Screening of Target-Specific Stress-Responsive Genes for the

Oct 28, 2005 - coli and to show the DNA microarray-assisted develop- ment of whole cell-based biosensors. Gene expression data from the DNA microarray...
0 downloads 0 Views 334KB Size
Anal. Chem. 2005, 77, 8020-8026

Screening of Target-Specific Stress-Responsive Genes for the Development of Cell-Based Biosensors Using a DNA Microarray Byoung Chan Kim,† Chul Hee Youn,† Joo-Myung Ahn,‡ and Man Bock Gu*,‡

Advanced Environmental Monitoring Research Center (ADEMRC), Gwangju Institute of Science and Technology (GIST), 1, Oryoung-dong, Puk-gu, Gwangju 500-712, Republic of Korea, and College of Life Science and Biotechnology, Korea University, Anam-dong, Sungbuk-gu, Seoul 136-701, Republic of Korea

In this study, we describe a straightforward strategy to develop whole cell-based biosensors using fusions of the bacterial bioluminescence genes and the promoters from chemically responsive genes within Escherichia coli, in which chemical target-responsive genes were screened by using the information of gene expression data obtained from DNA microarray analysis. Paraquat was used as a model chemical to trigger gene expression changes of E. coli and to show the DNA microarray-assisted development of whole cell-based biosensors. Gene expression data from the DNA microarray were obtained by time course analysis (10, 30, and 60 min) after exposure to paraquat. After clustering gene expression data obtained by time course analysis, a group of highly expressed genes over the all time courses could be classified. Within this group, three genes expressed highly for overall time points were selected and promoters of these genes were used as fusion partners with reporter genes, lux CDABE, to construct whole cell-based biosensors. The constructed biosensors recognized the presence of model inducer, paraquat, and structural analogue chemicals of paraquat with a high specificity, and the results were reconfirmed by using DNA microarray experiments for those structural analogues. This strategy to develop whole cell-based biosensors assisted bywith DNA microarray information should be useful in general for constructing chemicalspecific or stress-specific biosensors with a high-throughput manner. Bacteria have been used to derive a number of strategies for sensing, responding, and adapting to different environmental conditions. Understanding how the transcriptional mechanisms of bacteria are modulated in response to changing conditions has been a major issue.1 Especially, changes in gene expression patterns when an organism experiences different conditions can be investigated using the high-throughput nature of DNA mi* To whom correspondence should be addressed. Phone: +82-2-3290-3417. Fax: +82-2-928-6050. E-mail: [email protected]. † Gwangju Institute of Science and Technology (GIST). ‡ Korea University. (1) Martinez-Antonio, A.; Salgado, H.; Gama-Castro, S.; Gutierrez-Rios, R. M.; Jimenez-Jacinto, V.; Collado-Vides, J. Biotechnol. Bioeng. 2003, 84, 743749.

