Femtoliter Droplet Confinement of Streptococcus pneumoniae: Bacterial Genetic Transformation by Cell-cell Interaction in Droplets

Abstract

Streptococcus pneumoniae (pneumococcus), a deadly bacterial human pathogen, uses genetic transformation to gain antibiotic resistance. Genetic transformation begins when a pneumococcal strain in a transient specialized physiological state called competence, attacks and lyses another strain, releasing DNA, taking up fragments of the liberated DNA, and integrating divergent genes into its genome. While many steps of the process are known and generally understood, the precise mechanism of this natural genetic transformation is not fully understood and the current standard strategies to study it have limitations in specifically controlling and observing the process in detail. To overcome these limitations, we have developed a droplet microfluidic system for isolating individual episodes of bacterial transformation between two confined cells of pneumococcus. By encapsulating the cells in a 10 μm diameter aqueous droplet, we provide an improved experimental model of genetic transformation, as both participating cells can be identified, and the released DNA is spatially restricted near the attacking strain. Specifically, the bacterial cells, one rifampicin (R) resistant, the other novobiocin (N) and spectinomycin (S) resistant were encapsulated in droplets carried by the fluorinated oil FC-40 with 5% surfactant and allowed to carry out competence-specific attack and DNA uptake (and consequently gain antibiotic resistances) within the droplets. The droplets were then broken, and recombinants were recovered by selective plating with antibiotics. The new droplet system encapsulated 2 or more cells in a droplet with a probability up to 71%, supporting gene transfer rates comparable to standard mixtures of unconfined cells. Thus, confinement in droplets allows characterization of natural genetic transformation during a strictly defined interaction between two confined cells.

Introduction

Streptococcus pneumoniae (pneumococcus) is a Gram-positive bacterial human pathogen causing 1 million deaths of children, elderly and immunocompromised individuals due to pneumonia, septicemia, and meningitis annually.^1–4 Two clinical strategies used to combat such infections are antibiotic therapy to cure patients and vaccination to reduce the deadliest types of the bacteria. However, the effectiveness of both strategies has recently been challenged due to the spread of antibiotic resistant strains. The emergence of antibiotic resistance has become more serious due to the development of multi-drug resistant pneumococcal strains, which was first noted in South Africa and has since spread globally.^5–7 A fundamental phenomenon that contributes to the failure of both strategies, known as genetic transformation, is a complex mechanism that depends on a pathway for the uptake of exogenous DNA from the bacterial surroundings and the incorporation of the acquired DNA into the genome.^8–10 Consequently, a better understanding of this phenomenon could help development of new strategies to mitigate antibiotic resistance and vaccine escape.

Since its discovery in 1944, gene transfer in S. pneumoniae via the mechanism of genetic transformation has been studied primarily by using a simple experimental paradigm.^11–13 In this paradigm, a donor DNA extracted and purified from one strain carrying a gene marker of interest is introduced into a rapidly growing culture of recipient cells of a strain that lacks that marker, but that can enter into a transient specialized physiological state (K state, X state, competence) giving it the unusual ability to import DNA and catalyze its recombination into the recipient’s chromosome by homologous recombination.¹⁴ Using this paradigm, many remarkable aspects of the cell-DNA interaction have been discovered. These include the existence of 30–100 competence-specific genes, that are expressed only during a 20 min window of competence, a dedicated DNA transport pathway that cuts into double-strand DNA and imports a single strand starting from that initial break, and a specific recombination pathway that efficiently replaces recipient genes with related genes from the donor DNA.^{12, 15, 16}

More recently, the pneumococcal competence-specific genes were also found to encode proteins of a pathway that bypasses the DNA purification step of the classic paradigm. In this alternative pathway, dependent on seven proteins, CbpD, CibA, CibB, CibC, ComM, LytA and LytC, a competent cell kills non-competent pneumococci on contact, releasing DNA in a form suitable for immediate uptake by the competent cells.^17–21 Most recently, gene transfer by cell-cell contact was found to resemble more closely to DNA transfer in vivo than does the classic paradigm, especially in that the tracts of DNA that are transferred are much larger than had been seen within the classic paradigm.²² Thus, we infer that the elementary interaction of genetic transformation is only partly modeled in the classic paradigm, and in nature it usually occurs as an attack of one competent cell on one victim target cell. However, a complete description of the natural outcome of this mechanism of gene transfer is not yet available, especially as both donor and recipient carry two or more copies of the pneumococcal chromosome and the interacting cells are poorly identified in any presently used paradigm: i.e., either mixed liquid cultures, mixed biofilms, or so-called filter matings. The existing experimental paradigms suffer from two main limitations: their inability to identify the donor cell(s) within a bulk culture or a dense assemblage such as a biofilm, or even to know whether there was one donor or more than one donor cell, and an inability to capture all descendants from a productive interaction. At most, they capture one descendant of an unspecified interaction, based on transfer of a selective marker.

