In situ identification of toxin-producing Clostridioides difficile in stool samples based on single-cell Raman spectroscopy

<i>Clostridioides difficile</i> (CD) has emerged as one of the most prevalent nosocomial infections in hospitals and is the primary causative agent of antibiotic-associated diarrhea and pseudomembranous colitis. In recent years, <i>C. difficile</i>-induced infections have resulted in significant morbidity and mortality worldwide, with a particularly rapid increase in incidence observed in China. <i>C. difficile</i> strains are categorized into toxin-producing and non-toxin-producing based on their ability to synthesize toxins, with the pathogenicity of <i>C. difficile</i> being strictly dependent on the protein toxins produced by the toxin-producing strains. Therefore, early and rapid identification of toxin-producing <i>C. difficile</i> is crucial for the diagnosis and prevention of <i>Clostridioides difficile</i> infection (CDI). Currently, the detection methods of <i>C. difficile</i> infection carried out by clinical laboratories in China mainly include <i>C. difficile</i> toxin-producing culture, cell culture toxin assay, toxin assay by immunological methods, glutamate dehydrogenase (GDH) assay and nucleic acid amplification assay.However, current detection methods for CDI in clinical laboratories in China exhibit significant limitations, such as being time-consuming, operationally complex, and lacking in specificity and sensitivity. Raman microspectroscopy has been shown to have the potential for rapid and reliable identification in microbial diagnostics, with the method reducing the time to results to less than 1 hour, including the processing of clinical samples, the measurement of single-cell Raman spectra, and the final diagnosis through the use of training models. In this study, we aimed to predict <i>in situ</i> strain identification and virulent strain identification of 24 raw clinical stool samples by constructing a reference single-cell Raman spectroscopy (SCRS) database of common intestinal flora and <i>C. difficile</i>, as well as a reference SCRS database of toxin-producing and non-toxin-producing <i>C. difficile</i> strains. The results showed that the accuracy of <i>C. difficile</i> strain identification in clinical stool samples was 83%, and the accuracy of virulent strain prediction was 80%. These findings suggest that Raman spectroscopy may be a viable method for the rapid <i>in situ</i> identification of virulent and non-virulent <i>C. difficile</i> strains and holds promise for clinical application in the rapid diagnosis of CDI.