Google Scholar 

  14. Molinuevo JL, Gramunt N, Gispert JD, Fauria K, Esteller M, Minguillon C, et al. The ALFA project: A research platform to identify early pathophysiological features of Alzheimer’s disease. Alzheimers Dement (N Y). 2016;2(2):82–92.

    Article  PubMed  Google Scholar 

  15. Milà-Alomà M, Salvadó G, Gispert JD, Vilor-Tejedor N, Grau-Rivera O, Sala-Vila A, et al. Amyloid beta, tau, synaptic, neurodegeneration, and glial biomarkers in the preclinical stage of the Alzheimer’s continuum. Alzheimers Dement. 2020;16(10):1358–71.

    Article  PubMed  Google Scholar 

  16. Suárez-Calvet M, Karikari TK, Ashton NJ, Lantero Rodríguez J, Milà-Alomà M, Gispert JD, et al. Novel tau biomarkers phosphorylated at T181, T217 or T231 rise in the initial stages of the preclinical Alzheimer’s continuum when only subtle changes in Aβ pathology are detected. EMBO Mol Med. 2020;12(12):e12921.

    Article  PubMed  PubMed Central  Google Scholar 

  17. Milà-Alomà M, Brinkmalm A, Ashton NJ, Kvartsberg H, Shekari M, Operto G, et al. CSF Synaptic Biomarkers in the Preclinical Stage of Alzheimer Disease and Their Association With MRI and PET: A Cross-sectional Study. Neurology. 2021;97(21):e2065–78.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Milà-Alomà M, Salvadó G, Shekari M, Grau-Rivera O, Sala-Vila A, Sánchez-Benavides G, et al. Comparative Analysis of Different Definitions of Amyloid-β Positivity to Detect Early Downstream Pathophysiological Alterations in Preclinical Alzheimer. J Prev Alzheimers Dis. 2021;8(1):68–77.

    PubMed  Google Scholar 

  19. Salvadó G, Shekari M, Falcon C, Operto G, Milà-Alomà M, Sánchez-Benavides G, et al. Brain alterations in the early Alzheimer’s continuum with amyloid-β, tau, glial and neurodegeneration CSF markers. Brain Communications. 2022;4(3):fcac134.

  20. Klunk WE, Koeppe RA, Price JC, Benzinger TL, Devous MD Sr, Jagust WJ, et al. The Centiloid Project: Standardizing quantitative amyloid plaque estimation by PET. Alzheimer’s & Dementia. 2015;11(1):1-15.e4.

    Article  Google Scholar 

  21. Salvadó G, Molinuevo JL, Brugulat-Serrat A, Falcon C, Grau-Rivera O, Suárez-Calvet M, et al. Centiloid cut-off values for optimal agreement between PET and CSF core AD biomarkers. Alzheimers Res Ther. 2019;11(1):27.

    Article  PubMed  PubMed Central  Google Scholar 

  22. Ashburner J. A fast diffeomorphic image registration algorithm. Neuroimage. 2007;38(1):95–113.

    Article  PubMed  Google Scholar 

  23. Reitan R, Reitan R. Trail Making Test: Manual for Administration and Scoring. In 1992 [cited 2023 Oct 16]. Available from: https://www.semanticscholar.org/paper/Trail-Making-Test%3A-Manual-for-Administration-and-Reitan-Reitan/fb3612f133568883caa7c9c61e12d8cfe27807d3.

  24. Gramunt N, Buschke H, Sánchez-Benavides G, Lipton RB, Peña-Casanova J, Diéguez-Vide F, et al. Reference Data of the Spanish Memory Binding Test in a Midlife Population from the ALFA STUDY (Alzheimer’s and Family). J Alzheimers Dis. 2015;48(3):613–25.

    Article  PubMed  Google Scholar 

  25. Buschke H, Mowrey WB, Ramratan WS, Zimmerman ME, Loewenstein DA, Katz MJ, et al. Memory Binding Test Distinguishes Amnestic Mild Cognitive Impairment and Dementia from Cognitively Normal Elderly. Arch Clin Neuropsychol. 2017;32(1):29–39.

