- Email: [email protected]
Recovery of microarray-quality RNA from frozen EDTA blood samples
Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
Contents lists available at ScienceDirect
Journal of Pharmacological and Toxicological Methods j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / j p h a r m t o x
Original article
Recovery of microarray-quality RNA from frozen EDTA blood samples Johanna M. Beekman a,⁎, Joachim Reischl a, David Henderson a, David Bauer a, Rainer Ternes a, Carol Peña b, Chetan Lathia b, Jürgen F. Heubach a a b
Clinical Pharmacology, Bayer Schering Pharma AG, 13353 Berlin, Germany Clinical Pharmacology, Bayer HealthCare Pharmaceuticals, Montville, NJ 07045, USA
a r t i c l e
i n f o
Article history: Received 19 August 2008 Accepted 8 October 2008 Keywords: EDTA blood Human Methods Microarray analysis PAXgene Quality control RNA extraction RNA stabilizing additive RIN
a b s t r a c t Introduction: The availability of blood collection systems with RNA stabilizing additives (e.g. PAXgene) has opened the field for gene expression profiling in large multi-center clinical trials. Here, we investigated whether the PAXgene system also offers a method for extraction of RNA from frozen EDTA blood samples, which do not yield RNA of high enough quality for RNA expression profiling, when extracted with standard protocols. Methods: Whole blood was obtained from six healthy volunteers in conventional EDTA tubes and frozen. The thawed EDTA blood was transferred into PAXgene tubes, and the RNA was extracted using the PAXgene RNA extraction kit. Microarray analysis was performed to asses the effect of RNA quality on gene expression profiles. Results: The RNA yield of the transferred samples was 1.76 ± 0.88 µg/ml. This yield was clearly lower than the yield from a PAXgene reference group (2.84 ± 0.62 µg/ml), but considerably higher than the yield resulting from a standard protocol usually applied to fresh EDTA blood samples (0.07 ± 0.06 µg/ml). The RNA integrity number (RIN) of the transferred samples was 6.1 ± 0.8 as compared to 9.8 ± 0.1 for the PAXgene reference. Microarray analysis of the extracted RNA suggested that samples with RIN values above 5 produce data that fulfill the quality criteria defined by the manufacturer. Discussion: The transfer of thawed EDTA blood into PAXgene blood collection tubes offers a method to recover sufficient RNA of acceptable quality for microarray experiments. © 2008 Elsevier Inc. All rights reserved.
1. Introduction The use of gene expression analysis for biomarker discovery is widely practiced. Gene expression analysis of blood cells as an easily accessible surrogate tissue allows for minimally-invasive repeated measurements, and has been used in clinical studies for prospectively planned biomarker discovery. Blood cell RNA profiling has been used to identify candidate biomarkers for disease classification (Müller et al., 2002; Burczynski et al., 2006), prediction of prognosis (Burczynski et al., 2005) and drug responder/non-responder identification (McLean et al., 2004; for a review see also Burczynski & Dorner, 2006). The introduction of blood collection systems with RNA stabilizing additives and accompanying RNA extraction kits has significantly improved the quality of RNA isolated from blood samples collected in multicenter clinical trials (Rainen et al., 2002; Thach et al., 2003), and the convenience of processing the samples. Using such stabilization systems the RNA does not need to be extracted immediately; instead, blood samples can be collected, frozen and stored at the clinical site before batched dry ice shipment to a central
⁎ Corresponding author. Clinical Biomarkers, Clinical Pharmacology, Bayer Schering Pharma AG, Müllerstr. 178, D-13353 Berlin, Germany. Tel.: +49 30 468 12554; fax: +49 30 468 92554. E-mail address: [email protected] (J.M. Beekman). 1056-8719/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.vascn.2008.10.003
RNA extraction and analysis center, additionally reducing interoperator variability in RNA quality. Despite the availability of RNA stabilizing blood systems, there may be a need to perform RNA profiling experiments with legacy blood samples that were collected with common blood collection systems, such as EDTA tubes, and frozen after a variable time at ambient temperature. The main challenge of RNA extraction from frozen EDTA blood samples is to obtain sufficient RNA of acceptable quality for microarray hybridization. Whereas RNA stabilizing additives cause controlled cell lysis in an environment of RNA protection, the freezing of EDTA blood destroys a large fraction of blood cells, exposing the RNA to released enzymes and subsequent RNA degradation. A second challenge in using frozen EDTA blood as a starting material for RNA extraction is that standard protocols for RNA extraction from fresh blood include a hypotonic erythrocyte lysis step before white blood cells are harvested for RNA extraction. This step minimizes the adhesion of RNA to hemoglobin and cell debris but cannot be applied to thawed EDTA blood, since the cells are already destroyed by freezing and thawing. The RNA in these samples is trapped in the cell debris and is difficult to isolate. Here, we investigated the possibility of rescuing hybridizationquality RNA from frozen EDTA blood samples. We compared a standard RNA isolation protocol commonly used on fresh EDTA blood samples to a method where thawed EDTA blood was transferred
J.M. Beekman et al. / Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
45
Fig. 1. Flow chart of the four experimental protocols evaluated for RNA isolation from blood cells. Whole blood samples were collected from 6 healthy volunteers in blood collection tubes containing either Potassium EDTA or a RNA-stabilizing additive (PAXgene). RT: room temperature.
