Elsevier

Journal of Biotechnology

Volume 112, Issue 3, 9 September 2004, Pages 225-245
Journal of Biotechnology

Current issues for DNA microarrays: platform comparison, double linear amplification, and universal RNA reference

https://doi.org/10.1016/j.jbiotec.2004.05.006Get rights and content

Abstract

DNA microarray technology has been widely used to simultaneously determine the expression levels of thousands of genes. A variety of approaches have been used, both in the implementation of this technology and in the analysis of the large amount of expression data. However, several practical issues still have not been resolved in a satisfactory manner, and among the most critical is the lack of agreement in the results obtained in different array platforms. In this study, we present a comparison of several microarray platforms [Affymetrix oligonucleotide arrays, custom complementary DNA (cDNA) arrays, and custom oligo arrays printed with oligonucleotides from three different sources] as well as analysis of various methods used for microarray target preparation and the reference design. The results indicate that the pairwise correlations of expression levels between platforms are relative low overall but that the log ratios of the highly expressed genes are strongly correlated, especially between Affymetrix and cDNA arrays. The microarray measurements were compared with quantitative real-time-polymerase chain reaction (QRT-PCR) results for 23 genes, and the varying degrees of agreement for each platform were characterized. We have also developed and tested a double amplification method which allows the use of smaller amounts of starting material. The added round of amplification produced reproducible results as compared to the arrays hybridized with single round amplified targets. Finally, the reliability of using a universal RNA reference for two-channel microarrays was tested and the results suggest that comparisons of multiple experimental conditions using the same control can be accurate.

Introduction

DNA microarray technology has become an important tool in biological investigations by allowing researchers to measure the expression levels of thousands of genes simultaneously (Brown and Botstein, 1999, Choi et al., 2001, Lockhart and Winzeler, 2000, Schena et al., 1995. Generally, DNA microarrays are created in two basic forms: by DNA deposition or by in situ synthesis of oligonucleotide arrays. Deposited DNA materials can be in the form of polymerase chain reaction (PCR)-amplified complementary DNAs (cDNAs), pre-synthesized oligonucleotides, or genomic DNAs in the form of plasmids such as bacterial artificial chromosomes (BACs). Fabrication of in situ synthesized oligonucleotides by photolithographic masks was pioneered by Affymetrix Inc. (Santa Clara, CA). All these platforms employing cDNA or oligonucleotides use unique target amplification and labeling methods (Dorris et al., 2002, Eberwine et al., 1992, Feldman et al., 2002, Wang et al., 2000).

The availability of such multiple array platforms, which may also differ in probe preparation methods and array surface chemistry, raises the question of cross-platform agreement in gene expression measurements. Besides the many studies that have examined in detail the performance characteristics of single platforms [e.g., 50-mer oligonucleotides on glass (Kane et al., 2000) and cDNA arrays (Yue et al., 2001)], a number of comparative studies have been carried out. In Kuo et al. (2002), corresponding measurements from cDNA and Affymetrix GeneChip arrays were reported to show poor correlation for samples from human cancer cell lines; in Yuen et al. (2002), Affymetrix and laboratory-developed cDNA arrays were compared and concordant results were obtained on a number of genes with known regulation, although both platforms consistently underestimated the fold changes. In a comparison study between spotted 70-mer oligonucleotide arrays and Affymetrix for human samples, correlation coefficients of 0.8–0.9 were obtained for differential expression ratios (Barczak et al., 2003); similarly, between unmodified 70-mer oligonucleotide arrays on glass slides and cDNAs, a correlation coefficient of 0.80 (Wang et al., 2003) was observed. In Li et al. (2002), both sensitivity and specificities for selected genes were found to be very different between Affymetrix and commercial long cDNA arrays, and Affymetrix arrays appeared to perform more reliably. In Tan et al. (2003), Affymetrix, Agilent (cDNA probes) and Amersham (Codelink, 30-mer oligonucleotide probes) were shown to exhibit considerable divergence, with correlations in the range of 0.5–0.6 for both expression measurements and log ratios. All these reports have provided some answers as well as adding a barrage of new questions on the reliability of data from different microarray platforms.

