In Situ Strain-Level Detection and Identification of Vibrio

Jan 28, 2013 - Vibrio parahaemolyticus, a widely distributed halophilic Gram-negative ... Different strains of V. parahaemolyticus bacteria can be rap...
0 downloads 0 Views 929KB Size
Download PDF
Article pubs.acs.org/ac

In Situ Strain-Level Detection and Identification of Vibrio parahaemolyticus Using Surface-Enhanced Raman Spectroscopy Jiajie Xu,†,∥ Jeffrey W. Turner,‡ Matthew Idso,† Stanley V. Biryukov,‡ Laurel Rognstad,⊥ Heng Gong,∥ Vera L. Trainer,‡ Mark L. Wells,§ Mark S. Strom,*,‡ and Qiuming Yu*,† †

Department of Chemical Engineering, University of Washington, Seattle, Washington 98195, United States Northwest Fisheries Science Center, National Marine Fisheries Service, National Oceanic and Atmospheric Administration, Seattle, Washington 98112, United States § School of Marine Sciences, University of Maine, Orono, Maine 04469, United States ∥ The State Key Laboratory of Bioreactor Engineering, East China University of Science and Technology, Shanghai 200237, People’s Republic of China ⊥ Department of Chemical Engineering, Tennessee Technological University, Cookeville, Tennessee 38505, United States ‡

S Supporting Information *

ABSTRACT: The outer membrane of a bacterium is composed of chemical and biological components that carry specific molecular information related to strains, growth stages, expressions to stimulation, and maybe even geographic differences. In this work, we demonstrate that the biochemical information embedded in the outer membrane can be used for rapid detection and identification of pathogenic bacteria using surface-enhanced Raman spectroscopy (SERS). We used seven different strains of the marine pathogen Vibrio parahaemolyticus as a model system. The strains represent four genetically distinct clades isolated from clinical and environmental sources in Washington, U.S.A. The unique quasi-3D (Q3D) plasmonic nanostructure arrays, optimized using finitedifference time-domain (FDTD) calculations, were used as SERS-active substrates for sensitive and reproducible detection of these bacteria. SERS barcodes were generated on the basis of SERS spectra and were used to successfully detect individual strains in both blind samples and mixtures. The sensing and detection methods developed in this work could have broad applications in the areas of environmental monitoring, biomedical diagnostics, and homeland security.

S

Colloidal gold or silver nanoparticles have been used to carry out SERS bacteria detections, either by mixing the nanoparticles with bacteria in solution or by forming a thin film of nanoparticles on a substrate for bacterial attachment. Solutionbased nanoparticle SERS detection has been widely used10−13 due to its simplicity, because nanoparticles readily adsorb to bacterial surfaces when mixed together. The Raman vibrational modes of the biochemical components in the outer membrane are greatly enhanced by direct contact with the nanoparticles. However, because the nanoparticle attachment to bacterial surfaces is not uniform due to the aggregation of nanoparticles on the surfaces, forming irregular “hot spots”, it is very difficult to obtain reproducible and quantitative results using nanoparticle mixtures. Substrate-based nanoparticle SERS detection of bacteria has also been explored. One approach has been to chemically attach presynthesized gold or silver nanoparticles to a substrate on which bacterial solutions are placed and then detected using SERS.14 However, gold nanoparticles still are

urface-enhanced Raman spectroscopy (SERS) has recently attracted considerable attention as a powerful analytical and sensing tool in many biological applications.1−8 Biosensors based on the SERS platform enable the direct detection and discrimination of bacteria and offer many advantages over current approaches.9 First, SERS provides an intrinsic profile of microorganisms without any external labeling. Therefore, pathogenic bacteria can be identified and discriminated by specific SERS spectral characteristic peaks. Raman spectroscopy is the inelastic scattering of photons by molecular bonds. Raman spectra in the 500−2000 cm−1 range contain rich biological information on nucleic acids, proteins, polysaccharides, carbohydrates, and lipids, as well as complex molecular assemblages. This abundant spectral information offers a signature of the molecular structures, cellular compositions, and physiological states of microorganisms. Second, multiplexed sensing can be easily realized by the simultaneous measurement of multiple bacterial strains due to narrow SERS spectral peaks. Third, the online analysis and in-field applications are diverse because these spectral measurements are rapid and reagentless and require little to no sample preparation. © 2013 American Chemical Society

Received: August 3, 2012 Accepted: January 28, 2013 Published: January 28, 2013 2630

dx.doi.org/10.1021/ac3021888 | Anal. Chem. 2013, 85, 2630−2637

Analytical Chemistry

Article

and specificity is highly desired for environmental monitoring, biomedical diagnostics, and homeland security.24−26 In the work presented here, we designed the Q3D plasmonic nanostructure arrays with the maximum local electric field at the top gold nanoholes using 3D-FDTD electromagnetic calculations and fabricated these Q3D nanopatterns on the indium tin oxide (ITO) coated glass substrates via EBL followed by metallization. We demonstrate that the optimal Q3D nanostructures can act as SERS-active substrates and provide sensitive and reproducible detection of bacteria using SERS. Different strains of V. parahaemolyticus bacteria can be rapidly characterized and detected on the basis of the SERS barcodes generated from the SERS spectra. In addition, the capabilities to be able to detect blind samples and mixtures are demonstrated, and the limit of detection has been qualitatively evaluated.

