Open Access
Add to BibSonomyAbstract
We investigate the mode of action and classification of antibiotic agents (ceftazidime, patulin, and epigallocatechin gallate; EGCG) on Pseudomonas aeruginosa (P. aeruginosa) biofilm using Raman spectroscopy with multivariate analysis, including support vector machine (SVM) and principal component analysis (PCA). This method allows for quantitative, label-free, non-invasive and rapid monitoring of biochemical changes in complex biofilm matrices with high sensitivity and specificity. In this study, the biofilms were grown and treated with various agents in the microfluidic device, and then transferred onto gold-coated substrates for Raman measurement. Here, we show changes in biochemical properties, and this technology can be used to distinguish between changes induced in P. aeruginosa biofilms using three antibiotic agents. The Raman band intensities associated with DNA and proteins were decreased, compared to control biofilms, when the biofilms were treated with antibiotics. Unlike with exposure to ceftazidime and patulin, the Raman spectrum of biofilms exposed to EGCG showed a shift in the spectral position of the CH deformation stretch band from 1313 cm−1 to 1333 cm−1, and there was no difference in the band intensity at 1530 cm−1 (C = C stretching, carotenoids). The PCA-SVM analysis results show that antibiotic-treated biofilms can be detected with high sensitivity of 93.33%, a specificity of 100% and an accuracy of 98.33%. This method also discriminated the three antibiotic agents based on the cellular biochemical and structural changes induced by antibiotics with high sensitivity and specificity of 100%. This study suggests that Raman spectroscopy with PCA-SVM is potentially useful for the rapid identification and classification of clinically-relevant antibiotics of bacteria biofilm. Furthermore, this method could be a powerful approach for the development and screening of new antibiotics.
© 2014 Optical Society of America
1. Introduction
Pseudomonas aeruginosa (P. aeruginosa), a Gram-negative bacterium, is the most common pathogen in the lungs of patients with cystic fibrosis (CF) and immunodeficiency, and leads to lung damage, respiratory failure and death in these patients [1]. The persistent infections in the patients’ lung are often connected to the emergence of antibiotic-resistant strains. The bacteria in the lung form micro-colonies and a biofilm, a slimy layer composed of a hydrated matrix of polysaccharide and protein. Bioðlms are typically complex and heterogeneous systems mainly characterized by adhered growth, a channeled structure and extracellular matrices (ECMs). Biofilm formation is strongly correlated with the resistance to antibiotic chemotherapy [2, 3]. Bacteria embedded in biofilm can be 1,000-fold more resistant to antibiotics than planktonic cells, and most antibiotics are not able to eradicate the bacterial cells in the biofilm.
Considering that the biofilm is a critical factor for the efficacy of antimicrobials and also important for development of drug-resistance in vivo, an effective method to evaluate the effect of drugs on biofilms is required. However, the conventional tests of antibiotic susceptibilities, such as the Kirby-Bauer (disc diffusion) test or broth dilution measurements exclude the conditions of biofilm formation and thus do not reflect the effectiveness of antimicrobials in vivo. Moreover, fluorescence-induced confocal laser scanning microscopy (CLSM) has been shown to be a powerful technique that reveals a three-dimensional structural image with microscopic resolution and allows for quantitative assessment of the bioðlm constituents [4–7]. However, since the extracellular polymeric substance (EPS) is a complex mixture of different biopolymers with a large number of potential binding sites [8], it is difficult to design a suitable protocol to stain the entire EPS, which limits the application of CLSM in EPS identification and quantification. Other techniques, such as scanning electron microscopy (SEM) [9, 10], transmission electron microscopy (TEM) [11, 12] and Fourier transform infrared spectroscopy (FT-IR) [13, 14], have also been used to characterize the chemical composition of biofilms. However, these techniques are also limited by several disadvantages. Electron microscopy (SEM or TEM) is limited by the requirement of dehydration of the samples during preparation, which might alter the natural structure of the biofilm or create artifacts. For FT-IR, the limitation of spatial resolution may block its performance on smaller samples, such as the level of bacterial cells or micro-colonies [15].
Raman spectroscopy has recently attracted great interest as a prospective new optical technique for rapid microbial detection and classification. Buijtels et al. demonstrated the ability of Raman spectroscopy to accurately identification of Mycobacteria with a high specificity of 95.2% [16]. Chu et al. reported the rapid identification of microorganisms with a minimum of sample preparation using SERS-active silver nanorod array substrate [17]. Moreover, Grow et al. showed that pathogens can be individually identified by μSERS. The μSERS is a new biochip technology that uses SERS microscopy for label-free transduction [18]. The biochip itself comprises pixels of capture biomolecules immobilized on a SERS-active metal surface.
Raman spectroscopy has been also successfully tested for nondestructive biofilm analysis, including microbial constituents and EPS [19–23]. This technique provides detailed chemical information about microbial cells and complex biofilm matrices with the spatial resolution of an optical microscope, and it does not require staining of the entire sample. Furthermore, unlike IR spectroscopy, Raman spectroscopy is characterized by a low-water background, which is beneficial for analysis of biomatrices, and it can potentially be more accurate and sensitive than existing methods. Recently, a number of studies have been conducted using Raman spectroscopy to monitor the inactivation of bacteria exposed to antibiotics [24–26]. However, there have been few studies of the effect of antibiotics on biofilms [27]. This information is critical in order to accurately assess the significance of spectral changes induced by antibiotic drugs.
