Applications of Infrared and Raman Spectroscopy to Microorganisms

Both infrared and Raman spectroscopies have been extensively applied in various research areas (Thygesen et al. 2003), such as for detection of food toxicants and chemical adulteration (Cheung et al. 2010; Lin et al. 2008; Mauer et al. 2009), bioprocessing and fermentation monitoring (Goodacre and Jarvis 2005; Brewster et al. 2009; Clarke et al. 2005; Shaw et al. 1999), enzyme activity (Hollywood et al. 2010), microorganism identification and segregation (Burgula et al. 2007; Mariey et al. 2001; Preisner et al. 2007; Maquelin et al. 2002; Naumann 2001; Harz et al. 2009; Lu and Rasco 2010a; Jarvis and Goodacre 2008; Huang et al. 2010; Davis et al. 2010), microbial cell injury identification (Lin et al. 2004; Al-Qadiri et al. 2008a), virus identification (Fan et al. 2010), and prion structure elucidation (Beekes et al. 2007). Fourier transform-infrared (FT-IR) spectroscopy has also been used to monitor microbial spoilage in food systems, such as chicken breast (Ellis et al. 2002), beef (Ellis et al. 2004), and milk (Nicolaou and Goodacre 2008). This technique can be applied directly to the surface of the food and produce biochemically interpretable “fingerprints”. However, it is an indirect method to quantify total viable bacteria and the degree of spoilage in foods through the measurement of bulk foods (Ellis and Goodacre 2001).

Advantages of Spectroscopy for Bacterial Analysis

Rapid and precise detection, segregation, and quantification of bacteria in various matrices such as foods are important. The number of reported cases of foodborne illness and other bacteria-related disease has increased dramatically over the past several years due to improvements in rapid detection, enumeration of microbes and improved epidemiological capabilities in developed countries (FDA 2009). Improved analytical methods have been employed to investigate both phenotypic and genotypic features of microorganisms and are responsible in part for our ability to detect and then trace food and waterborne disease agents in the environment. Compared to the traditional morphological and biochemical tests, recently developed genetic methods, namely analysis of 16S ribosomal deoxyribonucleic acid (DNA) or 16S ribosomal ribonucleic acid (RNA), have been designated the “gold standard” for bacterial identification. However, these genetic methods have drawbacks, specifically, that they are time consuming and require expensive reagents and expendables.

Alternative methods would improve testing efficiency, especially in clinical microbiology where rapid identification is critical, and would greatly assist with diagnosing the source of an infection. Rapid identification would minimize patient risk and expedite treatment, potentially allowing a physician to prescribe a specific antibiotic against a target bacterium rather than a wide spectrum antibiotic.

The application of bioanalytical spectroscopy to study bacteria is a relatively new frontier. Both infrared and Raman spectroscopies are forms of vibrational spectroscopy and can provide “whole organism fingerprinting” (Timmins et al. 1998) through an examination of spectral features corresponding to a wide range of important functional groups that together can provide important information about the biochemical constituents of each bacterial cell and also identify and discriminate bacteria at a species level or strain level.

Introduction to Infrared and Raman Spectroscopic Properties of Bacteria

Infrared spectroscopy was firstly introduced to study membrane composition of bacteria in the early decades of the twentieth century. Structural features of macromolecules (protein, lipid, carbohydrate, nucleic acid, etc.) associated with bacterial membranes can be observed in the mid-infrared (4,000 to 400 cm−1) spectral regions (Jiang et al. 2004). The advantages of infrared spectroscopy were immediately recognized by chemists, particularly the potential this technology provides for reagentless analysis. However, vibrational spectroscopy only became useful after the invention of the interferometer, improvements in laser design, and with them the ability to increase light intensity. Most importantly, the utilization of chemometric analysis for high-dimensional spectral analysis, which began in the 1990s (Naumann et al. 1991), made complex analyses possible. The advantage of multivariate statistics is that it permits analysis of numerous spectra simultaneously, for example, to utilize Fourier transformation to decode interferograms collected from infrared light transmissions. Spectroscopic methods became popular in the 1980s in food analysis for prediction of the macronutrient content of foods such as the proximate composition of cereal grains, and from there to applications in food microbiology. The first work in clinical and environmental microbiology began at about the same time. These initial works resulted in a boom in industrial applications over the past 20 years. Recent advances in detector technology have made the analysis of biological samples simpler and faster (Richardson et al. 1998). The attenuated total reflectance (ATR) accessory began to replace the diffuse reflectance infrared detector around 1,993, and this made it easier to analyze aqueous samples by infrared spectroscopy. The ATR accessory shortened the time needed for sample preparation (Movasaghi et al. 2008), increased the durability of IR instruments, and reduced the analysis cost.

