Anal. Chem. 2006, 78, 2500-2506
Mid-IR Biosensor: Detection and Fingerprinting of Pathogens on Gold Island Functionalized Chalcogenide Films Chenxu Yu,† Ashtosh Ganjoo,‡ H. Jain,‡ C. G. Pantano,§ and Joseph Irudayaraj*,†
Department of Agricultural and Biological Engineering, Purdue University, West Lafayette, Indiana 47907, Center for Optical Technologies, Lehigh University, Bethlehem, Pennsylvania 18015, and Material Research Institute, Pennsylvania State University, State College, Pennsylvania 16802
Multicomponent arrays of immobilized biomolecules are rapidly becoming essential screening tools to study biointeractions.1 Array-based assay formats have been developed for the detection of biological targets both at the molecular level as in DNA/RNA2 and proteins3 and at the cellular level such as in bacteria.4 A majority of array-based assays currently employ fluorescent, enzymatic, or radiolabeled biomolecules for target recognition.1 Label-free detection is particularly advantageous for the detection of pathogenic bacteria in food or water samples that are exposed to potential contamination, since oftentimes it is almost impossible to label molecules that are to be detected before analysis. The mostly wide-used label-free technique is the surface plasmon resonance (SPR) biosensor, in which changes in optical
property (refractive index) caused by absorption/desorption of targets onto the array surface are measured. SPR-based methods have been used to detect DNA/RNA,5 peptides/proteins,6 carbohydrates,7 and bacteria.8 A more robust monitoring protocol could be achieved by mid-infrared biosensing. Among optical sensing schemes, mid-IR sensors are gaining attention because of their inherent molecular selectivity, which will allow qualitative and quantitative analysis of various chemical and biological species.9 Waveguides and fiber-optic materials transparent in the mid-IR spectral region offer access to fundamental vibrational fingerprint absorptions of organic molecules. The availability of mid-IR waveguides has helped convert benchtop-style optical instrument such as the FT-IR spectrometer into portable compact biological and biomedical optical sensors with higher sensitivity. For example, Afanasyeva et al.10 have used fiberoptic evanescent-wave sensor to help diagnose breast cancer. Bruch et al.11 have investigated various ways of utilizing mid-IR fiber-optic sensor in dermatology. Yu and Irudayaraj12 have demonstrated that bacterial targets can be identified and differentiated based on their IR signatures at both the species and the strain level. We hypothesize that mid-IR sensors, while providing characteristic vibrational fingerprints, could also perform the function of a biosensor and thus form the basis for a comprehensive label-free detection system. Chalcogenide glass in general has a high transmission in the infrared region of the spectrum, which makes them particularly attractive for the fabrication of active and passive IR devices. Although various III-V semiconducting materials have been used as infrared devices, various drawbacks, primarily related to the toxicity of the material and the cost of fabrication, limit their application. Amorphous chalcogenide on the other hand, especially Ge-containing chalcogenide are less toxic, can be fabricated by a
* Corresponding author. E-mail:
[email protected]. † Purdue University. ‡ Lehigh University. § Pennsylvania State University. (1) Miller, J. C.; Butler, E. B.; Teh, B. S.; Haab, B. B. Dis. Markers 2001, 17, 225-234. (2) Favis, R.; Day J. P.; Gerry, N. P.; Phelan, C.; Narod, S.; Barany, F. Nat. Biotechnol. 2000, 18 (5), 561-4. (3) Zhu, H.; Bilgin, M.; Bangham, R.; Hall, D.; Casamayor, A.; Bertone, P.; Lan, N.; Jansen, R.; Bidlingmaier, S.; Houfek, T.; Mitchell, T.; Miller, P.; Dean, R. A.; Gerstain, M.; Snyder, M. Science 2001, 293, 2101-2105. (4) Roth, S. B.; Jalava, J.; Ruuskanen, O.; Ruohola, A.; Nikkari, S. J. Clin. Microbiol. 2004, 42 (9), 4268-4274.
