Electrochemical Biosensor Array for the Identification of

ACS Applied Materials & Interfaces 2015 7 (50), 27755-27764. Abstract | Full Text ..... Microchimica Acta 2013 180, 1-13 ... BioNanoScience 2012 2, 19...
6 downloads 0 Views 75KB Size
Download PDF
Anal. Chem. 2001, 73, 4241-4248

Electrochemical Biosensor Array for the Identification of Microorganisms Based on Lectin-Lipopolysaccharide Recognition Peter Ertl and Susan R. Mikkelsen*

Department of Chemistry, University of Waterloo, Waterloo, Ontario, N2L 3G1 Canada

Rapid identification of bacterial strains remains a wellknown problem in applied medicine and, for viable pathogens, is an important diagnostic goal. We have investigated an electrochemical biosensor array, in which transduction is based on respiratory cycle activity measurements, where the microorganism’s native respiratory chain is interrupted with non-native external oxidants. The selective biochemical recognition agents employed in this study are lectins that, once immobilized, recognize and bind to cell surface lipopolysaccharides. Porous membranes with different surface properties were examined as potential immobilization supports for these lectins. Optimizations performed using concanavalin A and E. coli JM105 show that immobilization methods involving preactivated membranes significantly reduce the time required to create a functional lectin layer on the membrane surface. Overall, we found general agreement between agglutination test results and the electrochemical assessment of lectin-cell binding. Chronocoulometric measurements were made for cells captured on lectin-modified Immunodyne ABC membranes physically affixed to Pt working electrodes. This lectin-based sensor array was exposed to viable cells of Gram-negative and Grampositive bacteria as well as yeast, and chronocoulometric measurements were used to generate a pattern of responses for each organism toward each lectin. Principal component analysis was used to classify the chronocoulometric results for the different microbial strains. With this new method, six microbial species (Baccilus cereus, Staphylococcus aureus, Proteus vulgaris, Escherichia coli, Enterobacter aerogenes, Saccharomyces cerevisiae) were readily distinguished. The detection, identification, and quantitation of microorganisms play a vital role in fermentation technology, medical practice, and environmental monitoring. Bacterial pathogens are distributed in soil, marine waters, water contaminated with fecal matter, and the intestinal tracts of animals.1 Many bacterial species have been identified as significant food and waterborne pathogens that have profound effects on humans, and the incidence of pathogen-related * Corresponding author: (voice) (519)888-4567 ext 6871; (fax) (519) 746-0435; (e-mail) [email protected]. (1) Atlas, R. M. Principles of Microbiology, 2nd ed.; Wm. C. Braun Publishers: Dubuque, IA, 1997; Chapter 12. 10.1021/ac010324l CCC: $20.00 Published on Web 07/24/2001

© 2001 American Chemical Society

diseases among humans has not yet decreased.2 Generally, no single test provides definitive identification of an unknown bacterium. Traditional bacterial detection methods involve a preenrichment step and a selective enrichment step, followed by biochemical screening and serological confirmations;3 this complex series of tests can last up to 72 h. Thus a sensitive, reliable, and more rapid identification method is required for the detection and early treatment of infections. Many instrumental methods have been reported for microbial detection, including surface plasmon resonance, fiber optics, and flow cytometry.4 Infrared spectroscopy and the gas chromatographic mass spectrometric detection of fatty acids have also been used for the detection and identification of microorganisms.2 Other methods use piezoelectric crystals, impedimetry, calorimetry, and selective detection of cellular compounds such as ATP, DNA, protein, and lipid derivatives, as well as the detection of metabolic processes such as redox reactions.5-7 Lectins are structurally diverse proteins or glycoproteins that selectively and reversibly associate with mono- and oligosaccharide components of polysaccharide structures.8 Lectins are found in most organisms, ranging from microorganisms to plants and mammals, and are invaluable tools for the structural and functional investigation of complex carbohydrates.9,10 The lectin concanavalin A (Con A) has been used in a reversible immobilization strategy for glycoenzymes such as glucose oxidase and peroxidase in a FET-based biosensor for glucose.11 A competitive sensor for glucose has also been reported in which a fluorescent Con A derivative and a quenching dextran derivative are trapped in a hydrogel at the distal tip of an optical fiber, to allow free entry of glucose from the external solution.12 (2) Swaminathan, B.; Feng, P. Annu. Rev. Microbiol. 1994, 48, 401-426. (3) Hobsen, N. S.; Tothill, I. E.; Turner, A. F. P. Biosens. Bioelectron. 1996, 11, 455-477. (4) Cunningham, A. J. Introduction to Bioanalytical Sensors; John Wiley & Sons: New York, 1998. (5) Ivinski, D.; Abdel-Hamid, I.; Atanasov, P.; Wilkins, E. Biosens. Bioelectron. 1999, 14, 599-624. (6) Wood, K. V.; Gruber, M. G. Biosens. Bioelectron. 1996, 11, 207-214. (7) Mouritsen, C. L.; Hillyard, D. R. Anal. Chem. 1999, 71, 293R- 366R. (8) Sharon, N.; Lis, H. Lectins; Chapman and Hall: New York, 1984. (9) Liener, I. E., Sharon, N., Goldstein, I. J., Eds. The Lectins: Properties, Functions and Applications in Biology and Medicine, Academic Press: Orlando, FL, 1986. (10) Dwek, R. A. Chem. Rev. 1996, 96, 683-721. (11) Koeneke, R.; Menzel, C.; Ulber, R.; Schuegerl, K.; Scheper, T. Biosens. Bioelectron. 1996, 11, 1229-1236. (12) Russell, R. J.; Pishko, M. V.; Gefrides, C. C.; McShane, M. J.; Cote, G. L. Anal. Chem. 1999, 71, 1, 3126-32

