Miriam Cordovana (Bremen / DE), Norman Mauder (Bremen / DE), Arthur B. Pranada (Dortmund / DE), Frederik Pankok (Göttingen / DE), Ulrike Loederstaedt (Göttingen / DE), Simone Scheithauer (Göttingen / DE), Denise Dekker (Hamburg / DE), Andreas Erich Zautner (Magdeburg / DE), Walter Geißdörfer (Erlangen / DE), Judith Overhoff (Hamburg / DE), Miriam Werner (Hamburg / DE), Andreas Wille (Hamburg / DE), Hagen Frickmann (Hamburg / DE; Rostock / DE)
Background. The differentiation of Escherichia coli (Ec) from Shigella spp., the identification of Shigella at species level, as well as the discrimination of Ec at serotype level to delineate pathogenic lineages, are challenging with common routine methods. They require serological/genomic typing approaches, which present some disadvantages in terms of costs, ease-of-use and applicability in routine settings. In this study, we evaluated the discriminative power of Fourier-Transform Infrared (FT-IR) spectroscopy to distinguish E. coli isolates at serotype level and to delineate E. coli from Shigella species.
Material/Methods. 225 genomically or serologically typed strains (n=132 Ec belonging to 71 serotypes, n=93 Shigella spp.) were investigated. FT-IR analysis was performed applying the IR Biotyper® system (IRBT - Bruker Daltonics, Bremen, Germany) following the manufacturer"s instructions. Exploratory data analysis was performed by principal component analysis (PCA) and linear discriminant analysis (LDA). Machine learning was applied to create an automated classifier for the delineation of E. coli from Shigella spp., using the algorithms included in the IRBT software. One hundred and twenty-three isolates were used to build the training set, representing all 75 groups (Ec serotypes + the 4 Shigella species). The remaining 102 isolates were used as a testing set.
Results. Exploratory data analysis showed that IRBT clustering is correlated with the E. coli O-H serotypes and the Shigella species. PCA/LDA showed that the E. coli serotypes and the 4 Shigella species are separable. The classifier showed an accuracy of 99% (101/102), with one O157:H7 isolate misclassified as S. sonnei.
Conclusion. IR Biotyping showed the potential of delineating E. coli at serotype level, and of discriminating E. coli from Shigella spp., demonstrating its potential suitability for infection control, public health and epidemiological assessments. Further investigation is underway to confirm and strengthen these promising preliminary results.