Search

BackgroundThe ATN model represents a research framework used to describe in subjects the presence or absence of Alzheimer’s disease (AD) pathology through biomarkers. The aim of this study was to describe the prevalence of different ATN profiles using quantitative imaging biomarkers in two independent cohorts, and to evaluate the pertinence of ATN biomarkers to identify comparable populations across independent cohorts.MethodsA total of 172 subjects from the Geneva Memory Clinic and 113 volunteers from a study on healthy aging at the University Hospital of Zurich underwent amyloid (A) and tau (T) PET, as well as T1-weigthed MRI scans using site-specific protocols. Subjects were classified by cognition (cognitively unimpaired, CU, or impaired, CI) based on clinical assessment by experts. Amyloid data converted into the standardized centiloid scale, tau PET data normalized to cerebellar uptake, and hippocampal volume expressed as a ratio over total intracranial volume ratio were considered as biomarkers for A, T, and neurodegeneration (N), respectively. Positivity for each biomarker was defined based on previously published thresholds. Subjects were then classified according to the ATN model. Differences among profiles were tested using Kruskal-Wallis ANOVA, and between cohorts using Wilcoxon tests.ResultsTwenty-nine percent of subjects from the Geneva cohorts were classified with a normal (A−T−N−) profile, while the Zurich cohort included 64% of subjects in the same category. Meanwhile, 63% of the Geneva and 16% of the Zurich cohort were classified within the AD continuum (being A+ regardless of other biomarkers’ statuses). Within cohorts, ATN profiles were significantly different for age and mini-mental state examination scores, but not for years of education. Age was not significantly different between cohorts. In general, imaging A and T biomarkers were significantly different between cohorts, but they were no longer significantly different when stratifying the cohorts by ATN profile. N was not significantly different between cohorts.ConclusionStratifying subjects into ATN profiles provides comparable groups of subjects even when individual recruitment followed different criteria.

Amplicon high-throughput sequencing of 16S ribosomal RNA (rRNA) gene is currently the most widely used technique to investigate complex gut microbial communities. Microbial identification might be influenced by several factors, including the choice of bioinformatic pipelines, making comparisons across studies difficult. Here, we compared four commonly used pipelines (QIIME2, Bioconductor, UPARSE and mothur) run on two operating systems (OS) (Linux and Mac), to evaluate the impact of bioinformatic pipeline and OS on the taxonomic classification of 40 human stool samples. We applied the SILVA 132 reference database for all the pipelines. We compared phyla and genera identification and relative abundances across the four pipelines using the Friedman rank sum test. QIIME2 and Bioconductor provided identical outputs on Linux and Mac OS, while UPARSE and mothur reported only minimal differences between OS. Taxa assignments were consistent at both phylum and genus level across all the pipelines. However, a difference in terms of relative abundance was identified for all phyla (p < 0.013) and for the majority of the most abundant genera (p < 0.028), such as Bacteroides (QIIME2: 24.5%, Bioconductor: 24.6%, UPARSE-linux: 23.6%, UPARSE-mac: 20.6%, mothur-linux: 22.2%, mothur-mac: 21.6%, p < 0.001). The use of different bioinformatic pipelines affects the estimation of the relative abundance of gut microbial community, indicating that studies using different pipelines cannot be directly compared. A harmonization procedure is needed to move the field forward.