Most of the applications of IR spectroscopy focus on the mid-infrared region. Mid-IR spectra can be divided into four major regions. Aliphatic C―H stretching modes absorb in 3,000–2,800 cm−1 (region 1); although most components of food (proteins, carbohydrates, etc.) contain aliphatic C―H groups, this region is most frequently correlated with fatty acids. Region 2 (1,800–1,500 cm−1) contains the C═O stretching band of lipids at ca 1,740 cm−1 and the amide I and II bands of proteins and peptides at ca 1,650 and 1,550 cm−1, respectively. The amide bands provide structural information about α-helix, β-sheet and random coil conformations in proteins. Many cellular components have absorption bands between 1,500 and 1,200 cm−1 (region 3), for example, nucleic acids and phospholipids. The spectral interval between 1,200 and 900 cm−1 (region 4) provides information about the structure of polysaccharides.

However, there are some limitations for spectral analysis of biomolecules. Infrared absorption requires that a change in the intrinsic dipole moment occur with molecular vibrations; therefore, polar groups such as C═O, N―H, and O―H have strong IR stretching vibrations. Water is an example of a polar molecule that possesses strong infrared absorption and the interference of water with other spectral features can be compensated for when biological samples are examined. Furthermore, the detection limit with mid-IR methods is often not sufficiently low to quantify small levels of analytes compared with other techniques, such as immunochemical and nucleic acid analysis. A concentration of 103 colony forming units (CFU)/ml is generally the lowest limit of detection for identification of specific bacterial spectral features by infrared spectroscopy, similar to what is commonly anticipated for immunochemical detection. Detection limits can be improved by an order of magnitude if cells can be collected on a matrix such as a membrane filter under controlled cultivation conditions and in a relatively pure state, particularly if the cells can be partially dehydrated prior to analysis (Burgula et al. 2007). It is critical to concentrate bacteria and remove as much water from the sample as practical (Pearman and Fountain 2006) for qualitative as well as quantitative analysis. Recent developments in infrared spectroscopic characterization of bacteria have attempted to overcome some of these disadvantages by: (1) applying surface-enhanced infrared absorbance techniques to amplify the intensity of spectra derived from bacteria (Holman et al. 1998); and (2) using magnetic nanoparticles to bind to targeted microorganisms, isolate, and concentrate them, and then distinguish spectral features in the infrared region (Ravindranath et al. 2009).

Along with conventional infrared spectroscopy, FT-IR microscopy is an excellent tool extensively employed to identify and segregate bacteria. This technique combines FT-IR spectrometry with a microscope system to recover cells after enrichment on agar for spectral analysis (Maquelin et al. 2002; Ngo-Thi et al. 2003). This technique has been recently used to discriminate between various strains of pathogenic and spoilage microbes in food matrices, including Bacillus spp. in tryptic soy broth (Grasso et al. 2009b), Alicyclobacillus spp. in fruit juice (Grasso et al. 2009a), and Salmonella spp. in selected foods (Mannig et al. 2008).

In addition to IR spectroscopy, Raman spectroscopy is gaining increasing attention. Raman provides complementary information to conventional infrared spectroscopy, and coupling these techniques together is often advantageous as described below. Raman scattering relies on changes in the polarizability of functional groups as atoms vibrate; thus, nonpolar groups such as C―C and S―S have intense Raman bands. Raman spectroscopy measures the inelastic scattering of radiation of monochromatic light, yielding a spectral shift (called a “Raman” shift) that results from the interaction of light with electron clouds surrounding molecular bonds (Kneipp and Kneipp 2006). A Raman spectrum can often offer detailed and readily interpretable chemical information. Table 1 summarizes the major infrared and Raman bands that serve as the basis for a reference library for spectral interpretation in microbiological analysis (Maquelin et al. 2002; Movasaghi et al. 2008; Huang et al. 2010).

Table 1 Assignment of some bands frequently found in FT-IR and Raman spectra respective of biological specimens (a revision of Maquelin et al. 2002)
Full size table

Raman spectroscopy provides some major advantages over infrared spectroscopy for investigations of biological samples (Jarvis et al. 2004) since interference of water spectral features is less problematic. In addition, more spectral features are detectable in a Raman spectrum than in an infrared one over the same wavenumber range. Raman bands tend to be narrower than those in the mid-IR range because Raman laser/detector offers greater specificity. There is a wider range of potential analytical wavelengths and laser resources (UV, near infrared or visible) for Raman than for infrared (Jarvis and Goodacre 2004). However, it may not be prudent to choose Raman over infrared since IR spectrometers are generally simpler to operate and, at least for the present time, less costly than Raman instruments.