To characterize the outcome of the fundamental cell-cell attack event of the pneumococcal natural competence mechanism fully, we aim to isolate directly an attack on a single victim by a single competent cell, recover all descendants of the interacting competent cell (the survivor), and map all resulting recombination events (gene replacements) by whole-genome sequencing. Because both target and attacker each have at least two copies of the chromosome, a full description of the outcome of the exchange will require mapping of all events in descendants of the competent cell recovered after complete segregation of the four or more strands present at the time of attack. To accomplish this level of detailed mapping, we turn to droplet microfluidics. Droplet microfluidic technology has been previously proven to be a powerful platform for biological studies that different types of bacteria have been successfully encapsulated in droplets for numerous biological applications.^23–28 For example, Thibault et al recently developed transposon insertion sequencing (Tn-Seq) in droplets for bacteria including Streptococcus pneumoniae.²⁹ However, to date, none have demonstrated or characterized bacterial genetic transformation within micro-droplets.

In this report, we describe the design and evaluation of a prototype device for encapsulation of pneumococcal cell pairs and incubation for cell-cell attack under direct observation, showing that development of competence within droplets proceeds with kinetics matching unconfined cell mixtures, demonstrating cell-cell attack within confining droplets, and showing that recombinants can be recovered from the droplets afterwards. In future device designs we expect to introduce real-time imaging of cell attack/lysis, and image-directed sorting of droplets to allow capture of recombinants from imaged attack episodes, so as to enable start-to-finish analysis of this mechanism of horizontal gene transfer.

Experimental Section

Cell stocks and culture media

Two strains of pneumococcus, CP2204 (Rif^R) and CP2215 (Nov^RSpc^R), described previously²², were complementary in antibiotic resistance, to aid in recovering progeny from transformation events. For cell stocks, each strain was grown to an optical density of 0.2 at 550 nm (OD₅₅₀=0.2) in 12 mL of filtered chemically defined medium (CDM)^{30, 31} with 1% (v/v) of casein tryptone medium (CAT)³², mixed with 2 mL of 100% glycerol, and stored at −80°C. In bacterial culture, OD₅₅₀ is a convenient measure of population density. The strain CP2204, dependent for competence development on exogenous CSP pheromone, is designated as DNA recipient or attacker. In contrast, strain CP2215, lacking the CSP receptor, can only act as DNA donor. Phenotypes of strains used in this study are shown in Figure 1.

Figure 1. Summary of phenotypes of pneumococcal strains used for transformation in droplets and schematic of cell encapsulation for confined gene transfer between the strains.

CP2204, inducible with competence-stimulating peptide (CSP), is designated as attacker or DNA recipient; CP2215, genetically incompetent, is the natural donor of DNA. In an environment with CSP, CP2204 can attack and lyse CP2215 and take up liberated fragments of DNA, gaining resistance to novobiocin (N) or spectinomycin (S).

Before each experiment, CP2204 and CP2215 cell stocks were diluted separately in 12 mL of CDM with 1% (v/v) of CAT and maintained at 37°C until reaching an OD₅₅₀ of 0.25. Different inocula (50 μL of CP2215 vs. 350 μL of CP2204) usually allowed both cultures to reach the desired OD of 0.25 within 4–5 hours. Both cultures were then chilled on ice and centrifuged at 7000 g for 7 minutes in a 4°C centrifuge (Eppendorf model 5804R). The supernatant was poured off and the pellets were re-suspended with 750 μL of fresh cold CDM containing 1% (v/v) CAT. A 1:1 mixture of the two suspensions was then stored on ice to keep the cells dormant until introduction into the microfluidic device in the cold room.

Inducer cocktail.

