Droplet size and surface hydrophobicity enhance bacterial plasmid transfer rates in microscopic surface wetness

Conjugal plasmids constitute a major engine for horizontal gene transfer in bacteria, and are key drivers of the spread of antibiotic resistance, virulence, and metabolic functions. Bacteria in terrestrial habitats often inhabit surfaces that are not constantly water-saturated, where microscopic surface wetness (MSW), comprised of thin liquid films and microdroplets, permanently or intermittently occurs. How physical properties of microdroplets, and of the surfaces they reside on, affect plasmid transfer rates is not well understood. Here, building on microscopy-based microdroplet experiments, we examined the relation between droplet properties (size and spread) and plasmid transfer rates at single-cell and individual droplet resolution, using Pseudomonas putida as a model species. We show that transfer rates increase with droplet size, due to higher densities of cells on the surface in larger droplets, resulting from lower ratio between the area of the liquid-solid interface and droplet volumes. We further show that surface hydrophobicity promotes transfer rates via the same mechanism. Our results provide new insights into how physical properties of surfaces and MSW affect plasmid transfer rates, and more generally, microbial interactions mediated by cell-to-cell contact, with important implications for our understanding of the ecology and evolution of bacteria in unsaturated environments.


Bacterial strains and growth conditions
Pseudomonas putida KT2440 cells were used as donor and recipient pair. Donor cells were constructed by Sørensen & Smets as described elsewhere [1]. Briefly, donor cells are chromosomally tagged with constitutively expressed mCherry and lacIq production (and kanamycin resistance gene). Additionally, donor cells carry a broad host pKJK5 plasmid marked with gfpmut3b gene, which is expressed by a LacIq repressible promoter Plac (plasmid includes kanamycin and tetracycline resistance genes). P. putida KT2440 cells were routinely cultured in M9 medium (M9 Minimal Salts Base 5x, Formedium, UK) supplemented with 20 mM Glucose and 50 μg/mL Km + 10 μg/mL tetracycline (for the donor strain), under agitation set at 220 rpm, at 28°C.

Experiment 1: Plasmid transfer in droplets sprayed on untreated glass substrate
Donor and recipient strain were cultured (separately) in 50 mL Falcon tubes continuing 25 mL of medium and appropriate antibiotic. Overnight cultures were washed twice (centrifuge at 6,000 rcf for 5 min) in M9 medium, and the pellets were re-suspended in 2 mL medium in order to reach high-density cultures (OD600 ˃ 5). Next, the OD of the donor and recipient strains were adjusted to OD600 = 4, and the strains were mixed with a 1:3 (donor:recipient) ratio in a 5 mL tube that contained 1 µM of Alexa dye (Alexa Fluor 647, Invitrogen) that was used to fluorescently stain the sprayed droplets. The solution was loaded into 5 mL refillable spray bottles (purchased at a local cosmetics store), and a portion of the load was sprayed on a 12-well glass-bottom plate (P12-1.5H-N, Cellvis) in the following manner: a 12-well plate was placed (without the plastic lid cover) in a plastic bag, and then the solution was delivered by pressing the spray nozzle 4 times from a distance of about 15 cm above the plate. Tap water was added to the empty spaces between the wells of the plate, plates were covered with the plastic lid, and the plate's perimeter was sealed with a stretchable sealing tape to maintain a humid environment (>98% RH). The plate was incubated in the dark at 28 o C throughout the duration of the experiment (18 hrs).

Experiment 2: Plasmid transfer in droplets sprayed on hydrophobic and hydrophilic modified glass substrate
Experiment 2 was conducted similarly to Experiment I with the following modifications: (a) The donor and recipient mix solution was adjusted to OD600 = 3 at a 1:1 ratio (b) Spray was applied to each well separately through a cylinder made out of 50 mL Falcon tube from which the conical end was chopped 1.5 cm above the base of the tube (one squeeze of the spray nozzle per well) (c) Hydrophobic and hydrophilic modified glass well plates were used.

Glass modification
Procedure for hydrophobic and hydrophilic glass modification was adopted from [2] with a few adjustments: 12-well glass-bottom plates (P12-1.5H-N, Cellvis) were filled with 2% RBS35 solution (RBS™ 35 solution, Sigma) and sonicated for 5 min in an ultrasonic bath (model ACP-200H, MRC) followed by thorough rinsing with tap water, demineralized water, methanol, tap water, and finally demineralized water again to obtain a hydrophilic surface. The plates were dried in an oven for 2 hrs at 70 o C, and then either stored before use or further modified by applying a hydrophobic coating. To obtain a hydrophobic surface, 20 µL of 2% (v/v) dichlorodimethylsilane (CAS 75-78-5, Sigma) in trichloroethylene (CAS No 79-01-6, Sigma) were applied to the center of each well (avoiding contact between the siliconizing solution and the boundaries of the well, which would result in melting of the well plate coating) and left to dry for 1 hr in a chemical hood. The plates were then dried in an oven for 2 hrs at 70 o C, rinsed with tap water, demineralized water, and dried in the hood for 2 hrs before use.

