Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

‘Desperate house genes’: the dramatic example of hypoxia

This article has been updated



Microenvironmental conditions in normal or tumour tissues and cell lines may interfere on further biological analysis. To evaluate transcript variations carefully, it is common to use stable housekeeping genes (HKG) to normalise quantitative microarrays or real-time polymerase chain reaction results. However, recent studies argue that HKG fluctuate according to tissues and treatments. So, as an example of HKG variation under an array of conditions that are common in the cancer field, we evaluate whether hypoxia could have an impact on HKG expression.


Expression of 10 commonly used HKG was measured on four cell lines treated with four oxygen concentrations (from 1 to 20%).


Large variations of HKG transcripts were observed in hypoxic conditions and differ along with the cell line and the oxygen concentration. To elect the most stable HKG, we compared the three statistical means based either on PCR cycle threshold coefficient of variation calculation or two specifically dedicated software. Nevertheless, the best HKG dramatically differs according to the statistical method used. Moreover, using, as a reference, absolute quantification of a target gene (here the proteinase activating receptor gene 1 (PAR1) gene), we show that the conclusions raised about PAR1 variation in hypoxia can totally diverge according to the selected HKG used for normalisation.


The choice of a valid HKG will determine the relevance of the results that will be further interpreted, and so it should be seriously considered. The results of our study confirm unambiguously that HKG variations must be precisely and systematically determined before any experiment for each situation, to obtain reliable normalised results in the experimental setting that has been designed. Indeed, such assay design, functional for all in vitro systems, should be carefully evaluated before any extension to other experimental models including in vivo ones.


Gene expression analysis is becoming increasingly important as understanding gene expression patterns is expected to reveal complex regulatory networks involved in disease initiation or progression. Nowadays, two main techniques are used to evaluate gene expression: (1) microarray analysis, which allows qualitative parallel comparative analysis of thousands of target genes in two specific sets of RNA, (2) reverse-transcription (RT-PCR), which allows qualitative or semi-quantitative amplification analysis of limited sets of expressed genes in many different RNA samples. However, despite the greater interest raised by these techniques in the medical community, results may often be interpreted with caution as many conflicting reports regarding the same gene or set of gene expression have been published (Dheda et al, 2005; Strauss, 2006). One of their most important drawbacks is the lack of standardisation as no quality controls like the ones daily used routinely in clinical chemistry laboratories do exist (Eisenstein, 2006). Indeed, amount of starting material, RNA extraction threshold, enzyme efficiencies at the time of RT or PCR and differences in transcriptional activities of tissues or cell lines largely account for the final result but may undergo in-house variations that must be measured on a standardised basis (Nolan et al, 2006). Thus, standardisation protocols are mandatory (Brazma et al, 2001; Bustin et al, 2009).

One way to standardise is to report every gene expression to the extracted total RNA mass. However, as total RNA mass principally consisted in ribosomal RNA (rRNA) that rarely reflects messenger RNA (mRNA) amount, neither 18S nor 28S rRNA molecules are still used to standardise mRNA expression (Spanakis and Brouty-Boye, 1994). To date, internal control genes – also named housekeeping genes (HKG) whose expression is assumed to be and stay constant between cells under different experimental conditions – are frequently used to monitor mRNA amount. Most of HKG are chosen among cellular maintenance genes that are ubiquitously expressed and whose expression is generally considered as constant. Nevertheless, some authors have reported that HKG expression could rise and/or fall (Thellin et al, 1999). For example, two widely used HKG, namely glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and β-actin (ACTB), show great in vitro variations (Schmittgen and Zakrajsek, 2000; Lupberger et al, 2002; Aerts et al, 2004; Dheda et al, 2004; Synnergren et al, 2007). This holds also true in vivo as shown in human tissues such as kidney (Schmid et al, 2003; Biederman et al, 2004), melanoma (Giricz et al, 2008) or hepatocellular carcinoma (Cicinnati et al, 2008). In addition, reporting the uneven expression of 13 HKG in a panel of normal and tumour human tissues, de Kok et al (2005) have shown that the less variable HKG differed according to the tissue and suggested that it will be important, before experiments, to establish HKG variations clearly to select the best one for quantitative real-time PCR (qRT-PCR) standardisation.