(5) Brockman, J. M.; Frutos, A. G.; Corn, R. M. J. Am. Chem. Soc. 1999, 121, 8044-8051. (6) Wegner, G. J.; Lee, H. J.; Corn, R. M. Anal. Chem. 2002, 74, 5161-5168. (7) Frutos, A. G.; Brockman, J. M.; Corn, R. M. Langmuir 2000, 16, 21922197. (8) Swaminathan, A.; Irudayaraj, J.; Ryan, T. Biosens. Bioelectron. 2005, 40 (1), 61-69. (9) Mizaikoff, B. Anal. Chem. 2003, 75, 258A. (10) Afanasyeva, N. I.; Kolyakov, S. F.; Artjushenko, S. G.; Sokolov, V. V.; Frank, G. A. Proc. SPIE 1998, 3250, 140. (11) Bruch, R.; Afanasyeva, N.; Sukuta, S.; Brooks, A.; Makhine, V.; Kolyakov, S. Proc. SPIE Eur. Ser. 1998, 3564, 42. (12) Yu, C.; Irudayaraj, J. Biopolymers 2005, 77 (6), 368-377.
Antibody (human IgG, anti-E. coli O157:H7, and antiSalmonella) complexes on the surface of IR-transparent Ge-containing chalcogenide glass films were formed via thiol chemistry on 20-nm-thick gold islands. As a first step, the protocol was validated by monitoring fluorescently tagged targets to validate binding. FT-IR spectroscopy confirmed that the coating of the films with 20-nm gold did not have a significant effect on the propagation and penetration of IR evanescent waves through the film. The films functionalized with anti-E. coli O157:H7 and anti-Salmonella antibodies were used to detect E. coli O157:H7 and S. enteriditis through label-free IR fingerprinting. Highly selective detection of bacterial targets was achieved at both the species (E. coli vs. S. enteriditis) and strain level (E. coli O157:H7 vs E. coli K12). A midinfrared approach could thus be used as a biosensor as well as a molecular fingerprinting tool.
2500 Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
10.1021/ac051828c CCC: $33.50
© 2006 American Chemical Society Published on Web 03/17/2006
Figure 1. Schematic of an IR sensor using the functionalized GeSbSe film.
less costly evaporation technique, and are easy to handle. Further, the unique photoinduced property of amorphous chalcogenide provides additional functionality for optical fabrication and performance, e.g., through photodarkening of the absorption coefficient and modification of the optical band gap and the refractive index by illumination.13 Optical elements for application in communications and data storage could thus be patterned in chalcogenide glasses using this effect.9 The photoinduced change in the refractive index of the illuminated region of a film was also exploited for selective writing of channels in a planar waveguide. This process is much simpler and cost-effective than the chemical etching process employed with III-V semiconductor waveguides. A mid-IR-based biosensor could conceptually take advantage of the unique optical properties of chalcogenide (Figure 1) and provide information of the chemical groups in the mid-IR region and biosensing via optical transduction when an element of change occurs. First, a reliable and fool-proof protocol needs to be developed to functionalize chalcogenide glass with a linker that contains the capturing agents. Although various functionalization/activation methods have been developed for regular glass,14 most of them (based on Si chemistry) do not work for chalcogenide glasses due to their specific chemical properties. In this work, a functionalization procedure was developed utilizing the gold-thiol chemistry,14 which can be reliably replicated to produce the capturing layer for chalcogenide films without affecting their optical properties. Second, we have demonstrated that detection of targets that bind to the functionalized receptors is both possible and obtainable through IR fingerprinting. In the proposed design, a thin layer of bacterial cells attached to the surface upon binding will produce weak IR absorbance spectra. It is essential that the spectra measured are of high quality in order to ensure reliable identification. By a careful selection the characteristic IR bands (IR markers) of bacterial cells, we have demonstrated that IR measurements could confirm the presence/absence of target bacterial cells with adequate sensitivity and selectivity. EXPERIMENTAL SECTION Materials. 3-Mercaptopropionic acid (MPA), 16-mercaptohexadecanoic acid (MHA), absolute ethanol, sodium acetate, (13) Ganjoo, A.; Yoshida, N.; Shimakawa, K. In Recent Research Developments in Applied Physics; Kawasaki, M., Ashgriz, N., Anthony, R., Eds.; Research Signpost,1998. (14) Schaeferling, M.; Kambhampati, D. In Protein Array Technology; Kambhampati, D., Ed.; Wiley-VCH Verlag GmbH&Co.: Weinheim, Germany, 2004; pp 11-36.