Analytical Chemistry, Vol. 73, No. 17, September 1, 2001 4241

Bacteria and fungi have chemically distinct surface polysaccharide carbohydrate structures that can be recognized by lectins in agglutination studies.13 A recent report shows that Helicobacter pylori isolates can be differentiated based on their lectin agglutination patterns; 36 strains of this organism were grouped into 8 lectin reaction patterns, using an optimized panel of 5 lectins.14 Lectins have also been suggested for bacterial purification purposes due to their ability to selectively bind cell wall oligosaccharides and to retain microbial cells in affinity chromatography.13,15-17 Immobilized lectins have been incorporated into affinity surfaces used to isolate broad classes of bacterial samples for MALDI mass spectrometric analysis,18 and the same group has compared immobilized lectins with immobilized polysaccharides for affinity purification of microorganisms based on recognition of bacterial polysaccharides and surface-expressed lectins, respectively.19 In recent years, intensive research has been undertaken to develop portable, rapid, and sensitive biosensors capable of detecting microbes with high specificity and sensitivity. Bacterial detection strategies with biosensors have used biological recognition components such as receptors, nucleic acids, or antibodies, in intimate contact with an appropriate transducer.3,5,20 Although the concept of using lectins to recognize bacteria is not new,21,22 the application of lectins to bacterial identification in a biosensor array has never been reported. Furthermore, lectins are readily available and inexpensive recognition agents, well suited to exploitation in a sensor array targeted toward mass production. We have recently shown that the reduction of ferricyanide by viable Escherichia coli cells is particularly sensitive to growth conditions due to variations in the expression levels of enzymes such as the terminal cytochrome oxidases and the primary succinate dehydrogenase in the bacterial cell wall. In the presence of the respiratory substrates succinate and formate as well as the redox mediator ferricyanide, the respiratory cycle activity of a cell suspension can be measured electrochemically by chronocoulometry.23 We have also found that brief incubations of microorganisms with effective antibiotic compounds cause dramatic decreases in measured respiratory cycle activities, allowing rapid assessment of antibiotic drug susceptibility.24 In this study, lectins are employed as selective recognition elements in a sensor array that uses electrochemical signals from respiratory cycle activity to identify bacterial strains. Lectins were immobilized onto various membrane surfaces by adsorption with and without intermolecular cross-linking, through avidin-biotin (13) Mirelman, D.; Ofek, I. In Microbial Lectins and Agglutinins, Properties and Biological Activity; Mirelman, D., Ed.; John Wiley & Sons: New York, 1986. (14) Hynes, S. O.; Hirmo, S.; Wadstroem, T.; Moran, A. P. J. Clin. Microbiol. 1999, 37, 1994-1998. (15) Patchett, R. A.; Kelly, A. F.; Kroll, R. G. J. Appl. Bacteriol. 1991, 71, 277284. (16) Lis, H.; Sharon, N. Chem. Rev. 1998, 98, 637-675. (17) Oldenberg, K. R.; Loganathan, D.; Goldstein, I. J.; Schultz, P. G.; Gallop, M. A. Proc. Natl. Acad. Sci. U.S.A. 1992, 89, 5393. (18) Bundy, J. L.; Fenselau, C. Anal. Chem. 1999, 71, 1460-1463 (19) Bundy, J. L.; Fenselau, C. Anal. Chem. 2001, 73, 751-757. (20) Zhou, C.; Pivarnik, P.; Rand, G.; Letcher, S. V. Biosens. Bioelectron. 1998, 13, 495-500. (21) Payne, M. J.; Campell, S.; Kroll, R. G. J. Appl. Bacteriol. 1993, 74, 276283. (22) Pistole, T. G. Annu. Rev. Microbiol. 1981, 35, 85-112. (23) Ertl, P.; Unterladstaetter, B.; Bayer, K.; Mikkelsen, S. R. Anal. Chem. 2000, 72, 4949-4956. (24) Ertl, P.; Robello, E.; Battaglini, F.; Mikkelsen, S. R. Anal. Chem. 2000, 72, 4957-4964.

