From bulk samples to single cells: measurement strategies in Raman-based antibiotic susceptibility testing.

Raman-spectroscopy based antibiotic susceptibility testing (AST) enables rapid phenotypic analysis of bacterial responses to antibiotics. However, many established approaches rely on bulk bacterial material, limiting their applicability to samples with low and variable cell numbers, such as urine, where washing or medium exchange is often impractical. We systematically compare three Raman-based measurement strategies differing in sample preparation and cellular presentation: (i) dried bulk measurements on CaF2 substrates after centrifugation, (ii) cell cloud measurements in liquid using a dielectrophoresis (DEP) chip, and (iii) single-cell measurements on metal-coated filters after filtration. Escherichia coli ATCC 25922 was exposed to amoxicillin and levofloxacin across defined concentration ranges. Raman spectra were analysed using a PCA-LDA to classify between bacterial growth and inhibition, with validation against broth microdilution (BMD). Dried bulk measurements showed the highest agreement with BMD and the most robust classification performance, but require centrifugation and washing, limiting their suitability for low bacterial cell numbers. DEP-chip measurements enabled analysis in a closed system with reduced handling while maintaining good predictive performance, although they depend on sufficient cell density and involve higher chip costs. Single-cell measurements provided the simplest workflow and allowed analysis at clinically relevant low bacteria concentrations but showed reduced classification robustness due to increased spectral heterogeneity and longer acquisition times. Overall, this study highlights methodological trade-offs between robustness, experimental complexity, and suitability for low-cell-number samples. These findings provide practical guidance for selecting appropriate Raman-based AST strategies depending on sample type and application.