WhitePaper_RNASeq

An introduction to efficient planning and execution of RNA sequencing (RNA-Seq) experiments. RNA sequencing (RNA-Seq) refers to a method that is based on next genera - tion sequencing (NGS) technologies to study transcriptomes. While microar - ray-based technologies depend on measuring hybridization intensities to predesigned probes, RNA-Seq relies on discrete counting of sequenced mol - ecules. Thus, in contrast to microar - rays, RNA-Seq requires no prior knowl - edge about the genome and can be considered hypothesis free. In addi - tion, the counting of RNA molecules gives RNA-Seq experiments a much higher dynamic range compared to the hybridization intensities mea - sured in microarray experiments. Also considering the falling costs of NGS technologies, it is not surprising that RNA-Seq experiments have become the gold standard in transcriptome studies. Novel isoforms, alterna - tive splice sites, rare transcripts and gene-fusions, non-coding transcripts, and additional, even novel mecha - nism, can be detected all in a single experiment. Usually, the main goal of RNA-Seq studies is to obtain expres - sion profiles, pathways and gene net - works linked to the experimental con - dition studied. A generalized work?ow of a typical RNA-Seq study is depicted in Figure 1 . Given the complex orga - nization of genomes together with the huge amount of fragmented data, the interpretation of RNA-Seq experiments may appear a daunting task, especially for eukaryotic organ - isms [1]. For instance, the most recent human reference assembly – GRCh38 – has 244’550 unique exons with a mean length of 330 bases and a total amount of 80 million bases scattered across the 3 billion bases of the human genome. There are 2’879 annotated non-coding micro RNA (miRNA) [2], 52’000 transcripts from 26’475 genes, 22’302 associated gene ontology (GO) terms [3] and 330 KEGG pathways  [4] (as of August 2018). Thus, a good experimental strategy is important to reliably identify the desired genes and gene networks, for example the tran - scription targets of a stressor, or a spe - cific gene or pathway of interest after treatment, or gene-expression differ - ences between different genotypes. This white paper will give an overview on how to handle RNA-Seq data by presenting a selection of workflows. RNA Sequencing Figure 1. This figure depicts a generalized RNA sequencing workflow that may be used for differential expres - sion analysis. Motivation for RNA Sequencing Data AnalysisSamples Total RNA Isolation RNA Library Preparation Illumina Short Read Sequencing Report Generation Experimental Design Microsynth AG, SwitzerlandSch?tzenstrasse 15 ? P.O. Box ? CH - 9436 Balgach ? Phone + 41fi71fi722fi83 33 ? Fax + 41fi71fi722fi87 58 ? info @microsynth.ch ? www.microsynth.ch THE SWISS DNA COMPANY White Paper ? Next Generation Sequencing RNA Isolation Obtaining high quality RNA is critical. RNA degradation is detrimental to the experiments since it may introduce 3’ biases during polyadenylation (polyA) enrichment or may distort the tran - script profile by differentially affect - ing different RNAs. Thus, great care is needed to preserve the integrity of the input RNA. The acronym GIGO “garbage in, garbage out” holds true in this case as well. A notable excep - tion is RNA extracted from formalde - hyde-fixed paraffin-embedded (FFPE) tissue obtained by laser-assisted dis - section methods, where a certain amount of degradation is unavoid - able. Transcript profiles may also be distorted upon amplification of low amounts of input RNA, using either transcription-based or PCR-based amplification methods. However, with careful controls and a sufficient number of biological replicates, the adverse effects can be minimized. Sequencing Library Construction Depending on the desired RNA type (coding or non-coding) and type of organism studied (eukaryote or pro - karyote), different sequencing library types are constructed. For instance, to sequence mRNA, a polyA enrich - ment step is performed for eukary - otes, while a ribosomal RNA deple - tion step is carried out for prokaryotes. Kits for non-coding RNA library con - struction may use alternative tech - niques to enrich the relevant RNA fraction. The constructed libraries are stranded, meaning they retain the strand information of the sequenced molecule, which results in a more reli - able quantification of gene expression [7]. One typical method to keep the RNA strandness makes use of uracil instead of thymine for incorporation during second strand cDNA synthesis. After adapter ligation and before PCR amplification, uracil-DNA glycosylase (UNG) is added to degrade the second strand. As a result, all reads start in the same orientation, allowing the iden - tification of the transcribed strand. A schematic depiction of how total RNA is turned into a sequenceable Illumina cDNA library is shown in Figure 2 . Experimental Design For a successful experiment, many aspects, including experimental setup, sampling, and funding are to be con - sidered. In addition, the number of biological replicates and the number of reads produced for each replicate are essential parameters to produce valid results [5], especially to detect the maximal number of diff erentially expressed genes which includes rare transcripts. As gene expression anal - ysis builds on counting reads from the respective transcription unit, sin - gle-end reads of 75 bases length usually suffice for accurate mapping. However, paired-end sequencing (and in some cases longer reads, for instance as produced by Pacific Biosciences (PacBio) sequencing technologies) is required if highly accurate transcript quantification, determination of gene fusions or novel splice variant detec - tion is envisaged. In contrast to sin - gle-end sequencing, paired-end sequencing enables reading both ends of a (c)DNA fragment. Generally, it is recommended to work at least in trip - licates per experimental condition and sequence 30 million single-end