8020 Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

croarray platform technologies.2,3 The completion of genomic sequences permits one to gather huge amounts of gene expression profile data from these organisms when they are present within different environments and conditions.4 The detailed association of genes, and their expression, to an exposure with toxic substances has been widely investigated in parallel data analyses.5-7 The objective of identifying physiological changes in bacterial cells when exposed to toxic substances is to understand and identify the mechanistic action of the contaminants and to eventually utilize the toxicant-induced gene expression in the selection of a biomarker.6 Furthermore, the selection of appropriate biomarkers would be useful in the construction of biosensors that are capable of detecting toxic chemicals. Such biomarker genes may be used directly, but if their promoters, along with other regulatory regions, are used in collaboration with appropriate reporter genes, such biosensors can be applied within various environmental monitoring systems.8 Previous whole cell-based biosensors have been developed and used widely, owing primarily to their simplicity of use and high specificity.9,10 Some of the more important issues to consider when developing biosensors include the selection of an appropriate reporter gene/protein as well as the switching element, in this case, the promoter for a gene induced by the stress of interest. Several reporters, i.e., the green fluorescent protein (GFP),11-13 (2) Banerjee, D.; Pillai, B.; Karnani, N.; Mukhopadhyay, G.; Prasad, R. Biochem. Biophys. Res. Commun. 2004, 317, 406-413. (3) Ginis, I.; Luo, Y.; Miura, T.; Thies, S.; Brandenberger, R.; Gerecht-Nir, S.; Amit, M.; Hoke, A.; Carpenter, M. K.; Itskovitz-Eldor, J.; Rao, M. S. Dev. Biol. 2004, 269, 360-380. (4) Van Dyk, T. K.; Wei, Y.; Hanafey, M. K.; Dolan, M.; Reeve, M. J.; Rafalski, J. A.; Rothman-Denes, L. B.; LaRossa, R. A. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 2555-2560. (5) Espinoza, L. A.; Smulson, M. E. Toxicology 2003, 189, 181-190. (6) Neumann, N. F.; Galvez, F. Biotechnol. Adv. 2002, 20, 391-419. (7) Pichler, F. B.; Laurenson, S.; Williams, L. C.; Dodd, A.; Copp, B. R.; Love, D. R. Nat. Biotechnol. 2003, 21, 879-883. (8) Gu, M. B.; Mitchell, R. J.; Kim, B. C. Adv. Biochem. Eng. Biotechnol. 2004, 87, 269-305. (9) Belkin, S. Curr. Opin. Microbiol. 2003, 6, 206-212. (10) Daunert, S.; Barrett, G.; Feliciano, J. S.; Shetty, R. S.; Shrestha, S.; SmithSpencer, W. Chem. Rev. 2000, 100, 2705-2738. (11) Cha, H. J.; Srivastava, R.; Vakharia, V. N.; Rao, G.; Bentley, W. E. Appl. Environ. Microbiol. 1999, 65, 409-414. (12) Norman, A.; Hansen, L. H.; Sorensen, S. J. Appl. Environ. Microbiol. 2005, 71, 2338-2346. (13) Rothert, A.; Deo, S. K.; Millner, L.; Puckett, L. G.; Madou, M. J.; Daunert, S. Anal. Biochem. 2005, 342, 11-19. 10.1021/ac0514218 CCC: $30.25

© 2005 American Chemical Society Published on Web 10/28/2005

luxAB,14 luxCDABE, 4,15-17 and luc,18 has been used extensively for stress identification, functional transcription, and bioimaging studies. Of these, the luxCDABE genes have several merits since all the proteins needed for the generation of the bioluminescence are encoded in vivo and generate visible light, which can be detected easily and quantified within several minutes,4 although it is not as sensitive as using GFP fluorescence. Additionally, biosensors that employ bioluminescence offer simplicity, and dynamic advantages, especially where multiple samples are needed.8 DNA microarrays offer a means of selecting positive candidates from the whole genome for the second component needed for biosensors, namely, the regulatory regions of specific genes. In the past, to construct this type of biosensor, promoters were selected from genes that were previous described in the literature or after characterization of each individual gene’s function. Using the DNA microarray, several genes induced under specific conditions can now be selected for further analysis and the results used to develop different types of biosensors easily and in a high-throughput manner. In this study, a strategy that uses DNA microarray data for the selection of appropriate biomarker genes and the subsequent development of whole cell-based biosensors is described. For this, Escherichia coli was used as a model organism and paraquat as the model toxicant. Using time course expression data, several genes that showed a continuous induction in their expression levels after exposure to paraquat were selected. Using fusions of the promoter regions from these genes and the bacterial luciferase genes, i.e., the luxCDABE genes, several bacterial bioluminescent biosensors were constructed. Finally, the ability of each of these biosensors to respond to paraquat, and structurally analogous chemicals, was investigated. EXPERIMENTAL SECTION Bacterial Strains and Chemical Exposure Protocols. E. coli strain RFM443 (strR, galK2, lac∆74)19 was used for DNA microarray experiments and biosensor construction. The exposure was performed as follows for DNA microarray experiments: First, RFM443 was cultivated with 100 mL of fresh Luria-Bertani (LB) medium at pH 7 up to a cell optical density (OD600) of 0.35 at 600 nm. At this OD600, 4.5 mL of the seed culture transferred to 15mL Falcon tubes (Corning, MA) and 0.5 mL of the chemical stock solution was added. The control culture had 0.5 mL of distilled water added to it. Both the treated cells and untreated cells were then incubated for 1 h at 30 °C and 250 rpm in a shaking incubator. To check for growth inhibition due to the presence of paraquat, the OD600 was compared between the control and sample cultures. For assays using the recombinant strains, seed cultures (100 mL of LB medium with ampicillin) were grown with agitation at 250 rpm in a shaking incubator in 250-mL culture flasks at 30 °C. (14) Ramanathan, S.; Ensor, M.; Daunert, S. Trends Biotechnol. 1997, 15, 500506. (15) Kim, B. C.; Gu, M. B. Biosens. Bioelectron. 2003, 18, 1015-1021. (16) Francis, K. P.; Joh, D.; Bellinger-Kawahara, C.; Hawkinson, M. J.; Purchio, T. F.; Contag, P. R. Infect. Immun. 2000, 68, 3594-3600. (17) O’Connell-Rodwell, C. E.; Burns, S. M.; Bachmann, M. H.; Contag, C. H. Trends Biotechnol. 2002, 20, S19-S23. (18) Liu, J.; Escher, A. Gene 1999, 237, 153-159. (19) Drolet, M.; Phoenix, P.; Menzel, R.; Masse, E.; Liu, L. F.; Crouch, R. J. Proc. Natl. Acad. Sci. U.S.A. 1995, 92, 3526-3530.

