- Email: [email protected]
Hybridization buffer systems impact the quality of filter array data
Journal of Pharmacological and Toxicological Methods 50 (2004) 67 – 71 www.elsevier.com/locate/jpharmtox
Brief communication
Hybridization buffer systems impact the quality of filter array data Jie Liua,*, Nigel Walkerb, Michael P. Waalkesa a
Inorganic Carcinogenesis Section, LCC, NCI at NIEHS, 111 Alexander Drive, RTP, NC 27709, USA b Laboratory of Comparative Biology & Risk Analysis, NIEHS, USA Received 19 December 2003; accepted 3 February 2004
Abstract Introduction: cDNA microarray technology has greatly facilitated mechanistic studies in pharmacology and toxicology. A clean hybridization with minimal background is critical for successful microarray analysis and is highly desired. However, clean hybridization alone is not enough; verification is needed. Methods: Total RNA was isolated from the livers of acetaminophen-intoxicated mice and was subjected to cDNA microarray analyses using ExpressHyb, ULTRArrayHyb or MicroHyb on nylon membranes. Real-time RT-PCR analyses were performed for verification. Results: We have demonstrated in this paper that hybridization systems can significantly impact the quality of array data. MicroHyb produced very clean hybridizations, but some results could not be confirmed by real-time RT-PCR and in accord with biological responses. The hybridization images from ExpressHyb were not always clean, but were reliable. The sensitivity of ULTRArrayHyb was moderate. Conclusion: This study has indicated the importance of selecting hybridization buffers in membrane arrays and recommended real-time RT-PCR for follow-up analysis. Gene expression changes should also be correlated with biological significance. D 2004 Elsevier Inc. All rights reserved. Keywords: Hybridization buffer systems; Filter array data; Microarray technology
1. Introduction
2. Method
The cDNA microarray has been a huge technical advance in biological research because it allows the simultaneous measurement of hundreds to thousands of genes. Two main types of microarray, the chip-array (glass) and filter-array (nylon membrane), are currently used for gene expression analysis. A clean hybridization with minimal background is critical for successful microarray analyses, and the need for the optimization of hybridization systems is a major part of experimental protocols, which leads to the development of various hybridization systems. Microarray hybridizations can be influenced by many factors. Here, we compared the three commercially available hybridization systems in membrane arrays and found that different hybridization systems could make a substantial difference in the quality of data from membrane arrays, using gene expression analysis of acetaminophen hepatotoxicity as an example.
2.1. Animals and treatment
* Corresponding author. Tel.: +1-919-541-3951. E-mail address: [email protected] (J. Liu). 1056-8719/$ – see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.vascn.2004.02.001
Adult male CD-1 mice were obtained from Charles River Laboratories (Wilmington, MA) and were housed in facilities accredited by the American Association for the Accreditation of Laboratory Animal Care at the National Institute of Environmental Health Sciences, at 20 –22 jC with a 12-h light/dark cycle. The animals were given free access to rodent chow and tap water. The mice were given a hepatotoxic dose of acetaminophen (APAP, 600 mg/kg ip for 8 h), and the livers were removed for array analysis as described previously (Liu et al., 2003). 2.2. Microarray analysis Total RNA was isolated with Trizol reagent (Invitrogen, Carlsbad, CA), followed by purification with RNeasy columns (Qiagen, Valencia, CA). Approximately 5 mg of total RNA was converted to [a-32P]-dATP-labelled cDNA probe using Moloney murine leukemia virus (MuLV) reverse transcriptase and the cDNA synthesis primer mix (Clontech, Palo Alto, CA) and was purified with NucleoSpin columns.