Many publications dealing with cancer have reported gene expression studies in hypoxic conditions (Helczynska et al, 2008; Seifeddine et al, 2008; Sheikh et al, 2008; Chaudary and Hill, 2009; Zhang et al, 2009), but until now related HKG variations have not yet been much characterised (Zhong and Simons, 1999). However, analysing if HKG expression is stable under different oxygen concentrations remains an important issue especially in cancer field as it is well known that proliferating tumours often grow in hypoxic conditions (Semenza, 2003). Thus, the aim of this study was to examine HKG variations in cell lines of different origins, cultured under different oxygen concentrations. The expression pattern of 10 commonly used HKG (namely ATP synthase (ATP5G3), β-2-microglobulin (B2M), ACTB, β-glucuronidase (GUSB), cyclophilin A (PPIA), GAPDH, hypoxanthine ribosyltransferase (HPRT1), phosphoglycerokinase (PGK1), TATA-box-binding protein (TBP) and transferrin receptor (TFRC)) was investigated by qRT-PCR in different cell lines cultured in various hypoxic or aerobic conditions.

Materials and Methods

Cell culture

Human prostate LNCaP and PNT2 cell lines, as well as breast cancer MCF-7 and kidney HEK293 cell lines, were used for experiments and were originally obtained from the ATCC (Rockville, MD, USA). All cell lines were cultured in RPMI 1640 except HEK, which was cultured in DMEM. Cells were supplemented with 10% FBS (Sigma, St Louis, MO, USA), under different oxygen concentrations (1, 5, 10 and 20%) during 72 h (Sanyo, Osaka, Japan). All oxygen conditions for each cell line were performed independently in triplicate.

Total RNA isolation and reverse transcription

Total RNA from cultured cell lines was extracted using the cesium-chloride method. Briefly, cultured cells were homogenised in a 4 M guanidium thiocyanate/25 mM sodium citrate/0.5% sarcosyl solution. The homogenate was then ultracentrifuged (17 h, 35 000 r.p.m.) in a 5.7 mM cesium chloride/5 mM EDTA gradient. After centrifugation (15 000 r.p.m., 5 min) in a chloroform/isoamyl alcohol (24 : 1) solution, supernatant was collected then precipitated by centrifugation (15 000 r.p.m., 15 min) in 3 M sodium acetate (pH 5.2)/100% ethanol solution. Total RNA purity was controlled by Agilent Bioanalyzer (Massy, France). RT was systematically performed using 500 ng total RNA and universal primers at 72°C for 10 min, and then 1 h at 42°C with MMLv as reverse transcriptase (Invitrogen, Eragny, France). cDNA quality and concentration were controlled by RiboGreen fluorometric measurements (Invitrogen).

Primer design for housekeeping genes

A total of 10 HKG (Table 1), used in many studies, were selected for gene expression analysis. All genes are constitutively expressed, most of them lacking canonical TATA box. All show independent functions in cellular maintenance, and the regulation of their expression is assumed not to be directly related. Among those studied, only GAPDH and PGK1 share an identical cellular biochemical process, namely glycolysis. Specific primers (Eurofins, Edersberg, Germany) were designed by Primer3 software and validated in silico by Blast and Blat analysis. Each designed couple of primers was then tested in vitro onto human cDNAs by conventional PCR. Specificity of amplified products was checked by direct sequencing on 3130xl Genetic Analyzer (Applied Biosystems, Les Ulis, France). Designed and tested as above, forward 5′-IndexTermTCAATGAAACCCTGCTCGAA-3′ and reverse 5′-IndexTermGTTATTCAGGTTACTAGAGC-3′ primers were used to amplify the proteinase activating receptor gene 1 (PAR1).

Table 1 Characteristics of the different selected housekeeping genes

Quantitative real-time PCR

To measure HKG expression level, we performed qRT-PCR using the 7900 ABI system (Applied Biosystems). For experiment, 96-well PCR plates, divided into 12 columns of eight identical wells, were used. A total of 10 columns contained primers and probes for the detection of 10 HKG in duplicates. Six wells were systematically used for inter and intra-assays controls (500 and 50 ng of the same cDNA with identical set of primers). HKG amplification was detected by SYBR Green fluorescence using qRT-PCR mix containing 10 μ M of each primer, qPCR buffer (Applied Biosystems), and 500 ng cDNA in a 20 μl total reaction volume. Amplification was performed as follow: 95°C (10 min) for primer elongation, 40 cycles of amplification at 95°C (30 s) for denaturation and 60°C (60 s) for annealing and extension. The number of PCR cycles to reach the fluorescence threshold in each sample was defined as the cycle threshold (Ct). Ct values are proportional to the negative logarithm of the initial amount of input cDNA. Ct values of 10 HKG in each cell line were directly related. In addition, PAR1 mRNA expression was established on LNCaP cell line using a dedicated standard curve based on serial dilutions (10-fold) of known concentrations of PAR1 full-length transcript cloned into pCR2.1 plasmid (109 to one copy per μl) (Invitrogen).

All experiments were performed three times, and although Ct values did not fluctuate, the Ct mean of those three experiments was used for statistical analysis.

