RESEARCH REPORT Validation of housekeeping genes for normalizing RNA expression in real-time PCR Keertan Dheda1, Jim F. Huggett1, Stephen A. Bustin2, Margaret A. Johnson1, Graham Rook1, and Alimuddin Zumla1 BioTechniques 37:112-119 (July 2004) Analysis of RNA expression using techniques like real-time PCR has traditionally used reference or housekeeping genes to control for error between samples. This practice is being questioned as it becomes increasingly clear that some housekeeping genes may vary considerably in certain biological samples. We used real-time reverse transcription PCR (RT-PCR) to assess the levels of 13 housekeeping genes expressed in peripheral blood mononuclear cell culture and whole blood from healthy individuals and those with tuberculosis. Housekeeping genes were selected from conventionally used ones and from genes reported to be invariant in hu- man T cell culture. None of the commonly used housekeeping genes [e.g., glyceraldehyde-phosphate-dehydrogenase (GAPDH)] were found to be suitable as internal references, as they were highly variable (>30-fold maximal variability). Furthermore, genes previously found to be invariant in human T cell culture also showed large variation in RNA expression (>34-fold maximal vari- ability). Genes that were invariant in blood were highly variable in peripheral blood mononuclear cell culture. Our data show that RNA specifying human acidic ribosomal protein was the most suitable housekeeping gene for normalizing mRNA levels in human pulmonary tuberculosis. Validations of housekeeping genes are highly specific for a particular experimental model and are a cru- cial component in assessing any new model. INTRODUCTION the chosen housekeeping gene fluctu- build research capacity (http://www. ates randomly between samples, then eubusiness.com/funding/research/ It is essential to control for error be- small differences between genes of rtd01_en.htm). We therefore required tween samples when measuring RNA interest will be missed. Furthermore, a method with simplicity and relative- expression. This error can be intro- if the experimental condition causes a ly low cost. The housekeeping gene duced at a number of stages throughout directional change in the housekeep- method of normalization was chosen the experimental protocol (input sam- ing gene, the subsequent normaliza- for this reason. Our main genes of in- ple, RNA extraction, reverse transcrip- tion will cause an erroneous result. terest specify cytokines that have low tion, etc.). There are many methods to This was shown in a study of human RNA expression (6). Consequently, control for this error. One approach is asthma, where target gene expression we anticipate that differences between to normalize to total RNA. However, between experimental groups became study groups may be small. Therefore this requires a reliable RNA quanti- falsely different when β-actin was it was important to find an internal fication method and fails to take into used as a normalizer (4). It was in fact reference that had minimal variability. account the variability of the reverse the β-actin rather than the target gene Previous reports have used conven- transcription and other steps. A widely that was changing (4). More recently, tional housekeeping genes in models of used alternative is to normalize RNA it has become clear that housekeeping tuberculosis (TB) to normalize for dif- levels to an internal reference or house- genes like β-actin and glyceraldehyde- fering amounts of input RNA (7–9), but keeping gene (1). 3-phosphate-dehydrogenase (GAPDH) there has been no report of studies to The expression levels of reference may be inappropriate as internal refer- check the validity of these housekeep- genes should remain constant between ences because of their variability (5). ing genes in TB or in in vitro lympho- the cells of different tissues and under Appropriate validation of internal ref- cyte cultures. We used real-time reverse different experimental conditions (2). erences is therefore crucial to avoid transcription PCR (RT-PCR) to study If these requirements are not fulfilled, misinterpretations of study findings. the levels of 13 housekeeping genes ex- then normalization to varying inter- Our experimental protocols will pressed in whole blood and peripheral nal references can lead to increased be transferred to sites in the devel- blood mononuclear cell (PBMC) cul- “noise” or erroneous results (3). If oping world as part of a program to ture of healthy volunteers and patients 1Royal Free Medical School, London and 2Queen Mary’s School of Medicine and Dentistry, London, UK 112 BioTechniques Vol. 37, No. 1 (2004) Table 1. Housekeeping Genes Selected for Group 1 and Group 2 Panels as a requirement for suitability Group 1 Housekeeping Genes Group 2 Housekeeping Genes as a reference gene. We did not use ribosomal Human acidic ribosomal protein (HuPO) Human acidic ribosomal protein (HuPO) subunit RNAs as housekeeping β-Actin (BA) Elongation factor-1-α (EF-1-α) Cyclophylin (CYC) Metastatic lymph node 51 (MLN51) genes, as we used