Anal. Chem. 2004, 76, 1411-1418
Real-Time Detection of Bacterial Contamination in Dynamic Aqueous Environments Using Optical Sensors Jin Ji, J. Anthony Schanzle, and Mary Beth Tabacco*
Echo Technologies, Inc., 451 D Street, Boston, Massachusetts 02210
Here we describe on-line, real-time detection of waterborne bacteria using an optical sensor based on a starburst dendrimer film containing a lipophilic fluorophore. The sensor is constructed via covalent coupling between amine-terminated polyamidoamine dendrimer and silanized glass through an amide bond. The reporter molecule is embedded in the dendrimer layer through host-guest interaction. Real-time automated detection and quantitation of the bacteria are realized by using a charge-coupled detector camera and customized imaging and analysis software. The sensor responds to bacteria introduced to an aqueous flow system within 1 min. The limit of detection is ∼104cells/mL. The operational lifetime is more than 64 h, and the storage lifetime of the sensor is at least 7 months. Development of sensors for real-time detection of bacterial contamination in water supplies is a high priority with applications in national security and domestic preparedness (e.g., Biological Warfare Agent detection) and for ensuring the safety of municipal and recreational water supplies (presently two major EPA initiatives). To meet these needs, the sensors must be sensitive, rapid, and robust with long operational lifetime. In addition to these criteria, the sensors must be easily operated and require minimum maintenance. At present such sensors do not exist. The difficulties come in part from the limited availability of sensitive and robust bacteria recognition elements. Another challenge is development of suitable supporting matrixes to preserve the sensing elements and to attract and retain the bacterial contaminants in a dynamic aqueous environment. In recent years, detection methods using biomarkers, specific antibodies, and RNA and DNA probes have been demonstrated that allow the direct identification and quantification of microorganisms.1-3 Intrinsic biomarkers such as membrane lipids and their associated fatty acids have proven to be useful in obtaining a quick assessment of the microbial composition,2-3 but the method has low sensitivity and accuracy. Immunodetection is a * To whom correspondence should be addressed. E-mail: mtabacco@ erols.com. Fax: 617-204-3080. (1) Parkes, R. J. In Ecology of microbial communities; Fletcher M., Gray T. R. G., Jones, J. G., Eds.; Cambridge University Press: Cambridge, 1987; pp 147-177. (2) Bo ¨ttger, E. C. ASM News 1996, 62, 247-250. (3) Nishihara, M.; Akagawa-Matsushita, M.; Togo, Y.; Koga, Y. J. Ferment. Bioeng. 79, 400-402. 10.1021/ac034914q CCC: $27.50 Published on Web 01/20/2004
© 2004 American Chemical Society
powerful technique for the identification of a specific microorganism. With the recent advances in antibody-enzyme-linked immunosorbent assay,4 the technique is now able to quantify the target analyte. The use of new genetically engineered antibodies in biosensors shall help overcome the instability observed with the native form of the probes.5 Excellent work has also been done using RNA and DNA probes for detection of pathogenic microorganisms in water,6-7 Combined with polymerase chain reaction, this approach has also been demonstrated to detect microorganisms in a portable instrument in 7 min,8 a dramatically shortened analysis time given sample preparation requirements. However, these techniques are limited when faced with unknown or genetically modified organisms. Sensors based on molecular imprinted polymers ) is another new approach that has potential for real-time detection of microorganisms. Moldable polyurethane was imprinted with large yeast cells and coupled with a mass-sensitive quartz crystal microbalance (QCM) as the transducer to detect 1 × 104 cells/ mL Saccharomyces cerevisiae yeast with selectivity over other yeast species.9-10 The targets amenable to this technique, however, are currently limited to large yeast cell lines because sensor preparation heavily depends on the weight and size of the analyte. The use of the technique for detection of pathogenic organisms of interest such as Escherichia coli has been reported to be poor.9 The technique is also subject to interference from nonbiological particles that are of a size and shape similar to the contaminants. The concept of using fluorescent nucleic acid stains as the sensing elements to construct semiselective sensors has recently been demonstrated.11-13 This type of sensor is based on the property of increasing fluorescence quantum yield of some of (4) Bauer-Kreisel, P.; Eisenbeis, M.; Scholz-Muramatsu, H. Appl. Environ. Microbiol. 1996, 62, 3050-3052. (5) Colcher, D.; Pavlinkova, G.; Beresford, G.; Booth, B. J.; Choudhury, A.; Batra S. K. Q. J. Nucl. Med. 1998, 42(4), 225-241. (6) Amann, R. I.; Ludwig, W.; Schleifer, K.-H. J. Bacteriol. 1996, 178, 34963500. (7) Mozola, M. A. IFT Basic Symp. Ser. 1997, 12, 207-228 (Food Microbiological Analysis). (8) Belgrader, P.; Benett, W.; Hadley, D.; Richards, J.; Stratton, P.; Mariella, R.; Milanovich, F. Science 1999, 284, 449-450. (9) Hayden, O.; Dickert, F. L. Adv. Mater. 2001, 13, 1480-1483. (10) Dickert, F. L.; Hayden, O. Anal. Chem. 2002, 74(6), 1302-1306. (11) Chuang, H.; Macuch, P.; Tabacco, M. B. Anal. Chem. 2001, 73, 462-466. (12) Chang, A. C.; Gillespie, J. B.; Tabacco, M. B. Anal. Chem. 2001, 73, 467470. (13) Taylor, L. C.; Tabacco M. B.; Gillespie, J. B. Anal. Chim. Acta 2001, 435, 2309-246.