Immunohistochemistry experiments

Immunohistochemistry was performed on LNCaP cells cultured as already described for 72 h in aerobic (20%) and hypoxic (1%) conditions. After PBS washes, cells were scrapped, gently cytocentrifuged (Hettich, Kirchlengern, Germany) and slides were fixed with 10 min cold acetone treatment. After 10 min with blocking buffer (Pierce), slides were incubated 1 h with anti-PAR1 primary antibody (ATAP2; Santa Cruz, Heidelberg, Germany) then for 1 h with biotinylated secondary horse anti-mouse IgG (Vector Laboratories, Burlingame, CA, USA). Between incubation steps, slides were washed in PBS/Tween 0.1%. 3-3-Diaminobenzidine (DAB SK-4100; Vector Laboratories) was used as detection system. Slides were counterstained with Harris haematoxylin.

Statistical analysis

Statistical analyses were performed with both GraphPad Prism4 (GraphPad, Palo Alto, CA, USA) and SPSS10 (SPSS, Chicago, IL, USA) statistical software. Because most of the variables did not comply with normal Gaussian distribution as determined by Kolmogorov–Smirnov test, nonparametric Friedman test was applied to compare dependant samples of continuous variables. All P-values were based on two-tailed tests and the threshold to accept statistical significance was set at α level 0.05.

GeNorm and NormFinder, software specifically designed for normalisation, were used according to instructions. These are application tools for Microsoft Excel and are available online (NormFinder:; GeNorm:


Four phenotypically cell lines (PNT2, LNCaP, MCF-7 and HEK) were cultured under four oxygen concentrations (1, 5, 10% and aerobic). Total RNA of these cell lines was extracted and retrotranscribed into cDNA. For comparative purposes, 500 ng of total RNA and cDNA was systematically used for experiments. Ten HKG (Table 1) were then quantified by qRT-PCR.

HKG variability based on PCR cycle threshold is induced by hypoxic conditions

As an example of HKG variation, PPIA Ct fluctuations are represented (Figure 1). Results show that PPIA is not expressed at the same basal level in the different cell lines and varies regarding hypoxic conditions. Furthermore, it appears clearly that PPIA variations strongly differ according to cell lines, suggesting that PPIA is regulated in a cell-line-specific way. These variations of one of the most commonly used HKGs prompted us to verify whether other HKGs show same kind of variations.

Figure 1

Cyclophilin A (PPIA) Ct variations in PNT2, LNCaP, MCF-7 and HEK cell lines according to hypoxic culture conditions (1, 5, 10 and 20% oxygen). cDNA (500 ng) –provided by reverse transcription of 500 ng of total RNA – was systematically used for qRT-PCR.

A global graph (Figure 2) shows all tested HKGs in each cell line. As for PPIA, other HKGs fluctuate according to hypoxic conditions and cell lines. A Friedman test based on means of HKG expression for each oxygen concentration indicates that HKG variation is statistically significant between different oxygen concentrations in PNT2 (P=0.0006), LNCaP (P<0.0001) and HEK (P<0.0001) cell lines. However, the difference in MCF-7 cell line remains insignificant (P=0.356), as shown in Figure 2 where box plots are largely thinner than those in other cell lines.

Figure 2

Housekeeping genes variations in PNT2, MCF-7, LNCaP and HEK cell lines grown in different hypoxic conditions (1, 5, 10 and 20% oxygen). Box plots represent Ct variations measured for each gene.

Statistical analysis of hypoxia-based HKG variation

First, Ct coefficients of variations (CtCV%) were calculated for each HKG in each cell line (Table 2). As suggested by Figure 2, each HKG strongly varies regarding cell lines, as it is pinpointed for the ACTB gene whose CtCV% ranges from 1.41% in MCF-7 cell line to 9.77% in HEK cell line. Moreover, the best HKG differs from a cell line to another.

Table 2 Coefficients of variation (CtCV%) of housekeeping genes

Second, the expression stability of the different putative HKG was tested using mathematical models. Two are yet available thus data obtained for each sample and each HKG were analysed using both GeNorm and NormFinder (Vandesompele et al, 2002; Andersen et al, 2004). GeNorm provides a ranking of the tested genes based on the reference gene stability measure M, which is defined as the average pair-wise variation of a particular gene compared with all other control genes. Genes with higher M values have greater variations of expression (Vandesompele et al, 2002). NormFinder, whose strategy is rooted in a mathematical model of gene expression, provides a ranking of the tested genes based on a direct measure of both overall expression variation and the variation between sample subgroups of candidate reference genes. The stability of genes was shown as stability value, and genes with lower stability values have higher expression stability (Andersen et al, 2004).

