| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Briefs |
ViaGen Inc., Austin, TX
aaddress correspondence to this author at: ViaGen Inc., 12357-A Riata Trace Pkwy., Suite 100, Austin, TX 78727; fax 512-401-5919, e-mail wenli.zhou{at}viagen.com)
The SMART (switching mechanism at the 5' end of RNA templates of reverse transcriptase) method (1)(2), in combination with PCR, has been used to amplify minute samples obtained from sources such as laser capture microdissections and biopsies for gene expression analysis using oligonucleotide microarray gene chip technology (3)(4)(5)(6)(7). Amplifications >105-fold can be achieved in a short time. However, reports in the literature have rarely discussed the feasibility and reproducibility of PCR-based methods when the amount of starting total RNA in a sample is only a very few nanograms or less, which is often the case for clinical samples.
We developed a modified SMART-based PCR protocol with which reliable and robust gene expression data can be obtained from 1 ng of starting total RNA. We also investigated sources of array data variations.
The modified protocol (see Supplement 1 in the Data Supplement that accompanies the online version of this Technical Brief at http:www.clinchem.org/content/vol51/issue12) has three major changes compared with the published standard procedure (5): the volume of the reverse transcription reaction is reduced to 5 µL; 50 ng of poly(dG/dC) carrier (Sigma) is added to the reverse transcription reaction; and PowerScript reverse transcriptase (BD Biosciences Clontech) is used instead of SuperScript II reverse transcriptase (Invitrogen). When we used the modified protocol to reverse-transcribe and amplify 10, 1, 0.1, and 0.01 ng of human heart or liver total RNA (BD Biosciences Clontech), we obtained ample cDNA from one round of 23, 26, 29, and 32 PCR cycles, respectively (Fig. 1A
). The addition of poly(dG/dC) facilitates the reverse transcription of longer transcripts as indicated by larger ranges of cDNA sizes and clearer band patterns, which are important for getting consistent good array data, but it does not adversely interfere with the reactions. Use of Superscript II instead of PowerScript in the reverse transcription step under otherwise similar conditions failed to yield visible amounts of cDNA in the agarose gel when starting total RNA was 
10 ng, and a second round of PCR had to be performed to yield sufficient cDNA for synthesizing adequate amounts of cRNA for arrays. The number of genes detected on the arrays, however, was considerably lower than that obtained with PowerScript reverse transcriptase.
|
The cDNA obtained with the modified protocol from 1, 10, and 100 ng of human total heart or liver RNA was in vitro–transcribed into complementary RNA (cRNA), which was then labeled with Bio-11-CTP and Bio-16-UTP (Enzo Biochem) by use of MAXIscriptTM reagents (Ambion). The cRNA was then hybridized to the human GeneChip H133A (Affymetrix) as described previously (5). The numbers of genes detected as present (
30%) were comparable among arrays prepared with various starting amounts of total RNA. We determined the reproducibility of the expression signals detected from these minute samples by running 2 replicate assays. Shown in Fig. 1B
are the M-A plots of replicate arrays (M represents the log ratio of 2 replicates, and A represents the averaged logarithmic intensity). At high signal intensities (averaged raw signal intensity >1024 or logarithmic intensity >10), except for a few outliers, the data in all 3 plots were scattered around the line M = 0, indicating very good reproducibility even when starting the amount of total RNA was 1 ng. However, at lower signal intensities, particularly when the averaged raw signal intensity was <256 or the logarithmic intensity was <8, the data in the plots were scattered farther from the M = 0 line. The smaller the amount of starting total RNA used, the greater was the observed variation. When replicate arrays were prepared from 100 ng of starting total RNA, 1.2% of the examined probe sets exhibited more than 2-fold differences in expression values between 2 replicates; this number increased to 4.2% for arrays prepared from 10 ng of starting total RNA and 18.2% for 1 ng of starting total RNA.
We then conducted quadruple replicate assays for arrays prepared from 1 ng of starting total RNA. The 4 replicates were randomly divided into 2 groups of duplicate samples. Mean expression intensities from 2 arrays were calculated for each probe set in each group. The variation was greatly reduced, with 7.6% of probes sets exhibiting greater than 2-fold differences between duplicate groups. For the following studies, therefore, the mean of 2 replicates was used when the starting total RNA was 1 ng.