phosphorus buffer solution with Tween 20 (PBST, pH 7.4), phosphorus buffer solution (PBS, pH 7.4), sodium dodecyl sulfate, and ethanolamine were purchased from Sigma-Aldrich (St. Louis, MO). 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide hydrochloride (EDC), N-hydroxysuccinimide (NHS), ImmunoPure human IgG, and ImmunoPure goat anti-human IgG conjugated with rhodamine were purchased from Pierce Biotechnology (Rockford, IL). Purified anti-Escherichia coli O157:H7 MAbs (mouse) were purchased from Biodesign International (Saco, ME), and Bac Trace purified anti-Salmonella antibodies (Goat) were purchased from Kirkegaard and Perry Laboratories (Gaithersburg, MA). Water used for all experiments was 18.2 MΩ, filtered through a Barnstead Nanopure system. The chalcogenide glass waveguide sensor (Figure 1) consists of a Ge28Sb12Se60 chalcogenide core layer (thickness ∼4 µm) deposited on top of a GeSe2 clad layer (thickness ∼2 mm) on a clean Si substrate. A thin layer of gold (thickness ∼20 nm) was thermally evaporated, in a vacuum, on top of the core layer of a Ge28Sb12Se60 through a mask to contain the gold and nongold regions on top of the film. E. coli O157:H7, E. coli K12, and Salmonella enteriditis were obtained from the Gastroenteric Disease Center (GDC) at the Pennsylvania State University (University Park, PA). These species were cultured on soy agar plates for 24 h, and a single colony of each species was transferred onto 100-mL Luria-Bertani broth medium (5 g of yeast extract, 8 g of tryptone, and 5 g of NaCl in 500 mL of distilled water) at 35 °C and shaken at 100 rpm for 24 h. A 10-mL aliquot of each culture was centrifuged at 200g for 25 min, and the microbial cells were first collected and washed three times by distilled water to remove residual medium and resuspended in PBS for target binding experiments. Biofunctionalization of the Chalcogenide Glass Film Surface. Thin layers (∼20 nm) of gold islands were deposited on top of the core layer of the Ge28Sb12Se60 film. Anchor molecules (MPA and MHA) were then attached to the gold to form selfassembled monolayers (SAMs) to which antibodies could be attached. The anchor molecules with a SH tail binds to the gold surface with a COOH head to which NH2 groups of the capturing agents (i.e., antibodies) could react to form an amide bond, thus providing the needed biospecificity to the surface. Antibody molecules are large compared to the thiol compounds (MPA, MHA) used for SAM; hence, spatial blocking between antibodies might result in a nonevenly distributed antibody layer due to steric hindrance. A mixed SAM (mSAM) was thus used designed by mixing the two thiol compounds, MHA and MPA, in the ratio 1:4 (MHA/MPA) in ethanol and immersing the film for 24 h. mSAMs thus formed will have one MHA for every four MPA molecules on an average. The longer MHA will serve as an anchor for antibodies, and the shorter MPA will serve as spacer molecules to potentially minimize steric hindrance. The next step in functionalization is to attach appropriate receptors to the mSAM functionalized film surface. The film with mSAM was immersed into a freshly prepared mixture of 0.4 M EDC/0.1 M NHS for 7 min. Antibody (i.e., IgG) suspension at 0.1 mg/mL was prepared in sodium acetate (pH 5.5), and the films were immersed in the antibody suspension and incubated for 10 min. The films were then washed with PBST for 3 min to remove unbound antibodies. Reversibly bound antibodies were removed Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
2501
from the surface by rinsing with 10 mg/mL sodium dodecyl sulfate for 3 min. As a final step, the films were incubated in 1 M ethanolamine (adjusted to pH 8.5) for 5 min to block any unbound free COOH sites to minimize nonspecific binding by targets. Detection of Binding of Targets to the Functionalized Film by IR Fingerprinting. Measurement of the Bacterial Spectra. Bacterial cells collected by centrifugation of 10 mL of the cell cultures were transferred onto the surface of a gold-coated glass slide and slowly blow-dried by a stream of nitrogen gas. FT-IR measurements were conducted using a Digilab Excalibur FTS 6000 spectrometer fitted with a UMA 600 IR microscope (Digilab, Randolph, MA) with a liquid nitrogen-cooled mercury-cadmiumtelluride detector under a reflection acquisition mode. Ten spectra was measured at 10 different spots on the surface of each film sample and averaged. The measured spectra were then vectornormalized before analysis. Investigation of Specificity. Films functionalized with anti-E. coli O157:H7 antibodies and anti-Salmonella antibodies were incubated in 5 mL of a PBS suspension of E. coli O157:H7, E. coli K12, or S. enteriditis (108 cfu/mL) for individual measurements or a mixture consisting of 1 mL of each bacterial suspension in cocktail