Results of the different calculations are summarised in Table 3, where all HKGs are classified for each cell line and for each statistical model according to their rank. It appears clearly that HKGs varying the least differ according to the cell line considered, whatever the statistical approach used. Furthermore, as it is exposed with highlighted ATP5G3 in Table 3, GeNorm, NormFinder and CtCV% calculation do not give identical ranking of results. For example, in PNT2 cell line, when statistical analysis is performed either on CtCV% or NormFinder, ATP5G3 seems to be the best HKG for normalisation but when GeNorm is used, ATP5G3 is likely to be one of the worst. In LNCaP cells, GAPDH appears the best HKG for GeNorm and NormFinder, the second best using CtCV% but it is truly the worst for the three methods in MCF-7 cells. In addition, when HEK cells are analysed, whereas GAPDH seems a relevant HKG on the basis of CtCV%, GeNorm and NormFinder considered GAPDH as the worst HKG.

Table 3 HKG variations summary in PNT2, LNCaP, MCF-7 and HEK cell lines grown in different oxygen concentrations

Altogether, these results confirm that HKGs used for normalising gene expression in qRT-PCR may vary. Moreover, they emphasise that HKGs may dramatically fluctuate according to hypoxic conditions.

Gene expression varies according to the HKG used

To evaluate the impact of HKG choice may have on the study of a target gene of interest, we performed qRT-PCR on RNAs extracted from LNCaP cells cultured in 1 and 20% oxygen. From the two derived cDNAs, the target gene, encoding PAR1 receptor known to be expressed and regulated in prostate carcinoma and cell lines (Kaushal et al, 2006; Yuan and Lin, 2004), was quantified, as well as two dedicated HKGs, namely TBP and PPIA. Both were chosen for their discrepant expression under variable hypoxic conditions: TBP showed one of the greatest stability in LNCaP cell line whereas PPIA was one of the less stable HKG in this model system. To quantify PAR1 transcript expression, we performed an absolute quantification. For that purpose, a standard curve was established using serial dilutions of a recombinant plasmid containing verified full-length PAR1 clone. This plasmid was used to determine the exactitude of the technique (intra (n=30) and inter (n=15) CtCV% <3.5 and 4.2% respectively), and to calculate the number of PAR1 equivalent copies present in test samples. When HKG normalisation is used, the observed results (Figure 3) unambiguously show that PAR1 is upregulated in normoxia when TBP is the reference HKG (Figure 3B), whereas it appears strongly downexpressed when PPIA is used for calculations (Figure 3C). In addition, both PAR1 quantification by standard curve (Figure 3A) and PAR1 normalisation by TBP show the same profile, suggesting that using HKG is not necessarily the best method to study the variation of a specific target gene. To validate PAR1 downregulation in hypoxia, we performed immunohistochemistry experiments that confirm at the protein level the decrease of PAR1 onto the plasma membrane of hypoxic LNCaP cells (data not shown).

Figure 3

Gene expression varies according to the HKG used. In LNCaP cell line treated in two oxygen concentrations (1 and 20%), proteinase activating receptor gene 1 (PAR1) was (A) quantified alone with a standard curve, (B) normalised with the most stable HKG TBP or (C) normalised with one of the weakest stable HKG cyclophilin A (PPIA).


The development of qRT-PCR over the past decade allowed a rapid and easy way to study gene expression. All qRT-PCR assays are characterised by biological variation, mainly the inherent variability of mRNA levels in different tissues and individuals, but also analytical variations characterised by differences in RNA quality, variability in extraction protocols, RT and PCR efficiencies that can lead to erroneous and confusing results (Huggett et al, 2005). One of the main drawbacks of qRT-PCR technique remains the difficulty to normalise results conveniently. Several normalisation methods have been proposed but none has gained universal approval (Huggett et al, 2005). Normalisation can be performed by different means: (1) conventional normalisation may use accurate measurement then normalisation against total RNA concentration, and can produce biologically relevant quantitative results (Tricarico et al, 2002). However, this method does not provide internal control for either RT or PCR. In addition, before any RT–PCR measurement, it requires a very accurate quantification of the RNA sample that is difficult to achieve (Bustin et al, 2005). A derived method can use rRNA normalisation (Bhatia et al, 1994; Zhong and Simons, 1999), which may result in less variability (Bas et al, 2004), nevertheless some fluctuations in rRNA normalisation have also been reported (Spanakis and Brouty-Boye, 1994); (2) the widely used method consists in using one or several HKGs, whose expression is assumed to be constant whatever the experimental or physiological conditions. Hence, using such normalised gene expression allows correcting data for cellular variability, RT efficiency and RNA quality. It is now accepted that normalisation against a single reference mRNA is likely to be inappropriate (Tricarico et al, 2002; Andersen et al, 2004), except in very well-defined circumstances (Dheda et al, 2004), but normalisation against a panel of reference genes, containing at least the three less variable, should be preferred (Vandesompele et al, 2002). Unfortunately, even using a panel of HKG, many studies have pinpointed that HKG can fluctuate in cell lines as well as in tissue (Thellin et al, 1999; Schmittgen and Zakrajsek, 2000; Lupberger et al, 2002; Schmid et al, 2003; Aerts et al, 2004; Biederman et al, 2004; Dheda et al, 2004; de Kok et al, 2005; Synnergren et al, 2007; Cicinnati et al, 2008). Among the more recent ones, Derks et al (2008) showed that HKG expression was not constant enough to be used as internal control for gene normalisation in rat brains. The same conclusions were raised for melanoma and fibroblasts samples (Giricz et al, 2008).

