Münchberg Ute, Rösch Petra, Bauer Michael, Popp Jürgen
Institute of Physical Chemistry, Helmholtzweg 4, Friedrich Schiller University Jena, 07743, Jena, Germany.
Anal Bioanal Chem. 2014 May;406(13):3041-50. doi: 10.1007/s00216-014-7747-2. Epub 2014 Mar 21.
The identification of pathogenic bacteria is a frequently required task. Current identification procedures are usually either time-consuming due to necessary cultivation steps or expensive and demanding in their application. Furthermore, previous treatment of a patient with antibiotics often renders routine analysis by culturing difficult. Since Raman microspectroscopy allows for the identification of single bacterial cells, it can be used to identify such difficult to culture bacteria. Yet until now, there have been no investigations whether antibiotic treatment of the bacteria influences the Raman spectroscopic identification. This study aims to rapidly identify bacteria that have been subjected to antibiotic treatment on single cell level with Raman microspectroscopy. Two strains of Escherichia coli and two species of Pseudomonas have been treated with four antibiotics, all targeting different sites of the bacteria. With Raman spectra from untreated bacteria, a linear discriminant analysis (LDA) model is built, which successfully identifies the species of independent untreated bacteria. Upon treatment of the bacteria with subinhibitory concentrations of ampicillin, ciprofloxacin, gentamicin, and sulfamethoxazole, the LDA model achieves species identification accuracies of 85.4, 95.3, 89.9, and 97.3 %, respectively. Increasing the antibiotic concentrations has no effect on the identification performance. An ampicillin-resistant strain of E. coli and a sample of P. aeruginosa are successfully identified as well. General representation of antibiotic stress in the training data improves species identification performance, while representation of a specific antibiotic improves strain distinction capability. In conclusion, the identification of antibiotically treated bacteria is possible with Raman microspectroscopy for diverse antibiotics on single cell level.
病原菌的鉴定是一项经常需要进行的任务。当前的鉴定程序通常要么由于必要的培养步骤而耗时，要么在应用中成本高昂且要求苛刻。此外，患者先前使用抗生素治疗往往会使通过培养进行的常规分析变得困难。由于拉曼显微光谱法能够鉴定单个细菌细胞，因此可用于鉴定此类难以培养的细菌。然而到目前为止，尚未有关于抗生素处理细菌是否会影响拉曼光谱鉴定的研究。本研究旨在利用拉曼显微光谱法在单细胞水平上快速鉴定经过抗生素治疗的细菌。两种大肠杆菌菌株和两种假单胞菌已用四种抗生素进行处理，所有这些抗生素均靶向细菌的不同部位。利用未处理细菌的拉曼光谱建立线性判别分析（LDA）模型，该模型成功鉴定了独立的未处理细菌的种类。在用亚抑制浓度的氨苄青霉素、环丙沙星、庆大霉素和磺胺甲恶唑处理细菌后，LDA模型的种类鉴定准确率分别达到85.4%、95.3%、89.9%和97.3%。增加抗生素浓度对鉴定性能没有影响。一株耐氨苄青霉素的大肠杆菌菌株和一份铜绿假单胞菌样本也被成功鉴定。训练数据中抗生素应激的总体呈现提高了种类鉴定性能，而特定抗生素的呈现提高了菌株区分能力。总之，利用拉曼显微光谱法在单细胞水平上对多种抗生素处理过的细菌进行鉴定是可行的。