4242

Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

anchors using biotinylated lectins and by covalent coupling to activated membrane surfaces. Lectin-modified membranes were incubated in pure suspensions of microorganisms to allow selective cell attachment and were then rinsed and fixed at the surface of a Pt electrode. Chronocoulometric measurements were made in an assay buffer containing succinate, formate, and the redox mediators ferricyanide and menadione, for an array of 10 lectinmodified membranes, with each of 6 microorganisms. Factorbased principal component analysis (PCA)25-28 was used to analyze the chronocoulometric data. The results show clear groupings of replicate measurements for the six microorganisms in principal component plots, suggesting that rapid bacterial identification is possible using an array of lectin-modified electrodes. EXPERIMENTAL SECTION Materials and Instrumentation. Sigma supplied lectins from the species Artocarpus integrifolia, Arachis hypogaea, Galanthus nivalis, Phytolacca americana, Lens culinaris, Helix pomatia, Triticum vulgaris, and Codium fragile as well as Con A and biotinylated Con A (97% biotinylated). Bovine serum albumin (BSA, 99%), streptavidin (14 units/mg of protein), avidin (12.9 units/mg of protein), biotinamidocaproyl-labeled peroxidase (240 units/mg of protein), 2′,2′-azinobis(3-ethylbenzthioxoloine-6 sulfuric acid) (ABTS), citric acid, sodium cyanoborohydride, N-(2hydroxyethyl)piperazine-N′-(2-ethanesulfonic acid) (HEPES) buffer, 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC), menadione (2-methyl-1,4-naphthoquinone sodium bisulfite), and glutaraldehyde (25% aqueous solution) were also obtained from Sigma. Hydrogen peroxide (30% solution) was supplied by BDH and stored at 4 °C. Pierce supplied the cross-linking agents maleimidobenzyl-N-hydroxysulfosuccinimide ester (sulfo-MBS), dimethyl adipimidate-2 hydrochloride (DMA), and bis(sulfosuccinimidyl) suberate (BS3). Dialysis membrane (6-8000 kDa MWCO) was obtained from Fisher Scientific. Gibco BRL supplied nitrocellulose, supported nitrocellulose, neutral nylon (Biodyne A), and the positively charged nylon (Biodyne B) membranes. Pall Specialty Materials supplied the preactivated membranes ImmunodyneABC and UltraBind; both of these membranes were specified with a pore diameter of 0.45 µm. Bioanalytical Systems supplied the platinum disk working electrodes (1.8-mm diameter) and Ag/AgCl reference electrodes. Silver wire (1.0-mm diameter, 99.99%) was purchased from Aldrich. The microbial strains Saccharomyces cerevisiae ATCC9896, Bacillus cereus, Staphylococcus aureus ATCC6538P, Enterobacter aerogenes ATCC13048e, and Proteus vulgaris ATCC6380 were obtained from the strain collection of the Department of Biology, University of Waterloo. E. coli JM105 was obtained from Prof. G. Guillemette, Department of Chemistry, University of Waterloo. METHODS Cultivation of Microorganisms. All strains were cultivated under the same conditions using a growth medium with the (25) Jurs, P. C.; Bakken, G. A.; McClelland, H. E. Chem. Rev. 2000, 100, 26942678. (26) Kramer, R. Chemometric Techniques for Quantitative Analysis; Marcel Dekker: New York, 1998. (27) Schonkopf, S. Am. Lab. 1999, 31, 32-36. (28) Di Natale, C.; Davide, F. A. M.; D’Amico, A.; Sberveglieri, G.; Nelli, P.; Faglia, G.; Perego, C. Sens. Actuators, B 1995, 24-25, 801-805.