A new flask containing 50 mL of LB medium with ampicillin was inoculated with 2 mL of the seed culture. When the optical density (OD600) reached 0.35, the test chemical was added to the final test concentration. The biosensing activity of the fusion strains was determined using the relative bioluminescent response ratio level (RBL ) bioluminescent response to the tested chemical/ bioluminescent response of control sample). The bioluminescence emitted (arbitrary units, AU) was measured with 100-µL cell stocks using a highly sensitive luminometer (model 20e, Turner Design) or 96-well luminomenter (Dynex Technology, Inc.). All chemicals were purchased from the Sigma Co. except hydrogen peroxide, which was purchased from the Merck Co. cDNA Microarray. A total of 96 cDNAs from E. coli were used for this study. The genes were spotted in duplicates on the same slide (Genomictree, Daejeon, Korea). The genes included on the cDNA microarray are uspA, ushA, ahpC, rpoS, katE, katG, ompC, soxR, soxS, ubiA, pqiA, pta, grxA, rpoH, gor, dsbG, fnr, clpA, hemA, hemH, tnaA, ibpA, cysK, dsdA, dinF, dinG, dinI, phoH, polB, potA, potB, potD, trxA, marR, marA, rpoD, cyaA, crp, gadA, gadB, hns, atoC, atoS, speA, speC, speD, speG, pgi, zwf, sodA, sodB, sodC, icd, sdhA, mdh, cadA, arcA, arcB, cyoA, cyoB, cydA, rpoE, ptsG, hmp, hsdR, relA, spoT, ppk, nfo, aspA, fumC, fldA, ompA, ftsA, nhaA, aldA, dadX, inaA, malE, malK, nuoA, recA, umuD, uvrA, alkA, alkB, lacZ, nlpA, moaA, mtlR, oppA, sbmC, topA, acnA, oxyR, and rpoN. RNA Extraction and DNA Microarray Hybridization. One milliliter of the bacterial cell cultures was taken from treated and untreated cultures 10, 30, and 60 min after addition of the chemical or water, respectively, and 2 mL of bacterial RNA protectant (Qiagen) was added to inhibit further RNA synthesis or degradation within the samples. The total RNA was isolated from each sample using RNeasy RNA extraction kits (Qiagen). The quality of the extracted RNA was determined using the 260 nm/280 nm ratio. After mixing 10 µg of total RNA with 1 µL of random hexamer in RNAse-free water, the mixture was incubated at 65 °C for 10 min and immediately chilled on ice. Next, the reverse transcriptase reaction mixture (5 × first-strand buffer, 10 mM dATP, dCTP, and dGTP, 2 mM dTTP, 1 µL of 25 mM Cy3-dUTP for the unexposed sample or 1 µL of 25 mM Cy5-dUTP for the exposed sample) and 2 µL of Superscript II (200 unit/µL) reverse transcriptase (BD Biosciences) were added to each tube. The resulting 20-µL mixture was incubated at 42 °C for 2 h. After reverse transcription, to degrade the RNA, 0.1 M NaOH (15 µL) was added and the samples were incubated at 65 °C for 30 min after which the reaction samples were neutralized with 0.1 M HCl (15 µL). Next, the cDNA made with the exposed and unexposed culture RNA samples were labeled with Cy5-dUTP and Cy3-dUTP, respectively. The labeled cDNAs from replicate samples were mixed in one tube and then purified using a PCR purification kit (Qiagen) according to the manufacturer’s protocol. To precipitate the cDNA (200 µL), 20 µL of 3 M sodium acetate and 500 µL of 100% ethanol were added. After 10 min at room temperature, the labeled cDNA pools were centrifuged at 4 °C and 13 000 rpm for 10 min. Using 70% ethanol, the pellet was rinsed and then centrifuged again in the same manner for 3 min. The pellets were then dissolved into 30 µL of hybridization buffer (6 × SSC, 0.2% SDS, 5 × Denherdt’s solution, 0.1 mg/mL denatured salmon sperm DNA) and combined in one tube. Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