In this study, we investigate the effect of the three antibiotic agents (ceftazidime, patulin, and epigallocatechin gallate (EGCG)) on P. aeruginosa biofilm using Raman spectroscopy. Patulin and epigallocatechin gallate (EGCG) are the representative antagonists of quorum-sensing (QS) signal molecules, which induce biofilm dispersal and enhance the efficacy of antibiotics in the removal of the biofilms [28] The dispersal of biofilm is triggered by several bacterial signals including acyl-homoserine lactones, cell–cell autoinducing peptides, fatty acids and d-amino acids or the environmental cues such as availability of nutrients, oxygen nitric oxide (NO), iron levels and changes of temperature. Since biofilm is one of the main factors for virulence and resistance, recently QS inhibitors were attempted to control the biofilm dispersal as a novel therapeutic method against drug-resistant strains [29]. The biofilms were grown and treated with various agents in the microfluidic device, and then transferred onto gold-coated substrates for Raman measurement. In order to discriminate these antibiotic agents according to the spectral differences between them, the multivariate statistical method of principal component analysis (PCA)-support vector machine (SVM) analysis is used.
2. Materials and method
2.1. Bacterial strains and culture methods
P. aeruginosa (KCTC 1637) was purchased from the Korean Collection for Type Cultures (KCTC), Korea Research Institute of Bioscience & Biotechnology (Taejon, Korea). Bacterial cells were cultured overnight in Luria-Bertani (LB) broth at 37°C with constant shaking (220 rpm). For the biofilm experiments, bacteria at mid-log phase (OD600 nm = 0.5) were then injected into a microfluidic device. The device was incubated at room temperature for 3 hr to allow for the initial attachment of bacteria. LB broth was continuously flowed through the microfluidic channel at 2 μL/min using a syringe pump (KD Scientific, Holliston, MA, USA).
2.2. Microfluidic device for biofilm formation
The microfluidic chip was made from two layers of PDMS (polydimethylsiloxane, Sylgard 184, Dow-Corning, Cortland, NY, USA) using soft lithography [30]. In detail, the layer with chamber and channels was prepared at a ratio of 1:10 (v/v) of curing agent and PDMS, and both mixtures were then degassed. The structure was placed in the oven at 80°C for at least 4 hr, and holes through the layer were punched following this incubation. The microfluidic chamber was a 13 mm × 13 mm rectangular structure, 450-μm-deep and connected to inlet and outlet channels (Fig. 1). The holes onto the inlet and outlet channels were created using a hand punch (Miltex, Rietheim-Weilheim, Germany), and the surfaces of PDMS and a microscope slide were simultaneously treated in an O2 plasma cleaner (Femto Science laboratory, Hwasung, Korea) for 40 sec, after which the two surfaces were brought together to form an irreversible bond. Tygon tubing (Fisher Scientific International Inc., Hampton, NH, USA) was connected to small hollow metal tubes, which in turn were firmly inserted into the punched holes and the tubing connected to a 20 G syringe needle (Korea Vaccine Company, Seoul, Korea).
2.3. Antibiotics treatment
Ceftazidime, patulin and EGCG were purchased from Sigma Chemical Co. These were chosen to test the therapeutic effect on bacteria cells or bacterial biofilm. According to european consensus [31] the recommended 200 mg/kg/day ceftazidime was administrated to P. aeruginosa biofilm. The effective dosages for the antibacterial agents patulin was 2.5 mg/kg [32] and the average minimum inhibitory concentration (MIC) of EGCG against P. aeruginosa was 500 μg/mL (400-800 μg/ml) [33]. Even though the concentrations were the effective concentrations in vivo in order to be indisputable for the therapeutic effect of antimicrobials, the same concentrations considering the total volume of a micro-chamber (76.05 mm3) were applied in the microfluidic device including P. aeruginosa biofilm [32]. After formation of biofilm on the chip, the corresponding concentration of antimicrobial agents in phosphate buffered saline (PBS; pH 7.4) was continuously flowed through the microfluidic channel at 2 μL/min using a syringe pump for 18 hr (KD Scientific, Holliston, MA).
2.4. Quantification of biofilm using laser scanning microscopy
To determine the biofilm formation, two microfluidic device were fabricated and bacteria cells were incubated parallel under same low flow condition (2 μL/min) using the same syringe pump. After 3 days incubation, one microfluidic device with cultured biofilm including bacteria cells was applied for optical investigation using fluorescence microscopy. Another microfluidic device was opened and the biofilm was transferred onto gold-coated substrate for Raman spectroscopy. We performed experiment in triplicate. To quantify bacteria cells in biofilm, rabbit antibody against P. aeruginosa (AbCam), diluted 1:100 in PBS, was introduced into the chip and incubated for 30 min. After washing with PBS, fluorescein isothiocyanate (FITC)-labeled goat anti-rabbit antibody (Sigma Chemical Co.), diluted 1:100 in PBS, was added to the chip and incubated for 30 min. To quantify the biofilm by fluorescence, tetramethylrhodamine isothiocyanate (TRITC)-labeled concanavalin A (Sigma Chemical Co.), which specifically binds to d-(L)-glucose and d-(L)-mannose groups on EPS, was loaded into the chip and incubated for 30 min, then washed with PBS. Fluorescence images were obtained with a LSM700 confocal microscope system (Carl Zeiss) equipped with an argon/2 laser and a helium-neon laser with excitation wavelengths of 488 nm and 543 nm for FITC and TRITC, respectively. Images were analyzed using Imaris software (Bitplane).
