Choosing reference genes for qPCR normalisation

From OpenWetWare
Jump to navigationJump to search

Which reference genes should I use for my qPCR experiment? Quantifying mRNA via cDNA levels as in a quantitative reverse transcriptase PCR (QRT-PCR) hinges on the references you choose. If you pick only one reference gene and your pick is not constant across different conditions or samples, your results will be skewed. Pick several and check whether they satisfy the criteria for a good reference gene.

The ideal reference gene

A mRNA used as reference or standard of a QRT-PCR (and other experiments) should have the following properties:

  • expressed in all cells
  • constant copy number in all cells
  • medium copy number for more accuracy (or similar copy number to gene of interest)

Common reference genes

  • ribosomal proteins (e.g. RPLP0)
  • PPIA peptidylprolyl isomerase A = cyclophilin A
  • MHC I (major histocompatibility complex I)
  • TUBB β-tubulin (common cytoskeletal enzyme)
  • ACTB β-actin (common cytoskeletal enzyme) [1], [2]
  • YMHAZ tyrosine 3/tryptophan 5 -monooxygenase activation protein, zeta polypeptide
  • B2M β2-microglobulin
  • UBC ubiquitin C
  • TBP TATAA-box binding protein
  • GUSB β-glucuronidase
  • HPRT1 hypoxanthine-guanine phosphoribosyltransferase

Common but problematic references

GAPDH has many functions besides the most well known in the glycolytic pathway [5]. Its levels are not constant [Zhu 2001 PMID 11237753] and vary more than for other genes across different tissues [Radonic 2004 PMID 14706621]. Problems using GAPDH as a qPCR reference gene have been published previously [Ke 2000 PMID 10799275, Suzuki 2000 PMID 10948434].
  • ribosomal RNAs (28S or 18S)
Ajeffs 06:55, 21 April 2007 (EDT): 18S is generally a terrible choice for a reference gene thanks to the combination of (i) high abundance (creating a 1:100 dilution of template to run in parallel with neat template just for 18S is a complete drag); and (ii) having different degradation characteristics to mRNAs (it appears to be more resistant to degradation). However, if you can show that you have screened 5-10 reference genes, and 18S is still the best for your specific situation then so be it (but do try 28S if you or you PI is hung-up on 18S).
  • Panels of "housekeeping genes"
All genes used for normalization can show problems in one or the other condition. There are always conditions in which their expression differs significantly from their general level of expression. That is why reference genes should be validated for one's condition of interest. Ideally, one should choose from the complete genome rather than from a gene panel. Genevestigator RefGenes [6] is an open access online tool that uses a very large microarray database to identify genes that are most stable in conditions similar to that of your own experiment.

Reference genes across tissues

If you are comparing mRNA/cDNA levels from different tissue it is especially important that reference gene levels are close to constant across different tissues. Radonić et al compared 13 putative reference gene levels in 13 different human tissues [PMID 14706621]. The results are summarised below:

  • genes with the smallest range (most constant levels): TBP, RP2, Act, Tub, PLA
  • genes with the largest range (unsuitable for cross-tissue comparison): HPRT, Alb, PBGD, GAPDH, β2M
  • genes undetectable in tissue: Alb - colon; PPIA - ovaries; HPRT - prostate, testis, ovary, small intestine, colon; PBL, skeletal muscle; PBGD - skeletal muscle; TBP - lung, prostate, colon; G6PDH - colon
  • genes detected in all tissues: GAPDH, Act, Tub, β2M, L13, PLA, RP2

(note the source Fig 2 is sometimes impossible to read and the describing text is incomplete; that might have lead to some errors above)


  • Ajeffs 06:55, 21 April 2007 (EDT): In addition to the given requirements of good (well, acceptable) specificity and efficiency of the reference gene primers, the next most important aspect of reference gene selection is stability. I don't care if the CT value of my reference genes (yes, genes, not gene) is close to the target genes/s or not - as long as the efficiency of all the primers is similar, and they are all working within their respective limits of detection i.e. linear range, then the stability of the reference genes between samples, treatments, etc. is the most crucial aspect of generating believable qPCR results.