Rating of putative housekeeping genes for quantitative gene expression analysis in cyclic and early pregnant equine endometrium

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Animal Reproduction Science 125 (2011) 124– 132

Contents lists available at ScienceDirect

Animal Reproduction Science

journal homepage: www.elsevier.com/locate/anireprosci

Rating of putative housekeeping genes for quantitative geneexpression analysis in cyclic and early pregnant equine endometrium

Seyit A. Kayisa, Mehmet O. Atlib, Ercan Kurarc, Faruk Bozkayad, Ahmet Semacane,Selim Aslanf, Aydin Guzeloglue,∗

a Biometry-Genetics Unit, Department of Animal Science, Faculty of Agriculture, Selcuk University, Konya, Turkeyb Department of OBGYN, Faculty of Veterinary Medicine, Dicle University, Diyarbakir, Turkeyc Department of Genetics, Faculty of Veterinary Medicine, Selcuk University, Konya, Turkeyd Department of Genetics, Faculty of Veterinary Medicine, Harran University, Sanliurfa, Turkeye Department of OBGYN, Faculty of Veterinary Medicine, Selcuk University, Konya, Turkeyf Department of OBGYN, Faculty of Veterinary Medicine, Ankara University, Ankara, Turkey

a r t i c l e i n f o

Article history:Received 12 October 2010Received in revised form 3 January 2011Accepted 10 February 2011Available online 18 February 2011

Keywords:Housekeeping geneMixed modelgeNormNormFinderBestKeeperEquine endometrium

a b s t r a c t

The aim was an evaluation of a set of housekeeping genes (HKGs) to be used in the nor-malization of gene expression in the equine endometrium. Glyceraldehyde-3-phosphatedehydrogenase (GAPDH), hypoxanthine ribosyl transferase 1 (HPRT1), ubiquitin B (UBB),tubulin alpha 1 (TUBA1), ribosomal protein L32 (RPL32), beta-2-microglobulin (B2M), 18SrRNA (18S), and 28S rRNA (28S) HKGs were evaluated using real-time PCR and were com-pared in different physiological stages of the endometrium. Endometrial biopsies wereobtained from mares on day of ovulation (d0, n = 4), at late diestrus (LD, n = 4), after lute-olyis (AL, n = 4) of the cycle and on days 14 (P14; n = 3), 18 (P18, n = 3) and 22 (P22; n = 3)of pregnancy. A model based on REML with support of descriptive statistics was proposedin accordance with experimental design and was further confirmed with principal compo-nent analysis (PCA). Results were compared with widely used software including geNorm,BestKeeper, and NormFinder. Results indicated that GAPDH was the most stable HKG andRPL32 was ranked as the second best. 18S and 28S were found to be the least stable. Theproposed model, PCA, geNorm, and BestKeeper were in agreement in detecting the moststable and the least stable HKGs in the equine endometrium during the estrous cycle andearly pregnancy.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Real-time RT-PCR is a very sensitive method for thedetection of low abundance mRNAs (Bustin, 2000, 2002;Draghici et al., 2006; Pusterla et al., 2006) and can be usedfor different applications such as clinical diagnosis (Dantaset al., 2009; Solassol et al., 2010) and analysis of tissue-specific gene expressions (Bustin, 2000; Soria-Guerra et al.,2010).

∗ Corresponding author. Tel.: +90 332 2233597; fax: +90 332 2410063.E-mail address: [email protected] (A. Guzeloglu).

When comparing gene expressions in different samples,it is essential to take into account experimental varia-tions such as amount of starting material, RNA quality,reverse transcription and PCR efficiencies. This is achievedby normalization of gene of interest to an internal con-trol that is commonly named as housekeeping genes(HKGs) (Karge et al., 1998; Dheda et al., 2004). HKGs areexpected to be stably expressed at the mRNA level underany experimental and physiological conditions. However,many studies showed that there is no unique universal HKGthat could be employed for normalization of gene expres-sion under different experimental conditions (Thellin et al.,1999; Warrington et al., 2000; Sturzenbaum and Kille,

0378-4320/$ – see front matter © 2011 Elsevier B.V. All rights reserved.doi:10.1016/j.anireprosci.2011.02.019

Author's personal copy

S.A. Kayis et al. / Animal Reproduction Science 125 (2011) 124– 132 125

2001; Radonic et al., 2004; Walker et al., 2009). Inaccu-rate normalization results in inadequate quantification andspurious conclusions.

Uterus is a very dynamic organ governed at least bythe circulating steroid hormones such as estradiol from theovarian follicle and progesterone from the corpus luteumon the ovary. When the uterus is under the influence ofprogesterone, it is in the secretory phase while estradiolrules the proliferative phase. Therefore, gene expressionprofile is necessarily different between these two phases.Furthermore, a developing embryo inside the uterus fol-lowing the mating will also add another variable thataffects gene expressions in the uterine endometrium. Dur-ing the early pregnancy, there are even differences ona daily basis on the expression of specific genes evalu-ated before (Atli et al., 2010). As the uterine endometriumis under the influence of so many factors, studies com-paring differential gene expression among the days ofan estrous cycle and pregnancy and between pregnantand nonpregnant uterus should employ a stable andreliable HKG. A proper HKG must be stably and indepen-dently expressed from hormonal and embryonic effects.For example, previous studies demonstrated that estro-gen and progesterone could regulate expression of HKGsin mouse uterus (Schroder et al., 2009; Craythorn et al.,2009). Therefore, there is a need to determine a referencecontrol gene, which can be used under these physiologicalconditions in the equine endometrium.

