Nascent transcript sequencing visualizes transcription at nucleotide resolution

Abstract

Recent studies of transcription have revealed a level of complexity not previously appreciated even a few years ago, both in the intricate use of post-initiation control and the mass production of rapidly degraded transcripts. Dissection of these pathways requires strategies for precisely following transcripts as they are being produced. Here we present an approach (native elongating transcript sequencing, NET-seq), based on deep sequencing of 3' ends of nascent transcripts associated with RNA polymerase, to monitor transcription at nucleotide resolution. Application of NET-seq in Saccharomyces cerevisiae reveals that although promoters are generally capable of divergent transcription, the Rpd3S deacetylation complex enforces strong directionality to most promoters by suppressing antisense transcript initiation. Our studies also reveal pervasive polymerase pausing and backtracking throughout the body of transcripts. Average pause density shows prominent peaks at each of the first four nucleosomes, with the peak location occurring in good agreement with in vitro biophysical measurements. Thus, nucleosome-induced pausing represents a major barrier to transcriptional elongation in vivo.

Figure 1

a, Schematic diagram of NET-seq protocol. A yeast culture is flash frozen and cryogenically lysed. Nascent RNA is co-purified via an immunoprecipitation (IP) of the RNAPII elongation complex. Conversion of RNA into DNA results in a DNA library with the RNA as an insert between DNA sequencing linkers. The sequencing primer is positioned such that the 3′ end of the insert is sequenced. mG refers to the 7-methylguanosine cap structure at the 5′ end of nascent transcripts.b, The 3′ end of each sequence is mapped to the yeast genome and the number of reads at each nucleotide is plotted at the RPL30 locus for nascent RNA and lightly fragmented mature RNA. Note that for the nascent transcripts, the introns (grey box) and regions after the polyadenylation site (black arrow) are readily detected. c, Metagene analysis for well-expressed genes (n = 471, >1.5 reads per bp in both conditions) of the mean read density (arbitrary units, a.u.) in the presence and absence of transcription inhibitor, α-amanitin. TSS, transcription start site.

Figure 2

a, Nascent and mature transcripts initiating from URA1 and RPL5 promoters in the sense and antisense directions. Note that there are cryptic unstable transcripts (CUTs) in the antisense direction for URA1 but not RPL5. b, A histogram of the transcription ratio (antisense/sense transcription levels) for 1,875 genes. The green and grey boxes indicate the subset of genes with a ratio of less than 1:8 and less than 1:3, respectively. c, Antisense transcription levels are plotted versus sense transcription for each tandem gene (Spearman correlation coefficient, r_s = 0.34). d, The level of antisense transcription for each promoter is plotted versus the local enrichment for H4 hyperacetylation using available data (r_s = 0.65).

Figure 3

a, Examples of cryptic unstable transcripts (CUTs, orange data) upstream and antisense of DBF2,DRN1 and VAS1 promoters. The fold increase of CUT transcription in the rco1δ strain is marked. b, The transcription ratio (antisense/sense) in the rco1δ strain is plotted against the transcription ratio in the wild-type strain for each gene. c, A metagene analysis of well-expressed antisense transcription (n = 171, >1 read per bp).

Figure 4

a, NET-seq data at the GPM1 gene for biological replicates. b, A histogram of the mean distance between pauses for each well-expressed gene (n = 1,006, >2 reads per bp). c, The consensus sequence of the DNA coding strand surrounding pause sites found from all genes.

Figure 5

a, A schematic describing an existing model for how RNAPII pauses at an obstacle (red square), backtracks and is induced to cleave its transcript through binding to Dst1 (refs 32, 33). b, A comparison of NET-seq data for wild-type and dst1δ strains at the GPM1 gene. c, Mean cross-correlation between the dst1δ and wild-type data of well transcribed genes (n = 770, >2 reads per bp) (green line) was calculated by determining the Pearson’s correlation coefficient at each gene between fixed dst1δ data and shifted wild-type data followed by averaging over all genes. This analysis is compared to the mean autocorrelation of the wild-type data for well transcribed genes (blue line). d, The consensus sequence for all pauses observed in the dst1δ strain.

Figure 6

Plot of mean pause densities in dst1δ data relative to the first four nucleosomes after the transcription start site using available nucleosome positioning data. Error bars represent one standard deviation.