68
J. Liu et al. / Journal of Pharmacological and Toxicological Methods 50 (2004) 67–71
Table 1 Comparisons among three hybridization systems
Company Prehybridization Hybridization 1st Wash protocol 2nd Wash protocol
ExpressHyb
ULTRArrayHyb
MicroHyb
Clontech lot#0250B80 5 ml ExpressHyb, 50 Ag sperm DNA, 68 jC, 2 h Add 100 Al probe Cot-1 DNA 5 Ag, 68 jC, 16 h 2 SSC 1% SDS, 68 jC, 30 min 4 0.1 SSC 0.5% SDS, 68 jC, 30 min 2
Ambion lot#061P71A 10 ml ULTRArrayHyb, 50 Ag sperm DNA, 50 jC, 2 h Add 100 Al probe Cot-1 DNA 5 Ag, 50 jC, 16 h 2 SSC 0.5% SDS, 50 jC, 30 min 2
Research Genetics lot#R26478 5 ml MicroHyb, 50 Ag sperm DNA, 42 jC, 2 h Add 100 Al probe Cot-1 DNA 5 Ag, 42 jC, 16 h 2 SSC 1% SDS, 42 jC, 30 min 2
0.5 SSC 0.5% SDS, 50 jC, 30 min 2
0.5 SSC 1% SDS, 25 jC, 30 min 1
Three hybridization buffer systems, ExpressHyb from Clontech (lot#0250B80), ULTRArrayHyb from Ambion (Austin, TX; lot#061P71A), and MicroHyb from Research Genetics (Huntsville, AL, lot#R26478), were used according to conditions recommended by the manufactures (Table 1). The hybridizations were performed for 16 h using the same probe (6 106 cpm in 100 Al) and the same lot of the custom-made nylon membranes. The membranes were then washed according to the protocols for each hybridization system and then wrapped and exposed to a Molecular Dynamics Phosphoimage Screen (Sunnyvale, CA). The images were analyzed densitometrically using AtlasImage software (Clontech). The gene expression intensities were first corrected with the external background and then globally normalized.
as relative increases setting control as 100%. Assuming that the CT value is reflective of the initial starting copy and that there is 100% efficacy, a difference of one cycle is equivalent to a twofold difference in starting copy. 2.4. Statistics Means and S.E.M. (n = 4) were calculated for the Students’ t test. The level of significance was set at P < .05. The expression ratios (acetaminophen/control) were used for comparisons.
3. Results 3.1. Microarray images
2.3. Real-time RT-PCR analysis Expressions of the selected genes were quantified using real-time RT-PCR analysis (Walker, 2001). Briefly, total RNA was reverse transcribed with MuLV reverse transcriptase and oligo-dT primers. The forward and reverse primers for selected genes were designed using a Primer Express software (Applied Biosystems, Foster City, CA) and are listed in Table 2. The SYBR green DNA PCR kit (Applied Biosystems) was used for real-time PCR analysis. The relative differences in expression between groups were expressed using cycle time (CT) values, and the relative differences between groups were calculated and expressed
The representative hybridization images are shown in Fig. 1. Note the following: (1) The images were clean, but the patterns of the gene expression are quite different between the three hybridization systems. (2) The acetaminophen-induced expressions of genes encoding the acute phase proteins, such as c-Jun, heme oxygenase-1 (HO-1), early response growth factor-1 (ERG1), heat shock protein 70 (HSP70), the increased expression of DNA damage protein GADD45, and the decreased expression of CYP2E1, are readily seen with ExpressHyb, somewhat visible with ULTRArrayHyb, but are not evident with the MicroHyb system.
Table 2 Primer sequences used for quantitative real-time RT-PCR analysis Gene
Accession number
Forward primer
Reverse primer
c-Jun/AP-1 c-myc HO-1 EGR1 HSP70 GADD45 GADD153 CYP2E1 G3PDH h-actin
J04115 X00195 M33203 M20157 M35021 L28177 X67083 L11650 M32599 M12481
ACTCCGAGCTGGCATCCA CGCCGCTGGGAAACTTT CCTCACTGGCAGGAAATCATC AGGTTCCCATGATCCCTGACT GGACAAGCAAGCATTCCACA CAGATCCATTTCACCCTCATCC CTCCTGTCTGTCTCTCCGGAA CACAGCCAAGAACCCATGTACA AGTATGACTCCACTCACGGCAAAT GTATGACTCCACTCACGGCAAA
CCCACTGTTAACGTGGTTCATG TCCTGGCTCGCAGATTGTAA CCTCGTGGAGACGCTTTACATA GGTACGGTTCTCCAGACCCTG CTCAAGAGAGGCGGGATAAGG TCCAGTAGCAGCAGCTCAGC TACCCTCAGTCCCCTCCTCA CAGGAGCCCATATCTCAGAGTTG GTCTCGCTCCTGGAAGATGGT GGTCTCGCTCCTGGAAGATG
EGR1, early growth response protein-1; HSP70, heat shock protein (70 kD); GADD45 and GADD153, growth arrest and DNA damage inducible protein45 and 153; G3PDH, glyceraldehydes-3-phosphate dehydrogenase; HO-1, heme oxygenase-1.
