Analytical Biochemistry 405 (2010) 224229

Contents lists available at ScienceDirect

Analytical Biochemistry
journal homepage: www.elsevier.com/locate/yabio

Identication of suitable reference genes for measurement of gene expression

in human cervical tissues
Yuanming Shen a, Yang Li a, Feng Ye a, Fenfen Wang a, Weiguo Lu b, Xing Xie b,*
a

Womens Reproductive Health Laboratory of Zhejiang Province, Womens Hospital, School of Medicine, Zhejiang University, Hangzhou 310006, China
Womens Reproductive Health Laboratory of Zhejiang Province, Department of Gynecologic Oncology, Womens Hospital, School of Medicine, Zhejiang University,
Hangzhou 310006, China

a r t i c l e

i n f o

Article history:
Received 4 May 2010
Received in revised form 14 June 2010
Accepted 15 June 2010
Available online 19 June 2010
Keywords:
Cervical cancer
Real-time PCR
Reference gene
GeNorm
NormFinder

a b s t r a c t
For quantitative real-time reverse transcription-polymerase chain reaction (RT-qPCR), the most commonly used normalization strategy is to select a stable reference gene. However, no suitable reference
genes have been identied in cervical tissues to date. The aim of this study was to identify the most stable
gene or a set of genes as reference genes for RT-qPCR analysis in cervical tissues from a panel of 12 candidates (ALAS1, PPIA, GAPDH, HBB, TBP, ACTIN, B2M, MBNL2, PGKL, RPLP0, RPL-4, and EEF1A1). In total,
20 normal and 20 cervical cancer specimens were examined. Gene expression data were analyzed using
two different statistical models (geNorm and NormFinder). EEF1A1 was identied as the most stable and
reliable reference gene, followed by GAPDH and RPLP0, whereas EEF1A1 and GAPDH were the best twogene combination by NormFinder. The expression validity of EEF1A1 was further determined in 21 normal, 22 cervical intraepithelial neoplasia (CIN23), and 18 cancer tissues; no expression differences were
found among normal, CIN23, and cancer tissues (P > 0.05). Our results suggested that EEF1A1 can be used
as a reference gene for normalization in gene proling studies in clinic cervical samples, and the combination of EEF1A1 and GAPDH could be recommended as a much more reliable normalization strategy.
2010 Elsevier Inc. All rights reserved.

Quantitative real-time reverse transcription-polymerase chain

reaction (RT-qPCR)1 is an efcient tool that measures absolute transcript abundance and provides valuable quantitative information on
gene expression of biological samples from different sources. Thousands of research laboratories worldwide have embraced RT-qPCR as
a frequently used method for measuring genes expression in transcript levels because of its relatively low cost, high precision, and
high sensitivity as well as its exibility and scalability [13]. However, despite its popularity, more and more researchers have begun
to realize that accurate normalization of gene expression level is an
absolute prerequisite for reliable results, and several measures need
to be controlled to obtain reliable quantitative expression measures,
including RNA extraction, RNA integrity control, complementary
DNA (cDNA) synthesis, primer design, amplicon detection, and data
normalization [39]. To date, the most frequently used approach for
normalization of the above-mentioned parameters is to select a stable reference gene [1]. An ideal reference gene should be universally

* Corresponding author. Fax: +86 571 87036290.

E-mail address: xiex@mail.hz.zj.cn (X. Xie).
1
Abbreviations used: RT-qPCR, quantitative real-time reverse transcription-polymerase chain reaction; cDNA, complementary DNA; CIN, cervical intraepithelial
neoplasia; FIGO, International Federation of Gynecologic Oncology; H&E, hematoxylin
and eosin; HPV, human papillomavirus; ICC, invasive cervical cancer; rRNA, ribosomal
RNA; RIN, RNA integrity number.
0003-2697/$ - see front matter 2010 Elsevier Inc. All rights reserved.
doi:10.1016/j.ab.2010.06.029