randomly distributed across the surface of the substrate using this method, so it is very difficult to obtain quantitative measurements. A major challenge to enable the transfer of SERS from the research lab to practical applications is the development of SERS substrates having high sensitivity and reproducibility. Recently, the engineered 2D gold nanoparticle cluster arrays have been developed and applied for detecting bacteria.15 Novel quasi-3D (Q3D) plasmonic nanostructures have also been developed for detecting bacteria.16 The Q3D plasmonic nanostructures are composed of physically separated gold thin film with subwavelength nanoholes on the top and gold nanodiscs at the bottom of wells fabricated via one-step electron beam lithography (EBL) followed by metal evaporation. Localized surface plasmons (LSPs) can be excited at both 2D nanoholes17,18 and 0D nanodots.19 By engineering different plasmonic elements in a single Q3D nanostructure, additional degrees of freedom can be achieved for tuning plasmonic properties and SERS. Since the SERS enhancement factor (EF) due to the electromagnetic enhancement can be estimated by using the relationship of the fourth power of the ratio of the local maximum electric field at the laser excitation wavelength to the incident electric field (|Emax(ωlaser)/E0|4), the optimal SERS nanostructures can be predicted by the electromagnetic finitedifference time-domain (FDTD) calculations and then verified by EBL fabricated nanostructures and SERS experiments.16 The strength of the local electric field and the location of the strongest electric field (i.e., “hot spots”) at either top or bottom gold layer can be tailored by varying the diameter of nanoholes,16 the depth, and the materials of the substrate and dielectric medium.20 This unique property of the Q3D plasmonic nanostructures has not been achieved on any other type of SERS substrate. The strongest local electric field can be specifically tuned to the air/gold interface at the top holes of Q3D plasmonic nanostructures, thereby enabling the sensitive and reproducible detection of bacteria because the bacteria cells are larger (e.g., ∼2−3 μm in length and ∼500 nm in diameter) than the nanohole size (∼400−500 nm). These Q3D plasmonic nanostructures have shown great advantages as SERS-active substrates over colloid gold or silver nanoparticles for detecting bacteria in terms of the sensitivity and reproducibility. We previously demonstrated that Q3D gold nanostructures tuned for sensitive detection of bacteria can discriminate one strain of Gram-positive bacteria from three strains of Gram-negative bacteria by one principle component in the principle component analysis (PCA), and three strains of Gram-negative bacteria can be differentiated by additional two principle components.16 Vibrio parahaemolyticus, a widely distributed halophilic Gramnegative bacterium, is the leading etiologic agent of seafoodborne bacterial gastroenteritis worldwide.21,22 Infection commonly follows the consumption of raw or undercooked bivalves (especially oysters), which can bioaccumulate pathogenic strains through filter feeding.23 A variety of methods have been developed for the detection and enumeration of potentially virulent strains including selective-differential media, biochemical and immunological assays, DNA probes, and polymerase chain reaction (PCR) assays. Regardless of the wide application of these detection methodologies, V. parahaemolyticus remains a significant concern of seafood safety and human health.21 The development of biosensors for the rapid detection of potentially pathogenic strains with high sensitivity