2.5. Raman spectroscopic measurements
For the Raman spectroscopic measurements, we used gold-coated substrate. A thin 10 nm, Cr layer followed by a 50 nm gold layer were e-beam evaporated on the glass substrate. The Cr layer was included to increase the adhesion between the gold film and the glass substrate. This substrate was minimized spectral contributions from the sample substrate such as glass and enhanced the Raman signal compared to that. Pure metals are known to have no Raman spectral features and very low background signal. Untreated and treated P. aeruginosa biofilms were deposited onto the substrate, and Raman measurements were taken after drying at room temperature.
Raman spectra were acquired using the SENTERRA confocal Raman system (Bruker Optics Inc., Billerica, MA, USA) equipped with a 785 nm diode laser source (100 mW before objective) and a resolution of 3 cm−1. A 100 × air objective (MPLN N. A. 0.9, Olympus), which produced a laser spot size of ~1 μm, was used to collect Raman signals. The Raman spectra of the biofilm associated with the autofluorescence background were displayed in computer in real time and saved for further analysis. An automated algorithm for autofluorescence background removal was applied to the measured data to extract pure sample Raman spectra. The Raman spectrum of each biofilm was calculated as the average of fifteen measurements at different arbitrary sites on the biofilm. Baseline correction was performed by the rubber-band method, which was used to stretch between the spectrum endpoints. All Raman measurements were recorded with an accumulation time of 60 sec in the 650-1750 cm−1 range, and Raman spectral acquisition and preprocessing of preliminary data such as baseline subtraction, smoothing, and spectrum analysis were carried out using the OPUS software.
2.6. Data analysis
The PCA algorithm to extract the global spectral features of three antibiotic agents on P. aeruginosa biofilm formation was implemented with MATLAB computing software (MathWorks Inc., Natick, MA, USA) as the multivariable analytical tool. All spectral vectors normalized by Z-Score method were used as the transfer function to determine the principal component (PC) scores for each spectrum of each group. A total of seven Raman spectra, including one control spectra, three antibiotic agents treated spectra, and three the difference spectra between three antibiotic agents treated spectra and control spectra were designed as an input of the transfer function to evaluate the antibiotic effects on P. aeruginosa biofilm. A simple mapping method to project two global parameters selected manually among the seven PC scores acquired to a two-dimensional space was used. PC1 to PC4 were chosen as candidate global parameters. Multi-SVM classifiers with a one-against-all (OAA) approach [34] onto this two-dimensional plot were used to classify pure and ceftazidime-, patulin- and EGCG-treated biofilms, after the validation through leave-one-out cross-validation. A 60 data set consisting a training set (n = 42) and a testing set (n = 18) according to a common rule of data mining was used in this study. PCA-SVM assessments were performed by evaluating sensitivity, specificity and accuracy. Sensitivity was used to evaluate the ability of the PCA-SVM classifier to detect the presence of drug-derived effects, specificity was used to determine the ability of the PCA-SVM classifier to specify the effect of certain drugs and accuracy was used to evaluate the efficiency of the PCA-SVM classifier. Quantitative data are expressed as the mean ± standard deviation (SD). Statistical analysis was performed to compare relative intensities of the Raman peaks between groups using a multiple range of one-way ANOVA. Additional post hoc comparisons were performed using a Student-Newman-Keuls test where appropriate. P-values <0.05 were considered significant.
3. Results and discussion
Bacteria within biofilm can be 1,000 times more resistant to antibiotics than their planktonic counterparts. Thus, the monitoring of bacteria in biofilm and studies about the effect of antimicrobial materials on biofilm are required to understand and regulate the drug resistance of infectious bacteria. However, the structure of biofilm and its response to the antibacterial agents are different from those in static conditions [35, 36]. Therefore, in this study we utilized a microfluidic device and Raman spectroscopy for the study of antibiotic effects on P. aeruginosa biofilm.
Microfluidic systems are often used for biological studies because they allow precise analysis by exact controlled chemical gradients and conditions using a fluid flow of small liquid volume. Furthermore, the device enables 3-D development of the biofilm and is transparent, such that not only real-time imaging, but also the quantification of biofilm formation, is possible. For the biofilm studies the microfluidic system is an adequate platform, since the most important factor for the biofilm formation is hydrodynamic condition of fluid [37]. A schematic image of the experimental setup of the microfluidic device used in this study is presented in Fig. 1. The dimensions of the microfluidic chamber for biofilm formation were 13 mm by 13 mm and 450-μm-deep (Fig. 1). It provides favorable conditions for biofilm formation, where hydrodynamics have a direct effect on biofilm development. Thus, biofilm formation was promoted under low flow conditions (~1-5 μL/min) rather than static conditions [35, 38]. The proper condition for the biofilm formation has been reported in a comparable micro-chamber of microfluidic system [33]. According to the condition, a syringe pump was connected to the channel for introducing media into channel at 2 μL/min of low flow rate. After 3 days incubation under 2 μL/min flow condition biofilm reached to 18.4% of microfluidic chamber (total volume = 76.05 mm3), which covered almost the surface of micro-chamber (Fig. 2).
Fig. 2 Representative fluorescent images of P. aeruginosa cells and biofilm and 3D-reconstructed image of the cells and biofilm in the microfluidic device. (A) 3D-reconstructed image of biofilms in the microfluidic device for 1-3 days. (B) Bacteria cells were stained by fluorescein isothiocyanate-labeled goat anti-rabbit antibody (Ab-FITC) and biofilm by tetramethylrhodamine isothiocyanate-labeled concanavalin A (Concanavalin A-RITC). Scale bars represent 20 μm.
