Expression profiling with whole blood RNA samples has been difficult in the past due to the heterogeneous cellular nature of blood samples and the high concentration of globin mRNA present in the RNA. This article describes how Ambion's new technology, GLOBINclear-Human, significantly reduces the dilution effect of globin mRNA on expression profiling using microarrays by removing up to 95% of globin transcripts (>70% of the mRNA) from whole blood RNA samples prior to amplification.
Challenges of Blood Sample Analysis
Expression profiling is widely used to study physiological, pathogenic, and therapeutic responses of the transcriptome. For clinical research, blood is the most commonly used tissue; but blood samples present obstacles to gene expression analysis that are not encountered with most other tissues. Blood is made up of a heterogeneous population of erythrocytes, granulocytes, and other peripheral blood mononuclear cells (PBMC). The complex cellular nature of blood samples makes it difficult to detect differential gene expression that occurs in only a subset of cell types.
In addition, whole blood mRNA consists of a relatively large proportion of globin mRNA transcripts. Ambion scientists and others in the blood community, have estimated that globin transcripts represent as much as 70% of the total mRNA population. These “unwanted” globin transcripts decrease the detection sensitivity of less abundant mRNAs using microarray technology. Evidence of the dilution of mRNA by highly abundant globin mRNA has been observed on Affymetrix® GeneChip® arrays, where expression profiles from whole blood RNA show decreased Present calls and increased variability among biological replicates, relative to RNA from the white blood cell fraction of whole blood (which does not contain globin mRNA).
The GLOBINclear™-Human Kit employs a novel, nonenzymatic technology to remove >95% of the globin mRNA from whole blood total RNA. The procedure is rapid and robust (Figure 1). First, globin mRNA is removed from whole blood RNA using a novel hybridization capture technology. Then, the remaining RNA is further purified with a rapid magnetic bead-based clean-up procedure. The resulting RNA is a superior template for RNA amplification or for synthesis of labeled cDNA for array analysis. When used for expression profiling, GLOBINclear-treated blood RNA delivers a significant increase in sensitivity and a concomitant drop in variability, relative to untreated whole blood RNA.
The GLOBINclear Procedure
To remove globin mRNA, 1–10 µg of total RNA from human whole blood is mixed with a biotinylated Capture Oligo Mix in hybridization buffer (Figure 1). The mixture is incubated for 15 minutes to allow the biotinylated oligonucleotides to hybridize with the globin mRNA species. Streptavidin Magnetic Beads are then added, and the mixture is incubated for 30 minutes. During this incubation, streptavidin binds the biotinylated oligonucleotides, thereby capturing the globin mRNA on the magnetic beads.
The Streptavidin Magnetic Beads are then pulled to the side of the tube with a magnet and the RNA, depleted of the globin mRNA, is transferred to a fresh tube. The RNA is further purified using a rapid magnetic bead-based purification method. This simple and fast (~30 minutes) purification consists of adding an RNA Binding Bead suspension to the samples, and using magnetic capture to wash and elute the GLOBINclear processed RNA.
RNA Recovery and Amplification for Array Analysis
The GLOBINclear-Human procedure was primarily optimized for RNA amplification using Ambion's MessageAmp II aRNA Amplification Kit to synthesize RNA for microarray analysis. The goal was to develop a process to improve global gene expression screening of samples derived from human whole blood. Without GLOBINclear processing, aRNA amplified from whole blood samples shows a very distinct ~600 nt peak that represents amplified globin mRNA. Ambion scientists found that total RNA from whole blood that had been processed with GLOBINclear produced a very different aRNA profile, in which the distinctive globin peak was absent and the resulting aRNA produced a smooth curve (Figure 2). This curve resulted from the successful removal of both alpha- and beta- globin mRNA transcripts from the amplification reaction and was more similar to the aRNA profiles obtained from non-blood tissues.
Figure 3 shows yield data for RNA treated with both GLOBINclear and MessageAmp II RNA amplification.
The GLOBINclear procedure typically provides RNA yields of 70–90% of the starting material, depending primarily on the concentration and integrity of the input RNA. Since GLOBINclear depletes samples of >95% of their globin mRNA, (>70% of the mRNA present), GLOBINclear-processed RNA yields less aRNA from MessageAmp II amplification than an equivalent mass of unprocessed RNA. This is because GLOBINclear-processed RNA actually contains far less mRNA template for amplification in a given amount of total RNA than unprocessed whole blood RNA.
To determine the amount of RNA to process with GLOBINclear, consider the requirements of the downstream application. For example, we recommend processing between 1 and 10 µg of whole blood RNA with the GLOBINclear-Human Kit to prepare RNA for amplification with MessageAmp II products for array analysis. Note however, that more than 1 µg of input RNA may be required if a different amplification kit is used.
Effective Globin mRNA Removal
Throughout development of the GLOBINclear-Human Kit, qRT-PCR was used to assess depletion of alpha- and beta-globin mRNA. Figure 4 shows representative qRT-PCR data for the GLOBINclear RNA samples shown in Figure 3. TaqMan® probes for both alpha- and beta- globin were used to assess the level of globin mRNA before and after the GLOBINclear procedure. These data show that both alpha- and beta-globin mRNA were reduced substantially by GLOBINclear processing: alpha-globin mRNA was reduced between 40 and 60 fold (97.5–98.3% removal) and beta-globin mRNA was reduced between 80 and 180 fold (98.8–99.4% removal). Notice that there is an inverse relationship between globin mRNA depletion and the amount of whole blood RNA processed. However, even with this effect, globin mRNA depletion is still robust at the 7–10 µg whole blood RNA input level.