valid; it should neither vary among the tissues or cells nor respond
to experimental intervention. Recently, some literature studies have
shown that there probably is no universal reference gene with a constant expression in all tissues, particularly in clinical samples associated with malignant diseases [29]. Furthermore, an experimental
treatment may also affect the expression of any kind of reference
gene. Thus, it is necessary to carry out an extensive evaluation of
commonly used housekeeping genes and identify an optimal tissue-specic reference gene in restricted tissue types and restricted
experimental designs.
Cervical cancer is the second most common cancer affecting
women worldwide, with an estimated global incidence of
470,000 new cases and approximately 233,000 deaths per year
[10]. Human papillomavirus (HPV) infection has been rmly established as a central cause of invasive cervical cancer (ICC). Clinical
and epidemiological studies have clearly conrmed that 99.9% of
ICC cases present HPV positive [1012]. However, HPV infection
alone is not sufcient to induce the malignant transformation,
and other unidentied genetic alterations are also involved. During
recent years, a lot of gene expression studies in malignant cervical
tissue and its normal tissue counterpart have been performed to
nd new molecular markers that can promote the malignant transformation by HPV infection [13,14], and RT-qPCR is a frequently
used tool to detect the expression of those molecules. By this token, it seems to be necessary to scan normalization strategies used

Reference genes in human cervical tissues / Y. Shen et al. / Anal. Biochem. 405 (2010) 224229

in quantitative gene expression. The purpose of this study was to

identify the most stable reference control gene or a set of genes
as reference genes for RT-qPCR analysis in primary human cervical
tissues from a panel of 12 candidates (ALAS1, PPIA, GAPDH, HBB,
TBP, ACTIN, B2M, MBNL2, PGKL, RPLP0, RPL-4, and EEF1A1). These
12 candidates were commonly used as endogenous controls in the
context of, but not restricted to, cervical cancer [5,8,9,1519].
Materials and methods
Cervical sample collection
Informed consent was obtained according to the guidelines set
forth and approved by the ethical committee for human research at
Womens Hospital in the School of Medicine at Zhejiang University.
Between September 2008 and September 2009, we collected a total
of 101 tissue specimens comprising 41 normal cervical tissues, 38
cervical cancer tissues, and 22 cervical intraepithelial neoplasia
(CIN) tissues. Cervical cancer samples were collected under colposcopy from primary untreated cervical cancer patients (mean
age = 42 years, range = 3358). According to the classication of
International Federation of Gynecologic Oncology (FIGO), 25 cervical cancer patients were stage I, 10 were stage II, 2 were stage III,
and 1 was stage IV. Of the 38 tumors, 30 were histologically diagnosed as squamous cell carcinoma, 5 as adenocarcinoma, and 3 as
adenosquamous carcinoma. In addition, 6 tumors were grade 1, 28
were grade 2, and 4 were grade 3. Normal cervical samples (n = 41)
were derived from the patients (mean age = 44 years, range = 38
65) who underwent hysterectomy due to nonmalignant diseases.
CIN samples were obtained from primary untreated CIN23 patients
(mean age = 32 years, range = 2245) under colposcopy. Histological analysis of hematoxylin and eosin (H&E)-stained sections
showed that all cancer samples contained at least 90% tumor cells
without necrosis, and all normal samples were veried to be free of
any pathological lesions. HPV detection was performed using a
newly commercially available HPV GenoArray test kit (Hybribio,
Hong Kong); all cancer and CIN samples were veried to be HPV
positive, and all normal samples were veried to be negative. All
tissue samples were immediately snap-frozen in liquid nitrogen
after excision and stored at 180 C until RNA extraction.
To exclude run-to-run variation among the samples, we rst selected 40 representative samples (20 normal and 20 tumor) in the
initial experiment after statistical assessment. In 20 tumors, seven
were classied as stage I, 10 as stage II, two as stage III, and one as
stage IV. In addition, 12 tumors were histologically diagnosed as
squamous cell carcinoma, ve as adenocarcinoma, and three as
adenosquamous carcinoma. The histological grading was six G1,
10 G2, and four G3. The other 61 samples were used to further validate the data in initial experiment results.

