Principles of transcriptional bursting in mammalian cells

Abstract

In mammalian cells, transcription occurs in discrete bursts leading to fluctuations in transcripts from expressed genes. Although this behavior was first reported not long after the discovery of messenger RNA (mRNA), the methods to measure transcriptional bursting have been limited in throughput and scalability. To enable transcriptome wide analysis of transcriptional bursting, I have developed multiple methods to estimate transcriptional bursting behavior using deeply sequenced single-cell RNA-sequencing data.

In Paper I, we use a computational likelihood method based on the two-state model of transcriptional bursting to estimate allele-resolved bursting kinetics of mouse cells. The transcriptome wide estimates allow us to detect how the genomic regions of enhancers and promoters affect transcriptional bursts. To a first approximation, enhancers direct the frequency of bursts while promoters influence the number of transcripts per burst.

The fluctuations of the transcript alleles may cause phenotypic variability over time. In Paper II, we directly show that the bursting behavior of a gene determines how often monoallelic expression is observed from that gene. Moreover, we show that this can lead to false positive monoallelic observations in bulk experiments if not considered. This can be concluded for the genes present on autosomal chromosomes.

The X chromosome, however, has only one active copy in mature cells which causes complications in gene dosage. In Paper III, we report that the genes on the single active X chromosome are upregulated compared to the genes on the autosomal chromosomes, and that this upregulation is achieved through an increased burst frequency. Furthermore, this upregulation is coupled to X chromosome inactivation in females.

To study transcriptional bursting at a more resolved time scale, we developed novel single cell sequencing methods using metabolic labeling in Paper IV and V. These methods supply the nucleotide analog 4-thiouridine to cells during cell culture, which become incorporated during transcription. Due to the alkylation reaction during library preparation leading to the incorporation of the wrong nucleotide during reverse transcription, the incorporated 4-thiouridine can be computationally detected as mismatches to the reference genome during analysis. We use this approach to study responses to a perturbation (Paper IV) and to study transcriptional bursting during a 2-hour time window (Paper V). This data allows the further dissection of transcriptional bursting and the ability to study co-bursting in single cells. We show that the synthesis rate mainly determines burst size and not the transcriptional off rate. We do not find co-bursting to be a general phenomenon across the transcriptome but do find certain gene pairs that exhibit co-bursting.