Previous studies regarding evaluation of HKGs as inter-nal control in quantitative RT-PCR evaluated a set ofcandidate HKGs for gene expression studies in severalequine tissues (Zhang et al., 2009) and normal equine skinand sarcoids (Bogaert et al., 2006). Therefore, HKGs thatwere indicated as the most appropriate in the studies byZhang et al. (2009) and Bogaert et al. (2006) were employedalong with GAPDH, that was previously used in equineendometrium studies by Boerboom et al. (2004).

The objectives of this study were: (1) to evaluate aset of putative HKGs that could be used in the accu-rate quantitative evaluation of gene expressions in equineendometrium during the estrous cycle and early preg-nancy, (2) to compare results obtained from different HKGdetection methods and softwares, which are freely avail-able.

2. Materials and methods

2.1. Materials

TRIzol (TRIzol® Reagent, Invitrogen, Carlsbad, CA, USA)was obtained. cDNA Synthesis Kit (RevertAidTM First StrandcDNA Synthesis Kit), RNAse-free DNAse I, qPCR MasterMix (2×) for Real-time PCR (MaximaTM SYBR), dNTP setand Taq DNA polymerase were purchased from FermentasLife Sciences (Glen Burnie, MD, USA). Human chori-onic gonadotrophin (Chorulon; Intervet Inc, Boxeer, TheNetherlands) and PGF2� (Dinolytic®, Pfizer Manufacturing,Puurs, Belgium) were obtained. Specific oligonucleotideprimers were synthesized by Metabion International AG(Martinsried, Germany). DNAse and RNAse free sterile0.2 ml-tubes (Thermo Fisher, Waltham, MA, USA) and

1.5 ml-tubes (TreffLab, Degersheim, Switzerland) wereobtained. All other chemicals were obtained from Merck(Darmstadt, Germany).

2.2. Animals and experimental design

Ten Arabian mares and one Arabian stallion (aged from5 to 16 and weighed between 400 and 450 kg) were used.Mares were placed individually in the Equestrian Centerin the Faculty of Veterinary Medicine at Selcuk University,Turkey, provided haylage and water ad libitum, and werefed with grain pellets twice a day. Before the experiment,the genital tracts of mares were evaluated by rectal palpa-tion, ultrasonography, caslick index score, uterine culture,smear and biopsy inspections. Fertility examination of thestallion was performed by spermatological examinationand inspection of external genital tract. The biopsy scoresof mares were I or IIA according to Kenney and Doig (1986)classification. All experimental procedures were approvedby the Ethics Committee of Faculty of Veterinary Medicineat Selcuk University.

The complete experimental design was describedin detail previously (Atli et al., 2010). Briefly, mareswere either inseminated and used from pregnancy sam-ple collection group or not inseminated and used forcyclic/nonpregnant sample collection group. Ultrasonogra-phy was employed every 12 h until ovulation was detectedand the day of ovulation was designated as day zero (d0).Endometrial biopsies were obtained from mares on daysof ovulation (d0, n = 4), at late diestrus (LD, mean of theday 13.5–14, n = 4, high progesterone, Fig. 1A), after lute-olysis in estrus phase (AL, mean of the day 17.5–18, n = 4,low progesterone, Fig. 1B) of the estrous cycle and on days14 (P14; n = 3), 18 (P18, n = 3), and 22 of pregnancy (P22;n = 3). Pregnancy was diagnosed by the presence of viableembryo, which was defined by observation of non-echoicembryonic vesicle at least two days before biopsy and heartbeats of embryo on day 22 by means of ultrasonography.In an estrous cycle or pregnancy, only one biopsy samplingwas performed. Following biopsy, mares were rested forrecovery for one cycle period. Within a cyclic or pregnantgroup, biopsy samples were taken from different mares. Inany given sampling day, a mare was used only once, there-fore for each time point there were four different mares.Only one biopsy sample was obtained from each mare ina given cycle or pregnancy. Endometrial tissue was snapfrozen immediately in liquid nitrogen and stored at −70 ◦Cuntil RNA isolation.