Array Analysis with GLOBINclear-Human
To determine the effect of GLOBINclear processing on microarray expression profiling, blood samples were drawn from 6 healthy donors, and RNA was isolated with a modified PaxGene® (Qiagen) protocol. RNA (3 µg from each donor) was processed with the GLOBINclear Kit in triplicate. Figure 5 shows the amount of remaining globin mRNA for each sample after depletion of globin mRNA as described in Figure 4. In this experiment, alpha- and beta-globin mRNA were depleted by well over 95% for all donor samples.
Next, 1 µg of each GLOBINclear-RNA sample was amplified using the MessageAmp II-96 Kit to synthesize biotinylated aRNA. Whole blood RNA (1 µg, unprocessed) was also amplified in triplicate for each donor. Two of the three replicate amplification reactions were randomly chosen and hybridized to Affymetrix GeneChip U133 Plus 2.0 Arrays (6 donors x 2 replicates x 2 conditions = 24 arrays).
Increased Percent Present Calls: The sensitivity of GLOBINclear RNA was evaluated by looking at several metrics, including the percentage of genes called Present by Affymetrix GeneChip Operating Software (GCOS). The percentage of genes called Present increased dramatically (range, 9–17%; mean, 13.2%) as a result of GLOBINclear processing of the RNA from all the blood donors (Figure 6). An average of 7206 additional genes were called present on U133 Plus 2.0 Arrays after GLOBINclear processing (mean Present calls=21698 ± 606). This represents fully 50% more Present genes than were called for the unprocessed whole blood RNA samples (mean Present calls=14492 ± 1347). Interestingly, the variability in sensitivity (Percent Present call standard deviation) seen within and between donors was significantly reduced after GLOBINclear processing. The exact reasons for this decreased variability are unknown; however, the most likely cause is the low signal-to-noise ratio generally associated with untreated whole blood RNA. The variability does not seem to be caused by differences in the levels of globin mRNA concentration between donors (data not shown).
Decreased 3'/5' Ratio: Another instance of increased variability associated with untreated whole blood RNA seen during the course of this study was 3'/5' ratios in array analysis. aRNA from samples processed with the GLOBINclear-Human Kit showed a significant decrease in 3'/5' ratios of the housekeeping genes, GAPDH and actin, compared to aRNA from unprocessed samples (Figure 7). 3'/5' ratios decreased from an average of 4.43 for GAPDH and 8.03 for actin in whole blood samples, to 1.65 for GAPDH and 1.64 for actin in GLOBINclear-processed samples, respectively. In the last row of the table, results were recalculated after removing the single outlier array from Donor D, and 3'/5' ratios, and the standard deviations decreased even further. The high 3'/5' ratios seen with untreated whole blood samples are most likely due to decreased sensitivity.
Reproducibility: The GLOBINclear method was further assessed for reproducibility by examining both the technical replicate concordance values and the correlation coefficients between replicates. Total concordant calls [P-P (Present), M-M (Marginal), A-A (Absent)], and both technical and biological correlations, were demonstrably higher for GLOBINclear-processed samples, compared to untreated whole blood samples (Figure 8). Here, as demonstrated in Figure 7, if the Donor D outlier is removed from consideration, the average concordance and average technical replicate correlation is further increased to 90.2% and 0.995, respectively. Also shown in Figure 8 are correlation coefficients for samples from donors B–F compared with a baseline sample from Donor A. As with the technical replicate comparisons, processing with GLOBINclear also reduces the variation seen among biological replicates.
Figure 8. Correlation and Percent Concordance for GLOBINclear™ RNA and Untreated Whole Blood RNA. Pearson’s correlation coefficient (r) and total percent concordance are presented for biological and technical replicates. All correlation coefficients were derived from signal values estimated using Robust Multichip Average (RMA, www.bioconductor.org). Concordance is the number of concordant calls [P-P (Present), M-M (Marginal), or A-A (Absent)] divided by the total number of calls (or genes = 54,675). Present/Marginal/Absent calls were determined with Affymetrix® GeneChip Operating Software (GCOS).
Note 1: RMA signal values were averaged for donor technical replicates and correlation coefficients were computed for all donors vs. Donor A (baseline).
Note 2: Concordance and correlation coefficients for technical replicates were computed for each donor.
The GLOBINclear-Human Kit utilizes a biotin/streptavidin magnetic bead capture method that removes more than 95% of unwanted alpha- and beta-globin mRNA transcripts from human whole blood RNA. GLOBINclear-processed whole blood RNA provides increased sensitivity and decreased variability when used for microarray expression analysis. The method is simple, highly reproducible, and takes less than 1.5 hours to complete.
The kit comes with enough reagents to process 20 globin mRNA depletion reactions. Whole blood RNA isolation kits and magnetic stands are sold separately.
Penn Whitley, Sharmili Moturi, Jose Santiago, Charles Johnson, Robert Setterquist • Ambion, Inc.