DNA RESEARCH 16, 45–58, (2009) doi:10.

1093/dnares/dsn030

Database for mRNA Half-Life of 19 977 Genes Obtained by DNA

Microarray Analysis of Pluripotent and Differentiating
Mouse Embryonic Stem Cells

Lioudmila V. SHAROVA†, Alexei A. SHAROV†, Timur NEDOREZOV, Yulan PIAO, Nabeebi SHAIK,

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

and Minoru S.H. KO*

Developmental Genomics and Aging Section, Laboratory of Genetics, National Institute on Aging, NIH, 251 Bayview
Boulevard, Suite 100, Baltimore, MD 21224, USA
(Received 17 July 2008; accepted 21 October 2008; published online 11 November 2008)

Abstract
Degradation of mRNA is one of the key processes that control the steady-state level of gene expression.
However, the rate of mRNA decay for the majority of genes is not known. We successfully obtained the rate
of mRNA decay for 19 977 non-redundant genes by microarray analysis of RNA samples obtained from
mouse embryonic stem (ES) cells. Median estimated half-life was 7.1 h and only <100 genes, including
Prdm1, Myc, Gadd45 g, Foxa2, Hes5 and Trib1, showed half-life less than 1 h. In general, mRNA species
with short half-life were enriched among genes with regulatory functions (transcription factors),
whereas mRNA species with long half-life were enriched among genes related to metabolism and structure
(extracellular matrix, cytoskeleton). The stability of mRNAs correlated more significantly with the struc-
tural features of genes than the function of genes: mRNA stability showed the most significant positive cor-
relation with the number of exon junctions per open reading frame length, and negative correlation with
the presence of PUF-binding motifs and AU-rich elements in 30 -untranslated region (UTR) and CpG di-
nucleotides in the 50 -UTR. The mRNA decay rates presented in this report are the largest data set for
mammals and the first for ES cells.
Key words: mRNA decay; mRNA degradation; microarray; transcript; exon junction; AU-rich elements (ARE);
embryonic stem cells; leukemia inhibitory factor (LIF); retinoic acid (RA); cell differentiation; mouse strain

1. Introduction
5 –10% of all genes.2 Quantification of this process
Post-transcriptional regulation of gene expression is important for interpretation of gene-expression
includes mRNA processing, splicing, editing, transport, data obtained with microarrays as well as data from
stability and translation. Degradation of mRNA is an gene manipulation experiments. Changes in gene
important component of gene function that controls expression may be delayed by several days if the corre-
the steady-state concentration of functional transcript sponding transcript is stable. This may cause serious
levels in the cell.1 For example, in human mRNA, stab- errors in reconstruction of gene regulatory networks.
ility is estimated to control the mRNA level of about Studies in several organisms revealed that majority
of transcripts are stable.3 – 5 It appeared that the
Edited by Osamu Ohara
specific half-life of each mRNA is precisely related to
* To whom correspondence should be addressed. Tel. þ1 410- its physiological role.4 – 8 For example, most house-
558-8359. Fax. þ1 410-558-8331. E-mail: KoM@grc.nia.nih.gov keeping genes have long mRNA half-lives. In contrast,
† The first two authors contributed equally to this work.
proteins that are required only for limited time in the
Published by Oxford University Press 2008 cell, e.g. at a certain stage of the cell cycle, during
The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the
open access version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal
and Oxford University Press are attributed as the original place of publication with the correct citation details given; if an article is subsequently
reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use,
please contact journals.permissions@oxfordjournals.org
46 Half-life of mRNA in Mouse ES cells [Vol. 16,

development, growth or differentiation or in response development via Pop2-mediated deadenylation.28

to external stimuli, often have mRNAs with short Binding motif of PUF protein PUM2 has been ident-
half-lives.9 Transcriptionally inducible genes are dis- ified in mouse.29
proportionately represented in the class of genes Estimates of mRNA decay rates may depend on the
with rapid mRNA turnover.5 method of measurement. Functional degradation,
RNA stability can be altered in response to external defined as decrease of the numbers of functional
stimuli such as hormones10 or various types of stress. mRNA capable of producing correct protein, should
For example, response to UV-B exposure leads to be distinguished from physical degradation (or mass
stabilization of many short-lived mRNAs in mamma- decay), which represents the decrease of mRNA

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

lian cells.11 Conversely, in yeast, mild heat shock numbers measured by PCR amplification or hybridiz-
promotes rapid degradation of mRNAs encoding the ation intensity.30 Analysis of functional degradation
ribosomal proteins,12 while glucose starvation, amino is appropriate for individual genes, but it is not an
acid starvation, or sugar-induced osmotic stress causes option for high-throughput analysis targeted at the
stabilization of at least some yeast mRNAs.13 – 15 whole transcriptome. Thus, we will consider only
Changes in RNA stability can also play an important physical degradation of mRNA with understanding
role during embryonic development or cell differen- that this measurement may not capture some
tiation. In plasmodium it has been shown that a aspects of mRNA function. The most widely used
major determinant of mRNA decay rate appears to methods for high-throughput analysis of mRNA
be tightly linked to intra-erythrocytic development degradation is based on transcriptional inhibitors,
cycle, and to a lesser extent the functional category for example by actinomycin D, which inhibits tran-
of the mRNAs themselves.16 During primordial germ scription via binding to single-strand DNA at the
cells specification of Oryzias latipes (Japanese transcription initiation complex and preventing
medaka) some specific RNA molecules are protected elongation by RNA polymerase II.31
from degradation, thus establishing a specific mRNA Multiple studies attempted to measure mRNA
expression pattern controlled by post-transcriptional decay rates in human,3,5 but only one study has rela-
mechanisms.17 Expression of a wide variety of tran- tively large data of 5245 genes in hepatoma cells.4
scripts is controlled by changes in mRNA stability There are no data for mouse or any other mammals.
during neuronal development.18,19 During muscle Among eukaryotes, comparable data have been gen-
differentiation mRNA half-lives for muscle-specific erated for plasmodium,16 yeast7 and Arabidopsis.32
genes—myogenin and myoD—have been shown to be To obtain baseline data for the mRNA half-life in
the highest during differentiation, but declines when mouse, we have selected mouse embryonic stem
differentiation is completed.20 Abnormal changes in (ES) cells. ES cells are often used as experimental
RNA stability can be a cause of cell malfunction models for cell differentiation and for reconstruction
leading to cancer21,22 and other diseases like diabetic of gene regulatory networks. Genes that control
nephropathy,22 muscular atrophy,23 neurological mRNA decay rates may also affect the differentiation
disorders such as Alzheimer disease.9 of ES cells.33 Despite increasing interest in the study
Decay of mRNA is controlled by complex mechan- of ES cells, there have been no attempts to measure
isms that are not fully understood. This mechanism mRNA decay rates in these cells. Here, we have
is integrated with other mRNA-related molecular pro- described mRNA half-life of essentially all mouse
cesses including transcript elongation, splicing, polya- genes in mouse ES cells.
denylation, transport and translation.6,9 RNA decay
mechanisms include interaction between cis-acting
elements and trans-acting factors. Major cis-factors 2. Materials and methods
affecting decay rates include AU-rich elements (ARE),
iron-responsive elements, histone stem-loops, coding 2.1. Cell culture
region determinants, Jun-kinase response elements, We used two mouse ESC lines: MC1 derived from
retained introns, and exon junctions.24 – 26 Trans- mouse strain 129S6/SvEvTac and MC2-B6 derived
factors that control mRNA decay include RNAses from strain C57BL/6J. These lines were purchased
that are often assembled into exosomes, RNA- from The Transgenic Core Laboratory of the Johns
binding proteins ( poly-A binding, WD40 repeats, Hopkins University School of Medicine (Baltimore,
ARE-binding), siRNA, miRNA, piRNA. P-bodies in the MD, USA). To remove contaminating feeder cells, ES
cytoplasm can facilitate miRNA-mediated mRNA cells were first cultured for two passages on gelatin-
degradation induced by stress, and ribosomes coated culture dish in complete ES medium [DMEM,
control nonsense-mediated decay (NMD).27 Recent 15% FBS; Leukemia Inhibitory Factor (LIF) (ESGRO)
studies have shown that PUF-domain proteins can 1000 U/ml; 1 mM sodium pyruvate; 0.1 mM NEAA,
promote mRNA degradation of genes involved in 2 mM glutamate, 0.1 mM b-mercaptoethanol and
No. 1] L.V. Sharova et al. 47