Development of competence by CP2204 was induced by mixing with an inducer cocktail including CSP, bovine serum albumin (BSA), CaCl₂, and CDM. A 4.2 mg amount of fresh CSP-1³³ (NeoBiolab) was dissolved in 42 mL distilled water and filtered by 0.2 μm syringe filter (Sartorius) before 1-mL aliquots were stored at −20°C. BSA (Gold Biotechnology) was dissolved at 40 mg/mL in distilled water, sterile-filtered, and stored at 4°C. For short term uses, the BSA stock solution was diluted to make 1 mL of 0.4% BSA for inducer cocktail preparation. CaCl₂ solid was dissolved in distilled water at 0.1 mol/L, autoclaved, and stored at room temperature. The 0.1M CaCl₂ stock solution was diluted further to make 1 mL of 0.01M stock and stored at 4°C. An inducer cocktail, prepared directly before each encapsulation experiment by combining 28 μL of 100 μg/mL CSP, 14 μL of 0.4% BSA, 70 μL of 0.01M CaCl₂ and 588 μL of CDM, was stored on ice until use.

Antibiotic stocks.

Rifampicin powder (Biosciences) was dissolved in 99.8+% methanol (Alfa Aesar) to a concentration of 20 mg/mL. Novobiocin (Sigma Aldrich) and spectinomycin (Sigma Aldrich) were dissolved in distilled water at 10 mg/mL and 100 mg/mL, respectively. All antibiotic stocks were then sterilized with 0.2 μm syringe filters and stored at 4°C.

Selective Plating.

Todd Hewitt Broth with 2% yeast extract medium (THY) was used with 1.5% agar for plating. The agar plating method³⁴ used 60 × 15 mm petri dishes (Falcon). A 3-mL layer of THY agar was first filled as a base in all petri dishes and allowed to be harden. Then, for each dilution, 1.5 mL of cell sample was mixed into a culture tube containing 1.5 mL of THY agar at 55°C, mixed, and immediately poured into a petri dishes as a second layer. A third layer of 3 mL THY agar was added after the second layer completely solidified. The last layer, for selection, contained the selective antibiotics rifampicin (40 μg/ml), novobiocin (10 μg/ml) or spectinomycin (160 μg/ml)^35–38. All plates were incubated at 37°C for 40 hours before counting colonies.

Fabrication of a droplet generating microfluidic device

Droplet generating microfluidic devices were fabricated by standard soft-lithography.^39–41 SU-8 2015 negative photoresist (MicroChem Corp) was poured on a cleaned 100-mm silicon wafer and spun at 4000 RPM in a spin coater to generate a thickness of 10 μm. The SU-8 coated silicon wafer was soft-baked at 95°C for 5 minutes and then aligned in a Heidelberg μPG 501 Maskless Aligner for exposure to 390 nm radiation with the specific setup of 18 mW, 100% intensity, and 4× exposure time. After exposure, the wafer was baked at 95°C for 10 minutes and then 120°C for 30 minutes. Uncured SU-8 was washed away by SU-8 developer (MicroChem Corp) on a shaker for 30 minutes and the wafer was rinsed with acetone followed by isopropyl alcohol (IPA) and drying with nitrogen gas.

For PDMS replica molding, 4 grams of elastomer curing agent (Sylgard 184, Dow Corning) was added to 40 grams of PDMS base (Sylgard 184, Dow Corning) to make a final ratio of 1:10 and then mixed with a Thinky mixer (AR-100, Thinky) for 2 minutes. The mixture was poured on the patterned silicon wafer followed by degasification in vacuo (30 min) and cross-linking at 65°C (12 h). Then, the PDMS was carefully peeled off, cut to shape, and cleaned with Scotch tape on both sides before the inlet and outlet ports were punched with a 1.5-mm biopsy punch. The PDMS mold was placed in contact with a glass slide after 30 seconds of oxygen plasma exposure (Plasma Etch, Inc) and gently pressed for permanent bonding. The complete chip was baked overnight at 135°C to strengthen the bonding. The chip was cooled to room temperature and the channels were made hydrophobic with Novec 1720 (3M) by injecting the Novec through the device from the oil inlet; a 15-minute incubation was terminated by flushing with air. The device was again heated overnight at 135°C in ambient air for curing followed by storage at room temperature until being moved to a 4°C cold room for 2 hours before use in droplet generating experiments.