following components: KH2PO4 (2.88 g/L), K2HPO4‚3H2O (5.76 g/L), tryptone (2.4 g/L), yeast extract (1.2 g/L), trisodium citrate dihydrate (1.2 g/L), MgSO4‚7H2O (0.48 g/L), CaCl2‚2H2O (0.048 g/L), (NH4)2SO4 (1.63 g/L), NH4Cl (1.34 g/L), glucose (13.2 g/L), and 240 µL of the trace element stock solution per liter of medium. The trace element stock solution was prepared in 5 N HCl (Merck) and contained the following compounds: FeSO4‚7H2O (40 g/L), MnSO4‚H2O (10 g/L), AlCl3‚6H2O (10 g/L), CoCl2‚6H2O (4 g/L), ZnSO4‚7H2O (2 g/L), Na2MoO4‚2H2O (10 g/L), CuCl2‚2H2O (1 g/L), and H3BO3 (0.5 g/L). Glucose was prepared as a concentrated aqueous solution, sterilized separately, and added to the growth medium prior to inoculation. Cultivations were performed by adding 1 mL of 1:1 glycerol/exponential-phase cell culture mixture (previously stored at -80 °C) to 50 mL of growth medium in a shake flask. Growth proceeded for a maximum of 10 h in an incubator-shaker (200 rpm, 37 °C). Cell culture samples were centrifuged (5000 rpm, 5 min), resuspended in buffer consisting of growth medium without glucose, proteins, or trace elements but containing 10 mM succinate, and stored on ice. Agglutination Tests. An aliquot (50 µL) of the centrifuged and resuspended cell sample was combined with 150 µL of the buffered lectin stock solution containing 200 µg of protein/mL of buffer. Different microtiter wells were used to compare rows containing lectin/bacterial suspension samples to controls, where buffer was combined with the cell suspension. All lectins were investigated for their ability to form visible agglutinin after 6-h incubation at room temperature. The extent of the visible formation of agglutinin was used to classify binding patterns. Immobilization Methods. Adsorption. Lectin, avidin, and streptavidin solutions (100 µg/mL) were prepared in a buffer consisting of growth medium, as above, without glucose, proteins, or trace elements. Precut membrane disks (0.28 cm2) were immersed in 300 µL of these solutions and were incubated for 1 h at room temperature (22 ( 2 °C). Just prior to use, disks were removed from the protein solutions and rinsed with buffer. Cross-Linking. Unless otherwise noted, cross-linking agents were dissolved in 50 mM HEPES (pH 7.5-8.5) containing 5 mM MgSO4 to final concentrations of 5 (MBS and BS3) or 10 mM (DMA and EDC). Glutaraldehyde solutions (25% aqueous) were used as received. Membrane disks with adsorbed proteins were immersed in 300 µL of the cross-linking solution and allowed to react at room temperature (22 °C) for 30 min. An additional 15 min was used for glutaraldehyde cross-linking; this was done following addition of sodium cyanoborohydride to a final concentration of 10 mM. The disks were then removed and rinsed with buffer consisting of growth medium, as above, without glucose, proteins, or trace elements. Preactivated Membranes. Precut disks (0.28 cm2) of Immunodyne ABC and UltraBind membranes, featuring nucleophileselective and aldehyde-activated surfaces, respectively, were immersed in 100 µg/mL lectin, avidin, or streptavidin solutions and allowed to react for 30 min at room temperature. The disks were removed from the protein solutions, rinsed with buffer, and stored in buffer in an ice/water bath. Peroxidase Activity Assay. Biotinylated peroxidase was used as a label to detect membrane-bound avidin and streptavidin following adsorptive immobilization. Modified membranes were transferred to aliquots of a biotinylated peroxidase stock solution

(300 µL, 20 µg/mL) and incubated for 1 h at room temperature. Excess unbound peroxidase was washed away with buffer and the modified membrane was transferred into the ABTS reagent solution. The enzymatic reaction was then started by adding 3 µL of 30% H2O2 to the assay solution. The accumulation of oxidized dye over time (2-6 min) was measured spectrophotometrically at 412 nm using buffer as a blank. The calculation of ABTS conversion (µM/min) was performed using Beer’s law (A ) bC) with a molar absorptivity () of 32 400 M-1 cm-1.29 Cell Capture on Lectin-Modified Membranes. Exponentialphase bacteria were harvested, centrifuged at 5000 rpm (3000g), resuspended in buffer containing 10 mM succinate, and stored on ice. The lectin-modified membranes were added to the bacterial suspension and incubated for 100 min on ice (unless otherwise indicated). Membrane disks were then removed from the cell suspensions, rinsed with buffer, and affixed to the surfaces of the Pt working electrodes using nylon netting and rubber O-rings. Chronocoulometry. Unless otherwise noted, membranemodified electrodes and Ag/AgCl counter electrodes were immersed into an electrochemical cell containing 200 µL of the reagent solution (50 mM ferricyanide, 10 mM formate, 10 mM succinate, and 0.1 mM menadione in growth medium lacking glucose, proteins, or trace elements). After an incubation period of 2 min at 37 °C, the accumulation of ferrocyanide was detected by means of chronocoulometry. The potential of the Pt working electrode was set at + 0.50 V versus Ag/AgCl and the resulting current was integrated over 800 s to yield a plot of total charge against time. The difference in total charge between 200 and 800 s was recorded as the analytical signal, unless otherwise indicated. Chemometric Data Analysis. Normalized chronocoulometric data obtained for three replicate runs with each cell culture were used to generate a matrix for (PCA. One column in the data matrix consisted of the chronocoulometric signals for each of the 10 lectin-modified membranes, plus controls (membranes modified with BSA and unmodified membranes). Thus, three columns were generated for each microorganism. The matrix was converted into spreadsheet format using Microsoft Excel and was then converted into a Lotus file for incorporation into MATLAB programs. Factor analysis was performed using the Chemometrics Toolbox of MATLAB (Version 2.3, The MathWorks, Natick, MA, 1998) and involved the generation of reduced eigenvectors to determine the optimal number of factors, examination of the resulting residuals plots for randomness, and the generation of scores for the first three principal components. These scores were obtained for each of the three replicate trials for each microorganism and were plotted to determine whether qualitative groupings of microorganisms existed in the lectin-binding/chronocoulometric responses. RESULTS AND DISCUSSION Agglutination tests were performed using 10 lectins, to determine whether they interact selectively enough with outer membrane lipopolysaccharides of the six microorganism species to allow discrimination between cultures based on the formation of visible agglutinin after a 6-h incubation. Lectins for this study were chosen based on their differing known mono- and oligosaccharide binding selectivities.9 Negative controls (without lectin) were run alongside the lectin tests for each species, to ensure (29) Makinen, K. K.; Tenovuo, J. Anal. Biochem. 1982, 126, 100-108.

Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

4243

Table 1. Agglutination Test Results for the Reactions of Lectins with Microorganismsa agglutination observedb lectin

sugar specificity

E. coli

E. aerogenes

P. vulgaris

B. cereus

S. aureus

S. cerevisiae

Con A S. tuberosum L. culinaris A.integrifolia H. pomatia T. vulgaris A. hypogaea C. fragile P. americana G. nivalis

R-man, R-GLC (glcNAc)3 R-man R-gal galNAc (glcNAc)3, NeuNAc β-gal, galNAc galNAc (glcNAc)3 nonred R-man

+++ + + + + ++ + -

+++ +++ ++ ++ + +++ +++ +++ +++

++ ++ +++ +++ ++ ++ +

+ + + + -

++ + + + ++ + +++ ++ +++ ++

+ + + + + +

a Conditions: 50-µL aliquots of exponential-phase bacterial samples were combined with 150 µL of lectin solution (200 µg/mL) and incubated for 6 h at room temperature. b Key: (+++) strong, (++) medium, (+) low, and (-) no agglutination

the absence of autoagglutination.14 The results, shown in Table 1, indicate a range of selectivities, with each lectin causing no agglutination (-) of some cultures to strong agglutination (+++) of others. No identical agglutination patterns were observed with this group of lectins; however, similar patterns exist for S. tuberosum and C. fragile, A. integrifolia and H. pomatia, H. pomatia and C. fragile, and P. americana and G. nivalis, with only two minor differences in agglutination abilities noted for each pair. Similarities in the agglutination patterns of Solanum tuberosum, A. integrifolia, H. pomatia, and C. fragile lectins may be attributed to their known selectivities toward galactose or N-acetyl moieties, but the similar pattern observed for P. americana and G. nivalis lectins would not be predicted from their selectivities toward N-acetylglucosamine and nonreducing mannose, respectively. Possibly these two lipopolysaccharide components are coexpressed in the species studied. Table 1 also shows that two of the microorganisms studied, B. cereus and S. cerevisiae, yield only weak agglutination with about half of the lectins in this group. Despite the lack of strong agglutination, the agglutination patterns are sufficiently different to allow discrimination. One species, E. aerogenes, strongly agglutinates most of the lectins studied, suggesting the presence of diverse oligosaccharide structures in its outer membrane. Lectin immobilization studies were carried out using Con A, since it strongly agglutinates E. coli JM105, as shown in Table 1. Prior to Con A immobilization, preliminary experiments were conducted to select conditions for the detection of viable cells trapped at Pt electrode surfaces, using a dilution series of a resuspended, exponential phase culture of E. coli (OD600 ) 3.51). Previous calibration studies with this organism23 allowed correlation of optical density at 600 nm with viable cell counts, or cfu values. Aliquots of cell suspension (5 µL) were placed on unmodified nitrocellulose membrane disks, and these were fixed on the electrode surface using dialysis membrane and a rubber O-ring. The difference in chronocoulometric charge between 200 and 800 s is plotted against viable cell number in Figure 1, where an approximately linear increase in signal can be seen with an increasing numbers of trapped cells. On the basis of the area of the membrane disks used in immobilization studies (0.28 cm2), and an estimated requirement of 1 µm2 for each surface-bound cell, a close-packed monolayer of viable E. coli on one side of the membrane would be expected to involve 2.8 × 107 cfu. Figure 1 shows that 2 orders of magnitude fewer cells are readily detected 4244 Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

Figure 1. Chronocoulometric signal as a function of the number of cells (cfu) trapped at the working electrode surface. Modified electrodes were incubated at 37 °C for 10 min in 200 µL of a reagent solution containing 50 mM ferricyanide, 0.10 mM menadione, 10 mM succinate, and 10 mM formate; chronocoulometry was then performed for 800 s at + 0.5 V (Ag/AgCl).

by mediated chronocoulometric measurement of respiratory cycle activity, when the cells are trapped at the electrode surface rather than bound to a membrane. Initial immobilization studies involved avidin and streptavidin,4,30,38 which were adsorbed onto nitrocellulose, supported nitrocellulose, Biodyne A and Biodyne B membranes, exposed to cross-linking agents to stabilize the immobilized protein layer30-37 and then incubated with biotinylated peroxidase. This enzyme was used as a label to compare the quantities of active avidin or streptavidin on the membranes: ABTS conversion rates were measured, and these values (Table 2) indicate the relative quantities of peroxidase attached to the membranes.29,39,40 (30) Tatsuma, T.; Watanabe, T. Anal. Chem. 1992, 64, 625-630. (31) Staros, J.; Wright, W.; Swingle, D. Anal. Biochem. 1986, 156, 220-222. (32) Braun, B.; Klein, E.; Lopez, J. L. Biotechnol. Bioeng., 1996, 51, 327-342. (33) Oswald, P. R.; Evans, R. A.; Henderson, W.; Daniel, R. M.; Fee, C. J. Enzymol. Microb. Technol. 1998, 23, 14-20. (34) Cochrane, F. C.; Petach, H. H.; Henderson, W. Enzymol. Microb. Technol. 1996, 18, 373-379. (35) Yoshitake, S.; Imagawa, M.; Ishikava, E. Anal. Lett. 1982, 15, 147. (36) Saurina, J.; Hernandez-Cassou, S.; Fabregas, E.; Alegret, S. Anal. Chim. Acta 1998, 371, 49-57. (37) Moody, J.; Sanghera, G. S.; Thomas, J. D. R. Analyst 1986, 111, 605-704. (38) Madras, M. B.; Buck, R. P. Anal. Chem. 1998, 98, 637-641. (39) Childs, R. E.; Bardsley, W. G. Biochem. J. 1975, 145, 93-103. (40) Shindler, J. S.; Childs, R. E.; Bardsley, W. G. Eur. J. Biochem. 1976, 64, 325-331.

