Using Ribosomal Protein Genes as Reference: A Tale of Caution

Lieven Thorrez, Katrijn Van Deun, Léon-Charles Tranchevent et al. · 2008

Read Using Ribosomal Protein Genes as Reference: A Tale of Caution by Lieven Thorrez, Katrijn Van Deun, Léon-Charles Tranchevent et al. (2008) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/plos-pone-0001854.txt.

Introduction

Lieven Thorrez, Katrijn Van Deun, Léon-Charles Tranchevent, and their co-authors warned against a common lab shortcut in a 2008 PLOS ONE research article. Housekeeping genes are needed in every tissue, since their expression is required for cell survival, integrity, or duplication. They are commonly used as reference genes to normalize gene expression data, on the assumption that they are expressed at constant levels.

Using ribosomal protein genes as the test case, the team shows that assumption can fail: reference genes vary more than users expect. The full text gives the expression data across tissues. Bad normalization quietly distorts thousands of published gene measurements, so this cautionary tale has unusual reach.

Catalog record

CREATORLieven Thorrez, Katrijn Van Deun, Léon-Charles Tranchevent et al.
YEAR2008
COLLECTIONPAPERS
FORMATTXT
FILE SIZE43 KB
LICENSECreative Commons Attribution License
RIGHTSCC BY
SOURCEPLOS ONE

