Mutability of demographic noise in microbial range expansions

Strains and growth conditions

Single gene deletion strains were taken from the Keio collection [34] (Supplementary Table 1), which consists of all non-essential single gene deletions in E. coli K-12 strain BW25113. MreB and mrdA point mutant strains were from Ref. [35] (Supplementary Table 2). Plasmids pQY10 and pQY11 were created by Gibson assembly of Venus YFP A206K (for pQY10) or Venus CFP A206K (for pQY11) [31], and Spec^R from pKDsgRNA-ack (gift from Kristala Prather, Addgene plasmid # 62654, http://n2t.net/addgene:62654; RRID:Addgene_62654) [36]. Plasmids pQY12 and pQY13 were created similarly but additionally with Cm^R from pACYC184.

All E. coli experiments were performed in LB (Merck 110285, Kenilworth, New Jersey) with the appropriate antibiotics and experiments with S. cerevisae were performed in YPD [37]. All agar plates were prepared in OmniTrays (Nunc 242811, Roskilde, Denmark, 12.8 cm × 8.6 cm) or 12 cm × 12 cm square petri dishes (Greiner 688102, Kremsmuenster, Austria) filled with 70 mL media solidified with 2% Bacto Agar (BD 214010, Franklin Lakes, New Jersey). After solidifying, the plates were dried upside-down in the dark for 2 days and stored wrapped at 4 °C in the dark for 7–20 days before using.

Tracking lineages with fluorescent tracer beads

In order to track lineages, we spread fluorescent tracer beads with a similar size to the cells on the surface of an agar plate, allowed them to dry, then inoculated and grew a colony on top of the agar plate and imaged the tracer beads to track lineages. In this way, we are able to track lineages without genetic labels at low density (i.e. sparsely) in the colony so that we can distinguish individual lineages without needing high-resolution microscopy. We find that the bead trajectories track cell lineages over the course of one hour both at the colony front and behind the front (Figs. 1c, S1c, d, and S2). We chose to spread fluorescent tracer beads on the surface of the agar so that they could continue to be incorporated into the colony as it grew, which would allow us to track lineages even as existing beads and lineages get lost from the front. Even though behind the front many cells will be piled up on top of other cells rather than in contact with the agar, we don’t expect this to affect the ability of the beads to measure demographic noise, since lineages at the front (where cells are in a monolayer) are the most likely to contribute offspring to future generations [26].

Fluorescent tracer beads

For experiments with E. coli, 1 µm red fluorescent polystyrene beads from Magsphere (PSF-001UM, Pasadena, CA, USA) were diluted to 3 µg/mL in molecular grade water and 920 µL was spread on the surface of the prepared OmniTray agar plates with sterile glass beads. Excess bead solution was poured out, and the plates were dried under the flow of a class II biosafety cabinet (Nuaire, NU-425-300ES, Plymouth, MN, USA) for 45 min. The bead density was chosen to achieve ~250 beads in a 56x field of view. For experiments with S. cerervisiae, 2 µm dragon green fluorescent polystyrene beads from Bang’s labs (FSDG005, Fishers, IN, USA) were used at a similar surface density.

Measurement of the distribution of demographic noise

We randomly selected 352 single gene deletion strains from the Keio collection. For each experiment, cells were thawed from glycerol stock (see Supplementary Methods), mixed, and 5 µL was transferred into a 96-well flat bottom plate with 100 µL LB and the appropriate antibiotics. Plates were covered with Breathe-Easy sealing membrane (Diversified Biotech BEM-1, Doylestown, PA, USA) and grown for 12 h at 37 °C without shaking. A floating pin replicator (V&P Scientific, FP12, 2.36 mm pin diameter, San Diego, CA, USA) was used to inoculate a 2–3 mm droplet from each well of the liquid culture onto a prepared OmniTray covered with fluorescent tracer beads. Droplets were dried and the plates were incubated upside down at 37 °C for 12 h before timelapse imaging.