8021

Table 1. Primers Used To Amplify the Promoter Regions for Each of the Genesa primer

sequence

target gene

1 2 3 4 5 6 7 8

5′-agcagcgaattccgaatctttttcgcggcgta-3′ 5′-acttaaggatccttggcgcagggaagtattaa-3′ 5′-agcagcgaattcccgctcataccattacggc-3′ 5′-acttaaggatccgggtaaaagcggtgtcaagc-3′ 5′-acttaaggatccgcggctggtcaatatgctc-3′ 5′-agcagcgaattcgctgcgtttcgccactt-3′ 5′-agcagcgaattcgcggctggtcaatatgctc-3′ 5′-acttaaggatccgctgcgtttcgccactt-3′

fumC′ inaA′ soxR′ soxS′

a The underlined bases correspond to the restriction enzyme sites used in cloning.

Before hybridization, the chips were stored in a prehybridization solution (1.75 × SSC, 0.1% SDS, 10 mg/mL of bovine serum albumin) for 30 min at 65 °C. They were then washed using distilled water for 1 min and then 100% 2-propanol for 1 min at room temperature and finally dried by centrifugation (600 rpm, 5 min, and room temperature). The hybridization mixture from above was boiled for 2 min at 100 °C and then injected into the gap between the chips and slide cover glass. All hybridizations were performed for 16 h on the DNA microarray chip at 65 °C. After this, the chips were dipped once into buffer 1 (2 × SSC) at room temperature to remove the cover glass and washed once with buffer 2 (2 × SSC, 0.2% SDS), which was prewarmed to 65 °C, for 10 min, and twice with buffer 3 (0.05 × SSC) for 5 min. Afterward, the chip was dried by centrifugation (1200 rpm, 3 min, room temperature). Scanning and Data Mining. After hybridization, the washed microarray chips were scanned with a Genepix 4000B laser scanner (Axon Instruments, Inc.). The scanned images were captured as a 16-bit TIFF file format with 10-µm resolution, while intensity analysis was performed with GenepixPro 3.0 software (Axon Instruments, Inc.). The Cy3 and Cy5 intensity values were derived from a median value of the pixel-by-pixel intensity after subtracting the background fluorescence. The Cy5/Cy3 intensity ratios were calculated and normalized with a median value for all the spots. Clustering analysis for gene expression patterns were performed using Cluster 3.0.20 Clustering methods were performed with average linkage (average distance, UPGMA) clustering based upon correlation measure-based distance (uncentered). Cloning Strategy. To construct the recombinant biosensor cells, the promoter regions were amplified by PCR using the genomic DNA from E. coli strain RFM443. The DNA sequences were obtained from the National Center for Biotechnology Information (NCBI), and the primer pairs used to amplify the promoter regions are described in Table 1 for each biomarker gene that was selected. Using these primer pairs, the promoter regions, including part of the translated regions of each gene, were amplified and then ligated into plasmid pUCD615, which carries the promoterless luxCDABE genes.21 The ligated plasmids were then transformed into E. coli strain RFM443. RESULTS AND DISCUSSION Strategy for the Selection of Potential Biomarker Genes from DNA Microarray Results. The cDNA microarray was (20) Eisen, M. B.; Spellman, P. T.; Brown, P. O.; Botstein, D. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 14863-14868.