reads per replicate for eukaryotic organisms and 10 million single-end reads for each replicate for prokaryotic organ - isms. For miRNA the read numbers may be halved. It is also worth men - tioning that the External RNA Controls Consortium (ERCC) has developed a set of external RNA controls designed to mimic natural eukaryotic mRNA sequences [6]. These sequences may be spiked in after RNA isolation and can be used to estimate the uncertainty in the subsequent measurements. Sequencing and Differential Expression Analysis of Coding and Non-coding RNA Selected Applications of RNA-Seq Transcriptome studies are well suited to understand disease mech - anisms, developmental mecha - nisms, or response to various stress - ors. Differential expression analysis of RNA-Seq data relies on the compari - son of data sets obtained from exper - imental conditions (e.g. drug treat - ments) and controls to determine the difference in transcript abun - dance. The focus here is on messen - ger RNA (mRNA). In addition, non- coding miRNAs, which often have gene regulatory purposes, may be used to develop biomarkers specific to a medical condition. Such differ - entially expressed miRNA, can then be experimentally verified to develop diagnostic qPCR kits for instance. Microsynth AG, SwitzerlandSch?tzenstrasse 15 ? P.O. Box ? CH - 9436 Balgach ? Phone + 41fi71fi722fi83 33 ? Fax + 41fi71fi722fi87 58 ? info @microsynth.ch ? www.microsynth.ch THE SWISS DNA COMPANY White Paper ? Next Generation Sequencing Bioinformatics Analysis As an example, let us consider an RNA-Seq experiment aimed at detect - ing changes in the gene expression profile of a human cell line after expo - sure to a drug. With control samples and additional samples taken at two different time points after drug appli - cation, each in three replicates, the total sample number would amount to nine. It is strictly recommended to not pool different replicates or condi - tions into a single sample for sequenc - ing as this would eliminate all statisti - cal power and the experiment would become useless. Assuming each sample in the outlined RNA-Seq exper - iment generates 30 million single-end, 75 bases long reads, the whole exper - iment produces 270 million reads or 20 billion bases. This large amount of data requires a dedicated analysis pipeline to extract meaningful infor - mation. Bioinformatics analysis of RNA-Seq data generally consists in: 1) quality control, optional size selection (e.g. to specifically separate non-cod - ing RNA fractions) and filtering of the sequenced reads, 2) splice-aware mapping of the reads to the refer - ence genome, 3) counting of uniquely mapped reads for each gene, 4) normal - ization of read counts across the exper - iment and 5) statistical evaluation of the normalized values comparing the different conditions (such as treatment and control) to each other to iden - tify significant fold changes and up- or down-regulation of the genes [10]. Table 1 presents an excerpt of a mRNA differential gene expression analy - sis. Figure 4 shows how expression values of replicates group differ from the values of the respective controls. Next Generation Sequencing Illumina short-read sequencing by syn - thesis (SBS) technology, as depicted in Figure 3 , is especially well suited for RNA-Seq, as it is fast, accurate and cost effective [8]. Sequenced reads are pro - duced in the standard fastq format [9] that incorporates both sequence infor - mation and quality scoring and can be further processed in downstream anal - yses. Figure 4. Principal Component Analysis (PCA) plot to visualize grouping of samples in an RNA-Seq experiment. The three conditions depicted are clearly separated, indicating significant, differential gene expression patterns of the three analyzed con - ditions. Flow Cell 5‘3‘ Flow Cell 5‘3‘ Flow Cell 5‘3‘ Flow Cell 5‘3‘ First strand sequencing Indexing Flipping and second strand sequencing Index read Sequence read 2 will be sense Sequence read 1 will be antisense Sense DNA Sense DNA Sense DNA Anti-sense DNA Figure 3: Schematic of Illuminas paired-end sequencing workflow. Figure 2. Schematic description of a poly-A enriched RNA Illumina library ready for sequencing. Image: David Corney. Microsynth AG, SwitzerlandSch?tzenstrasse 15 ? P.O. Box ? CH - 9436 Balgach ? Phone + 41fi71fi722fi83 33 ? Fax + 41fi71fi722fi87 58 ? info @microsynth.ch ? www.microsynth.ch THE SWISS DNA COMPANY White Paper ? Next Generation Sequencing Figure 6. A depiction of a significant de novo miRNA motif discovered in a miRNA Seq analysis. A miRNA motif is a region that is well conserved in many of the analyzed sequences. Figure 5. Excerpt from the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway graph “TRANSCRIPTIONAL MISREGULATION IN CANCER”, where colored nodes represent significantly up- or downregulated genes in the selected pathway. Based on the differential gene expres - sion results and depending on the content of gene information published in databases, gene set and pathway analysis may be carried out to illumi - nate the larger context of the involved metabolic processes as exemplified in Figure 5 . A useful additional anal - ysis in the case of miRNAs comprises a motif search to identify potential miRNA targets and to uncover addi - tional, novel miRNAs [11]. The results of such analyses may be submitted to public databases such as miRNet [12] for further network-based visual analy - sis. Figure 6 depicts such a motif iden - tified by a miRNA analysis. Table 1. This excerpt of a table shows the main output of a differential gene expression analysis. In this experiment two conditions with three replicates are compared to each other. The table lists from left to right the gene identifier, boxplots represent - ing expression level distributions of the replicates, the log 2 fold change of gene expression between condition 1 and condition 2, the propability value (p-value) of the log 2 fold change and the p-value adjusted for multiple testing.