Analytical Chemistry, Vol. 76, No. 5, March 1, 2004 1411
fluorophores when bound to nucleic acids such as DNA or RNA in microorganisms. Unlike sensors using antibodies and RNA/ DNA probes, these sensors classify microorganisms so that the exact nature of the agent does not need to be known in advance. The sensors provide presumptive detection and early warning or function as a “smart trigger” to signal another analytical instrument to sample and identify the contaminants. To date, nucleic acid stains in solution phase have been demonstrated to be chemically robust, rapid, and sensitive to bacteria. However, immobilization of nucleic acid stains in a membrane for extended on-line detection of bacterial contamination in water supplies has not yet been realized. The difficulty is that most of the nucleic acid stains are hydrophilic and therefore not suitable for applications in dynamic aqueous environments. In addition, a suitable polymer support for the reagents has not been developed. The specific conditions demand that the support material act as a bioadhesive and be stable under flow conditions. This poses a dilemma because most of tissue adhesive materials are hydrophilic and therefore may not be chemically or mechanically stable under dynamic aqueous conditions. In this paper, we demonstrate the use of an amine-modified fourth generation starburst dendrimer and a membrane-reactive fluorophore, FAST DiA, to construct optical sensors for real-time detection of bacteria in water sources. Similar to some of the nucleic acid stains, the quantum yield of FAST DiA is enhanced when the fluorophore is introduced into a lipid environment such as a bacteria cell wall. However, unlike most of the nucleic acidstaining fluorophores, the lipophilic FAST DiA exhibits good hydrolytic stability in the aqueous phase. Dendrimers are well-defined, monodisperse, globular macromolecules constructed around a core unit. During synthesis, each successive reaction step leads to an additional “generation” of branching. The macromolecule also contains tunable inner cavities and can be modified with a high concentration of surface functional groups such as amine and carboxylate groups. This unique structure enables covalent bonding and host-guest interaction between dendrimers and many other reagents. Dendritic quaternary ammonium compounds have also been reported to promote disruption of cell membranes through their polycationic structure.14 Due to their interesting and potentially useful properties that derive from their unique structural characteristics, the macromolecules have been studied intensively for use in gene delivery,15-16 targeted delivery of toxic drugs used in chemotherapy,17 DNA biosensors fabrication,18-19 for use as antibacterial agents,14,20 and to deliver fluorescence probes into live bacteria.12 We report here the use of dendrimer films for cell capture and collection. In our study, fourth generation amine-terminated polyamidoamine (PAMAM) dendrimer (G4-NH2), which contains 64 surface amine groups, was coupled to silanized glass through an amide bond. The dendrimer layer functions as the biological (14) Chen, C. Z.; Tan, N. C. B.; Cooper, S. L. Chem. Commun. 1999, 15851586. (15) Zeng, F. W.; Zimmerman, S. C. Chem. Rev. 1997, 97, 1681. (16) Maksimenko, A. V.; Mandrouguine, V.; Gottikh, M. B.; Bertrand, J. R.; Majoral, J. P.; Malvy, C. J. Gene Med. 20035 (1), 61-71. (17) Freemantle, M. Science/Technology 1999, 77 (44), 27-35. (18) Watson, S. W.; Novitsky, T. J.; Quinby, H. L.; Valois, F. W Appl. Environ. Microbiol. 1977, 33, 940. (19) Benters, R.; Niemeyer, C. M.; Wohrle, D. Chembiochem 2001, 2 (9), 686694. (20) Chen, C. Z.; Cooper, S. L. Biomaterials 2002, 23 (16), 3359-68.