225

RNA extraction and cDNA synthesis

RNA was isolated by the Trizol method (Invitrogen, Carlsbad,
CA, USA) following the manufacturers protocol and was dissolved
in diethylpyrocarbonate-treated water. Finally, the quantity and
concentration of RNA were spectrophotometrically assessed by
measuring absorbance at OD260/280, and gel electrophoresis was assessed by measuring the integrity of ribosomal RNA (rRNA) subunits (28S and 18S). The criterion to include RNA samples was
260/280  2 (1.92.2) and 28S/18S ratio P1.7. The concentration
of RNA was adjusted to 0.5 lg/ll with diethylpyrocarbonate-treated water. Total RNA (1 lg) was DNase treated with Turbo DNAfree (Ambion, Austin, TX, USA) following the manufacturers
instructions. Of the 101 total samples, 40 samples RT was performed using 1.5 lg of DNase-treated RNA in a 60-ll reaction volume by a PrimeScript RT reagent kit (perfect real-time) according
to the manufacturers instructions (TaKaRa Biotechnology, Shiga,
Japan). The other 61 samples RT was performed using 0.5 lg of
DNase-treated RNA in a 20-ll reaction volume. Briey, the RT reactions were initiated with a 15-min incubation at 37 C, followed by
5 s at 85 C to inactivate enzymes. The cDNA product was stored at
80 C until use.
Selection of candidate reference genes for normalization
The 12 candidate genes were selected based on their known
expression in cervical tissues and/or based on their common use
as endogenous controls in the context of, but not limited to, cervical cancer: ALAS1, PPIA, GAPDH, HBB, TBP, ACTIN, B2M, MBNL2,
PGKL, RPLP0, RPL-4, and EEF1A1. Known information about the
candidate genes is listed in Tables 1 and 2. The primer sequences
were designed with primer 5 software according to the sequences
obtained from GenBank data, and the specicity of the primers was
conrmed by BLAST searches.
Real-time PCR
Following the RT step, SYBR Green real-time PCR was performed
with an Applied Biosystems 7900HT Fast Real-Time PCR system
using the SYBR Premix Ex Taq (perfect real time) (TaKaRa Biotechnology). The PCR volume was 20 ll and contained 2 ll of cDNA,
10 ll of SYBR Green I, 0.4 ll of 50 ROX standard, 1.0 ll of 10
PCR forward and reverse primer mix (Invitrogen), and 6.6 ll of
diethylpyrocarbonate-treated water. The amplication cycling
conditions were as follows: 95 C for 10 s (95 C for 10 s, 60 or
63 C for 30 s)  40 cycles. A no-RT control was amplied from each
sample to control for remaining DNA contaminations. To exclude
between-run variations, all reactions were run in duplicate with
a no-template control for each run, and the no-RT controls were

Table 1
Putative reference genes evaluated.
Gene symbol

Accession number

Gene name

Genomic localization

Molecular function

ACTIN
ALAS1
GAPDH
PPIA
TBP
PGK1
EEF1A1
HBB
B2M
MBNL2
RPLP0
RPL4

NM-001101
NM-000688
NM-002046
NM-021130
NM-003194
NM_000291
NM_001402
NG_000007
NM_004048
NM_144778
NM_001002
NM_000968

Beta-actin
Delta-aminolevulinate synthase 1
Glyceraldehyde-3-phosphate dehydrogenase
Peptidylprolyl isomerase A
TATA box binding protein
Phosphoglycerate kinase 1
Eukaryotic translation elongation factor 1 alpha 1
Hemoglobin beta
Beta-2-microglobulin
Muscleblind-like 2
Large ribosomal protein P0
Ribosomal protein L4

7p15p12
3p21.1
12p13
7p13
6q27
Xq13
6q14.1
11p15.5
15q21q22.2
13q32.1
12q24.2
15q22

Cytoskeletal structural protein

5-Aminolevulinate synthase
Oxidoreductase in glycolysis and gluconeogenesis
Cyclosporine binding protein
General transcription protein
Glycolytic enzyme
Translation elongation factor
Hemoglobin
Microglobulin
C3H-type zinc nger protein
Ribosomal protein
Structural constituent of ribosome

226

Reference genes in human cervical tissues / Y. Shen et al. / Anal. Biochem. 405 (2010) 224229
Table 2
Details of primers and PCR efciency for the 12 evaluated genes.
Gene

Forward primer sequences 50 30

Reverse primer sequences 50 30

Amplicon length (bp)

PCR efciency (%)

ACTIN
ALAS1
GAPDH
PPIA
TBP
PGK1
EEF1A1
HBB
B2M
MBNL2
RPLP0
RPL4

AGAAAATCTGGCACCACACC
GGCAGCACAGATGAATCAGA
GACAGTCAGCCGCATCTTCT
AGACAAGGTCCCAAAGAC
TGCACAGGAGCCAAGAGTGAA
ATTAGCCGAGCCAGCCAAAATAG
TGCGGTGGGTGTCATCAAA
CAGGTACGGCTGTCATCACTTAGA
CACCCCCACTGAAAAAGATG
GTCACTGTCCCGGGCTCAACTGC
GCTGCTGCCCGTGCTGGTG
GCTCTGGCCAGGGTGCTTTTG