Figure 2 shows the biofilm after 3 days of incubation in the microfluidic chamber as well as the effect of antibiotic agents on biofilm formation. Bacterial cells and biofilms were stained with FITC- and TRITC-labeled concanavalin A, respectively. After 3 days of incubation, the biofilm covered the microfluidic chamber and the total volumes of bacteria and biofilm in the microfluidic chamber were estimated at 8.7 ± 1.4 mm3 and 14.0 ± 3.2 mm3, calculated by a three-dimensional reconstruction process based on the fluorescent images. After treatment with ceftazidime for 18 hr, 20.4% and 27.3% of the bacteria and biofilm remained, respectively, compared to biofilm without antibiotic treatment. Treatments with patulin and EGCG led to 43.8% and 40.8% of bacteria remaining, and 43.8% and 19.8% of biofilm remaining, respectively.
Figure 2 shows the biofilm after 3 days of incubation in the microfluidic chamber as well as the effect of antibiotic agents on biofilm formation. Bacterial cells and biofilms were stained with FITC- and TRITC-labeled concanavalin A, respectively. After 3 days of incubation, the biofilm covered the microfluidic chamber and the total volumes of bacteria and biofilm in the microfluidic chamber were estimated at 8.7 ± 1.4 mm3 and 14.0 ± 3.2 mm3, calculated by a three-dimensional reconstruction process based on the fluorescent images. After treatment with ceftazidime for 18 hr, 20.4% and 27.3% of the bacteria and biofilm remained, respectively, compared to biofilm without antibiotic treatment. Treatments with patulin and EGCG led to 43.8% and 40.8% of bacteria remaining, and 43.8% and 19.8% of biofilm remaining, respectively.
In this study, we analyzed the effect of antibiotic agents on the biofilm of P. aeruginosa using Raman spectroscopy with a PCA-SVM bioinformatics technique. To analyze the biofilm and therapeutic effect of antimicrobial agents, the biofilms were transferred onto gold-coated substrates for Raman measurement. Since the biofilm on the gold surface did not uniform, the Raman spectra of biofilm were measured at 15 different sites on a biofilm and calculated as the average of fifteen measurements.
Figure 3 shows averaged Raman spectra for P. aeruginosa biofilm. It is characterized by vibrational bands typical for nucleic acids, proteins, lipids, carbohydrates and carotenoids (see Table 1) [19, 39–42]. The prominent Raman bands of biofilm belong to proteins: 1002 cm−1 (phenylalanine), 1243 cm−1 (amide III) and 1658 cm−1 (amide I). The strong bands at 1313 cm−1 and 1449 cm−1 (CH deformations) can be assigned to polysaccharides and lipids, as well as to protein, which can be found primarily in the cell membrane. The Raman spectrum obtained from the nucleus of P. aeruginosa biofilm is characterized by bands at 726 cm−1 (A), 780 cm−1 (U, C and T), 1100 cm−1 (PO2- stretching) and 1574 cm−1 (G and A). The band at 1530 cm−1 (C = C stretching) in the Raman spectra of biofilm could be associated with carotenoids. The peak at 1124 cm−1 could be associated with the stretching vibration from symmetric glycosidic linkages (C-O-C) and ring breathing of polysaccharides or C-C stretching vibrations. Polysaccharides, especially polyanionic macromolecules, are thought to predominantly account for the multiple bands in the bioðlm Raman spectra.
To investigate the effect of antibiotic agents themselves in Raman mode of antibiotic agent-treated biofilm, we measured the three antibiotic agents themselves (ceftazidime, patulin, and EGCG) as controls. The Raman spectra of three antibiotic solutions show unique peak and shape corresponding to the antibiotics (Fig. 4.). There were no overlapped between the Raman spectra of the antibiotic agents and antibiotic-treated biofilms. This result indicated that the Raman spectra of antibiotic agent-treat biofilms were independent of the added agents themselves Raman spectra.
Figure 5(A) shows the averaged Raman spectra of P. aeruginosa biofilm undergoing control and treatment with three antibiotic agents. To signify the difference in spectral profiles more clearly, the difference spectra were also extracted by subtracting ceftazidime-, patulin- and EGCG-treated spectra from control spectra, respectively, see Figs. 5(B), 5(C), and 5(D). Figure 5 clearly shows a difference in behavior between normal and antibiotic agent-treated biofilms, as well as differences between films treated with the different antibiotic agents.
Fig. 5 Raman spectra of P. aeruginosa biofilms. (A) Averaged spectra for control and antibiotic-treated biofilms; (B-D) (a) Overlaid spectra of control and antibiotic-treated biofilm, (b) spectral differences of control biofilm and biofilm treated with antibiotic: (B) ceftazidime, (C) patulin, (D) EGCG.
In order to quantitatively identify the effect of treatment with antibiotic agents on the variation in P. aeruginosa biofilm components, we investigated specific Raman peaks and their intensities corresponding to DNA/RNA, proteins, lipids, carbohydrates, and carotenoids (Fig. 6). P. aeruginosa biofilm, upon exposure to antibiotics, showed a decrease in the magnitude of Raman intensity at 1002 cm−1, 1243 cm−1 and 1658 cm−1, corresponding to proteins. This result might indicate denaturation and conformational changes in proteins related to cell death. It was previously proposed that decrease of these Raman intensity were related to the denaturation and conformational changes of proteins, which correlated with the cell death [43–45]
Fig. 6 Relative intensities of the Raman peaks for control biofilm and biofilm treated with three antibiotic agents. *P < 0.001 (one-way ANOVA) with post hoc comparisons (Table 2).