■

EXPERIMENTAL SECTION 3D-FDTD Calculations. The 3D-FDTD method27 was used to calculate the electric field profiles of Q3D gold nanostructure arrays fabricated on ITO coated glass substrates upon irradiation with a 785 nm laser.16,20 The optimal Q3D plasmonic nanostructure array with the strongest local electric field at the top gold nanoholes was determined for sensitive detection of V. parahaemolyticus bacteria. Simulation details are provided in the Supporting Information. Fabrication of SERS-Active Substrates. The Q3D gold nanostructure arrays with the dimensions determined by the 3D-FDTD calculations, shown in Figure 1, were fabricated via EBL using an FEI Sirion scanning electron microscope (SEM) with Nabity NPGS software. A 300 nm thick layer of polymethyl methacrylate (PMMA) electron-sensitive resist was spin-coated on an ITO coated glass substrate. The PMMA coated substrate was then exposed to an electron beam to create 50 μm by 50 μm nanohole arrays with a 400 nm hole diameter and 100 nm edge-to-edge distance. Four arrays were generated on each chip. Holes were generated after development in 1:3 methyl isobutyl ketone/isopropanol (MIBK/IPA) PMMA developer for 70 s followed by an IPA rinse and a postbake at 95 °C for 30 min. Evaporating a 50 nm thick gold film onto the patterned substrate completed the Q3D gold nanostructures. The lateral and vertical dimensions of nanostructures were characterized by an FEI Sirion SEM and tapping mode atomic force microscopy (AFM) using a Vecco Dimension 3100 AFM equipped with a Nanoscope IVa controller. Before use, the chips were cleaned in UV ozone for 20 min, rinsed with 18.2 MΩ·cm deionized water, and dried with nitrogen.28 Preparation of Bacterial Samples. Seven V. parahaemolyticus strains (50, 3355, 605, 551, 846, 12310, and 3256) were selected from a culture collection maintained at NOAA’s Northwest Fisheries Science Center (NWFSC), Seattle, WA (Table 1). Strains originated from clinical and environmental sources in Washington, U.S.A. and represent four genetically similar clades as determined by repetitive extragenic palindromic polymerase chain reaction (REP-PCR) and multilocus sequence typing (MLST)29 analysis (Table 1). Log-phase V. parahaemolyticus cultures were grown overnight (12−16 h) in tryptic soy broth (TSB) (1.7 g/L casein, 0.3 g/L peptone, 2.0 g/L NaCl, and 0.25 g/L phosphate) at 30 °C with shaking (150 rpm) (Cellstar, Model Q4730ABA, Queue Systems Inc., Asheville, NC, USA). The following day, exponential-phase cultures were achieved by inoculating 5 mL 2631

dx.doi.org/10.1021/ac3021888 | Anal. Chem. 2013, 85, 2630−2637

Analytical Chemistry

Article

sample. Exponential-phase cultures were standardized to an OD600 of 0.5 by dilution with the appropriate volume of sterile 0.5% NaCl (5 g/L), which corresponded to a final cell density of 108 cfu/mL according to a standard bacterial growth curve for the V. parahaemolyticus type strain (RIMD2210633) grown at identical conditions (data not shown). Cultures (1 mL) were then transferred to 1.5 mL microcentrifuge tubes and washed 3 times with 1 mL of 0.5% NaCl by centrifugation (2300g for 5 min at room temperature) using an Eppendorf microcentrifuge (Model 5415D, Hamburg, Germany). The washed cultures were then suspended in a 1 mL volume of 0.5% NaCl and vortexed. Standardized cell cultures were immediately analyzed by SERS. Acquisition of SERS Spectra. SERS spectra were collected using a Renishaw inVia Raman microspectroscope equipped with a Leica DMIRBE inverted optical microscope. As shown in Figure 2a, a SERS-active substrate was put into a well (1 cm × 1 cm opening and 5 mm depth) of a custom-made Teflon holder. After a 5 s vortex, one drop of 500 μL of the V. parahaemolyticus solution was added into the well, and a microscope cover slide (∼1 mm thick) was carefully laid over the Teflon well to avoid forming air bubbles between the substrate and the cover slide. The bacterial cells and the Q3D plasmonic nanostructure array can be seen under the dark field optical images (Figure 2b,c). A 785 nm laser was focused into a rectangular spot (2 μm × 25 μm) on one SERS-active pattern using a 50× (N.A. 0.8) Leica objective. The laser power after passing the objective was 5 mW. Each SERS spectrum was collected from 400 to 2000 cm−1 with the exposure time of 10 s and a single accumulation. The observed spectral resolution was 1.1 cm−1. A total of 8 SERS spectra were collected from different areas of each of four Q3D patterns, resulting in a total of 32 SERS spectra collected on one SERS-active chip (Figure 2d). Two SERS-active chips were used for collecting SERS spectra of each V. parahaemolyticus strain in order to evaluate the chip-to-chip reproducibility. Therefore, a total of 64 spectra were collected for each V. parahaemolyticus strain in this study. Generation of SERS Barcodes. Each SERS spectrum, consisting of 1260 data points, was baseline-corrected using cubic spline curve fitting based on 10 local minimum points of the original spectrum using the software of Renishaw WiRE 2.0. After normalizing each baseline-corrected spectrum by the largest spectral intensity to 1000, 64 spectra of each strain were input into Matlab30 as a 1260 × 64 matrix. The matrix was visualized as a SERS barcode using the code Surface in Matlab.31 The bar location, width, and color contrast of the SERS barcode represent the wavenumber shift, the width, and the intensity, respectively, of the vibrational bands of the SERS spectra.