GCTGGAATTACCGCGGCT
CCTCCATCGGTTTTCACACT
TTAAAAGCAGCCCTGGTGAC
ACCACCCTGACACATAAA
CACATCACAGCTCCCCACCA
TCATCAAAAACCCACCAGCCTTCT
AAGAGTGGGGTGGCAGGTATTG
CATGGTGTCTGTTTGAGGTTGCTA
ATATTAAAAAGCAAGCAAGCAGAA
ATGGTGCTGTCTGCGGGGTGTG
TGGTGCCCCTGGAGATTTTAGTGG
ATGGCGTATCGTTTTTGGGTTGT

186
150
127
118
132
152
123
183
167
137
130
154

100
98
100
98
99
98.5
100
95.0
98
99.5
100
100

analyzed with pooled samples by ve PCR replicates. A melting

curve was constructed for each primer pair to conrm product
specicity.
The comparative Cq method was used to determine quantitative
comparison of the amplication of the candidate genes. The cycle
threshold (Cq) was converted into quantities relative to normal
(QRel), and corrected for PCR amplication efciency (E), using
the following formula: QRel = EDCq, where DCq = Cq test sample  Cq calibrator sample. The PCR amplication efciencies (E)
were calculated for each candidate reference genes RT-qPCR assay
using the formula E = (10  1/slope  1)  100 using the slope of
the semilog regression plot of Cq versus log input of cDNA (10-fold
dilution series of ve points) [20]. PCR amplication efciencies for
each reference gene candidate are shown in Table 2.

with high quality were selected in this study. The absorbance ratio
at 260/280 nm of the RNA samples was in the range from 1.85 to
2.12 (1.9 0.045) and indicated pure and protein-free RNA. An
RNA integrity number (RIN) was generated based on the ratio of
ribosomal bands (28S/18S ratio > 1.7) on 1% agarose gels.
The amplication efciencies of RT-qPCR (E) of 12 primer pairs
ranged from 95 to 100%, and the correlation coefcients (R2) ranged from 0.995 to 0.999. The amplication specicity for each
RT-qPCR analysis was conrmed by a single peak in melt curve
analysis, and no primer dimer was detected. Furthermore, the
identities of real-time products were conrmed by a single band
with the expected size in 1.5% agarose gel electrophoresis.

Analysis of reference gene stability

To identify suitable reference genes for cervical tissue gene

expression studies in fresh-frozen clinical samples, a panel of 12
candidate genes commonly used as reference genes was selected
from the literature for analysis of stability: ALAS1, PPIA, GAPDH,
HBB, TBP, ACTIN, B2M, MBNL2, PGKL, RPL, RPL-4, and EEF1A1.
Genes were analyzed in randomly selected samples (20 tumor
samples and 20 normal samples) using RT-qPCR. The 12 candidate
control genes displayed a relatively wide range of expression level
with mean Cq values between 14.68 and 32.50. The DAgostino
Pearson tting procedure was used to conrm that all genes were
normally distributed in both malignant and nonmalignant tissue
samples. Among these genes, EEF1A1 was the most abundant transcript, with mean (standard deviation) Cq values of 16.59 0.73 in
malignant samples and 16.39 0.54 in nonmalignant samples. HBB
was the lowest expressed gene, with mean Cq values of
30.16 2.03 in malignant samples and 30.85 1.34 in nonmalignant samples. Of the 12 candidate reference genes, four (GAPDH
[P = 0.09, t test], HBB [P = 0.28, t test], RPLP0 [P = 0.48, t test], and
EEF1A1 [P = 0.48, t test]) were equivalently expressed between
the normal and tumor groups, whereas the expressions of the other
eight candidate reference genes (ALAS1, PPIA, TBP, ACTIN, B2M,
MBNL2, PGKL, and RPL-4) were signicantly increased in malignant samples (all Ps <0.001, t test). We also used Bonferroni correction to correct the P value of each gene, and the corrected P values
of these eight genes all were less than 0.05, as shown in Fig. 1.
In addition, the expressions of these 12 candidate genes did not
depend on age (R2 values = 0.1210.245, P values = 0.1021.000,
Spearmans rank test), histological type (R2 values = 0.1030.312,
P values = 0.3651.000, Spearmans rank test), FIGO stage (R2 values = 0.1360.213, P values = 0.1470.932, Spearmans rank test),
or tumor grade (R2 values = 0.0930.232, P values = 0.2540.957,
Spearmans rank test).
Because of the lower expression of HBB in cervical tissues, we
excluded it as a suitable reference gene. The stabilities of the other
11 candidate reference genes were analyzed using two statistical

