Sergi Borrego-Écija’s research while affiliated with University of Barcelona and other places

INTRODUCTION We aimed to determine whether cognitively unimpaired (CU) amyloid‐ beta‐positive (Aβ+) individuals display decreased practice effects on serial neuropsychological testing. METHODS We included 209 CU participants from three research centers, 157 Aβ− controls and 52 Aβ+ individuals. Participants underwent neuropsychological assessment at baseline and annually during a 2‐year follow‐up. We used linear mixed‐effects models to analyze cognitive change over time between the two groups, including time from baseline, amyloid status, their interaction, age, sex, and years of education as fixed effects and the intercept and time as random effects. RESULTS The Aβ+ group showed reduced practice effects in verbal learning (β = −1.14, SE = 0.40, p = 0.0046) and memory function (β = −0.56, SE = 0.19, p = 0.0035), as well as in language tasks (β = −0.59, SE = 0.19, p = 0.0027). DISCUSSION Individuals with normal cognition who are in the Alzheimer's continuum show decreased practice effects over annual neuropsychological testing. Our findings could have implications for the design and interpretation of primary prevention trials. Highlights This was a multicenter study on practice effects in asymptomatic Aβ+ individuals. We used LME models to analyze cognitive trajectories across multiple domains. Practice‐effects reductions might be an indicator of subtle cognitive decline. Implications on clinical and research settings within the AD field are discussed.

Objectives Frontotemporal dementia (FTD) usually shows more asymmetric atrophy patterns than Alzheimer’s disease (AD). We aim to quantify this asymmetry to differentiate FTD, AD, and FTD subtypes. Methods We studied T1-MRI scans, including FTD (different phenotypes), AD, and healthy controls (CTR). We defined the Cortical Asymmetry Index (CAI) using measures based on a metric derived from information theory with the cortical thickness measures. Some participants had additional follow-up MRIs, cerebrospinal fluid (CSF), or plasma measures. We analysed differences at cross-sectional and longitudinal levels. We then clustered FTD and AD participants based on the CAI values and studied the patients’ fluid biomarker characteristics within each cluster. Results A total of 101 FTD patients (64 ± 8 years, 53 men), 230 AD patients (65 ± 10 years, 84 men), and 173 CTR (59 ± 15 years, 67 men) were studied. CAI differentiated FTD, AD, and CTR. It also distinguished the semantic variant primary progressive aphasia (svPPA) from the other FTD phenotypes. In FTD, the CAI increased over time. The cluster analysis identified two subgroups within FTD, characterised by different neurofilament-light (NfL) levels, and two subgroups within AD, with different plasma glial fibrillary acidic protein (GFAP) levels. In AD, CAI correlated with GFAP and Mini-Mental State Examination (MMSE); in FTD, the CAI was associated with NfL levels. Conclusions The proposed method quantifies asymmetries previously described visually. The CAI could define clinically and biologically meaningful disease subgroups in the differential diagnosis of AD and FTD and its subtypes. CAI could also be of interest in tracking disease progression in FTD. Key Points Question There is a need to find quantitative metrics from MRI that can identify disease subgroups, and that could be useful for diagnosis and tracking. Findings We propose a Cortical Asymmetry Index that differentiates Alzheimer’s disease (AD) from Frontotemporal dementia (FTD), distinguishes FTD subtypes, correlates with NFL and GFAP levels, and monitors FTD progression. Clinical relevance Our proposed index holds the potential to support clinical applications for diagnosis and disease tracking in AD and FTD, using a quantitative summary metric from MRI data. It also contributes to the understanding of these diseases.