experiments to assess the specificity of mixed cultures for 5 min. The films were then removed from the suspension and washed with PBS for 3 min followed by a 3-min distilled water rinse. The films were then blow-dried by N2 and analyzed by FT-IR microspectrometry. Investigation of Sensitivity. Films functionalized with antiSalmonella antibodies were incubated in 5 mL of PBS suspensions of S. enteriditis at concentrations between 108 and 106 cfu/mL for 5 min and 103 cfu/mL for 5, 45, and 90 min to allow binding. The films were then washed as above and interrogated by FT-IR spectrometry using the same settings as in the specificity study to determine whether the methodology is sensitive enough to detect low numbers of bacterial cells. RESULTS Biofunctionalization of the Chalcogenide Film. The functionalization chemistry was first confirmed by tethering human IgG to capture goat anti-human IgG labeled with rhodamine 6 (R6G). The IgG-functionalized films were incubated in goat-anti human IgG suspension (0.1 mg/mL in PBS) for 5 min, washed with PBS for 3 min and distilled water for 3 min, and blow-dried with N2. An FT-IR microspectrometric investigation indicated that the amides I (1690-1650 cm-1), II (∼1550 cm-1), and III (13001220 cm-1) all fall in the region where atmospheric water vapor has strong absorbance. The weak bands in the fingerprint region could be attributed to very small amount of protein, indicating the presence of biomolecules on the film surface. Figure 2 shows the FT-IR spectrum of the film functionalized with human IgG. It is important to observe the peak around 1032 cm-1, which is due to the silicon substrate (as illustrated in Figure 1), implying that the thin gold layer does not prevent the evanescent wave from penetrating and reaching the silicon substrate, especially when the IR microscope objective is focused to observe from the top in the reflection mode. The observation of the silicon peak demonstrates that the thin gold layer (∼20 nm) does not inhibit the transmission of the IR waves and the standard functionalization protocol for gold could be used. Other peaks related to atmo2502 Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
Figure 2. FT-IR spectra of the functionalized chalcogenide film with human IgG.
spheric CO2 (∼2349 cm-1) and H2O vapor (1300 ∼ 1700 cm-1) had an overwhelming effect on the protein IR bands. As a first step, the integrity of the functionalization protocol was tested using fluorescence, which in subsequent experiments will be extended to label-free monitoring of both sensitivity and specificity of detection of select foodborne pathogens. Initial testing of the protocol to monitor a target capture was conducted using an Olympus BX-60 EPI-fluorescent microscopy equipped with a Hamamatsu Orca-100 camera (Olympus America Inc., Melville, NY). Figure 3 shows the binding results of R6G-labeled goat anti-human IgG to films treated with and without human IgGs. Figure 3a shows that the film was not treated with human IgG; hence no specific binding to R6G-labeled goat anti human IgG was observed other than a few anti IgG due to physisorption (nonspecific absorption). On the other hand, films treated with human IgG showed a significantly higher anti-IgG binding activity (Figure 3b). This observation positively confirmed that an IgG layer was successfully formed on the gold island containing chalcogenide films. Figure 4 demonstrates that direct physisorption of the fluorescently labeled target (anti IgG) to the naked and gold islandcontaining films was minimal, indicating that the target will only bind to the films containing the receptor molecule. Figure 5 further confirms that nonspecific absorption was minimum by comparing the gold island containing IgG tethered film (upper half) with the naked chalcogenide film (bottom half) for target (anti IgG) binding. Fluorescence was observed only on the gold islandcontaining films except for some minor nonspecific adsorption, demonstrating the validity of the biofunctionalization protocol. It should be noted that examination of fluorescently labeled molecules was done on IR films as protocol validation steps. Experiments below will fully utilize the IR capabilities of functionalized films for pathogen detection applications. To independently validate the functionalization protocol, an XPS analysis was conducted to assess the elemental composition on the sensor surface before and after the IgG attachment. Before an IgG attachment, only mSAMs was formed on the surface, consisting of C, H, and O; after IgG attachment, N should be expected. XPS results shown in Figure 6 clearly demonstrate the presence of nitrogen, suggesting that IgG was attached. Up to
Figure 3. Comparison of binding properties between films without (a) and with (b) human IgG conjugation.
