Using the power of RNA-seq to characterize brain cell types

Out of many, one: the brain as a heterogeneous whole

There are as many brain cells as there are stars in our galaxy—approximately 100 billion—with each cell possessing a unique function, driven by its unique transcriptional signature. These signatures are, however, lost in traditional bulk RNA-seq experiments, potentially masking the identity of specific cell type(s) that may drive gene expression differences in disease.

To address this challenge, the Allen Institute for Brain Science (AIBS) used our SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing to develop a straightforward workflow for generating transcriptional data from single nuclei derived from human brain tissue.

Separation anxiety: isolating and labeling single nuclei from human brain

Working with postmortem human brain tissue comes with several challenges, most notably the need for rapid preservation of human brain tissue (typically by flash-freezing) to maintain sample quality. Due to the incompatibility of flash-freezing with the isolation of intact cells from postmortem human brain, AIBS researchers generated nuclei preparations from cortical (middle temporal gyrus, MTG) and lateral geniculate nucleus (LGN) human brain tissue samples. Nuclei were then stained with a PE-conjugated antibody against NeuN—a nuclear-enriched marker specific to neuronal cells and nuclei—and DAPI. Single-nuclei preps were generated using DAPI and PE (NeuN) signals for FACS.

We go to 11: sequencing of SMART-Seq v4 amplified libraries

Due to the low amount of material in single nuclei, and the SMART-Seq v4 Ultra Low Input RNA Kit for Sequencing's track record for robust and highly sensitive amplification of as little as one cell (10 pg) of total RNA, AIBS used our kit to amplify RNA prior to library generation. Following amplification with SMART-Seq v4 and library construction of 1,576 LGN and 15,928 MTG, libraries were sequenced to a depth of ~10 million reads. These libraries passed QC metrics with flying colors, with a mean of 87% of reads aligning successfully and an average of 28.3% and 38.5% of reads mapping to exons and introns, respectively.

Nuclei were then clustered using principal component analyses followed by nearest-neighbor analysis (Bakken et al. 2017) and categorized into broad cell types (GABAergic interneuron, glutamatergic neuron, astrocyte, microglia, etc.) based on marker gene expression. Broad cell types containing more than one cluster were further subdivided based upon the complement of genes most specific to individual clusters. A summary of sequencing metrics by cluster can be seen in Tables 9 and 11 of the AIBS protocol.

It pays to be smart: robust and sensitive amplification of ultra-low-input samples

Working with single cells and nuclei for sequencing represents a significant challenge due to extremely small RNA input sizes, but these sample types will only grow in importance as we shift from tissue- to cell-level gene expression profiling. Choosing the right tools for RNA-seq library preparation from single-cell and single-nuclei inputs is crucial to generating reliable data, and this study exemplifies the sensitivity and strength of our SMART-Seq v4 Ultra Low Input Kit for Sequencing.

References

Bakken, T. E. et al. Equivalent high-resolution identification of neuronal cell types with single-nucleus and single-cell RNA-sequencing. bioRxiv 239749, doi:10.1101/239749 (2017).