To investigate whether the modified protocol can faithfully amplify massive genes from 2 samples under different conditions so that the expression ratio for each gene in the 2 samples will be preserved and correlate to that obtained from linear amplification, we obtained independent gene expression ratios for human heart and liver, using both the linear amplification scheme with MessageAmpTM reagents (Ambion) and the modified protocol developed for nanograms of total RNA, with 2 µg each of heart or liver total RNA as starting material. Both methods generated strong comparable hybridization signals in the gene chips (see Supplement 2 in the online Data Supplement). The Pearson correlation coefficients for the gene expression ratios in human heart and liver obtained with the 2 methods were 0.95, 0.94, and 0.92 when the amounts of starting total RNA for the PCR-based amplification were 100, 10, and 1 ng, respectively. Despite the amplification preference associated with each method (3)(4)(5), we achieved good correlation between expression ratios obtained from both methods for the majority of transcripts.
The fact that this modified PCR-based protocol reproducibly preserves expression ratios can be further demonstrated by comparing gene expression ratios obtained under different amplification conditions, i.e., various amounts of starting total RNA with various numbers of PCR cycles. Expression arrays were prepared from 100 ng of total RNA amplified for 20 PCR cycles, 10 ng of total RNA amplified for 23 PCR cycles, or 1 ng of total RNA amplified for 26 PCR cycles. The correlation coefficients (r) for the 100- and 10-ng, 100- and 1-ng, and 10- and 1-ng array pairs were 0.97, 0.95, and 0.92, respectively. The preservation of expression ratios over such a wide range of PCR cycles and template concentrations can be explained theoretically (see Supplement 3 in the online Data Supplement) and makes more convenient the study of clinical samples, in which the amount of starting material may be very different.
During preparation of cRNA for chip hybridization from nanograms of total RNA, reverse transcription and PCR are the only steps involving minute amounts of reactants. The fact that the array data became more variable when the amount of starting total RNA was <10 ng indicates that this variation is introduced primarily in these 2 early reactions. To further understand which of the 2 reactions, reverse transcription or PCR, generates the larger variation, we performed 3 sets of replicate assays. In the first assay, we reverse-transcribed 100 ng of total heart RNA. We then used 1/100th of the first-strand cDNA, which was quantitatively equivalent to the cDNA reverse-transcribed from 1 ng of total starting RNA, for PCR amplification and cRNA synthesis. In the second assay, we reverse-transcribed 100 ng of total heart RNA and used all the first-strand cDNA for PCR amplification and cRNA synthesis. The third assay was the same as the second assay except that the starting total RNA used was 1 ng. Replicate cRNA samples for each set of arrays were then hybridized to gene chips, and the Pearson correlation coefficients were 0.87, 0.96, and 0.88, respectively, for the 3 sets of replicate arrays (see Supplement 4 in the online Data Supplement). If we regard 1 ng as minute, the second set of arrays (r = 0.96) does not have any minute amounts of reactants. In the first set of arrays, only the PCR step involved minute amounts of reactants. In the third set of arrays, both reverse transcription and PCR used minute amounts of reactants, but the Pearson correlation coefficients of these 2 sets of arrays were basically the same, 0.87 and 0.88, suggesting that the PCR step contributes the most to the variation between arrays prepared from 1 ng of starting total RNA. Our findings are consistent with mathematical simulations of PCR amplification illustrating that when transcripts with small copy numbers are amplified, the amplified gene population distribution is broad and has a large standard deviation (8)(9).
In conclusion, the modified protocol greatly increases the sensitivity of array detection when as little as 1 ng of total RNA is amplified. Highly reproducible array data were obtained from different quantities of starting RNA that were amplified with different numbers of PCR cycles. The variation associated with minute samples can be effectively reduced with multiple replicates. The findings that when a minute sample is reverse-transcribed and amplified the sensitivity is determined mainly at the step of reverse transcription and that the reproducibility is largely controlled by the PCR reaction could guide the development of more powerful enzymes and methods.
Acknowledgments
We thank Dr. Kelcey Walker for editing the manuscript.
References
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
