Gene expression profiling is a key tool used to study cellular processes and the associated complex interaction network. Studies on mRNA abundance have typically been the method of choice to obtain basic insights into the underlying molecular mechanisms. The development of whole-transcriptome microarrays 1 and, more recently, next-generation sequencing of RNA (RNA-seq) 2-4 fueled this approach. While these technologies have revolutionized our understanding of the complexity of cellular gene expression, they face major limitations due to intrinsic properties of their template sample, i.e. total cellular RNA. First, short-term changes in total RNA levels do not match changes in transcription rates, but are inherently dependent on the RNA half-life of the respective transcripts. While a fivefold induction of a short-lived transcript, e.g. encoding for a transcription factor, will be readily detectable in total RNA within an hour, the same induction of a long-lived transcript, e.g. encoding for a metabolic enzyme, will remain virtually invisible. In addition, even a complete shut-down (>1,000-fold down-regulation) in the transcription rate of an average gene with an RNA half-life of five hours will simply take five hours for its total RNA levels to decrease by only twofold. Therefore, analysis of total RNA favors the detection of up-regulation of short-lived transcripts, many of which encode for transcription factors and genes with regulatory functions 5. In addition, the true kinetic cascade of regulation is obscured and primary signaling events cannot be differentiated from secondary. Both, in turn, may result in substantial bias in downstream bioinformatics analyses. Second, alterations in total RNA levels cannot be attributed to changes in RNA synthesis or decay. Measurements of the latter require cell invasive approaches, e.g. blocking transcription using actinomycin D 6, and extended monitoring of ongoing RNA decay over time. With a mean mRNA half-life in mammalian cells of 5 - 10 hr 5,7, mRNA levels of most genes will only have decreased by less than twofold following several hours of transcriptional arrest. These rather small differences result in grossly imprecise measurements of mRNA half-lives for the majority of cellular genes due to the exponential nature of the underlying mathematical equations. Finally, while RNA-seq of total cellular RNA revealed that approximately half of our genes are subject to alternative splicing events8, the underlying kinetics as well as the dynamic mechanisms guiding tissue- and context-specific regulation of RNA processing remain poorly understood. In addition, the contribution of RNA processing to differential gene expression, particularly for non-coding RNAs, remains to be determined. Altogether, these limitations represent major obstacles for bioinformatic kinetic modeling of the underlying molecular mechanisms.
We recently developed an approach, termed high resolution gene expression profiling, to overcome these problems 5,7,9. It is based on metabolic labeling of newly transcribed RNA using 4-thiouridine (4sU-tagging), a naturally occurring uridine derivative, and provides direct access to newly transcribed transcripts with minimal interference in cell growth and gene expression (see Figure 1) 5,10-12. Exposure of eukaryotic cells to 4sU results in its rapid uptake, phosphorylation to 4sU-triphosphate, and incorporation into newly transcribed RNA. Following isolation of total cellular RNA, the 4sU-labeled RNA fraction is thiol-specifically biotinylated generating a disulfide bond between biotin and the newly transcribed RNA. 'Total cellular RNA' can then be quantitatively separated into labeled ('newly transcribed') and unlabeled ('pre-existing') RNA with high purity using streptavidin-coated magnetic beads. Finally, labeled RNA is recovered from the beads by simply adding a reducing agent (e.g. dithiothreitol) cleaving the disulfide bond and releasing the newly transcribed RNA from the beads.
Newly transcribed RNA depicts the transcriptional activity of every gene during the timeframe of 4sU exposure. 4sU-tagging in the timescale of minutes thus provides a snapshot picture of eukaryotic gene expression and an ideal template for down-stream bioinformatic analyses (e.g. promoter analysis). In cases where steady-state conditions can be assumed, the ratios of newly transcribed/total, newly transcribed/unlabeled and unlabeled/total RNA provide non-invasive access to precise RNA half-lives 7,13. In addition, it is important to note that newly transcribed RNA purified after as little as 5 min of 4sU-tagging (5 min 4sU-RNA) is younger than 15 and 60 min 4sU-RNA. When performing both ultra-short and progressively longer 4sU-tagging in a single experimental setting combined with RNA-seq, the kinetics of RNA processing are revealed at nucleotide resolution 9. Finally, time-course analyses of newly transcribed and total RNA combined with computational modeling allow an integrative analysis of RNA synthesis and decay 14.
In conclusion, this approach allows for the direct analysis of the dynamics of RNA synthesis, processing, and degradation in eukaryotic cells. It is applicable in all major model organisms including mammals, insects (Drosophila), amphibians (Xenopus), and yeast 5,15,16. It is directly compatible with microarray analysis 5,17, RNA-seq 9,13,14, and is applicable in vivo12,15. Here, we detail the methodology to label, isolate, and purify newly transcribed RNA in cultured mammalian cells. In addition, potential problems and pitfalls are discussed.