Table 2. ABTS Conversion Rates for Modified Membranes Labeled with Biotinylated Peroxidase cross-linking agent

a

species adsorbed

Biodyne A

ABTS conversion rate, µM/min, 412 nma Biodyne B nitrocellulose

supported NC

glutaraldehyde EDC MBS DMA BS3

avidin avidin avidin avidin avidin

7.6 7.3 2.8 8.2 9.8

10.7 3.2 1.2 6.4 9.1

3.7 8.8 6.8 8.4 0.0

4.6 5.6 1.7 6.7 0.0

glutaraldehyde

streptavidin

5.8

6.4

1.2

0.6

Average of three replicate measurements, RSD < 10%.

With adsorbed avidin, the highest ABTS conversion rates (Table 2) were found following cross-linking with either glutaraldehyde or BS3. Signals obtained with streptavidin-glutaraldehyde membranes are significantly lower. Results obtained with the two nitrocellulose membranes suggest that avidin adsorption occurs but that deactivation results from cross-linking, especially with glutaraldehyde and BS3. With these membranes, EDC and DMA cross-linking procedures provide the highest remaining biotinbinding activity. Since supported nitrocellulose yielded the lowest ABTS conversion rates, it was not used in further experiments. Chronocoulometric measurements were used to determine which combination of membrane and immobilization chemistry yields the greatest signals following capture of E. coli JM105 by surface-bound Con A. In these experiments, two preactivated membrane materials were studied (Immunodyne ABC and UltraBind) in addition to Biodyne A, Biodyne B, and nitrocellulose. Following immobilization of Con A, avidin, or streptavidin as well as cross-linking and incubation with biotinylated Con A (where necessary), membranes were incubated in resuspended, exponential-phase E. coli at 0 °C for 60 min. Membranes were then rinsed and fixed to the surfaces of Pt working electrodes and incubated 10 min with reagent solution (as in Figure 1). Table 3 shows the results of these experiments, where the signal corresponds to the total charge accumulated between 200 and 800 s during the chronocoulometric run. With this 10-min integration period, results ranged from 1.2 to 44.6 µC and clearly illustrate the effect of immobilization chemistry on surface activity. The results shown in Table 3 mainly represent experiments where Con A was bound directly to the membrane surfaces, unlike those shown in Table 2, where avidin or streptavidin was immobilized. However, for comparison, Biodyne B was used to immobilize avidin and was then cross-linked with glutaraldehyde prior to biotin-Con A and E. coli exposure, since this method yielded the highest ABTS conversion rate for biotinylated peroxidase binding, as shown in Table 2. This immobilization method yielded a large chronocoulometric signal (30.6 µC), indicating good Con A immobilization and good E. coli binding, but still larger signals were obtained using the preactivated membranes, particularly Immunodyne ABC with direct immobilization of Con A (44.6 µC) and UltraBind with the streptavidin-biotin-Con A configuration (37.0 µC). Further experiments involved the direct binding of lectins to Immunodyne ABC membranes. Con A concentration was varied (5-500 µg/mL) during the 30-min Immunodyne immobilization step, and chronocoulometric signals obtained over 600 s were largest using 100 µg/mL Con A. These traces are shown in Figure 2, where a background trace

Table 3. Chronocoulometric Signals Obtained Following E. coli JM105 Binding to Con A-Modified Membranes immobilized species

additional membrane treatment

charge, µCa

nitrocellulose

Con A

Biodyne A

Con A

Biodyne B

Con A

9.9 ( 4.3 4.3 ( 0.8 2.4 ( 0.1 17.7 ( 0.4 17.3 ( 1.5 4.6 ( 0.8 1.2 ( 0.4 2.2 ( 0.1 17.4 ( 2.4 7.8 ( 2.0 9.3 ( 3.3 5.3 ( 1.9 3.4 ( 0.2 16.7 ( 0.8 13.0 ( 3.0 30.6 ( 3.5

17.4 ( 2.6 16.7 ( 1.6 21.0 ( 5.0 18.0 ( 4.8 37.0 ( 4.5

membrane

Immunodyne ABC

Con A

none EDC BS3 MBS DMA none EDC BS3 MBS DMA none EDC BS3 MBS DMA glutaraldehyde, biotinylated Con A none

UltraBind

avidin streptavidin Con A avidin streptavidin

biotinylated Con A biotinylated Con A none biotinylated Con A biotinylated Con A

avidin

44.6 ( 6.5

a Values represent the average difference in charge measured between 200 and 800 s of each chronocoulometric run. Uncertainty is one standard deviation.