    Article  PubMed  PubMed Central  Google Scholar 

  26. Farrer LA, Cupples LA, Haines JL, Hyman B, Kukull WA, Mayeux R, et al. Effects of age, sex, and ethnicity on the association between apolipoprotein E genotype and Alzheimer disease. A meta-analysis. APOE and Alzheimer Disease Meta Analysis Consortium. JAMA. 1997;278(16):1349–56.

  27. Mosconi L, Rinne JO, Tsui WH, Berti V, Li Y, Wang H, et al. Increased fibrillar amyloid-{beta} burden in normal individuals with a family history of late-onset Alzheimer’s. Proc Natl Acad Sci U S A. 2010;107(13):5949–54.

    Article  PubMed  PubMed Central  Google Scholar 

  28. Jack CR, Bennett DA, Blennow K, Carrillo MC, Feldman HH, Frisoni GB, et al. A/T/N: An unbiased descriptive classification scheme for Alzheimer disease biomarkers. Neurology. 2016;87(5):539–47.

    Article  PubMed  PubMed Central  Google Scholar 

  29. Kärkkäinen M, Prakash M, Zare M, Tohka J, For The Alzheimer’s Disease Neuroimaging Initiative null. Structural Brain Imaging Phenotypes of Mild Cognitive Impairment (MCI) and Alzheimer’s Disease (AD) Found by Hierarchical Clustering. Int J Alzheimers Dis. 2020;2020:2142854.

  30. Hernández-Lorenzo L, Gil-Moreno MJ, Ortega-Madueño I, Cárdenas MC, Diez-Cirarda M, Delgado-Álvarez A, et al. A data-driven approach to complement the A/T/(N) classification system using CSF biomarkers. CNS Neurosci Ther. 2023.

  31. Dong A, Toledo JB, Honnorat N, Doshi J, Varol E, Sotiras A, et al. Heterogeneity of neuroanatomical patterns in prodromal Alzheimer’s disease: links to cognition, progression and biomarkers. Brain. 2017;140(3):735–47.

    PubMed  Google Scholar 

  32. Gamberger D, Lavrač N, Srivatsa S, Tanzi RE, Doraiswamy PM. Identification of clusters of rapid and slow decliners among subjects at risk for Alzheimer’s disease. Sci Rep. 2017;7(1):6763.

    Article  PubMed  PubMed Central  Google Scholar 

  33. Ritchie C, Smailagic N, Noel-Storr AH, Ukoumunne O, Ladds EC, Martin S. CSF tau and the CSF tau/ABeta ratio for the diagnosis of Alzheimer’s disease dementia and other dementias in people with mild cognitive impairment (MCI). Cochrane Database Syst Rev. 2017;3(3):CD010803.

  34. Heneka MT, Carson MJ, El Khoury J, Landreth GE, Brosseron F, Feinstein DL, et al. Neuroinflammation in Alzheimer’s Disease. Lancet Neurol. 2015;14(4):388–405.

    Article  PubMed  PubMed Central  Google Scholar 

  35. Ewers M, Biechele G, Suárez-Calvet M, Sacher C, Blume T, Morenas-Rodriguez E, et al. Higher CSF sTREM2 and microglia activation are associated with slower rates of beta-amyloid accumulation. EMBO Mol Med. 2020;12(9): e12308.

    Article  PubMed  PubMed Central  Google Scholar 

  36. Dicks E, Vermunt L, van der Flier WM, Visser PJ, Barkhof F, Scheltens P, et al. Modeling grey matter atrophy as a function of time, aging or cognitive decline show different anatomical patterns in Alzheimer’s disease. Neuroimage Clin. 2019;22: 101786.

    Article  PubMed  PubMed Central  Google Scholar 

  37. Jack CR, Knopman DS, Jagust WJ, Petersen RC, Weiner MW, Aisen PS, et al. Update on hypothetical model of Alzheimer’s disease biomarkers. Lancet Neurol. 2013;12(2):207–16.