2.3. RNA extraction and cDNA synthesis

Fifty mg of endometrial tissue was minced with ascalpel and was homogenized in TRIzol and total RNAextraction was performed according to the manufacturer’sprotocol. RNA purity was controlled with optic density of260/280 (2.0 ± 0.1) with NanoDrop ND-100 (Thermo Scien-tific, Wilmington, DE, USA) and RNA integrity was verifiedby agarose gel electrophoresis. Two �g of total RNA wastreated with DNAse I to clean genomic DNA contamina-tion and was then reverse transcribed in the presence ofboth random hexamer and oligodT primers in equal vol-

Author's personal copy

126 S.A. Kayis et al. / Animal Reproduction Science 125 (2011) 124– 132

0

a

b

1

2

3

4

5

6

7

8

-48 h -36 h -24 h -12 h 0

Pro

gest

eron

e (n

g/m

l)

LD

0

1

2

3

4

5

6

7

8

-48 h -36 h -24 h -12 h 0

Pro

gest

eron

e (n

g/m

l)

AL

Fig. 1. Plasma progesterone concentrations were measured two daysbefore every 12 h from biopsy sampling in LD (A) and AL (B) of cyclicmares.Taken from Atli et al. (2010).

ume by using RevertAidTM First Strand cDNA Synthesis Kitaccording to the manufacturer’s protocol.

2.4. Primers and real time PCR

Glyceraldehyde-3-phosphate dehydrogenase (GAPDH),hypoxanthine ribosyl transferase 1 (HPRT1), ubiquitin B(UBB), tubulin alpha 1 (TUBA), ribosomal protein L32

(RPL32), beta-2-microglobulin (B2 M), 18S rRNA (18S), and28S rRNA (28S) primers were designed according to primersequences published previously (Boerboom et al., 2004;Bogaert et al., 2006; Zhang et al., 2009) (Table 1). Realtime PCR was set up as follow: 10 �l SYBR Green MasterMix (2×), 5 pMol of each primer, 1 �l cDNA and ddH2O upto 20 �l of final volume. Thermal cyclic conditions wereinitial denaturation at 95 ◦C for 10 min followed by 40cycles of denaturation, annealing and amplification (95 ◦C30 s, 60 ◦C 1 min, 72 ◦C 30 s) on a Mx3005PTM 3005 Real-Time PCR System (Agilent Technologies Inc., Santa ClaraCA, USA). Melting curve analysis was performed as follow:95 ◦C 1 min, then fluorescence measurement was done atevery 1◦ increments between 55 ◦C and 95 ◦C. In each run,a negative control with no cDNA template was included.To verify reaction specificity, amplification products wererun on 2% agarose gel. From the RNA extraction to thereal-time PCR, whole procedure was performed twice astechnical replicate. Data used were Ct values obtained fromthe qRT-PCR analysis based on SYBR Green detection. Thespecificity of all qRT-PCRs was confirmed via melting curveanalysis of amplification products (Fig. 2). Primer efficiencywas obtained according to Schefe et al. (2006) by using 5data points of log transformed fluorescence graph of theexponential phase of the PCR kinetic curve.

2.5. Statistical approach for HKG selection

Several methods were proposed for selection of bestHKGs. Vandesompele et al. (2002) utilized variation (i.e.standard deviation of pairwise log2-transformed expres-sion ratios) and proposed gene-stability measure Mj forcontrol gene j as the arithmetic mean of all pairwise varia-tions when 10 HKGs were compared via the gene-stabilitymeasure. Andersen et al. (2004) proposed a model to eval-uate systematic variation across the sample subgroupsbeside overall expression variation. Pfaffl et al. (2004) pro-posed an index method to combine multiple HKGs andevaluated each candidate HKG expression stability basedon standard deviation (SD) and coefficient of variance(CV). Rubie et al. (2005) utilized fold differences to com-pare expression level of 21 HKGs in human malignant and

Table 1List of primers used.

Gene Gen Bank accessioncode

Primer (5′–3′) Ampliconlength (bp)

Reference

18S AJ311673 F: ATGCGGCGGCGTTATTCCR: GCTATCAATCTGTCAATCCTGTCC

204 Zhang et al. (2009)

28S EU554425 F: CGGGTAAACGGCGGGAGTAACR: TAGGTAGGGACAGTGGGAATCTCG

109 Zhang et al. (2009)

B2M X69083 F: GGCTACTCTCCCTGACTGGR: ACACGGCAACTATACTCATCC

271 Zhang et al. (2009)

HPRT1 AY372182 F: GAGATGTGATGAAGGAGATGR: TGACCAAGGAAAGCAAGG

300 Zhang et al. (2009)

UBB AF506969 F: GCAAGACCATCACCCTGGAR: CTAACAGCCACCCCTGAGAC

206 Bogaert et al. (2006)

TUBA1 AW260995 F: GCCCTACAACTCCATCCTGAR: ATGGCTTCATTGTCCACCA

78 Bogaert et al. (2006)

RPL32 CX594263 F: AGCCATCTACTCGGCGTCAR: TCCAATGCCTCTGGGTTTC

149 Bogaert et al. (2006)

GAPDH NM 001163856 F:ATCACCATCTTCCAGGAGCGAGAR:GTCTTCTGGGTGGCAGTGATGG

341 Boerboom et al. (2004)

Author's personal copy

S.A. Kayis et al. / Animal Reproduction Science 125 (2011) 124– 132 127

70 75 80 85 90

050

010

0015

0020

0025

0030

00

Temperature (Celsius) Temperature (Celsiu s) Temperature (Celsius)