Candidate reference gene stability was evaluated by using two

popular algorithms. First, geNorm (version 3.5 available online at
http://medgen.ugent.be/genorm) calculates a gene stability measure (M) as the standard deviation of the log2-transformed expression ratios of each housekeeping gene with all other tested
candidate reference genes throughout a given sample panel [20].
Genes with higher M values (>1.5) have greater variations of
expression. In addition, the assessment of the pairwise variations
(Vn/n+1) between each combination of sequential normalization factors allows the identication of the optimal number of reference
genes. Second, NormFinder uses a model-based approach to estimate expression stability based on intra- and intergroup variations
for candidate reference genes [21]. A low stability value indicates a
low combined variation and reveals high expression stability.
Statistical analyses were performed with SPSS 15.0 statistical
software, and P < 0.05 was considered as statistically signicant.
The DAgostinoPearson omnibus normality test and parametric
tests were used to determine the distribution of data. Two-sample
t tests were used to compare calibrator-scaled reference gene
quantities between nonmalignant and malignant tissue groups.
One-way analysis of variance (ANOVA) was used to compare the
means between the groups. A Bonferroni correction test was used
to correct the P value. Correlations between gene expression level
(a continuous scale variable) and tumor stage and FIGO stage as
well as histological type were characterized by the Spearmans
rank test.
Results
Quality control
All RNA samples were examined for their concentration, purity,
and integrity. To avoid erroneous conclusions, only RNA samples

Expression stability of candidate reference genes

Reference genes in human cervical tissues / Y. Shen et al. / Anal. Biochem. 405 (2010) 224229

227

Fig. 1. Expression levels of candidate reference genes in normal cervical and cervical cancer samples. Shown are quantities of candidate references in cervical cancer (n = 20)
and normal (n = 20) tissues as expressed as quantication cycle (Cq) values. Box plots (blank: cancer; cross-striated: normal) depict median lines of interquartile-range boxes.
Error bars represent range of values. No signicant difference (P > 0.05, t test) was found within ve reference genes between tumor and normal tissues (), whereas a
signicant difference (P > 0.05, Bonferroni correction, t test) was found within seven reference genes between tumor and normal tissues ().

analysis tools, geNorm and NormFinder, which employ different

statistical models to dene the most reliable reference genes for
normalization. All 11 candidate reference genes were included in
the program geNorm and ranked according to their M values. The
M values for ALAS1, PPIA, TBP, MBNL2, PGKL, GAPDH, B2M, RPLP0,
RPL-4, and EEF1A1 were lower than the geNorm default threshold
of 1.5, whereas only the one remaining gene, ACTIN, showed an M
value greater than the threshold (Table 3). PPIA and TBP (both
Ms  0.6) were identied as the two most stable genes according
to geNorm analysis (Fig. 2A). After excluding the eight genes that
differed in expression between the normal and tumor groups, the
analysis by geNorm suggested that EEF1A1 is the most stable gene,
followed by GAPDH and RPLP0. In addition, geNorm also identied
the optimal number of reference genes by assessing the pairwise
variations (Vn/n+1) between each combination of sequential normalization factors. As shown in Fig. 2B, the combination of PPIA,
TBP, ALAS1, and MBNL2 was suitable for normalizing gene expression data among 40 clinical cervical samples, yielding a V4/5 value
of 0.156, which is close to the cutoff value of 0.15. The trend of the
value became roughly stable after the addition of the fth gene.
After excluding the eight genes that differed in expression between

Table 3
Ranking and best combination of candidate reference genes based on expression
stability values calculated by NormFinder and geNorm programs.
Rank

NormFinder
Gene
name

1
2
3
4
5
6
7
8
9
10
11
Best combination

geNorm
Stability
value

EEF1A1
0.340
GAPDH
0.430
PPIA
0.488
ALAS1
0.509
PGKL
0.515
MBNL2
0.563
TBP
0.598
RPL-4
0.845
B2M
0.910
RPLP0
1.291
ACTIN
2.438
EEF1A1 and GAPDH