into blood collection tubes with RNA stabilizing additive (PAXgene™ blood RNA system) followed by RNA extraction with the respective RNA extraction kit. A state-of-the-art reference group of blood samples was collected directly into PAXgene tubes. We demonstrate that the transfer of thawed EDTA blood into PAXgene tubes and subsequent RNA extraction allows the isolation of RNA of sufficient yield and acceptable quality for microarray analysis. 2. Methods 2.1. Sample collection and storage The research was carried out in accordance with the principles of the current version of the Helsinki Declaration. After approval of the Institutional Review Board, blood from six healthy volunteers (aged 35 to 55 years) was collected into Potassium EDTA tubes (2.7 ml, Sarstedt) and PAXgene™ blood RNA tubes (Becton-Dickinson) according to the manufacturer's directions. Volunteers provided written informedconsent. The filled EDTA tubes were either put immediately into a −20 °C freezer or left at room temperature for 3 h prior to freezing. As a modification to the procedures suggested by the manufacturer, the blood collected in PAXgene tubes was placed immediately into the −20 °C freezer. 2.2. RNA isolation and quality assessment Frozen EDTA tubes were thawed on ice during approximately 2 h. RNA was then extracted using the QIAamp kit (Qiagen) or by transferring 2.5 ml of EDTA blood into a PAXgene tube followed by 16–21 h incubation at room temperature and RNA extraction as described below for the PAXgene samples. The RNA was eluted in two steps utilizing 40 µl buffer for each step. Frozen PAXgene tubes were thawed and incubated for 16–21 h at room temperature followed by RNA isolation using the PAXgene kit including QIAshredder homogenization and on-column DNase digestion as described in the manufacturer's handbook (Qiagen). RNA concentration was deter-
mined using a NanoDrop spectrophotometer (NanoDrop Technologies), and RNA integrity was assessed by determining the RIN (Schroeder et al., 2006) on a 2100 Bioanalyzer (Agilent technologies). 2.3. RNA expression analysis Total RNA, isolated from thawed EDTA blood using the QIAamp kit, was labeled according to the Affymetrix GeneChip Expression Analysis Technical Manual (Affymetrix). To obtain enough RNA for labeling, 2 RNA samples from each donor (2 extractions of 2.7 ml blood each) had to be pooled. The RNA amount of the pooled samples ranged from 23 to 751 ng. Total RNA isolated using the PAXgene system was labeled accordingly, however using BLOCK reagent (GeneLogic) in the firststrand cDNA synthesis reaction. BLOCK reagent is a mixture of hemoglobin-specific oligonucleotides and a T7-Oligo(dT) promoter primer. It is used to increase overall assay sensitivity by blocking cDNA synthesis from hemoglobin transcripts, which originate from the reticulocytes in the whole blood samples. The labeled cRNA was quantified using a NanoDrop spectrophotometer and checked for quality on the Agilent Bioanalyzer. Resulting cRNA was hybridized to Affymetrix HG U133 Plus 2.0 GeneChip microarrays (containing approximately 54,000 probe sets representing 47,400 transcripts and variants; Affymetrix), and the arrays were washed and stained using the GeneChip Fluidics Station 450 (Affymetrix). The stained microarrays were scanned at 570 nm using a confocal laser scanner (GeneChip-3000 Scanner, Affymetrix), and the resulting ⁎.DAT-files were quantified using GCOS software (Affymetrix). The resulting ⁎.CEL-files were condensed using the MAS 5.0 algorithm (Affymetrix). Selected quality criteria were extracted from the MAS 5.0 reports. Analysis of expression data was performed by using the Expressionist Analyst Pro Version 4.5 software (GeneData AG). In the first analysis step a probe set expression value was considered valid if it reached a p-value below 0.04 in the one-sided Wilcoxon signed rank test of the MAS 5.0 condensing algorithm (Liu et al., 2002). The overall data structure was investigated using hierarchical clustering.
46
J.M. Beekman et al. / Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
3. Results In the present study we compared the yield and quality as well as the suitability for microarray analysis of RNA obtained from blood cells according to four different experimental protocols, summarized in Fig. 1. Three protocols utilized frozen EDTA blood to simulate the condition of legacy samples and to develop an efficient method to recover RNA from this inappropriate starting material. EDTA blood frozen immediately, thawed, and extracted using the QIAamp RNA extraction kit was defined as the Standard Protocol. Two investigational protocols where EDTA blood was frozen, thawed, transferred to PAXgene tubes, and extracted using the PAXgene RNA isolation kit are referred to as Transfer Protocol A (samples frozen immediately) and B (samples frozen after 3 h at room temperature). The room temperature incubation in Transfer Protocol B was designed to mimic possible delay until freezer storage of samples at clinical sites. In a fourth protocol, referred to as the Reference Protocol, blood was collected into containers prepared with RNA stabilizing additive (PAXgene tubes), frozen immediately, and extracted using the PAXgene RNA isolation kit. This Reference Protocol is expected to produce high yield, high quality RNA. 3.1. RNA extraction: yield and quality The Standard Protocol resulted in poor RNA yields as compared to the Reference Protocol (0.07 ± 0.06 µg/ml versus 2.84 ± 0.62 µg/ml;
Fig. 3. Comparison of microarray quality parameters for blood samples processed using the four different experimental protocols. (A) GAPDH 3′/5′, (B) scaling factor, and (C) percent presence calls. Results from individual microarrays are shown (n = 6 per protocol).
p = 5.7 ⁎ 10− 7; T-test; Fig. 2A). Transfer Protocols A and B produced intermediate RNA amounts with a tendency towards lower yields after 3 h of room temperature storage prior to freezing (1.76 ± 0.88 µg/ml versus 1.29 ± 0.45 µg/ml; p = 0.27; T-test; Fig. 2A). The RNA integrity number (RIN), determined using the algorithm provided in the Agilent Bioanalyzer 2100 expert software provides an objective assessment of RNA integrity using the entire electrophoretic trace of the RNA sample instead of using the ratio of the 18S and 28S ribosomal RNAs alone. The RIN values for RNA resulting from the two Transfer Protocols were close to 6 (6.1 ± 0.8 and 6.4 ± 0.8) and lower than those from the Reference Protocol (9.8 ± 0.1; Fig. 2B). Due to the low concentration of RNA obtained from the Standard Protocol, quality assessment was not possible for this group. 3.2. GeneChip analysis
Fig. 2. Yield (A) and quality (B) of RNA extracted from whole blood samples using the four different protocols. The graphs show mean values ± standard deviations from n = 6 samples. RIN: RNA integrity number. N.D., not done.