Figure 1 presents different vibrational spectra of Streptomyces pseudovenezuela that the Raman spectrum has a higher intensity at the specific wavenumber compared to IR (Fig. 1a, b) and different laser sources will provide different scattering cross sections at specific wavenumber regions (Fig. 1b, c). Using a combination of both infrared and Raman spectroscopies is powerful since the techniques are complementary. The spectral variations of different bacteria are shown in Fig. 2. A comparison of different types of Raman spectroscopy is provided in Table 2 (Ellis and Goodacre 2006). For example, UV-Raman and surface-enhanced techniques used together can increase sensitivity because interference from fluorescence can be reduced (Huang et al. 2004; Jarvis and Goodacre 2004). This is extremely useful when studying bacteria, which exhibit a high fluorescence background under excitation in the near infrared to visible regions of the electromagnetic spectrum, resulting in a troublesome analysis because of this form of interference (Knauer et al. 2010).

Fig. 1
figure 1

Different vibrational spectroscopic spectra of S. pseudovenezuela (a) IR absorptions spectrum, (b) micro-Raman spectrum with an excitation wavelength of 532 nm, (c) UV-resonance Raman spectrum with an excitation wavelength of 244 nm (citing Harz et al. 2009)

Full size image
Fig. 2
figure 2

Representative FT-IR spectra (4,000–800 cm−1) of control (apple juice) (a), E. coli O157:H7 ATCC 35150 (b), Alicyclobacillus acidoterrestris 1016 (c), Alicyclobacillus spp. C-Fuji -6 (d), and the mixed culture of both Alicyclobacillus acidoterrestris 1016 and Alicyclobacillus spp. C-Fuji-6 (e) recovered from inoculated apple juice (citing Al-Qadiri et al. 2006a)

Full size image
Table 2 Features of various Raman spectroscopies (a revision of Ellis and Goodacre 2006)
Full size table

Raman signals are weak, but new advances have made it possible to increase the resultant signal. Roughly one in 108 incident photons are inelastically scattered, but the resultant signal can be enhanced several orders of magnitude if a laser wavelength possesses an intense electronic absorption band for a particular chromophore of interest. Also, if an analyte is attached to, or is perhaps microscopically close to, a suitably roughened surface (substrate), vibrational mode coupling can also be enhanced. This technique is called surface-enhanced Raman scattering (SERS) (Sengupta et al. 2006; Sujith et al. 2009; Premasiri et al. 2005). SERS substrates consist of either silver or gold nanomaterials—either colloidal or extended surfaces with nanoscale morphologies, both of which are appropriate for use with biological samples, making it possible to increase the Raman intensity by perhaps as much as 1015-fold (Chu et al. 2008). Theoretically, Raman spectroscopy can detect the spectral properties of a single bacterial cell (Patel et al. 2008a, b).

Like IR, a Raman spectrometer can be coupled with a microscope. The confocal Raman microscope was invented in 1990 and has been used for recording Raman spectra for a single human cell (Puppels et al. 1990). The invention of the confocal microscope provides many advantages, and integration of this capability with Raman spectroscopic detection allows for spatially precise measurements since the scattered radiation would be collected from light very close to the focal plane of the microscope. Raman microscopy allows for examination of defined optical sections of two-dimensional or three-dimensional mapping of very precise areas within a sample (Schuster et al. 2000; Kalasinsky et al. 2007), for example: detection of single bacteria within a biofilm, differentiation of bacterial cell wall components from those in the cytoplasm, and monitoring bacterial cellular response due to environmental stress (Xie et al. 2003; Jarvis et al. 2008).

A Short Introduction to Chemometrics

Advanced data preprocessing and statistical analysis are required to interpret infrared and Raman spectra so that minor differences in the spectral features in a biological specimen can be distinguished. It is important to apply advanced data preprocessing and statistical analysis to interpret the raw spectra. Analysis of a “whole organism fingerprint” with vibrational spectroscopic methods yields a high dimension vector containing multiple dependent variables (or wavenumbers) (Shaw et al. 1999).