The choice of a relevant HKG has to rely on observed unambiguous data regarding the experimental system where it is supposed to be used. Unfortunately, for most of the HKG used in published results, no exhaustive studies of their variations in dedicated conditions have been performed yet. For example, a lot of reports deal with transcripts variation under hypoxia (Helczynska et al, 2008; Seifeddine et al, 2008; Sheikh et al, 2008; Chaudary and Hill, 2009; Zhang et al, 2009). Authors formulate relevant conclusions regarding gene expression on the basis of the use in most cases of a unique or more rarely a specific set of HKG. However, the variations of that unique gene or set of genes have not been much studied, especially in the case of hypoxic conditions. If, as it was shown for some instances of HKG in very controlled conditions, HKG vary also under hypoxia, then a large number of results analysing transcript regulation under hypoxia have to be interpreted with the utmost care. Thus, as potential users of HKG to evaluate gene expression under hypoxia, we aimed to evaluate to which extent widely used HKG in qRT-PCR may vary in cell lines cultured in different oxygen concentrations. Our study results unambiguously show that HKG expression dramatically fluctuates depending on air oxygen content. Whatever the HKG measured, slight to important variations in expression can be observed in an oxygen-concentration-dependent manner. Moreover, although these variations differ from one HKG to another, we observed surprisingly that variation of a dedicated HKG may differ from one cell line to another. Indeed, HKG harbouring a good CtCV% in a given cell line may show a higher coefficient of variation in another cell line (Table 2). This holds particularly true for GAPDH, which oscillates with the second larger amplitude after B2M according to hypoxia conditions and cell lines studied. This confirms previous results that have already been reported GAPDH variations in various prostate and endothelial cell lines (Zhong and Simons, 1999). Nevertheless, among the cell lines we tested, some seem to be more or less responsive to hypoxic treatments than others, as it is suggested for MCF-7 cell line where HKG variations appears to be not significant. Our study results corroborate the conclusions raised by Haller et al (2004) on HKG variations in several carcinomas who stated that a stable expression of a reference gene in one tumour type does not predict a stable expression in another.

Indeed, top choice of HKG is challenging because of the impact it could have on the results reliability and, as a consequence, on the results interpretations. Indeed, using HKG whose expression will vary in opposite directions under the same stimulus will undoubtedly led to errors in interpretation as it was shown above on PAR1 transcripts that are upregulated or downregulated by hypoxia according to the HKG used for the experiment (Figure 3). Here, we found by both absolute quantification and best HKG normalisation (using TBP in LNCaP hypoxic system) that PAR1 transcripts are likely to be downregulated in that prostate cancer cell line. This was further validated at the protein level confirming PAR1 global downregulation in LNCaP cells treated in hypoxic conditions. However, a recent report showed that, in MDAMDB321 aggressive breast cancer cells, hypoxia (2% oxygen) enhance PAR1 expression on the basis of conventional qRT-PCR experiments using ACTB as HKG, gel analysis for transcript quantification (Naldini et al, 2009), but no reported data regarding the choice of ACTB. Nevertheless, in our hands, MCF-7 breast cancer cells appear to have the most stable HKGs in hypoxia. Among them, ACTB displays a middle range CtCV% (1.41%) and is relatively stable. However MCF-7 could not be superimposed to MDAMDB321 cells as great expression variability can be observed between cell lines of the same tissue or embryonic origin as shown by prostate PNT2 and LNCaP cells. Thus, no clear conclusions concerning the rationale of the use of ACTB can be extrapolated to MDAMB321 cells system. Hence, PAR1 regulation appears strongly modulated by hypoxia during natural course of cancer, emphasising the need of a very careful quantification of its variation at either transcript or protein levels. For a unique specific target or a small set of specific targets, confirmation of transcript variation by further evaluation of protein variation by either immunohistochemistry or western blot will undoubtedly validate the initial choice of the best HKG.