Gene
name

Stability
value

PPIA
TBP
ALAS1
MBNL2
PGKL
EEF1A1
RPL-4
GAPDH
B2M
RPLP0
ACTIN
PPIA and TBP

0.603
0.615
0.681
0.732
0.801
0.902
1.080
1.154
1.295
1.412
1.521

the normal and tumor groups, geNorm also determined that the
optimal number of reference genes was two (EEF1A1 and GAPDH).
The results of the analysis by NormFinder appeared to be identical to those determined using geNorm, with EEF1A1 being the
most stable gene, followed by GAPDH (Table 3). The stable value
of the best combination of two genes determined by NormFinder
for paired tissues was EEF1A1 and GAPDH.
Validation EEF1A1 as a suitable reference gene in clinical cervical
samples
To further validate the stability of EEF1A1 in clinical cervical
specimens as a reference gene, we applied a model composed of
three groups representing the full spectrum of clinic cervical samples: normal (n = 21), moderate to severe dysplasia (CIN23, n = 22),
and tumor (n = 18). The EEF1A1 gene displayed a relatively narrow
range of expression levels, with mean Cq values between 15.87 and
16.11, and there was no expression difference among normal,
CIN23, and cancer tissues (P = 0.48, ANOVA) (Fig. 3).
Discussion
Accurate data normalization is an important but underappreciated aspect of quantitative gene expression analysis [14]. Reference genes are generally employed in RT-qPCR experiments to
normalize variability between different samples. However, the pitfall of this approach is that no universal reference gene exists because many of them are varied in different conditions, which
may lead to altered ndings and wrong experimental conclusions
[6,7,16]. So, researchers need to choose the most appropriate reference gene for a given tissue or disease and, therefore, prove the
suitability of these genes in each specic experimental situation.
To our knowledge, this is the rst systematic evaluation of the
reliability of a large number of genes used as reference genes for
RT-qPCR analysis in clinic human cervical tissues. Both GeNorm
and NormFinder were used to evaluate the expression stability of
reference genes, that is, intragroup stability and intergroup stability in this study [20,21]. The HBB gene was excluded because of its
too low expression in cervical tissues. geNorm identied that the

228

Reference genes in human cervical tissues / Y. Shen et al. / Anal. Biochem. 405 (2010) 224229

Fig. 2. GeNorm analysis of the candidate reference genes. Results are presented according to the output le of the geNorm program. (A) Stepwise exclusion of the least stable
genes by calculating the average expression stability measure M. The value of M was calculated for each gene, and the least stable gene with the highest M value was
automatically excluded for the next calculation round. The x axis, from left to right, indicates the ranking of the reference genes according to expression stability, and the y
axis indicates the stability measure M. (B) Determination of the optimal number of reference genes for normalization.

other 10 candidate reference genes, except ACTIN, could be considered as stable enough for normalization purposes. PPIA and TBP
were nally identied as the two most stably expressed reference
genes. On the other hand, NormFinder indicated that EEF1A1 and
GAPDH were the two genes with the best stability, followed by
PPIA. However, the previously performed t test and equivalence
test analyses of gene expression in tumor and normal groups revealed that only the EEF1A1, RPLP0, and GAPDH genes showed
intergroup stability. After excluding eight genes that differed in
expression between the normal and tumor groups, the analysis

by geNorm suggested that EEF1A1 is the most stable gene, followed by GAPDH and RPLP0. Thus, we could conclude that only
three genes (EEF1A1, GAPDH, and RPLP0) fullled the criterion of
expression stability (both intra- and intergroup) and that EEF1A1
can be considered as the most suitable RT-qPCR normalizer for relative gene quantication in cervical samples.
To further validate the stability of EEF1A1 in cervical specimens,
we applied a model composed of three groups representing the full
spectrum of cervical lesions: normal, CIN23, and invasive cancer.
Finally, the expression of EEF1A1 was nearly unchanged among

Reference genes in human cervical tissues / Y. Shen et al. / Anal. Biochem. 405 (2010) 224229

16.4

229

Acknowledgments

16.2
16
15.8
15.6

We acknowledge nancial support by grants from the National

Natural Science Foundation of China (30672229 and 30672230)
and the Special Research Fund for the Doctoral Program of Higher
Education of China (20070335054).