To account for systematic differences between plates, we also put 8 wild type BW25113 wells in each 96-well plate in different positions on each plate. The mean squared displacement (MSD, see below) of each gene deletion colony was normalized to the weighted average MSD of the wild type BW25113 colonies on that plate, 〈MSD〉_WT, and this “relative MSD” is reported. We performed three biological replicates for each strain (grown from the same glycerol stock, Fig. S3), and their measurements were averaged together weighted by the inverse of the square of their individual error in relative MSD. The reported error for the strain is the standard error of the mean. During the experiment, several experimental challenges impede our ability to measure demographic noise, including the appearance of beneficial sectors (identified as diverging bead trajectories that correspond to bulges at the colony front) either due to de novo beneficial mutations or standing variation from glycerol stock (see Supplementary Section 2.4, Figs. S4 and S5), slow growth rate leading to bead tracks that were too short for analysis, no cells transferred during inoculation with our pinning tool, inaccurate particle tracking due to beads being too close together, or out of focus images. In order to keep only the highest quality data points, we focused on the 191 strains that had at least 2 replicates free of such issues.

Timelapse imaging of fluorescent beads

Plates were transferred to an ibidi stagetop incubator (Catalog number 10918, Gräfelfing, Germany) set to 37 °C for imaging. Evaporation was minimized by putting wet Kim wipes in the chamber and sealing the chamber with tape. The fluorescent tracer beads at the front of the colony were imaged with a Zeiss Axio Zoom.V16 (Oberkochen, Germany) at 56x magnification. A custom macro program written using the Open Application Development for Zen software was used to find the initial focal position for each colony and adjust for deterministic focus drift over time due to slight evaporation. Timelapse imaging was performed at an interval of 10 min for 12 h, during which time the colony grew about halfway across the field of view. Two z slices were taken for each colony and postprocessed to find the most in-focus image to adjust for additional focus drift. Subpixel-resolution particle tracking of the bead trajectory was achieved using a combination of particle image velocimetry and single particle tracking [38] and is described in detail in the Supplementary Methods.

Measurement of bead trajectory mean squared displacement

The measurement of mean squared displacement (MSD) is adapted from [31] and is illustrated in Figs. 1a and S1a. Points in a trajectory that fall within a window of length L are fit to a line of best fit. The MSD is given by

where Δw(L’) is the displacement of the bead trajectory from the line of best fit at each point, 〈〉_windows is an average over all possible definitions of a window with length L along the trajectory (window definitions are overlapping), and 〈〉_trajectories is a weighted average over all trajectories in a field of view, where the weight is the inverse squared standard error of the mean for each trajectory’s MSD(L) (Fig. S1a). We use 200 linearly spaced window sizes from L = 6 to 1152 μm. Window sizes that fit in fewer than 5 trajectories are dropped due to the noisiness in calculating the averaged MSD(L). The combined MSD(L) for all trajectories reflects that of bead trajectories at the colony front, which will have the largest contribution to the strength of demographic noise [26] (Fig. S6). Because we expect the trajectories to follow an anomalous random walk [31], the combined MSD(L) for all trajectories across the field of view is fit using weighted least squares to a power law, where the weight is the inverse square of the propagated standard error of the mean. Colonies with data in fewer than 5 window sizes are dropped due to the noisiness in fitting to a power law. The fit is extrapolated or interpolated to L = 50 µm to give a single summary statistic for each colony, and this quantity is reported as MSD(L = 50 µm) (see Supplementary Section 2.2, Figs. S7 and S8), and the error is calculated as half the difference in MSD (L = 50 µm) from using the upper and lower bounded coefficients to the fit. For calculating the distribution of demographic noise effects, only MSD values where the error is less than half of the value are kept.

Measurement of phenotypic traits

For the phenotypic trait measurements, in addition to the 191 single gene deletions, we also measured 41 additional strains of E. coli which included 4 strain backgrounds, 1 mreB knockout in the MC1000 background, 2 adhesin mutants, and 34 single gene knockouts from the Keio collection that we predicted may have large changes to demographic noise because of an altered biofilm forming ability in liquid culture [39] or altered cell shape from the wild type (using the classification on the Keio website, https://shigen.nig.ac.jp/ecoli/strain/resource/keioCollection/list). We normalized all phenotypic trait values to the average value measured from the wild type colonies on the same plate. The reported values for each strain are averages across 2–3 replicate colonies on different plates and the errors are the standard error of the mean. See the Supplementary Methods for more details of the specific phenotypic trait measurements.

Measurement of neutral fraction of diversity preserved

Neutral fluorescent pairs were created by transforming background strains with plasmids pQY10 (YFP, Spec^R) or pQY11 (CFP, Spec^R). Cells were streaked from glycerol stock and a single colony of each strain was inoculated into a 96 well plate with 600 µL LB and 120 µg/mL spectinomycin for plasmid retention. Plates were covered with Breathe-Easy sealing membrane and grown for 12 h at 37 °C without shaking. 50 µL of culture from each strain in a neutral pair were mixed and a floating pin replicator was used to inoculate a 2–3 mm droplet from the liquid culture onto a prepared OmniTray covered with fluorescent tracer beads. Droplets were dried and the plates were incubated at 37 °C.