Nevertheless, for large-scale studies of gene expression (transcripts variation analysis by microarrays, microfluidic qRT-PCR and so on) for which validation at the protein level is quite impossible, a clear validation step of suitable reference genes remains a crucial problem. Different statistical approaches (GeNorm, NormFinder) have been developed to identify and validate reference genes appropriate for normalisation in qRT-PCR assays (Andersen et al, 2004; Haller et al, 2004; Vandesompele et al, 2002). However, the use of these devoted software does not elucidate the problem of the best HKG choice, as shown in Table 3. Indeed, both statistical approaches do not validate the same genes according to a given cell line. Moreover, as it was observed in CtCV% calculation, these approaches pinpoint HKG undulations according to cell lines and hypoxia conditions.

Taken together, the results of our study suggest that neither a single gold standard HKG nor a set of HKG can be widely used for qRT-PCR normalisation in hypoxia experiments. Therefore, it seems essential to test HKG expression before any experiment based on hypoxic culture. Besides for each cell line, the best HKG should also be specifically determined and validated before conducting hypoxic studies. An additional and fundamental question that remains to be determined before any analysis concerns the choice of the most convenient statistical approach that should be used to identify the best HKG or set of HKG. Whereas GeNorm and NormFinder have been reported to be interesting statistical means to choose the best HKG in specific normal or tumour tissues (Cicinnati et al, 2008; Lyng et al, 2008), a careful evaluation of their use in different in vitro experiments with a multiplicity of experimental conditions is necessary.

However, in the case of a unique target gene of interest, in very well-defined in vitro conditions where it is possible to have equivalent amount of starting material, the use of a standardised curve for absolute quantification is likely to be the best way to study its variations. Otherwise, in all other cases where multiple set of target genes in various conditions have to be analysed, each set of HKG must be tested before any normalisation in the cell lines and under the conditions the assay will be performed. Till universal validation of relevant statistical software, results have to be carefully analysed regarding the CtCV% of each HKG. Once the more stable HKG or set of HKG in that condition has been evidenced as the best for normalisation, then it will be really possible to begin qRT-PCR experiments.

The choice of a valid HKG set will undoubtedly determine the relevance of the results that will be further interpreted, and so it should be seriously considered. Indeed, such kind of assay design, functional for all in vitro systems, should be implemented before any extension to other experimental models including in vivo ones. Nevertheless, one should keep in mind that defining a universal and unanimous method to choose the best HKG in a system model is still strongly mandatory.

Accession codes



Change history

  • 16 November 2011

    This paper was modified 12 months after initial publication to switch to Creative Commons licence terms, as noted at publication


  1. Aerts JL, Gonzales MI, Topalian SL (2004) Selection of appropriate control genes to assess expression of tumor antigens using real-time RT-PCR. Biotechniques 36: 84–86, 88, 90-1

    CAS  Article  PubMed  Google Scholar 

  2. Andersen CL, Jensen JL, Orntoft TF (2004) 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: 5245–5250

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  3. Bas A, Forsberg G, Hammarstrom S, Hammarstrom ML (2004) Utility of the housekeeping genes 18S rRNA, beta-actin and glyceraldehyde-3-phosphate-dehydrogenase for normalization in real-time quantitative reverse transcriptase-polymerase chain reaction analysis of gene expression in human T lymphocytes. Scand J Immunol 59: 566–573

    CAS  Article  Google Scholar 

  4. Bhatia P, Taylor WR, Greenberg AH, Wright JA (1994) Comparison of glyceraldehyde-3-phosphate dehydrogenase and 28S-ribosomal RNA gene expression as RNA loading controls for northern blot analysis of cell lines of varying malignant potential. Anal Biochem 216: 223–226

    CAS  Article  PubMed  Google Scholar 

  5. Biederman J, Yee J, Cortes P (2004) Validation of internal control genes for gene expression analysis in diabetic glomerulosclerosis. Kidney Int 66: 2308–2314

    CAS  Article  PubMed  Google Scholar 

  6. Brazma A, Hingamp P, Quackenbush J, Sherlock G, Spellman P, Stoeckert C, Aach J, Ansorge W, Ball CA, Causton HC, Gaasterland T, Glenisson P, Holstege FC, Kim IF, Markowitz V, Matese JC, Parkinson H, Robinson A, Sarkans U, Schulze-Kremer S, Stewart J, Taylor R, Vilo J, Vingron M (2001) Minimum information about a microarray experiment (MIAME) – toward standards for microarray data. Nat Genet 29: 365–371

    CAS  Article  PubMed  Google Scholar 

  7. Bustin SA, Benes V, Garson JA, Hellemans J, Huggett J, Kubista M, Mueller R, Nolan T, Pfaffl MW, Shipley GL, Vandesompele J, Wittwer CT (2009) The MIQE guidelines: minimum information for publication of quantitative real-time PCR experiments. Clin Chem 55: 611–622

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  8. Bustin SA, Benes V, Nolan T, Pfaffl MW (2005) Quantitative real-time RT–PCR – a perspective. J Mol Endocrinol 34: 597–601