1412
Analytical Chemistry, Vol. 76, No. 5, March 1, 2004
adhesive element to capture suspended bacteria in a flow system. FAST DiA, which functions as the reporter fluorophore in the sensor, was introduced into the dendrimer layer through hostguest interaction. Real-time detection and quantitation of bacteria were realized by imaging the sensor surface using a chargecoupled detector (CCD) controlled by customized imaging and analysis software (IAS). The use of CCD imaging enables enumeration of individual bacteria on the sensor, dramatically increasing the sensitivity of the sensor. The sensor is simple to fabricate, robust, and stable under dynamic flow conditions for at least 64 h. The storage lifetime of the sensor is more than 7 months. The limit of detection is 1 × 104 cells/mL in flowing water, and the response time is less than 1 min. EXPERIMENTAL SECTION Chemicals. FAST DiA was obtained from Molecular Probes, Inc. (Eugene, OR). PAMAM dendrimer (G4-NH2, 10 wt % solution in methanol), glutaraldehyde (50%), aminopropyltriethoxylsilane (APTS), sodium bicarbonate, and toxin stimulant from albumin chicken egg were obtained from Sigma-Aldrich. Tryptic soy broth is obtained from Becton Dickinson and Co. (Cockeysville, MD). Urban dust (organic) was obtained from the U.S. Department of Commerce, National Institute of Standards and Technology (NIST, Gaithersburg, MD). The 25 × 75 mm cleaned microscope glass slides and 18 × 18 mm glass coverslips were purchased from VWR. Deionized-distilled water (DDI-H2O) was prepared from a Corning MP-6A Mega-pure system. When needed, the DDI-H2O was passed through a 0.2-µm TreckTech filter (VWR) for sterilization purpose. Microorganisms Preparation. All the microorganisms were obtained from the American Type Culture Collection (ATCC, Rockville, MD). E. coli cells (strain no. 15597), Gram-negative bacteria, were cultured in tryptic soy agar (TSA) media at 37 °C. TSA medium was composed of 30 g of tryptic soy broth, 15 g of agar/L of DDI-H2O. The medium was replaced once a week. Before use, cells were harvested from the culture plates by scraping, washed twice with sterilized DDI-H2O by centrifugation, and suspended in sterilized DDI-H2O at a concentration of (1-3) × 1010 cells/mL. The final concentration of bacteria was determined by acridine orange direct count (AODC)21 using a fluorescence microscope. Bacillus subtilis 23095, Gram-positive bacteria, were prepared from plates cultured in a similar manner, suspended in sterilized DDIH2O, washed two times by centrifugation, and counted via AODC method. The T7 virus (used in cross-reactivity studies) was prepared by propagating the virus in a host bacteria strain of E. coli using a standard protocol.22 The phage (T7) was added to an actively growing broth culture of the host strain (E. coli). The broth was mixed with soft agar and cultured on an agar plate. The propagated virus in the soft agar was scraped off the plate and centrifuged. The virus suspension was counted using standard plate count method. Penicillium roquefortii fungal spores (used in cross-reactivity studies) were also cultured according to the aforementioned method and harvested from plate cultures, suspended, washed in sterilized DDI-H2O two times by centrifugation, and counted by AODC method. (21) Mittleman, M. W.; Greesey, G. C.; Hite, R. R. Microcontamination 1983, 1 (2), 32-37. (22) Pienta, P., Tang, J., Cote, R., Eds. ATCC Bacteria&Bacteriophages, 19th ed.; American Culture Collection: Rockville, MD, 1996; pp 468-469.