Chemometrics describes the multivariate statistical analysis that can be used to reduce this multidimensional information to a few, independent latent variables that preserve the most relevant and representative information (designated as the first several principal components). These principal components can then be used to segregate and quantify analytes based upon specific calibration models (Ellis and Goodacre 2006; Lin et al. 2009a, b). Chemometric methods are usually divided into unsupervised and supervised methods (Goodacre 2003). Unsupervised chemometrics are employed in food microbiology to segregate different bacterial species and strains into distinct groupings, either in the form of a cluster (principal component analysis, PCA) (Lu et al. 2010a) or a dendrogram (hierarchical cluster analysis, HCA) based upon biochemically distinctive characteristics reflected with the major latent variables (Fig. 3, PCA and HCA). The latent variables reflect which specific biochemical constituents within bacterial cells contributed to the segregation (Jarvis et al. 2006a). However, precise results are sometimes not achievable, and thus supervised chemometrics models are necessary. Supervised models require that reference values (i.e., concentration of specific lipid or protein components, nucleic acid sequences) be available for correlation with spectra. Supervised chemometrics can provide either qualitative analysis, for example, to predict if a specific foodborne pathogen is present or not, or to identify what microorganisms are present within a food sample at the strain level. Generally, supervised chemometric models are used for quantitative analysis, such as predicting the number of CFU in a biological sample such as food (Goodacre et al. 2004; Lu et al. 2010e). Each type of a supervised chemometric model is built upon a prior unsupervised chemometric one. For example, a PCA model for latent variable selection is commonly used to select the number of latent variables by either discriminant function analysis (DFA) or a pass–fail test, such as soft independent modeling of class analog (SIMCA; Lu et al. 2010b). Quantitative prediction is usually achieved by developing a linear regression model using appropriate reference data (partial least squares regression, PLSR; Fig. 4, DFA, SIMCA and PLSR).

Fig. 3
figure 3

a Principal component analysis (PCA) for control (blank without bacteria) (A), E. coli ATCC 25922 (B), P. aeruginosa (C), and mixed (1:1, v/v) culture of E. coli and P. aeruginosa (D). Groups were tightly segregated and significantly different with each other (P < 0.05; citing Al-Qadiri et al. 2006b). b Dendrogram of hierarchical cluster analysis (HCA) performed on FT-IR spectral data of eight Alicyclobacillus isolates (citing Lin et al. 2007)

Full size image
Fig. 4
figure 4

a A composite dendrogram generated by HCA using the combined PC-DFA space from the training (red color) and validation (blue color) replicates for different clinical isolates of E coli. (citing Jarvis and Goodacre 2004). b SIMCA classification of E. coli ATCC 25922 (B) as compared to control (blank without bacteria) (A), P. aeruginosa (C), and mixed (1:1, v/v) culture of E. coli and P. aeruginosa (D) (citing Al-Qadiri et al. 2006a, b). c Comparison between measured and predicted bacterial count (log10 cfu/ml) for PLSR of S. enterica serotype Typhimurium ATCC 14028: control (no thermal treatment) (A), 2D (B), 4D (C), 6D (D), and 8D (E) (citing Al-Qadiri et al. 2008a, b). *D refers to decimal reduction time: the time required at a certain temperature to kill 90% of the organisms being studied

Full size image

The rigor used for the development of supervised models is critical. To obtain a reliable prediction result, each model needs to be first calibrated, followed by cross-validation then prediction of the analyte concentration in the unknown samples. For calibrations, a set of samples with predetermined reference values must be available for relatively large numbers of samples, which contain concentrations of the analyte over the range of interest and are preferably evenly spaced across the concentration range. This is necessary if the calibration model is to provide sufficient predictive ability (Alsberg et al. 1998). It is worth noting that overfitting spectra can provide a cross-validation model with a high correlation coefficient (Lu and Rasco 2010b); however, the ability to predict the analyte concentration within a new sample will be compromised or lost (Jarvis et al. 2006b). Overfitting often involves developing a chemometric model that contains extra and irrelevant components, and this reduces the ability of the analyst to detect errors in the database that would lead to prediction errors or could lead to the loss of valuable data. Furthermore, adding additional predictive coefficients into a model may mean that random variation is incorporated into future predictions (Hawkins 2004).

In the future, advanced chemometric skills are continuously needed to combine multidimensional data from different resources to analyze. A representative example was to combine data from vibrational spectroscopy and DNA microarray results to comprehensively study bacteria inactivation and injury on both chemical and genetic levels (Moen et al. 2005; Oust et al. 2006; Moen et al. 2009).