Figure 1. (a) 3D illustration of the Q3D gold nanostructure array composed of a separated gold thin film with nanoholes on top and gold nanodiscs at the bottom of each well. The dimensions of the key features are labeled. (b) A top-view SEM image of a Q3D gold nanostructure array showing the diameter of ∼400 nm. (c) A 3 μm × 1.5 μm AFM image of a Q3D array and a line profile showing the depth of ∼300 nm. (d) The FDTD-calculated electric field distributions along the x−y plane of the Q3D nanostructure array at the air-gold interfaces of the top gold nanoholes with the ratio of | Emax/E0|4 = 2 × 105 (left) and bottom gold nanodiscs with the ratio of | Emax/E0|4 = 1 × 104 (right) air−gold interfaces. The amplitude of the incident electric field is 1 V/m.

of fresh TSB (30 °C) with 100 μL of the log-phase culture and incubated at 30 °C with shaking (150 rpm) for 4−6 h until the optical density at 600 nm (OD600) approached 0.8 to 1.0 (LBK Biochrome Ultraspec, Model 4050, Cambridge, England). This phase of growth is necessary to ensure the uniformity of bacterial envelope and yield consistent SERS spectra for each Table 1. Seven Strains of V. parahaemolyticus Used in This Study

V. parahaemolyticus strain RepPCRa MLSTb tdh+ c trh+ d source

50

3355

605

551

846

12310

3256

4 ST34 + + water

10 ST65 − − clinical

2 ST3 + − plankton

2 ST3 + − water

1 ST36 + + oyster

1 ST36 + + clinical

1 ST36 + + clinical

a

REP-PCR: repetitive extragenic palindromic (REP) PCR targets the repetitive sequences in bacterial genomes. bMLST: multi locus sequencing type. ctdh+: thermostable direct hemolysin (protein). dtrh+: thermostable related hemolysin (gene). 2632

dx.doi.org/10.1021/ac3021888 | Anal. Chem. 2013, 85, 2630−2637

Analytical Chemistry

Article

nanostructure arrays could be tuned by adjusting hole diameter and depth.16,20 On the basis of 3D-FDTD calculated electric field distributions, the intensity and the location of the maximum local electric fields induced by LSPR can be predicted with precision. We also demonstrated that the optimal SERS sensitivity for detecting small molecules was achieved on the Q3D nanostructure arrays with the maximum local electric field at the bottom gold nanodiscs, an area that small molecules can access. In contrast, the optimal SERS sensitivity for detecting large microorganisms was achieved on the Q3D nanostructure arrays with the maximum local electric field along the nanohole rims of the top gold film.16 In this work, the Q3D nanostructure arrays were fabricated on the ITO coated glass substrates with the ITO layer of ∼140 nm. The 3D-FDTD calculations predicted that the maximum local electric fields are located along the rims of nanoholes at the top gold film while the electric field at the bottom gold nanodiscs are weak when the nanohole diameter is 400 nm, the edge-to-edge spacing is 100 nm, and the depth is 300 nm. The 3D-FDTD calculated electric field profiles along the x−y plane at the top and bottom Au−air interfaces of such Q3D nanostructure arrays is shown in Figure 1d. The SERS EF due to electromagnetic enhancement was estimated according to the relationship of (|Emax(ωlaser)/E0|4), where ωlaser is 785 nm, and the factor of 2 × 105 was obtained. The total EFs reported from SERS experiments include both electromagnetic and chemical enhancement, which is typically in the range of 102− 104 depending on molecules. We estimated the total EF by forming a self-assembled monolayer (SAM) of 4-mercaptopyridine on the Q3D substrate, and the factor of ∼2 × 108 was determined. Both FDTD simulation and experimental results indicate that the Q3D nanostructure arrays made on the ITO coated glass substrates with 400 nm diameter, 100 nm spacing, and 300 nm depth are suitable SERS-active substrates for detecting large microorganisms. Therefore, such Q3D gold plasmonic nanostructure arrays were fabricated on the ITO coated glass substrates and were tested as SERS-active substrates for detecting V. parahaemolyticus. The lateral and vertical dimensions of the Q3D nanostructure are determined by SEM and AFM shown in Figure 1b,c, respectively. SERS Spectra and Barcoding. Since SERS is a near-field effect, bacterial SERS spectra obtained from Q3D nanostructure arrays contain the molecular information primarily from the areas of the bacteria in close proximity to the top nanohole surfaces, where the maximum local electric fields locate. In order to acquire the bacterial SERS spectra that can be used for the rapid identification of bacteria, it is important that the molecular integrity of the cell membrane be maintained and not damaged during the analysis. Previous SERS investigations14,32 used dried bacterial cells affixed to a substrate, introducing the possibility of cellular damage, and thus, SERS data may not be representative of cells under natural conditions. Here, the in situ spectral acquisition method was chosen, and the experimental setup for the measurement is shown in Figure 2a. This method allows for the detection of live bacteria in their natural state, and the bacterial outer membrane experiences the most significant enhancement. Therefore, the degree of spectral variability (wavenumber, intensity, and band shape) reflects the unique biochemical components embedded in the outer membrane of different strains due to bacterial cell inhomogeneities. The dark field optical microscopy images of a SERS bacterial sample taken from this setup are shown in Figure 2b,c. The bright spots are the scattered light from individual live

Figure 2. (a) Schematic illustration of the experimental setup for the in situ acquisition of SERS spectra. (b, c) Dark-field optical microscopy images of V. parahaemolyticus cells taken in an area without and with a SERS-active pattern using a 100× and a 50× objective, respectively. (d) Schematic illustrations of a SERS-active chip with four Q3D nanostructure arrays on an ITO coated glass substrate (left) and the positions of eight SERS spectra collected at each array (right). (e) A 3D illustration showing a bacteria solution on the Q3D nanostructure array.