Microscopy
On the indicated time points (see Main Text), 12-well plates were mounted on a stage top chamber (H301-K-FRAME, Okolab) set at 28°C. Microscopic inspection and image acquisition were performed using an Eclipse Ti-E inverted microscope (Nikon) equipped with Plan Apo 40x/095 N.A. air objective and the Perfect Focus System for maintenance of focus. An LED light source (SOLA SE II, Lumencor) was used for fluorescence excitation. GFP fluorescence was excited with a 470/40 filter, and emission was collected with a T495lpxr dichroic mirror and a 525/50 filter. mCherry fluorescence was excited with a 545/25 filter, and emission was collected with a T565lpxr dichroic mirror and a 605/70 filter. Alexa 647 fluorescence was excited with a 620/60 filter, and emission was collected with a T660lpxr dichroic mirror and a 700/75 filter. Filters and dichroic mirror were purchased from Chroma, USA. A motorized encoded scanning stage (Märzhäuser Wetzlar, DE) was used to collect multiple positions of the well bottom surface. In each well, two random positions were chosen and imaged by scanning 5 × 5 adjacent fields of view (with a 5% overlap, 1.6 × 1.6 mm per scan). Images were acquired with an SCMOS camera (ZYLA 19 4.2PLUS, Andor, Oxford Instruments, UK). NIS Elements 5.02 software was used for acquisition.

Image processing
Image processing and analyses were performed to quantify the area of the liquid-surface interface (droplet area), total number of cells per droplet, number of donor cells per droplet, and plasmid transfer events per droplet. NIS Elements 5.02 software was used for image processing. Droplet masks were generated by intensity threshold segmentation of the Alexa 647 channel. Binary masks were converted to region of interest (ROI) elements delineating droplets' boundaries. Bright field channel was used to identify the total cell (donor, recipient, and transconjugate) entities within single droplets (i.e., identified ROIs). Rolling ball background correction (0.49 µm) was applied on the entire image, and the 'spot detection tool' was applied to enumerate cell number. mCherry channel was used to identify the donor cell entities within single droplets. Rolling ball background correction (0.49 µm) was applied on the entire image, and 'spot detection tool' was applied to enumerate donor cell number. GFP channel was used to identify trans-conjugant cells within single droplets. Rolling ball background correction (0.49 µm) was applied on the entire image, and intensity threshold was applied to identify single GFP expressing cells. The 'dilate' tool was operated on the resulting binary mask in order to cluster adjacent GFP expressing cells into a single object counted as a single 'plasmid transfer event' (i.e., we assumed that the plasmid was acquired prior to cell division). See also Fig. S10.

Mechanistic model
Density-based mechanistic model for the number of transfers per droplet (Te): A naïve mechanistic model assumes that the number of plasmid transfer events is a multiplication of the densities of donor and acceptor cells in each droplet and the droplet area, and some factor k.

(1) Te= k (Dd · Dr · A)
Where: Dd is the donor cell density (units: 1/μm 2 ) Dr is the recipient cell density (units: 1/μm 2 ) A is droplet area (units: μm 2 ) k is a constant (units: 1/μm 2 ) To estimate k, we used a simple regression model Y ~ kX where Y is the number of transfer events per droplet (Te) and X equals Dd · Dr · A. Assuming that p is the fraction of donors of all cells, and the recipients' fraction is (1-p), and that D=N/A, we get: where N is the total number of cells (donor + recipient) in the droplet.
N is estimated as a power function of the form: (3) N= β1A α1 Overall Te model:

Experiment 1 (untreated glass)
We first estimated k (see (1) below and Fig. S5) based on our data, and then replaced N based on our empirical model of cell number as a function of droplet area (see (2) and Fig.  1C). For comparison, we also computed a best fit model (this model is based on fitting to the data, not a mechanistic model) of Te as a function of A (see (3)). Power function coefficients were estimated using 'fitnlm' function to a simple power function of the form: Y = βX α .
Overall Te model development:   Although we chose to model Te, we also present here a model for Tc. (1) Estimating k: Only droplets wherein the average bin value crossed the threshold to show a positive number of transfers where used to calculate the slope (see Fig. S5).  S1. Droplet evolution between t = 0 hr (spray deposition) and t = 6 hrs post deposition. Left column shows representative images captured immediately after spray application on untreated glass (Experiment 1), hydrophilic, and hydrophobic treated glass (Experiment 2). Right column shows the same sections of the surface 6hrs after spray delivery. Note that in Experiment 1 (left panels), droplets become larger due to condensation or spread, while in Experiment 2 (middle and right panels) droplet size remained more stable on the hydrophilic surface, and shrank a bit due to evaporation on the hydrophobic surface. Note that evaporation was somewhat higher in smaller droplets. All images show an 800 X 800µm section.  . The X-axis depicts the naïve 'population-based' mechanistic model that we evaluated in the main text. Drops without any cells were removed from the analysis. Power function coefficients were estimated using 'fitnlm' function to a simple power function of the form: Y = β'X α' . This yielded the following fit: Y = 349·X 0.93 with an R 2 of 0.97. This analysis shows that the density-based model provides a good approximation for the number of donor-recipient pairs in close proximity.