PAT-seq: a method to study the integration of 3′-UTR dynamics with gene expression in the eukaryotic transcriptome

Abstract

A major objective of systems biology is to quantitatively integrate multiple parameters from genome-wide measurements. To integrate gene expression with dynamics in poly(A) tail length and adenylation site, we developed a targeted next-generation sequencing approach, Poly(A)-Test RNA-sequencing. PAT-seq returns (i) digital gene expression, (ii) polyadenylation site/s, and (iii) the polyadenylation-state within and between eukaryotic transcriptomes. PAT-seq differs from previous 3 ′ focused RNAseq methods in that it depends strictly on 3 ′ adenylation within total RNA samples and that the full-native poly(A) tail is included in the sequencing libraries. Here, total RNA samples from budding yeast cells were analyzed to identify the intersect between adenylation state and gene expression in response to loss of the major cytoplasmic deadenylase Ccr4. Furthermore, concordant changes to gene expression and adenylation-state were demonstrated in the classic Crabtree-Warburg metabolic shift. Because all polyadenylated RNA is interrogated by the approach, alternative adenylation sites, noncoding RNA and RNAdecay intermediates were also identified. Most important, the PAT-seq approach uses standard sequencing procedures, supports significant multiplexing, and thus replication and rigorous statistical analyses can for the first time be brought to the measure of 3 ′ -UTR dynamics genome wide.

Coordination of Cell Cycle Progression and Mitotic Spindle Assembly Involves Histone H3 Lysine 4 Methylation by Set1/COMPASS

“…For RNA-sequencing (RNA-seq) analysis, 1 mg of total RNA was processed for 39-end focused RNA-seq by the poly(A) tail sequencing (PAT-seq) approach (Harrison et al 2015). To identify genes differentially regulated between WT and Dset1, the equivalent of paired samples t-test, with samples paired by time, was performed.…”

“…If there was a difference between strains that changes over time, this was viewed as noise and reduced the significance of the result. Read counts were transformed and weighted using limma's voom function, then they were tested using Fitnoise as described in Harrison et al (2015). Like limma, Fitnoise uses an empirical Bayes method to moderate the estimate of residual variance for each gene based on residual variance observed in other genes, increasing the statistical power.…”

“…Because of that, we synchronized WT and Dset1 strains through alpha-factor block and analyzed gene expression changes following release into the cell cycle using PAT-seq, a custom RNA-seq approach ( Figure 3A; Harrison et al 2015). This identified transcripts that deviated between both strains with high significance over the 2-hr duration of the experiment (P-value ,0.05; File S1).…”

Nuclear poly(A) tail size is regulated by Cnot1 during the serum response

“…In addition, the poly(A) tail profiles suggest that these tails are not gradually removed during their lifetimes. Indeed, it has been shown, by us and others, that many stable housekeeping mRNAs accumulate with relatively short poly(A) tails, but this has been attributed to cytoplasmic deadenylation of longer initial poly(A) tails [16][17][18][19]50]. Together, these data suggest that constitutively expressed mRNAs have a distinct poly(A) metabolism from serum response mRNAs and may lack delayed decay.…”

“…These sizes suggest that the constitutively expressed mRNAs can bind two copies of the cytoplasmic poly(A) binding proteins, while serum response mRNAs generated after the transient induction can only bind one poly(A) binding protein [36,37]. A high throughput determination of poly(A) tail sizes in nucleus and cytoplasm or on briefly metabolically labelled and total RNA should be able to establish definitively how common short initial poly(A) tails are and how many different classes of initial poly(A) tail size can be distinguished [17,18,50].…”