Figure 4. Fluorescent signals from untreated surfaces: (a) chalcogenide surface and (b) gold surface.
Figure 5. Edge between IgG activated gold island and naked chalcogenide film.
five replications were done, and the XPS and FT-IR results consistently confirmed the presence of receptor and validated the assay, suggesting that the functionalization protocol is highly reproducible and can provide consistent film-sensor surface for subsequent pathogen detection. Detection of Bacteria by IR Fingerprinting. Specificity Study. Because of the strong interference due to atmospheric water vapor and carbon dioxide, the amide band absorbance due to protein was clouded (discussed previously). When the targets are bacterial cells, peaks associated with other key cellular constituents, such as DNA/RNA, polysaccharides, and lipids, could serve as potential markers for target identification, especially if they are not overwhelmed by water absorption peaks. Peaks observed (>1300 cm-1) in the regular FT-IR spectrum (in the attenuated total reflectance mode using the standard sampling accessory) of the
respective organisms shown in Figure 7 could be used as markers for label-free biosensing through IR fingerprinting using chalcogenide films. Consequently, the spectra in Figure 7 will be compared with the spectra of the organisms bound to the receptor when actual films are used. Figure 8 shows the spectra of the film functionalized with antiE. coli O157:H7 antibodies and treated with E. coli O157:H7, E. coli K12, and S. enteriditis suspended in PBS buffer, in the region between 800 and 1300 cm-1. Distinct IR bands from E. coli O157: H7 and anti-E. coli O157:H7 receptor tethered film were observed, while the spectrum of the anti-E. coli O157:H7 tethered film surface treated with the other two organisms was similar to buffertreated blank films, thus confirming specificity. Evanescent waves can penetrate ∼3 µm due to attenuated total reflectance. If microbial cells, which have a dimension of ∼3 µm, Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
2503
Figure 6. Elemental analysis of the films for nitrogen by XPS without IgG. Figure 8. Mid-IR spectra of anti-E. coli O157:H7 antibody functionalized films treated with different target bacteria
Figure 7. FT-IR mid-infrared spectra of the three selected bacteria using a standard ATR accessory. Figure 9. Comparison of the spectra of film treated with E. coli O157:H7 to the spectra of bacteria.
are to be captured on the surface due to binding specificity, a strong absorbance due to the cells should be noted. Results of anti-E. coli O157:H7 tethered film treated with these three microorganisms were compared with the spectra of these organisms obtained independently using the standard ATR accessory (Figure 7). From Figures 7 and 8, it can be clearly seen that the spectra of films treated with E. coli K12 and S. enteriditis showed no similarity to the spectra of their respective bacteria; hence, the absence of both bacteria was confirmed. Figure 9 shows the comparison between the spectrum of film treated with E. coli O157:H7 and the spectrum of E. coli O157:H7. The PdO peak demonstrating symmetric stretching (∼1085 cm-1) and the PsO single bond from P-O-C (∼950 cm-1) denote contributions from DNA/RNA backbones.12,15,16 In the region below 900 cm-1, peaks associated with the “fingerprints” of nucleotides16 could be identified. These observations confirm that E. coli O157:H7 cells (15) Naumann, D. In Encyclopedia of Analytical Chemistry; Meyers, R. A., Ed.; John Wiley & Sons Ltd.: Chichester, U.K., 2000; pp 102-131. (16) Hannah, R. W. In Course Notes on the Interpretation of Infrared and Raman Spectra; Mayo, A. W., Miller, F. A., Hannah, R. W., Eds.; John Wiley & Sons: Hoboken, NJ, 2004; pp 217-247.