Temperature (Celsius) Temperature (Celsiu s) Temperature (Celsius)

Temperature (Celsius) Temperature (Celsius )

Fluo

resc

ence

(-R

'(T))

18S

70 75 80 85 90

050

010

0015

0020

0025

0030

00

Fluo

resc

ence

(-R

'(T))

28S

70 75 80 85 90

050

010

0015

0020

0025

0030

00

Fluo

resc

ence

(-R

'(T))

B2M

70 75 80 85 90

010

0020

0030

00

Fluo

resc

ence

(-R

'(T))

HPRT1

70 75 80 85 90

050

010

0015

0020

0025

00

Fluo

resc

ence

(-R

'(T))

TUBA1

70 75 80 85 90

050

010

0015

0020

0025

0030

0035

00

Fluo

resc

ence

(-R

'(T))

RPL32

70 75 80 85 90

050

010

0015

0020

0025

0030

00

Fluo

resc

ence

(-R

'(T))

UBB

70 75 80 85 90

050

010

0015

0020

0025

0030

00

Fluo

resc

ence

(-R

'(T))

GAPDH

Fig. 2. Graph of melting curve analysis of amplification products. The rate of change (first derivative) of the fluorescence units in time (−R′(T)) peaks at themelting temperature (Tm). Similar peaks indicate that there is no contamination, mispriming, primer–dimer artifact, etc.

Author's personal copy

128 S.A. Kayis et al. / Animal Reproduction Science 125 (2011) 124– 132

Table 2Mean and standard deviations (±SD) of Ct values of genes experimental days and overall days.

Genes Experimental days Min.SD

Max.SD

SDrange

d0 LD AL P14 P18 P22 Overall

18S 14.13 (±1.63) 13.60 (±0.85) 14.82 (±0.81) 14.02 (±0.60) 14.59 (±0.32) 14.02 (±0.47) 14.19 (±0.93) 0.32 1.63 1.3128S 10.58 (±0.76) 10.26 (±0.90) 10.99 (±0.86) 11.12 (±0.64) 11.88 (±0.39) 11.26 (±0.33) 10.96 (±0.81) 0.33 0.9 0.57B2M 19.88 (±0.32) 19.57 (±0.43) 19.19 (±0.46) 19.82 (±0.87) 19.93 (±0.43) 20.04 (±0.69) 19.71 (±0.55) 0.32 0.87 0.55HPRT1 27.37 (±0.84) 26.59 (±0.25) 26.32 (±0.20) 27.10 (±0.60) 27.47 (±0.58) 27.62 (±0.86) 27.03 (±0.71) 0.20 0.86 0.66TUBA1 25.56 (±0.58) 25.21 (±0.42) 24.32 (±0.42) 25.17 (±1.21) 24.88 (±0.33) 23.87 (±0.49) 24.86 (±0.79) 0.33 1.21 0.88RPL32 22.99 (±0.64) 23.02 (±0.25) 23.18 (±0.20) 22.88 (±0.59) 23.23 (±0.37) 22.74 (±0.47) 23.02 (±0.42) 0.20 0.64 0.44UBB 20.89 (±0.75) 20.40 (±0.44) 20.72 (±0.14) 20.89 (±0.71) 21.01 (±0.62) 20.84 (±0.49) 20.77 (±0.52) 0.14 0.75 0.61GAPDH 20.45 (±0.26) 20.44 (±0.44) 19.91 (±0.39) 20.44 (±0.47) 20.68 (±0.14) 20.42 (±0.29) 20.39 (±0.39) 0.14 0.47 0.33

normal tissues. de Kok et al. (2005) used a principal compo-nent analysis (PCA) approach to group housekeeping geneswith similar expression variation patterns for compari-son of 13 HKGs in different human tissues (colon, breast,prostate, skin, bladder). Zhang et al. (2009) employed one-way ANOVA and Kruskal–Wallis tests to compare stabilityof 6 HKGs expressions in equine colon, heart, kidney, liver,lung, lymph node, small intestine and spleen.

Expectation from a HKG is stability when the organismis exposed to different external effects. In general, a linearmodel with a common mean across treatment groups anda high coefficient of determination (R2) that implies havingsmall standard deviation (SD) within and across groups isimagined as a measure of stability. This can be expressedas:

yj = � + ej (1)

where yj = threshold cycles (Ct) value of jth individual,� = over-all mean, ej = residual term for jth individual.

Eq. (1) implies that there is no statistically significantdifference between the group (experimental day) means.However, if any two group means are statistically signifi-cantly different, then Eq. (1) can be extended as below totake into account group effects:

yij = � + ˛i + eij (2)

where yij = Ct value of jth individual which is placed inith group, � = over-all mean, ˛i = fixed effect of ith group,eij = residual term for jth individual which is placed in theith group.