Identification and Speciation of Microorganisms

Because vibrational spectroscopy segregates microorganisms based upon the biochemical composition of bacterial cell membranes, even minor differences among the same species (different strains) can be determined. Infrared and Raman spectroscopies coupled with supervised and unsupervised chemometric models have been used to study vegetative bacterial cells, including Listeria monocytogenes (Al-Holy et al. 2006, IR; Green et al. 2009, Raman), Salmonella enterica (Mannig et al. 2008, IR; Liu et al. 2009a, b), Escherichia coli O157:H7 (Al-Qadiri et al. 2006a, IR; Efrima and Zeiri 2009, Raman), Clostridium botulinum (Kirkwood et al. 2006, IR), Pseudomonas aeruginosa (Al-Qadiri et al. 2006b, IR; Huang et al. 2007, Raman), Enterococcus faecium (Goodacre et al. 1996, IR; Guibet et al. 2003, IR; Kirschner et al. 2001, Raman), Staphylococcus sp. (Harz et al. 2005, Raman; Amiali et al. 2007, IR), Brucella sp. (Miguel Gomez et al. 2003, IR; Kalasinsky et al. 2007), Acinetobacter sp. (Maquelin et al. 2006, Raman; Winder et al. 2004, IR), Enterobacter sakazakii (Lin et al. 2009a,b, IR), Yersinia enterocolitica (Castro et al. 2010, IR), Alicyclobacillus (Lin et al. 2005, IR; Lin et al. 2007, IR; Grasso et al. 2009a, IR), lactobacilli (Oust et al. 2004) and yeast (Kummerle et al. 1998, IR; Sayin et al. 2009, Raman). In addition, Bacillus cereus spores (Alexander and Le 2007, Raman; Goodacre et al. 2000, IR; He et al. 2008, Raman; Daniels et al. 2006, Raman; Cheng et al. 2009, Raman) and microbial biofilms (Ivleva et al. 2008, Raman; Bosch et al. 2006, IR) have also been studied by vibrational spectroscopy. Detailed information for the analysis of individual bacteria by IR and Raman is summarized in Table 3 with representative publications cited in this table for infrared and Raman spectroscopic methods.

Table 3 Vibrational spectroscopic investigation for selected microorganisms
Full size table

Biofilm Detection by Vibrational Spectroscopy

A biofilm is an assemblage of microorganisms in which cells are associated with each other and adhere to a surface. Those adherent cells are embedded within a matrix of extracellular polymeric substance (EPS). An EPS biofilm provides a suitable environment for the survival, growth, and genetic material exchange among the cells contained within it. The cells of microorganisms growing in a biofilm are physiologically distinct from planktonic counterparts of the same organisms (Donlan and Coesterton 2002).

FT-IR coupled with chemometrics has been used to determine macromolecular compositions of pathogenic microorganism biofilm matrices and also for monitoring the kinetics of development and maturation of biofilms from bacteria and microeukaryotes, including fungi, algae, and protozoa (Nivens et al. 1995). The spectra obtained from the biofilm contain biochemical information of both the sessile cells and the polymeric matrix of the embedded microbe (Geesey and Suci 2000). Nichols et al. (1985) used diffused reflectance/FT-IR spectroscopy to study the adhesion and biofilm formation of Caulobacter crescentus by ATR (Nivens et al. 1995; Nivens et al. 1993a,b). They also studied the role of alginate, an extracellular polysaccharide, and its O-acetylated derivative on the formation of P. aeruginosa biofilms (Nivens et al. 2001). Until recently, little work has been conducted on the application of spectral analysis of food pathogen colonization within biofilms. However, it is anticipated that the mechanism of attachment to abiotic surfaces and bacterial growth within biofilms may be similar between Gram-negative pathogens, for example, between Bordetella sp. described below and Enterobacter sp., Escherichia sp. or Pseudomonas sp.

Bordetella pertussis, a Gram-negative pathogen, which colonizes the respiratory tract and is responsible for whooping cough, has been the subject of a flurry of recently intensive studies of its infection mechanism, especially the initial phase that involves adherence of the microbes to epithelial tissues and their growth as a biofilm. Bosch et al. (2000) used FT-IR to investigate how B. pertussis can grow attached to an abiotic surface and how this attachment can foster production of exopolymers that then lead to formation of a biofilm. Later on, Bosch et al. (2006) studied the characterization of B. pertussis growing as a biofilm by FT-IR. The IR spectra of the biofilm cells showed higher intensity in the absorption bands assigned to the vibrational modes of polysaccharides (1,200–900 cm−1) and carboxylate groups (1,627, 1,405, and 1,373 cm−1) compared with the spectra of planktonic cells. Uronic acid containing polysaccharides, proteins, and lipopolysaccharides were detected in the biofilm matrix. Serra et al. (2007) verified that FT-IR is not only a useful method for studying the dynamics of B. pertussis biofilm growth through specific biomass and biofilm marker absorption bands, but that this technique can also be employed to monitor the maturation of biofilms by following increases of the carbohydrate to protein ratio. Serra et al. (2008) combined proteomic approaches with matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry and ATR/FT-IR spectroscopy to study the biosynthesis of a putative acidic-type polysaccharide polymer as the most distinctive trait of B. pertussis life inside a biofilm and also how biofilm development impacts B. pertussis pathogenesis.

