1 Introduction

Rapid identification and typing microorganisms is an increasingly important task in food safety and epidemiological diagnosis ^[1]. Traditional bacteria classification methods, such as serological test, PCR, DNA-DNA hybridization, and DNA sequencing^[2], are always time-consuming and need laborious culturing procedures. Moreover, these phenotype or gene based methods have difficulty in discriminating closely related strains which have similar phenotypic or genotypic properties ^[1].

Fourier transform infrared (FTIR) spectroscopy has emerged as a real alternative of molecular biology and biochemical methods to describe microorganisms dependent on chemical components ^[3-5]. Due to the advantages of short analysis time and low cost ^[6], since it was introduced to characterize microorganisms by Naumann and co-workers ^{[7, 8]}, FTIR spectroscopy was successfully applied to discriminate bacteria at genus, species, and even subspecies level ^[9]. The commonly used infrared spectra (Figure S1, Supporting Information) include a lipid region, protein region, mixed region, and polysaccharide region (3000-2800, 1800-1500, 1500-1200, and 1200-900 cm^-1, respectively) ^[10]. Since infrared spectra contain complex chemical information, multivariate statistical analysis methods, such as principal component analysis (PCA) and hierarchical cluster analysis (HCA) are usually applied to explain their diversities and to further type FTIR spectra of bacteria ^[11-13].

The development of a synchrotron source for infrared spectroscopy has greatly improved the accuracy and signal-to-noise ration of the measurements ^[14].In the previous work, we have successfully applied synchrotron radiation based FTIR (SR-FTIR) spectroscopy to discriminate 10 bacterial strains. Herein, we tried to further classify and type 16 foodborne bacteria from seven genera by the combination of SR-FTIR and chemometric methods. Whole spectra (3000-2800, 1800-900 cm^-1) and spectra of four divided regions (lipid region, protein region, mixed region, and polysaccharide region) were used, respectively. We found that the spectra region between 1300 and 1000 cm^-1 can be recognized as the signature region to cluster the bacteria from the same taxonomic levels into one group. With high reproducibility and a low sample amount, SR-FTIR spectroscopy shows great potential in fast identifying and typing microorganisms based on the specific infrared signature.

2 Experimental Section

3 Results and Discussion

Spectra were recorded from 16 different bacterial isolates (Table S1, Supporting Information). A typical SR-FTIR spectrum of Vibrio parahaemolyticus was shown in Fig.S1 (Supporting Information). The recorded spectra of these 16 bacterial strains showed similar bands in the whole spectral region (3000-2800 and 1800-900cm^-1) and little differences were observed (Fig.S2a, Supporting Information). To further resolve the spectral differences among these strains, the first derivative transformations were performed. As shown in Fig.S2b (Supporting Information), the first derivatives of reduced FTIR spectra displayed minor differences in the mixed region (1500-1200 cm^-1) and in the polysaccharide (1200-900 cm^-1) region.

To find out the better discriminating results, the first derivative spectra of the 16 bacterial strains were used and multivariate analysis methods PCA and HCA were carried out on the whole spectral region and specific regions (such as lipid region, protein region, mixed region and polysaccharide region), respectively.

We firstly tried to carry out PCA analysis on whole spectra of 16 bacteria while performing HCA to explain their affinity-disaffinity. As shown in Fig.1, it can be seen that all Listeria strains are well separated by principle component one and two. The dendrograms from HCA of the average spectra showed close clustering of these species, suggesting that Listeria strains have a similar biochemical profile and all of them are from the same order. Two Staphylococcus strains were also grouped together. However, Both Vibrio strains and Salmonella strains were decentralized and couldn't cluster into their corresponding group. These results indicated the whole spectra could be difficult to distinguish all the bacterial strains at distinct taxonomic levels.

Fig.1Full-screen View

Score plots from a PCA (first and second components, left) and HCA (right) of the whole infrared spectra of the 16 bacterial strains.

Since the infrared spectra of the microorganism reflect the information of several classes of biomolecules, PCA and HCA were carried out on these specific spectral regions, respectively. For the lipid region (3000-2800cm^-1) and protein region (1800-1500cm^-1), most strains with the same genus didn’t cluster correctly and only Listeria strains were partly grouped together (Fig.S3 and Fig.S4, Supporting Information). These results indicated that either the lipid region or protein region contained low specific information among these strains and couldn’t be used to classify bacteria.

We further typed the bacteria based on the spectra of the mixed region 1500-1200 cm^-1 (Fig.2). PCA displayed that four Listeria strains and three Vibrio strains were significantly separated from other bacteria. HCA illustrated that Listeria, Vibrio, Staphylococcus, and most Salmonella (except S. dysenteriae) were formed separate genus-specific clusters. Although not all strains were classified into their corresponding sub-clusters, this result was very close to the phylogenetic relatedness. Interestingly, similar results were obtained after PCA and HCA analysis of the spectral region 1200-900cm^-1, which mainly represent the information about nucleic acids and polysaccharides components (Fig. 3). For example, four Listeria strains were clearly differentiated from the others and three of four Salmonella species were classified into a same cluster. However, some strains, such as Staphylococcus epidermidis, S. typhimurium (CICC10420), and Vibrio vulnificus, were misclassified.

Fig.2Full-screen View

PCA (left) and HCA (right) of the 16 bacterial strains based on the infrared spectra of the mixed region (1500-1200cm^-1).

Fig.3Full-screen View

PCA (left) and HCA (right) of the 16 bacterial strains based on the infrared spectra of the polysaccharide region (1200-900cm^-1).

The above results suggested that the spectra information in the mixed region (1500-1200cm^-1) and polysaccharide region (1200-900cm^-1) had relatively good discriminating capability, which have been previously reported as the specific regions for characterizing these kinds of bacterial strains ^{[17, 18]}. To acquire the most appropriate spectra for typing bacteria within their corresponding taxonomy, the wavenumber, ranging from 1500 cm^-1 to 900cm^-1, was carefully optimized and selected. We found that the region 1300-1000cm^-1 were of particular interest because their proximity to the best clustering results (Fig.4). This region mainly reflected the chemical information of P=O symmetric and asymmetric stretching vibrations of >PO₂^- in nucleic acids or phospholipids and C-O-C, C-O vibrations in glycogen and carbohydrates ^[19]. The results showed that these bacteria were typing into three larger clusters, including four. Listeria strains, three Vibrio strains and two Staphylococcus strains which were clearly classified. Although the Salmonella species, Yersinia enterocolitica and Shigella dysenteriae were grouped together, the fact that all three of them belong to members of Enterobacteriales could explain the difficulty of discrimination, which was consistent with previous reports ^{[12, 20]}.

Fig.4Full-screen View

PCA (left) and HCA (right) of 16 bacterial strains based on the infrared spectral region 1300-1000cm^-1.

4 Conclusion

We reported on the use of SR-FTIR spectroscopy to type 16 foodborne bacteria from seven genera. A whole spectral region and four subdivided spectral regions were analyzed respectively. We found that the wavenumber range of 1300-1000 cm^-1, which mainly reflects the information of nucleic acids and polysaccharides, contained more specific information and could be referred to as the signature region. The spectral signature proceeded with PCA and HCA obtained a delineation among the bacterial classes and cluster results matched with their phylogenetic relationship. These results proved that SR-FTIR spectroscopy with high spectral resolution has the greatest potential in fast identifying and typing microorganisms based on the specific infrared signature.