MicroRNA expression profiling to identify and validate reference genes for relative quantification in colorectal cancer

Abstract

Background: Advances in high-throughput technologies and bioinformatics have transformed gene expression profiling methodologies. The results of microarray experiments are often validated using reverse transcription quantitative PCR (RT-qPCR), which is the most sensitive and reproducible method to quantify gene expression. Appropriate normalisation of RT-qPCR data using stably expressed reference genes is critical to ensure accurate and reliable results. Mi(cro)RNA expression profiles have been shown to be more accurate in disease classification than mRNA expression profiles. However, few reports detailed a robust identification and validation strategy for suitable reference genes for normalisation in miRNA RT-qPCR studies.

Methods: We adopt and report a systematic approach to identify the most stable reference genes for miRNA expression studies by RT-qPCR in colorectal cancer (CRC). High-throughput miRNA profiling was performed on ten pairs of CRC and normal tissues. By using the mean expression value of all expressed miRNAs, we identified the most stable candidate reference genes for subsequent validation. As such the stability of a panel of miRNAs was examined on 35 tumour and 39 normal tissues. The effects of normalisers on the relative quantity of established oncogenic (miR-21 and miR-31) and tumour suppressor (miR-143 and miR-145) target miRNAs were assessed.

Results: In the array experiment, miR-26a, miR-345, miR-425 and miR-454 were identified as having expression profiles closest to the global mean. From a panel of six miRNAs (let-7a, miR-16, miR-26a, miR-345, miR-425 and miR-454) and two small nucleolar RNA genes (RNU48 and Z30), miR-16 and miR-345 were identified as the most stably expressed reference genes. The combined use of miR-16 and miR-345 to normalise expression data enabled detection of a significant dysregulation of all four target miRNAs between tumour and normal colorectal tissue.

Conclusions: Our study demonstrates that the top six most stably expressed miRNAs (let-7a, miR-16, miR-26a, miR-345, miR-425 and miR-454) described herein should be validated as suitable reference genes in both high-throughput and lower throughput RT-qPCR colorectal miRNA studies.

Figure 1

Relative quantity and variation associated with each candidate reference gene. (a) Quantity of candidate reference miRNAs in colorectal tumour (n = 35) and normal (n = 39) tissues as expressed as quantification cycle (Cq) values. Boxplots depict median lines, interquartile-range boxes and outliers (*). Error bars represent range of values. No significant difference (p > 0.05, t-test) was found within all reference genes between tumour and normal tissues, thus supporting further evaluation of these genes as references. (b) Variation associated with each candidate reference gene. There was a significant difference in variance (p < 0.001, Bartlett's test) associated with each reference gene indicating differing stabilities. RNU48 and Z30 showed greater variance than miR-16, miR-345 and miR-425.

Figure 2

Equivalence test for candidate reference genes. Each line indicates the difference in logarithmic (base 2) expression level between tumour and normal tissues, with the upper and lower bars representing the upper and lower limits of symmetrical confidence intervals respectively. All genes were equivalently expressed with confidence intervals within fold change of 2 (deviation area 1, -1).

Figure 3

GeNorm analysis of candidate reference genes. (a) Ranking of candidate reference genes according to average expression stability. The least stable genes with the highest stability measure, M were excluded in a stepwise manner until the two most stable genes remained: miR-16 and miR-345. (b) Determination of optimal number of reference genes for normalisation. The GeNorm programme calculates a normalisation factor (NF) which is used to determine the optimal number of reference genes required for accurate normalisation. This factor is calculated using the variable V as the pairwise variation (Vn/Vn + 1) between two sequential NFs (NFn and NFn + 1). The number of reference genes is deemed optimal when the V value achieves the lowest, at which point it is unnecessary to include additional genes in the normalisation strategy. In this instance, the GeNorm output file indicated that optimal normalisation of gene expression could be achieved using the top five most stable reference genes.

Figure 4

Effect of reference gene choice on relative expression of oncogenic target miRNAs in colorectal tumour (n = 35) and normal (n = 39) tissues. Boxplots depict median lines, interquartile-range boxes and outliers (*). Error bar represent range of values. Relative expression of oncogenic miRNAs: (a) miR-21 and (b) miR-31 between colorectal tumour and normal tissues normalised to different reference genes with p values indicated. The use of the two most stable reference genes: miR-16 and miR-345 detected significant dysregulation both target miRNAs between colorectal tumour and normal tissues. Dysregulation of miR-31 was observed regardless of the choice of reference indicating that it's highly differentially expressed in colorectal cancer.

Figure 5

Effect of reference gene choice on relative expression of tumour-suppressor target miRNAs in colorectal tumour (n = 35) and normal (n = 39) tissues. Boxplots depict median lines, interquartile-range boxes and outliers (*). Error bar represent range of values. Relative expression of tumour-suppressor miRNAs: (a) miR-143 and (b) miR-145 between colorectal tumour and normal tissues normalised to different reference genes with p values indicated. The use of the two most stable reference genes: miR-16 and miR-345 detected significant dysregulation of both miRNAs between colorectal tumour and normal tissues.