Pseudomonas fluorescens and Streptococcus pneumoniae biofilms have been studied in situ by FT-IR (Donlan et al. 2004; Delille et al. 2007). Suci et al. (1994, 1997, 1998) investigated the interactions between antimicrobial agents (i.e., ciprofloxacin) and P. aeruginosa biofilms by FT-IR. They provided detailed information on the transport limitations of antibiotics to bacteria embedded in the biofilm and the corresponding chemical modifications of the biofilm. Schmitt et al. (1995) studied P. putida biofilm response to toluene. The spectra showed a significant increase of EPS-polysaccharide and carboxyl group formation at a toluene level of higher than 5 mg L−1.

Studying Antibiotic Properties and the Development of Antibiotic Resistance by Vibrational Spectroscopy

Development of spectroscopic methods has the potential to shorten diagnosis time and allow physicians to quickly determine appropriate antibiotic treatments, potentially reducing the risk of creating antibiotic resistance in microbes. This line of research is also important for elucidating mechanisms of pathogen susceptibility to antibiotics and as a means to improve inhibitory efficiency of selective antibiotics. The resistance of certain pathogens to antibiotics is increasing from over-utilization of wide-spectrum antibiotics, causing potential risk to human health. This is of concern in food science since it is possible that food could be a source of transmission of antibiotic-resistant bacteria. Traditional microbiological diagnostic methods require a long time (2–3 days) for detection, involving extracting pathogen cells from either blood or urine, enrichment culturing for several hours to a couple of days, then testing by various methods, such as traditional plating or polymerase chain reaction (PCR) techniques. Ideally, a shortened diagnostic time could help doctors to prescribe a specific antibiotic that can inhibit the growth of specific pathogens and decrease the resistance of pathogens to antibiotics (multidrug resistance). Maquelin et al. (2000) and Choo-Smith et al. (2001) demonstrated that by using confocal microscopy coupled with Raman spectroscopy, spectra can be acquired from a small biomass of microorganisms cultured in 6 h or from solid culture media. A similar study performed at the Robert Koch Institute (Berlin, Germany) using FT-IR yielded equally good results within ~8 h (Kirschner et al. 2001; Maquelin et al. 2003). Recently, this technique with the aid of HCA was applied to rapidly monitor hospital-acquired infections (118 Staphylococcus aureus isolates) within a day, achieving early detection of potential outbreaks and leading to the initiation of routine epidemiological monitoring of bacterial infections in hospitals (Willemse-Erix et al. 2009).

The mode of action of specific antibiotics to targeted pathogens has been studied by both Raman and infrared spectroscopies by monitoring changes of various macromolecular constituents on cell membranes, which are important for antibiotic binding and intracellular transport. UV resonance Raman was conducted at 244 nm to monitor the concentration effect of amikacin on P. aeruginosa cells (Lopez-Diez et al. 2005). The clustering pattern in the discriminant factors space correlated directly to the concentration of amikacin. Using PLSR analysis of the UV resonance Raman spectra, these investigators were able to predict the concentration of amikacin to which bacterial cells had been exposed.

Excitation of bacteria at wavelengths between 222 and 257 nm provide Raman spectra with prominent nucleic acid and aromatic amino acid features (Lopez-Diez and Goodacre 2004; Huang et al. 2004) indicative of the synthesis of proteins and nucleic acids that are targets of antimicrobial drugs. Therefore, it is reasonable to expect that UV resonance Raman spectra may provide useful information in studies of the mode of action of various antibiotics. For example, the mode of action of drugs from the fluoroquinolone group to Bacillus pumilus was studied using UV resonance Raman at 244 nm (Neugebauer et al. 2006). Small changes in spectral features could be monitored due to the interaction of the drug with its biological targets of DNA and the enzyme gyrase within the cell. The same research group continued similar investigations on the mode of action of moxifloxacin to Staphylococcus epidermidis by both infrared and UV resonance Raman (532 nm for DNA and 244 nm for aromatic amino acids) (Neugebauer et al. 2007). The wavenumbers mostly affected by the action of the drug could be assigned to protein and DNA moieties.