Experiments using parallel droplet-confined cells and unconfined controls

Droplet formation and pneumococcal encapsulation.

All experiments were conducted in a 4°C cold room to hold the bacterial strains dormant during droplet generation. After centrifugation and re-suspension, the mixed culture was maintained at 0°C by placing the mixed culture tube in ice-water. The droplet system was controlled by syringe pumps (Chemyx Inc). An oil phase syringe containing Pico-Surf (Dolomite), which is 5% surfactant in FC-40 fluorinated oil, operated at ~250 μL/hr. Both aqueous phase syringes (culture mixture and inducer cocktail) operated at a flow rate of ~70 μL/hr. Once droplets were stably forming, a 20-cm long PTFE tubing (~1-mm ID) connected the outlet of the device to a droplet-collecting vial containing 300 μL of CAT and maintained in ice. In characterization experiment, a ~5-cm-long tube connected the droplet generator outlet to the inlet of the imaging chamber, and then the tube was removed to keep the droplets static inside of the chamber. Cell count experiment was performed via live imaging of the droplets in the imaging chamber.

Control of competence by manipulation of culture temperature and post-droplet dilution.

For this experiment, the droplets were accumulated for 1 hour and two equivalent droplet samples were collected on ice. Each sample vial contained 300 μL of CAT to prevent droplet evaporation. After droplet formation and collection, one droplet sample vial was incubated for 25 minutes at 37°C while the other remained at 4°C. After 25 minutes, both droplet vials were centrifuged gently (100 g, 30 seconds) and the bottom Pico-Surf layer was pipetted out. The remaining emulsion of each vial was mixed by inversion with 200 μL of Pico-Break (Dolomite) and then centrifuged at 1000 g for 1 minute. The volume of upper phase containing CAT plus the contents released from broken droplets from each vial was measured and pipetted out to new test tubes. The upper-phase samples were diluted in CAT by dilution factors of 10-, 100-, and 1000-fold, and incubated at 37°C for one hour to allow gene integration and expression. The resulting cultures were then serially diluted further in CAT for plating and antibiotic selection. Since CP2204 is itself rifampicin-resistant, to test for its genetic transformation, the RN combination of antibiotics was used. In addition, single antibiotic plating of R and N was used for estimating viable cell recovery. The final dilution factors for antibiotics selection for all experiments were 10⁻³ and 10⁻⁴ for plates with the double selective combination (RN). For viable cell recovery, dilutions were made of 10⁻⁵ and 10⁻⁶ for each single selection (R/N).

Comparison of kinetics of competence development within droplets vs. planktonic cultures.

For this experiment, droplets were collected sequentially for 10 minutes and stored separately in 7 droplet-collecting vials which all contained 200 μL of CAT and were maintained in ice. These 7 samples were then incubated at 37°C for 0, 10, 15, 20, 25, 30, or 35 minutes, respectively, to activate cell-cell attack and competence. After incubation, each sample vial was stored on ice until all samples finished the incubation. Emulsion breaking steps described above were performed for all droplet vials. Finally, the upper phase of the Pico-Break centrifugation containing CAT and aqueous droplets contents was diluted with additional CAT medium for a total dilution of 1000-fold. After incubation for one hour at 37°C to allow gene integration and expression, all samples were diluted further in CAT for plating in selective THY agar with RN double selection combination as described above.

Comparison of bacterial transformation in droplets, in a planktonic positive control, and in a Pico-exposed control.

For this experiment, droplets were collected in the cold room for 30 minutes. Simultaneously, two unconfined culture control experiments which were then used as positive and Pico-exposed control, proceeded equivalently by mixing equal volumes of cell mixture and inducer cocktail to match the predicted final volume of aqueous phases in droplets collected after 30 min and maintained on ice with the droplet-collecting vial until the droplet collection process was finished. Two controls have similar reaction contents at this step and would be experimentally differentiated in the emulsion breaking steps. After 30 min, the droplets sample and both of the control samples were incubated at 37°C for 25 minutes for cell-cell attack and transformation process. Thereafter, the positive control rested in room temperature. The Pico-exposed control was mixed with 125 μL of Pico-Surf followed by 300 μL of CAT. Then, both the Pico-exposed control and the experimental droplets experienced the emulsion breaking steps and the remaining suspensions were measured and pipetted out. All samples were then added to additional CAT for a total 1000-fold dilution, and all samples were then incubated at 37°C for one hour and diluted for plating in selection THY agar. In this experiment, two different double selection combinations, RN and RS, were used.