penicillin/streptomycin (50 U/50 mg per ml)]. Cells because of low Cy5/Cy3 signal ratio (.3-fold
were then seeded in gelatin-coated 6-well plates at difference) and high variance (.0.05) of log10(Cy3)
the density of 1 – 2  105 cells/well (1 –2  104/ (chain effect is described in Section 2.4).
cm2) and cultured for 3 days before the test for Degradation rate of mRNA was estimated using
mRNA stability. Cells were incubated at 378C and linear regression of log-transformed (base 10) signal
5% CO2. Medium was changed daily. Differentiated intensity values y versus time t:
ES cells were obtained from MC1 cells by LIF withdra-
wal (complete medium without LIF, below referenced y ¼ a  bt; ð1Þ
as LIF2) or by retinoic acid (RA) treatment (complete

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

medium with 1 mM RA, below referenced as RA) for 7 where t is time, b is the slope, a is intercept and d =
days before the test. To measure mRNA stability, b*ln(10) is instantaneous decay rate. The decrease of
transcription was blocked by adding actinomycin D mRNA amount measured by microarray usually
(Sigma, cat # A-9415) to the medium at the concen- matched well with this model at all time points
tration of 5 mg/ml. Cells were harvested at 0, 1, 2, 4 (Fig. 1A). However, for some mRNA species, the decay
and 8 h after addition of actinomycin D. leveled off at some time point, which may be attributed
to either heterogeneity in degradation rates (i.e. short-
2.2. Microarray experiments lived fraction degraded first) or non-specific hybridiz-
ation to probes on the array (Fig. 1B). For these mRNA
RNA was extracted using Phase lock gelTM
species we estimated decay rates after omitting the
(Eppendorf/Brinkman) columns according to the
later time points to obtain the most accurate estimate
manufacturer’s protocol after adding 1 ml of TrizolTM
of degradation rate. Time points were removed sequen-
(Invitrogen) per well. RNA was precipitated with iso-
tially from the end of the time course till the lower con-
propanol, washed with 70% ethanol and dissolved
fidence limit for b was maximized and the first time
in DEPC dH2O. RNA was hybridized to the NIA
point (t = 0) was within the confidence interval of the
Mouse 44K Microarray v2.3 (whole genome 60-mer
regression. Out of 32 601 probes on the arrays, for
oligonuleotide probe; manufactured by Agilent
which mRNA decay rates were estimated, 95.5%
Technologies, #014951)34 in duplicate for each cell
matched well to the exponent and all five time points
type and time point. Fluorescently labeled microarray
were used for analysis; in 3.4, 0.9 and 0.2% cases
targets were prepared from 2.5 mg aliquots of total
decay rates were estimated using four, three and two
RNA samples using a Low RNA Input Fluorescent
time points, respectively. The same number of time
Linear Amplification Kit (Agilent). A reference target
points was used for different cell types and culture con-
(Cy5-CTP-labeled) was produced from Stratagene
ditions (MC2-B6-LIF+, MC1-LIF+, MC1-LIF2 and MC1-
Universal Mouse Reference RNA (UMR), and all
RA) to ensure proper comparison of mRNA degradation
other targets were labeled with Cy3-CTP. Targets
rates. Because the same amount of RNA was used for
were purified using an RNeasy Mini Kit (Qiagen)
array hybridization, degradation of some mRNA
according to the manufacturer’s protocol, and quanti-
species after block of transcription resulted in the
fied on a NanoDrop scanning spectrophotometer
increase of relative abundance of other stable mRNA
(NanoDrop Technologies). All hybridizations were
species. Thus, additional correction was needed to
carried out by combining a Cy3-CTP-labeled exper-
account for global mRNA degradation. Yang et al.4
imental target and a Cy5-CTP-labeled UMR target.
used b-actin for normalization assuming that it is very
Microarrays were hybridized and washed according
stable. However, Actb appeared not very stable in our
to Agilent protocol (G4140-90030; Agilent 60-mer
experiments (half-life = 7.9 h). Thus, for global normal-
oligonucleotide microarray processing protocol—SSC
ization, we used 200 most stable non-redundant genes
Wash, v1.0). Slides were scanned on an Agilent DNA
with average log intensity of .2.5 and for which decay
Microarray Scanner, using standard settings, including
rates were successfully measured for all four types of
automatic photo-multiplier tube adjustment.
cells. Average mRNA degradation rate for these genes
was estimated for each cell type using Equation (1)
2.3. Estimation of mRNA decay rates (e.g. d = 20.1012 for undifferentiated MC1 cells) and
Outliers in microarray data were detected using cri- then was subtracted from all estimated mRNA degra-
terion of z . 7 for log-transformed Cy3 and Cy5 data dation rate for this cell type. Half-life of each mRNA
separately and removed (0.17% of the data). Then, species was estimated as H = min [24, ln(2)/d] for posi-
Cy3 signals were normalized by Cy5 signals for the tive d, and H = 24 h for negative d. We truncated half-
same probe in log scale, except for 4.34% of probes life estimates by 24 h, because it is difficult to project
in which strong correlation between log Cy5 and mRNA half-life beyond this time point based on the
Cy3 signals (slope of log10(Cy5/Cy3) . 0.25) was experiment that lasted for 8 h. We also estimated the
likely a hybridization artifact (termed ‘chain effect’), average decay rate of each gene in all examined cell
48 Half-life of mRNA in Mouse ES cells [Vol. 16,

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

Figure 1. Stability of mRNA in mouse ES cells. (A and B) Regression of gene expression level (microarray signal intensity in log scale) versus
time after suppressing transcription by actinomycin D ( y = a + bx) is used to estimate the rate of mRNA decay, d = b*ln(10). All 10 data
points are used for an example (A) and only six data points are used for and example (B). (C) Comparison between mRNA decay rates
estimated for MC1 ES cells and those estimated for MC2-B6 ES cells. Rates are log-transformed, log10(d + 0.1). (D) Comparison between
mRNA decay rates estimated using different probes (#1 and #2) on the microarray, which match to the main transcript of the same
gene. Only the data for MC1 cells are shown. (E) Comparison of mRNA decay rates identified with the best probe (#1) from the
main transcript of each gene with decay rates estimated using other probe (#2) that matched the same gene; pairs of probes were
first classified into groups according to the difference in estimated mRNA decay rate (group size is shown below bars, n), and then
the proportion of probes #2 that matched different types of transcripts and/or introns was estimated and plotted for each group.
(F) Frequency distribution of mRNA half-life among all genes. (G) Frequency distribution of mRNA half-life in different functional
categories of genes; color circles indicate that the functional group was previously identified as being over-represented among stable
or unstable mRNA species.