For each of the four protocols, RNA samples from six donors were processed to cRNA, and the cRNA yield and distribution were analyzed. The cRNA size distribution resulting from the Reference Protocol showed a maximum around 1500 bases (data not shown). The three protocols utilizing frozen EDTA blood resulted in cRNA size distributions between 250 and 1000 bases, indicating lower RNA integrities. A trend towards longer cRNA was seen using the Transfer Protocols (as
J.M. Beekman et al. / Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
compared to the Standard Protocol), but this was not quantifiable. None of the cRNA profiles showed a hemoglobin peak, confirming a successful erythrocyte lysis in the standard protocol and efficient use of the BLOCK reagent for the transfer and reference protocol samples. To obtain enough RNA from the Standard Protocol, two RNA samples from each donor, i.e. two extractions of 2.7 ml blood each, had to be pooled. Despite pooling the minimum amount of 10 µg cRNA necessary for GeneChip hybridization was not achieved for two individuals. In these cases the total amount of resulting cRNA was used for hybridization (7.3 µg for Stan5 and 5.4 µg for Stan6; see Fig. 4A for reference). After hybridization, washing, staining and scanning, the gene expression data was processed using MAS 5.0, and a data quality check was performed. The noise and background levels of all GeneChips were within quality limits defined by the microarray manufacturer (data not shown). The GAPDH 3′/5′ ratio, the scaling factors, and the percent present calls, however, demonstrated clear differences among the four experimental protocols. The GAPDH 3′/5′ ratio, which is a quality measure of the cRNA applied to the GeneChip (reflecting RNA degradation, with larger numbers indicating increased degradation), increased in the order Reference b Transfer A/B b Standard Protocol, indicating decreasing RNA integrity (Fig. 3A). All samples processed according to the Standard Protocol showed GAPDH ratios higher than the quality limit of 3 recommended by Affymetrix, whereas no samples from the Reference Protocol and Transfer Protocol A and only one sample from Transfer Protocol B exceeded this limit. In addition, the Standard Protocol produced the GeneChips with the lowest overall signal, reflected by the high scaling factors necessary for normalization (Fig. 3B). Application of the Transfer Protocols to frozen EDTA blood resulted in a marked reduction of scaling factors. Lastly, the Transfer Protocols clearly increased the percent present calls as compared to the Standard Protocol, allowing the detection and quantification of a
47
considerably higher number of transcripts (Fig. 3C). For all three quality parameters, blood processing using the Reference Protocol achieved the best assay performance with low interindividual variability. Hierarchical clustering and principal component analysis of the microarray data are shown in Fig. 4. Three clusters are obvious (Fig. 4A): The topmost cluster with little interindividual variability contains the six samples processed according to the Reference Protocol. The cluster in the middle is composed of samples processed using the Transfer Protocols, where A and B samples from the same donor clustered in pairs. In other words, little separation is seen between the immediately frozen samples and those which were frozen after 3 h at room temperature; the interindividual differences of gene expression profiles exceeded the variability introduced by the 3 h incubation at room temperature. The cluster at the bottom encompasses the samples processed using the Standard Protocol. The principal component analysis shows the same cluster formation and the same variability within the clusters (Fig. 4B). Additional technical parameters such as call concordance and false change may offer a more quantitative comparison of the outcome of different protocols applied. Call concordance is defined as the average percentage of probe sets giving concordant Present and concordant Absent calls between pair-wise comparisons (intrasubject comparisons). In comparison to the Reference Protocol, the mean call concordances for the Standard Protocol and for Transfer Protocols A and B were 77.1 ± 5.5%, 87.5 ± 3.6% and 85.2 ± 3.2%, respectively. When compared to the Standard Protocol, the mean call concordances for the Transfer Protocols A and B were 81.9 ± 5.5% and 83.7 ± 4.3% respectively. The two Transfer Protocols A and B showed a mean call concordance of 90.2 ± 1.0%, thereby even fulfilling the Affymetrix requirement for technical replicates with a call concordance above 90%, when the same hybridization cocktail is applied on two different chips.
Fig. 4. (A) Hierarchical clustering and (B) principal component analysis of all microarray data obtained from the Reference Protocol (Ref), Transfer Protocol A (TransA), Transfer Protocol B (TransB) and Standard Protocol (Stan). Numbers specify the different blood donors.
48
J.M. Beekman et al. / Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
A false change is defined as a two-fold or greater change in a comparison analysis between replicate samples. For technical replicates, a false change of 1% or less is accepted, according to Affymetrix standards. In comparison to the Reference Protocol, the mean false change rates for the Standard Protocol and for Transfer Protocols A and B were 8.8 ± 1.1%, 9.9 ± 1.8% and 11.0 ± 1.4%, respectively. When compared to the Standard Protocol, the mean false changes for the Transfer Protocols A and B were both 7.1 ± 0.9% and 7.1 ± 1.0%, respectively. The two Transfer Protocols A and B themselves showed a mean false change of only 1.3 ± 0.3%, which again almost fulfilled the requirement for technical replicates. 3.3. Correlation between RNA quality and GeneChip performance Transfer Protocol A was successfully applied to extract RNA from a set of legacy EDTA blood samples collected at multiple sites and thus expected to have high variability in quality due to differing sample processing, storage conditions, and storage times. Four samples were selected for GeneChip analysis from each of the following RIN classes: RIN 2–3, RIN 3–4, RIN 4–5, RIN 5–6, RIN 6–7 and RIN N 7. Again, upon hybridization background and noise levels were not affected by the RNA quality (data not shown). Analysis of the 24 GeneChips, however, indicated a clear negative correlation between the RIN score and the GAPDH 3′/5′ ratio (Fig. 5A). All samples with RIN values above 5 resulted in GAPDH 3′/5′ ratios below 3. In addition, there was a negative correlation between the RIN and the scaling factor (Fig. 5B),
Fig. 5. Correlation between RNA integrity number (RIN) and (A) GAPDH 3′/5′ ratio, (B) scaling factor, and (C) percent present call.
as well as a positive correlation between RIN values and the percent present call (Fig. 5C). 4. Discussion 4.1. The Transfer Protocol for RNA recovery The purpose of this work was to develop a method for RNA extraction from frozen EDTA blood samples, providing sufficient yield and acceptable quality for gene expression analysis. As many groups have shown, the PAXgene method utilized as the Reference Protocol yielded superior RNA quantity and quality (Thach et al., 2003; Debey et al., 2004; Kim et al., 2007). We do not propose the Transfer Protocol as an approach for prospective blood collection with the intention of performing expression analysis, but as a way to extract RNA from legacy blood samples collected using EDTA tubes, provided that the patient informed consent allows for RNA analyses. Our recovery scenario was based on EDTA blood but the same approach may be feasible for other types of blood samples, for example citrate or heparin preparations. 