Results

Droplet generation and encapsulation of pneumococcal cells in droplets

To enable gene transfer between pairs of cells isolated by containment within microdroplets, a droplet microfluidic device was developed as shown in Figure 2. The droplets were generated by flow-focusing^42–48; separate aqueous phase streams of inducer and cell mixture were forced to flow through a 10 μm width narrowed neck, then an outer Pico-Surf oil stream exerted pressure and viscous stresses to break the mixed aqueous streams into uniform droplets via the narrowed orifice.⁴² The use of flow-focusing in droplet formation produces more stable monodisperse droplets at a low oil-water flow rates ratio compared to other geometries.^49–51 In our droplet device, the dimensions of the droplet generating region were designed to produce uniform 10 μm diameter droplets including the 10 μm height of the channels and the width of the narrowed neck, so the droplet generation process was only dependent on controlling flow rates of the inlet streams. In addition, passive microfilters with 10-μm and 6-μm constrictions were placed at inlets of both aqueous streams and upstream of the narrowed neck, respectively, to prevent debris from clogging the 10 μm droplet generating neck.⁵² Downstream of the device is a separate imaging chamber for use in imaging stationary droplets in characterization experiments.

Figure 2. Microfluidic device for encapsulation, observation, and recovery of bacterial cells.

a) Droplet microfluidic device with droplet generator and imaging chamber. Droplets are generated by flow-focusing from three inlets; one oil inlet for Pico-Surf (FC-40 fluorinated oil with 5% surfactant) and two aqueous inlets, for cell mixture and inducer cocktail, respectively. b) Detailed design of the droplet generating device and the focusing nozzle. In-line 10-μm-gap filters are placed at the aqueous inlets to trap debris. Two in-line 6-μm-gap filters are also placed at oil inlets to prevent debris from clogging the droplet generating narrowed neck. c) Detailed design of the imaging chamber, which stores droplets for imaging and cell count in characterization experiments.

In order to stabilize the fluid-fluid interfaces of the droplets against uncontrolled coalescence and to keep the bacterial culture intact inside of the droplets, the use of surfactant with good biocompatibility is necessary.^{42, 53–55} In this work, commercial Dolomite Pico-Surf, composed of 5% polyfluorinated surfactant in FC-40, was used. A high concentration of surfactant was preferred in our experiment as the rich culture medium, CDM, disrupted droplet forming abilities at lower surfactant concentrations. This Pico-Surf with surfactant was previously demonstrated to have good function in droplet microfluidic platform with high biocompatibility to cell encapsulation that can resist cell adhesion and allow cell proliferation in droplets.^56–58

We first characterized our system for its ability to encapsulate multiple cells per droplet for modelling bacterial interactions. Droplets were generated by use of the device with syringe pumps in the setup illustrated in Figure 3, using cell feedstock prepared at a nominal OD of 2.0 per strain in the mixture. Freshly generated droplets were collected in the imaging chamber, where 100 representative droplets were selected for cell counts in each of 18 trials. As shown in Figure 4, the droplet generation system formed droplets with 43% probability of encapsulating 2 cells per droplet and a total of 71% of droplets containing multiple cells. Average occupancy was 2 cells per droplet; as expected, droplet occupancy approximated a Poisson distribution.⁵⁹

Figure 3. Experimental design for pneumococcus encapsulation and genetic transformation in droplets.

Two strains of pneumococcus were grown and re-suspended in fresh medium so that each strain had an OD of 4. The strains were mixed with a volume ratio of 1:1 and stored on ice. In the cold room, oil, inducer, and cell mixture were loaded into 1-mL syringes with 23G x 1’’ blunt needles and controlled by syringe pumps. Droplet-collecting vials, prepared with 300 μL of CAT medium to prevent droplet evaporation, were maintained on ice. For a specific experiment, when droplets were stably formed, two parallel unconfined culture control experiments, positive control and Pico-exposed control, proceeded by mixing an equal volume of cell mixture and inducer in each control so that the total volume of control would be similar to the total volume of collected aqueous droplets. The droplet-collecting vials and control vials were maintained on ice during the droplet generating process, then incubated at 37°C to activate competence and allow cell-cell attack. After opening of droplets with Pico-Break, cells were diluted, incubated again at 37°C for 1 hour for gene integration and expression, and plated for selection.