types and the corresponding half-life. Statistical and SE2 are standard errors of these decay rates esti-
significance of the differential mRNA decay rates mated from regression. FDR, which is necessary for
in various cell types was determined using false discov- multiple hypothesis testing, was estimated using the
ery rate (FDR pffiffi, 0.05), estimated from z-values: method of Benjamini and Hochberg.35 In addition to
z ¼ ðd1  d2 Þ= ðSE1 2 þ SE2 2 Þ; where d1 and d2 are FDR, we used the criterion of .2-fold difference
decay rates in cell types 1 and 2, respectively, and SE1 between half-lives to be significant.
No. 1] L.V. Sharova et al. 49

2.4. Calibration of microarray signal intensities GO terms that had a correlation of gene presence
Reliable estimation of mRNA degradation rates 0.7. Second, in the list of GO terms ordered by increas-
requires accurate scaling of expression values to ing P-values, pi, estimated from the hypergeometric
ensure that the ratio of hybridization signal intensities distribution, GO term i was considered redundant, if it
is equal to the ratio of mRNA abundance uniformly was redundant to at least one preceding term.
across the entire range of values. We tested the
scaling using a series of dilutions (5, 1, 0.2 and 2.6. Real-time quantitative reverse transcriptase –
0.04) of the mRNA pool (obtained from newborn polymerase chain reaction
mice of C57BL/6J strain) labeled with Cy3 and Total RNAs were extracted from ES cells using TrizolTM

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

hybridized together with the constant amount of (Invitrogen) and Phase lock gelTM (Eppendorf/
the UMR sample labeled with Cy5. Regression of Brinkman) columns according to the manufacturers’
log-transformed Cy3 signals versus log RNA input protocols. RNAs were precipitated with isopropanol,
was close to 1 for most probes with log signal .1.5 washed with 70% ethanol and dissolved in DEPC
(Supplementary Fig. S1), which indicates proper dH2O. Primers for quantitative reverse transcriptase–
scaling. The results confirmed accurate calibration of PCR (qRT–PCR) were designed using the Vector NTI
intensity signals across the entire range of values. Advance 9.1 software (Invitrogen) and tested for SYBR
Unexpectedly, Cy5 signals of some probes on the Green chemistry using an established in-house proto-
microarray also showed strong positive dependence col.37 Reactions were run on the ABI 7500 Sequence
on RNA input despite the fact that the amount of Detection Systems using the default cycling program,
Cy5-labeled UMR was constant in all the arrays and data were processed using SDS 2.2 software
(Supplementary Fig. S2). This effect was observed (Applied Biosystems). Gene expression levels were
only when Cy3 signal was much stronger than Cy5 normalized to a Cyclin D3 as an internal control and
signal. The mechanism of this effect is unknown, but to total RNA amount.
we hypothesize that it stems from the chain hybridiz-
ation when free ends of Cy3-labeled mRNA bound to
probes on the array can hybridize with Cy5-labeled 3. Results and discussion
mRNA either non-specifically or via rare antisense
segments (termed ‘chain effect’). This effect may 3.1. Measuring mRNA decay rates in mouse ES cells
cause underestimation of gene expression difference To quantify mRNA decay rates in mouse ES cells
between cell types if values are normalized by UMR we measured changes in gene expression with
signal. To avoid distortion, we did not use UMR for whole-genome microarrays in a time course (0, 1, 2,
normalization for a small portion of probes where 4 and 8 h) after treating cells with actinomycin
we expected chain effect. D. Experiments were done for two ES cell lines—
Another challenge in interpretation of microarray MC1 (129S6/SvEvTac) and MC2-B6 (C57BL/6J) cul-
measurements is that expression values often do not tured in the standard condition in the presence of
fall below some ‘background’ level, which most likely LIF. To increase the number of genes that can be
results from non-specific hybridization. If a gene has assayed, we also carried out the analysis of MC1 ES
low expression and is measured close to probe’s back- cells undergoing differentiation into two different
ground level, it may show little change in expression culture conditions—in the absence of LIF and in the
value after actinomycin D treatment, which can be presence of RA for 7 days. Expression data for all
erroneously viewed as evidence of high mRNA stab- genes are available in Supplementary Table S1.
ility. To avoid these errors, we estimated mRNA Proper calibration of signal intensities in the full
decay rates only for sets of data with expression range of gene expression values was confirmed using
values that were either high, .2.5 in log10 scale, or hybridization of different RNA amounts to microar-
substantially above the background, .2 fold. rays (see details in Section 2.2). Expression values for
most genes were normalized using UMR, except for
1824 probes (4.3%) where normalization was not
2.5. Microarray data analysis by gene ontology terms possible because UMR signal was low and unstable
Analysis of gene ontology (GO) terms in a selected list (see details in Section 2.4). Decay rates were esti-
of genes was done using the hypergeometric distri- mated using linear regression of log-transformed
bution and FDR  0.05 criterion using NIA Mouse data (Fig. 1A and B) and normalized by average
Gene Index (ver. mm8) software.36 Only non-redundant expression change of 200 most stable genes (see
genes with gene symbols were used for analysis. Because details in Section 2.4; Supplementary Table S1).
many GO categories are redundant (contain similar lists Because the earlier work did not fully address
of genes), we adjusted the calculation of FDR in the the data reproducibility issue, where analysis was
following way. First, we identified all redundant pairs of limited to estimating statistical (technical) error and
50 Half-life of mRNA in Mouse ES cells [Vol. 16,

validation of expression measurements by PCR for a between 30 -UTR and ORF. For example, oligonucleo-
small number of genes, we felt that it was important tide probes recognized different transcripts, which
to evaluate the reproducibility of the measurements. are produced by alternative transcription start sites
To this end, we used the following two methods: (i) (TSSs), splicing and polyA signals, may lead to the esti-
comparing mRNA decay rates in two independent mation of different half-lives for the same gene. To
experiments with different ES cell lines (MC1 and determine the effect of alternative transcription and
MC2-B6) cultured in the standard conditions and splicing on mRNA stability, we compared decay rates
(ii) comparing mRNA decay rates estimated using identified with the best probe from the main tran-
different oligonucleotide probes on the microarray script of each gene with decay rates estimated using