J. Liu et al. / Journal of Pharmacological and Toxicological Methods 50 (2004) 67–71 Fig. 1. Representative phosphorimage of microarray results. Top, control normal liver samples; Bottom, acetaminophen-intoxicated liver samples. Arrows show acetaminophen-induced increases in expression of growth arrest and DNA damage inducible protein GADD45, HSP70, early growth response protein-1 (EGR1), c-jun and heme oxygenase-1 (HO-1), as well as the decrease in expression of cytochrome P450 2E1 (CYP2E1).
69
70
J. Liu et al. / Journal of Pharmacological and Toxicological Methods 50 (2004) 67–71
Table 3 Comparisons in acetaminophen-induced gene expression fold changes between real-time RT-PCR and hydridiztion systems Gene
Real-time RT-PCR analysis
ExpressHyb
ULTRArrayHyb
MicroHyb
c-Jun/AP-1 c-myc HO-1 EGR1 HSP70 GADD45 GADD153 CYP2E1 G3PDH h-actin
12.5 11.8 11.5 35.5 3.12 16.3 11.5 0.20 1.01 1.06
12.9 10.4 8.6 21.2 2.52 7.27 6.21 0.48 1.24 1.21
1.98 3.77 2.94 2.95 1.97 1.68 1.84 0.51 1.55 1.42
0.98 0.79 1.12 1.02 0.71 0.61 0.48 0.72 1.47 1.37
The mice were given acetaminophen (600 mg/kg ip for 8 h), and hepatic gene expression was assessed by real-time RT-PCR and microarray with three hybridization systems. The fold changes are expressed as mean ratios of acetaminophen/controls.
3.2. Real-time RT-PCR confirmation Expression of the selected genes was additionally quantified using real-time RT-PCR analysis. A comparison of gene expression changes by microarrays and real-time RT-PCR analyses is shown on Table 3. Using ExpressHyb system, the marked increases in acute phase proteins, such as c-Jun ( 12), c-myc ( 10), HO-1 ( 9), early response protein1 (EGR1, 21), and heat shock protein-70 (HSP70, 2.5), the genes encoding DNA damage and cell death, such as GADD45 ( 10) and GADD153 ( 6.2), and the suppression of constitutive metabolic enzymes, such as CYP2E1 ( 0.48), were well correlated with real-time RT-PCR analysis. The ULTRArrayHyb system also produced similar trend, but to a less extent, while MicroHyb system did not match RT-PCR results. For example, the expression of c-Jun was 12.5, 12.9, 1.98, and 0.98 for real-time RTPCR, ExpressHyb, ULTRArrayHyb, and MicroHyb, respectively. The expressions of the housekeeping genes G3PDH and h-actin were about the same across these hybridization buffers.