           
       Repression of tumor supressors Inhibition of apoptosis Cell survival Apoptosis Inhibition of apoptosis Proliferation? < Neuroendocrine cancers> Neuroblastoma Amplifcation Carcinoid Mutation MYCN MEN1 MLL MAX SP1ZBTB17(TF) (TF) TF DNA DNA DNA MDM2 PTK2 TP53 COMMD3-BMI1 NTRK1 NGFR CDKN1B 2.4 0 2.3 fold change Microsynth AG, SwitzerlandSch?tzenstrasse 15 ? P.O. Box ? CH - 9436 Balgach ? Phone + 41fi71fi722fi83 33 ? Fax + 41fi71fi722fi87 58 ? info @microsynth.ch ? www.microsynth.ch THE SWISS DNA COMPANY White Paper ? Next Generation Sequencing Diferential gene expression in eukaryotes Analysis Method Experimental Setup Material and Resources Sample Preparation Sequencing Data Analysis Diferential gene expression in prokaryotes Non-coding RNA diferential expression analysis in eukaryotes Alternative splice-sites and gene-fusion detection (novel isoforms) De novo transcriptome assembly At least two conditions in replicates, no pooling of diferent conditions Pooling of diferent tissues, growth stadiums, etc. to capture the transcriptome in its entirety mRNA and availability of annotated reference genome mRNA and availability of annotated reference genome e.g. miRNA and availability of annotated non-coding RNA and reference genome mRNA and availability of annotated reference Genome mRNA, missing annotated reference Genome Total RNA isolation; stranded polyA enriched sequencing library Total RNA isolation; stranded ribo-depleted sequencing library Total RNA isolation; non-coding RNA enriched sequencing library Total RNA isolation; stranded polyA enriched sequencing library Total RNA isolation; normalized mRNA sequencing library 30 Mio single-end reads, 75 bp length 10 Mio single-end reads, 75 bp length 15 Mio single-end reads, 75 bp length 50 Mio paired-end reads, 2 x 150 bp length 20 Mio paired-end reads, 2 x 300 bp length Diferential gene expression analysis; pathway analysis if pathway database for the organism in question is available De novo transcriptome assembly and annotation Common Scientifc Question Study the transcriptome of a yet uncharted species Study cancer specifc mechanisms Develop biomarkers for specifc medical conditions Explore the infuence of a treatment on prokaryotic gene expression Explore the infuence of a treatment on eukaryotic gene expression At least two conditions in replicates, no pooling of diferent conditions At least two conditions in replicates, no pooling of diferent conditions Diferential gene expression analysis; pathway analysis if pathway database for the organism in question is available Diferential expression analysis; motif search Statistical appraisal of detected alternative splice sites and gene fusions At least two conditions in replicates, no pooling of diferent conditions Table 2. This table provides an overview of common scientifc questions in the field of RNA-Seq and gives a brief overview of the most important points that need to be considered in a RNA-Seq project. The table is intended as a quick reference guide. Summary Obviously, RNA-Seq is not limited to dealing with questions of differen - tial gene expression or identification of miRNA, which have been discussed in the previous sections. Table 2 lists common RNA-Seq applications. The table can serve as a guide for selecting an appropriate approach to a research question. Another application of RNA Seq technology is, for example, de novo transcriptome assembly and annota - tion, which is useful when no anno - tated reference genome is available. In short, RNA is collected from as many different stages and tissues as possi - ble. The entire RNA is then enriched for polyadenylated mRNA. The pool of mRNA, which ideally represents all transcribed genes, is then normal - ized to reduce abundant mRNAs and enrich