We used an untargeted mass spectrometric approach, tandem mass tag proteomics, for the identification of proteomic signatures in genetic frontotemporal dementia (FTD). A total of 238 cerebrospinal fluid (CSF) samples from the Genetic FTD Initiative were analyzed, including samples from 107 presymptomatic (44 C9orf72 , 38 GRN , and 25 MAPT ) and 55 symptomatic (27 C9orf72 , 17 GRN , and 11 MAPT ) mutation carriers as well as 76 mutation-negative controls (“noncarriers”). We found shared and distinct proteomic alterations in each genetic form of FTD. Among the proteins significantly altered in symptomatic mutation carriers compared with noncarriers, we found that a set of proteins including neuronal pentraxin 2 and fatty acid binding protein 3 changed across all three genetic forms of FTD and patients with Alzheimer’s disease from previously published datasets. We observed differential changes in lysosomal proteins among symptomatic mutation carriers with marked abundance decreases in MAPT carriers but not other carriers. Further, we identified mutation-associated proteomic changes already evident in presymptomatic mutation carriers. Weighted gene coexpression network analysis combined with gene ontology annotation revealed clusters of proteins enriched in neurodegeneration and glial responses as well as synapse- or lysosome-related proteins indicating that these are the central biological processes affected in genetic FTD. These clusters correlated with measures of disease severity and were associated with cognitive decline. This study revealed distinct proteomic changes in the CSF of patients with genetic FTD, providing insights into the pathological processes involved in the disease. In addition, we identified proteins that warrant further exploration as diagnostic and prognostic biomarker candidates.

Plasma tau phosphorylated at threonine 181 (p-tau181) and 217 (p-tau217) have demonstrated high accuracy for Alzheimer’s disease (AD) diagnosis, defined by CSF/PET amyloid beta (Aβ) positivity, but most studies have been performed in research cohorts, limiting their generalizability. We studied plasma p-tau217 and p-tau181 for CSF Aβ status discrimination in a cohort of consecutive patients attending an academic memory clinic in Spain (July 2019–June 2024). All patients had CSF AD biomarkers performed as part of their routine clinical assessment. Aβ positivity was defined with a local cut-off of CSF Aβ1–42 < 600 pg/mL; in patients with borderline Aβ1–42 values or when there was a mismatch between the Aβ and the T status (T + if CSF p-tau181 ≥ 65 pg/mL), a ratio Aβ1–42/Aβ1–40 < 0.07 was used. Plasma p-tau217 and p-tau181 were measured retrospectively, from blood samples collected at first visit, with Fujirebio Lumipulse and Quanterix Simoa assays, respectively. We included 468 patients (mean age 67 years, 50% female, 61% Aβ positive). Plasma p-tau217 outperformed plasma p-tau181 in discriminating CSF Aβ status (AUC 0.95 vs 0.90, p = 0.005). A 97.5% sensitivity and specificity plasma p-tau217 algorithm, classifying patients into three groups of Aβ probability (Low, Intermediate and High), resulted in 67% of patients in the Low and High groups, having their Aβ status predicted (as negative and positive, respectively) with 96% accuracy. The remaining 33% in the Intermediate group were candidates to undergo CSF/PET testing. A model with a 10% variation in p-tau217 levels yielded small changes in accuracy (95%). In conclusion, plasma p-tau217 could have discriminated CSF Aβ status in two-thirds of patients with very high accuracy in a memory clinic cohort. These results support the implementation of plasma p-tau217 as an initial diagnostic tool in memory clinics for AD diagnosis, reducing the need for more invasive/expensive testing.

