MAPCap allows high-resolution detection and differential expression analysis of transcription start sites

Abstract

The position, shape and number of transcription start sites (TSS) are critical determinants of gene regulation. Most methods developed to detect TSSs and study promoter usage are, however, of limited use in studies that demand quantification of expression changes between two or more groups. In this study, we combine high-resolution detection of transcription start sites and differential expression analysis using a simplified TSS quantification protocol, MAPCap (Multiplexed Affinity Purification of Capped RNA) along with the software icetea . Applying MAPCap on developing Drosophila melanogaster embryos and larvae, we detected stage and sex-specific promoter and enhancer activity and quantify the effect of mutants of maleless (MLE) helicase at X-chromosomal promoters. We observe that MLE mutation leads to a median 1.9 fold drop in expression of X-chromosome promoters and affects the expression of several TSSs with a sexually dimorphic expression on autosomes. Our results provide quantitative insights into promoter activity during dosage compensation.

Fig. 1

MAPCap (Multiplexed Affinity Purification of Capped RNA) efficiently captures transcript 5’-end signal. a Overview of the MAPCap protocol. After fragmentation and ribo-depletion, the Capped RNA is immunoprecipitated using an antibody, and the s-oligos are attached, afterwards the samples are pooled for PCR and library preparation steps. b Nucleotide content in read positions. CAGE and RAMPAGE show a high artificial G-bias due to their cloning steps, while MAPCap shows low bias for any specific nucleotide. c, d Correlation of signal (log2(counts per million + 1)) between MAPCap and CAGE and between MAPCap and RNA-seq on 5’-untranscribed region of genes. e Genome snapshot of the de-duplicated counts from MAPCap, RAMPAGE and CAGE on transcription start site. For MAPCap and RAMPAGE, the de-duplication was performed using 5’-position of the reads and the UMIs, while for CAGE it was performed using only 5’-position. RNA-seq track is shown for comparison. f Metagene profile comparing signal enrichment between nAnTiCAGE and MAPCap on all genes in mouse embryonic stem cells (CPM = counts per million). g Metagene profile of signal from the MAPCap experiment performed in S2 cells, using different quantity of total RNA (5 μg, 1 μg, 500 ng, 100 ng) as starting material. h Added relative concentration of capped spike-ins ((amount of spike-in RNA/amount of total RNA) × 100) vs recovered relative counts ((reads mapped to spike-ins/total mapped reads) × 100) for the embryos. The samples were added with proportionally increasing relative concentration of ERCCs

Fig. 2

Replicate-based analysis improves accuracy of transcription start site (TSS) detection. a Evaluation of true and false positives on MAPCap (Multiplexed Affinity Purification of Capped RNA) data of embryos. TSS was detected by paraclu using individual replicates (embryos 1–4), pooled replicates (embryo_pooled), or individual replicates followed by intersection (embryo_intersect) and compared with the “local enrichment” method (replicate based) that uses all four samples. The detection accuracy improves using this method. b Precision-recall curve (PRC) of subsampled MAPCap data, comparing paraclu (embryos 1–4, pooled and intersect) with replicate-based method (AUPRC = area under PRC). c Comparative Gene Ontology enrichment of “sharp” (upto 20 bp) and “broad” (>20 bp) TSSs detected by the local enrichment method. d MAPCap signal (counts per million), as well as motif presence (detected by FIMO) for the “sharp” and “broad” TSS. e, f Genome snapshot of MAPCap signal and the detected TSS using the “local enrichment” method on a developmental gene (hunchback, hb) and a housekeeping gene (proteosomal subunit, Prosbeta5). g TSS detected on validated enhancer regions (eRNAs) in embryo, L3 larvae brains (MAPCap), and adult heads (modENCODE CAGE). h Overlap of eRNAs detected between male and female larvae brains and adult heads show that most eRNAs are stage-specific (yellow) rather than sex-specific (blue)

Fig. 3

Analysis of differential promoter activity in the male and female larvae brains. a Background: The dosage compensation complex (DCC) contains an RNA helicase MLE (maleless) that incorporates roX RNAs into the complex. DCC first targets certain high-affinity sites (HAS) on the X chromosome, followed by the spread to the low affinity sites (LAS). This spread is guided by three-dimensional conformation of the X chromosome and requires the recognition of roX RNAs by MLE. b Experimental set-up: We extracted brains from L3 larvae of flies that are mutant (leading to Knock-out) of MLE and compared them to wild-type flies. c Comparison of gene-level expression estimates (log-CPM) obtained from MAPCap (Multiplexed Affinity Purification of Capped RNA) and ribo-depleted RNA-seq in wild-type males (see Supplementary Fig. 3c for females). Expression of most genes are similar except for small nucleolar RNAs and small nuclear RNAs (exclusively depleted in MAPCap). d Barcode plot of RNA-seq differential expression results over t statistics obtained from MAPCap. The vertical red and blue bars mark differentially expressed (DE) gene in RNA-seq at false discovery rate (FDR) < 0.05 and the lines above and below the bars represent their enrichment in MAPCap data. e Promoter usage of roX1 between sexes and stages. f MA plot of DE TSSs obtained from MAPCap after spike-in normalization. X chromosome promoters are marked in blue while autosomes are marked in red. g Volcano plot of differential expression estimates on TSSs from MAPCap data in female (top) and male (bottom) MLE mutants. TSS with female-specific activity (at FDR < 0.1) are downregulated in female mutants (marked in red) while TSSs with male-specific activity (marked in blue) are downregulated in male mutants

Fig. 4

MLE (maleless) sensitivity of X chromosome promoters. a Histogram of fold-changes in transcription start site activity in male knockouts (KOs) compared to wild type, for genes on X (blue) and autosomes (grey). b Effect of KO on the expression of eRNAs (n = 63), lncRNAs (n = 16) and protein-coding genes (n = 1721) on the X chromosome (boxplots = median and IQR, dashed line = 2-fold). c Mean expression (replicates) of MLE sensitive (605) and insensitive (1047) X-chromosomal promoters. d Distance to high-affinity sites (HAS) for a subset of MLE sensitive (377) and insensitive (136) X-chromosomal promoters (in bp). The subset was selected such that mean expression difference between the two groups become insignificant. e Distance to topological domain (TAD) boundary for the same subset of promoters (in bp). f A summary of the functionalities of the “icetea” bioconductor package. Outputs of three functions: plotReadStats, annotateTSS, and detectDiffTSS, are shown (left to right)