4. Discussion This study provides a lesson we have learned from our own mistakes. In an attempt to get a clear hybridization image, we tested five different hybridization systems and found that MicroHyb system was the best and always gave very clean hybridization images, while the ExpressHyb did not always produce clean images. We have thus switched to MicroHyb system for over 150 arrays in five different projects. The hybridizations of every membrane with MicroHyb were very clean and highly reproducible. However, the obtained gene expression profiles did not match biological responses, and we were unable to make a coherent interpretation of these 150 arrays. This frustrating experience pro-
moted us to compare hybridization systems using the same probes. The study demonstrated that the choice of hybridization systems could make a dramatic difference in data obtained from membrane arrays. To our surprise, gene expression patterns are quite different among the three hybridization systems (Fig. 1). Some visible spots on the membrane with ExpressHyb or UltrarrayHyb are invisible when MicroHyb is used, and vice versa. For example, there is a 10-fold difference in the basal expression level of CYP2E1. To define the reason(s) for such a huge discrepancy is beyond our capability, as the recipes for these hybridization buffers are confidential. The only thing we could do was to vary the temperature of hybridization and washing conditions, but the differences still remained (data not shown). Thus, only the standard procedures recommended by the manufactures were used for comparison in this paper. All three hybridization systems produced clean, reproducible hybridization images, but the gene expression patterns were quite different, meaning that it was not possible to ascertain which results represented reality. To address this question, real-time RT-PCR analysis was performed on selected genes. As shown in Table 3, there is an excellent match between real-time RT-PCR and ExpressHyb, a good match between real-time RT-PCR and ULTRAarrayHyb, and a poor match between real-time RT-PCR and MicroHyb. The real-time RT-PCR analysis supports the results obtained from ExpressHyb, although its hybridization image is not always clean. It is very important to correlate gene expression changes with biological function. We thus compared the three hybridization systems in two separate studies on acute acetaminophen hepatotoxicity (Liu et al., 2003; Shankar et al., 2003). Both studies gave similar, consistent gene expression patterns in response to acetaminophen-induced liver injury. For example, a 12-fold induction of c-Jun was seen with ExpressHyb and real-time RT-PCR, while it was moderately increased (2-fold) with ULTRArrayHyb or unchanged with MicroHyb. In general, the results obtained with ExpressHyb and RT-PCR correlated well with biological responses to acetaminophen, while ULTRArrayHyb produced similar trends of gene expression alterations, but to a lesser extent, while MicroHyb did not match the biological response to acetaminophen. A similar phenomenon was also observed in a transplacental arsenic carcinogenesis study, in which 72 arrays with MicroHyb were useless (data not shown), while the results obtained with ExpressHyb were valid (Liu, Xie, Ward, Diwan, & Waalkes, 2004). It is very important to verify array results by other means. Real-time RT-PCR analysis is more sensitive than array analysis and more accurately represents the gene expression changes in response to acetaminophen. In this regard, realtime RT-PCR is recommended in performing gene expression analysis, and is the present favored technique (Walker, 2002). In our experience, the consistency rate between realtime RT-PCR and ExpressHyb is about 85% (Liu et al.,
J. Liu et al. / Journal of Pharmacological and Toxicological Methods 50 (2004) 67–71
2004), while the consistency between real-time RT-PCR and MicroHyb is about 15% (data not shown). Thus, based on its simplicity, accuracy, and the time required, we strongly recommend the use of real-time RT-PCR as a follow-up study in microarray analysis. In summary, this benchmark presents a lesson we have learned from using different hybridization systems for membrane array analysis. Clean hybridization is not the only goal of arrays. It is of great importance that array data accuracy should undergo confirmation by other means such as real-time RT-PCR, and gene expression changes should be shown to accord with biological responses.
Acknowledgements The authors thank Drs. Kartik Shankar, Tong Zhou, Yaxiong Xie, and Alex Merrick for their valuable help and comments for this technical appraisal.
71
References Liu, J., Li, C., Waalkes, M. P., Clark, J., Myers, P., Saavedra, J. E., & Keefer, L. K. (2003). The nitric oxide donor, V-PYRRO/NO, protects against acetaminophen-induced hepatotoxicity in mice. Hepatology, 37, 324 – 333. Liu, J., Xie, Y., Ward, J. M., Diwan, B. A., & Waalkes, M. P. (2004). Toxicogenomic analysis of aberrant gene expression in liver tumors and non-tumorous livers of adult mice exposed in utero to inorganic arsenic. Toxicological Sciences, 77, 249 – 257. Shankar, K., Vaidya, V. S., Corton, J. C., Bucci, T. J., Liu, J., Waalkes, M. P., & Mehendale, H. M. (2003). Activation of PPAR-alpha in streptozotocin-induced diabetes is essential for resistance against acetaminophen toxicity. The FASEB Journal, 17, 1748 – 1750. Walker, N. J. (2001). Real-time and quantitative PCR: Applications to mechanism-based toxicology. Journal of Biochemical and Molecular Toxicology, 15, 121 – 127. Walker, N. J. (2002). Tech.Sight. A technique whose time has come. Science, 296, 557 – 579.