Identification of Blind Bacterial Samples. Three of the seven (3/7) V. parahaemolyticus strains were selected blind such that the person conducting SERS analysis was unaware of the identity of these strains. Bacterial samples were cultured and prepared following the same procedures aforementioned. Instead of collecting 64 spectra for one strain, a total of 32 SERS spectra were collected on one SERS-active chip using the same spectral acquisition method. Measurement of Mixtures of Two Isolates. Mixtures of the standardized cell cultures of V. parahaemolyticus 551 and 3256 with the volume ratio of 1:1, 2:1, and 1:2 (551:3256) were prepared for the total volume of 500 μL and the cell concentration of 108 cfu/mL. After a 5 min vortex, the 500 μL mixture solution was added to the well of the Teflon holder and a total of 32 spectra were taken for each mixture on one SERSactive chip using the same spectral acquisition method.

■

RESULTS AND DISCUSSION Optimal SERS Substrates Predicted by 3D-FDTD Calculations. In our previous studies, we demonstrated that the electric field distributions of Q3D gold plasmonic 2633

dx.doi.org/10.1021/ac3021888 | Anal. Chem. 2013, 85, 2630−2637

Analytical Chemistry

Article

Figure 3. (a) Averaged SERS spectra of seven V. parahaemolyticus test strains obtained on the Q3D SERS-active substrates. Each spectrum is baseline corrected and is the average of 64 spectra. Vertical lines represent the positions of main peaks presented by all seven V. parahaemolyticus test strains. (b) The normalized SERS barcodes of the seven V. parahaemolyticus test strains. The x-axis is the wavenumber shift. Each barcode contains 64 spectra along the y-axis taken from eight different Q3D gold nanostructure arrays on two SERS-active chips. SERS barcodes of V. parahaemolyticus strains belonging to the same REP-PCR clade are displayed inside boxes.

bacterial cells in solution, and the SERS-active nanopattern area can be seen by a brighter area in Figure 2c. The bacterial cells in solution on a Q3D SERS-active pattern are illustrated in Figure 2e. A total of 64 spectra were collected from two separate SERS-active chips for each V. parahaemolyticus strain as illustrated in Figure 2d. Figure 3a shows the baseline-corrected, averaged, and normalized 64 SERS spectra of each of seven V. parahaemolyticus strains obtained on Q3D gold nanostructured surfaces. Vertical lines indicate 35 common peaks among these seven V. parahaemolyticus strains. The 64 spectra of each V. parahaemolyticus strain were also displayed in a color-contrast spectral barcode as shown in Figure 3b. The Raman shift is represented by the x-axis, and the 64 spectra were stacked along the y-axis. On each barcode, the lines represent the Raman shift of the vibrational peaks and the color contrast and the line width represent the relative intensity and the width of the

peaks, respectively. Due to the highly reproducible spectra collected on the Q3D SERS substrates, the color contrast is very uniform along each line. The striking SERS characteristics of each strain are seen straightforwardly from the barcode. Table S1 of the Supporting Information summarizes the wavenumber and the relative peak intensities of 35 common peaks observed in these strain-specific average SERS spectra. Bolded, underlined, and normal formatting indicate the relative peak intensities of 600−1000, 300−600, and below 300, respectively. These peaks are most likely attributed to the vibrational modes of carbohydrates, proteins, and phospholipids presenting at the outer membrane of V. parahaemolyticus. The cell surface or envelope of V. parahaemolyticus, like many Gram-negative bacteria, is composed (from interior to exterior) of a cytoplasmic inner membrane, the periplasmic space containing a thin peptidogylcan cell wall, and a lipid-rich outer membrane.33,34 As aforementioned, SERS is a near-field 2634

dx.doi.org/10.1021/ac3021888 | Anal. Chem. 2013, 85, 2630−2637

Analytical Chemistry

Article

barcoding for the rapid and accurate identification of specific bacterial strains, three of the seven V. parahaemolyticus strains were randomly selected and prepared as three blind samples where the identity of the strains were only disclosed post-SERS analysis. Following the same cell culturing, sample preparation, spectral collection, and data processing methods established in this study, SERS barcodes (based on 32 spectra) were obtained for each blind sample. Comparison with our seven 64 spectra barcodes (described above) revealed that each blind barcode could be correctly and unambiguously matched to the unknown strains (Figure 4). These results clearly demonstrate