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

for the main transcript of the same gene. The first other probes that matched the same gene. Probes
method yielded very high consistency between from alternatively spliced transcripts often showed
estimates of mRNA degradation rate (r = 0.925), indi- shorter mRNA half-lives compared with the best
cating biological reproducibility of our database probe in the main transcript (Fig. 1E). This effect
(Fig. 1C). The second method gave somewhat lower was strongest for probes located in the retained
match (r = 0.744), but the correlation was still introns of alternative transcripts. Similar effect of
within the range expected from the comparison of alternative splicing was reported in Arabidopsis;
gene expression data obtained with different array however, it was not statistically significant because
platforms (Fig. 1D). of the small number of alternatively spliced forms
Discrepancy between mRNA half-lives estimated captured by the array.32 Thus, we, for the first time,
using different oligonucleotide probes for the same statistically confirm that alternative transcripts,
gene was likely related to the biological complexity of especially those with retained introns, have shorter
mRNA processing rather than to the error in measure- mRNA half-life compared with main transcripts.
ment. For example, probes located in 30 -UTR often This phenomenon can be simply explained by the
showed shorter mRNA half-lives than probes located possibility that alternative transcripts have a higher
in the open reading frame (ORF) (Supplementary degradation rate due to specific sequences located
Fig. S3A). This may be caused by either alternative ter- in retained introns or alternatively spliced exons.
mination of transcription coupled with a differential Alternatively, it is also possible that the degradation of
degradation rate or active mRNA truncation at some alternative transcripts appears to be accelerated by
early step of mRNA processing. Interestingly, some the splicing that may continue even after the blocking
genes of histone H1 family had the opposite pattern of transcription by actinomycin D. Because RNAs were
of degradation: probes located in the short 30 -UTR extracted from whole cells in our experiments, it may
showed longer half-lives than probes located in the contain heterogeneous nuclear RNA with some
ORF (Supplementary Fig. S3B). However, because of a retained introns, which are able to hybridize with
high sequence similarity between histone genes and intron-matching probes in the array. If splicing con-
a partial cross-hybridization of probes, this effect may tinues for a while after the blocking of transcription
be an artifact. by actinomycin D, the number of heterogeneous
To validate the microarray data, we carried out nuclear RNA would decrease. Further experiments are
qRT–PCR analysis and found that qRT– PCR results needed to distinguish between the two possible mech-
matched well to those obtained by DNA microarrays anisms. All subsequent analysis was done only for main
for six out of seven selected genes (Supplementary transcripts of each gene. We would like to note that for
Fig. S4). An exception was Sox17, whose degradation 1391 genes, mRNA decay rates were not consistent
rates measured by qRT-PCR were much lower than between different microarray probes (.2-fold differ-
those measured by microarrays. The closer inspection ence). Therefore, mRNA half-lives for these genes
of probe locations revealed that all microarray probes estimated with the best oligonucleotide should be
for Sox17 were located in the 30 -UTR of the gene, considered with caution (marked by an asterisk in
whereas qRT-PCR primers were located in the ORF. Supplementary Table S2).
This result was, therefore, consistent with differential
degradation rates of Sox21 measured with probes
located in 30 -UTR and ORF (Supplementary Fig. S3A). 3.3. General trends in stability of mRNAs in ES cells
We were able to estimate mRNA half-lives of 19 977
non-redundant genes (Supplementary Table S2).
3.2. Effect of alternative transcription and splicing For 14 815 genes, estimates are available for two ES
on estimation of RNA stability cell lines and three culture conditions. We were not
It is conceivable that the influence of probe able to estimate mRNA half-lives of 7260 genes due
locations on the estimation of mRNA stability to the low level of expression in ES cells and those of
goes beyond the aforementioned simple difference 541 genes due to the lack of acceptable probes in the
No. 1] L.V. Sharova et al. 51

main transcript. Most of the genes had long half-lives, .2-fold change of half-life) (Supplementary Table S5)
which was in concordance with published data.3 – 5 In and only 10 of these genes overlapped with strain-
our analysis, median estimated half-life for all genes specific signature genes. Although differential gene
was 7.1 h (Fig. 1F), which was less than reported for expression of at least some strain-specific genes can
human hepatoma cells (10 h);4 however, this differ- be explained partially by differences in mRNA stability,
ence may be related to normalization method, as dis- we were not able to identify correlation between differ-
cussed below (see Section 3.9). In our analysis, only ential gene expression and mRNA stability between
54 genes (including Prdm1, Myc, Gadd45 g, Foxa2, strains on a global scale.
Hes5 and Trib1) showed a half-life shorter than 1 h.

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

Published data indicate that more genes showed RNA
half-life shorter than 30 min.3 Therefore, it is possible 3.6. Alteration of mRNA half-life in differentiating
that the number of genes with very short mRNA life ES cells
may be underestimated in our study, because we did Differentiation of ES cells in the absence of LIF or in
not collect samples at the time points earlier than 1 h. the presence of RA caused changes in mRNA stability
in a large set of genes (n = 3332 in LIF2 and n = 2257
3.4. Functional annotation of stable and unstable in RA) (Fig. 2A and B, Supplementary Tables S6 and
transcripts S7). These changes were much stronger than the
strain difference in mRNA stability between MC1
We analyzed over representation of GO terms in the
and MC2-B6 ES cells (Fig. 1C). In general, decay
610 genes with half-life shorter than 2 h and in 4310
rates of mRNA increased after LIF withdrawal: the
genes with half-life longer than 12 h to identify func-
median half-life declined from 7.07 to 5.48 h; and
tional categories of genes with unstable and stable
out of 3332 genes with significant change of mRNA
mRNA, respectively (Supplementary Tables S3 and S4).
stability, 77.9% had increased decay rates compared
Using GO database and additional information on
with undifferentiated ES cells (Fig. 2A and C). In con-
protein domains, we manually assembled lists of genes
trast, differentiation of ES cells in RA conditions
with functions that were characteristic to stable and
caused a general increase of mRNA stability: the
unstable mRNA species (Supplementary Table S2). The
median half-life increased from 7.07 to 8.60 h; and
major functional categories of genes with unstable
out of 2257 genes with significant change of mRNA
mRNA were ‘Regulation of transcription’, ‘Cell cycle’,
stability, 87.1% had decreased decay rates compared
‘Apoptosis’, ‘Signal transduction’ and ‘Development’,
with undifferentiated ES cells (Fig. 2B and C). This
whereas the major functional categories of genes with
indicates a substantial difference between two proto-
stable mRNA were ‘Metabolism’, ‘Protein biosynthesis’,
cols of ES cell differentiation. But despite these global
‘Extracellular matrix’ and ‘Cytoskeleton’ (Fig. 1G). Most
differences, gene-specific changes in mRNA stability in
of these results confirm findings published earlier;4,5,32
these differentiating ES cells showed a good corre-
however, overrepresentation of ‘Extracellular matrix’
lation: the change in decay rates of mRNA (in log
and ‘Cytoskeleton’ groups among stable mRNA species
scale) in RA conditions was positively correlated with
is reported here for the first time. Many transcription
the change of decay rates (in log scale) in LIF2 con-
factors showed very short mRNA half-life, which may
ditions (r = 0.593) (Supplementary Fig. S5A). This
be important for the quick expression changes of
means that changes in mRNA decay rates can be
target genes. Genes with half-life from 1 to 2 h (n =
viewed as a superposition of two effects: (i) gene-
557) included early response and cell cycle regulation
specific effect that is consistent between LIF2 and
and also some genes important for pluripotency and
RA conditions and (ii) non-specific effect associated
early development: for example, Sox2, Klf4, Foxd3,
with the culture medium that caused general
Sox7, Nodal, Ctgf, Id1, Id2, Id3 and Wnt signaling.
decrease in stability in LIF2 conditions and general
increase in mRNA stability in RA conditions.
3.5. Strain-dependent stability of mRNA To select genes whose mRNA were either destabilized
In the previous study, we have shown that there are or stabilized in differentiated cells in both LIF2 and RA
genes differentially expressed between ES cells derived conditions, we used one standard deviation cut-off of
from 129 mouse stain and ES cells derived from log decay rates (Supplementary Fig. S5A). Genes with
C57BL/6 mouse strain (strain-specific signature increased rate of mRNA decay upon differentiation
genes).38 We, therefore, asked if the observed differ- were enriched in GO terms ‘transcription regulation’,
ences were related to differences in mRNA half-life of ‘nervous system development’, ‘apoptosis’, ‘ubiquitin
corresponding genes. We found that a small number cycle’, ‘ribosome’ (mostly mitochondrial ribosomal
of genes (n = 92) had statistically significant differences proteins), ‘chromatin binding’ and ‘methyltransferase
in mRNA decay rates between MC1 (129S6/SvEvTac) activity’ (Supplementary Table S8). Genes with
and MC2-B6 (C57BL/6J) ES cells (FDR ,0.05 and decreased rate of mRNA decay upon differentiation
52 Half-life of mRNA in Mouse ES cells [Vol. 16,