Brief communication
Hybridization buffer systems impact the quality of filter array data Jie Liua,*, Nigel Walkerb, Michael P. Waalkesa a
Inorganic Carcinogenesis Section, LCC, NCI at NIEHS, 111 Alexander Drive, RTP, NC 27709, USA b Laboratory of Comparative Biology & Risk Analysis, NIEHS, USA Received 19 December 2003; accepted 3 February 2004
Abstract Introduction: cDNA microarray technology has greatly facilitated mechanistic studies in pharmacology and toxicology. A clean hybridization with minimal background is critical for successful microarray analysis and is highly desired. However, clean hybridization alone is not enough; verification is needed. Methods: Total RNA was isolated from the livers of acetaminophen-intoxicated mice and was subjected to cDNA microarray analyses using ExpressHyb, ULTRArrayHyb or MicroHyb on nylon membranes. Real-time RT-PCR analyses were performed for verification. Results: We have demonstrated in this paper that hybridization systems can significantly impact the quality of array data. MicroHyb produced very clean hybridizations, but some results could not be confirmed by real-time RT-PCR and in accord with biological responses. The hybridization images from ExpressHyb were not always clean, but were reliable. The sensitivity of ULTRArrayHyb was moderate. Conclusion: This study has indicated the importance of selecting hybridization buffers in membrane arrays and recommended real-time RT-PCR for follow-up analysis. Gene expression changes should also be correlated with biological significance. D 2004 Elsevier Inc. All rights reserved. Keywords: Hybridization buffer systems; Filter array data; Microarray technology
1. Introduction
2. Method
The cDNA microarray has been a huge technical advance in biological research because it allows the simultaneous measurement of hundreds to thousands of genes. Two main types of microarray, the chip-array (glass) and filter-array (nylon membrane), are currently used for gene expression analysis. A clean hybridization with minimal background is critical for successful microarray analyses, and the need for the optimization of hybridization systems is a major part of experimental protocols, which leads to the development of various hybridization systems. Microarray hybridizations can be influenced by many factors. Here, we compared the three commercially available hybridization systems in membrane arrays and found that different hybridization systems could make a substantial difference in the quality of data from membrane arrays, using gene expression analysis of acetaminophen hepatotoxicity as an example.
2.1. Animals and treatment
* Corresponding author. Tel.: +1-919-541-3951. E-mail address: [email protected] (J. Liu). 1056-8719/$ – see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.vascn.2004.02.001
Adult male CD-1 mice were obtained from Charles River Laboratories (Wilmington, MA) and were housed in facilities accredited by the American Association for the Accreditation of Laboratory Animal Care at the National Institute of Environmental Health Sciences, at 20 –22 jC with a 12-h light/dark cycle. The animals were given free access to rodent chow and tap water. The mice were given a hepatotoxic dose of acetaminophen (APAP, 600 mg/kg ip for 8 h), and the livers were removed for array analysis as described previously (Liu et al., 2003). 2.2. Microarray analysis Total RNA was isolated with Trizol reagent (Invitrogen, Carlsbad, CA), followed by purification with RNeasy columns (Qiagen, Valencia, CA). Approximately 5 mg of total RNA was converted to [a-32P]-dATP-labelled cDNA probe using Moloney murine leukemia virus (MuLV) reverse transcriptase and the cDNA synthesis primer mix (Clontech, Palo Alto, CA) and was purified with NucleoSpin columns.