2504 Analytical Chemistry, Vol. 78, No. 8, April 15, 2006
were bound to the functionalized film, demonstrating strainspecific selectivity and label-free detection of the target (E. coli O157:H7). Additional validation of the infrared detection protocol was demonstrated using anti-Salmonella antibodies and was tested against the three bacteria as well as cocktail mixtures. Since the film was functionalized with anti-Salmonella antibodies, E. coli samples (K12 and O157:H7) are nontargets; consequently, only Si band at ∼1032 cm-1 could be observed, suggesting that there was no binding to the Salmonella antibody-containing films (Figure 10). Results from a more complex experiment containing microorganism cocktail (mixtures of the three bacterial cells) also confirms the specificity of detection as demonstrated in Figure 10. Samples that contained no S. enteriditis yielded a spectrum almost identical to blank films, while the cocktail with S. enteriditis yielded a spectrum similar to that obtained for S. enteriditis. Spectral fingerprints depicted in Figure 11 show PdO symmetric stretching (∼1085 cm-1), PsO single bond from P-O-C (∼950 cm-1), and the PdO asymmetric stretching band (∼1240 cm-1)12,15
the other hand, more specific spectral signatures were acquired after 45 and 90 min of incubation, suggesting that, given adequate incubation time (45 min or longer), the sensor surface with a specific receptor could attract adequate amount of its corresponding target (S. enteriditis cells in this experiment) and further extension of the incubation time (90 min) was not necessary. Figure 12b further shows that the IR fingerprint of the films exposed to different S. enteriditis concentrations was not a determining factor of sensitivity, rather the incubation time was found to be more critical. Hence, optimization of incubation time is essential in a practical setting.
Figure 10. Mid-IR spectra of anti-Salmonella antibody functionalized films treated with different target bacteria and bacterial cocktails.
Figure 11. Comparison of the spectra of film treated with S. enteriditis to the spectra of bacteria.
relevant to the bound molecules. A sensing methodology where a target-receptor concept is complemented by molecular fingerprinting is clearly evidenced in this experiment. Sensitivity Study. S. enteriditis suspensions were prepared at three different concentrations (108, 106, and 103 cfu/mL) through a series of dilutions. The cells that could bind to the sensor surface should ultimately be determined by the surface density of the viable antibodies, provided that the sensor surface is accessible to all of the cells in the suspension for a sufficient period of time to allow binding to occur. However, if the sensor surface is available to the target cells only for a short period of time, then only a small number of cells in the suspension are able to reach the sensor, and the strength of the spectral signals would be greatly affected. Our observations indicate that this uncertainty was more predominant at lower concentrations (∼103 cfu/mL). Spectral signatures measured off the sensor surface after incubating the anti-Salmonella containing film in a suspension of 103 cfu/mL for 5, 45, and 90 min is shown in Figure 12a. After 5 min, almost no meaningful signal was obtained, suggesting that only a very small number of cells were bound to the surface. On
DISCUSSION IR fingerprinting has been effectively used to identify and differentiate a variety of microorganisms.12,15,17-23 The differentiation relies on the inherent spectroscopic differences of cellular and cell membrane constituents due to proteins, DNA/RNA, polysaccharides, lipids, and other elements. It should be noted that the differences in cellular composition of different microbial cells are small, especially in terms of functional groups, i.e., sNH, sOH. PdO, which are the units identified by IR spectroscopy. To identify such delicate differences, the IR spectra must be of high quality. Statistical analysis is oftentimes necessary to extract the species/strain-specific spectroscopic signatures. Therefore, the numbers of microbial cells being analyzed cannot be too low. To date, the lowest concentration reported that can be differentiated by FT-IR spectroscopy was ∼103 cfu/mL.23 When the numbers are too low, distinct IR peaks could still be identified but not significant enough to differentiate microorganisms. Also, at very low concentration (