Figure 4. Quantitative image analysis of encapsulation of pneumococci in droplets.

a) Droplet occupancy distribution of the system operated with cell feedstocks at OD 4.0 for each strain: 100 representative droplets were collected and imaged at each trial and manually counted during live video imaging for number of cells in each droplet; 18 trials total. b) Image of pneumococci in ~10 μm diameter droplets.

Control of competence by manipulation of culture temperature

S. pneumoniae is not constitutively competent for DNA uptake but enters the competent state after expressing several dozen genes triggered by exposure to the peptide pheromone CSP.²⁰ In the new device, cells are exposed to CSP only at the moment of droplet formation, and the waves of competence-specific gene expression are postponed by working at 4°C during droplet formation and accumulation, but unleashed by returning the cells/droplets to the standard culture temperature of 37°C. In standard culture conditions at 37°C competence for DNA uptake (and cell attack) reaches a maximum approximately 20 min later, and then declines nearly to zero by 30–35 min, as result of a feedback inhibition of the CSP receptor by a recombination protein, DprA.⁶⁰ After uptake of a single strand of donor DNA by a competent cell, two processing steps must be completed before a new gene can be fully expressed – recombination (insertion of a donor strand in place of one strand of the recipient chromosome) and replication (copying the new allele to form a template-strand copy suitable for transcription). A post-transfer incubation of 60 minutes at 37°C accommodates both processes, while a further in-agar delay before exposure to selective agents allows complete development of the transferred phenotype.³⁴

To learn whether any of the confined cell pairs could develop competence and exchange genes within droplets after extensive pausing at 0–4°C, gene transfer between attacker and victim cells was determined for cell pairs before and after periods of incubation at 37°C. Detail of the experimental setup in cold room is illustrated in Figure 3. As shown in Figure 5, the cells that were encapsulated as above but never warmed for development of competence while in droplets retained an ability to exchange genes at a high rate when released from droplets into CAT medium and warmed to 37°C for 1 hour, showing that cells emerging from droplets retained full responsiveness through the encapsulation experience and recovery, and remained capable of developing competence. At a post-release dilution of 1:10, out-of-droplet competence and gene transfer was activated and completed during 1 hour at 37°C after release from droplets, much as reported previously.⁶¹ This post-droplet cell attack and gene transfer was completely abrogated by simply increasing the post-droplet dilution ratio to 1:1000. Thus, at this dilution ratio, the confined cells developed competence and carried out lytic attack and gene transfer within droplets when they were warmed to 37°C while still enclosed within droplets.

Figure 5. Dependence of gene transfer on droplet incubation at 37°C.

Comparison of number of transformants and yield of recombinants in 25-minute incubated droplets (Droplets, 25 mins, 37°C) and non-incubated droplets (Droplets, 0 min, 37°C) at post-droplet dilution factors of 10, 100 or 1000. Yield of recombinants is calculated as a percentage of Nov^RRif^R transformants among initial input of CP2204 (Rif^R) recipients. The error bar shown in the figure is the standard deviation for number of CFU/mL.

Competence development kinetics

Because the materials used in droplet formation and handling had not been used in competence studies previously, it was unknown whether the progress of this complex developmental program would mimic its progress in standard cultures so as to provide a cell-scale model representative of the natural competence phenomenon. To compare the temporal pattern of competence development in droplets with that in bulk cultures, gene transfer was determined as a function of the time of incubation at 37°C, for parallel halves of a culture split for droplet formation in one half and simple incubation of unconfined cells in the other. Recombinants were formed in both cases at similar rates, as shown in Figure 6. The similar temporal patterns establish good biocompatibility of this droplet forming device and the selected microfluidic materials.

Figure 6. Comparison of kinetics of competence development within droplets vs. in unconfined planktonic cultures.

Seven equivalent droplet samples were collected sequentially for every 10 minutes in the cold room and then incubated at 37°C for different amount of time (0, 10, 15, 20, 25, 30, 35 minutes). Competence development is expressed as the number of Nov^RRif^R transformants per mL and yield of recombinants, as a function of incubation time. Yield of recombinants is calculated as a percentage of Nov^RRif^R transformants among initial input of CP2204 (Rif^R) recipients. The error bar shown in the figure is the standard deviation for number of CFU/mL.