Background Primary progressive aphasia (PPA) is a language‐based dementia linked with underlying Alzheimer’s disease (AD) or frontotemporal dementia. Clinicians often report difficulty differentiating between the logopenic (lv) and nonfluent/agrammatic (nfv) subtypes, as both variants present with disruptions to “fluency” yet for different underlying reasons. In English, acoustic and linguistic markers from connected speech samples have shown promise in machine learning (ML)‐based differentiation of nfv from lv. To our knowledge, this approach has not been evaluated in other languages nor in the context of bilingualism. Method Twenty‐four Spanish‐Catalan bilingual patients (lv=15, nfv=9) were asked to describe a picture (WAB Picnic Scene) in both their dominant and non‐dominant language. From the participant’s recorded response, 10 acoustic features were derived with PRAAT and 15 linguistic features were derived with the Natural Language Processing (NLP) tools SpaCy and CLAN. A similarity score between the image and patient’s transcription was derived with the Vision‐Language model CLIP. The acoustic features, linguistic features, and CLIP scores, were separately fed into ML classification algorithms for differentiating nfv from lv in participants’ dominant and non‐dominant samples. Result The acoustic‐based classifiers achieved classification accuracy (F1 score) of 59% in the dominant and 86% in the non‐dominant language, respectively. The linguistic‐based classifiers achieved F1 scores of 73% in the dominant and 77% in the non‐dominant language, respectively. The CLIP‐based classifier achieved F1 scores of 82% in the dominant and 82% in the nondominant language, respectively. The acoustic and linguistic classifier performed 25% (p=0.077) and 4% (p=0.18) better given only non‐dominant samples compared to only dominant samples. Conclusion Taking advantage of recent advances in multilingual NLP, we achieved promising and effective differentiation of nfv from lv for Spanish‐Catalan bilingual patients using a nearly automated pipeline. Interestingly, our acoustic and linguistic‐based classifiers performed better given responses from a patient’s non‐dominant language, and the acoustic feature set was more accurate in discriminating between nfv and lv compared to the linguistic model. Future directions include examining patterns in a larger sample size and comparison of performance on different types of connected speech tasks.

Background Alzheimer's disease (AD) features a complex interplay of factors influencing cognitive decline. While CSF and plasma biomarkers have widely demonstrated their diagnostic utility, whether they may add prognostic value remains unrevealed. With this longitudinal study we aim to address this knowledge gap by evaluating the predictive value of several fluid biomarkers over cognitive decline in a cohort of biomarker‐confirmed AD individuals. Method We included 139 participants with biologically‐confirmed AD (A+T+N+). Four cerebrospinal fluid (CSF) biomarkers (Amyloid‐Beta1‐42 [Aβ1‐42], tau phosphorylated at threonine 181 [p‐tau181], total tau [t‐tau], and neurofilament light chain [NfL]) were determined with enzyme immunoassay, and three plasma biomarkers (p‐tau181, NfL and glial fibrillary acidic protein [GFAP]) were determined with single‐molecule array. Biomarkers were stratified into tertiles. Comprehensive neuropsychological assessments were administered at baseline (n=139) and annually (Year 1 n=108, Year 2 n=78, Year 3 n=25, Year 4 n=3 and Year 5 n=3; mean follow‐up time 1.7 years [SD 0.3]). Mixed Models for Repeated Measures explored the effectof CSF and blood biomarkers on Mini‐Mental State Examination (MMSE) score progression. Result Participants had a mean age at onset of 65.7 (SD 6.4) years, 17% were non‐amnestic, 58% were APOEε4 carriers. Higher baseline plasma p‐tau181 and GFAP concentrations correlated with MMSE score decline (p=0.009 and p=0.002, respectively) (Table 1, Table 2, Figure 1). Conversely, no significant associations were observed between plasma NfL or CSF biomarkers concentrations and MMSE decline. Conclusion This longitudinal study highlights the potential prognostic value of baseline plasma p‐tau181 and GFAP concentrations for cognitive decline progression in AD.