68
J. Liu et al. / Journal of Pharmacological and Toxicological Methods 50 (2004) 67–71
Table 1 Comparisons among three hybridization systems
Company Prehybridization Hybridization 1st Wash protocol 2nd Wash protocol
ExpressHyb
ULTRArrayHyb
MicroHyb
Clontech lot#0250B80 5 ml ExpressHyb, 50 Ag sperm DNA, 68 jC, 2 h Add 100 Al probe Cot-1 DNA 5 Ag, 68 jC, 16 h 2 SSC 1% SDS, 68 jC, 30 min 4 0.1 SSC 0.5% SDS, 68 jC, 30 min 2
Ambion lot#061P71A 10 ml ULTRArrayHyb, 50 Ag sperm DNA, 50 jC, 2 h Add 100 Al probe Cot-1 DNA 5 Ag, 50 jC, 16 h 2 SSC 0.5% SDS, 50 jC, 30 min 2
Research Genetics lot#R26478 5 ml MicroHyb, 50 Ag sperm DNA, 42 jC, 2 h Add 100 Al probe Cot-1 DNA 5 Ag, 42 jC, 16 h 2 SSC 1% SDS, 42 jC, 30 min 2
0.5 SSC 0.5% SDS, 50 jC, 30 min 2
0.5 SSC 1% SDS, 25 jC, 30 min 1
Three hybridization buffer systems, ExpressHyb from Clontech (lot#0250B80), ULTRArrayHyb from Ambion (Austin, TX; lot#061P71A), and MicroHyb from Research Genetics (Huntsville, AL, lot#R26478), were used according to conditions recommended by the manufactures (Table 1). The hybridizations were performed for 16 h using the same probe (6 106 cpm in 100 Al) and the same lot of the custom-made nylon membranes. The membranes were then washed according to the protocols for each hybridization system and then wrapped and exposed to a Molecular Dynamics Phosphoimage Screen (Sunnyvale, CA). The images were analyzed densitometrically using AtlasImage software (Clontech). The gene expression intensities were first corrected with the external background and then globally normalized.
as relative increases setting control as 100%. Assuming that the CT value is reflective of the initial starting copy and that there is 100% efficacy, a difference of one cycle is equivalent to a twofold difference in starting copy. 2.4. Statistics Means and S.E.M. (n = 4) were calculated for the Students’ t test. The level of significance was set at P < .05. The expression ratios (acetaminophen/control) were used for comparisons.
3. Results 3.1. Microarray images
2.3. Real-time RT-PCR analysis Expressions of the selected genes were quantified using real-time RT-PCR analysis (Walker, 2001). Briefly, total RNA was reverse transcribed with MuLV reverse transcriptase and oligo-dT primers. The forward and reverse primers for selected genes were designed using a Primer Express software (Applied Biosystems, Foster City, CA) and are listed in Table 2. The SYBR green DNA PCR kit (Applied Biosystems) was used for real-time PCR analysis. The relative differences in expression between groups were expressed using cycle time (CT) values, and the relative differences between groups were calculated and expressed
The representative hybridization images are shown in Fig. 1. Note the following: (1) The images were clean, but the patterns of the gene expression are quite different between the three hybridization systems. (2) The acetaminophen-induced expressions of genes encoding the acute phase proteins, such as c-Jun, heme oxygenase-1 (HO-1), early response growth factor-1 (ERG1), heat shock protein 70 (HSP70), the increased expression of DNA damage protein GADD45, and the decreased expression of CYP2E1, are readily seen with ExpressHyb, somewhat visible with ULTRArrayHyb, but are not evident with the MicroHyb system.
Table 2 Primer sequences used for quantitative real-time RT-PCR analysis Gene
Accession number
Forward primer
Reverse primer
c-Jun/AP-1 c-myc HO-1 EGR1 HSP70 GADD45 GADD153 CYP2E1 G3PDH h-actin
J04115 X00195 M33203 M20157 M35021 L28177 X67083 L11650 M32599 M12481
ACTCCGAGCTGGCATCCA CGCCGCTGGGAAACTTT CCTCACTGGCAGGAAATCATC AGGTTCCCATGATCCCTGACT GGACAAGCAAGCATTCCACA CAGATCCATTTCACCCTCATCC CTCCTGTCTGTCTCTCCGGAA CACAGCCAAGAACCCATGTACA AGTATGACTCCACTCACGGCAAAT GTATGACTCCACTCACGGCAAA
CCCACTGTTAACGTGGTTCATG TCCTGGCTCGCAGATTGTAA CCTCGTGGAGACGCTTTACATA GGTACGGTTCTCCAGACCCTG CTCAAGAGAGGCGGGATAAGG TCCAGTAGCAGCAGCTCAGC TACCCTCAGTCCCCTCCTCA CAGGAGCCCATATCTCAGAGTTG GTCTCGCTCCTGGAAGATGGT GGTCTCGCTCCTGGAAGATG
EGR1, early growth response protein-1; HSP70, heat shock protein (70 kD); GADD45 and GADD153, growth arrest and DNA damage inducible protein45 and 153; G3PDH, glyceraldehydes-3-phosphate dehydrogenase; HO-1, heme oxygenase-1.