Direct comparison of yields of gene transfer events in planktonic vs. confined cells

To compare the number of successful attacks and DNA transfer events inside vs. outside droplets, we made direct comparisons using cultures split, as above, immediately before starting droplet generation. Also, to test further the bio-compatibility of Pico solutions (Pico-Surf and Pico-Break) for pneumococcal transformation and viability, a parallel Pico-exposed control was prepared in which unconfined control cells were subjected to the same Pico-Surf and Pico-Break treatment as were droplets (Figure 7). A somewhat lower viable cell count was observed after recovery from droplets than for both of the unconfined controls. The viable counts for both strains recovered from droplets were approximately 10–20% of the positive control. Viable cell recovery in the Pico-exposed control was also significantly reduced (by 40% and 60% for DNA recipient and donor, respectively) compared to the positive control. The longer and more intimate contact with Pico reagents during encapsulation may explain the reduced apparent viability for confined cells. We expect that if viable cells recovery is increased in future work, the transformation yield might also increase further.

Figure 7. Nov^R and Spc^R gene transfer in micro-droplets and in unconfined cultures.

Nov^R and Spc^R gene transfer in micro-droplets and in unconfined cultures are shown as total transformants and yield of recombinants in droplets, in a positive (+) control, and in a Pico-exposed control (Pico Control) experiment. Yield of recombinants is calculated as percentage of total transformants for each antibiotic resistance marker among initial input CP2204 (Rif^R) recipients. a) Number of transformants and yield of recombinants for Nov^R marker. b) Number of transformants and yield of recombinants for Spc^R marker. c) Recovery of viable cells of each strain. CP2204 recipient determined as Rif^R colonies; CP2215 donor determined independently as Nov^R or Spc^R colonies. The error bar shown in the figure is the standard deviation for number of CFU/mL.

The number of the competent cells transformed for the single-gene Nov^R and Spc^R markers and the yield of recombinants in droplets by the was approximately 50% of the controls (Fig. 7a–b). However, the difference in number of transformants and yield of recombinants between droplets and controls was not statistically significant.

Development of competence in pneumococcus is known to be sensitive to interference by different environment factors. By encapsulating the cells in femtoliter-scale droplets, the conditions for the cells are changed significantly. The large number of recombinants for Nov^R and Spec^R donor genes shown in Figure 7 demonstrates that both development of competence and the cell-cell attack capabilities of competent cells are compatible with encapsulation in droplets with Pico solutions. Considering that we only assayed transfer of a single gene at a time out of the totality of 2000 genes, we estimate that a substantial fraction of co-encapsulated donor/attacker cell pairs are exchanging genes within the droplets of our prototype device.

Discussion

Competence for natural genetic transformation in pneumococcus is controlled by CSP,⁶² one of the components in the inducer cocktail prepared with CSP, BSA, CaCl₂ and CDM. In the presence of CSP, development of the competent state is activated, and transformation occurs through successive stages as summarized in Figure 1. Nearby non-competent cells are lysed through contact with competent cells, followed by binding of released double-stranded DNA, uptake of single-stranded DNA fragments by the competent cells, and homologous recombination.^63–65

Thus, natural transformation of pneumococcus reflects the presence and pairing of both competent and non-competent cells. By confining two different strains together within femtoliter droplets, we aimed to achieve a better model of this process. Our system generated droplets with 1–5 (average 2) cells per droplet. By starting with 1:1 mixture of competent and non-competent cells, we expected to have both strains in an individual droplet a practicable frequency, to model cell-cell attack. With an average of 2 cells per droplet, the probability to have 2 cells in a droplet is approximately 30%.⁵⁹ To optimize our system for observing and controlling the specific cell-cell attack between competent and non-competent cells, initial cell density or droplet size can be further adjusted, or a droplet sorting method could be incorporated in a later iteration of the current system.

