EXPRSS: an Illumina based high-throughput expression-profiling method to reveal transcriptional dynamics

Abstract

Background: Next Generation Sequencing technologies have facilitated differential gene expression analysis through RNA-seq and Tag-seq methods. RNA-seq has biases associated with transcript lengths, lacks uniform coverage of regions in mRNA and requires 10-20 times more reads than a typical Tag-seq. Most existing Tag-seq methods either have biases or not high throughput due to use of restriction enzymes or enzymatic manipulation of 5' ends of mRNA or use of RNA ligations.

Results: We have developed EXpression Profiling through Randomly Sheared cDNA tag Sequencing (EXPRSS) that employs acoustic waves to randomly shear cDNA and generate sequence tags at a relatively defined position (~150-200 bp) from the 3' end of each mRNA. Implementation of the method was verified through comparative analysis of expression data generated from EXPRSS, NlaIII-DGE and Affymetrix microarray and through qPCR quantification of selected genes. EXPRSS is a strand specific and restriction enzyme independent tag sequencing method that does not require cDNA length-based data transformations. EXPRSS is highly reproducible, is high-throughput and it also reveals alternative polyadenylation and polyadenylated antisense transcripts. It is cost-effective using barcoded multiplexing, avoids the biases of existing SAGE and derivative methods and can reveal polyadenylation position from paired-end sequencing.

Conclusions: EXPRSS Tag-seq provides sensitive and reliable gene expression data and enables high-throughput expression profiling with relatively simple downstream analysis.

Figure 1

Schematic diagram representing EXPRSS Tag-seq. (A) Library preparation (B) Data analysis pipeline.

Figure 2

Characteristics of EXPRSS tag-sequences. (A) Uniquely aligned tags to the sense strand of cDNA and genome sequences from all Arabidopsis genes are used to plot tag alignment position as a distance from 3′ end of annotated genes against the frequency of reads mapped. Rectangle selection is shown as inset picture. (B) Example of alternative polyadenylation. A distinct cluster of tags from both full length (blue circle) and short (red circle) sense transcripts of FCA (AT4G16280), which are about 6.5 kb apart. Evidence from cDNA sequences (green arrows) in the TAIR database corroborates short transcripts. Reads from such alternative polyadenylation transcripts result in a long tail in tag alignment frequency distribution presented in A. (C) Frequency distribution of antisense tags aligning to 10 selected transcripts plotted against mapping position as a distance from 5′ end of annotated genes. Each individual abundantly expressing an antisense transcript has a distinct peak of tag alignment suggesting a defined polyadenylation site for antisense transcripts identified through EXPRSS method.

Figure 3

Higher reproducibility and better dynamics of differential expression detection using EXPRSS. (A-C) Pairwise scatter plots of gene counts from treatment replicates, expressed as tags per million in log10 scale. Correlations between four independent technical replicates of EXPRSS Tag-seq (A), made from same RNA sample; four independent biological replicates of EXPRSS Tag-seq (B) and NlaIII-DGE (C) are presented. Pearson correlation of log10 transformed tag counts per million plus 1 is depicted at left hand top corner of each comparison. Right hand bottom corner indicates replicate number depicted on X and Y-axis, respectively. (D-E) Example of differential expression using reads aligned (small black arrows) to WRKY22, a flg22 responsive gene are presented from control (D) and treatment (E).

Figure 4

Comparison of differential expression between EXPRSS, Nla III-DGE and Affymetrix ATH1 array. (A-B) Venn diagrams showing overlap of differential expression identified from the same RNA samples (EXPRSS and NlaIII-DGE) and similar RNA sample (ATH1 microarray). Overlap of (A) sense transcripts (numbers in black - genes spotted on ATH1 array and numbers in red- genes not spotted on ATH1 array) (B) antisense transcripts identified by EXPRSS and NlaIII-DGE. (C-D) Q-PCR confirmation of differential expression observed through EXPRSS. Three up-regulated genes (C) and five down-regulated genes (D) that are found differentially expressed with EXPRSS are verified with QPCR. Error bars indicate standard deviation from three biological replicates. FDR values are provided for EXPRSS log2 fold changes.

Figure 5

Sensitivity of EXPRSS in differential expression detection for specific responses to treatment. (A-C) Venn diagram representing overlap of differential expression for flg22 treatment from four different experiments (Microarray on seedling, leaf, leaf disc and EXPRSS on leaf disc) (A) Up-regulated genes between the four methods are compared. (B) Down-regulated genes between the four methods are compared. (C) Depiction of the overlap among the sections of Venn diagram. Number under each experiment represents number of genes differentially expressed. (Numbers in black - genes spotted on ATH1 array and numbers in red- genes not spotted on ATH1 array) (D-E) Scatter plots showing fold change distribution for commonly detected genes between two experiments. (D) Comparison of log2 fold changes between leaf and seedling microarray shows more restricted dynamics on negative scale (~ − 3) than positive scale (~ + 7). (E) Similar comparison of log2 fold changes between leaf disc data from EXPRSS and NlaIII-DGE showing more even distribution (−6 to +7).