IMAGE ANALYSIS OF COLONY SIZE FOR INVESTIGATING SUBLETHAL INJURY IN LISTERIA MONOCYTOGENES

Efficient microbial colony growth dynamics quantification with ColTapp, an automated image analysis application

Populations of genetically identical bacteria are phenotypically heterogeneous, giving rise to population functionalities that would not be possible in homogeneous populations. For instance, a proportion of non-dividing bacteria could persist through antibiotic challenges and secure population survival. This heterogeneity can be studied in complex environmental or clinical samples by spreading the bacteria on agar plates and monitoring time to growth resumption in order to infer their metabolic state distribution. We present ColTapp, the Colony Time-lapse application for bacterial colony growth quantification. Its intuitive graphical user interface allows users to analyze time-lapse images of agar plates to monitor size, color and morphology of colonies. Additionally, images at isolated timepoints can be used to estimate lag time. Using ColTapp, we analyze a dataset of Staphylococcus aureus time-lapse images including populations with heterogeneous lag time. Colonies on dense plates reach saturation early, leading to overestimation of lag time from isolated images. We show that this bias can be corrected by taking into account the area available to each colony on the plate. We envision that in clinical settings, improved analysis of colony growth dynamics may help treatment decisions oriented towards personalized antibiotic therapies.

Effect of environmental stresses on the mean and distribution of individual cell lag times of Escherichia coli O157:H7

The effect of starvation, heat or acid stress on duration of individual cell lag time (tau) and standard deviation (SD) of tau was investigated using Escherichia coli O157:H7. Cells were stressed by exposure to acid (pH 3.5), heat (50 degrees C), or starvation in either glucose-free mineral medium (MOPS), tryptic soy broth (TSB) or Luria broth (LB). Stressed cells were then diluted into wells of a Bioscreen plate to obtain single cells per well. Replicate time to detection (td) values were obtained using the Bioscreen and used to calculate the tau and SD. Significant (P< or =0.05) increases in tau over untreated controls were found for the following treatments: 14 days in acid; 2 h of heating; 3 days starvation in MOPS; and 2 days starvation in either TSB or LB. The largest increase in tau was >2-fold from 2.5 to 5.6 h observed with the heat treatment. MOPS starvation was more detrimental to the cells than was acid treatment over the same time period. A significant increase in SD was found with 21 days acid treatment, and 2 days starvation in either TSB or LB. No significant increase in SD was found for MOPS starvation or heat treatment. Lognormal, Gamma, ExtremeValue and Weibull distributions were fitted to the tau data using BestFit. The results suggest that the Lognormal distribution is suitable for fitting tau data from either stressed or unstressed cells.

Assessment of distributions for fitting lag times of individual cells in bacterial populations

To develop mathematical models describing lag times of individual bacterial cells (tau), experimental tau data were fitted to a variety of continuous distributions using BestFit. Six strains of Escherichia coli O157:H7 were used, and serial dilutions were made in Bioscreen multi-well plates to get single cells per well. Detection times (td) for individual wells were converted to tau using the maximum specific growth rate (mu) for each strain. All strains were subject to 25 trials, with up to 100 replicate wells per trial. Some strains had significantly longer t(d), and lower mu, but the tau values were not significantly different. Distributions were best fit in the order Pearson V>Pearson VI>Extreme Value>Lognormal>Lognormal2>Inverse Gaussian based on the Anderson-Darling statistic. The Lognormal distribution was selected because there was less variability in the parameter values, and parameters have specific biological meanings. Distributions could be fit to sample populations as low as six, with fittings and parameter values comparable to those obtained with larger samples (up to 89). Extreme Value, Pearson V, and Pearson VI distributions were more suitable for fitting tau values generated from a Lognormal distribution when the numbers of sample points were few, which suggested that there are similarities between the distributions. The results suggest that a Lognormal distribution can be used successfully to characterize tau.

Automated image analysis of bacterial colony growth as a tool to study individual lag time distributions of immobilized cells

A method to determine the individual lag time (lag) distributions of immobilized bacteria was presented. The method was based on the image analysis of the bacterial colony growth. The lag distributions were retrieved from the distributions of the detection times (Td) required to form macroscopically visible colonies. Using this method, the lag distributions on agar for Listeria monocytogenes cells previously subjected to two situations reproducing conditions encountered during the contamination of cheese, were determined. The results were presented and compared with lag distributions obtained with an established method based on the time to detection of turbidity in broth. An original method to retrieve lag in broth and agar without any knowledge of the growth rate was also proposed. In order not to bias the distributions of lag on agar the impact of spatial separation between colonies on colony growth rates was quantified. Means and standard deviations of lag distributions for the two different stresses were found to be similar in broth and on agar. Extreme Value type II distributions fitted the best the different datasets of lag distributions.