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

Figure 2. Change in mRNA decay rates in ES cells after differentiation. (A) Change of mRNA decay rates after 7 days of differentiation in
LIF2 conditions; significant changes (FDR ,0.05 and .2-fold change of half-life) is shown by red and green dots; rates are log-
transformed, log10(d + 0.1). (B) Change in mRNA decay rates after 7 days of differentiation in RA conditions. (C) Frequency
distribution of mRNA half-life in undifferentiated MC1 cells (LIF+) compared with cells differentiated after LIF withdrawal (LIF2) and
retinoic acid (RA) treatment. (D) Half-life of mRNA of genes associated with pluripotent state of ES cells in undifferentiated cells
(LIF+) and upon differentiation (LIF2 and RA).

were enriched in GO terms ‘extracellular space’, related genes and stabilization of lineage-specific
‘extracellular matrix’, ‘calcium ion binding’, ‘actin cytos- genes. Indeed, our data showed that mRNAs of some
keleton’ and ‘lysosome’ (Supplementary Table S9). pluripotency-related genes (Sall4, Eed, Esrrb, Notch4,
During the ES cell differentiation, unstable mRNA Mras, Tbx3, Btbd14b and Sap30) were destabilized
species became even less stable, whereas stable mRNA upon differentiation (Fig. 2D). Interestingly, the stron-
species became more stable (Supplementary Fig. S5B). gest increase in mRNA decay rate was observed for
It is believed that in general the differentiation of ES Btbd14b (Nac1), which is a known cofactor of
cells is associated with destabilization of pluripotency- NANOG and POU5F1 proteins.39,40 Similarly, mRNAs
No. 1] L.V. Sharova et al. 53

of some differentiation-related genes (e.g. Bmp1, correlations. We incorporated as many structural

Acta1, Igfbp1, Igfbp4, Igfbp6, Frzb, Cited1, Flt1, Slit2, characteristics of genes as possible, including the
Krt8, Krt14, Cnn1, Mest, Wnt3 and Amot) were stabil- length of ORF, the length of 50 -UTR, the length of 30 -
ized. On the other hand, the change in mRNA stability UTR, the number of exons, the occurrence of different
for all genes had no correlation with the change in kinds of ARE, the presence of CpG di-nucleotides and
gene expression levels (Supplementary Fig. S5C and the presence of PUF binding motifs in all parts of the
D), indicating that mRNA stability has no global role transcript (Supplementary Table S11). AREs were
in controlling gene expression changes during ES detected by using three types of reported consensus
cells differentiation. sequences: AUUUA (ARE1), UUAUUUAWW (ARE2),

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

One way to control mRNA degradation is to modu- WWWUAUUUAUWWW (ARE3)45,47 and non-specific
late the expression levels of particular ribonucleases. A/U 12-mers with no more than one mismatch
In metazoans, it is known that the modulation of the (ARE4). We used UGUANAUA as PUF consensus
expression of ribonucleases and/or associated factors motifs.29,48,49 CpG elements were added to the list
play an important role in the control of mRNA stability after we unexpectedly found that they can affect
during development.6 It has also been shown that mRNA stability. We did not analyze the effects of
many ribonucleases and associated factors are differen- miRNA on the mRNA stability, because it would
tially regulated during development. For example, in require extensive data mining and integration with
Drosophila, the expression of ribonucleases pacman/ the existing miRNA databases.
XRN1, twister/SKI2 and tazman/DIS3/RRP44, varies Multiple regression analysis showed that mRNA
during development at both RNA and protein stability was correlated positively with the number
levels.41 – 43 We, therefore, focused on genes that are of exons and negatively with the ORF length,
involved in the mRNA degradation or binding and number of ARE2, ARE4 and PUF in 30 -UTR, number
found that 103 genes changed their expression at of ARE4 and PUF in ORF, and number of CpG di-
.2-fold between undifferentiated and differentiated nucleotides in 50 -UTR (Supplementary Table S11).
ES cells (Supplementary Table S10). The multiplicity The effect of ORF length was nearly equal to the
of mRNA degradation pathways makes it difficult to effect of the number of exons with the opposite
relate expression of these genes with observed patterns sign, which indicated that the number of exon junc-
of mRNA degradation. However, we noted that the tions per unit ORF length may be a better predictor
expression levels of Lsm3, Lsm5, Lsm6, Lsm7, Lsm12, of mRNA stability than the number of exons per
Lsmd1, Cnot4, Cnot10, Dcp2, Exosc1, which are known gene. This is also supported by the fact that most
to promote mRNA degradation, decreased in the exon junctions are located in the ORF region of
differentiated ES cells, especially in the RA condition. genes. Indeed, log-transformed rate of mRNA decay
This may explain increased stability of mRNA in cells correlated stronger with log-transformed number of
differentiated in the RA condition. In contrast, exon junctions per unit ORF length (r = 20.326)
degradation-promoting genes—Zfp36l1 and Cnot6l— than with log-transformed number of exons
showed increased expression in differentiated ES cells, (r = 20.111). Therefore, below we always used the
which may have contributed to lower mRNA stability ratio of exon junctions per 1 kb of ORF length as a
in the LIF2 condition, but it does not explain increased predictor of mRNA stability.
mRNA stability in the RA condition. The final form of regression after removing non-
significant terms for MC1 cells in the standard
culture conditions (LIF+) was
3.7. Multivariate analysis of gene structural features
that affect mRNA stability y ¼ 0:592  0:179x1  0:020x2 þ 0:054x3
Decay rates of mRNA vary substantially between
þ 0:022x4 þ 0:040x5 þ 0:051x6 þ 0:071x7
genes (Fig. 1F); however, little is known about factors
and mechanisms responsible for these differences. þ 0:044x8 ð2Þ
Two major structural factors that affect mRNA
stability are known: (i) the presence of introns in where y is log-transformed rate of mRNA degradation,
genes (i.e. multi-exon genes), which makes mRNA log10(b+0.1); x1 is log-transformed number of exon
more stable;44 and (ii) the presence of AREs in the 30 - junctions per 1 kb ORF length; x2 is log-transformed
UTR of genes, which makes mRNA less stable.45,46 ORF length; x3 is log-transformed number of ARE4
However, all the previous results have been obtained in 30 -UTR; x4 is log-transformed number of ARE4 in
using single-factor analysis, which may easily lead to ORF; and x5 is log-transformed number of ARE2 in
erroneous conclusions if factors are correlated. We, 30 -UTR; x6 is log-transformed number of CpG di-
therefore, used multivariate regression analysis, which nucleotides in 50 -UTR; x7 is log-transformed number
helps to distinguish direct causal effects from mere of PUF motifs in 30 -UTR; and x8 is log-transformed
54 Half-life of mRNA in Mouse ES cells [Vol. 16,

PUF motifs in ORF. Log-transformation was used life (Fig. 3A). Although the role of introns in stability
because it increased the predictive power of the of mRNA has been reported both in experimental
regression and made all regression coefficients and statistical studies,44 here we present the first evi-
dimensionless and comparable with each other; dence that the number of exon junctions per unit ORF
log10(n + 1) was used to handle zero counts as argu- length is the most important known structural factor
ments. This regression accounted for 19.8% of vari- that increases mRNA stability. After intron is excised
ation in mRNA degradation rates, which was only by spliceosome, the exon junction location is
slightly lower than the regression with all 19 factors marked by binding of the exon junction complex
(20.0%). Equation (2) was not precise enough to (EJC), which consists of at least 10 proteins.50 EJC

