Widespread choroid plexus contamination in sampling and profiling of brain tissue

Abstract

The choroid plexus, a tissue responsible for producing cerebrospinal fluid, is found predominantly in the lateral and fourth ventricles of the brain. This highly vascularized and ciliated tissue is made up of specialized epithelial cells and capillary networks surrounded by connective tissue. Given the complex structure of the choroid plexus, this can potentially result in contamination during routine tissue dissection. Bulk and single-cell RNA sequencing studies, as well as genome-wide in situ hybridization experiments (Allen Brain Atlas), have identified several canonical markers of choroid plexus such as Ttr, Folr1, and Prlr. We used the Ttr gene as a marker to query the Gene Expression Omnibus database for transcriptome studies of brain tissue and identified at least some level of likely choroid contamination in numerous studies that could have potentially confounded data analysis and interpretation. We also analyzed transcriptomic datasets from human samples from Allen Brain Atlas and the Genotype-Tissue Expression (GTEx) database and found abundant choroid contamination, with regions in closer proximity to choroid more likely to be impacted such as hippocampus, cervical spinal cord, substantia nigra, hypothalamus, and amygdala. In addition, analysis of both the Allen Brain Atlas and GTEx datasets for differentially expressed genes between likely “high contamination” and “low contamination” groups revealed a clear enrichment of choroid plexus marker genes and gene ontology pathways characteristic of these ciliated choroid cells. Inclusion of these contaminated samples could result in biological misinterpretation or simply add to the statistical noise and mask true effects. We cannot assert that Ttr or other genes/proteins queried in targeted assays are artifacts from choroid contamination as some of these differentials may be due to true biological effects. However, for studies that have an unequal distribution of choroid contamination among groups, investigators may wish to remove contaminated samples from analyses or incorporate choroid marker gene expression into their statistical modeling. In addition, we suggest that a simple RT-qPCR or western blot for choroid markers would mitigate unintended choroid contamination for any experiment, but particularly for samples intended for more costly omic profiling. This study highlights an unexpected problem for neuroscientists, but it is also quite possible that unintended contamination of adjacent structures occurs during dissections for other tissues but has not been widely recognized.

Thyroid Hormone T4 Mitigates Traumatic Brain Injury by Dynamically Remodeling Cell Type Specific Genes, Pathways, and Networks in Hippocampus and Frontal Cortex

“…64 These cells are not part of the hippocampus but are typically included during hippocampal dissection. 33 All these changes occurred in or near the hippocampus, again supporting that this brain region and its related cell types are main T4 targets. To supplement the cell proportion analysis, we used a Euclidean distance-based method to measure global transcriptomic changes in individual cell types (Fig.3b).…”

“…Notably, choroid plexus is not part of the hippocampus but can easily be included during hippocampus dissection. 33 Endothelial and mural cells are important for blood-brain-barrier (BBB) integrity and function while choroid plexus plays a role in neural repair. 34 Therefore, T4 treatment altered the proportions of multiple cell types in or near the hippocampus that either counteracted mTBI-induced changes or enhanced protective and reparative cell types in response to injury.…”

An integrated single-nucleus and spatial transcriptomics atlas reveals the molecular landscape of the human hippocampus

“…45 , Extended Data Fig. 46 ) [35], which were excluded from marker gene detection and enrichment analyses.…”

“…We also identified 1 pattern that appeared to capture variance related to choroid plexus “smearing” and contamination, a known phenomenon in brain transcriptomics studies ( Extended Data Fig. 46 ) [35]. These patterns were excluded from downstream analyses.…”

What has single-cell transcriptomics taught us about long non-coding RNAs in the ventricular-subventricular zone?

“…This can be in part overcome using a combination of known gene markers for specific anatomical locations, although this is dependent on a priori knowledge of these markers and may have limited anatomical resolution. This is particularly relevant in the context of scRNA-seq in the V-SVZ as samples are frequently contaminated with surrounding tissue, for example from the striatum or choroid plexus ( Olney et al., 2022 ). There are many emerging variations of scRNA-seq that consider spatial location of cells, such as “cell2location” that allows spatial information to be integrated into the data analysis ( Kleshchevnikov et al., 2022 ).…”