Table 2. Effect of treatment with antibiotic agents on the variation in P. aeruginosa biofilm components. Each Raman peak indicates a significant difference between two groups after ANOVA (Student-Newman-Keuls test, P < 0.05).
The reduction of Raman intensity corresponding to DNA, such as the bands at 780 and 1100 cm−1, arose from the destruction of the ring structure, indicating degradation of the DNA. It indicates another evidence for cell death. Apart from biochemical changes related to proteins, cell death involves significant changes in the cell nucleus.
The magnitude of Raman intensity at 1530 cm−1 increased following treatment with ceftazidime and patulin antibiotics, which is assigned to the C = C stretching mode in carotenoids. The accumulation of carotenoids is thought to be triggered by various environmental conditions and stimuli, especially oxidative stress [46]. It could be speculated that since the oxidative stress in P. aeruginosa is an important cue for the production of diversity, the increased level of carotenoids is the first step for developing drug resistant strains in the biofilm [47].
Unlike with ceftazidime and patulin antibiotics, the EGCG-treated P. aeruginosa biofilm showed a shift in the 1313 cm−1 (CH deformation) stretch band toward 1333 cm−1. Furthermore, there was no difference in the band intensity at 1530 cm−1, which indicates a lack of induced accumulation of carotenoids.
EGCG is a quorum sensing inhibitor and suppressed biofilm formation in Pseudomonas putida and Burkholderia cepacia by interfering with the AI-1 signaling system [48]. However, EGCG can also generate the hydroxyl radical (H2O2) and cause damage to the cell walls as visualized by atomic force microscopy (AFM) [49]. The perforation of the cell wall may be the reason for the absence of carotenoid accumulation.
Although the interpretation of the spectral changes observed in Fig. 5 is not straightforward, these changes might be sufficient to detect or identify the effect of antibiotics. To confirm this hypothesis, PCA-SVM analysis was applied to the Raman spectral data. Figure 7(a) presents the score plot of PC1 versus PC2 from PCA for normal and treatment with three antibiotic agents on P. aeruginosa biofilm. PCA demonstrated an analytical grouping of spectra according to normal, ceftazidime-, patulin- and EGCG-treated biofilms. The variation in the pure biofilm in the score plot is small, as seen by the clustering of the spectra in the score plot. However, the loading profiles of biofilm groups treated with antibiotic agents were distributed along a wide region. The loading profiles of PC1 and PC2 explain 72% and 20% of the total variation in the spectra. The optimal separating lines (Fig. 7(a)) obtained from the SVM classification algorithm (linear kernel) using the training data from each group clearly showed the difference between the normal and treatment groups. The proposed PCA-SVM classifiers achieved a high sensitivity (93.33%), high specificity (100%) and an average accuracy of 98.33% for classifying antibiotic-treated biofilm and normal biofilm. Therefore, this finding suggests that the separating lines calculated on PC1 and PC2 could be used as a marker to monitor and detect the presence of antibiotic agents on P. aeruginosa biofilm. Figure 7(b) shows the PCA-SVM scores (Gaussian kernel) corresponding to PC3 and PC4 for the discrimination between the effects of different antibiotic agents on the biofilm. In fact, the loading profiles of PC3 and PC4 are relatively low at 7% and 1% of the total variation in the spectra. However, the score plot of PC3 versus PC4 obviously showed the individual characteristics of the normal and antibiotic agent groups, which yielded a sensitivity of 100% and a specificity of 100% for classifying between the antibiotics. In fact, we performed the classification with three kernels such as linear, Gaussian, and polynomial. Gaussian and low-order polynomial kernels could separate the data well, whereas linear and high-order (p>4) polynomial kernels showed poor cross-validation performance. This lower classification accuracy is responsible for over-fitting for the polynomials of higher order. Gaussian kernel showed the best performance. The representative Raman spectral peaks to play a major role in PC1 to PC4 are summarized in Table 3. It is obvious that high Raman spectral information has a strong influence on PC1 and PC2, whereas low Raman spectral information has a strong influence on PC3 and PC4. Finally, this optimal boundary information could be used as a marker to find the distinctive functions of newly-developed antibiotic agents on biofilm. Particularly, the database based on PCA-SVM analysis including more sample sizes and disease cases as well as various kernel functions and combining with other bioinformatic method is more efficient for classification of various infectious diseases.
Fig. 7 PCA-SVM scores: (a) PC1-PC2 plot of control biofilm and biofilm treated with antibiotic agents, (b) PC3-PC4 plot of control and biofilms treated with ceftazidime, patulin and EGCG. The optimal boundary lines could be used as a marker to monitor and detect the presence of antibiotic agents and to find the distinctive functions of newly-developed antibiotic agents on P. aeruginosa biofilm.
Table 3. Representative five Raman spectral peaks with high contribution at PC1 to PC4.
Raman spectroscopy and PCA-SVM analysis not only distinguished between normal and antibiotic treatment biofilms, but can also discriminate between three different antibiotic agents based on the cellular biochemical and structural changes induced by antibiotic drugs. Thus, the high discriminatory power of the Raman spectroscopy, combined with the PCA-SVM method, has great potential for providing an effective and accurate diagnostic schema for infectious diseases correlated drug resistance within biofilms.