In the present study, we systematically compared three different microarray platforms constructed from three different oligonucleotide sources (Affymetrix MG-U74A array, a custom cDNA array, and custom oligo arrays printed with oligonucleotides from three different sources). Quantitative real-time RT-PCR (QRT-PCR) on tens of selected genes was also performed to confirm the results obtained with each platform. We carry out an extensive analysis of the data. In addition to the correlation analysis on matched genes for overall agreement, our analysis includes estimation of coefficient of variations through regression, examination of dynamic ranges, comparisons of log ratios at different signal intensity levels, characterization of systematic under-estimation of the ratios relative to the RT-PCR results, and comparisons of probes mapping to the same gene in a given platform. In particular, the typical analysis by the Pearson correlation coefficient on the log ratios can be unstable, depending heavily on the details of the filtering criteria and simply due to the inherent properties of ratios. We therefore examine the correlations among the platforms as a function of signal intensity. We also suggest how the fold ratios should be modified for each platform based on the extent of under-estimation for log ratios.

DNA microarray hybridization using conventional methods where mRNA or total RNA is labeled and hybridized without amplification is particularly challenging when only a small amount of RNA is available. Using conventional labeling methods, even 20 μg of total RNA is often insufficient. This can ultimately lead to diminished signal intensity and thus introduce a great deal of spot to spot variation. PCR methods have been used to amplify signals (Iscove et al., 2002, Livesey et al., 2000, Puskas et al., 2002). However, it was not clear if the number of transcripts amplified was proportional to the original copy numbers due to the exponential amplification nature of PCR itself. A linear amplification method using T7 promoter has been developed and popularly used in past years (Dorris et al., 2002, Puskas et al., 2002, Wang et al., 2000). As starting materials extracted from various experiments become enormously scarce such as on the tens of nanogram scale, the need for a more significant amplification method is required. Here, we have developed and tested a double amplification method which allows the use of much smaller amounts of starting material.

Furthermore, a consensus has not yet been reached regarding the type of RNA reference sample most suitable for two color microarray experiments. Currently, a universal standard RNA reference sample, which combines total RNA from several cell lines, is available for use from Stratagene (La Jolla, CA) or BD Biosciences Clontech (Palo Alto, CA). Nonbiased testing using this universal standard reference will provide researchers with meaningful information to incorporate in future microarray experimental plans. To see if a standard for microarray RNA reference can be applied to research practice, a universal RNA reference for microarrays was tested.

Section snippets

Microarray fabrication

Amongst all arrays compared, only the Affymetrix GeneChip array is currently commercially available. The Affymetrix array used in this experiment was the Murine Genome U74A Version 2 GeneChip. Each gene represented on the Affymetrix array contains twenty 25-mer probes encompassing 200–300 bps derived from the gene. The remaining arrays included in our comparisons were all custom-designed arrays. The cDNA array is a 16K array designed by our laboratory. The cDNA clones are based on the RIKEN

Scanning

After hybridization and washing, cDNA and oligo arrays were scanned by the Agilent Scanner G2505A (Agilent) while Affymetrix arrays were scanned by the Agilent GeneArray Scanner (Agilent).

Double amplification method

Single amplification was performed as described in cDNA arrays. First strand cDNA synthesis for the second cycle of amplification began by incubating the amplified cRNA with 1 μl of 1 μg/μl random primer and sufficient amount of nuclease-free water at 70 °C for 10 min. Then, it was incubated with 4 μl of 5× first strand buffer, 2 μl of 0.1 M DTT, 1 μl of 10 mM dNTP, and 1 μl of 40 U/μl RNase inhibitor at 42 °C for 2 min. Immediately afterwards, 1 μl of 200 U/μl SuperScript II was added to the mixture and

Quantitative real-time RT-PCR

Quantitative real-time RT-PCR was performed using GeneAmp 5700 Sequence Detector System (Applied Biosystems, Foster City, CA). The measurement was normalized to an 18S ribosomal RNA control. To measure the copy number of each transcript, PCR amplified segment of each gene was cloned into pGEM-Teasy (Promega Corp., Madison, WI) and then cRNA was linearly amplified from NdeI-digested plasmid using MEGAscript T7 Kit (Ambion). cRNA was measured with spectrophotometer DU640 (Beckman Coulter Inc.,