    CAS  Article  PubMed  Google Scholar 

  9. Chaudary N, Hill RP (2009) Increased expression of metastasis-related genes in hypoxic cells sorted from cervical and lymph nodal xenograft tumors. Lab Invest 89: 587–596

    CAS  Article  PubMed  Google Scholar 

  10. Cicinnati VR, Shen Q, Sotiropoulos GC, Radtke A, Gerken G, Beckebaum S (2008) Validation of putative reference genes for gene expression studies in human hepatocellular carcinoma using real-time quantitative RT-PCR. BMC Cancer 8: 350

    Article  PubMed  PubMed Central  Google Scholar 

  11. de Kok JB, Roelofs RW, Giesendorf BA, Pennings JL, Waas ET, Feuth T, Swinkels DW, Span PN (2005) Normalization of gene expression measurements in tumor tissues: comparison of 13 endogenous control genes. Lab Invest 85: 154–159

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  12. Derks NM, Muller M, Gaszner B, Tilburg-Ouwens DT, Roubos EW, Kozicz LT (2008) Housekeeping genes revisited: different expressions depending on gender, brain area and stressor. Neuroscience 156: 305–309

    CAS  Article  PubMed  Google Scholar 

  13. Dheda K, Huggett JF, Bustin SA, Johnson MA, Rook G, Zumla A (2004) Validation of housekeeping genes for normalizing RNA expression in real-time PCR. Biotechniques 37: 112–114, 116, 118-9

    CAS  Article  PubMed  Google Scholar 

  14. Dheda K, Huggett JF, Chang JS, Kim LU, Bustin SA, Johnson MA, Rook GA, Zumla A (2005) The implications of using an inappropriate reference gene for real-time reverse transcription PCR data normalization. Anal Biochem 344: 141–143

    CAS  Article  PubMed  Google Scholar 

  15. Eisenstein M (2006) Microarrays: quality control. Nature 442: 1067–1070

    CAS  Article  PubMed  Google Scholar 

  16. Giricz O, Lauer-Fields JL, Fields GB (2008) The normalization of gene expression data in melanoma: investigating the use of glyceraldehyde 3-phosphate dehydrogenase and 18S ribosomal RNA as internal reference genes for quantitative real-time PCR. Anal Biochem 380: 137–139

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  17. Haller F, Kulle B, Schwager S, Gunawan B, von Heydebreck A, Sultmann H, Fuzesi L (2004) Equivalence test in quantitative reverse transcription polymerase chain reaction: confirmation of reference genes suitable for normalization. Anal Biochem 335: 1–9

    CAS  Article  PubMed  Google Scholar 

  18. Helczynska K, Larsson AM, Holmquist Mengelbier L, Bridges E, Fredlund E, Borgquist S, Landberg G, Pahlman S, Jirstrom K (2008) Hypoxia-inducible factor-2alpha correlates to distant recurrence and poor outcome in invasive breast cancer. Cancer Res 68: 9212–9220

    CAS  Article  PubMed  Google Scholar 

  19. Huggett J, Dheda K, Bustin S, Zumla A (2005) Real-time RT-PCR normalisation; strategies and considerations. Genes Immun 6: 279–284

    CAS  Article  PubMed  Google Scholar 

  20. Kaushal V, Kohli M, Dennis RA, Siegel ER, Chiles WW, Mukunyadzi P (2006) Thrombin receptor expression is upregulated in prostate cancer. Prostate 66: 273–282

    CAS  Article  PubMed  Google Scholar 

  21. Lupberger J, Kreuzer KA, Baskaynak G, Peters UR, le Coutre P, Schmidt CA (2002) Quantitative analysis of beta-actin, beta-2-microglobulin and porphobilinogen deaminase mRNA and their comparison as control transcripts for RT–PCR. Mol Cell Probes 16: 25–30

    CAS  Article  PubMed  Google Scholar 

  22. Lyng MB, Laenkholm AV, Pallisgaard N, Ditzel HJ (2008) Identification of genes for normalization of real-time RT-PCR data in breast carcinomas. BMC Cancer 8: 20

    Article  PubMed  PubMed Central  Google Scholar 

  23. Naldini A, Filippi I, Ardinghi C, Silini A, Giavazzi R, Carraro F (2009) Identification of a functional role for the protease-activated receptor-1 in hypoxic breast cancer cells. Eur J Cancer 45: 454–460

    CAS  Article  PubMed  Google Scholar 

  24. Nolan T, Hands RE, Bustin SA (2006) Quantification of mRNA using real-time RT-PCR. Nat Protoc 1: 1559–1582

    CAS  Article  PubMed  Google Scholar 

  25. Schmid H, Cohen CD, Henger A, Irrgang S, Schlondorff D, Kretzler M (2003) Validation of endogenous controls for gene expression analysis in microdissected human renal biopsies. Kidney Int 64: 356–360