4. Conclusion
In this work, we have demonstrated that Raman spectroscopy can be used to identify spectral differences in P. aeruginosa biofilms following treatment with three antibiotic agents. The Raman spectra of biofilms exposed to the three antibiotic agents each exhibited different behaviors. For biofilms exposed to EGCG, the Raman peak showed a shift from 1313 cm−1 to 1333 cm−1, and there was no difference in the band intensity associated with carotenoids at 1530 cm−1 (C = C stretching). A discriminant model-based PCA and SVM applied to the Raman spectra was able to group those samples with good sensitivity and high specificity, as well as separate spectra resulting from treatment with the three antibiotic agents with 100% sensitivity and specificity.
Raman spectroscopy combined with multivariate analysis may lead to applications in pharmacology and infectious medicine because it can be used to analyze the action of antibiotics and correlate with drug resistance within biofilms.
Acknowledgments
This work was supported by a grant from Kyung Hee University in 2013 (KHU-20131086).
References and links
1. J. R. Govan and V. Deretic, “Microbial pathogenesis in cystic fibrosis: mucoid Pseudomonas aeruginosa and Burkholderia cepacia,” Microbiol. Rev. 60(3), 539–574 (1996). [PubMed]
2. P. K. Singh, A. L. Schaefer, M. R. Parsek, T. O. Moninger, M. J. Welsh, and E. P. Greenberg, “Quorum-sensing signals indicate that cystic fibrosis lungs are infected with bacterial biofilms,” Nature 407(6805), 762–764 (2000). [CrossRef] [PubMed]
3. P. S. Stewart and J. W. Costerton, “Antibiotic resistance of bacteria in biofilms,” Lancet 358(9276), 135–138 (2001). [CrossRef] [PubMed]
4. R. Wakamatsu, S. Takenaka, T. Ohsumi, Y. Terao, H. Ohshima, and T. Okiji, “Penetration kinetics of four mouthrinses into Streptococcus mutans biofilms analyzed by direct time-lapse visualization,” Clin. Oral Investig. 18(2), 625–634 (2014). [CrossRef] [PubMed]
5. N. Mangwani, S. K. Shukla, T. S. Rao, and S. Das, “Calcium-mediated modulation of Pseudomonas mendocina NR802 biofilm influences the phenanthrene degradation,” Colloids Surf. B Biointerfaces 114, 301–309 (2014). [CrossRef] [PubMed]
6. M. van Gennip, L. D. Christensen, M. Alhede, K. Qvortrup, P. Ø. Jensen, N. Høiby, M. Givskov, and T. Bjarnsholt, “Interactions between polymorphonuclear leukocytes and Pseudomonas aeruginosa biofilms on silicone implants in vivo,” Infect. Immun. 80(8), 2601–2607 (2012). [CrossRef] [PubMed]
7. M. Y. Chen, D. J. Lee, Z. Yang, X. F. Peng, and J. Y. Lai, “Fluorecent staining for study of extracellular polymeric substances in membrane biofouling layers,” Environ. Sci. Technol. 40(21), 6642–6646 (2006). [CrossRef] [PubMed]
8. N. P. Ivleva, M. Wagner, H. Horn, R. Niessner, and C. Haisch, “In situ surface-enhanced Raman scattering analysis of biofilm,” Anal. Chem. 80(22), 8538–8544 (2008). [CrossRef] [PubMed]
9. Y. Wang, W. Zhang, Z. Wu, X. Zhu, and C. Lu, “Functional analysis of luxS in Streptococcus suis reveals a key role in biofilm formation and virulence,” Vet. Microbiol. 152(1-2), 151–160 (2011). [CrossRef] [PubMed]
10. X. J. Du, F. Wang, X. Lu, B. A. Rasco, and S. Wang, “Biochemical and genetic characteristics of Cronobacter sakazakii biofilm formation,” Res. Microbiol. 163(6-7), 448–456 (2012). [CrossRef] [PubMed]
11. P. N. Tawakoli, A. Al-Ahmad, W. Hoth-Hannig, M. Hannig, and C. Hannig, “Comparison of different live/dead stainings for detection and quantification of adherent microorganisms in the initial oral biofilm,” Clin. Oral Investig. 17(3), 841–850 (2013). [CrossRef] [PubMed]
12. T. A. Smirnov, L. V. Didenko, I. G. Tiganova, S. G. Andreevskaya, N. V. Alekseeva, T. V. Stepanova, and Y. M. Romanov, “Study of the structures of biofilms formed by Salmonella typhimurium bacteria on abiotic surfaces by the methods of light and transmission electron microscopy,” Appl. Biochem. Microbiol. 46(7), 706–711 (2010). [CrossRef]
13. M. Mecozzi, M. Pietroletti, and A. Tornambè, “Molecular and structural characteristics in toxic algae cultures of Ostreopsis ovata and Ostreopsis spp. evidenced by FTIR and FTNIR spectroscopy,” Spectrochim. Acta A Mol. Biomol. Spectrosc. 78(5), 1572–1580 (2011). [CrossRef] [PubMed]