Background Primary progressive aphasia (PPA) is a language‐based dementia linked with underlying Alzheimer’s disease (AD) or frontotemporal dementia. Clinicians often report difficulty differentiating between the logopenic (lv) and nonfluent/agrammatic (nfv) subtypes, as both variants present with disruptions to “fluency” yet for different underlying reasons. In English, acoustic and linguistic markers from connected speech samples have shown promise in machine learning (ML)‐based differentiation of nfv from lv. To our knowledge, this approach has not been evaluated in other languages nor in the context of bilingualism. Method Twenty‐four Spanish‐Catalan bilingual patients (lv = 15, nfv = 9) were asked to describe a picture (WAB Picnic Scene) in both their dominant and non‐dominant language. From the participant’s recorded response, 10 acoustic features were derived with PRAAT and 15 linguistic features were derived with the Natural Language Processing (NLP) tools SpaCy and CLAN. A similarity score between the image and patient’s transcription was derived with the Vision‐Language model CLIP. The acoustic features, linguistic features, and CLIP scores, were separately fed into ML classification algorithms for differentiating nfv from lv in participants' dominant and non‐dominant samples. Result The acoustic‐based classifiers achieved classification accuracy (F1 score) of 59% in the dominant and 86% in the non‐dominant language, respectively. The linguistic‐based classifiers achieved F1 scores of 73% in the dominant and 77% in the non‐dominant language, respectively. The CLIP‐based classifier achieved F1 scores of 82% in the dominant and 82% in the nondominant language, respectively. The acoustic and linguistic classifier performed 25% (p = 0.077) and 4% (p = 0.18) better given only non‐dominant samples compared to only dominant samples. Conclusion Taking advantage of recent advances in multilingual NLP, we achieved promising and effective differentiation of nfv from lv for Spanish‐Catalan bilingual patients using a nearly automated pipeline. Interestingly, our acoustic and linguistic‐based classifiers performed better given responses from a patient’s non‐dominant language, and the acoustic feature set was more accurate in discriminating between nfv and lv compared to the linguistic model. Future directions include examining patterns in a larger sample size and comparison of performance on different types of connected speech tasks.

Background Epigenetic mechanisms as a potential underlying pathogenic mechanism of neurodegenerative diseases have been the scope of several studies performed so far. However, there is a gap in analyzing different forms of early‐onset dementia to minimize the effect of aging and the use of Lymphoblastoid cell lines (LCLs) as a possible disease model for earlier clinical phases. Method We performed a genome‐wide DNA methylation analysis in 64 samples (from prefrontal cortex and lymphoblastoid cell lines) from Alzheimer’s Disease (AD) and Frontotemporal dementia (FTD) using the Illumina Infinium MethylationEPIC V2.0 array. The studied cohort included sporadic early‐onset (sEOAD, sFTD‐TP43, sFTD‐Tau) and genetic subgroups of AD (PSEN1) and FTD (MAPT, GRN, C9orf72), with n = 5 subjects/group. We analyzed the differentially methylated positions (DMPs) using the Beta regression model, with age and sex as covariates, and all p‐values adjusted by False Discovery Rate (FDR). Venn diagrams to visualize common genes between pairwise comparisons and heatmaps were performed to further explore the most important DMPs. Elastic Net logistic regression was used to obtain epigenetic diagnostic signatures. We also performed a correlation analysis of DNA methylation levels with Clariom D array gene expression data for the same cohort. Result Results showed hypermethylation in patients’ groups as the most frequent finding in both tissues studied (Fig. 1). We identified common DMPs when comparing patients with healthy controls (CTRL) for each respective disease (Fig. 2, 3). Biological significance analysis revealed common pathways altered in AD and FTD affecting neuron development, metabolism, signal transduction and immune system pathways. These alterations were also found in LCLs, even some related to neuron development. We obtained diagnostic signatures to differentiate patients from CTRL. In the brain, CpG methylation presented an inverse correlation with gene expression, while in LCLs we observed mainly a positive correlation. Conclusion This study enhances our understanding about the biological pathways that are associated with neurodegeneration, describes differential methylation patterns, and suggests LCLs are a potential cell model for studying neurodegenerative diseases in early clinical phases.

