Spatial Multi-omic Profiling of Alzheimer’s Disease in the Human Inferior Temporal Cortex

Spatial Multi-omic Profiling of Alzheimer’s Disease in the Human Inferior Temporal Cortex


Author(s): Sowmya Parthiban,Sang Ho Kwon,Madhavi Tippani,Heena R Divecha,Jashandeep S Lobana,Stephen Williams,Michelle Mark,Guixia Yu,Julianna Avalos-Gracia,Rahul A Bharadwaj,Joel E Klenman,Thomas M Hyde,Stephanie C Page,Stephanie Hicks,Keri Martinowich,Kristen R Maynard,Leonardo Collado Torres

Affiliation(s): Johns Hopkins School of Public Health, Department of Biostatistics

Social media: https://twitter.com/sowmyapartybun

The Visium SPG (Spatial Proteogenomics) platform, namely 10x Visium coupled with IF (immunofluorescence) protein detection, answers biological questions about spatial association between RNA and protein expression within identical tissue sections. In this talk, I will first introduce the Visium SPG platform broadly. This platform allows us to identify genome-wide transcriptome changes associated with the neighboring expression of proteins of interest. Spatial clustering methods alone cannot reliably detect in the absence of spatially-resolved protein information. Then, I will give a specific example where we used the platform to address the correspondence between various AD (Alzheimer’s Disease) related neuropathologies and transcriptomic expression at the spot level in human tissues. I will also highlight statistical and machine learning methods produced by the Bioconductor community in application to this novel data type. The spatial dispersion of neuropathological hallmarks of AD amyloid beta (Abeta), hyperphosphorylated tau (pTau) - and gene expression levels were evaluated using IF and Visium respectively on cryosections from 3 AD and 1 neurotypical donor (n = 38287 spots across n = 10 tissue sections). After performing quality control, and batch correction using Harmony on gene expression and image data, spots in the AD samples were labeled as one of (Abeta, pTau, next-Abeta, next-pTau, both, next-both, none) based on a manual thresholded approach. Downstream analyses including unsupervised clustering, pseudo-bulking, differential expression, and spatial registration were performed to identify distinct transcriptional signatures associated with spatially-resolved AD neuropathologies within AD subjects, with Bioconductor packages including BayesSpace, scuttle, and spatialLIBD. This work demonstrates how innovative multi-omic statistical approaches can lead to molecular insights into neuropathology on spatially-defined brain microenvironments. Finally, we generated a user-friendly web resource to explore the dataset further: https://libd.shinyapps.io/Visium_IF_AD_Kwon2022/