14. J. J. Ojeda, M. E. Romero-Gonzalez, H. M. Pouran, and S. A. Banwart, “In situ monitoring of the biofilm formation of Pseudomonas putida on hematite using flow-cell ATR-FTIR spectroscopy to investigate the formation of inner-sphere bonds between the bacteria and the mineral,” Mineral. Mag. 72(1), 101–106 (2008). [CrossRef]
15. K. C. Schuster, E. Urlaub, and J. R. Gapes, “Single-cell analysis of bacteria by Raman microscopy: spectral information on the chemical composition of cells and on the heterogeneity in a culture,” J. Microbiol. Methods 42(1), 29–38 (2000). [CrossRef] [PubMed]
16. P. C. A. M. Buijtels, H. F. Willemse-Erix, P. L. C. Petit, H. P. Endtz, G. J. Puppels, H. A. Verbrugh, A. van Belkum, D. van Soolingen, and K. Maquelin, “Rapid Identification of Mycobacteria by Raman Spectroscopy,” J. Clin. Microbiol. 46(3), 961–965 (2008). [CrossRef] [PubMed]
17. H. Chu, Y. Huang, and Y. Zhao, “Silver Nanorod Arrays as a Surface-Enhanced Raman Scattering Substrate for Foodborne Pathogenic Bacteria Detection,” Appl. Spectrosc. 62(8), 922–931 (2008). [CrossRef] [PubMed]
18. A. E. Grow, L. L. Wood, J. L. Claycomb, and P. A. Thompson, “New Biochip Technology for Label-Free Detection of Pathogens and Their Toxins,” J. Microbiol. Methods 53(2), 221–233 (2003). [CrossRef] [PubMed]
19. N. P. Ivleva, M. Wagner, H. Horn, R. Niessner, and C. Haisch, “Towards a nondestructive chemical characterization of biofilm matrix by Raman microscopy,” Anal. Bioanal. Chem. 393(1), 197–206 (2009). [CrossRef] [PubMed]
20. Y. Chao and T. Zhang, “Surface-enhanced Raman scattering (SERS) revealing chemical variation during biofilm formation: from initial attachment to mature biofilm,” Anal. Bioanal. Chem. 404(5), 1465–1475 (2012). [CrossRef] [PubMed]
21. B. Virdis, D. Millo, B. C. Donose, and D. J. Batstone, “Real-time measurements of the redox states of c-type cytochromes in electroactive biofilms: a confocal resonance Raman Microscopy study,” PLoS ONE 9(2), e89918 (2014). [CrossRef] [PubMed]
22. B. Virdis, F. Harnisch, D. J. Batstone, K. Rabaey, and B. C. Donose, “Non-invasive characterization of electrochemically active microbial biofilms using confocal Raman microscopy,” Energy Environ. Sci. 5(5), 7017–7024 (2012). [CrossRef]
23. T. Schwartz, C. Jungfer, S. Heissler, F. Friedrich, W. Faubel, and U. Obst, “Combined use of molecular biology taxonomy, Raman spectrometry, and ESEM imaging to study natural biofilms grown on filter materials at waterworks,” Chemosphere 77(2), 249–257 (2009). [CrossRef] [PubMed]
24. L. Rizzo, D. Sannino, V. Vaiano, O. Sacco, A. Scarpa, and D. Pietrogiacomi, “Effect of solar simulated N-doped TiO2 photocatalysis on the inactivation and antibiotic resistance of an E. coli strain in biologically treated urban wastewater,” Appl. Catal. B 144, 369–378 (2014). [CrossRef]
25. T. J. Moritz, D. S. Taylor, C. R. Polage, D. M. Krol, S. M. Lane, and J. W. Chan, “Effect of cefazolin treatment on the nonresonant Raman signatures of the metabolic state of individual Escherichia coli cells,” Anal. Chem. 82(7), 2703–2710 (2010). [CrossRef] [PubMed]
26. A. Walter, M. Reinicke, T. Bocklitz, W. Schumacher, P. Rösch, E. Kothe, and J. Popp, “Raman spectroscopic detection of physiology changes in plasmid-bearing Escherichia coli with and without antibiotic treatment,” Anal. Bioanal. Chem. 400(9), 2763–2773 (2011). [CrossRef] [PubMed]
27. X. Lu, D. R. Samuelson, B. A. Rasco, and M. E. Konkel, “Antimicrobial effect of diallyl sulphide on Campylobacter jejuni biofilms,” J. Antimicrob. Chemother. 67(8), 1915–1926 (2012). [CrossRef] [PubMed]
28. D. McDougald, S. A. Rice, N. Barraud, P. D. Steinberg, and S. Kjelleberg, “Should we stay or should we go: mechanisms and ecological consequences for biofilm dispersal,” Nat. Rev. Microbiol. 10(1), 39–50 (2012). [PubMed]
29. T. B. Rasmussen, M. E. Skindersoe, T. Bjarnsholt, R. K. Phipps, K. B. Christensen, P. O. Jensen, J. B. Andersen, B. Koch, T. O. Larsen, M. Hentzer, L. Eberl, N. Hoiby, and M. Givskov, “Identity and effects of quorum-sensing inhibitors produced by Penicillium species,” Microbiology 151(5), 1325–1340 (2005). [CrossRef] [PubMed]
30. S. W. Nam, S. T. Kim, K. M. Lee, S. H. Kim, S. Kou, J. Lim, H. Hwang, M. K. Joo, B. Jeong, S. H. Yoo, and S. Park, “N-methyl-D-aspartate receptor-mediated chemotaxis and Ca2+ signaling in Tetrahymena pyriformis,” Protist 160(2), 331–342 (2009). [CrossRef] [PubMed]