obtained with an unmodified membrane that was not exposed to E. coli is also shown. Subsequent immobilizations were performed using 100 µg/mL protein. The effect of E. coli incubation time was studied using Immunodyne ABC membranes that had been modified with each of the 10 lectins as well as BSA, which was included as a control to deactivate reactive groups on the membrane surface (agglutination tests with BSA and E. coli were negative). Using these membranes, chronocoulometric measurements were made as a function of their time of exposure to resuspended, exponentialphase E. coli, for a minimum of 3 h. Results for three representative lectin-modified and control (unmodified) membranes are shown in Figure 3. Six of the 10 lectins (Con A, S. tuberosum, L. culinaris, A. hypogaea, G. nivalis and P. americana) as well as BSA caused increasing chronocoulometric signals as a function of E. coli capture time; the L. culinaris response is shown in Figure 3. Three lectins (H. pomatia, T. vulgaris, C. fragile) produced signals that increased to a maximum and then declined at longer Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

4245

Figure 2. Chronocoulometric traces obtained for E. coli JM105 captured on Immunodyne ABC membranes modified using Con A concentrations of (a) 0, (b) 50, (c) 100, (d) 300, and (e) 500 µg/mL.

Figure 3. Chronocoulometric signals as a function of cell capture time for E. coli JM105 on unmodified (b) and lectin-modified Immunodyne ABC membranes. Lectins are (∆) L. culinaris, ([) H. pomatia, and (1) A. integrifolia. Chronocoulometry was performed at 37 oC and a run time of 800 s.

E. coli incubation times; the H. pomatia response is shown in Figure 3. One lectin (A. integrifolia) yielded a slowly decreasing chronocoulometric signal. The unmodified membrane yielded reasonably large signals (40 µC), but these did not vary significantly over more than 2-h incubation with E. coli. From these results, an incubation time of 100 min was selected for further studies. Finally, the 10 lectin-modified Immunodyne membranes along with the BSA-modified and an unmodified membrane were used in a model 12-element array with each of the 6 microorganisms (E. coli, E. aerogenes, P. vulgaris, S. aureus, B. cereus, S. cerevisiae). These were grown under identical conditions, harvested in the exponential phase (OD600 between 2.5 and 3.5), centrifuged and resuspended in buffer, and stored on ice. Membranes were incubated individually in aliquots (300 µL) of the cell suspension in microtiter wells for 100 min at 0 °C. Membranes were then fixed onto Pt electrode surfaces, and the modified electrodes were incubated at 37 °C for 30 min in 200 µL of the assay solution prior to an 800-s chronocoulometric measurement at +0.5 V (Ag/AgCl). Three replicate measurements of respiratory cycle activity were made with each microorganism/membrane combination. 4246 Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

Table 4 shows the background-subtracted chronocoulometric results, as the difference in total charge consumed between 200 and 800 s, for each replicate of the 72 microorganism-membrane combinations. The range of the chronocoulometric signals is large: after background subtraction, values ranged from 0 to 389 µC. Within most of the replicate groupings, significant variation between individual measurements can be seen. However, overall agreement was observed between these results and the results of agglutination tests (Table 1). With this group of lectins, the strongest agglutination was observed with E. aerogenes, P. vulgaris, and S. aureus (see Table 1), and these organisms also yielded the highest chronocoulometric signals at the electrochemical lectin array (Table 4). Similarly, the organisms that showed only weak agglutination with these lectins, B. cereus and S. cerevisiae, gave the lowest electrochemical signals. Occasional noncorrespondence of agglutination and electrochemical results occurred. For example, the very large electrochemical signals observed with P. vulgaris at a C. fragile lectinmodified membrane (Table 4) does not correspond to the agglutination test result (Table 1), which was negative for this combination of microorganism and lectin. Similarly, with E. aerogenes, reasonably strong signals were observed with Con A-modified membranes, while weak signals were seen with S. tuberosum (Table 4); in contrast, agglutination test results were negative for this organism with Con A and strongly positive with S. tuberosum (Table 1). We have attributed these discrepancies to changes in lectin structure that may be expected to occur upon covalent immobilization on the Immunodyne ABC membranes. The large chronocoulometric signals obtained with some of the organisms at unmodified and BSA-modified Immunodyne ABC membranes were unexpected, since these membranes were included as controls for which significant cell binding was expected to be minimal. It is possible that unmodified membranes, possessing aldehyde functional groups, react with components of the outer membranes of the cells to affect covalent attachment of cells to the membrane surfaces. Although this was not evident with S. cerevisiae, large chronocoulometric signals were observed with E. aerogenes, P. vulgaris, and S. aureus (Table 4). BSAmodified membranes, which were intended as deactivated controls, showed large chronocoulometric signals for E. coli, P. vulgaris, and S. aureus; this has been attributed to hydrophobic, noncovalent interaction of the cells with the immobilized protein. Chemometric PCA was applied to the chronocoulometric results shown in Table 4, in an attempt to apply pattern recognition methods41 to classify the response patterns according to species. Individual measurements were incorporated as a 12-element column into the data matrix, so that each column contained one measurement with each of the 12 membranes for one microorganism. Thus, three replicate columns existed for each microorganism, with each column corresponding to one complete array measurement. PCA yielded three factors containing more than 90% of the total variance across the matrix: PC 1 (76%), PC 2 (6%), and PC 3 (5%). Scores for each column were computed for the first two principal components, and these are plotted in Figure 4. In this scores plot, each point represents 12 chronocoulometric experiments (10 different lectins, BSA, and control) made with (41) Malinkowski, E. R. Factor Analysis in Chemistry, 2nd ed.; Wiley-Interscience: New York, 1991.

