Single-Cell Identification and Characterization of Viable but Nonculturable Campylobacter jejuni Using Raman Optical Tweezers and Machine Learning

Campylobacter jejuni is a leading foodborne pathogen that may enter a viable but nonculturable (VBNC) state to survive under environmental stresses, posing a significant health concern. VBNC cells can evade conventional culture-based detection methods, while viability-based assays are usually hindered by low sensitivity, insufficient specificity, or technical challenges. There are limited studies analyzing VBNC cells at the single-cell level for accurate detection and an understanding of their unique behavior. Here, we present a culture-independent approach to identify and characterize VBNC C. jejuni using single-cell Raman spectra collected by optical tweezers and machine learning. C. jejuni strains were induced into the VBNC state under osmotic pressure (7% w/v NaCl solution) and aerobic stress (atmospheric condition). Using single-cell Raman spectra and a convolutional neural network (CNN), VBNC C. jejuni cells were distinguished from their culturable counterparts with an accuracy of similar to 92%. There were no significant spectral differences between the VBNC cells formed under different stressors or induction periods. Furthermore, we utilized gradient-weighted class activation mapping to highlight the spectral regions that contribute most to the CNN-based classification between culturable and VBNC cells. These regions align with previously identified changes in proteins, nucleic acids, lipids, and peptidoglycan in VBNC cells, providing insights into the molecular characterization of the VBNC state of C. jejuni.