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

predict mRNA degradation rates for individual genes, remains bound to mRNA after transport to cytoplasm,
but it worked well to estimate the average degra- can stimulate translation, and mediate NMD of mRNA
dation rate in groups of genes identified in microarray during the pioneer round of translation.50 It is indeed
experiments. possible that some components of EJC can remain
According to the regression equation [Equation bound to mRNA after the first round of translation
(2)], we conclude that the number of exon junctions and decrease the rate of mRNA degradation.44 But
per unit ORF length shows the strongest and the the mechanism of this process remains unknown.
most consistent effect on mRNA stability between According to the regression equation [Equation (2)],
cells types. This conclusion is further supported by the second strongest factor affecting the mRNA stability
the alternative data presentation shown in was the number of PUF motifs in 30 -UTR, which
Fig. 3A. The proportion of genes with 5 exon junc- decreased the mRNA stability. The third strongest
tions per unit ORF length was higher in genes with factor affecting the mRNA stability was the presence of
longer mRNA half-life, whereas the proportion of AREs. Unexpectedly, the sequence non-specific ARE4 in
genes associated with the destabilizing factors (CpG 30 -UTR (to a smaller extent, in ORF) decreased mRNA
di-nucleotides in 50 -UTR, and ARE4 and PUF in 30 - half-life more strongly than the consensus-matching
UTR) was higher in genes with shorter mRNA half- AREs. Among different consensus matching AREs, only

Figure 3. Effect of structural and functional characteristics of genes on the rate of mRNA decay. (A) Proportion of genes with increased
number of exon junctions, CpG di-nucleotides in 50 -UTR, and ARE4 and PUF motifs in 30 -UTR in groups of genes with different range
of mRNA half-life. (B) Regression coefficients for the effect of structural elements (cis-factors) and gene function on the mRNA decay
rate; abundance of structural elements and mRNA decay rates are log-transformed. (C) Correlation between functional
characteristics of genes and abundance of structural elements (log-transformed).
No. 1] L.V. Sharova et al. 55

ARE2 in 30 -UTR showed a significant effect on mRNA number of exon junctions per ORF length and low
stability. Several mechanisms are known to be involved number of ARE4 and PUF in 30 -UTR and CpG in 50 -
in ARE-mediated mRNA degradation. For example, UTR (Fig. 3C). However, several functional categories
ARE-binding proteins can recruit PARN or exosomes.51 of genes (e.g. ‘Regulation of transcription’ or
It is also known that the stability of mRNA can be modi- ‘Extracellular matrix’) had a significant effect on
fied by immune factors via AREs in 30 -UTR.52 mRNA stability based on regression (Fig. 3B). This indi-
The fourth strongest factor that affects mRNA stability cates that the effect of gene function cannot fully be
was the number of CpG di-nucleotides in 50 -UTR. explained by structural characteristics of transcripts
Because this was not known before, this is indeed the included in the regression.

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

first report about the effect of CpG di-nucleotides on As mentioned above, differentiation of ES cells in
the mRNA stability. We suspected that the occurrence both LIF2 and RA conditions caused changes in
of CpG di-nucleotides in 50 -UTR may be correlated mRNA stability in a large set of genes (Fig. 3A and B,
with the presence of CpG islands at TSS. We, therefore, Supplementary Tables S6 and S7). To identify struc-
estimated the number of CpG di-nucleotides near TSS tural and functional factors that affected these
(from 2400 to +600 bp) and included it into the global changes in mRNA stability, we compared the
regression analysis. We found that the numbers of CpG coefficients of multiple regression that related decay
di-nucleotides in 50 -UTR appeared to have a 5-fold rates (log scale) with various factors in undifferen-
stronger effect on mRNA stability than their numbers tiated and differentiated ES cells in LIF2 and RA
in the CpG island adjacent to TSS (Supplementary conditions (Fig. 3B, Supplementary Table S13).
Table S12), which indicates that CpG di-nucleotides Degradation-promoting effects of CpG di-nucleotides
need to be located within 50 -UTR to affect mRNA in 50 -UTR and ARE in 30 -UTR appeared much stronger
decay. The mechanism of this effect is unknown, but in cells differentiated in the LIF2 condition compared
considering a strong effect of splicing on mRNA stability with undifferentiated cells and cells differentiated in
we can hypothesize that it may be mediated by methyl the RA condition. Cells differentiated in the RA con-
CpG-binding proteins that regulate splicing.53 dition had weaker degradation-promoting effects of
We have, thus far, shown that both functional ARE4 in 30 -UTR than in undifferentiated cells. These
categories of genes (Section 3.4) and the structural effects may explain the increased mRNA decay rates
features of genes (this section) affect the mRNA stab- in ES cells differentiated in LIF2 and the decreased
ility. It will be, thus, interesting to see which factors decay rates in ES cells differentiated in the RA con-
affect mRNA stability more strongly. To this end, we dition. Protection of mRNA from degradation by
carried out the regression analysis after including exon junctions was stronger in both types of differen-
both structural features and functional categories tiated ES cells than in undifferentiated cells.
(Fig. 3B, Supplementary Table S13). Among functional
groups of genes, the strongest destabilizing effect was
shown in the category of ‘Regulation of transcription’ 3.9. Comparison with human data
followed by ‘Signal transduction’ and ‘Cell cycle’. In Interspecies comparison of mRNA decay rates has
contrast, ‘Metabolism’, ‘Extracellular matrix’ and been done only among distant evolutionary lineages
‘Cytoskeleton’ functional categories increased mRNA ( plant, yeast and human). The new mouse data set
stability, which is consistent with our previous results provides an opportunity to compare the two
(Fig. 1G). However, we found that all the effects of mammalian species: mouse and human. Using
structural factors listed in Equation (2) remained Homologene database54 we identified 4153 pairs of
strong and statistically significant, and the sign of orthologous genes for which mRNA rates were esti-
regression coefficients did not change after adding mated both in mouse (this study) and human
functional gene categories. This indicates that the stab- (Supplementary Table S3). For compatibility with
ility of mRNAs is more significantly correlated with the our data, we converted estimates of mRNA decay
structural features of genes than the function of genes. rates in human from log2 to loge. Because human
Furthermore, the differential mRNA stability in data are normalized using b-actin, whose mRNA is
various functional groups of genes can be partially not highly stable (based on our data), a large set of
explained by structural features of genes. For example, human genes (11.6%) shows negative degradation
low stability of mRNA among genes in the functional rates.4 We, therefore, re-normalized the estimates of
category of ‘Regulation of transcription’ and human mRNA decay rates using the method we
‘Development’ can be explained by the low number of used for mouse (by average decay rate of 200
exon junctions per ORF length and high number of genes with most stable mRNA). Median rate of
ARE4 and PUF in 30 -UTR and CpG in 50 -UTR (Fig. 3C). mRNA decay for all orthologous genes in mouse
Similarly, high stability of mRNA among genes in the (d ¼ 0.098 h21) was lower than in human (d ¼
category of ‘Metabolism’ can be explained by the high 0.137 h21). Because these experiments differed not
56 Half-life of mRNA in Mouse ES cells [Vol. 16,

of any mammalian species. (iii) We have analyzed

the difference in mRNA decay rates between cells
derived from two different mouse strains and
between undifferentiated and differentiated cells.
(iv) The set of genes for which mRNA decay rates esti-
mated is 3.8 times larger than in the most extensive
earlier study for human. (v) Close match between
data from different cell lines has confirmed high
reproducibility of mRNA decay rate estimates. (vi)