4.2. RNAs from the four protocols show different gene expression profiles The two experimental Transfer Protocols are hybrid methods utilizing EDTA tubes for blood collection (as in the Standard Protocol) and PAXgene processing for RNA isolation (as in the Reference Protocol). The two procedures gave similar RNA yields of comparable quality. Hierarchical clustering demonstrated that the gene expression profiles of samples processed using the Transfer Protocols clustered separately from the profiles of the two other protocols. Within this group, samples from the same donor stored in the freezer either immediately or after 3 h at room temperature clustered in pairs. This indicates that a 3 h room temperature incubation, intended to mimic possible real-world delays in sample freezing, does not have a major impact on the global gene expression profiles. Overall, it is hard to judge whether the Transfer Protocol gene expression profiles share more similarity with samples processed according to the Standard or Reference Protocol. The hierarchical clustering suggested a somewhat higher similarity with Standard Protocol samples, but this was not reflected in the PCA. Comparison of the technical parameters was inconclusive: the higher call concordance suggested greater similarity with Reference Protocol samples (85.2 and 87.5%), whereas the lower false change values (7.1%) were in support of greater similarity with Standard Protocol samples. Given the technical feasibility of processing frozen EDTA samples for microarray analysis, there remains the question whether such microarray experiments may produce meaningful results. It is difficult to give a general answer to this question. First, we do not know the “true” gene expression profile at the moment of blood collection. Therefore, we do not have a clear indication how accurate the gene expression profiles obtained with the Transfer Protocols are as compared to the “true” profiles. It is highly likely that even the Reference Protocol introduces some bias, e.g. due to the RNA stabilizing additive or due to the subsequent RNA processing. Nevertheless, the gene expression profiles obtained with the Reference Protocol provide some important orientation. Without doubt, the “reference” RNA is of high quality, and the microarrays produce gene expression profiles with low interindividual variability and good assay performance characteristics. The transfer of thawed EDTA blood into the PAXgene system clearly increased RNA yield and quality and may therefore provide the means to generate biologically meaningful data from sub-optimal clinical samples. The procedure can, however, obviously not be expected to reverse gene expression changes and variability, or RNA quality loss that may have occurred after blood sampling in the absence of an RNA stabilizing additive. It has been shown by several groups that gene transcription continues
J.M. Beekman et al. / Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
ex vivo and reflects environmental changes such as hypothermia, hypoxia, stress (Debey et al., 2004; Kim et al., 2007; Tanner et al., 2002; Pahl & Brune, 2002) and contact with foreign surfaces (Härtel et al., 2001). When compared to the Reference Protocol group gene expression profiles resulting from the Transfer Protocols demonstrate a larger interindividual variability. Therefore, there is a higher chance to overlook subtle gene expression changes, whereas strong effects may still be detectable. 4.3. RNA integrity numbers and outcome of microarrays High quality RNA is a prerequisite for reliable and robust microarray experiments. It has been demonstrated by many groups that RNA quality has a large influence on gene expression data (Madabusi et al., 2006; Copois et al., 2007). There is, however, little published data addressing objective criteria for RNA quality control. There is a general recommendation to use RNA with RIN scores of 6 or higher, but data to support this cutoff have to our knowledge not been published. We observed an inverse correlation between RIN values and microarray outcome with regard to GAPDH 3′/5′ ratios and scaling factor, whereas lower RIN values were associated with lower percent present calls. A RIN cutoff of 6 is easily met when processing blood samples according to the Reference Protocol. Following the Affymetrix microarray recommendations of GAPDH 3′/5′ b 3 and percent presence call N25%, we find that RNA samples with RIN values of 5 or higher still fulfill the criteria. Such a reduction of the RIN cutoff would allow all samples processed according to the Transfer Protocol to be used for microarray hybridization. 4.4. Summary RNA from whole blood samples originally collected without RNA stabilizing additive and frozen before RNA extraction can be extracted in quantities and qualities that allow for gene expression profiling. In brief, an aliquot of thawed EDTA blood is transferred into a PAXgene blood collection tube followed by RNA preparation using the PAXgene kit. Subsequent microarray experiments largely fulfilled the Affymetrix quality requirements. The Transfer Protocol provides a recovery method for extraction of RNA from highly valuable legacy samples. Any prospective blood collection for RNA use should consider RNA stabilizing additives, which clearly ensure superior results. Acknowledgements The excellent technical assistance of Sandra Patkovic and Karin Bressler is gratefully acknowledged.
49
References Burczynski, M. E., & Dorner, A. J. (2006). Transcriptional profiling of peripheral blood cells in clinical pharmacogenomic studies. Pharmacogenomics, 7, 187−202. Burczynski, M. E., Peterson, R. L., Twine, N. C., Zuberek, K. A., Brodeur, B. J., Casciotti, L., et al. (2006). Molecular classification of Crohn's disease and ulcerative colitis patients using transcriptional profiles in peripheral blood mononuclear cells. Journal of Molecular Diagnostics, 8, 51−61. Burczynski, M. E., Twine, N. C., Dukart, G., Marshall, B., Hidalgo, M., Stadler, W. M., et al. (2005). Transcriptional profiles in peripheral blood mononuclear cells prognostic of clinical outcomes in patients with advanced renal cell carcinoma. Clinical Cancer Research, 11, 1181−1189. Copois, V., Bibeau, F., Bascoul-Mollevi, C., Salvetat, N., Chalbos, P., Bareil, C., et al. (2007). Impact of RNA degradation on gene expression profiles: Assessment of different methods to reliably determine RNA quality. Journal of Biotechnology, 127, 549−559. Debey, S., Schoenbeck, U., Hellmich, M., Gathof, B. S., Pillai, R., Zander, T., et al. (2004). Comparison of different isolation techniques prior gene expression profiling of blood derived cells: Impact on physiological responses on overall expression and the role of different cell types. Pharmacogenomics Journal, 4, 193−207. Härtel, C., Bein, G., Müller-Steinhardt, M., & Klüter, H. (2001). Ex vivo induction of cytokine mRNA expression in human blood samples. Journal of Immunological Methods, 249, 63−71. Kim, S. J., Dix, D. J., Thompson, K. E., Murrell, R. N., Schmid, J. E., Gallagher, J. E., et al. (2007). Effects on storage RNA extraction GeneChip type and donor sex on gene expression profiling of human whole blood. Clinical Chemistry, 53, 1038−1045. Liu, W. M., Mei, R., Di, X., Ryder, T. B., Hubbell, E., Dee, S., et al. (2002). Analysis of high density expression microarrays with signed-rank call algorithms. Bioinformatics, 18 (12), 1593−1599. Madabusi, L. V., Latham, G. J., & Andruss, B. F. (2006). RNA extraction for arrays. Methods in Enzymology, 411, 1−14. McLean, L. A., Gathmann, I., Capdeville, R., Polymeropoulous, M. H., & Dressmann, M. (2004). Pharmacogenomic analysis of cytogenetic response in chronic myeloid leukemia patients treated with imatinib. Clinical Cancer Research, 10, 155−165. Müller, M. C., Merx, K., Weisser, A., Kreil, S., Lahaye, T., Hehlmann, R., et al. (2002). Improvement of molecular monitoring of residual disease in leukemias by bedside RNA stabilization. Leukemia, 16, 2395−2399. Pahl, A., & Brune, K. (2002). Gene expression changes in blood after phlebotomy: Implications for gene expression profiling. Blood, 100, 1094−1095. Rainen, L., Oelmueller, U., Jurgensen, S., Wyrich, R., Ballas, C., Schram, J., et al. (2002). Stabilization of mRNA expression in whole blood samples. Clinical Chemistry, 48, 1883−1890. Schroeder, A., Mueller, O., Stocker, S., Salowsky, R., Leiber, M., Gassmann, M., et al. (2006). RNA integrity number for assigning integrity values to RNA measurements. BMC Molecular Biology, 7, 3−17. Tanner, M. A., Berk, L. S., Felten, D. L., Blidy, A. D., Bit, S. L., & Ruff, D. W. (2002). Substantial changes in gene expression level due to the storage temperature and storage duration of human whole blood. Clinical and Laboratory Haematology, 24, 337−341. Thach, D. C., Lin, B., Walter, E., Kruzelock, R., Rowley, R. K., Tibbetts, C., et al. (2003). Assessment of two methods for handling blood in collection tubes with RNA stabilizing additive for surveillance of gene expression profiles with high density microarrays. Journal of Immunological Methods, 283, 269−279.