rare mRNAs. The normalized transcripts are sequenced, then assem - bled in a second step and annotated with various databases in a third step, resulting in a ready-to-use de novo transcriptome [13]. RNA-Seq provides a snapshot of the transcriptome in cells and cell pop - ulations, making it a very attractive and powerful method. However, the results of the RNA-Seq experiments are complex because they produce a large amount of fragmented data. However, with the right approach, the challenge of extracting knowledge is reduced to a manageable task. Microsynth AG, SwitzerlandSchützenstrasse 15 · P.O. Box · CH - 9436 Balgach · Phone + 41f71f722f83 33 · Fax + 41f71f722f87 58 · info @microsynth.ch · www.microsynth.ch THE SWISS DNA COMPANY White Paper · Next Generation Sequencing References [1] Steven L. Salzberg. Open ques - tions: How many genes do we have? BMC Biology. 2018;16(94). doi:10.1186/ s12915-018-0564-x [2] Sam Griffiths-Jones, Russell J. Grocock, Stijn van Dongen, Alex Bateman, Anton J. Enright; miRBase: microRNA sequences, targets and gene nomenclature, Nucleic Acids Research, Volume 34, Issue suppl_1, 1 January 2006, Pages D140–D144, https://doi. org/10.1093/nar/gkj112 [3] The Gene Ontology Consortium; Expansion of the Gene Ontology knowl - edgebase and resources, Nucleic Acids Research, Volume 45, Issue D1, 4 January 2017, Pages D331–D338, https://doi. org/10.1093/nar/gkw1108 [4] Minoru Kanehisa, Miho Furumichi, Mao Tanabe, Yoko Sato, Kanae Morishima; KEGG: new perspectives on genomes, pathways, diseases and drugs, Nucleic Acids Research, Volume 45, Issue D1, 4 January 2017, Pages D353–D361, https://doi.org/10.1093/nar/gkw1092 [5] Schurch NJ, Schofield P, Gierli?ski M, et al. How many biological replicates are needed in an RNA-seq experiment and which differential expression tool should you use? RNA. 2016;22(6):839- 851. doi:10.1261/rna.053959.115 [6] Lemire A, Lea K, Batten D, et al. Development of ERCC RNA Spike-In Control Mixes. Journal of Biomolecular Techniques : JBT. 2011;22(Suppl):S46 [7] Zhao S, Zhang Y, Gordon W, et al. Comparison of stranded and non- stranded RNA-seq transcriptome profil - ing and investigation of gene overlap. BMC Genomics. 2015;16(1):675. doi:10.1186/s12864-015-1876-7 [8] Online at: https://emea.illumina. com/systems/sequencing-platforms/ nextseq/applications.html?langsel=/ ch/, accessed 14.09.2018 [9] Online at: http://maq.sourceforge. net/fastq.shtml, accessed 14.09.2018 [10] Sandrine Borgeaud, Lisa C. Metzger, Tiziana Scrignari, Melanie Blokesch, The type VI secretion system of Vibrio chol - erae fosters horizontal gene transfer, Science 02 Jan 2015: Vol. 347, Issue 6217, pp. 63-67. DOI: 10.1126/science.1260064 [11] Bhupesh K. Prusty, Nitish Gulve, Suvagata Roy Chowdhury, Michael Schuster, Sebastian Strempel, Vincent Descamps, Thomas Rudel. HHV-6 encoded small non-coding RNAs define an intermediate and early stage in viral reactivation. npj Genomic Medicine. 2018;3(25). 10.1038/s41525-018-0064-5 [12] Yannan Fan, Keith Siklenka, Simran K. Arora, Paula Ribeiro, Sarah Kimmins, Jianguo Xia; miRNet - dissecting miR - NA-target interactions and functional associations through network-based visual analysis, Nucleic Acids Research, Volume 44, Issue W1, 8 July 2016, Pages W135–W141, https://doi.org/10.1093/ nar/gkw288 [13] Neves, R.C., Guimaraes, J.C., Strempel, S. et al., Transcriptome pro - filing of Symbion pandora (phylum Cycliophora): insights from a differential gene expression analysis, Org Divers Evol (2017) 17: 111. https://doi.org/10.1007/ s13127-016-0315-1 Microsynth AG, SwitzerlandSch?tzenstrasse 15 ? P.O. Box ? CH - 9436 Balgach ? Phone + 417172283 33 ? Fax + 417172287 58 ? info @microsynth.ch ? www.microsynth.ch THE SWISS DNA COMPANY White Paper ? Next Generation Sequencing