Liu et al. (2009a,b) employed an extensive study of antibiotics (oxacillin, ampicillin, vancomycin, cefotamicin and tetracycline) inducing chemical changes of cell walls in selective bacteria, many of which are commonly associated with foods (Staphylococcus aureus, Enterococcus feacalis, L. monocytogenes, E. coli, Serratia marcescens and Klebsiella pneumoniae) by SERS. ATR-FT-IR was used to investigate the dynamic changes of P. aeruginosa cell membrane treated with the antibiotic imipenem (Sockalingum et al. 1997). These authors validated the primary mechanism of imipenem resistance, ascribing it to outer membrane impermeability owing to a loss of expression of certain proteins. It would be interesting to investigate the effect of bioactive compounds such as those contained within nutraceutical ingredients and to study the ability of these natural compounds to suppress pathogens with or without the aid of antibiotics (Kyung and Lee 2001) using vibrational spectroscopy (Lu et al. 2010f). For example, the bioactive sulfur compounds extracted from vegetables have been shown to have greater inhibitory activity on the multiplication of Helicobacter pylori than antibiotics, and it is likely that bioactive components, including sulfur-containing compounds from Allium sp. and cruciferous vegetable, may have a synergistic effect (O’Gara et al. 2000) and exhibit antimicrobial activity.

Studies of Bacterial Injury and Inactivation by Vibrational Spectroscopy

Many factors can cause microbial cells injury, including sonication, acid or alkaline treatment, exposure to chemical sanitizers and antimicrobial agents such as bioactive peptides, thermal treatment, freezing, dehydration, irradiation, high hydrostatic pressure, etc. Injured bacterial cells pose a potential threat to food safety since they can resuscitate under suitable environmental conditions and grow in food during storage (Wu 2008). In addition, traditional microbiological procedures may not provide positive results for sublethally injured cells, and laborious overlay methods are required to confirm that the injured microorganisms are “viable but not culturable”. Recent research indicates that bacterial cell injury can be studied using vibrational spectroscopy (Lin et al. 2004; Al-Qadiri et al. 2008a,b). Lin et al. (2004) discriminated intact L. monocytogenes cells from sonication-injured ones by infrared spectroscopy using PCA segregation clusters, noting that injury was caused by macromolecular shearing and subsequently redistribution of cell wall components along with possible denaturation of intracellular proteins. Later, sublethal thermal injuries in S. enterica serotype Typhimurium and L. monocytogenes were detected by ATR-FT-IR spectroscopy (Al-Qadiri et al. 2008a). The mechanism of bacterial cell injury is different for Gram-positive and Gram-negative microorganisms based upon variance in spectral features between injured and intact cells. In addition, cluster analysis and SIMCA can be employed to sort groups of microbes based upon the degree of heat-injury with linear regression applied to quantify and predict the actual number of injured microbial cells with a high correlation coefficient (R = 0.97 (S. enterica serotype Typhimurium) and 0.98 (L. monocytogenes)). Al-Qadiri et al. (2008b) also studied the effect of chlorine-induced bacterial injury in water by infrared spectroscopy, with both PCA and SIMCA methods being successful in differentiating treatment groups according to the various degrees of injury caused by different chlorine concentrations.

Bacterial injury under different forms of environmental stresses can be studied using FT-IR and the appropriate chemometric methods. Sublethal injuries of Campylobacter jejuni, E. coli O157:H7 and P. aeruginosa under cold and freeze stresses have been studied by infrared spectroscopy. Different segregation analyses (cluster analysis, dendrogram analysis, class analog analysis and loading plot analysis) indicate that polysaccharides were produced by bacteria in response to harsh environmental stress, and the content of lipids and proteins decreased substantially under freeze stress (Lu et al. 2010c). The inhibitory effect of metal nanoparticle addition (i.e., zinc oxide nanoparticles) on the growth of L. monocytogenes, E. coli O157:H7 and C. jejuni has also been studied. Cell membrane leakage and structural variation under different concentrations of metal nanoparticles could be monitored using FT-IR (Lu et al. 2010d). Furthermore, the sublethal injury of Listeria under organic acid can be monitored in the mid-IR range. PCA and SIMCA were performed to segregate pathogens according to different acid injury levels (Wu et al. in prep).

Recently, nanoparticles made from metal oxides (less than 100 nm) inhibited the growth of pathogens, and the mechanism was studied using vibrational spectroscopy (Liu et al. 2009a,b). The damage of bacterial cell membrane and the leakage of intracellular contents of E. coli O157:H7 induced by zinc oxide nanoparticles have been monitored by Raman spectroscopy. The changes of spectral features (mainly in lipid regions) indicated that cell damage was proportional to nanoparticle concentration. The authors also observed increased Raman intensity for lipids and some proteins but not for nucleic acids as a function of nanoparticle concentration indicating that nanoparticle-induced injury is likely caused by a different mechanism than antibiotic treatment in which protein and nucleic acid synthesis are inhibited. In other studies, FT-IR was used to study heat-induced lethal and sublethal injuries in Lactococcus lactis (Kilimann et al. 2006), osmotic and thermally induced injuries of E. coli (Mille et al. 2002), UV radiation-induced injury of Staphylococcus aureus (Krishnamurthy et al. 2010), lethal injury in E. coli caused by dehydration (Beney et al. 2004), radical-induced damage of Micrococcus luteus (Lorin-Latxague and Melin 2005), acid tolerance response of Streptococcus macedonicus (Papadimitriou et al. 2008), and toxicity response of yeast (Corte et al. 2010). Recently, the membrane phase behavior of E. coli during desiccation and rehydration was also studied by FT-IR (Scherber et al. 2009).