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

The microarrays have been calibrated. (vii) We have
applied multivariate statistics for the analysis of
factors affecting mRNA decay rates, which identifies
direct effects and distinguishes them from mere cor-
relations. (viii) We have explored the effect of
Figure 4. Comparison of mRNA decay rates for orthologous genes
various structural features of transcripts and gene
in mouse ES cells and in human hepatocytes (from the function on mRNA stability in both general and
reference4) Rates are log-transformed, log10(d + 0.1). Human during ES cell differentiation. The major result is that
mRNA decay rates were re-normalized using average decay the number of exon junctions per ORF is the strongest
rate of 200 most unstable mRNA species for proper comparison.
factor promoting mRNA stability, whereas decay is
promoted by CpG di-nucleotides in 50 -UTR region as
just in species (i.e. mouse versus human), but also in well as by non-canonical ARE (without specific con-
cell types (ES cells versus hepatoma), and were done sensus sequence) in 30 -UTR and in ORF. We believe
in different laboratories using different microarray that the presented data on mRNA decay rates pro-
platforms, it was impossible to attribute this differ- vides a valuable resource for future experiments in
ence to any specific factor. However, correlation interpreting various gene manipulation and gene
between mRNA decay rates in mouse and human expression studies as well as the computational mod-
was high (r = 0.610, Fig. 4), which was only slightly eling and simulation on gene regulatory networks.
lower than correlation between estimates of mRNA
decay rates obtained with different probes that Supplementary data: Supplementary data are avai-
matched the same mouse transcript (r = 0.744, see lable online at www.dna.research.oxfordjournals.org.
Section 3.1). This finding indicates that pathways of
mRNA degradation are mostly conserved between
Funding
mammalian species. Interestingly, mRNA decay rates
The work was supported by the Intramural
in human hepatoma were more similar to decay
Research Program of the National Institute on Aging,
rates in differentiated ES cells (r = 0.605 for LIF2
NIH (Z01 AG000656).
and r = 0.616 for RA) than to decay rates in undiffer-
entiated ES cells (r = 0.534, SE = 0.013, P = 0.0001).
This may be explained by the fact that hepatoma
References
cells are differentiated and, thus, the profile of
mRNA decay rate in hepatoma matched better with
1. Raghavan, A. and Bohjanen, P. R. 2004, Microarray-
that in differentiated ES cells. However, this similarity based analyses of mRNA decay in the regulation of
needs to be confirmed with further experiments that mammalian gene expression, Brief Funct. Genomic
use the same cell types in both species. Proteomic., 3, 112– 124.
2. Bolognani, F. and Perrone-Bizzozero, N. I. 2008, RNA –
protein interactions and control of mRNA stability in
3.10. Conclusion neurons, J. Neurosci. Res., 86, 481– 489.
In this paper we present a whole-genome analysis 3. Raghavan, A., Ogilvie, R. L., Reilly, C., et al. 2002,
of mRNA decay rates in mouse ES cells. Decay rates Genome-wide analysis of mRNA decay in resting and
were reliably measured for 19 977 non-redundant activated primary human T lymphocytes, Nucleic Acids
Res., 30, 5529– 5538.
genes from time course of microarray data in undif-
4. Yang, E., van Nimwegen, E., Zavolan, M., et al. 2003,
ferentiated ES cells originated from different mouse Decay rates of human mRNAs: correlation with func-
strains and in differentiated ES cells obtained by LIF tional characteristics and sequence attributes, Genome
withdrawal (LIF2) or RA treatment. The novelty of Res., 13, 1863– 1872.
this study can be summarized as follows: (i) this is 5. Lam, L. T., Pickeral, O. K., Peng, A. C., et al. 2001,
the first study of mRNA decay rates in mouse. (ii) Genomic-scale measurement of mRNA turnover and
This is the first study of mRNA decay rates in ES cells the mechanisms of action of the anti-cancer drug
No. 1] L.V. Sharova et al. 57
flavopiridol, Genome Biol., 2, research0041.1 – for the mRNA stabilization process, Leuk. Res., 30,
research0041.11. 1407– 1416.
6. Newbury, S. F. 2006, Control of mRNA stability in eukar- 22. Benjamin, D. and Moroni, C. 2007, mRNA stability and
yotes, Biochem Soc. Trans., 34, 30– 34. cancer: an emerging link?, Expert Opin. Biol. Ther., 7,
7. Garcia-Martinez, J., Aranda, A. and Perez-Ortin, J. E. 1515– 1529.
2004, Genomic run-on evaluates transcription rates 23. Heier, C. R., Gogliotti, R. G. and DiDonato, C. J. 2007,
for all yeast genes and identifies gene regulatory mech- SMN transcript stability: could modulation of messen-
anisms, Mol. Cell, 15, 303– 313. ger RNA degradation provide a novel therapy for
8. Wang, J., Chang, Y. F., Hamilton, J. I. and Wilkinson, M. F. spinal muscular atrophy?, J. Child Neurol., 22,
2002, Nonsense-associated altered splicing: a frame- 1013– 1018.