    Article  PubMed  PubMed Central  Google Scholar 

  38. Convit A, de Asis J, de Leon MJ, Tarshish CY, De Santi S, Rusinek H. Atrophy of the medial occipitotemporal, inferior, and middle temporal gyri in non-demented elderly predict decline to Alzheimer’s disease. Neurobiol Aging. 2000;21(1):19–26.

    Article  PubMed  Google Scholar 

  39. Scott MR, Hampton OL, Buckley RF, Chhatwal JP, Hanseeuw BJ, Jacobs HI, et al. Inferior temporal tau is associated with accelerated prospective cortical thinning in clinically normal older adults. Neuroimage. 2020;15(220): 116991.

    Article  Google Scholar 

  40. Halawa OA, Gatchel JR, Amariglio RE, Rentz DM, Sperling RA, Johnson KA, et al. Inferior and medial temporal tau and cortical amyloid are associated with daily functional impairment in Alzheimer’s disease. Alzheimers Res Ther. 2019;11(1):14.

    Article  PubMed  PubMed Central  Google Scholar 

  41. Jack CR, Wiste HJ, Weigand SD, Therneau TM, Lowe VJ, Knopman DS, et al. Defining imaging biomarker cut-points for brain aging and Alzheimer’s disease. Alzheimers Dement. 2017;13(3):205–16.

    Article  PubMed  Google Scholar 

  42. Fjell AM, Westlye LT, Amlien I, Espeseth T, Reinvang I, Raz N, et al. High consistency of regional cortical thinning in aging across multiple samples. Cereb Cortex. 2009;19(9):2001–12.

    Article  PubMed  PubMed Central  Google Scholar 

  43. Bakkour A, Morris JC, Wolk DA, Dickerson BC. The effects of aging and Alzheimer’s disease on cerebral cortical anatomy: Specificity and differential relationships with cognition. Neuroimage. 2013;1(76):332–44.

    Article  Google Scholar 

  44. Chetelat G, Baron JC. Early diagnosis of Alzheimer’s disease: contribution of structural neuroimaging. Neuroimage. 2003;18(2):525–41.

    Article  PubMed  Google Scholar 

  45. Da Cunha E, Plonka A, Arslan S, Mouton A, Meyer T, Robert P, et al. Logogenic Primary Progressive Aphasia or Alzheimer Disease: Contribution of Acoustic Markers in Early Differential Diagnosis. Life (Basel). 2022;12(7):933.

    PubMed  Google Scholar 

  46. Jobson DD, Hase Y, Clarkson AN, Kalaria RN. The role of the medial prefrontal cortex in cognition, ageing and dementia. Brain Commun. 2021;3(3):fcab125.

    Article  PubMed  PubMed Central  Google Scholar 

  47. Albert MS, DeKosky ST, Dickson D, Dubois B, Feldman HH, Fox NC, et al. The diagnosis of mild cognitive impairment due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement. 2011;7(3):270–9.

    Article  PubMed  Google Scholar 

  48. Sánchez-Cubillo I, Periáñez JA, Adrover-Roig D, Rodríguez-Sánchez JM, Ríos-Lago M, Tirapu J, et al. Construct validity of the Trail Making Test: role of task-switching, working memory, inhibition/interference control, and visuomotor abilities. J Int Neuropsychol Soc. 2009;15(3):438–50.

    Article  PubMed  Google Scholar 

  49. Tideman P, Stomrud E, Leuzy A, Mattsson-Carlgren N, Palmqvist S, Hansson O, et al. Association of β-Amyloid Accumulation With Executive Function in Adults With Unimpaired Cognition. Neurology. 2022;98(15):e1525–33.

    Article  PubMed  PubMed Central  Google Scholar 

  50. Jack CR, Knopman DS, Weigand SD, Wiste HJ, Vemuri P, Lowe V, et al. An operational approach to National Institute on Aging-Alzheimer’s Association criteria for preclinical Alzheimer disease. Ann Neurol. 2012;71(6):765–75.

    Article  PubMed  PubMed Central  Google Scholar 

  51. Idland AV, Sala-Llonch R, Watne LO, Brækhus A, Hansson O, Blennow K, et al. Biomarker profiling beyond amyloid and tau: cerebrospinal fluid markers, hippocampal atrophy, and memory change in cognitively unimpaired older adults. Neurobiol Aging. 2020;93:1–15.