    CAS  Article  PubMed  Google Scholar 

  26. Schmittgen TD, Zakrajsek BA (2000) Effect of experimental treatment on housekeeping gene expression: validation by real-time, quantitative RT-PCR. J Biochem Biophys Methods 46: 69–81

    CAS  Article  PubMed  Google Scholar 

  27. Seifeddine R, Dreiem A, Blanc E, Fulchignoni-Lataud MC, Le Frere Belda MA, Lecuru F, Mayi TH, Mazure N, Favaudon V, Massaad C, Barouki R, Massaad-Massade L (2008) Hypoxia down-regulates CCAAT/enhancer binding protein-alpha expression in breast cancer cells. Cancer Res 68: 2158–2165

    CAS  Article  PubMed  Google Scholar 

  28. Semenza GL (2003) Targeting HIF-1 for cancer therapy. Nat Rev Cancer 3: 721–732

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  29. Sheikh AY, Chun HJ, Glassford AJ, Kundu RK, Kutschka I, Ardigo D, Hendry SL, Wagner RA, Chen MM, Ali ZA, Yue P, Huynh DT, Connolly AJ, Pelletier MP, Tsao PS, Robbins RC, Quertermous T (2008) In vivo genetic profiling and cellular localization of apelin reveals a hypoxia-sensitive, endothelial-centered pathway activated in ischemic heart failure. Am J Physiol Heart Circ Physiol 294: H88–H98

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  30. Spanakis E, Brouty-Boye D (1994) Evaluation of quantitative variation in gene expression. Nucleic Acids Res 22: 799–806

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  31. Strauss E (2006) Arrays of hope. Cell 127: 657–659

    CAS  Article  PubMed  Google Scholar 

  32. Synnergren J, Giesler TL, Adak S, Tandon R, Noaksson K, Lindahl A, Nilsson P, Nelson D, Olsson B, Englund MC, Abbot S, Sartipy P (2007) Differentiating human embryonic stem cells express a unique housekeeping gene signature. Stem Cells 25: 473–480

    CAS  Article  PubMed  Google Scholar 

  33. Thellin O, Zorzi W, Lakaye B, De Borman B, Coumans B, Hennen G, Grisar T, Igout A, Heinen E (1999) Housekeeping genes as internal standards: use and limits. J Biotechnol 75: 291–295

    CAS  Article  PubMed  Google Scholar 

  34. Tricarico C, Pinzani P, Bianchi S, Paglierani M, Distante V, Pazzagli M, Bustin SA, Orlando C (2002) Quantitative real-time reverse transcription polymerase chain reaction: normalization to rRNA or single housekeeping genes is inappropriate for human tissue biopsies. Anal Biochem 309: 293–300

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  35. Vandesompele J, De Preter K, Pattyn F, Poppe B, Van Roy N, De Paepe A, Speleman F (2002) Accurate normalization of real-time quantitative RT-PCR data by geometric averaging of multiple internal control genes. Genome Biol 3: RESEARCH0034

    Article  PubMed  PubMed Central  Google Scholar 

  36. Yuan TC, Lin MF (2004) Protease-activated receptor 1: a role in prostate cancer metastasis. Clin Prostate Cancer 3: 189–191

    CAS  Article  PubMed  Google Scholar 

  37. Zhang C, Song QH, Li J, Tian Y (2009) Hypoxia-induced expression of RTEF-1 (related transcriptional enhancer factor-1) in endothelial cells is independent of HIF-1 (hypoxia-inducible factor-1). Biochem Biophys Res Commun 381: 333–338

    CAS  Article  PubMed  Google Scholar 

  38. Zhong H, Simons JW (1999) Direct comparison of GAPDH, beta-actin, cyclophilin, and 28S rRNA as internal standards for quantifying RNA levels under hypoxia. Biochem Biophys Res Commun 259: 523–526

    CAS  Article  PubMed  Google Scholar 

Download references


We thank Bioquanta (Aurora, CO, USA) for research grants and constant support.

Author information



Corresponding author

Correspondence to S Loric.

Rights and permissions

From twelve months after its original publication, this work is licensed under the Creative Commons Attribution-NonCommercial-Share Alike 3.0 Unported License. To view a copy of this license, visit

Reprints and Permissions

About this article

Cite this article

Caradec, J., Sirab, N., Keumeugni, C. et al. ‘Desperate house genes’: the dramatic example of hypoxia. Br J Cancer 102, 1037–1043 (2010).

Download citation


  • housekeeping genes
  • normalisation
  • RNA
  • PCR
  • quantification

Further reading


Quick links