We have further extended Eq. (2) as below accordingto our experimental design to take into account technicalreplicate:

yijk = � + ˛i + mj(˛i) + eijk (3)

where yijk = Ct value of the kth technical replicate of jthmare which is nested in the ith group, � = overall mean,˛i = fixed effect of ith group, mj(˛i) = random effect of jthmare nested in the ith group, eijk = residual for kth technicalreplicate of jth mare which is nested in the ith group.

In this study, first step of HKG selection was per-formed according to model fitting in Eq. (3). Any HKG withstatistically significantly different group means was elim-inated. Then, remaining HKGs were ranked according tofollowing descriptive statistics (SD, SD range [=maximum(SD) − minimum (SD)], and CV) using average of dupli-cates of each sample. The second step of evaluation was

performed according to rank of remaining HKGs. For fur-ther confirmation, principal component analysis (PCA) wasemployed. All statistical analyses were carried out by usingGenStat (Release 7, Payne et al., 2003).

Finally, the data were also analysed by methods ofVandesompele et al. (2002) [geNorm; http://medgen.ugent.be/∼jvdesomp/genorm], Andersen et al. (2004)[NormFinder; http://www.mdl.dk/publicationsnormfinder.htm], and Pfaffl et al. (2004) [BestKeeper; http://www.gene-quantification.de/bestkeeper.html] and the resultswere compared.

3. Results

Eight different HKGs were amplified in equineendometrium during the experimental days of estrouscycle and early pregnancy by using specific primers(Table 1) and their expression stabilities were evalu-ated by using different statistical approaches. Resultsfrom analyses according to Eq. (3) showed that out of 8HKGs only the group means of TUBA were statisticallysignificantly different (P = 0.024) and it was excludedfrom further analysis in the proposed model. Descriptivestatistics including mean and standard deviation of Ctvalues for each experimental day and overall experimentaldays (Table 2) and CV (Table 3) for each candidate HKGwere presented.

The 28S had the lowest mean Ct value (10.96) while theHPRT had the highest (27.03). The lowest CV was obtainedfrom RPL32 while the second lowest CV value was obtainedfrom GAPDH. The 28S and 18S had the highest CV. ThegeNorm detected that all these HKGs had stable expressionlevels. Among 8 HKGs, the most stable genes were GAPDH(M = 0.488) and RPL32 (M = 0.488), while the least stablegenes were 18S (M = 0.886) and TUBA (M = 0.846, Table 4).

The BestKeeper determined that all candidate HKGs aresuitable for using as reference control gene and RPL32was found to be the best HKG (SD = 0.32, CV = 1.38) whileGAPDH was the second (SD = 0.33, CV = 1.62, Table 4). The18S (SD = 0.70, CV = 4.95) and 28S (SD = 0.67, CV = 6.14)found to be the least stable genes (Table 4). The NormFinderdetected UBB as the best stably expressed gene (stabil-ity value = 0.013) while the best combination of two geneswere UBB and GAPDH (stability value = 0.013). On the otherhand, 28S was found to be the least stable gene (stabilityvalue = 0.040).

Author's personal copy

S.A. Kayis et al. / Animal Reproduction Science 125 (2011) 124– 132 129

Table 3Coefficient of variation (CV [%]) of Ct values of genes in experimental days and overall days.

Genes Experimental days

d0 LD AL P14 P18 P22 Overall

18S 11.53 6.28 5.50 4.28 2.21 3.32 6.5328S 7.22 8.76 7.85 5.72 3.30 2.97 7.41B2M 1.58 2.22 2.41 4.38 2.14 3.44 2.81HPRT1 3.06 0.93 0.77 2.22 2.09 3.11 2.64TUBA1 2.27 1.65 1.73 4.80 1.34 2.06 3.19RPL32 2.80 1.10 0.87 2.57 1.61 2.07 1.82UBB 3.59 2.14 0.70 3.41 2.94 2.34 2.51GAPDH 1.25 2.13 1.95 2.28 0.68 1.41 1.91

4. Discussion

Expression stabilities of 8 candidate HKGs (Table 1)in equine endometrium during the experimental days ofestrous cycle and early pregnancy were evaluated by usingdifferent statistical approaches. The smallest values formaximum SD, SD range, and overall SD were obtainedfrom GAPDH, while the second smallest values for thosethree descriptive statistics were obtained from RPL32. Thehighest values for maximum SD, SD range, and overall SDwere obtained from 18S and the second highest overall SDwas observed in 28S. The lowest CV was obtained fromRPL32 while the second lowest CV value was obtained fromGAPDH. The 28S and 18S had the highest CV.

Here, we note that SD would be more reliable than CVsince SD of Ct does not increase with increasing mean Ctvalue. To facilitate visual comparison, 95% confidence inter-val (95% CI) of all candidate HKGs for all experimental daysare given in Fig. 3.

TUBA was eliminated in the first step of evaluation(Table 4). In the second step, out of remaining seven HKGs,the best HKG was selected with the support of descriptivestatistics. Thus, it can be suggested that GAPDH used in thisstudy as the most stable HKG for evaluating gene expres-sions in equine endometrium while RPL32 was ranked asthe second best HKG.