8022

Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

printed using 96 genes from E. coli to investigate changes in gene expression after exposure of E. coli to model chemicals. Although the whole genome of E. coli is not represented in this study, these 96 genes are sufficient for the downstream selection of target genes and to clearly demonstrate the feasibility of the protocol presented in this study, namely, the development of whole cellbased biosensors based upon variations in E. coli’s gene expression patterns after exposure to chemicals. Although information on all the genes represented is readily available, this was initially not taken into consideration and each spot was randomly numbered from 1 to 96. In this manner, “blind” tests were performed so as not to bias the results when appropriately responsive genes were selected. Paraquat was used as a model chemical since it was known to generate superoxide radicals in E. coli.22 At first, growth inhibition tests were performed with cultures of E. coli strain RFM443 with various concentrations of paraquat to determine the most suitable concentration. It was found that 35 ppm was the highest sublethal concentration tested and thus was used in all subsequent experiments. To monitor changes in gene expression, cDNA probes were prepared using cultures that were exposed to 35 ppm paraquat for 10, 30, and 60 min. Two independent replicates of the cDNA microarray experiments were done in parallel with Cy3 (untreated sample)- and Cy5 (treated sample)-labeled cDNA probes to determine the reliability of the hybridization signals. Scatterplots of the Cy3 versus Cy5 fluorescent values revealed a tight distribution (data not shown). After converting the signal intensities to Log2 ratio values, i.e., Log2 (Cy5/Cy3), gene clustering was carried out. The expression profiling and its gene clustering results are shown in Figure 1a. Based on the clustering results, the genes were separated into seven classifications (or groups) according to the changes in their expression levels. The average expression levels for each group are presented in Figure 1b. In selecting an appropriate biomarker gene, a gene whose expression is highly downregulated, upregulated, or undergoes both up- and downregulation during the time course is a potential candidate. Furthermore, genes that highly upregulated are obviously good candidates for sensing target chemicals or stresses. As well, those that are strongly downregulated may also be of interest, but their use in biosensors includes some risks since the loss of a reporter protein’s activity is harder to quantify. For instance, the basal level bioluminescence of the fusion strains is typically near zero15,23,24 but differs based on the selection of the promoters and their regulatory characteristics inside the cells.4 Furthermore, a decrease in the bioluminescence could be due to general cellular toxicity and not truly represent the mechanistics of the toxicity.25 For these reasons, the use of genes that are highly upregulated is considered to be more favorable. Therefore, candidate genes for biomarkers were selected from those that were highly upregulated, i.e., at least 2-fold, which did not include any of the genes in groups 1, 2, 3, or 4. (21) Rogowsky, P. M.; Close, T. J.; Chimera, J. A.; Shaw, J. J.; Kado, C. I. J. Bacteriol. 1987, 169, 5101-5112. (22) Hassan, H. M.; Fridovich, I. J. Biol. Chem. 1978, 253, 8143-8148. (23) Belkin, S.; Smulski, D. R.; Vollmer, A. C.; Van Dyk, T. K.; LaRossa, R. A. Appl. Environ. Microbiol. 1996, 62, 2252-2256. (24) Vollmer, A. C.; Belkin, S.; Smulski, D. R.; Van Dyk, T. K.; R. A., L. Appl. Environ. Microbiol. 1997, 63, 2566-2571. (25) Kim, B. C.; Park, K. S.; Kim, S. D.; Gu, M. B. Biosens. Bioelectron. 2003, 18, 821-826.