Contents lists available at ScienceDirect
Journal of Pharmacological and Toxicological Methods j o u r n a l h o m e p a g e : w w w. e l s e v i e r. c o m / l o c a t e / j p h a r m t o x
Original article
Recovery of microarray-quality RNA from frozen EDTA blood samples Johanna M. Beekman a,⁎, Joachim Reischl a, David Henderson a, David Bauer a, Rainer Ternes a, Carol Peña b, Chetan Lathia b, Jürgen F. Heubach a a b
Clinical Pharmacology, Bayer Schering Pharma AG, 13353 Berlin, Germany Clinical Pharmacology, Bayer HealthCare Pharmaceuticals, Montville, NJ 07045, USA
a r t i c l e
i n f o
Article history: Received 19 August 2008 Accepted 8 October 2008 Keywords: EDTA blood Human Methods Microarray analysis PAXgene Quality control RNA extraction RNA stabilizing additive RIN
a b s t r a c t Introduction: The availability of blood collection systems with RNA stabilizing additives (e.g. PAXgene) has opened the field for gene expression profiling in large multi-center clinical trials. Here, we investigated whether the PAXgene system also offers a method for extraction of RNA from frozen EDTA blood samples, which do not yield RNA of high enough quality for RNA expression profiling, when extracted with standard protocols. Methods: Whole blood was obtained from six healthy volunteers in conventional EDTA tubes and frozen. The thawed EDTA blood was transferred into PAXgene tubes, and the RNA was extracted using the PAXgene RNA extraction kit. Microarray analysis was performed to asses the effect of RNA quality on gene expression profiles. Results: The RNA yield of the transferred samples was 1.76 ± 0.88 µg/ml. This yield was clearly lower than the yield from a PAXgene reference group (2.84 ± 0.62 µg/ml), but considerably higher than the yield resulting from a standard protocol usually applied to fresh EDTA blood samples (0.07 ± 0.06 µg/ml). The RNA integrity number (RIN) of the transferred samples was 6.1 ± 0.8 as compared to 9.8 ± 0.1 for the PAXgene reference. Microarray analysis of the extracted RNA suggested that samples with RIN values above 5 produce data that fulfill the quality criteria defined by the manufacturer. Discussion: The transfer of thawed EDTA blood into PAXgene blood collection tubes offers a method to recover sufficient RNA of acceptable quality for microarray experiments. © 2008 Elsevier Inc. All rights reserved.
1. Introduction The use of gene expression analysis for biomarker discovery is widely practiced. Gene expression analysis of blood cells as an easily accessible surrogate tissue allows for minimally-invasive repeated measurements, and has been used in clinical studies for prospectively planned biomarker discovery. Blood cell RNA profiling has been used to identify candidate biomarkers for disease classification (Müller et al., 2002; Burczynski et al., 2006), prediction of prognosis (Burczynski et al., 2005) and drug responder/non-responder identification (McLean et al., 2004; for a review see also Burczynski & Dorner, 2006). The introduction of blood collection systems with RNA stabilizing additives and accompanying RNA extraction kits has significantly improved the quality of RNA isolated from blood samples collected in multicenter clinical trials (Rainen et al., 2002; Thach et al., 2003), and the convenience of processing the samples. Using such stabilization systems the RNA does not need to be extracted immediately; instead, blood samples can be collected, frozen and stored at the clinical site before batched dry ice shipment to a central
⁎ Corresponding author. Clinical Biomarkers, Clinical Pharmacology, Bayer Schering Pharma AG, Müllerstr. 178, D-13353 Berlin, Germany. Tel.: +49 30 468 12554; fax: +49 30 468 92554. E-mail address: [email protected] (J.M. Beekman). 1056-8719/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.vascn.2008.10.003
RNA extraction and analysis center, additionally reducing interoperator variability in RNA quality. Despite the availability of RNA stabilizing blood systems, there may be a need to perform RNA profiling experiments with legacy blood samples that were collected with common blood collection systems, such as EDTA tubes, and frozen after a variable time at ambient temperature. The main challenge of RNA extraction from frozen EDTA blood samples is to obtain sufficient RNA of acceptable quality for microarray hybridization. Whereas RNA stabilizing additives cause controlled cell lysis in an environment of RNA protection, the freezing of EDTA blood destroys a large fraction of blood cells, exposing the RNA to released enzymes and subsequent RNA degradation. A second challenge in using frozen EDTA blood as a starting material for RNA extraction is that standard protocols for RNA extraction from fresh blood include a hypotonic erythrocyte lysis step before white blood cells are harvested for RNA extraction. This step minimizes the adhesion of RNA to hemoglobin and cell debris but cannot be applied to thawed EDTA blood, since the cells are already destroyed by freezing and thawing. The RNA in these samples is trapped in the cell debris and is difficult to isolate. Here, we investigated the possibility of rescuing hybridizationquality RNA from frozen EDTA blood samples. We compared a standard RNA isolation protocol commonly used on fresh EDTA blood samples to a method where thawed EDTA blood was transferred
J.M. Beekman et al. / Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
45
Fig. 1. Flow chart of the four experimental protocols evaluated for RNA isolation from blood cells. Whole blood samples were collected from 6 healthy volunteers in blood collection tubes containing either Potassium EDTA or a RNA-stabilizing additive (PAXgene). RT: room temperature.