effect. Therefore, the variation of SERS spectral characteristics shown in Figure 3 is hypothesized to reflect the biochemical variation in the outer membrane. Structures presenting at the outer membrane include flagella, fimbriae (pili), outer membrane proteins, and lipo- and capsular polysaccharides (LPS and CPS).35 In host, these structures can elicit an antigenic response, and serotyping is a means of classifying strains based on natural variation in LPS (13 known O antigens) and CPS (71 known K antigens).36 Adherence to epithelial cells of the intestinal tract is an essential step in pathogenesis, and flagella, fimbriae, LPS, and CPS are known to mediate the colonization of epithelial cell lines.22 The seven tested strains of V. parahaemolyticus were selected from four genetically similar clades, based on MLST by Turner et al.29 (Table 1). The SERS spectra and barcodes were also grouped according to their genetic similarity (Figure 3b). The SERS barcodes clearly show that the SERS characteristic peaks are dramatically different for strains in different clades while some common SERS characteristic peaks are found in the same group. For example, V. parahaemolyticus 605 and 551 are characteristic by strong and merged peaks at 1562 and 1539 cm−1 accompanied by a narrow, moderate strong peak at 1444 cm−1. All peaks appearing in the SERS barcodes of these two strains are relatively narrow. In contrast, the SERS barcodes of V. parahaemolyticus 846, 12310, and 3256 show a qualitatively similar and significant peak at 525 cm−1 accompanied by two moderate strong peaks at 669 and 738 cm−1. All these peaks are associated with chemical structures that constitute membranebound proteins. In addition, the spectra of this group of V. parahaemolyticus strains also show a very intense peak at ∼1319 cm−1, which is related to the C−H deformation of proteins. These features indicate the presence of several specific proteins on the outer membrane of V. parahaemolyticus strains from this group. Although these three strains were isolated from different sources, they exhibit very analogous spectral features. V. parahaemolyticus 50 and 3355 belong to two other different groups. They exhibit unique spectral features. They both have a significant peak around 1319 cm−1 and another one around 1540 cm−1 but with a little shift from each other. V. parahaemolyticus 3355 has a relative strong, narrow peak at 525 cm−1 while V. parahaemolyticus 50 almost has no peaks observed in the range of 500−800 cm−1. V. parahaemolyticus 3355 is also characterized by three peaks in the range of 900− 1050 cm−1. All these results indicate that strain-level identification and differentiation of bacteria is feasible based on the SERS barcodes generated from the highly reproducible SERS spectra collected from the Q3D plasmonic nanostructure arrays. In a related study, the NOAA group has initiated the genome sequencing of 30 V. parahaemolyticus strains, including the seven strains presented in this investigation (Turner et al.,30 manuscript in preparation). We hypothesize that comparative genomic analyses will reveal strain-level differences in genes encoding structures associated with the cell surface such as flagella, fimbriae, outer membrane proteins, LPS, and CPS. Such comparisons could aid in the identification of the specific cell surface components that correlate with the observed SERS spectra. We anticipate that the eventual integration of SERS and genomic data will correlate phenotypic and genetic variation at the cell surface and further lend support to the strain-level SERS barcoding of bacterial pathogens. Rapid Identification of V. parahaemolyticus Strains via SERS Barcoding. To verify the capability of the use of SERS

Figure 4. SERS barcodes of three blind V. parahaemolyticus cultures randomly selected from seven test strains. Compared to those generated from seven known strains, these three blind samples can be easily identified: unknown 1, 2, and 3 are strains 12310, 50, and 3355, respectively.

the validity and reproducibility of our methodology. We conclude that the use of SERS-active substrates enhances sensitivity and reproducibility and suggest that SERS-active substrates are essential for the rapid and accurate detection of specific bacterial strains. Tests of Bacterial Mixtures and Limit of Detection. In the natural environment, potentially pathogenic bacterial strains will coexist with a complex microbial community. Therefore, the simultaneous detection of multiple strains becomes an important feature for in-field application. In order to verify that V. parahaemolyticus strains can be detected and distinguished via SERS barcoding when multiple strains are present in one sample, we prepared and analyzed mixed solutions of V. parahaemolyticus 551 and 3256 in 2:1, 1:1, and 1:2 volume/ volume ratios at a final concentration of 108 cfu/mL. The barcodes of the mixtures were directly compared to the singlestrain barcodes of V. parahaemolyticus 551 and 3256 (Figure 5). The spectra of the mixtures clearly display both the most intense peaks of V. parahaemolyticus 551 at 1002, 1177, and 1532 cm−1 indicated by the boxes marked with triangles and the most intense peaks of V. parahaemolyticus 3256 at 525, 738, 1319, and 1639 cm−1 indicated by the boxes marked with circles. Other characteristic peaks from each of two strains can also be observed in the spectra of the mixtures. Not only can the individual strains forming the mixtures be identified by the SERS barcodes but also the quantitative analysis of the cells of each individual strain in the mixtures can be determined by the color contrast of the SERS barcodes. Figure 5 shows that the intensity of the characteristic peaks of V. parahaemolyticus 551 is stronger than that of the characteristic peaks of V. 2635

dx.doi.org/10.1021/ac3021888 | Anal. Chem. 2013, 85, 2630−2637

Analytical Chemistry

Article

demonstrated that the Q3D SERS substrates are capable of detecting the mixtures of two isolates and of measuring the bacterial samples quantitatively. Future work will be focused on the genetic identification of the strain-level differences in SERS spectra through the incorporation of genomic data. The capability of identifying pathogenic strains using SERS could have broader impact to environmental monitoring, biomedical diagnostics, and homeland security.