Results from principal component analysis (PCA)showed that first two principal components (PCs)explained 72.6% of total variation (PC1: 51.7%, PC2: 20.9%).Minimum contributions to the PC1 were made by GAPDHand RPL32 (loadings were 0.120 and 0.173, respectively).The 18S and 28S made maximum contributions to the PC1(loadings were 0.566 and 0.514, respectively) while TUBA

Table 4Ranks of HKGs (the best [1] to the least [8]) according to different methods.

Genes Ranks

Proposedmodel

PCA (PC1) geNorm BestKeeper NormFinder

18S 7 8 8 8 728S 6 7 6 7 8B2M 4 3 3 4 5HPRT1 5 6 5 5 2TUBA1 Eliminated 4 7 6 6RPL32 2 2 1 1 4UBB 3 5 4 3 1GAPDH 1 1 1 2 3

and 18S made the largest contributions to the PC2 (load-ings were 0.571 and -0.518, respectively). PCA is a lineardimension reduction technique. PCs are linear combina-tions of the original variables and ordered with respectto their variance. Thus, the first PC contains the largestvariance while the last PC contains the smallest variance.Loadings of PCs are the coefficients for variables and canbe interpreted as contribution weights of the original vari-ables (HKGs) when calculating the principal components.Therefore, having smallest loadings for GAPDH and RPL32in PC1 while having largest loadings for 18S and 28S arevery well aligned with the results from descriptive statis-tics supporting that both methods produce similar resultsin terms of HKGs selection process (Table 4).

The results were also analysed with widely used soft-wares such as geNorm, BestKeeper and GeneFinder. ViageNorm, expression stability values, M, of all HKGs wereless than 1.5 indicating that these candidate genes havestable expression levels. Among 8 HKGs, the most sta-ble genes were GAPDH (M = 0.488) and RPL32 (M = 0.488),while the least stable genes were 18S (M = 0.886) and TUBA(M = 0.846, Table 4). These results are in agreement with theresults of proposed model where GAPDH and RPL32 werethe most stable HKGs. Although TUBA was eliminated inthe proposed method as its group means were statisticallydifferent, geNorm included TUBA as an acceptable HKG.

BestKeeper suggest that a HKG with a SD value lessthan 1 is stably expressed while any studied gene with theSD higher than 1 is considered to be inconsistent (Pfafflet al., 2004). Consequently, all candidate HKGs in thisstudy passed the criteria to be a HKG similarly to geNorm.RPL32 was found to be the best HKG (SD = 0.32, CV = 1.38)while GAPDH was the second (SD = 0.33, CV = 1.62, Table 4).The 18S (SD = 0.70, CV = 4.95) and 28S (SD = 0.67, CV = 6.14)found to be the least stable genes (Table 4). Here, we notethat Pfaffl et al. (2004) calculate SD as average of absolutevalue of deviates of each sample’s Ct from their arithmeticmean. Consequently, SD and CV of Pfaffl et al. (2004) differfrom our SD and CV results (Tables 2 and 3).

In contrast to the results mentioned above, NormFinderdetected UBB as the best stably expressed gene (stabil-ity value = 0.013) while the best combination of two geneswere UBB and GAPDH (stability value = 0.013). On the otherhand, 28S was found to be the least stable gene (sta-bility value = 0.040). The significant differences in findingof the most appropriate HKGs between NormFinder andother methods are due to the algorithms used. Apart from

Author's personal copy

130 S.A. Kayis et al. / Animal Reproduction Science 125 (2011) 124– 132

1015

2025

30

Ct

d0 LD AL P14 P18 P22

18S Mean Ct and 95 % CI

1015

2025

30

Ct

d0 LD AL P14 P18 P22

28S Mean Ct and 95 % CI

1015

2025

30

Ct

d0 LD AL P14 P18 P22

B2M Mean Ct and 95% CI10

1520

2530

Ct

d0 LD AL P14 P18 P22

HPRT1 Mean Ct and 95 % CI

1015

2025

30

Ct

d0 LD AL P14 P18 P22

TUBA1 Mean Ct and 95 % CI

1015

2025

30

Ct

d0 LD AL P14 P18 P22

RPL3 2 Mean Ct and 95% CI

1015

2025

30

Ct

d0 LD AL P14 P18 P22

UBB Mean Ct and 95 % CI

1015

2025

30

Ct

d0 LD AL P14 P18 P22

GAPDH Mean Ct and 95 % CI

Fig. 3. Mean Ct values and 95% CI of candidate HKGs in different stages. Dotted horizontal bar shows overall mean Ct value for each candidate HKG.

Author's personal copy

S.A. Kayis et al. / Animal Reproduction Science 125 (2011) 124– 132 131

NormFinder, we also note that there were slight differencesamong the methods used in this study. These are due to thealgorithms used in different methods.