Recent examinations of the properties of microbial spores have been conducted using vibrational spectroscopy. In one study, confocal micro-Raman spectrometry was employed to explore the inactivation effect of 20% formaldehyde, 2% peracetic acid, and autoclaving on single Bacillus sp. endospores (Stockel et al. 2010). Protein damage, but not DNA damage, was observed in injured spores with a considerable decrease of calcium dipicolinate as treatment intensity increased. Subramanian et al. (2006, 2007) used infrared spectroscopy coupled with PCA and PLSR to determine spore inactivation and predicted viable spore concentrations (Clostridium tyrobutyricum, Bacillus sphaericus, and three strains of B. amyloliquefaciens) in samples treated by pressure-assisted thermal processing (PATP) and by thermal processing alone. The changes in dipicolinic acid (DPA) concentration and α-helix and β-sheet conformations of secondary proteins were evident during PATP. The DPA content in spores is closely related to its resistance to PATP. FT-IR spectroscopy and microspectroscopy were also used to study the inactivation effect of pharmaceuticals to ubiquitous environmental aquatic microorganisms, through which the potential ecological risk could be validated (Patel et al. 2008b; Wharfe et al. 2010).

During the recent several years, DNA microarray techniques were extensively used to study bacterial injury resulting from various types of environmental stress, such as acid, heat and cold treatments. Moen et al. (2005) and Oust et al. (2006) performed DNA microarray analysis and FT-IR to study the survival mechanims of C. jejuni to various unfavorable environmental stresses and found that survival was associated with downregulation of genes associated with metabolic functions to save energy and to produce polysaccharides and oligosaccharides for protection against environmental stress. They also demonstrated that FT-IR could provide important information about stress response that was not detectable by DNA microarrays. Recently, this group studied responses of E. coli to ten different adverse conditions by DNA microarray and FT-IR (Moen et al. 2009), and they asserted that the gene expression and biomolecular response of E. coli to most stress and inactivation were not well correlated, which indicated that multifactorial analysis via a combination of different types of techniques based on genetic analysis, chemistry and cell morphology, provides the strongest basis for studies on bacterial injury and survival in response to unfavorable environmental conditions.

Conclusion and Future Trends

Vibrational spectroscopy is a technique for both the identification of bacteria including bacterial spores and biofilms and for studies involving cell injury and inactivation, which have important ramifications in food analysis. Raman and FT-IR provide complementary information about the biochemical properties of cells. FT-IR has the greatest promise for rapid analysis when a relatively large mass of cells is available (~103–104 cells). In contrast, Raman spectroscopy coupled with microscopy offers the ability to obtain substantial information about the biochemical properties of a single cell. Spectroscopy provides a basis for studies of bacterial injury and inactivation but is substantially enhanced when complementary techniques are used that can provide greater insight regarding complicated interactions of a living organism with its environment. Future trends will involve multifactorial approaches to investigate survival and injury of bacteria following environmental stress, such as would be experienced by refrigeration, chemical treatment, or a food pasteurization treatment.

Utilizing a combination of different methods will make it possible to better understand the underlying causes of bacterial cell injury. Clearly, vibrational spectroscopy is an established technique for identification and discrimination of bacterial strains with a number of comprehensive reviews available (Lu and Rasco 2010a; Maquelin et al. 2002; Naumann 2001; Harz et al. 2009) indicating which biochemical properties of microbial cells can be best determined using this method. However, vibrational spectroscopy is most valuable if it is coupled with other microbial analytical methods. Traditional plating techniques involving both selective and non-selective media would provide baseline data as to the quantity of injured cells with other methods incorporated with plating to increase our understanding of the mechanisms of cell injury and develop successful “hurdle technologies” to maximize both food safety and quality. DNA microarray methods could be used along with either traditional methods or vibrational spectroscopy to elucidate gene expression patterns, or electron microscopy for examination of cell morphology. Applications of these methods to study the microbial ecology of food systems are still developing and should provide new advances in our understanding of the effectiveness of minimal and coupled processes, biofilm formation, and the use of natural antimicrobials, all of which are important considerations for improving the safety of our food supply.