In the natural interaction, pneumococcus achieves the competent state by relying on quorum sensing (QS), a mechanism by which bacterial cells communicate at the intercellular level to monitor their density.⁶⁶ For sensing a high density of cells, naturally competent pneumococcal strains secrete CSP via an ABC transporter (ComAB) and sense its level to initiate expression of multiple CSP-responsive genes termed early and late competence (com) genes. However, in this work, to be able to control the competence process in the laboratory, a synthetically produced CSP was used, in conjunction with the comA mutant CP2204, which cannot export CSP itself, ensuring that its competence in droplets is totally dependent on the exogenous synthetic CSP. Synthetic CSP has been used widely in pneumococcus transformation studies for the same reasons.^{33, 67–69} Upon exposure to CSP, competence reaches maximum by about 20 minutes at 37°C.⁶² Because direct comparison of kinetics of competence development showed a similar temporal pattern, both inside and outside droplets, it appears that this microdroplet environment will provide a good model for exchange events occurring in vivo.

Using the new device, we encapsulated pneumococcal cells in 10 μm diameter droplets. There are two advantages to confine the cells in a specific 10 μm diameter droplet: 1) to have equivalent cell densities in encapsulating droplets and in a control bulk mixed culture, at a density ~5×10⁸ cells/mL previously shown optimal for pneumococcal cell-cell lytic attack, and 2) to establish a confinement suitable for droplet micro-manipulation to confine the free DNA released from lysed non-competent cells from diffusing away into a large volume and for direct cell imaging. In the limited space of droplets, pneumococcal cells remained capable of developing competence and successfully transformed when the droplets are warmed to 37°C. When we released the cells from droplets and diluted them with a modest post-droplet dilution of 1:10, the competent cells remained capable of developing competence and conducting, cell-cell attack and gene exchange. For the bacterial transformation within droplets, a post-droplet dilution factor of 1000-fold was routinely used, as it forestalled post-droplet transformation simply by greatly increasing cell-cell separation.

The numbers of both Nov^R and Spc^R recombinants in droplets were 50% less than those in the control, a modest difference considering the 75% reduction in the viable cell counts in droplets. The reduction in viable cell counts may reflect several factors involved in the design for encapsulation bacterial cells in droplets, such as limited resources in droplets, the high shear experienced during droplet formation or a modest bio-incompatibility of Pico solutions with the bacterial strains, or a modest level of clumping, which would reduce the viable cell counts more than recombinant yields. To test if the Pico solutions themselves affect cell viability, we mixed a control cell culture with Pico-Surf and Pico-Break, labeled as Pico-exposed control, and performed the standard emulsion breaking steps with droplets. Based on our results (Figure 7), it appears that the Pico solutions, as contacted directly with our bacterial cells, can affect the cell survival rate. In observing cell movements within the droplets in the device, we noticed that in some droplets, the cells tended to clump together, possibly reducing CFU counts. If the viable cell recovery can be improved in droplets by optimizing culture parameters, it is possible that the yield may also be increased further.

The confinement of pneumococcal cells in 10 μm diameter droplets achieved here offers several advantages at the cellular level for the study of genetic transformation. Competent and non-competent cells are close to each other and free fragments of DNA released from the non-competent cells are confined near the competent cells. Because DNA release of non-competent cells depends on heterolysis, in which the non-competent cells are killed by the lysins attached on competent cells surface,⁶⁵ reducing the space between cells can create a better environment for observing cell-cell contact. The DNA release from lysed non-competent cell was demonstrated to be time and space dependent as maximum DNA release and DNA uptake reach at the same time. Most important, all progeny from a single cell-cell transformational interaction can potentially be captured and characterized, to define the complete set of resulting recombination events. Thus, encapsulating competent and non-competent bacterial cells in droplets can improve cell-cell contact and competence, DNA release, and uptake between strains for transformation process, establishing a better model of transformation process compared to bulk mixed culture.

Conclusions

We demonstrated for the first time that it is possible to model pneumococcal genetic transformation in droplets by encapsulating competent and non-competent strains within femtoliter-scaled droplets and allowing DNA release and uptake to occur therein. The droplet generating system encapsulated multiple cells per droplet, with 71% probability of 2 or more cells per droplet. Transformation events within these droplets yielded a high number of recombinants, representing transfer of Nov^R and Spc^R markers to competent cells. The effectiveness of droplet microfluidics for the studies of gene transfer between pneumococcus strains provides several advantages, including an improved ability to monitor individual cell behaviors in isolated droplets whose contents can be recovered for downstream genetic analysis, promising an improved understanding of the precise mechanism of bacterial transformation, which is a critical link in finding ways to mitigate antibiotic resistance.