J. Liu et al. / Journal of Pharmacological and Toxicological Methods 50 (2004) 67–71 Fig. 1. Representative phosphorimage of microarray results. Top, control normal liver samples; Bottom, acetaminophen-intoxicated liver samples. Arrows show acetaminophen-induced increases in expression of growth arrest and DNA damage inducible protein GADD45, HSP70, early growth response protein-1 (EGR1), c-jun and heme oxygenase-1 (HO-1), as well as the decrease in expression of cytochrome P450 2E1 (CYP2E1).
69
70
J. Liu et al. / Journal of Pharmacological and Toxicological Methods 50 (2004) 67–71
Table 3 Comparisons in acetaminophen-induced gene expression fold changes between real-time RT-PCR and hydridiztion systems Gene
Real-time RT-PCR analysis
ExpressHyb
ULTRArrayHyb
MicroHyb
c-Jun/AP-1 c-myc HO-1 EGR1 HSP70 GADD45 GADD153 CYP2E1 G3PDH h-actin
12.5 11.8 11.5 35.5 3.12 16.3 11.5 0.20 1.01 1.06
12.9 10.4 8.6 21.2 2.52 7.27 6.21 0.48 1.24 1.21
1.98 3.77 2.94 2.95 1.97 1.68 1.84 0.51 1.55 1.42
0.98 0.79 1.12 1.02 0.71 0.61 0.48 0.72 1.47 1.37
The mice were given acetaminophen (600 mg/kg ip for 8 h), and hepatic gene expression was assessed by real-time RT-PCR and microarray with three hybridization systems. The fold changes are expressed as mean ratios of acetaminophen/controls.
3.2. Real-time RT-PCR confirmation Expression of the selected genes was additionally quantified using real-time RT-PCR analysis. A comparison of gene expression changes by microarrays and real-time RT-PCR analyses is shown on Table 3. Using ExpressHyb system, the marked increases in acute phase proteins, such as c-Jun ( 12), c-myc ( 10), HO-1 ( 9), early response protein1 (EGR1, 21), and heat shock protein-70 (HSP70, 2.5), the genes encoding DNA damage and cell death, such as GADD45 ( 10) and GADD153 ( 6.2), and the suppression of constitutive metabolic enzymes, such as CYP2E1 ( 0.48), were well correlated with real-time RT-PCR analysis. The ULTRArrayHyb system also produced similar trend, but to a less extent, while MicroHyb system did not match RT-PCR results. For example, the expression of c-Jun was 12.5, 12.9, 1.98, and 0.98 for real-time RTPCR, ExpressHyb, ULTRArrayHyb, and MicroHyb, respectively. The expressions of the housekeeping genes G3PDH and h-actin were about the same across these hybridization buffers.