Download references

Acknowledgements

The authors would like to express their sincerest gratitude to the ALFA project participants, without whom this research would have not been possible. The Roche NeuroToolKit is a panel of exploratory prototype assays designed to robustly evaluate biomarkers associated with key pathologic events characteristic of AD and other neurological disorders, used for research purposes only and not approved for clinical use. Elecsys β-amyloid (1–42) CSF, Elecsys Phospho-Tau (181P) CSF and Elecsys Total-Tau CSF assays are approved for clinical use. This publication is part of the ALFA study (ALzheimer and FAmilies). Collaborators of the ALFA study are: Federica Anastasi, Annabella Beteta, Anna Brugulat-Serrat, Raffaele Cacciaglia, Irene Cumplido-Mayoral, Alba Cañas, Marta del Campo, Carme Deulofeu, Ruth Dominguez, Maria Emilio, Karine Fauria, Ana Fernández-Arcos, Sherezade Fuentes, Patricia Genius, Armand González-Escalante, Laura Hernández, Felipe Hernández-Villamizar, Jordi Huguet, David López-Martos, Ferran Lugo, Paula Marne, Tania Menchón, Carolina Minguillon, Paula Ortiz, Wiesje Pelkmans, Albina Polo, Sandra Pradas, Blanca Rodríguez-Fernandez, Mahnaz Shekari, Anna Soteras, Laura Stankeviciute, Marc Vilanova and Natalia Vilor-Tejedor.

Funding

The research leading to these results has received funding from “la Caixa” Foundation (ID 100010434), under agreement LCF/PR/GN17/50300004, the Alzheimer’s Association, and an international anonymous charity foundation through the TriBEKa Imaging Platform project (TriBEKa-17–519007). Support has also been received from the Universities and Research Secretariat, Ministry of Business and Knowledge of the Catalan Government under grant no. 2021 SGR 00913.

The research leading to these results has received funding by the Ministry of Science and Innovation (PID2019-111514RA-I00) and the Alzheimer’s Association research grants (AARG 2019-AARG-644641, AARG 2019-AARG-644641-RAPID), to EMA-U. EMA-U is supported by the Spanish Ministry of Science and Innovation—State Research Agency (RYC2018-026053-I), co-funded by the European Social Fund (ESF). EMA-U receives support from EU Joint Programme-Neurodegenerative Disease Research (JPND2022-138). E.P is funded by the Spanish Ministry of Science and Innovation (PRE2020-095827). GA receives support from the Alzheimer’s Association research fellowship (23AARF-1028873).

HZ is a Wallenberg Scholar supported by grants from the Swedish Research Council (#2022–01018 and #2019–02397), the European Union’s Horizon Europe research and innovation program under grant agreement No 101053962, Swedish State Support for Clinical Research (#ALFGBG-71320), the Alzheimer Drug Discovery Foundation (ADDF), USA (#201,809–2016862), the AD Strategic Fund and the Alzheimer's Association (#ADSF-21–831,376-C, #ADSF-21–831,381-C, and #ADSF-21–831,377-C), the Bluefield Project, the Olav Thon Foundation, the Erling-Persson Family Foundation, Stiftelsen för Gamla Tjänarinnor, Hjärnfonden, Sweden (#FO2022-0270), the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No 860197 (MIRIADE), the European Union Joint Program – Neurodegenerative Disease Research (JPND2021-00694), the National Institute for Health and Care Research University College London Hospitals Biomedical Research Centre, and the UK Dementia Research Institute at UCL (UKDRI-1003). MSC receives funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (Grant agreement No. 948677), Project "PI19/00155″, funded by Instituto de Salud Carlos III (ISCIII) and co-funded by the European Union, and from a fellowship from “la Caixa” Foundation (ID 100010434) and from the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No 847648 (LCF/BQ/PR21/11840004).

