- Comparison of RNA isolation methods on RNA-Seq: implications for differential. - In this study, we examined the effects of RNA isolation method as a possible source of batch effects in RNA- seq design.. - Results: Based on the different chemistries of “ classic ” hot phenol extraction of RNA compared to common commercial RNA isolation kits, we hypothesized that specific mRNAs may be preferentially extracted depending upon method, which could masquerade as differential expression in downstream RNA-seq analyses. - Comparing technical replicates that only differed in RNA isolation method, we found over one thousand transcripts that appeared “ differentially ” expressed when comparing hot phenol extraction with the two kits. - Strikingly, transcripts with higher abundance in the phenol-extracted samples were enriched for membrane proteins, suggesting that indeed the chemistry of hot phenol extraction better solubilizes those species of mRNA.. - control versus treatment), the method of RNA isolation had little effect on the ability to identify differentially expressed transcripts. - However, we suggest that researchers performing meta-analyses across different experimental batches strongly consider the RNA isolation methods for each experiment.. - The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. - Even in the case where the batch effect is not a complete confounder, account- ing for batch can reduce our power to detect true bio- logical signal [5]. - In this study, we examined the effects of RNA isolation method as a possible source of batch effects in RNA-seq design. - Newly synthesized pre-mRNAs are processed in the nucleus before being exported. - Despite the widespread acknowledgement that mRNAs are differentially localized within the cell, there has been a paucity of studies examining whether “common” RNA extraction methods are equivalent in their abilities to ex- tract differentially localized RNA species, and whether the method of RNA isolation affects our ability to detect differentially expressed transcripts. - Sultan and colleagues compared two RNA isolation methods (Qiagen RNeasy kit and guanidinium-phenol (TRIzol) extraction) and two library selection schemes (poly-A enrichment and rRNA depletion) on downstream transcript abundance estimates, and found that rRNA depletion was particu- larly sensitive to the RNA extraction method [2]. - Thus, we sought to systematically examine whether three common RNA isolation methods led to differences in transcript abundance and/or our ability to detect dif- ferential expression between two experimental condi- tions in the form of the Saccharomyces cerevisiae heat shock response. - The different RNA isolation methods were the classic “hot acid phenol” method, and the two most commonly-used types of kits [7]—a silica-based column kit (Qiagen RNeasy Kit) and a guanidinium- phenol (TRIzol)-based kit (Zymo Research Direct-zol), hereafter referred to as the Phenol, RNeasy, and Direct-. - To test this hypothesis, and whether the choice of RNA isolation method had downstream effects on our ability to detect differentially expression transcripts, we collected four biological replicates of the model yeast Saccharomy- ces cerevisiae before and after a 20-min heat shock. - Im- portantly, each biological sample was split into three identical technical replicates that differed only in their mode of RNA isolation. - This allowed us to systematically test whether the RNA isolation method affects relative transcript abundance between technical replicates, and whether that matters for differential expression analysis.. - Based on these results, we strongly recommend that meta-analyses be performed on groups of experiments with common RNA isolation methods.. - All of the RNA isolation methods yielded generally high quality RNA, as defined by a RIN of 9.0 or above, though the phenol extracted RNA averaged significantly higher RIN values than those isolated from the Direct-zol kit (9.96 vs. - There were larger differences in the percentage of uniquely mapped reads across RNA isola- tion methods (Supplementary Table 2). - We were particular interested in whether differences in the RNA isolation method could masquerade as “dif- ferential” expression due to differences in transcript quantification. - The second principal component corresponded to RNA isolation method and explained 26.9% of the variation. - To visualize differences in transcript abundance across RNA isolation methods, we performed hierarchical clus- tering on the TPMs of the unstressed samples (Fig. - To quantify these differ- ences, we used edgeR to identify transcripts with signifi- cantly differential abundance in pairwise comparisons of each RNA isolation method (FDR <. - Pairwise comparisons of the Phenol method with each Kit method identified a large number of transcripts with differential abundance: 2430 transcripts (Phenol vs.. - 0.01) in any pairwise comparison of RNA isolation method (Fig. - We found striking functional gene ontology (GO) en- richments for transcripts with higher or lower abun- dance in the phenol-extracted samples compared to both kits. - Looking more closely at the cellular component GO enrichments, transcripts with higher abundance in the phenol samples were strongly enriched for those encoding intrinsic membrane proteins (P <. - In contrast, mRNAs with lower relative abundance in the phenol samples were enriched for nuclear in localization (P <. - “repressed” in the Phenol samples compared to both Kits also had significantly lower expression relative to the genomic average (60.7 TPMs vs. - That Phenol-isolated samples have higher transcript abundance for mRNAs encoding membrane proteins fits with the hypothesis that the Phenol method better solu- bilizes that species of mRNA. - We did find a significant difference in the Phenol vs. - Direct- zol comparison, but not for the Phenol vs. - This turned out to be correct—we identified 788 “differ- entially” expressed genes in the Phenol vs. - Because the RNeasy-isolated samples had relatively high RIN values relative to the Direct-zol- isolated samples, the vast majority of transcripts with differential expression were retained as significant when accounting for RIN in the edgeR QL model . - 2 Principal component analysis (PCA) strongly implicates RNA isolation method as a batch effect. - Kit samples were more similar to each other than they were to the Phenol samples. - eliminated for the Phenol vs. - However, the surviving differentially expressed transcripts with higher