Figure 1. (a) Hierarchical gene clustering of the expression data after exposure to 35 ppm paraquat for 10, 30, and 60 min. Based on the clustering, each gene was classified within one of seven groups. Group 6 is shown in greater detail, and the numbering corresponds to the different genes. (b) Average Log2 expression levels of the genes within each group were obtained from two independent experiments. The standard variations in the expression levels for each group are shown as error bars.

Whereas the expression levels of the genes in groups 5 and 7 increased as time progressed, no significant expression was seen for an exposure of only 10 min (Figure 1b). In contrast, the genes in group 6 showed a significant increase in their expression levels 10 min after addition of paraquat with an average response ratio of ∼2-fold. Furthermore, the expression levels were still elevated 30 and 60 min after initiating the exposure. This prolonged gene expression with group 6 genes suggests that they are strongly upregulated by the presence of paraquat. Although the genes in groups 5 and 7 are also potential candidates, their expression was delayed, while their average expression levels were generally lower than seen with the group 6 genes. Considering the characteristics of biosensors, i.e., specificity, speed, and simplic-

ity,26 the biomarker genes from group 6 would be more favorable. Therefore, three genes from this group were selected for constructing the whole cell-based biosensors and include, from Figure 1a, genes 9, 42, and 83. Validation of the Biological Responses with Recombinant Bacterial Bioluminescent Biosensors. Before constructing the whole cell-based biosensors, the genes corresponding to numbers 9, 42, and 83 were identified as soxS, inaA, and fumC, respectively. Not surprisingly, all three of these genes were previously found to be induced by paraquat exposure and are within the SoxRS (26) Rogers, K. R.; Gerlach, C. L. Environ. Sci. Technol. 1996, 30, 486A-491A.

Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

8023

Figure 2. Schematic diagram showing the construction of whole cell-based biosensors.

Figure 3. Change in the bioluminescent expression after exposure to 35 ppm paraquat. The strains tested are the soxS (a), soxR (b), fumC (c), and inaA (d) fusion strains. These assays were performed in flask cultures. Paraquat was added to the cultures when the OD600 was 0.35. The values presented are the relative bioluminescent levels.

regulon.27 The fumC and inaA genes are also regulated by the SoxS protein. In addition, the soxR gene (from group 4) also was selected since the active form of SoxR activates transcription of the soxS gene.28 To construct the biosensing cells, the promoter regions from each of these genes were amplified by PCR with specific primer pairs (Table 1) and ligated into plasmid pUCD615, giving pBCfumC′Lux, pBCinaA′Lux, pBCsoxS′Lux, and pBCsoxR′Lux (Figure 2). After selection of individual colonies carrying the appropriate plasmid, each was screened for its ability to respond through an enhanced bioluminescence to 35 ppm paraquat. Independent verification of the transcriptional responses with the fusion strains was done in batch cultivations using the same conditions as for the DNA microarray experiments. The bioluminescent responses from each of the fusion strains are shown in Figure 3. For the three strains carrying fusions of the lux genes with soxS′, fumC′, and inaA′, the bioluminescent levels increased significantly after addition of 35 ppm paraquat (Figure 3a, c, and d). The maximum bioluminescent signal was seen after ∼80 min. (27) Pomposiello, P. J.; Bennik, M. H.; Demple, B. J. Bacteriol. 2001, 183, 38903902. (28) Pomposiello, P. J.; Demple, B. Trends. Biotechnol. 2001, 19, 109-114.