15.4
15.2

References

15
14.8
Fig. 3. EEF1A1 gene expression in different cervical lesions. The x axis, from left to
right, indicates a spectrum of clinic cervical samples (normal, CIN23, and cancer),
and the y axis indicates real-time PCR cycle threshold numbers (Cq values). The
EEF1A1 gene expression level was not signicantly different among normal, CIN23,
and cancer tissues (P = 0.48, ANOVA).

normal, CIN23, and cancer tissues (P = 0.48, ANOVA), further identifying EEF1A1 as the most suitable reference gene in cervical tissue samples.
Regarding the optimal number of reference genes to be used for
calculating a normalization factor, some researchers recently also
advocated using multiple reference genes rather than a single
RNA transcript [4,22]. This is a robust method for providing accurate normalization and is consequently favorable if ne measurements are to be made. In the current study, we found that the
combination of PPIA, TBP, ALAS1, and MBNL2 was suitable for normalizing gene expression data among 40 cervical samples, yielding
a V4/5 value of 0.156, which is close to the cutoff value 0.15. However, these four genes differ in their expression between the normal
and tumor groups, so it is not consistent with the criterion of intergroup expression stability. After excluding the eight genes that differed in expression between the normal and tumor groups, the
analysis by geNorm suggested that the optimal number of reference
genes was two (EEF1A1 and GAPDH). The results of the analysis by
NormFinder appeared to be identical to those determined using
geNorm. Therefore, EEF1A1 is a reliable gene for normalizing data
generated from RT-qPCR analysis in cervical tissue. In addition,
EEF1A1 and GAPDH are also the best two-gene combination.
It has been reported that the expression of candidate reference
genes can be inuenced by several physiological or pathological
conditions such as tissue types, age, tumor stage, and tumor grade
[23]. Our data proved that the expression of these genes in this cohort was independent of age, FIGO stage, and tumor grade. However, considering the limited size of the cohort of patients in this
study, a denite correlation between gene expression and clinic
pathological features should be further determined in a larger sample. In addition, a few studies have suggested that the expression of
candidate reference genes in cervical samples can also be inuenced by HPV infection status [19]. Nevertheless, as a limitation
of our study, we selected HPV-positive cervical cancer, HPV-positive CIN23, and HPV-negative normal cervical samples as representatives in the full spectrum of clinical cervical samples in this
study because nearly 100% of cervical cancer patients and approximately 8090% of CIN23 patients were HPV positive in the clinic
[12]. The applicability of the three recommended reference genes
(EEF1A1, GAPDH, and RPLP0) in HPV-negative cancer and CIN samples was not tested, and further studies are needed to conrm their
potential use.
In conclusion, our current study has demonstrated that EEF1A1
is the most stable gene that can be used as a reference gene for normalization in gene proling studies in HPV-positive clinical cervical samples (cancer and preinvasive cervical neoplasia
specimens) and that the combination of two genes (EEF1A1 and
GAPDH) should be recommended as a much more reliable normalization strategy.