Source: https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0001854

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Using Ribosomal Protein Genes as Reference: A Tale of Caution Authors: Lieven Thorrez, Katrijn Van Deun, Léon-Charles Tranchevent, Leentje Van Lommel, Kristof Engelen, Kathleen Marchal, Yves Moreau, Iven Van Mechelen, Frans Schuit Published: 2008, PLOS ONE DOI: 10.1371/journal.pone.0001854 ABSTRACT Housekeeping genes are needed in every tissue as their expression is required for survival, integrity or duplication of every cell. Housekeeping genes commonly have been used as reference genes to normalize gene expression data, the underlying assumption being that they are expressed in every cell type at approximately the same level. Often, the terms “reference genes” and “housekeeping genes” are used interchangeably. In this paper, we would like to distinguish between these terms. Consensus is growing that housekeeping genes which have traditionally been used to normalize gene expression data are not good reference genes. Recently, ribosomal protein genes have been suggested as reference genes based on a meta-analysis of publicly available microarray data. We have applied several statistical tools on a dataset of 70 microarrays representing 22 different tissues, to assess and visualize expression stability of ribosomal protein genes. We confirmed the housekeeping status of these genes, but further estimated expression stability across tissues in order to assess their potential as reference genes. One- and two-way ANOVA revealed that all ribosomal protein genes have significant expression variation across tissues and exhibit tissue-dependent expression behavior as a group. Via multidimensional unfolding analysis, we visualized this tissue-dependency. In addition, we explored mechanisms that may cause tissue dependent effects of individual ribosomal protein genes. Here we provide statistical and biological evidence that ribosomal protein genes exhibit important tissue-dependent variation in mRNA expression. Though these genes are most stably expressed of all investigated genes in a meta-analysis they cannot be considered true reference genes. INTRODUCTION A challenge for the accurate quantification of differences in gene expression level across biological conditions is to normalize for potential artifacts caused by sample preparation or gene expression detection. A common technique in RT-PCR, northern blots or western blots is to normalize data for such artifacts by measuring in the same samples the expression of a reference gene in parallel. The reference gene(s) are assumed to be expressed at constant levels across all the experimental conditions, tissues or cell lines. When only one tissue or cell line is studied, it suffices to look at genes that are constantly expressed in that particular tissue, but need not be expressed in other tissues. In the study of the relative levels of gene expression in various tissues, such as in the study of tissue-specific regulatory elements, a gene that is expressed at constant levels in many tissues is needed. The choice for such reference gene(s) has been a subject of debate for many years. Typical choices were beta-actin, GAPDH, HPRT , or 18S RNA . These genes were thought to be stably expressed since they are considered as “housekeeping genes”. Housekeeping genes have been defined functionally as “constitutively expressed to maintain cellular function” [1] . Being constitutively expressed however does not necessarily meet the prerequisites for a good reference gene, which also needs to display a sufficiently small variation in expression among different tissues. Many of the commonly used reference genes exhibit considerable variability in expression over different tissues and/or experimental conditions [2] – [5] and therefore are not a good choice as reference. More recently, several attempts were performed based on microarray or large-scale sequencing technologies to find more stably expressed reference genes. A meta-analysis of 13629 human Affymetrix arrays was conducted to identify the most stably expressed genes [6] . As a result, a list of 15 genes was suggested with the most constant expression level, based on a coefficient of variation smaller than 4%, a maximum fold change smaller than 2 and a mean expression level lower than the maximum expression level minus 2 times the standard deviation. Thirteen of these fifteen genes are coding for ribosomal proteins. We assessed the expression stability of ribosomal proteins across 22 tissues and suggest that, from a biological and statistical perspective, one should be careful to use these genes as reference genes. RESULTS It was suggested before [7] that ribosomal genes are good housekeeping genes as they are expressed in all cell types to direct biogenesis of new ribosomes. To validate this, we analyzed the probesets for 81 different ribosomal protein genes represented on the Affymetrix mouse 430 2.0 expression arrays in a set of 22 different mouse tissues. In total, 6951 probesets were called present over all individual arrays which was 15% of all probesets on the 430 2.0 array and corresponded to 4845 unique genes (20% of estimated number of mouse genes). With a few exceptions, all ribosomal protein mRNAs were called present over all individual arrays and two were exceptionally called absent ( Figure 1A ). Transcripts from these two genes, Rpl39 -like and Rpl3 -like, were only expressed in a few tissues. Therefore, based on their ubiquitous expression profiles, we could endorse the status of most ribosomal protein genes as housekeeping genes. The next question arising was whether these genes would be suitable as reference genes. (A) Percentage of present expression calls tested in 22 different mouse tissues (3–5 replicates per tissues, in total 70 arrays). Most ribosomal protein genes were present in all tissues examined and thus can be called housekeeping genes. Exceptions were Rpl39l and Rpl3l . (B) Variance of expression levels within replicates of tissues, representing biological variation between animals and technical error on measurements (triangles) compared to variance of expression between tissues (circles). Variance within replicated measurements was significantly smaller than variance of expression between different conditions (Wilcoxon's rank sum statistic: p<0.0001). 1.2 VARIATION ANALYSIS WITHIN VERSUS BETWEEN DIFFERENT TISSUES For all 81 probes representing the ribosomal protein genes, we found with one-way analysis of variance (ANOVA) that the expression levels differed significantly between tissues at a simultaneous significance level of .01 (using a Bonferroni correction to account for multiple testing). To get a clearer picture of the different sources of variation, we performed a two-way ANOVA with the 22 tissues and the 81 probesets representing ribosomal protein genes considered as the factors of variation. The gene effect was most significant, reflecting different average expression levels of ribosomal protein genes over all tissues (F 80,5669 = 5598.61, p<0.0001). After correcting for gene variation, the tissue effect was also highly significant (F 21,5669 = 4094.43, p<0.0001), reflecting that in general ribosomal proteins as a group were more highly expressed in certain tissues. In addition to these main effects (genes as a group or tissues as a group), significant variation could be attributed to the gene-tissue interaction effect (F 1680,5669 = 38.55, p<0.0001), reflecting gene specific deviations in expression across various tissues. In order to assess the importance of expression variation of the ribosomal protein genes among different mouse tissues, we first compared the variance amongst replicates of the same tissue from different animals (representing technical variation plus inter-individual variation) with tissue variance ( Figure 1B ). We performed this analysis for 81 ribosomal protein gene probesets for the 22 different tissues, with 3–5 replicates per tissue. Frequency distribution of the variance on the 1782 (81×22) sets of biologically replicated data is shown in Figure 1B .

Read the complete text