Author information

Author notes

  1. Georgette Argiris and Muge Akinci contributed equally to this work.

Authors and Affiliations

  1. Department of Neuroscience, Columbia University, New York, NY, USA

    Georgette Argiris

  2. Barcelona Institute of Global Health (ISGlobal), Health and Environment Over the Lifecourse Programme, Barcelona, Spain

    Muge Akinci, Eleni Palpatzis & Eider M. Arenaza-Urquijo

  3. University of Pompeu Fabra (UPF), Barcelona, Spain

    Muge Akinci, Eleni Palpatzis, Mahnaz Shekari, Eider M. Arenaza-Urquijo & for the ALFA study

  4. Barcelonaβeta Brain Research Center (BBRC), Pasqual Maragall Foundation, Barcelona, Spain

    Muge Akinci, Cleofé Peña-Gómez, Eleni Palpatzis, Marina Garcia-Prat, Mahnaz Shekari, Karine Fauria, Gonzalo Sánchez-Benavides, Oriol Grau-Rivera, Marc Suárez-Calvet & Eider M. Arenaza-Urquijo

  5. Hospital del Mar Research Institute, Barcelona, Spain

    Mahnaz Shekari, Gonzalo Sánchez-Benavides, Oriol Grau-Rivera & Marc Suárez-Calvet

  6. Department of Psychiatry and Neurochemistry, Institute of Neuroscience and Physiology, University of Gothenburg, Mölndal, Sweden

    Kaj Blennow, Henrik Zetterberg, Nicholas J. Ashton, Thomas K. Karikari, Ann Brinkmalm-Westman & Juan Lantero-Rodriguez

  7. Clinical Neurochemistry Laboratory, Sahlgrenska University Hospital, Mölndal, Sweden

    Kaj Blennow & Henrik Zetterberg

  8. UK Dementia Research Institute at UCL, London, UK

    Henrik Zetterberg

  9. Department of Neurodegenerative Disease, UCL Institute of Neurology, London, UK

    Henrik Zetterberg

  10. Hong Kong Center for Neurodegenerative Diseases, Hong Kong, China

    Henrik Zetterberg

  11. Wisconsin Alzheimer’s Disease Research Center, University of Wisconsin School of Medicine and Public Health, University of Wisconsin-Madison, Madison, WI, USA

    Henrik Zetterberg

  12. Roche Diagnostics GmbH, Penzberg, Germany

    Gwendlyn Kollmorgen

  13. Roche Diagnostics International Ltd, Rotkreuz, Switzerland

    Clara Quijano-Rubio

  14. Wallenberg Centre for Molecular and Translational Medicine, University of Gothenburg, Gothenburg, Sweden

    Nicholas J. Ashton

  15. King’s College London, Institute of Psychiatry, Psychology & Neuroscience, Maurice Wohl Clinical Neuroscience Institute, London, UK

    Nicholas J. Ashton

  16. NIHR Biomedical Research Centre for Mental Health & Biomedical Research Unit for Dementia at South London & Maudsley NHS Foundation, London, UK

    Nicholas J. Ashton

  17. Department of Psychiatry, University of Pittsburgh, Pittsburgh, PA, USA

    Thomas K. Karikari

  18. Centro de Investigación Biomédica en Red de Fragilidad y Envejecimiento Saludable (CIBERFES), Instituto de Salud Carlos III, Madrid, Spain

    Gonzalo Sánchez-Benavides, Oriol Grau-Rivera & Marc Suárez-Calvet

  19. Servei de Neurologia, Hospital del Mar, Barcelona, Spain

    Oriol Grau-Rivera & Marc Suárez-Calvet

  20. Department of Radiology, Mayo Clinic, Rochester, MN, USA

    Eider M. Arenaza-Urquijo

Authors
  1. Georgette Argiris
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. Muge Akinci
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Cleofé Peña-Gómez
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Eleni Palpatzis
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Marina Garcia-Prat
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. Mahnaz Shekari
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Kaj Blennow
    View author publications