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Authors

*Phone: 206-860-3377 (M.S.S.); Fax: 206-860-3467 (M.S.S.); E-mail: [email protected] (M.S.S.). Phone: 206-543-4807 (Q.Y.); Fax: 206-685-3451 (Q.Y.); Email: [email protected] (Q.Y.).

Figure 5. Comparison of the SERS barcodes of V. parahaemolyticus 551, mixtures of 2:1, 1:1, and 1:2 ratios of V. parahaemolyticus 551 to V. parahaemolyticus 3256, and V. parahaemolyticus 3256. The most intense peaks of both V. parahaemolyticus 551 and 3256 strains displayed in the SERS barcodes of the mixtures are marked in boxes labeled with triangular and circular dots, respectively.

Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS This work was supported in part by the University of Washington (UW) faculty start-up funds, the National Science Foundation (NSF CBET 1158609), and the West Coast Center for Oceans and Human Health, the NOAA Oceans and Human Health Initiative, and National Marine Fisheries Service. J.J.X. acknowledges a fellowship from the China Scholarship Council. Nanofabrication and studies of SERS were performed at the Nanotech User Facility, the UW site of the National Nanotechnology Infrastructure Network (NNIN) supported by the NSF. L.R. acknowledges the NNIN Research Experience for Undergraduates (REU) program supported by the NSF.

parahaemolyticus 3256 for the mixture with the volume ratio of 2:1 of the strains 551 to 3256 and vice versa. We performed a qualitative assessment of the limit of detection (LOD) using five V. parahaemolyticus 551 dilutions, ranging from 104 to 108 cfu/mL. This approach is qualitative in that we do not know what fraction of bacteria in these samples became attached to the SERS surface or whether it was consistent among the dilutions. The most intense peak (1539 cm−1) in the SERS spectra was selected to test the LOD. At concentrations below 108 cfu/mL, the normalized intensity of peak 1539 cm−1 decreased nonlinearly with cell concentration (Figure S1 in the Supporting Information). This nonlinearity likely is due largely to changes in the relative proportions of cells on the SERS surface as total cell concentrations decrease, though we were unable to assess this aspect. The findings indicate that at concentrations below 105 cfu/mL there were too few bacterial cells adhering to the surface to generate a measurable signal. Increasing cellular adhesion to the surface, or increasing the thickness of the detection zone above the SERS surface, should lead to significant improvement in the analytical detection limit.

■

REFERENCES

(1) Kneipp, J.; Kneipp, H.; Kneipp, K. Chem. Soc. Rev. 2008, 37, 1052−1060. (2) Kneipp, J.; Wittig, B.; Bohr, H.; Kneipp, K. Theor. Chem. Acc. 2010, 125, 319−327. (3) Vo-Dinh, T.; Wang, H. N.; Scaffidi, J. J. Biophotonics 2010, 3, 89− 102. (4) Scaffidi, J. P.; Gregas, M. K.; Seewaldt, V.; Vo-Dinh, T. Anal. Bioanal. Chem. 2009, 393, 1135−1141. (5) Tripp, R. A.; Dluhy, R. A.; Zhao, Y. Nano Today 2008, 3, 31−37. (6) Ma, K.; Yuen, J. M.; Shah, N. C.; Walsh, J. T.; Glucksberg, M. R.; Van Duyne, R. P. Anal. Chem. 2011, 83, 9146−9152. (7) Bantz, K. C.; Meyer, A. F.; Wittenberg, N. J.; Im, H.; Kurtulus, O.; Lee, S. H.; Lindquist, N. C.; Oh, S. H.; Haynes, C. L. Phys. Chem. Chem. Phys. 2011, 13, 11551−11567. (8) Fan, M. K.; Andrade, G. F. S.; Brolo, A. G. Anal. Chim. Acta 2011, 693, 7−25. (9) Jarvis, R. M.; Goodacre, R. Chem. Soc. Rev. 2008, 37, 931−936. (10) Efrima, S.; Bronk, B. J. Phys. Chem. B 1998, 102, 5947−5950. (11) Jarvis, R.; Brooker, A.; Goodacre, R. Anal. Chem. 2004, 76, 5198−5202. (12) Jarvis, R.; Brooker, A.; Goodacre, R. Faraday Discuss. 2006, 132, 281−292. (13) Sengupta, A.; Mujacic, M.; Davis, E. Anal. Bioanal. Chem. 2006, 386, 1379−1386. (14) Premasiri, W.; Moir, D.; Klempner, M.; Krieger, N.; Jones, G., II; Ziegler, L. J. Phys. Chem. B 2005, 109, 312−320.