31. G. Döring, S. P. Conway, H. G. Heijerman, M. E. Hodson, N. Høiby, A. Smyth, D. J. Touw, and the Consensus Committee, “Antibiotic therapy against Pseudomonas aeruginosa in cystic fibrosis: a European consensus,” Eur. Respir. J. 16(4), 749–767 (2000). [CrossRef] [PubMed]
32. S. W. Nam, X. Chen, J. Lim, S. H. Kim, S. T. Kim, Y. H. Cho, J. Yoon, and S. Park, “In vivo Fluorescence Imaging of Bacteriogenic Cyanide in the Lungs of Live Mice Infected with Cystic Fibrosis Pathogens,” PLoS ONE 6(7), e21387 (2011). [CrossRef] [PubMed]
33. Y. Yoda, Z. Q. Hu, W. H. Zhao, and T. Shimamura, “Different susceptibilities of Staphylococcus and Gram-negative rods to epigallocatechin gallate,” J. Infect. Chemother. 10(1), 55–58 (2004). [CrossRef] [PubMed]
34. S. Choi and Z. Jiang, “Cardiac sound murmurs classification with autoregressive spectral analysis and multi-support vector machine technique,” Comput. Biol. Med. 40(1), 8–20 (2010). [CrossRef] [PubMed]
35. J. Lim, K. M. Lee, S. H. Kim, S. W. Nam, Y. J. Oh, H. S. Yun, W. Jo, S. Oh, S. H. Kim, and S. Park, “Nanoscale Characterization of Escherichia coli Biofilm Formed under Laminar Flow Using Atomic Force Microscopy (AFM) and Scanning Electron Microscopy (SEM),” Bull. Korean Chem. Soc. 29(11), 2114–2118 (2008). [CrossRef]
36. Y. T. Zheng, M. Toyofuku, N. Nomura, and S. Shigeto, “Correlation of Carotenoid Accumulation with Aggregation and Biofilm Development in Rhodococcus sp. SD-74,” Anal. Chem. 85(15), 7295–7301 (2013). [CrossRef] [PubMed]
37. M. Simões, M. O. Pereira, S. Sillankorva, J. Azeredo, and M. J. Vieira, “The effect of hydrodynamic conditions on the phenotype of Pseudomonas fluorescens biofilms,” Biofouling 23(4), 249–258 (2007). [CrossRef] [PubMed]
38. V. Janakiraman, D. Englert, A. Jayaraman, and H. Baskaran, “Modeling Growth and Quorum Sensing in Biofilms Grown in Microfluidic Chambers,” Ann. Biomed. Eng. 37(6), 1206–1216 (2009). [CrossRef] [PubMed]
39. U. Neugebauer, U. Schmid, K. Baumann, W. Ziebuhr, S. Kozitskaya, V. Deckert, M. Schmitt, and J. Popp, “Towards a detailed understanding of bacterial metabolism-spectroscopic characterization of staphylococcus epidermidis,” ChemPhysChem 8(1), 124–137 (2007). [CrossRef] [PubMed]
40. H. Schulz, M. Baranska, and R. Baranski, “Potential of NIR-FT-Raman spectroscopy in natural carotenoid analysis,” Biopolymers 77(4), 212–221 (2005). [CrossRef] [PubMed]
41. K. Maquelin, C. Kirschner, L. P. Choo-Smith, N. van den Braak, H. P. Endtz, D. Naumann, and G. J. Puppels, “Identification of medically relevant microorganisms by vibrational spectroscopy,” J. Microbiol. Methods 51(3), 255–271 (2002). [CrossRef] [PubMed]
42. N. P. Ivleva, M. Wagner, A. Szkola, H. Horn, R. Niessner, and C. Haisch, “Label-free in situ SERS imaging of biofilms,” J. Phys. Chem. B 114(31), 10184–10194 (2010). [CrossRef] [PubMed]
43. I. Notingher, S. Verrier, S. Haque, J. M. Polak, and L. L. Hench, “Spectroscopic study of human lung epithelial cells (A549) in culture: living cells versus dead cells,” Biopolymers 72(4), 230–240 (2003). [CrossRef] [PubMed]
44. C. A. Owen, J. Selvakumaran, I. Notingher, G. Gell, L. L. Hench, and M. M. Stevens, “In vitro toxicology evaluation of pharmaceuticals using Raman micro-spectroscopy,” J. Cell. Biochem. 99(1), 178–186 (2006).
45. G. B. Jung, Y. J. Lee, G. H. Lee, and H. K. Park, “A simple and rapid detection of tissue adhesive-induced biochemical changes in cells and DNA using Raman spectroscopy,” Biomed. Opt. Express 4(11), 2673–2682 (2013). [CrossRef] [PubMed]
46. P. Bhosale, “Environmental and cultural stimulants in the production of carotenoids from microorganisms,” Appl. Microbiol. Biotechnol. 63(4), 351–361 (2004). [CrossRef] [PubMed]
47. B. R. Boles and P. K. Singh, “Endogenous oxidative stress produces diversity and adaptability in biofilm communities,” Proc. Natl. Acad. Sci. U.S.A. 105(34), 12503–12508 (2008). [CrossRef] [PubMed]
48. T. R. De Kievit, R. Gillis, S. Marx, C. Brown, and B. H. Iglewski, “Quorum-sensing genes in pseudomonas aeruginosa biofilms: their role and expression patterns,” Appl. Environ. Microbiol. 67(4), 1865–1873 (2001). [CrossRef] [PubMed]
49. Y. Cui, Y. J. Oh, J. Lim, M. Youn, I. Lee, H. K. Pak, W. Park, W. Jo, and S. Park, “AFM study of the differential inhibitory effects of the green tea polyphenol-epigallocatechin-3-gallate (EGCG) against Gram-positive and Gram-negative bacteria,” Food Microbiol. 29(1), 80–87 (2012). [CrossRef] [PubMed]