Background Practice effects are a well‐known cognitive phenomenon that is reduced in patients with Alzheimer’s disease (AD). We aimed to investigate whether cognitively unimpaired (CU) individuals within the Alzheimer’s continuum (i.e., positive amyloid‐β biomarker) display decreased practice effects on serial neuropsychological testing. Methods We included 310 CU from four Spanish research centers, classified into controls (n = 250) or Aβ+ (n = 60). In the main cohort (Cohort A; n = 209), participants underwent neuropsychological assessment at baseline and annually during a 2‐year follow‐up (FU1 and FU2). A “long‐term cohort” (Cohort B; n = 101) was employed to assess practice effects over longer time periods (i.e., two follow‐up sessions at years 3 and 6 from baseline). Practice effects were defined as simple discrepancy scores (SDS) by subtracting Z‐scores at FU2 from Z‐scores at baseline for each neuropsychological variable, and linear mixed effects models (LME) were run to assess the temporal evolution of practice effects according to the two groups. Results There were no cross‐sectional differences between the control and Aβ+ groups in none of the neuropsychological scores at baseline (Fig. 1). The Aβ+ group displayed lower practice effects than the controls in terms of SDS in several neuropsychological outcomes (Fig. 2). In Cohort A, LME showed negative slopes by the Aβ+ group in verbal memory measures such as the free learning score (β = ‐0.37, SD = 0.12, p = 0.0034), delayed free recall (β = ‐0.43, SD = 0.15, p = 0.0047) and delayed total recall (β = ‐0.46, SD = 0.17, p = 0.0069) from the Free and Cued Selective Reminding Test; as well as in language tasks (Boston Naming Test; β = ‐0.26, SD = 0.087, p = 0.0025) and executive function measures (Trail Making Test; β = ‐0.33, SD = 0.12, p = 0.0094) (Fig. 3A). In Cohort B, similar findings were observed in visual memory measures, such as the Rey‐Osterrieth Complex Figure immediate (β = ‐0.80, SD = 0.35, p = 0.024) and delayed (β = ‐1.25, SD = 0.34, p = 0.00038) recall (Fig. 3B). Conclusions Individuals with normal cognition who are in the Alzheimer’s continuum show decreased practice effects over serial neuropsychological testing. Our findings suggest the reduction of practice effects, particularly in memory measures, as an indicator of subtle cognitive decline in the earliest phase of the Alzheimer’s continuum and could be particularly relevant for the design and interpretation of primary prevention trials on disease‐modifying therapies.

Background and objectives: Pathogenic variants in the GRN gene cause frontotemporal dementia (FTD-GRN) with marked brain asymmetry. This study aims to assess whether the disease progression of FTD-GRN depends on the initial side of the atrophy. We also investigated the potential use of brain asymmetry as a biomarker of the disease. Methods: Retrospective examination of data from the prospective Genetic Frontotemporal Initiative (GENFI) cohort study that recruits individuals who carry or were at risk of carrying a pathogenic variant causing FTD. GENFI participants underwent a standardized clinical and neuropsychological assessment, MRI, and a blood sample test yearly. We generated an asymmetry index for brain MRI to characterize brain asymmetry in participants with or at risk of FTD-GRN. Depending on the side of the asymmetry, we classified symptomatic GRN patients as right-GRN or left-GRN and compared their clinical features and disease progression. We generated generalized additive models to study how the asymmetry index evolves in carriers and noncarriers and compare its models with others created with volumetric values and plasma neurofilament light chain. Results: A total of 399 participants (mean age 49.7 years, 59% female) were included (63 symptomatic carriers, 177 presymptomatic carriers, and 159 noncarriers). Symptomatic carriers showed higher brain asymmetry (11.6) than noncarriers (1.0, p < 0.001) and presymptomatic carriers (1.0, p < 0.001), making it possible to classify most of them as right-GRN (n = 21) or left-GRN (n = 36). Patients with right-GRN showed more disease severity at baseline (β = 6.9, 95% CI 2.4-11.0, p = 0.003) but a lower deterioration by year (β = -1.5, 95% CI -2.7 to -0.31, p = 0.015) than patients with left-GRN. Brain asymmetry could be found in GRN carriers 10.4 years before the onset of the symptoms (standard difference 0.85, CI 0.01-1.68). Discussion: FTD-GRN affects the brain hemispheres asymmetrically and causes 2 anatomical asymmetry patterns depending on the side of the disease onset. We demonstrated that these 2 anatomical asymmetry patterns present different symptoms, severity at the time of the first visit, and different disease courses. Our results also suggest brain asymmetry as a possible biomarker of conversion in GRN carriers.