■

CONCLUSIONS We demonstrate here that seven clinical and environmental V. parahaemolyticus strains from four phylogenetic clades can be rapidly detected and identified using SERS. The in situ spectral acquisition method allows for the detection of live bacteria in their natural state, and the bacterial outer membrane experiences the most significant enhancement. The unique SERS barcode of each strain carries the SERS spectral characteristics of the biochemical information embedded in the outer membrane, making possible the direct identification of bacterial strains by comparing the SERS barcodes. Using these unique SERS barcodes as references, we were able to rapidly and accurately identify “blind” (i.e., unknown) preparations of our bacterial strains. Additionally, our results 2636

dx.doi.org/10.1021/ac3021888 | Anal. Chem. 2013, 85, 2630−2637

Analytical Chemistry

Article

(15) Yang, L. L.; Yan, B.; Premasiri, W. R.; Ziegler, L. D.; Dal Negro, L.; Reinhard, B. M. Adv. Funct. Mater. 2010, 20, 2619−2628. (16) Xu, J. J.; Zhang, L.; Gong, H.; Homola, J.; Yu, Q. M. Small 2011, 7, 371−376. (17) Ebbesen, T. W.; Lezec, H. J.; Ghaemi, H. F.; Thio, T.; Wolff, P. A. Nature 1998, 391, 667−669. (18) Garcia-Vidal, F. J.; Martin-Moreno, L.; Ebbesen, T. W.; Kuipers, L. Rev. Mod. Phys. 2010, 82, 729−787. (19) Willets, K. A.; Van Duyne, R. P. Annu. Rev. Phys. Chem. 2007, 58, 267−297. (20) Xu, J. J.; Kvasnička, P.; Idso, M.; Jordan, R. W.; Gong, H.; Homola, J.; Yu, Q. M. Opt. Express 2011, 19, 20493−20505. (21) Su, Y. C.; Liu, C. C. Food Microbiol. 2007, 24, 549−558. (22) Yeung, P. S. M.; Boor, K. J. Foodborne Pathog. Dis. 2004, 1, 74. (23) Daniels, N. A.; MacKinnon, L.; Bishop, R.; Altekruse, S.; Ray, B.; Hammond, R. M.; Thompson, S.; Wilson, S.; Bean, N. H.; Griffin, P. M.; Slutsker, L. J. Infect. Dis. 2000, 181, 1661−1666. (24) Connelly, J. T.; Baeumner, A. J. Anal. Bioanal. Chem. 2012, 402, 117−127. (25) Lazcka, O.; Del Campo, F. J.; Munoz, F. X. Biosens. Bioelectron. 2007, 22, 1205−1217. (26) Nayak, M.; Kotian, A.; Marathe, S.; Chakravortty, D. Biosens. Bioelectron. 2009, 25, 661−667. (27) Lumerical FDTD Solution, FDTD Solutions 6.5, http://www. lumerical.com/fdtd.php. (28) Yu, Q. M.; Guan, P.; Qin, D.; Golden, G.; Wallace, P. M. Nano Lett. 2008, 8, 1923−1928. (29) Turner, J. W.; Paranjpye, R. N.; Landis, E.; Gonzales-Escalona, N.; Nilsson, W. B.; Biryukov, S. V.; Strom, M. S. PLoS One 2012, 8, e55726. (30) Turner, J. W.; Paul, S.; Berthiaume, C.; Sokurenko, E. V.; Morales, R.; Armbrust, E. V.; Strom, M. S. Proceedings of the National Academy of Sciences 2013, to be submitted. (31) MathWorks, Matlab 7.0, www.mathworks.com. (32) Zhang, L.; Xu, J.; Mi, L.; Gong, H.; Jiang, S.; Yu, Q. Biosens. Bioelectron. 2012, 31, 130−136. (33) Costerton, J. W.; Ingram, J. M.; Cheng, K.-J. Bacteriological Rev. 1974, 38, 87−110. (34) Deneke, C. F.; Colwell, R. R. Can. J. Microbiol. 1973, 19, 241− 245. (35) Enos-Berlage, J. L.; McCarter, L. L. J. Bacteriol. 2000, 182, 5513−5520. (36) Hsieh, Y. C.; Liang, S. M.; Tsai, W. L.; Chen, Y. H.; Liu, T. Y.; Liang, C. M. Infect. Immun. 2003, 71, 3329−3336.

2637

dx.doi.org/10.1021/ac3021888 | Anal. Chem. 2013, 85, 2630−2637