into blood collection tubes with RNA stabilizing additive (PAXgene™ blood RNA system) followed by RNA extraction with the respective RNA extraction kit. A state-of-the-art reference group of blood samples was collected directly into PAXgene tubes. We demonstrate that the transfer of thawed EDTA blood into PAXgene tubes and subsequent RNA extraction allows the isolation of RNA of sufficient yield and acceptable quality for microarray analysis. 2. Methods 2.1. Sample collection and storage The research was carried out in accordance with the principles of the current version of the Helsinki Declaration. After approval of the Institutional Review Board, blood from six healthy volunteers (aged 35 to 55 years) was collected into Potassium EDTA tubes (2.7 ml, Sarstedt) and PAXgene™ blood RNA tubes (Becton-Dickinson) according to the manufacturer's directions. Volunteers provided written informedconsent. The filled EDTA tubes were either put immediately into a −20 °C freezer or left at room temperature for 3 h prior to freezing. As a modification to the procedures suggested by the manufacturer, the blood collected in PAXgene tubes was placed immediately into the −20 °C freezer. 2.2. RNA isolation and quality assessment Frozen EDTA tubes were thawed on ice during approximately 2 h. RNA was then extracted using the QIAamp kit (Qiagen) or by transferring 2.5 ml of EDTA blood into a PAXgene tube followed by 16–21 h incubation at room temperature and RNA extraction as described below for the PAXgene samples. The RNA was eluted in two steps utilizing 40 µl buffer for each step. Frozen PAXgene tubes were thawed and incubated for 16–21 h at room temperature followed by RNA isolation using the PAXgene kit including QIAshredder homogenization and on-column DNase digestion as described in the manufacturer's handbook (Qiagen). RNA concentration was deter-
mined using a NanoDrop spectrophotometer (NanoDrop Technologies), and RNA integrity was assessed by determining the RIN (Schroeder et al., 2006) on a 2100 Bioanalyzer (Agilent technologies). 2.3. RNA expression analysis Total RNA, isolated from thawed EDTA blood using the QIAamp kit, was labeled according to the Affymetrix GeneChip Expression Analysis Technical Manual (Affymetrix). To obtain enough RNA for labeling, 2 RNA samples from each donor (2 extractions of 2.7 ml blood each) had to be pooled. The RNA amount of the pooled samples ranged from 23 to 751 ng. Total RNA isolated using the PAXgene system was labeled accordingly, however using BLOCK reagent (GeneLogic) in the firststrand cDNA synthesis reaction. BLOCK reagent is a mixture of hemoglobin-specific oligonucleotides and a T7-Oligo(dT) promoter primer. It is used to increase overall assay sensitivity by blocking cDNA synthesis from hemoglobin transcripts, which originate from the reticulocytes in the whole blood samples. The labeled cRNA was quantified using a NanoDrop spectrophotometer and checked for quality on the Agilent Bioanalyzer. Resulting cRNA was hybridized to Affymetrix HG U133 Plus 2.0 GeneChip microarrays (containing approximately 54,000 probe sets representing 47,400 transcripts and variants; Affymetrix), and the arrays were washed and stained using the GeneChip Fluidics Station 450 (Affymetrix). The stained microarrays were scanned at 570 nm using a confocal laser scanner (GeneChip-3000 Scanner, Affymetrix), and the resulting ⁎.DAT-files were quantified using GCOS software (Affymetrix). The resulting ⁎.CEL-files were condensed using the MAS 5.0 algorithm (Affymetrix). Selected quality criteria were extracted from the MAS 5.0 reports. Analysis of expression data was performed by using the Expressionist Analyst Pro Version 4.5 software (GeneData AG). In the first analysis step a probe set expression value was considered valid if it reached a p-value below 0.04 in the one-sided Wilcoxon signed rank test of the MAS 5.0 condensing algorithm (Liu et al., 2002). The overall data structure was investigated using hierarchical clustering.
46
J.M. Beekman et al. / Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
3. Results In the present study we compared the yield and quality as well as the suitability for microarray analysis of RNA obtained from blood cells according to four different experimental protocols, summarized in Fig. 1. Three protocols utilized frozen EDTA blood to simulate the condition of legacy samples and to develop an efficient method to recover RNA from this inappropriate starting material. EDTA blood frozen immediately, thawed, and extracted using the QIAamp RNA extraction kit was defined as the Standard Protocol. Two investigational protocols where EDTA blood was frozen, thawed, transferred to PAXgene tubes, and extracted using the PAXgene RNA isolation kit are referred to as Transfer Protocol A (samples frozen immediately) and B (samples frozen after 3 h at room temperature). The room temperature incubation in Transfer Protocol B was designed to mimic possible delay until freezer storage of samples at clinical sites. In a fourth protocol, referred to as the Reference Protocol, blood was collected into containers prepared with RNA stabilizing additive (PAXgene tubes), frozen immediately, and extracted using the PAXgene RNA isolation kit. This Reference Protocol is expected to produce high yield, high quality RNA. 3.1. RNA extraction: yield and quality The Standard Protocol resulted in poor RNA yields as compared to the Reference Protocol (0.07 ± 0.06 µg/ml versus 2.84 ± 0.62 µg/ml;
Fig. 3. Comparison of microarray quality parameters for blood samples processed using the four different experimental protocols. (A) GAPDH 3′/5′, (B) scaling factor, and (C) percent presence calls. Results from individual microarrays are shown (n = 6 per protocol).
p = 5.7 ⁎ 10− 7; T-test; Fig. 2A). Transfer Protocols A and B produced intermediate RNA amounts with a tendency towards lower yields after 3 h of room temperature storage prior to freezing (1.76 ± 0.88 µg/ml versus 1.29 ± 0.45 µg/ml; p = 0.27; T-test; Fig. 2A). The RNA integrity number (RIN), determined using the algorithm provided in the Agilent Bioanalyzer 2100 expert software provides an objective assessment of RNA integrity using the entire electrophoretic trace of the RNA sample instead of using the ratio of the 18S and 28S ribosomal RNAs alone. The RIN values for RNA resulting from the two Transfer Protocols were close to 6 (6.1 ± 0.8 and 6.4 ± 0.8) and lower than those from the Reference Protocol (9.8 ± 0.1; Fig. 2B). Due to the low concentration of RNA obtained from the Standard Protocol, quality assessment was not possible for this group. 3.2. GeneChip analysis
Fig. 2. Yield (A) and quality (B) of RNA extracted from whole blood samples using the four different protocols. The graphs show mean values ± standard deviations from n = 6 samples. RIN: RNA integrity number. N.D., not done.