8024

Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

Furthermore, the time course bioluminescent expression pattern for all three of the biosensing strains was similar to the microarray expression data, suggesting that fusions using the soxS, fumC, or inaA promoters can be used to develop biosensing cells that can sense paraquat. In contrast, the strain carrying a fusion of the lux genes with the soxR promoter was unresponsive after addition of paraquat (Figure 3b). Although the protein encoded by this gene is involved in sensory and regulatory functions after exposure to paraquat and other superoxide-generating compounds, transcription of this gene remains unchanged even after exposure (soxR is classified as a group 4 gene in Figure 1b). Therefore, this lack of response from the fusion strain is not surprising. For the strains that were responsive, the results from the DNA microarray and bioluminescent assays differed greatly in their magnitude.For example, the expression ratios at 60 min for the microarray after exposure of E. coli strain RFM443 to 35 ppm paraquat were 14.5, 3.9, and 2.6 for soxS, fumC, and inaA, respectively, while those of the bioluminescent assay were 29 665, 2138, and 38.4 (relative values) for the soxS′, fumC′ , and inaA′ fusion strains, respectively. These differences in the ratios are thought to be due to the biomolecules used in each assay. In the DNA microarray, the relative quantity of mRNA in the cells is used as an indicator of induced expression. On the other hand, the bioluminescent assay relies upon the results of an enzymatic reaction. The expression of the Lux proteins, however, is tied to the promoter strength (as is the production of mRNA) but also other characteristics, such as the mRNA stability and translation efficiency, which are inherent to the mRNA being produced. Therefore, the differences, in particular, the high bioluminescent level of the soxS′-lux fusion strain, were not unexpected since the promoter is highly induced when the cells are exposed to paraquat. However, the results from both the DNA microarray and bioluminescence assays clearly show that it is feasible to develop bioluminescent bacterial biosensors from the DNA microarray results. Specificity of Responses of the Bacterial Bioluminescent Biosensors. One of the more important issues in biosensing is the specificity of the biosensor.22 Although the genes selected to construct the fusions used in this study responded to paraquat, these responses do not guarantee the specificity of these strains to only one given target chemical or stress, in this case, the generation of the superoxide radical. Therefore, each of the four fusion strains was exposed to various concentrations of mitomycin C and hydrogen peroxide, in addition to paraquat. Mitomycin C is a DNA damaging agent with a very high specificity toward DNA since it leads to interstrand DNA cross-links.29 Hydrogen peroxide is a well-known oxidative damaging agent that, in E. coli, leads to an OxyR-regulated response.18 Figure 4 shows the maximum relative bioluminescent values from each of the four strains after exposure to paraquat, mitomycin C, and hydrogen peroxide. For paraquat, a dose-dependent response was observed up to 35 ppm for the soxS′, fumC′, and inaA′ fusion strains. In contrast, these strains showed no induction after addition of mitomycin C (50 ppb) or hydrogen peroxide (0.000 15%). As expected, the soxR′ fusion strain was unresponsive to any of the chemicals. The expression levels of these three gene were reconfirmed using DNA microarray experiments. None of these three genes showed (29) Tomasz, M.; Palom, Y. Pharmacol. Ther. 1997, 76, 73-87.

Figure 4. Relative bioluminescence values 1 h after induction with various concentrations of paraquat, mitomycin C, and hydrogen peroxide. The chemicals were added to the cultures when the OD600 was 0.35.

a significant change in their expression when E. coli strain RFM443 was exposed to either mitomycin C (50 ppb) or hydrogen peroxide (0.000 15%). In contrast, the recA and katG genes, two genes present on the microarray, were highly induced by MMC and hydrogen peroxide, respectively. (Data not shown.) This is expected since the recA gene is strongly induced by DNA damaging chemicals while the katG gene is induced when E. coli is exposed to hydrogen peroxide. Biosensing of Chemicals Structurally Similar to Paraquat. Four biosensor cells (soxS′-lux, soxR′-lux, fumC′-lux, inaA′-lux), all constructed in this study, were used to investigate direct sensing of chemicals structurally analogous to paraquat (methyl viologen,

