Candida species account for a significant proportion of invasive fungal infections and candidaemia has been cited as the fourth most common cause of nosocomial bloodstream infections. Rapid diagnosis and drug susceptibility testing are required for better patient outcomes. Infrared spectroscopy has demonstrated promising capacity to distinguish between fungal species, strains and morphological structures. This study explored the capacity of Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy (ATR-FTIR) coupled with machine learning techniques to detect chemical signatures of Candida spp. that could be used for rapid antifungal susceptibility testing.
A rapid susceptibility assay was developed using susceptible and resistant C. albicans isolates. Isolates were cultivated at 30°C in liquid media in varying concentrations of fluconazole (control, 1 µg/ml, 2 µg/ml 4 µg/ml and 8 µg/ml). Spectra were acquired from cells harvested at timepoints (2, 4, 6 and 8 hours) using a Bruker Alpha FTIR instrument (Bruker Optics, USA) equipped with a diamond crystal Attenuated Total Reflection (ATR) element and were analysed with Principal Component Analysis (PCA). These data were used to inform an optimal timepoint and fluconazole concentration for further validation using a larger number of isolates.
PCA revealed a potential dose dependent relationship in spectral pattern changes within the susceptible isolates grown in varying concentrations of fluconazole. Spectral changes were observed as early as 4 hours post drug exposure in a fluconazole concentration as low as 4 µg/ml. PCA performed on the refined protocol with additional isolates demonstrated that susceptible and resistant isolates were spectroscopically distinct based on various chemical markers. Resistant isolates were correlated with various protein, lipid, polysaccharide and DNA bands whilst susceptible isolates were correlated with a distinct polysaccharide band.
Clinically, patient outcomes are heavily reliant on rapid drug susceptibility testing to reduce the time to effective antifungal therapy. Our preliminary data provides evidence that ATR-FTIR coupled with machine learning can identify differences in chemical signatures of susceptible and resistant Candida isolates grown under antifungal selection pressure within 4 hours of drug exposure. These data provide new insights into rapid antifungal susceptibility testing that could be further developed to optimise patient outcomes from a life-threatening infection.