Single-cell scattering and auto-fluorescence-based fast antibiotic susceptibility testing for gram-negative and gram-positive bacteria

Abstract

In this study, we assess the scattering of light and auto-fluorescence from single bacterial cells to address the challenge of fast (<2 h), label-free phenotypic antimicrobial susceptibility testing (AST). Label-free flow cytometry is used for monitoring both the respiration-related auto-fluorescence in two different fluorescence channels corresponding to FAD and NADH, and the morphological and structural information contained in the light scattered by individual bacteria during incubation with or without antibiotic. Large multi-parameter data are analyzed using dimensionality reduction methods, based either on a combination of 2D binning and Principal Component Analysis, or with a one-class Support Vector Machine approach, with the objective to predict the Susceptible or Resistant phenotype of the strain. For the first time, both Escherichia coli (Gram-negative) and Staphylococcus epidermidis (Gram-positive) isolates were tested with a label-free approach, and, in the presence of two groups of bactericidal antibiotic molecules, aminoglycosides and beta-lactams. Our results support the feasibility of label-free AST in less than 2 h and suggest that single cell auto-fluorescence adds value to the Susceptible/Resistant phenotyping over single-cell scattering alone, in particular for the mecA+ Staphylococcus (i.e., resistant) strains treated with oxacillin.

Keywords: Escherichia coli; Staphylococcus epidermidis; aminoglycoside; antibiotic susceptibility testing; beta-lactam; flow cytometry; growth-free; label-free.

Figure 1

Daily experimental protocol, following an overnight culture on COS agar plate. ① Pre-culture in MHB, ② adjustment to 0.1 McF with fresh MHB, ③ start of incubation in 27 Eppendorf tubes at 37°C: 9 tubes without antibiotics (0), 9 tubes with the antibiotic molecule at the low breakpoint concentration (c), 9 tubes with the antibiotic molecule at the high breakpoint concentration (C); every 15 min, ④ readjustment to 0.1 McF with fresh MHB of one tube at 0, one tube at c, and one tube at C prior to ⑤ measurement in the flow cytometer: recording of 50′000 events per tube, in the FSC, SSC, FAD, and NADH channels.

Figure 2

Illustration of the two data analysis pipelines, using a susceptible E. coli (EC1) incubated with amoxicillin. (A,B) 2D raw scatter plots. (A–E) Measurements are shown at 0 min (l-h-s), 60 min (center), and 120 min (r-h-s) of incubation, for 0 (first row) and the high breakpoint C = 32 μg/mL (second row) of amoxicillin: (A) raw SSC versus FSC scattering data, and (B) raw NADH versus FAD auto-fluorescence data. Each subplot shows 50′000 events, and the grid illustrates how 2D-distribution of events is eventually represented as a vector for the subsequent population-based PCA analysis (a 5 × 5 grid is shown here for clearer readability, but a 20 × 20 grid was used for the actual data analysis). (C–E) Population-based PCA analysis. Pc2 versus pc1 scores resulting from the PCA analysis of the 27 distribution vectors – nine time points (from red dots at 0 min to yellow dots at 120 min) and three amoxicillin concentrations, 0 μg/mL (black curve), c = 8 μg/mL (grey curve), C = 32 μg/mL (blue curve): (C) for the SSC versus FSC input scattering data, (D) for the NADH versus FAD auto-fluorescence input data, and (E) for the concatenated scattering and auto-fluorescence input data. (F–J): Single-cell OC-SVM analysis. (F–G) Example of reference support (blue zone) determined by SVM analysis and containing 50% of the events at t = 0 min: (F) for the 2D scattering data at 0 μg/mL, and (G) for the 2D auto-fluorescence data at 0 μg/mL; a reference support is calculated at t = 0 min for each of the three antibiotic concentrations. (H–J) Fraction of events remaining in the reference support over time: (H) for the 2D scattering input data, (I) for the 2D auto-fluorescence input data, and (J) for the 4D scattering-auto-fluorescence input data (corresponding 4D reference supports not shown); at each time point, the fraction of in-support events for 0 μg/mL has been systematically subtracted.

