Applying Large-Scale Transcriptomic Data to Analyze In Vitro Human Hypothalamic Differentiation Protocols

Abstract

The increased prevalence of obesity is causing challenges on individual and socioeconomic levels. Neuronal subtypes of the hypothalamus control appetite regulation through the secretion of peptide hormones and neurotransmitters. Our limited knowledge of early molecular patterning of the human hypothalamus sets a challenge to design human in vitro models for cellular studies and drug screening. In Kirkeby lab, we have established in vitro human hypothalamic differentiation protocols from stem cells to produce regional subtypes of the hypothalamus for studying neuronal control of appetite. Here, publicly available single-cell RNA sequencing (scRNA-seq) and spatial transcriptomics data are used as reference to study the cell type composition of our human stem cell differentiation protocols. scANVI – the best-performing integration method in our benchmarking study – was used to integrate two publicly available scRNA-seq hypothalamus datasets into a reference atlas of human fetal hypothalamus. The respective scANVI model accurately predicts neural progenitors, neurons, and arcuate nucleus, a hypothalamic region having a crucial role in appetite regulation, from the stem cell data. The results were enhanced by using spatial mapping in which the single-cell data from the stem cell-derived hypothalamic progenitors spatially aligned precisely to the corresponding anatomical location of the hypothalamus in the neural tube, confirming the accuracy of the differentiation protocols to recapitulate human hypothalamic development in vitro. Nonetheless, we observed the integrated single-cell human fetal hypothalamic atlas containing both currently available developmental human hypothalamus datasets was not comprehensive enough to predict rare transcriptionally similar cell types correctly. Currently, only a subset of the neurons in the integrated dataset are divided into subpopulations. In cell type prediction majority of the subpopulations are predicted as neurons. Hence dividing the remaining neurons into their respective subclusters might increase the prediction accuracy.

Popular Abstract

Reference-based analysis of in vitro developing human hypothalamus

The human hypothalamus located at most ventral part of the forebrain is a complex brain region controlling multiple fundamental processes. The increased prevalence of obesity has caused a worldwide epidemic posing a serious challenge for society. In our lab, we have established in vitro differentiation protocols from human embryonic stem cells to produce three distinct regions (nuclei) of the human hypothalamus. These protocols provide a disease modeling and drug screening platform for obesity. With single-cell RNA sequencing, the results from these protocols are transformed to a data format interpretable by computer. Here, we apply novel computational techniques to analyze the protocol data using reference-based approaches, i.e. utilizing biological information stored in reference data.

In the present study, two publicly available single-cell RNA sequencing datasets are integrated into a human fetal hypothalamus atlas, serving as our reference. Through integration, datasets from different sources become comparable and provide a more comprehensive resource. To ensure the usage of the most appropriate integration algorithm, we tested 13 different algorithms with our benchmarking tool, which revealed scANVI was the top performer. Thereby, scANVI integrated reference was used to make cell type predictions from data generated from our in vitro human hypothalamic differentiation protocols.

Single-cell RNA sequencing enables studying cellular processes at single-cell resolution, but it doesn’t provide the spatial locations of cells in a tissue. Spatial transcriptomics, a recent bioinformatics technique, overcomes the issue by providing the anatomical location of each cell. Spatial methods allow addressing how closely our protocols modeling the hypothalamus corresponds to its counterpart dissected from a human embryo. Therefore, as a validation step of our differentiation protocol, its respective data were spatially aligned using the BoneFight algorithm towards human fetal neural tube spatial data.

In the present study, we described the first integrated single-cell atlas of the human fetal hypothalamus. Based on marker gene expression and exploration of hierarchical connections of cell types, we can conclude the integrated reference conserved the expected biology. The reference predicted the cell types accurately from our protocol data and identified an arcuate nucleus population from a protocol whose purpose is to generate neurons of the arcuate nucleus. By applying spatial mapping we aligned data from the hypothalamic differentiation protocols to anatomically correct location.

These results suggest that our in-vitro hypothalamic protocols recapitulate hypothalamic development in vivo. According to the reference-based cell diversity analysis, specifically, the arcuate nucleus protocol produces neurons of the arcuate nucleus as intended.


Master’s Degree Project in Bioinformatics 30 credits 2023
Department of Biology, Lund University

Advisor: Agnete Kirkeby
Department of Experimental Medicine Lund, Human Neural Development group

• author: Hänninen, Erno
• supervisor: Agnete Kirkeby
• organization: Degree Projects in Bioinformatics
• course: BINP50 20231
• year: 2023
• type: H2 - Master's Degree (Two Years)
• subject: Biology and Life Sciences
• language: English
• id: 9122489
• date added to LUP: 2023-06-09 12:07:04
• date last changed: 2023-06-09 12:07:04