For each of the four protocols, RNA samples from six donors were processed to cRNA, and the cRNA yield and distribution were analyzed. The cRNA size distribution resulting from the Reference Protocol showed a maximum around 1500 bases (data not shown). The three protocols utilizing frozen EDTA blood resulted in cRNA size distributions between 250 and 1000 bases, indicating lower RNA integrities. A trend towards longer cRNA was seen using the Transfer Protocols (as
J.M. Beekman et al. / Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
compared to the Standard Protocol), but this was not quantifiable. None of the cRNA profiles showed a hemoglobin peak, confirming a successful erythrocyte lysis in the standard protocol and efficient use of the BLOCK reagent for the transfer and reference protocol samples. To obtain enough RNA from the Standard Protocol, two RNA samples from each donor, i.e. two extractions of 2.7 ml blood each, had to be pooled. Despite pooling the minimum amount of 10 µg cRNA necessary for GeneChip hybridization was not achieved for two individuals. In these cases the total amount of resulting cRNA was used for hybridization (7.3 µg for Stan5 and 5.4 µg for Stan6; see Fig. 4A for reference). After hybridization, washing, staining and scanning, the gene expression data was processed using MAS 5.0, and a data quality check was performed. The noise and background levels of all GeneChips were within quality limits defined by the microarray manufacturer (data not shown). The GAPDH 3′/5′ ratio, the scaling factors, and the percent present calls, however, demonstrated clear differences among the four experimental protocols. The GAPDH 3′/5′ ratio, which is a quality measure of the cRNA applied to the GeneChip (reflecting RNA degradation, with larger numbers indicating increased degradation), increased in the order Reference b Transfer A/B b Standard Protocol, indicating decreasing RNA integrity (Fig. 3A). All samples processed according to the Standard Protocol showed GAPDH ratios higher than the quality limit of 3 recommended by Affymetrix, whereas no samples from the Reference Protocol and Transfer Protocol A and only one sample from Transfer Protocol B exceeded this limit. In addition, the Standard Protocol produced the GeneChips with the lowest overall signal, reflected by the high scaling factors necessary for normalization (Fig. 3B). Application of the Transfer Protocols to frozen EDTA blood resulted in a marked reduction of scaling factors. Lastly, the Transfer Protocols clearly increased the percent present calls as compared to the Standard Protocol, allowing the detection and quantification of a
47
considerably higher number of transcripts (Fig. 3C). For all three quality parameters, blood processing using the Reference Protocol achieved the best assay performance with low interindividual variability. Hierarchical clustering and principal component analysis of the microarray data are shown in Fig. 4. Three clusters are obvious (Fig. 4A): The topmost cluster with little interindividual variability contains the six samples processed according to the Reference Protocol. The cluster in the middle is composed of samples processed using the Transfer Protocols, where A and B samples from the same donor clustered in pairs. In other words, little separation is seen between the immediately frozen samples and those which were frozen after 3 h at room temperature; the interindividual differences of gene expression profiles exceeded the variability introduced by the 3 h incubation at room temperature. The cluster at the bottom encompasses the samples processed using the Standard Protocol. The principal component analysis shows the same cluster formation and the same variability within the clusters (Fig. 4B). Additional technical parameters such as call concordance and false change may offer a more quantitative comparison of the outcome of different protocols applied. Call concordance is defined as the average percentage of probe sets giving concordant Present and concordant Absent calls between pair-wise comparisons (intrasubject comparisons). In comparison to the Reference Protocol, the mean call concordances for the Standard Protocol and for Transfer Protocols A and B were 77.1 ± 5.5%, 87.5 ± 3.6% and 85.2 ± 3.2%, respectively. When compared to the Standard Protocol, the mean call concordances for the Transfer Protocols A and B were 81.9 ± 5.5% and 83.7 ± 4.3% respectively. The two Transfer Protocols A and B showed a mean call concordance of 90.2 ± 1.0%, thereby even fulfilling the Affymetrix requirement for technical replicates with a call concordance above 90%, when the same hybridization cocktail is applied on two different chips.
Fig. 4. (A) Hierarchical clustering and (B) principal component analysis of all microarray data obtained from the Reference Protocol (Ref), Transfer Protocol A (TransA), Transfer Protocol B (TransB) and Standard Protocol (Stan). Numbers specify the different blood donors.
48
J.M. Beekman et al. / Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
A false change is defined as a two-fold or greater change in a comparison analysis between replicate samples. For technical replicates, a false change of 1% or less is accepted, according to Affymetrix standards. In comparison to the Reference Protocol, the mean false change rates for the Standard Protocol and for Transfer Protocols A and B were 8.8 ± 1.1%, 9.9 ± 1.8% and 11.0 ± 1.4%, respectively. When compared to the Standard Protocol, the mean false changes for the Transfer Protocols A and B were both 7.1 ± 0.9% and 7.1 ± 1.0%, respectively. The two Transfer Protocols A and B themselves showed a mean false change of only 1.3 ± 0.3%, which again almost fulfilled the requirement for technical replicates. 3.3. Correlation between RNA quality and GeneChip performance Transfer Protocol A was successfully applied to extract RNA from a set of legacy EDTA blood samples collected at multiple sites and thus expected to have high variability in quality due to differing sample processing, storage conditions, and storage times. Four samples were selected for GeneChip analysis from each of the following RIN classes: RIN 2–3, RIN 3–4, RIN 4–5, RIN 5–6, RIN 6–7 and RIN N 7. Again, upon hybridization background and noise levels were not affected by the RNA quality (data not shown). Analysis of the 24 GeneChips, however, indicated a clear negative correlation between the RIN score and the GAPDH 3′/5′ ratio (Fig. 5A). All samples with RIN values above 5 resulted in GAPDH 3′/5′ ratios below 3. In addition, there was a negative correlation between the RIN and the scaling factor (Fig. 5B),
Fig. 5. Correlation between RNA integrity number (RIN) and (A) GAPDH 3′/5′ ratio, (B) scaling factor, and (C) percent present call.
as well as a positive correlation between RIN values and the percent present call (Fig. 5C). 4. Discussion 4.1. The Transfer Protocol for RNA recovery The purpose of this work was to develop a method for RNA extraction from frozen EDTA blood samples, providing sufficient yield and acceptable quality for gene expression analysis. As many groups have shown, the PAXgene method utilized as the Reference Protocol yielded superior RNA quantity and quality (Thach et al., 2003; Debey et al., 2004; Kim et al., 2007). We do not propose the Transfer Protocol as an approach for prospective blood collection with the intention of performing expression analysis, but as a way to extract RNA from legacy blood samples collected using EDTA tubes, provided that the patient informed consent allows for RNA analyses. Our recovery scenario was based on EDTA blood but the same approach may be feasible for other types of blood samples, for example citrate or heparin preparations. 