MV). Several of the chemicals have quaternary amines, and include 1,1′-diethyl-4,4′-bipyridinium (ethyl viologen, EV), 1,1′diheptyl-4,4′-bipyridinium (heptyl viologen, HV), and 1,1′-dibenzyl4,4′-bipyridinium (benzyl viologen, BV). Two additional chemicals, 4,4′-dipyridyl (DP) and 4,4′-dipyridyl N,N′-dioxide (DD), were also investigated. These have structures similar to paraquat but lack the quaternary amine (Figure 5a). Changes in the bioluminescent levels after a 1-h exposure for different concentrations of each chemical (0.05, 0.5, 5, and 50 mM) are presented in Figure 5b. For both DP and DD, which lack the quaternary amine, no observable change in the bioluminescence is seen, regardless of the biosensor tested. Interestingly, MV, EV, HV, and BV, all of which have quaternary amines, led to induced bioluminescent levels in the soxS′-lux, fumC′-lux, and inaA′-lux biosensors. In all the tests, the soxR′-lux strain did not respond. From these results, it appears likely that the quaternary amines of these chemicals play a major role in generating the superoxide radical. MV and EV led to an induction of the soxS′-lux, fumC′-lux, and inaA′-lux genes over a broad range of concentrations, while HV and BV led to an increased bioluminescence from the strains carrying these genes only at the lowest concentration tested (0.05 mM). Furthermore, the bioluminescent levels of the cultures exposed to 0.05 mM HV or BV were higher than those from the cells exposed to 0.05 mM MV or EV. It is thought that higher concentrations of HV and BV lead to significant amounts of cellular toxicity, which then result in a loss of growth. The results from the biosensors after being exposed to these chemicals were confirmed using a DNA microarray analysis. To do this, each strain was exposed to 0.05 mM concentrations each chemical for 1 h, after which changes in each genes expression levels were determined with the microarray. As shown in Figure 5c, the expression of these genes did not significantly change after exposure to DP and DD. Meanwhile, MV, EV, HV, and BV, which are structurally similar, all highly induced expression of the soxS, fumC, and inaA genes. Furthermore, the microarray results

Figure 5. Biosensing of paraquat (MV) and structurally similar chemicals. (a) Shows the structures of the chemicals, (b) shows the bioluminescence levels seen (assay was performed in 96-well plates), and (c) shows the gene expression results from DNA microarray analyses. The number in each box is Log2 value representing the gene expression ratio. N.D.: not detectable.

Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

8025

correlate well with the biosensor results, with BV showing the highest induction in both assays for all the strains. Overall, these results indicate that the biosensors constructed using the microarray results can be used for sensing similarly reactive chemicals. They also show the possibility of discriminating between chemical species based upon the bioluminescent level of the biosensors. In conclusion, we report a straightforward strategy for the construction of bacterial bioluminescent biosensors, which are specific to a target chemical and structurally analogous chemicals. This was achieved by screening biomarker genes based on DNA microarray analysis results. The use of DNA microarray informatics, which enables researchers to monitor and screen changes in a massive number of genes within a high-throughput manner, simplifies the screening to only genes of interest.30,31 As such, this study shows a direct approach employing in vivo transcription results for the development of bacterial bioluminescent biosensors. Although the number of genes used here was limited, they were sufficient to demonstrate the ease with which one screens and selects target genes based upon their expression results. (30) Conzone, S. D.; Pantano, C. G. Mater. Today 2004, 7, 20-26. (31) Shoemaker, D. D.; Linsley, P. S. Curr. Opin. Microbiol. 2002, 5, 334-337.

8026

Analytical Chemistry, Vol. 77, No. 24, December 15, 2005

This study also evaluated and screened target genes according to their time-dependent response after chemical exposure. The clustering results showed several genes that were highly induced after exposure to paraquat, indicating their potential use as biosensing components and biomarkers. Several of these genes were selected and their promoter regions used in the development of bacterial biosensors. Characterization of these strains found them to respond in a dose-dependent manner, as well as be specific to the target chemical or stress, in this case the generation of the superoxide radical. Therefore, the strategy to develop whole cell-based biosensors presented in this study, as assisted by DNA microarray data, is useful when constructing chemical-specific or stress-specific biosensors. ACKNOWLEDGMENT This work was supported by the Korea Science and Engineering Foundation (KOSEF) through ADEMRC at GIST. The authors are grateful for the support. Received for review August 8, 2005. Accepted September 23, 2005. AC0514218