Fabrication of Optical Sensors. Precleaned 25 × 75 mm microscope glass slides were cut into 18 × 18 mm squares to fit into custom sensor mounts. Prior to sensor fabrication, the glass coupons were carefully cleaned by heating in piranha solution at 100 °C for 2h. The glass coupons were then silanized with 4% APTS in acetone for 30 min. After silanization, the glass coupons were rinsed with acetone and DDI-H2O and oven cured at 140 °C overnight. Cured silanized glass coupons were allowed to cool to room temperature before being treated with 10% glutaraldehyde in 0.1 M sodium bicarbonate buffer (pH 8.3) for 2 h. Glass coupons were then rinsed with 0.1 M sodium bicarbonate buffer to remove unbound glutaraldehyde. A dendrimer-FAST DiA mixture was prepared by mixing 572 µM dendrimer (G4-NH2) in DDI-H2O and 1 mM FAST DiA in DMSO by thorough vortexing (v/v 1:1). A 20-µL aliquot of the mixture was deposited on to a glass coupon and covered with an 18 × 18 mm glass coverslip to form a sandwich. The sensors were incubated in a moisture-rich culture plate overnight at room temperature in the dark. The sensor fabrication is completed by rinsing off extra dendrimer-FAST DiA mixture on the glass coupon surface using DDI-H2O. The sensors were stored in argonpacked, light, and moisture tight aluminum foil packages at 4 °C until use. Experimental Setup. The sensors were tested using a prototype instrument developed in-house, the microbiological water monitoring system (MWMS) as shown in Figure 1a. The inset in Figure 1a is a photo of the flow cell. A schematic diagram of the system is shown in Figure 1b. The light source is a 473 ( 5 nm blue laser (Intelite Inc., Minden, NV). Aqueous samples were introduced to a custom-designed flow cell using a MasterFlexL/S pump (Cole-Palmer Instrument Co.) at flow rates ranging from 0.5 to 10 mL/min. Sensors were mounted at the backside of the flow cell and imaged through a 10× objective onto a CCD (DVC 1310-M, DVC Corp.). A 500-nm dichroic mirror and a 515-nm longpass emission filter are used in the optical path to improve the image quality. The system was controlled by a 1U (1.75-in. height) rack mount PC (CRM 102, Diversified Technology). IAS developed using Labview and IMAQ Vision software (National Instruments, Austin, TX) was used for image acquisition and automated cell counting. It should be noted that the whole system is portable (36.5 kg), compact (∼50 cm × 50 cm × 25 cm), and rugged, and is well suited for laboratory or field use. Digital photos presented in this paper were taken through a Nikon Coolpix 990 digital camera coupled with/without a Nikon Eclipse E400 fluorescence microscope. Sample Detection and Quantitation. The general experimental protocol is described here. Samples were introduced into the custom-designed flow cell using a MasterFlexL/S pump at a designated flow rate. After introducing an aliquot of bacteria sample, DDI-H2O was continuously pumped through the flow cell. Image acquisition using the IAS was automated to occur every minute or at another desired frequency. Cells that were captured and stained on the sensor appear on the inverted images as dark spots. The number of cells captured and stained on the sensor was determined by enumerating the dark spots in the image using the IAS. The cell count at 5 min after introduction of the bacteria spike was used as the time standard for sensor comparison.
RESULTS AND DISCUSSION Choice of Fluorescence Reporter. The goal of our work is to build a simple and robust optical sensor to monitor waterborne bacterial contaminants under dynamic aqueous flow conditions in real time. This requires the sensing element to remain stable during continuous exposure to flowing water. The sensor membrane should also possess the following attributes: (1) high chemical stability and photostability; (2) rapid cell staining; (3) low intrinsic fluorescence; and (4) high quantum yield upon interaction with bacterial contaminants. Based on these criteria, we screened a number of commercially available nucleic acid stains and cell membrane stains. We found that most of nucleic acid-staining fluorophores, such as SYTO 13 and SYBR Green, are hydrophilic. Sensors incorporating these dyes as the fluorescence reporter rapidly lost their sensing ability because of the loss of sensing element in continuous water flow. Probes with functional groups such as SYBR 101, succinimidyl ester (SE), SYBR 102, SE, and SYBR 103, SE were also evaluated. Although the loss of dye could be eliminated through covalent attachment via the SE groups, the photostability of these probes decreased significantly after covalent bonding. This resulted in poor sensitivity of the sensor (data not shown). FAST DiA, a lypophilic membrane-staining fluorophore, was found to be an excellent candidate optical reporter. The fluorophore displays good hydrolytic stability, rapid response (