Data analysis

The Affymetrix GeneChip information was extracted and data were computationally compared using the Affymetrix Microarray Suite Version 5.0. Genes flagged NC/MI/MD (not changed/marginal increase/marginal decrease) were removed. Genes with two or more replicate values were averaged and used for the analysis. The oligo and cDNA array information were extracted using the Agilent G2566AA Extraction Software Version A.6.1.1. Several criteria were used to filter the oligo and cDNA array data. Genes

Within-platform variability

A basic property of a good microarray platform is high reproducibility in repeated experiments. One way to measure reproducibility within a platform is to measure a correlation coefficient between the fold ratios of all genes in replicate chips (all fold ratios or fold changes hereafter refer to base 2 log ratios of spleen and liver comparisons). When only non-competitively hybridized arrays are considered, finding correlation between the actual expression measurements is natural, but we

Acknowledgements

We are grateful to the members of the Molecular Biology Laboratory of The Alliance for Cellular Signaling (http://www.signaling-gateway.org/, Alliance for Cellular Signaling) for their assistance. In addition, the authors would like to thank Drs. Mel Kronick, Douglas Amorese, and Stephanie Fulmer-Smentek for providing us access to the Agilent inkjet printing technology by arranging for printing of Caltech-supplied materials. We also thank Drs. Melvin I. Simon, Mel Kronick, and Winston Kuo for

References (27)

  • F.J. Livesey et al.

    Microarray analysis of the transcriptional network controlled by the photoreceptor homeobox gene Crx

    Curr. Biol.

    (2000)
  • A. Barczak et al.

    Spotted long oligonucleotide arrays for human gene expression analysis

    Genome Res.

    (2003)
  • J.P. Brody et al.

    Significance and statistical errors in the analysis of DNA microarray data

    Proc. Natl. Acad. Sci. U.S.A.

    (2002)
  • P.O. Brown et al.

    Exploring the new world of the genome with DNA microarrays

    Nat. Genet.

    (1999)
  • S. Choi et al.

    Gene expression profiles of light-induced apoptosis in arrestin/rhodopsin kinase-deficient mouse retinas

    Proc. Natl. Acad. Sci. U.S.A.

    (2001)
  • D.R. Dorris et al.

    A highly reproducible, linear, and automated sample preparation method for DNA microarrays

    Genome Res.

    (2002)
  • A.M. Dudley et al.

    Measuring absolute expression with microarrays with a calibrated reference sample and an extended signal intensity range

    Proc. Natl. Acad. Sci. U.S.A.

    (2002)
  • J. Eberwine et al.

    Analysis of gene expression in single live neurons

    Proc. Natl. Acad. Sci. U.S.A.

    (1992)
  • Feldman, A.L., Costouros, N.G., Wang, E., Qian, M., Marincola, F.M., Alexander, H.R. and Libutti, S.K., 2002....
  • A.J. Holloway et al.

    Options available—from start to finish—for obtaining data from DNA microarrays II

    Nat. Genet.

    (2002)
  • N.N. Iscove et al.

    Representation is faithfully preserved in global cDNA amplified exponentially from sub-picogram quantities of mRNA

    Nat. Biotechnol.

    (2002)
  • T.K. Jenssen et al.

    Analysis of repeatability in spotted cDNA microarrays

    Nucleic Acids Res.

    (2002)
  • M.D. Kane et al.

    Assessment of the sensitivity and specificity of oligonucleotide (50mer) microarrays

    Nucleic Acids Res.

    (2000)
  • Cited by (85)

    • Aging elevates metabolic gene expression in brain cholinergic neurons

      2008, Neurobiology of Aging
      Citation Excerpt :

      To demonstrate “basal” gene expression differences between the BF and BS (i.e., in the young groups), we used the “single-channel” processed (globally normalized) dataset. The validity of analyzing channels independently has been discussed previously (Park et al., 2004; Attoor et al., 2004). It appears that even though the hybridization is competitive, the amount of spotted probes is sufficiently large that the hybridization in each channel is nearly independent.

    • Volcano plots in analyzing differential expressions with mRNA microarrays

      2012, Journal of Bioinformatics and Computational Biology
    View all citing articles on Scopus

    The microarray data used in this paper were deposited into GEO (http://www.ncbi.nlm.nih.gov/geo/, Gene Expression Omnibus) under the accession numbers of GSM4652–4670.

    1

    Contributed equally to this work.

    View full text