Zhang et al. (2009) evaluated 8 different HKGs in dif-ferent tissues of equine such as colon, heart, kidney, liver,lung, lymph node, small intestine and spleen and reportedthat 18S was the most appropriate HKG to normalizegene expressions compared among these different tis-sues. Using the same 18S primers, it was determined that18S was the least stable HKG in the equine endometrium(Table 4). All these tissues are expected to have differentrates of expression as their metabolic and physiologi-cal stages should be different. Therefore, 18S appearedto tolerate these differences among different tissues andsuggested as reference control gene, but it was not thebest HKG to be used in normalization of gene expres-sion in equine endometrium. Furthermore, Bogaert et al.(2006) compared HKGs between normal skin and sarcoidsin horses and found that a set of 3 HKGs were stablyexpressed in normal skin while a different set of 3 HKGswere stable in sarcoids. In that study, an arithmetic meanof these genes were used as control reference gene inrespective tissues. In the present study, equine endome-trial tissues were collected in different days of estrous cycleand early pregnancy. Although the tissue is the same, thephysiological stages of cycle are under the control of cir-culating hormones from the ovarian follicle and corpusluteum. Furthermore, developing embryo inside the uterussecretes many different molecules such as growth fac-tors and cytokines. This milieu inside the uterus is knownto affect endometrial gene expression in favor of preg-nancy or cyclical changes. To make reliable comparisonof these differential expression profile, selection of properHKG is a prerequisite as it must be free of the effects men-tioned above. Results from this study indicated that GAPDHprimers used yielded the most stable expression profileamong 8 different HKGs. RPL32 could also be used in anexperimental setting similar to the one reported in thisstudy.

5. Conclusion

The proposed model, geNorm, BestKeeper, and PCA butnot NormFinder were in agreement detecting the most sta-ble HKGs. On the other hand, all the methods studied werewell aligned in finding the least stable genes. In agreementwith Zhang et al. (2009), it appears that although it is basedon simple descriptive statistical calculations in addition toANOVA, the proposed model could identify the most appro-priate HKG to be used in normalization of gene expressions.Beside, when the first two PCs explained large amount ofvariation as in this study, PCA is another alternative todetect the best and least stable HKGs.

Acknowledgements

This study was financially supported by TUBITAK grant(TOVAG 107O035 to AG). The authors would like to thankProf. Dr. Sefa Celik for his help in real-time analysis.

References

Andersen, C.L., Jensen, J.L., Orntoft, T.F., 2004. Normalization of real-time quantitative reverse transcription-PCR data: a model-basedvariance estimation approach to identify genes suited for normaliza-tion, applied to bladder and colon cancer data sets. Cancer Res. 64,5245–5250.

Atli, M.O., Kurar, E., Kayis, S.A., Aslan, S., Semacan, A., Celik, S., Guzel-oglu, A., 2010. Evaluation of genes involved in prostaglandin action inequine endometrium during estrous cycle and early pregnancy. Anim.Reprod. Sci. 122, 124–132.

Boerboom, D., Brown, K.A., Vaillancount, D., Poitras, P., Goff, A.K., Watan-abe, K., Dore, M., Sirois, J., 2004. Expression of key prostaglandinsynthases in equine endometriumduring late diestrus and early preg-nancy. Biol. Reprod. 70, 391–399.

Bogaert, L., Van Poucke, M., De Baere, C., Peelman, L., Gasthuys, F., Martens,A., 2006. Selection of a set of reliable reference genes for quantita-tive real-time PCR in normal equine skin and in equine sarcoids. BMCBiotechnol. 6, 24.

Bustin, S.A., 2000. Absolute quantification of mRNA using real-timereverse transcription polymerase chain reaction assays. J. Mol.Endocrinol. 25, 169–193.

Bustin, S.A., 2002. Quantification of mRNA using real-time reverse tran-scription PCR (RT-PCR): trends and problems. J. Mol. Endocrinol. 29,23–39.

Craythorn, R.G., Girling, J.E., Hedger, M.P., Rogers, P.A.W., Winnall, W.R.,2009. An RNA spiking method demonstrates that 18S rRNA is reg-ulated by progesterone in the mouse uterus. Mol. Hum. Reprod. 15,757–761.

Dantas, S.F.I.M., de Rezende, T.C.V., Bailao, A.M., Taborda, C.P., Santos,R.D., de Castro, K.P., Soares, C.M.D., 2009. Identification and char-acterization of antigenic proteins potentially expressed during theinfectious process of Paracoccidioides brasiliensis. Microbes Infect. 11,895–903.

de Kok, J.B., Roelofs, R.W., Giesendorf, B.A., Pennings, J.L., Waas, E.T., Feuth,T., Swinkels, D.W., Span, P.N., 2005. Normalization of gene expres-sion measurements in tumor tissues: comparison of 13 endogenouscontrol genes. Lab. Invest. 85, 154–159.

Dheda, K., Huggett, J.F., Bustin, S.A., Johnson, M.A., Rook, G., Zumla, A., 2004.Validation of housekeeping genes for normalizing RNA expression inreal-time PCR. BioTechniques 37, 112–119.