Colonies were imaged after 24 h with fluorescence microscopy using a Zeiss Axio Zoom.V16 and the number of sectors of each color was manually counted. The fraction of diversity preserved was calculated as in Ref. [26] by dividing the number of neutral sectors by one-half times the estimated initial number of cells at the inoculum front (see Fig. 1b). The factor of one-half accounts for the probability that two neighboring cells at the inoculum front share the same color label. The initial number of cells is estimated by measuring the inoculum size of each colony (manually measured by fitting a circle to a brightfield backlight image at the time of inoculation) divided by the effective cell size for E. coli (sqrt(length*width) taken to be 1.7 µm, Ref. [26]).

Colony fitness

The colony fitness coefficient between two strains was measured using a colony collision assay as described in Refs. [26, 40] by growing colonies next to one another and measuring the curvature of the intersecting arc upon collision. Cells were streaked from glycerol stock and a single colony for each strain was inoculated into LB with 120 µg/mL spectinomycin for plasmid retention and incubated at 37 °C for 15 h. The culture was back diluted 1:500 in 1 mL fresh LB with 120 µg/mL spectinomycin and grown at 37 °C for 4 h. 1 µL of the culture was then inoculated onto the prepared 12cmx12cm square petri dishes containing LB with different concentrations of chloramphenicol (0 µg/mL, 1 µg/mL, 2 µg/mL, 3 µg/mL) in pairs that were 5 mm apart, with 32 pairs per plate, then the colonies were incubated at 37 °C. After half of a day, bright field backlight images are taken and were used to fit circles to each colony to determine the distance between the two colonies. After 6 days, the colonies were imaged with fluorescence microscopy using a Zeiss Axio Zoom.V16. The radius of curvature of the intersecting arc between the two colonies was determined with image segmentation and was used to calculate the fitness coefficient between the two strains (Fig. S9a).

Measurement of non-neutral establishment probability

We transformed 9 gene deletion strains from the Keio collection (gpmI, recB, pgm, tolQ, ychJ, lpcA, dsbA, rfaF, tatB) and 3 strain backgrounds (BW25113, MG1655, DH5α) with pQY11 (CFP, Spec^R) or pQY12 (YFP, Spec^R, Cm^R). Cells were streaked from glycerol stock and a single colony of each strain was inoculated into media with 120 µg/mL spectinomycin for plasmid retention, then incubated at 37 °C for 16 h. The culture was back-diluted 1:1000 in 1 mL fresh media with 120 µg/mL spectinomycin and grown at 37 °C for 4 h. YFP chloramphenicol-resistant and CFP chloramphenicol-sensitive cells from the same strain background were mixed respectively at approximately 1:500, 1:200, and 1:50 and distributed in a 96-well plate. A floating pin replicator was used to inoculate a 2–3 mm droplet from the liquid culture onto prepared OmniTrays with varying concentrations of chloramphenicol (0 µg/mL, 1 µg/mL, 2 µg/mL, 3 µg/mL). Droplets were dried and the plates were incubated at 37 °C for 3 days, then imaged by fluorescence microscopy using a Zeiss Axio Zoom.V16.

The establishment probability of the resistant strain can be measured by counting the number of established resistant sectors normalized by the initial number of resistant cells at the inoculum front [26], which gives the probability that any given resistant cell in the inoculum escaped genetic drift and grew to a large enough size to create a sector. Briefly,

where N_sectors is the number of resistant sectors after 3 days (counted by eye) and N₀ is the estimated initial number of cells of the resistant type at the inoculum front. Because the establishment probability can only be accurately measured when the initial number of resistant cells is low enough that the resistant sectors do not interact with one another, we only keep colonies where neighboring resistant sectors are distinguishable at the colony front. In cases where we could see that a sector had coalesced from multiple sectors, we counted the number of sectors pre-coalescence. We also did not find a clear downward bias in the establishment probability as a function of initial mutant fraction (Fig. S10), suggesting that the probability of sector coalescence is low in the regime of these experimental parameters. The initial number N₀ of cells of the resistant type is estimated by multiplying the initial number of cells at the inoculum front (see measurement of neutral fraction of diversity preserved) by the fraction of resistant cells in the inoculum (measured by plating and counting CFUs).

Source: Ecology - nature.com