Quantification of bacteria in water using PLS analysis of emission spectra of fluorescence and excitation-emission matrices

Abstract Bacterial contamination of drinking water is a considerable concern for public health. Tryptophan-like fluorescence (TLF) has been widely suggested to enable fast and inexpensive monitoring and quantification of bacterial contamination of water. Typically, TLF is determined at a certain excitation (ex)/emission (em) wavelengths pair. The aim of this study was to assess fluorescence spectroscopy supported with partial least squares (PLS) algorithms as a tool for a rapid evaluation of the microbial quality of water, by comparing the use of a single ex/em wavelengths pair, of the spectrum of emission obtained at a single excitation wavelength to that of whole excitation-emission matrices (EEMs). For that, laboratory-grown Escherichia coli, Bacillus subtilis and Pseudomonas aeruginosa were studied as the model systems, as well as, 90 groundwater samples from 6 different wells in Israel. The groundwater samples were characterized for fluorescence emission, coliforms, fecal coliforms, fecal streptococci and heterotrophic plate counts (HPC). The PLS analysis of emission spectra and, especially, of EEMs was capable of meaningfully reducing the detection limit of microorganisms in model systems, as compared with the single ex/em wavelengths pair-based determination commonly used, reaching a detection threshold as low as 10 CFU/ml. Use of PLS-analyzed EEMs becomes beneficial also in terms of correlation and similarity between the actual and predicted bacterial concentrations. Similarly, improved detection of bacteria was also achieved in groundwater samples. Furthermore, at a level of >104 CFU/ml, use of EEMs coupled with PLS enabled discrimination between E. coli, B. subtilis and P. aeruginosa.