    You can also search for this author inPubMed Google Scholar

  8. Henrik Zetterberg
    View author publications

    You can also search for this author inPubMed Google Scholar

  9. Gwendlyn Kollmorgen
    View author publications

    You can also search for this author inPubMed Google Scholar

  10. Clara Quijano-Rubio
    View author publications

    You can also search for this author inPubMed Google Scholar

  11. Nicholas J. Ashton
    View author publications

    You can also search for this author inPubMed Google Scholar

  12. Thomas K. Karikari
    View author publications

    You can also search for this author inPubMed Google Scholar

  13. Ann Brinkmalm-Westman
    View author publications

    You can also search for this author inPubMed Google Scholar

  14. Juan Lantero-Rodriguez
    View author publications

    You can also search for this author inPubMed Google Scholar

  15. Karine Fauria
    View author publications

    You can also search for this author inPubMed Google Scholar

  16. Gonzalo Sánchez-Benavides
    View author publications

    You can also search for this author inPubMed Google Scholar

  17. Oriol Grau-Rivera
    View author publications

    You can also search for this author inPubMed Google Scholar

  18. Marc Suárez-Calvet
    View author publications

    You can also search for this author inPubMed Google Scholar

  19. Eider M. Arenaza-Urquijo
    View author publications

    You can also search for this author inPubMed Google Scholar

  20. for the ALFA study
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

G.A., M.A., C.P.-G, and E.M.A.-U. were involved in conceptualization, design, statistical analysis, interpretation, and drafting of the study. M.G.-P., M.S.-C., and G.S.-B. played a major role in the acquisition of data. E.P., M.S., K.B., H.Z., G.K., C.Q.-R., N.J.A., T.K.K., A.B-W., J.L.-R., K.F., and O.G.-R. contributed to the acquisition or interpretation of the data. All authors have revised the manuscript for important intellectual content.

Corresponding authors

Correspondence to Karine Fauria or Eider M. Arenaza-Urquijo.

Ethics declarations

Ethics approval and consent to participate

The ALFA + study (ALFA-FPM-0311) was approved by the Independent Ethics Committee “Parc de Salut Mar,” Barcelona, and registered at Clinicaltrials.gov (identifier: NCT02485730) on June 10, 2015. All participants signed the informed consent form that had also been approved by the Independent Ethics Committee “Parc de Salut Mar,” Barcelona. The study was performed in accordance with the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments and comparable ethical standards.

Consent for publication

NA.

Competing interests

H.Z. has served at scientific advisory boards and/or as a consultant for Abbvie, Acumen, Alector, Alzinova, ALZPath, Annexon, Apellis, Artery Therapeutics, AZTherapies, Cognito Therapeutics, CogRx, Denali, Eisai, Merry Life, Nervgen, Novo Nordisk, Optoceutics, Passage Bio, Pinteon Therapeutics, Prothena, Red Abbey Labs, reMYND, Roche, Samumed, Siemens Healthineers, Triplet Therapeutics, and Wave, has given lectures in symposia sponsored by Alzecure, Biogen, Cellectricon, Fujirebio, Lilly, and Roche, and is a co-founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program (outside submitted work). G.K. is a full‑time employee of Roche Diagnostics GmbH. C.Q.-R. is a full‑time employee of Roche Diagnostics International Ltd. O.G-R. has given lectures in symposia sponsored by Roche Diagnostics, and receives support for research (to the institution) from F- Hoffmann La Roche. M.S.-C. has given lectures in symposia sponsored by Roche Diagnostics, S.L.U, Roche Farma, S.A and Amirall. He has served as a consultant and at advisory boards for Roche Diagnostics International Ltd and Grifols S.L. He was granted with a project funded by Roche Diagnostics International Ltd; payments were made to the institution (BBRC). He received in-kind support for research (to the institution) from Roche Diagnostics International Ltd, Avid Radiopharmaceuticals, Inc., Eli Lilly and Janssen Research & Development.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Argiris, G., Akinci, M., Peña-Gómez, C. et al. Data-driven CSF biomarker profiling: imaging and clinical outcomes in a cohort at risk of Alzheimer’s disease. Alz Res Therapy 16, 274 (2024). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s13195-024-01629-y

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s13195-024-01629-y

Keywords

Alzheimer's Research & Therapy

ISSN: 1758-9193