Table 4. Chronocoulometric Signals Obtained after Selective Cell Capture on Lectin-Modified Electrodes with Six Microbial Strains modified electrodes Con A S. tuberosum L. culinaris A. integrifolia H. pomatia T. vulgaris A. hypogaea C. fragile P. americana G. nivalis control BSA

E. coli

E. aerogenes

P. vulgaris

33 30 31 16 13 15 29 46 25 32 13 8 25 13 19 52 32 42 19 12 15 27 14 20 32 26 26 15 19 11 77 8 74 138 271 18

188 109 101 34 37 36 139 151 106 116 144 223 139 176 159 116 230 173 160 113 104 77 61 110 92 157 125 44 159 83 116 133 143 35 35 35

203 124 148 213 95 204 182 253 163 210 257 102 132 49 65 68 46 57 209 118 310 268 389 253 177 196 105 128 237 197 172 113 110 243 66 122

charge, µCa B. cererus 0 0 0 0 0 0 16 23 18 40 73 23 17 0 20 6 1 2 23 52 38 16 37 20 0 1 0 11 8 3 16 16 10 8 18 15

S. aureus

S. cerevisiae

139 248 193 151 69 177 72 98 102 35 97 86 128 130 143 100 113 120 193 169 140 58 198 130 187 239 213 144 114 140 144 131 105 118 100 135

6 3 0 5 4 0 0 0 0 3 0 0 19 9 7 27 4 0 0 2 1 3 0 6 44 0 0 58 16 11 0 0 0 0 3 2

a Values are corrected for background observed with unmodified membranes that had not been exposed to microorganisms and represent the difference in charge consumed between 200 and 800 s of the chronocoulometric run.

Figure 4. Pattern recognition plot obtained using data shown in Table 4. Each point represents a column of 12 individual measurements (10 lectins, BSA and control membranes) with (X) E. coli, (b) S. aureus, (0) S. cerevisiae, (2) B. cereus, (+) P. vulgaris, and (∇) E. aerogenes.

each microorganism. In Figure 4, replicate array measurements have been given the same symbol and have been circled to show the groupings of replicate measurements that resulted from PCA. Figure 4 shows that the array results for all six microbial strains form subpopulations that can be distinguished from one another.

Variances between microbial strains appear to be high enough with this two-dimensional plot to allow classification, while individual replicates are similar enough to cluster together. The third principal component, which contains variance similar to the second, can be used in a three-dimensional plot (not shown), which allows further discrimination of E. coli, B. cereus, and S. cerevisiae, which are the three organisms grouped on the left side of Figure 4. This work demonstrates the feasibility of lectin arrays used in conjunction with electrochemical methods for microorganism identification in pure cultures. Development of the arrays for practical identification would require the application of computational methods, such as the calculation of Euclidean distances and the determination of confidence intervals in these values, and would require a much larger data set to clearly identify the volume occupied by each organism on the three-dimensional plot.42 Applications of lectin arrays to mixtures of microorganisms may also be possible, if large lectin arrays are used in conjunction with computational methods such as the generalized rank annihilation (42) Massart, B.; Guo, Q.; Questier, F.; Massart, D.; Boucon, C.; de Jong, S.; Vandeginste, B. G. M. Trends Anal. Chem. 2001, 20, 35-41.

Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

4247

method (GRAM)43 to distinguish (and remove) signals from individual species. Further work is underway to achieve this goal. CONCLUSIONS Membrane-bound lectins were used as recognition agents in a sensor array for the identification of microorganisms. The selective, but not specific, binding of lectins to oligosaccharide residues present on the exterior surface of the microorganisms was exploited to generate characteristic patterns of binding of the microorganisms to the different lectin-modified membranes. Bound microorganisms were detected electrochemically, by mediated chronocoulometric measurement of respiratory cycle activity, so that only viable cells bound to the membranes were detected. (43) Sanchez, E.; Kowalski, B. R. Anal. Chem. 1986, 58, 496-499.

4248

Analytical Chemistry, Vol. 73, No. 17, September 1, 2001

Principal component analysis of the array measurements showed clear groupings of six microorganisms and will be pursued as a novel method of microorganism species identification. ACKNOWLEDGMENT Financial support from NSERC (Canada) and an International Scholarship from the Austrian Ministry of Science (P.E.) are gratefully acknowledged.

Received for review March 19, 2001. Accepted June 19, 2001. AC010324L