4.2. RNAs from the four protocols show different gene expression profiles The two experimental Transfer Protocols are hybrid methods utilizing EDTA tubes for blood collection (as in the Standard Protocol) and PAXgene processing for RNA isolation (as in the Reference Protocol). The two procedures gave similar RNA yields of comparable quality. Hierarchical clustering demonstrated that the gene expression profiles of samples processed using the Transfer Protocols clustered separately from the profiles of the two other protocols. Within this group, samples from the same donor stored in the freezer either immediately or after 3 h at room temperature clustered in pairs. This indicates that a 3 h room temperature incubation, intended to mimic possible real-world delays in sample freezing, does not have a major impact on the global gene expression profiles. Overall, it is hard to judge whether the Transfer Protocol gene expression profiles share more similarity with samples processed according to the Standard or Reference Protocol. The hierarchical clustering suggested a somewhat higher similarity with Standard Protocol samples, but this was not reflected in the PCA. Comparison of the technical parameters was inconclusive: the higher call concordance suggested greater similarity with Reference Protocol samples (85.2 and 87.5%), whereas the lower false change values (7.1%) were in support of greater similarity with Standard Protocol samples. Given the technical feasibility of processing frozen EDTA samples for microarray analysis, there remains the question whether such microarray experiments may produce meaningful results. It is difficult to give a general answer to this question. First, we do not know the “true” gene expression profile at the moment of blood collection. Therefore, we do not have a clear indication how accurate the gene expression profiles obtained with the Transfer Protocols are as compared to the “true” profiles. It is highly likely that even the Reference Protocol introduces some bias, e.g. due to the RNA stabilizing additive or due to the subsequent RNA processing. Nevertheless, the gene expression profiles obtained with the Reference Protocol provide some important orientation. Without doubt, the “reference” RNA is of high quality, and the microarrays produce gene expression profiles with low interindividual variability and good assay performance characteristics. The transfer of thawed EDTA blood into the PAXgene system clearly increased RNA yield and quality and may therefore provide the means to generate biologically meaningful data from sub-optimal clinical samples. The procedure can, however, obviously not be expected to reverse gene expression changes and variability, or RNA quality loss that may have occurred after blood sampling in the absence of an RNA stabilizing additive. It has been shown by several groups that gene transcription continues
J.M. Beekman et al. / Journal of Pharmacological and Toxicological Methods 59 (2009) 44–49
ex vivo and reflects environmental changes such as hypothermia, hypoxia, stress (Debey et al., 2004; Kim et al., 2007; Tanner et al., 2002; Pahl & Brune, 2002) and contact with foreign surfaces (Härtel et al., 2001). When compared to the Reference Protocol group gene expression profiles resulting from the Transfer Protocols demonstrate a larger interindividual variability. Therefore, there is a higher chance to overlook subtle gene expression changes, whereas strong effects may still be detectable. 4.3. RNA integrity numbers and outcome of microarrays High quality RNA is a prerequisite for reliable and robust microarray experiments. It has been demonstrated by many groups that RNA quality has a large influence on gene expression data (Madabusi et al., 2006; Copois et al., 2007). There is, however, little published data addressing objective criteria for RNA quality control. There is a general recommendation to use RNA with RIN scores of 6 or higher, but data to support this cutoff have to our knowledge not been published. We observed an inverse correlation between RIN values and microarray outcome with regard to GAPDH 3′/5′ ratios and scaling factor, whereas lower RIN values were associated with lower percent present calls. A RIN cutoff of 6 is easily met when processing blood samples according to the Reference Protocol. Following the Affymetrix microarray recommendations of GAPDH 3′/5′ b 3 and percent presence call N25%, we find that RNA samples with RIN values of 5 or higher still fulfill the criteria. Such a reduction of the RIN cutoff would allow all samples processed according to the Transfer Protocol to be used for microarray hybridization. 4.4. Summary RNA from whole blood samples originally collected without RNA stabilizing additive and frozen before RNA extraction can be extracted in quantities and qualities that allow for gene expression profiling. In brief, an aliquot of thawed EDTA blood is transferred into a PAXgene blood collection tube followed by RNA preparation using the PAXgene kit. Subsequent microarray experiments largely fulfilled the Affymetrix quality requirements. The Transfer Protocol provides a recovery method for extraction of RNA from highly valuable legacy samples. Any prospective blood collection for RNA use should consider RNA stabilizing additives, which clearly ensure superior results. Acknowledgements The excellent technical assistance of Sandra Patkovic and Karin Bressler is gratefully acknowledged.
49
References Burczynski, M. E., & Dorner, A. J. (2006). Transcriptional profiling of peripheral blood cells in clinical pharmacogenomic studies. Pharmacogenomics, 7, 187−202. Burczynski, M. E., Peterson, R. L., Twine, N. C., Zuberek, K. A., Brodeur, B. J., Casciotti, L., et al. (2006). Molecular classification of Crohn's disease and ulcerative colitis patients using transcriptional profiles in peripheral blood mononuclear cells. Journal of Molecular Diagnostics, 8, 51−61. Burczynski, M. E., Twine, N. C., Dukart, G., Marshall, B., Hidalgo, M., Stadler, W. M., et al. (2005). Transcriptional profiles in peripheral blood mononuclear cells prognostic of clinical outcomes in patients with advanced renal cell carcinoma. Clinical Cancer Research, 11, 1181−1189. Copois, V., Bibeau, F., Bascoul-Mollevi, C., Salvetat, N., Chalbos, P., Bareil, C., et al. (2007). Impact of RNA degradation on gene expression profiles: Assessment of different methods to reliably determine RNA quality. Journal of Biotechnology, 127, 549−559. Debey, S., Schoenbeck, U., Hellmich, M., Gathof, B. S., Pillai, R., Zander, T., et al. (2004). Comparison of different isolation techniques prior gene expression profiling of blood derived cells: Impact on physiological responses on overall expression and the role of different cell types. Pharmacogenomics Journal, 4, 193−207. Härtel, C., Bein, G., Müller-Steinhardt, M., & Klüter, H. (2001). Ex vivo induction of cytokine mRNA expression in human blood samples. Journal of Immunological Methods, 249, 63−71. Kim, S. J., Dix, D. J., Thompson, K. E., Murrell, R. N., Schmid, J. E., Gallagher, J. E., et al. (2007). Effects on storage RNA extraction GeneChip type and donor sex on gene expression profiling of human whole blood. Clinical Chemistry, 53, 1038−1045. Liu, W. M., Mei, R., Di, X., Ryder, T. B., Hubbell, E., Dee, S., et al. (2002). Analysis of high density expression microarrays with signed-rank call algorithms. Bioinformatics, 18 (12), 1593−1599. Madabusi, L. V., Latham, G. J., & Andruss, B. F. (2006). RNA extraction for arrays. Methods in Enzymology, 411, 1−14. McLean, L. A., Gathmann, I., Capdeville, R., Polymeropoulous, M. H., & Dressmann, M. (2004). Pharmacogenomic analysis of cytogenetic response in chronic myeloid leukemia patients treated with imatinib. Clinical Cancer Research, 10, 155−165. Müller, M. C., Merx, K., Weisser, A., Kreil, S., Lahaye, T., Hehlmann, R., et al. (2002). Improvement of molecular monitoring of residual disease in leukemias by bedside RNA stabilization. Leukemia, 16, 2395−2399. Pahl, A., & Brune, K. (2002). Gene expression changes in blood after phlebotomy: Implications for gene expression profiling. Blood, 100, 1094−1095. Rainen, L., Oelmueller, U., Jurgensen, S., Wyrich, R., Ballas, C., Schram, J., et al. (2002). Stabilization of mRNA expression in whole blood samples. Clinical Chemistry, 48, 1883−1890. Schroeder, A., Mueller, O., Stocker, S., Salowsky, R., Leiber, M., Gassmann, M., et al. (2006). RNA integrity number for assigning integrity values to RNA measurements. BMC Molecular Biology, 7, 3−17. Tanner, M. A., Berk, L. S., Felten, D. L., Blidy, A. D., Bit, S. L., & Ruff, D. W. (2002). Substantial changes in gene expression level due to the storage temperature and storage duration of human whole blood. Clinical and Laboratory Haematology, 24, 337−341. Thach, D. C., Lin, B., Walter, E., Kruzelock, R., Rowley, R. K., Tibbetts, C., et al. (2003). Assessment of two methods for handling blood in collection tubes with RNA stabilizing additive for surveillance of gene expression profiles with high density microarrays. Journal of Immunological Methods, 283, 269−279.