expression in the Phenol-isolated samples relative to the Direct-zol iso- lated samples were still strongly enriched for those en- coding intrinsic membrane proteins (P <. - Moreover, because of the substantial overlap between genes called as differentially expressed in the Phenol vs.. - Direct-zol comparisons, we hypothesize that the differing chemistries in the. - Differences in RNA isolation method have little effect on the ability to detect differential expression with a batch The striking differences in transcript abundance depend- ing on RNA isolation could conceivably affect the ability to detect differential expression. - We identified ~ 3800 differentially expressed transcripts for all three RNA isolation methods, with substantial overlap for all three (Fig. - Clustering on relative transcript abundance (TPMs) reveals differences depending upon RNA isolation method, while clustering on sample identity shows that the Phenol method diverges from both Kits. - 0.01) in any pairwise comparisons between each RNA isolation method. - Phenol in the P v. - RNeasy in the P v. - 0.01) in the Phenol v.. - We also detected zero transcripts that had significant fold change differences in their heat shock response in any pairwise comparison between RNA isolation methods (Supplementary File 2).. - One possibility is that the transcripts encoding the processes most affected by differences in the extraction methods (i.e. - membrane-associated proteins) are lowly represented during the heat shock response, and thus the method of RNA isolation could affect differential expression under different conditions. - Thus, we hypothesize that at sufficient sequencing depth, the abil- ity to detect differential expression is robust to the mod- est differences in transcript counts caused by differences in RNA isolation method.. - In this study, we tested whether differences in RNA iso- lation method affect relative transcript abundance be- tween samples, and whether the RNA isolation method impacts our ability to detect differential expression. - Our results suggest that differences in RNA isolation method can substantially affect relative transcript abundance, and we saw thousands of differences in transcript abun- dance when comparing hot acid phenol extraction with an RNeasy or Direct-zol kit. - That transcripts with higher abundance in the Phenol-isolated samples are strongly enriched for encoding membrane proteins suggests the Phenol method better solubilizes those mRNAs. - First, while we do see differences in RIN values across the different RNA isolation methods, the differences are relatively small, and our RIN values are all much higher than the points where other studies identified them as confounding RNA-seq analysis [3, 12].. - The right portion shows differences in abundance in pairwise comparisons between each RNA isolation method, with brown indicating higher expression than the comparison group, and violet indicating lower expression than the comparison group. - The Venn Diagram depicts overlap between differentially expressed genes in the Phenol, RNeasy, and Direct-zol isolated samples. - Only one of the Phenol vs. - And while transcripts with higher relative abundance in the phenol-extracted samples versus the kits had higher GC content and gene length, which both correlate with higher in vivo degrad- ation rates [3], the correlation between those parameters and fold-change differences was not strong (Supplemen- tary Table 3). - The method of RNA isolation had little effect on the ability to identify differentially expressed tran- scripts in our heat shock test case. - Thus, experiments within a single lab are unlikely to be affected by the choice of RNA isolation method as long as the same method is used throughout an experiment. - For meta- analyses however, we recommend that researchers avoid comparing experiments where the RNA isolation methods differ.. - To compare RNA isolation methods, we collected three identical 10-ml ‘technical’ replicates for each biological replicate (4 biological replicates in total). - RNA isolation methods Hot phenol isolation. - Cells were lysed and RNA was isolated using a standard hot phenol method as described [17], and a detailed proto- col can be found on the protocols.io repository under DOI dx.doi.org/10.17504/protocols.io.inwcdfe. - The phenol extracted RNA was then ‘cleaned’ using an RNeasy Miniprep Kit with optional on-column DNase treatment according to the manufacturer’s instructions.. - RNA isolation with two different Miniprep kits. - through an automated Eppendorf epMotion 5075 liquid handling robot, and a detailed a protocol can be found on protocols.io under DOI dx.doi.org/10.17504/proto- cols.io.uueewte. - To account for differences in RIN across sam- ples, we also performed a separate analysis that included sample type, replicate, and RIN as factors in the model.. - Principal component analysis (PCA) was performed using ClustVis [21] on ln-transformed TPM values for all transcripts included in the differential expression ana- lysis, using unit variance scaling and singular value de- composition. - Six genes with either significantly higher (LAS17, SED1, PRY3) or lower (JNM1, EAF7, RRP36) abundance in the Phenol vs. - A detailed protocol is available on protocols.io under DOI dx.doi.org/10.17504/. - Supplementary information accompanies this paper at https://doi.org/10.. - Properties of transcripts with differential abundance depending upon RNA isolation method.. - The analyses generated during this study are included in the supplementary in- formation files.. - https://doi.org/10.1038/nbt.3000 PMID:. - https://doi.org PMID: 25113896.. - https://doi.. - https://doi.org/10.12688/. - https://doi.org/. - https://doi.org/10.. - https://doi.org/10.13070/mm.en.2.201 PMID:.. - https://doi.org/10.1016/S PMID:. - https://doi.org/10.1261/. - https://doi.org/10.1038/nature19309 PMID: 27487213.. - https://doi.org PMID:. - https://doi.org PMID: 20696062.. - https://doi.org/10.1016/s PMID: 10570978.. - doi.org/10.1186/gb-2010-11-2-r14 PMID: 20132535.. - https://doi.org/10.1093/. - doi.org/10.1093/bioinformatics/bts635 PMID: 23104886.. - https://doi.org PMID: 21816040.. - https://doi.org/10.1093/nar/gkv468 PMID:. - doi.org/10.1093/bioinformatics/bth456 PMID: 15297299.. - https://doi.org PMID.. - https://doi.org/10.1371/journal.pgen.. - https://doi.org/10.1006/meth.2001.1262 PMID: 11846609.. - https://doi.org/10.1091/mbc E07-07-0680 PMID: 18753408.
Xem thử không khả dụng, vui lòng xem tại trang nguồn hoặc xem
Tóm tắt