Figure 3

Illustration of the data set and analysis results for the resistant E. coli (EC2) incubated with amoxicillin. (A,B): 2D raw scatter plots. Measurements are shown at 0 min (l-h-s), 60 min (center), and 120 min (r-h-s) of incubation, for 0 (first row) and the high breakpoint C = 32 μg/mL (second row) of amoxicillin: (A) raw SSC versus FSC scattering data, and (B) raw NADH versus FAD auto-fluorescence data. (C–E) Population-based PCA analysis. Pc2 versus pc1 scores resulting from the PCA analysis of the 27 distribution vectors – nine time points (from red dots at 0 min to yellow dots at 120 min) and three amoxicillin concentrations, 0 μg/mL (black curve), c = 8 μg/mL (grey curve), C = 32 μg/mL (blue curve): (C) for the SSC versus FSC input data, (D) for the NADH versus FAD input data, and (E) for the concatenated scattering and auto-fluorescence input data. (F–H): Single-cell OC-SVM analysis. Fraction of events remaining in the reference support over time: (F) for the input 2D scattering data, (G) for the input 2D auto-fluorescence data, and (H) for the input 4D scattering-auto-fluorescence data; at each time point, the fraction of in-support events for 0 μg/mL has been systematically subtracted.

Figure 4

PCA and OC-SVM analysis of the cytometry data recorded for the E. coli + gentamicin model: the susceptible EC1 strain (A,B) versus the resistant EC2 strain (C,D), for gentamicin at 0 μg/mL (black curves), c = 4 μg/mL (grey curves) and C = 16 μg/mL (blue curves). (A,C) pc2 versus pc1 scores, for the 2D scattering data (l-h-s), the 2D auto-fluorescence data (center), and the concatenation of the 2D scattering and the 2D auto-fluorescence data (r-h-s). (B,D) Fraction of events remaining in the reference (t = 0 min) SVM support over time, for the 2D scattering data (l-h-s), the 2D auto-fluorescence data (center), and the 4D scattering-auto-fluorescence data (r-h-s).

Figure 5

PCA and OC-SVM analysis of the cytometry data recorded for the S. epidermidis + gentamicin model: the susceptible SE2 strain (A,B) versus the resistant SE3 strain (C,D), for gentamicin at 0 μg/mL (black curves), c = 4 μg/mL (grey curves) and C = 16 μg/mL (blue curves). (A,C) pc2 versus pc1 scores, for the 2D scattering data (l-h-s), the 2D auto-fluorescence data (center), and the concatenation of the 2D scattering and the 2D auto-fluorescence data (r-h-s). (B,D) Fraction of events remaining in the reference (t = 0 min) SVM support over time, for the 2D scattering data (l-h-s), the 2D auto-fluorescence data (center), and the 4D scattering-auto-fluorescence data (r-h-s).

Figure 6

PCA and OC-SVM analysis of the cytometry data recorded for the S. epidermidis + oxacillin model: the susceptible SE1 strain (A,B), the resistant SE3 strain (C,D), the susceptible SE2 strain (E,F), and the resistant SE4 strain (G,H), for gentamicin at 0 μg/mL (black curves), c = 0.25 μg/mL (grey curves), and C = 0.5 μg/mL (blue curves). (A,C,E,G) pc2 versus pc1 scores, for the 2D scattering data (l-h-s), the 2D auto-fluorescence data (center), and the concatenation of the 2D scattering and the 2D auto-fluorescence data (r-h-s). (B,D,F,H) Fraction of events remaining in the reference (t = 0 min) SVM support over time, for the 2D scattering data (l-h-s), the 2D auto-fluorescence data (center), and the 4D scattering-auto-fluorescence data (r-h-s).