[1] T. Nolan, R.E. Hands, S.A. Bustin, Quantication of mRNA using real-time PCR,
Nat. Protoc. 1 (2006) 15591582.
[2] L. Gutierrez, M. Mauriat, J. Pelloux, C. Bellini, V.O. Wuytswinkel, Towards a
systematic validation of references in real-time RT-PCR, Plant Cell 20 (2008)
17341735.
[3] C.L. Andersen, J.L. Jensen, T.F. Orntoft, Normalization of real time quantitative
reverse transcription-PCR data: a model based variance estimation approach
to identify genes suited for normalization, applied to bladder and colon cancer
data sets, Cancer Res. 64 (2004) 52455250.
[4] J. Vandesompele, K. De Preter, F. Pattyn, B. Poppe, N. Van Roy, A. De Paepe,
Accurate normalization of real-time quantitative RT-PCR data by geometric
averaging of multiple internal control genes, Genome Biol. 3 (2002)
research0034.1research0034.11.
[5] V.R. Cicinnati, Q. Shen, G.C. Sotiropoulos, A. Radtke, G. Gerken, S. Beckebaum,
Validation of putative reference genes for gene expression studies in human
hepatocellular carcinoma using real-time quantitative RT-PCR, BMC Cancer 8
(2008) 350355.
[6] S.A. Bustin, T. Nolan, Pitfalls of quantitative real-time reverse transcription
polymerase chain reaction, J. Biomol. Technol. 15 (2004) 155166.
[7] J. Huggett, K. Dheda, S. Bustin, A. Zumla, Real-time RT-PCR normalization:
strategies and considerations, Genes Immun. 6 (2005) 279284.
[8] Y.L. Li, F. Ye, Y. Hu, W.G. Lu, X. Xie, Identication of suitable reference genes for
gene expression studies of human serous ovarian cancer by real-time
polymerase chain reaction, Anal. Biochem. 394 (2009) 110116.
[9] E.M. Roisin, M. Nicola, J.K. Michael, Evaluation and validation of candidate
endogenous control genes for real-time quantitative PCR studies of breast
cancer, BMC Mol. Biol. 8 (2007) 107120.
[10] D.M. Parkin, C.S. Muir, Cancer incidence in ve continents: comparability and
quality of data, IARC Sci. Publ. 120 (1992) 45173.
[11] F.X. Bosch, M.M. Manos, N. Munoz, M. Sherman, A.M. Jansen, J. Peto, Prevalence
of human papillomavirus in cervical cancer: a worldwide perspective, J. Natl.
Cancer Inst. 87 (1995) 796802.
[12] N. Muoz, F.X. Bosch, S. de Sanjos, R. Herrero, X. Castellsagu, K.V. Shah, P.J.
Snijders, C.J. Meijer, For the International Agency for Research on Cancer
Multicenter Cervical Cancer Study Group, Epidemiologic classication of
human papillomavirus types associated with cervical cancer, N. Engl. J. Med.
348 (2003) 518527.
[13] C.M. Martin, K. Astbury, L. McEvoy, S. OToole, O. Sheils, J.J. OLeary, Gene
expression proling in cervical cancer: identication of novel markers for
disease diagnosis and therapy, Methods Mol. Biol. 511 (2009) 333359.
[14] C.M. Martin, L. Kehoe, C.O. Spillane, J.J. OLeary, Gene discovery in cervical
cancer: towards diagnostic and therapeutic biomarkers, Mol. Diagn. Ther. 15
(2007) 277290.
[15] F. Haller, B. Kulle, S. Schwaqer, B. Gunawan, A. von Heydebreck, H. Sltmann, L.
Fzesi, Equivalence test in quantitative reverse transcription polymerase chain
reaction: conrmation of reference genes suitable for normalization, Anal.
Biochem. 335 (2004) 19.
[16] K. Dheda, J.F. Huggett, J.S. Chang, L.U. Kim, S.A. Bustin, M.A. Johnson, G.A. Rook,
A. Zumla, The implications of using an inappropriate reference gene for realtime reverse transcription PCR data normalization, Anal. Biochem. 344 (2005)
141143.
[17] N. Janssens, M. Janicot, T. Perera, A. Bakker, Housekeeping genes as internal
standards in cancer research, Mol. Diagn. 8 (2004) 107113.
[18] M. Steinau, M.S. Rajeevan, E.R. Unger, DNA and RNA references for RT-qPCR
assays in exfoliated cervical cells, J. Mol. Diagn. 8 (2006) 113118.
[19] I.I. Daud, M.E. Scott, Validation of reference genes in cervical cell samples from
human papillomavirus-infected and -uninfected women for quantitative
reverse transcription-PCR assays, Clin. Vaccine Immunol. 15 (2008) 1369
1373.
[20] J. Vandesompele, K. De Preter, F. Pattyn, B. Poppe, N. Van Roy, A. De Paepe, F.
Speleman, geNorm software manual, update 6 2004, http://medgen.ugent.be/
~jvdesomp/genorm.
[21] NormFinder software, http://www.mdl.dk/publicationsnormnder.htm.
[22] C. Tricarico, P. Pinzani, S. Bianchi, M. Paglierani, V. Distante, M. Pazzagli, S.A.
Bustin, C. Orlando, Quantitative real time reverse transcription polymerase
chain reaction: normalization to rRNA or single housekeeping genes is
inappropriate for human tissue biopsies, Anal. Biochem. 309 (2002) 293300.
[23] J.B. de Kok, R.W. Roelofs, B.A. Giesendorf, J.L. Pennings, E.T. Waas, T. Feuth,
D.W. Swinkels, P.N. Span, Normalization of gene expression measurements in
tumor tissues: comparison of 13 endogenous control genes, Lab. Invest. 85
(2005) 154159.