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

dependent response distinct from nonsense-mediated 24. Stoecklin, G., Mayo, T. and Anderson, P. 2006, ARE-
decay, Mol. Cell, 160, 951 –957. mRNA degradation requires the 50 – 30 decay pathway,
9. Hollams, E. M., Giles, K. M., Thomson, A. M. and EMBO Rep., 7, 72– 77.
Leedman, P. J. 2002, MRNA stability and the control of 25. Le Hir, H., Gatfield, D., Izaurralde, E. and Moore, M. J.
gene expression: implications for human disease, 2001, The exon– exon junction complex provides a
Neurochem. Res., 27, 957– 980. binding platform for factors involved in mRNA export
10. Sahmi, M., Nicola, E. S. and Price, C. A. 2006, Hormonal and nonsense-mediated mRNA decay, EMBO J., 20,
regulation of cytochrome P450 aromatase mRNA stab- 4987– 4997.
ility in non-luteinizing bovine granulosa cells in vitro, 26. Le Hir, H., Izaurralde, E., Maquat, L. E. and Moore, M. J.
J. Endocrinol., 190, 107 –115. 2000, The spliceosome deposits multiple proteins
11. Gowrishankar, G., Winzen, R., Bollig, F., et al. 2005, 20-24 nucleotides upstream of mRNA exon – exon
Inhibition of mRNA deadenylation and degradation by junctions, EMBO J., 19, 6860 –6869.
ultraviolet light, Biol. Chem., 386, 1287– 1293. 27. Yamasaki, S., Stoecklin, G., Kedersha, N., Simarro, M. and
12. Herruer, M. H., Mager, W. H., Raue, H. A., Vreken, P., Anderson, P. 2007, T-cell intracellular antigen-1 (TIA-
Wilms, E. and Planta, R. J. 1988, Mild temperature 1)-induced translational silencing promotes the decay
shock affects transcription of yeast ribosomal protein of selected mRNAs, J. Biol. Chem., 282, 30070– 30077.
genes as well as the stability of their mRNAs, Nucleic 28. Goldstrohm, A. C., Hook, B. A., Seay, D. J. and Wickens, M.
Acids Res., 16, 7917– 7929. 2006, PUF proteins bind Pop2p to regulate messenger
13. Greatrix, B. W. and van Vuuren, H. J. 2006, Expression of RNAs, Nat. Struct. Mol. Biol., 13, 533 –539.
the HXT13, HXT15 and HXT17 genes in Saccharomyces 29. White, E. K., Moore-Jarrett, T. and Ruley, H. E. 2001,
cerevisiae and stabilization of the HXT1 gene transcript PUM2, a novel murine puf protein, and its consensus
by sugar-induced osmotic stress, Curr. Genet., 49, RNA-binding site, RNA, 7, 1855– 1866.
205 –217. 30. Kennell, D. 2002, Processing endoribonucleases
14. Benard, L. 2004, Inhibition of 50 to 30 mRNA degra- and mRNA degradation in bacteria, J. Bacteriol., 184,
dation under stress conditions in Saccharomyces cerevi- 4645– 4657. Discussion 4665.
siae: from GCN4 to MET16, RNA, 10, 458– 468. 31. Sobell, H. M. 1985, Actinomycin and DNA transcription,
15. Jona, G., Choder, M. and Gileadi, O. 2000, Glucose star- Proc. Natl Acad. Sci. USA, 82, 5328 –5331.
vation induces a drastic reduction in the rates of both 32. Narsai, R., Howell, K. A., Millar, A. H., O’Toole, N., Small, I.
transcription and degradation of mRNA in yeast, and Whelan, J. 2007, Genome-wide analysis of mRNA
Biochim. Biophys. Acta, 1491, 37– 48. decay rates and their determinants in Arabidopsis thali-
16. Shock, J. L., Fischer, K. F. and DeRisi, J. L. 2007, Whole- ana, Plant Cell, 19, 3418– 3436.
genome analysis of mRNA decay in Plasmodium falci- 33. Wegmuller, D., Raineri, I., Gross, B., Oakeley, E. J. and
parum reveals a global lengthening of mRNA half-life Moroni, C. 2007, A cassette system to study embryonic
during the intra-erythrocytic development cycle, stem cell differentiation by inducible RNA interference,
Genome Biol., 8, R134. Stem Cells, 25, 1178– 1185.
17. Herpin, A., Rohr, S., Riedel, D., Kluever, N., Raz, E. and 34. Carter, M. G., Sharov, A. A., VanBuren, V., et al. 2005,
Schartl, M. 2007, Specification of primordial germ Transcript copy number estimation using a mouse
cells in medaka (Oryzias latipes), BMC Dev. Biol., 7, 3. whole-genome oligonucleotide microarray, Genome
18. Deschenes-Furry, J., Perrone-Bizzozero, N. and Biol., 6, R61.
Jasmin, B. J. 2006, The RNA-binding protein HuD: a reg- 35. Benjamini, Y. and Hochberg, Y. 1995, Controlling the
ulator of neuronal differentiation, maintenance and false discovery rate—a practical and powerful approach
plasticity, Bioessays, 28, 822– 833. to multiple testing, J. R. Stat. Soc. Ser. B, 57, 289 –300.
19. Ratti, A., Fallini, C., Cova, L., et al. 2006, A role for the 36. Sharov, A. A., Dudekula, D. B. and Ko, M. S. 2005,
ELAV RNA-binding proteins in neural stem cells: stabiliz- Genome-wide assembly and analysis of alternative
ation of Msi1 mRNA, J. Cell Sci., 119, 1442– 1452. transcripts in mouse, Genome Res., 15, 748–754.
20. Figueroa, C. and Vojtek, A. B. 2003, Akt negatively regu- 37. Carter, M. G., Hamatani, T., Sharov, A. A., et al. 2003,
lates translation of the ternary complex factor Elk-1, In situ-synthesized novel microarray optimized for
Oncogene, 22, 5554– 5561. mouse stem cell and early developmental expression
21. Champelovier, P., Pautre, V., Elatifi, M., et al. 2006, profiling, Genome Res., 13, 1011 –1021.
Resistance to phorbol ester-induced differentiation 38. Sharova, L. V., Sharov, A. A., Piao, Y., et al. 2007, Global
in human myeloid leukemia cells: a hypothetic role gene expression profiling reveals similarities and
58 Half-life of mRNA in Mouse ES cells [Vol. 16,
differences among mouse pluripotent stem cells of 47. Bakheet, T., Frevel, M., Williams, B. R., Greer, W. and
different origins and strains, Dev. Biol., 307, 446– 459. Khabar, K. S. 2001, ARED: human AU-rich element-con-
39. Wang, J., Rao, S., Chu, J., et al. 2006, A protein inter- taining mRNA database reveals an unexpectedly diverse
action network for pluripotency of embryonic stem functional repertoire of encoded proteins, Nucleic Acids
cells, Nature, 444, 364 –368. Res., 29, 246–254.
40. Wang, J., Levasseur, D. N. and Orkin, S. H. 2008, 48. Xie, X., Lu, J., Kulbokas, E. J., et al. 2005, Systematic dis-
Requirement of Nanog dimerization for stem cell self- covery of regulatory motifs in human promoters and 30
renewal and pluripotency, Proc. Natl Acad. Sci. USA, UTRs by comparison of several mammals, Nature, 434,
105, 6326– 6331. 338– 345.
41. Cairrao, F., Arraiano, C. and Newbury, S. 2005, Drosophila 49. Cora, D., Di Cunto, F., Caselle, M. and Provero, P. 2007,

Downloaded from https://academic.oup.com/dnaresearch/article/16/1/45/364974 by Demography Library user on 13 June 2024

gene tazman, an orthologue of the yeast exosome com- Identification of candidate regulatory sequences in
ponent Rrp44p/Dis3, is differentially expressed during mammalian 30 UTRs by statistical analysis of oligonu-
development, Dev. Dyn., 232, 733–737. cleotide distributions, BMC Bioinformatics, 8, 174.
42. Seago, J. E., Chernukhin, I. V. and Newbury, S. F. 2001, 50. Chang, Y. F., Imam, J. S. and Wilkinson, M. F. 2007, The
The Drosophila gene twister, an orthologue of the nonsense-mediated decay RNA surveillance pathway,
yeast helicase SKI2, is differentially expressed during Annu. Rev. Biochem., 76, 51– 74.
development, Mech. Dev., 106, 137– 141. 51. Garneau, N. L., Wilusz, J. and Wilusz, C. J. 2007, The
43. Till, D. D., Linz, B., Seago, J. E., et al. 1998, Identification and highways and byways of mRNA decay, Nat. Rev. Mol.
developmental expression of a 50 –30 exoribonuclease Cell Biol., 8, 113 –126.
from Drosophila melanogaster, Mech. Dev., 79, 51–55. 52. Tebo, J., Der, S., Frevel, M., Khabar, K. S., Williams, B. R.
44. Zhao, C. and Hamilton, T. 2007, Introns regulate the and Hamilton, T. A. 2003, Heterogeneity in control of
rate of unstable mRNA decay, J. Biol. Chem., 282, mRNA stability by AU-rich elements, J. Biol. Chem.,
20230– 20237. 278, 12085– 12093.
45. Chen, C. Y. and Shyu, A. B. 1995, AU-rich elements: 53. Young, J. I., Hong, E. P., Castle, J. C., et al. 2005, Regulation
characterization and importance in mRNA degradation, of RNA splicing by the methylation-dependent transcrip-
Trends Biochem. Sci., 20, 465– 470. tional repressor methyl-CpG binding protein 2, Proc.
46. Shaw, G. and Kamen, R. 1986, A conserved AU Natl Acad. Sci. USA, 102, 17551–17558.
sequence from the 30 untranslated region of GM-CSF 54. Wheeler, D. L., Barrett, T., Benson, D. A., et al. 2008,
mRNA mediates selective mRNA degradation, Cell, 46, Database resources of the National Center for
659– 667. Biotechnology Information, Nucleic Acids Res, 36, D13–21.