oligo(dT) as Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) Ubiquitin conjugating enzyme (UbcH5B) a primer for cDNA synthesis, Phosphoglycerokinase (PGK) and compared to specific down- β2-Microglobulin (B2M) stream primers, random hexamer β-Glucuronidase (GUS) primers have been shown to Hypoxanthine phosphoribosyltransferase (HPRT) overestimate mRNA copy num- Transcription factor IID TATA binding protein (TBP) bers by up to 19-fold (15). Transferrin receptor (TfR) Isolation of RNA and cDNA with TB. Housekeeping genes were and penicillin-streptomycin) at a cell Synthesis selected from those usually used and concentration of 1 × 106 cells/mL (final others found to be invariant in human volume of 0.5 mL) in 24-well plates. RNA was isolated from whole blood T cell culture (10). Here we report a Cells stimulated with TB antigen (12) collected in PAXgene tubes using the validation exercise to identify the most and harvested at baseline and days 2/3. PreAnalytiX PAXgene blood RNA kit suitable housekeeping gene in studies Days 4/5 were used for housekeeping (Qiagen) and from harvested PBMCs of pulmonary tuberculosis (PTB). gene expression studies. Control wells using the RNeasy® Mini kit (Qiagen). were challenged with phytohemagglu- All samples were DNase (Qiagen) tinin, Mycobacterium vaccae sonicate, treated. The RNA template was quali- MATERIALS AND METHODS or no antigen. Cell viability was as- tatively assessed and quantified us- sessed at each harvest with trypan blue, ing an Agilent 2100 Bioanalyzer with ® Patients and Samples and a proliferation assay was performed the RNA 6000 Nano Labchip kit for on days 4/5 to quantify the proliferative blood-derived RNA and the RNA 6000 Clinical samples (n = 28) were taken response to antigen challenge [bromo- Pico Labchip kit for culture-derived from four patients with smear-positive deoxyuridine (BrdU) cell proliferation RNA (all from Agilent Technologies, PTB, four healthy individuals, and the enzyme-linked immunosorbent assay Palo Alto, CA, USA). Total RNA ex- cells harvested from PBMC cultures of (ELISA); Roche Applied Science, In- traction varied from approximately 1–5 four additional patients with PTB. In dianapolis, IN, USA]. μg for blood and 0.1–0.8 μg from 5 × order to maximize variability, we chose 105 cells. To study the effect on house- subjects of different ages (from 26 to 50 Selection of Housekeeping Genes keeping gene expression, we used a years), sex, and ethnicity (Caucasian, fixed amount of input RNA for each Somalian, Indian, Chinese, Filipino, We first studied the gene expression cDNA reaction. Limited RNA quanti- and Black African). Informed consent levels of 10 housekeeping genes (des- ties dictated input RNA amounts to be was obtained from patients, and the ignated group 1) using a commercially 600 ng for PTB patients, 400 ng for relevant hospital ethics committees ap- available assay (TaqMan® human en- healthy volunteers, and 3 ng for the proved the study. dogenous control plate; Applied Bio- PBMC culture reactions. Reverse tran- Whole blood (20 mL) was taken, systems, Foster City, CA, USA) (13). scription reactions were performed fol- and 2.5 mL was immediately trans- The 10 genes investigated in this as- lowing the manufacturer’s instructions ferred into PreAnalytiX PAXgene™ say are shown in Table 1. We selected using Omniscript® Reverse Transcrip- blood RNA tubes (Qiagen, Valencia, group 2 genes as they were among 47 tase (Qiagen) for blood-derived RNA CA, USA) to fix the mRNA expression out of 535 maintenance genes found by and Sensiscript® Reverse Transcriptase profile (11). The remaining blood was microarray data to be relatively stable (Qiagen) for culture-derived RNA in transferred to a heparinized container in 11 different human tissues (14), and 60-μL reactions. and transported in a thermo flask at since they were found to be invariant in 37°C. Heparinized blood was incubat- human CD4 T cell cultures from cord Real-Time PCR ed at 37°C for a further 4 h on reach- blood (10). ing the laboratory, after which a further Our aim was to identify a house- The PCRs for group 1 genes were 2.5-mL sample was transferred to a keeping gene with minimal variability performed using the ABI Prism® 7700 PAXgene tube. under our different experimental condi- Sequence Detection System (Applied Blood taken from four of the pa- tions. As some of the genes of interest Biosystems). In each reaction, approxi- tients with PTB was layered over a Fi- are low-copy number cytokines, we mately 15 and 0.1 ng of reverse-tran- coll-Paque® gradient (Amersham Bio- anticipated that a significant change scribed RNA (based on the initial RNA sciences, Piscataway, NJ, USA) and the between study groups was likely to be concentration) was used for blood and PBMCs isolated. PBMCs were cultured small. Therefore, a standard deviation cell culture PCRs, respectively. The in triplicate in RPMI (supplemented of less than 2-fold from the mean ex- TaqMan endogenous control plate as- with 5% human AB serum, glutamine, pression level of the gene was chosen say was used according to the manufac- Vol.