4. Discussion This study provides a lesson we have learned from our own mistakes. In an attempt to get a clear hybridization image, we tested five different hybridization systems and found that MicroHyb system was the best and always gave very clean hybridization images, while the ExpressHyb did not always produce clean images. We have thus switched to MicroHyb system for over 150 arrays in five different projects. The hybridizations of every membrane with MicroHyb were very clean and highly reproducible. However, the obtained gene expression profiles did not match biological responses, and we were unable to make a coherent interpretation of these 150 arrays. This frustrating experience pro-
moted us to compare hybridization systems using the same probes. The study demonstrated that the choice of hybridization systems could make a dramatic difference in data obtained from membrane arrays. To our surprise, gene expression patterns are quite different among the three hybridization systems (Fig. 1). Some visible spots on the membrane with ExpressHyb or UltrarrayHyb are invisible when MicroHyb is used, and vice versa. For example, there is a 10-fold difference in the basal expression level of CYP2E1. To define the reason(s) for such a huge discrepancy is beyond our capability, as the recipes for these hybridization buffers are confidential. The only thing we could do was to vary the temperature of hybridization and washing conditions, but the differences still remained (data not shown). Thus, only the standard procedures recommended by the manufactures were used for comparison in this paper. All three hybridization systems produced clean, reproducible hybridization images, but the gene expression patterns were quite different, meaning that it was not possible to ascertain which results represented reality. To address this question, real-time RT-PCR analysis was performed on selected genes. As shown in Table 3, there is an excellent match between real-time RT-PCR and ExpressHyb, a good match between real-time RT-PCR and ULTRAarrayHyb, and a poor match between real-time RT-PCR and MicroHyb. The real-time RT-PCR analysis supports the results obtained from ExpressHyb, although its hybridization image is not always clean. It is very important to correlate gene expression changes with biological function. We thus compared the three hybridization systems in two separate studies on acute acetaminophen hepatotoxicity (Liu et al., 2003; Shankar et al., 2003). Both studies gave similar, consistent gene expression patterns in response to acetaminophen-induced liver injury. For example, a 12-fold induction of c-Jun was seen with ExpressHyb and real-time RT-PCR, while it was moderately increased (2-fold) with ULTRArrayHyb or unchanged with MicroHyb. In general, the results obtained with ExpressHyb and RT-PCR correlated well with biological responses to acetaminophen, while ULTRArrayHyb produced similar trends of gene expression alterations, but to a lesser extent, while MicroHyb did not match the biological response to acetaminophen. A similar phenomenon was also observed in a transplacental arsenic carcinogenesis study, in which 72 arrays with MicroHyb were useless (data not shown), while the results obtained with ExpressHyb were valid (Liu, Xie, Ward, Diwan, & Waalkes, 2004). It is very important to verify array results by other means. Real-time RT-PCR analysis is more sensitive than array analysis and more accurately represents the gene expression changes in response to acetaminophen. In this regard, realtime RT-PCR is recommended in performing gene expression analysis, and is the present favored technique (Walker, 2002). In our experience, the consistency rate between realtime RT-PCR and ExpressHyb is about 85% (Liu et al.,
J. Liu et al. / Journal of Pharmacological and Toxicological Methods 50 (2004) 67–71
2004), while the consistency between real-time RT-PCR and MicroHyb is about 15% (data not shown). Thus, based on its simplicity, accuracy, and the time required, we strongly recommend the use of real-time RT-PCR as a follow-up study in microarray analysis. In summary, this benchmark presents a lesson we have learned from using different hybridization systems for membrane array analysis. Clean hybridization is not the only goal of arrays. It is of great importance that array data accuracy should undergo confirmation by other means such as real-time RT-PCR, and gene expression changes should be shown to accord with biological responses.
Acknowledgements The authors thank Drs. Kartik Shankar, Tong Zhou, Yaxiong Xie, and Alex Merrick for their valuable help and comments for this technical appraisal.
71
References Liu, J., Li, C., Waalkes, M. P., Clark, J., Myers, P., Saavedra, J. E., & Keefer, L. K. (2003). The nitric oxide donor, V-PYRRO/NO, protects against acetaminophen-induced hepatotoxicity in mice. Hepatology, 37, 324 – 333. Liu, J., Xie, Y., Ward, J. M., Diwan, B. A., & Waalkes, M. P. (2004). Toxicogenomic analysis of aberrant gene expression in liver tumors and non-tumorous livers of adult mice exposed in utero to inorganic arsenic. Toxicological Sciences, 77, 249 – 257. Shankar, K., Vaidya, V. S., Corton, J. C., Bucci, T. J., Liu, J., Waalkes, M. P., & Mehendale, H. M. (2003). Activation of PPAR-alpha in streptozotocin-induced diabetes is essential for resistance against acetaminophen toxicity. The FASEB Journal, 17, 1748 – 1750. Walker, N. J. (2001). Real-time and quantitative PCR: Applications to mechanism-based toxicology. Journal of Biochemical and Molecular Toxicology, 15, 121 – 127. Walker, N. J. (2002). Tech.Sight. A technique whose time has come. Science, 296, 557 – 579.