Draghici, S., Khatri, P., Eklund, A.C., Szallasi, Z., 2006. Reliability and repro-ducibility issues in DNA microarray measurements. Trends Genet. 22,101–109.

Karge, W.H.I.I.I., Schaefer, E.J., Ordovas, J.M., 1998. Quantification of mRNAby polymerase chain reaction (PCR) using an internal standard and anonradioactive detection method. Methods Mol. Biol. 110, 43–61.

Kenney, R.M., Doig, P.A., 1986. Equine endometrial biopsy. In: Marrow,D.A. (Ed.), Current Therapy in Theriogenology. W.B. Saunders Com-pany, Philadelphia, USA.

Payne, R.W., Murray, D.A., Harding, S.A., Baird, D.B., Soutar, D.M., 2003.GenStat for Windows. Introduction, 7th edition VSN International,Hemel Hempstead.

Pfaffl, M.W., Tichopad, A., Prgomet, C., Neuvians, T.P., 2004. Determinationof stable housekeeping genes, differentially regulated target genes,and sample integrity: BestKeeper–Excel-based tool using pair-wisecorrelations. Biotechnol. Lett. 26, 509–515.

Pusterla, N., Madigan, J.E., Leutenegger, C.M., 2006. Real-time polymerasechain reaction: a novel molecular diagnostic tool for equine infectiousdiseases. J. Vet. Int. Med. 20, 3–12.

Radonic, A., Thulke, S., Mackay, I.M., Landt, O., Siegert, W., Nitsche, A., 2004.Guideline to reference gene selection for quantitative real-time PCR.Biochem. Biophys. Res. Commun. 313, 856–862.

Rubie, C., Kempf, K., Hans, J., Su, T.F., Tilton, B., Georg, T., Brittner, B., Lud-wig, B., Schilling, M., 2005. Housekeeping gene variability in normaland cancerous colorectal, pancreatic, esophageal, gastric and hepatictissues. Mol. Cell Probes 19, 101–109.

Schefe, J.H., Lehmann, K.E., Buschmann, I.R., Unger, T., Funke-Kaiser, H.,2006. Quantitative real-time RT-PCR data analysis: current conceptsand the novel “gene expression’s CT difference” formula. J. Mol. Med.84, 901–910.

Schroder, A.L., Pelch, K.E., Nagel, S.C., 2009. Estrogen modulates expressionof putative housekeeping genes in the mouse uterus. Endocrine 35,211–219.

Solassol, J., Burcia, V., Costes, V., Lacombe, J., Mange, A., Barbotte, E., deVerbizier, D., Cartier, C., Makeieff, M., Crampette, L., Boulle, N., Maude-londe, T., Guerrier, B., Garrel, R., 2010. Pemphigus vulgaris antigenmRNA quantification for the staging of sentinel lymph nodes in headand neck cancer. Br. J. Cancer 102, 181–187.

Author's personal copy

132 S.A. Kayis et al. / Animal Reproduction Science 125 (2011) 124– 132

Soria-Guerra, R.E., Rosales-Mendoza, S., Chang, S., Haudenshield, J.S.,Padmanaban, A., Rodriguez-Zas, S., Hartman, G.L., Ghabrial, S.A., Kor-ban, S.S., 2010. Transcriptome analysis of resistant and susceptiblegenotypes of Glycine tomentella during Phakopsora pachyrhizi infec-tion reveals novel rust resistance genes. Theor. Appl. Genet. 120,1315–1333.

Sturzenbaum, S.R., Kille, P., 2001. Control genes in quantitative molecularbiological techniques: the variability of invariance. Comp. Biochem.Physiol. Part B: Biochem. Mol. Biol. 130, 281–289.

Thellin, O., Zorzi, W., Lakaye, B., De Borman, B., Coumans, B., Hen-nen, G., Grisar, T., Igout, A., Heinen, E., 1999. Housekeepinggenes as internal standards: use and limits. J. Biotechnol. 75,291–295.

Vandesompele, J., De Preter, K., Pattyn, F., Poppe, B., Van Roy, N., De Paepe,A., Speleman, F., 2002. Accurate normalization of real-time quantita-tive RT-PCR data by geometric averaging of multiple internal controlgenes. Genome. Biol., 3 RESEACH0034.

Walker, C.G., Meier, S., Mitchell, M.D., Roche, J.R., Littlejohn, M., 2009. Eval-uation of real-time PCR endogenous control genes for analysis of geneexpression in bovine endometrium. BMC Mol. Biol. 10, 100.

Warrington, J.A., Nair, A., Mahadevappa, M., Tsyganskaya, M., 2000. Com-parison of human adult and fetal expression and identification of 535housekeeping/maintenance genes. Physiol. Genomics 2, 143–147.

Zhang, Y.W., Davis, E.G., Bai, J., 2009. Determination of internal controlfor gene expression studies in equine tissues and cell culture usingquantitative RT-PCR. Vet. Immunol. Immunopathol. 130, 114–119.