Prolonged bacterial lag time results in small colony variants that represent a sub-population of persisters

Persisters are a subpopulation of bacteria that are not killed by antibiotics even though they lack genetic resistance. Here we provide evidence that persisters can manifest as small colony variants (SCVs) in clinical infections. We analyze growth kinetics of Staphylococcus aureus sampled from in vivo conditions and in vitro stress conditions that mimic growth in host compartments. We report that SCVs arise as a result of a long lag time, and that this phenotype emerges de novo during the growth phase in various stress conditions including abscesses and acidic media. We further observe that long lag time correlates with antibiotic usage. These observations suggest that treatment strategies should be carefully tailored to address bacterial persisters in clinics.


Supplementary Figure 2: Analyzing growth rates of colonies with different lag times
At millimeter scale (top), appearance time of the colonies (the time when a colony reached a radius of 80 µm) was predictive of the colony radius at 24h (R 2 =0.83) (A). The initial growth rate of colonies (B) was slowly decreasing as a function of their appearance time as expected because of nutrients depletion on a plate. The rate at which a given colony grew also decreased over time (red line; green lines are 95% confidence intervals). These observations were supported by microscopic analysis (bottom): the colony radius initially grew exponentially (C). The initial exponential growth rate of colonies (D) did not correlate with the time of first division.

Supplementary Figure 3: Testing speed of reversion to wild type colony size
Small colonies, obtained from bacteria grown in DMEM pH 5.5 for 3 days, sampled 24 hours after plating from the tail of a wide distribution (A) will restore a wild type colony size distribution if plated immediately (B). The distributions that are obtained are peaked with no small colony variants, similarly to the distribution obtained by sampling bacteria from an exponentially growing culture. C, D, E are mean distributions from three replated SCVs; each represents one biological replicate. This shows that in addition to the restoration of the growth rate phenotype immediately after first divisions (supplementary material 2), the width of colony size distribution is also restored upon sub-cultivation.

Supplementary Figure 4: Colony growth model: estimation of lag time from 24h colony size
One can estimate initial lag time of the bacterium that initiated a colony by extrapolation based on the colony radius at 24 h. For this, we assumed a twostep growth dynamics. First, the colony grew exponentially at a rate of 0.41 h -1 (blue line, as observed under the microscope) until the radius reached Rlin=130 µm. Once the colony has reached this radius, cells in the center cannot access nutrients anymore, and only a fixed band of bacteria at the colony edge will be dividing, with the consequence that radial growth is linear (39) (orange line). We fitted this growth rate on time lapse movies from 3 different plates to be on average 55 µm/h. We use this model to estimate lag time from colony size data. Note that two problems emerge for the "size ratio" commonly used to define SCVs. First, this definition is based on using the most common colony size as the denominator, which will change with SCV proportion (see Fig. 1D for example). As a consequence, the SCV proportion can be underestimated. Second, since the growth of colonies is linear, it results that the ratio of sizes between "large" colonies and small colonies is time dependent: the ratio between the radii of two colonies will tend towards 1 for infinite time (in linear growth). Below a comparison of two colonies with different lag times:

R lin
Exponential growth Linear growth

Supplementary Figure 5: Survival of pre-exposed bacteria upon antibiotics exposure
Time kill curves after pre-exposure to acidic or neutral media. The curves depict the fraction of bacteria that were killed as a function of time upon exposure to the antibiotics flucloxacillin (A) and ciprofloxacin (B) at 40 fold the minimum inhibitory concentration. C and D depict the data for the antibiotic levofloxacin at 40 times and 10 times MIC, respectively. Bacteria were sampled from cultures grown in neutral (pH 7.4) or acidic (pH 5.5) medium for 3 days and transferred to medium at pH 7.4 containing antibiotics. Bacterial survival was analyzed by washing and plating at different time points. Each data point was calculated as the ratio between the number of colony forming units (CFU) at a specific time point after antibiotic addition and the number of CFU in the inoculum. The P value at the bottom right of each figure is the result of a onephase association fit to the time kill data points followed by an extra sum of squares F test to compare Y0 of the two fitted curves, their plateau (i.e., the percentage of killed bacteria as time approaches infinity) and K values (rate constant, expressed in reciprocal of the X axis time units). Estimates of the total number of bacterial CFU recovered from mouse samples (blue circles) and the number of SCVs recovered (green stars). The bacteria were sampled from both pus and tissue material surrounding the abscess, whose volume was estimated based on its weight. The horizontal blue line represents the initial inoculum and the green line represents the maximal initial number of inoculated SCVs, as per our detection limit. The samples from mouse 3 and 4 resulted in the shifted distributions (dashed lines) on the left panel.