Quantitative analysis of bacterial cell-cell communication at the single-cell level using microdroplet arrays Authors: Corine Jackman Burden, Matthew Hou, Anne Liu, Lydia Eutsey, Rory Eutsey, Joel Greenhouse, Frederick Lanni, Shelley L Anna, N Luisa Hiller Published: 2026 ABSTRACT AbstractCell-cell communication (CCC) contributes to bacterial survival and adaptability. Gram-positive bacteria employ secreted peptides to coordinate CCC. While the molecular pathways activated by these peptides are well studied, little is known about how individual cells contribute to initiating the signaling response. To address this question, we used microdroplet arrays to examine the major human pathogen Streptococcus pneumoniae and its TprA/PhrA regulator/peptide CCC system, which promotes colonization and virulence. We measured phrA promoter activity in wild-type (WT) cells and in a phrA deletion mutant, using populations seeded before signaling began. As signaling emerged, we observed heterogeneity in S. pneumoniae signaling within and across microdroplets. Addition of exogenous PhrA increased both the magnitude of signal and the percentage of signaling cells, yet it did not reduce the heterogeneity of signal. When examining whether PhrA peptide produced from WT cells was shared with ΔphrA cells, we found a preference for self-signaling over signaling to neighboring cells. Overall, we developed a platform to quantify cell-cell signaling at the single-cell level and determined that at early stages TprA/PhrA signaling is highly heterogeneous and primarily targets producing cells. We propose that this heterogeneity and its amplification through autoinduction may confer a fitness advantage to the population.Subject terms: Pathogens, Bacterial pathogenesis Introduction The organization of cells into communities allows for the development of complexity and new functions1. In multicellular organisms, cells differentiate into specialized types enabling 3D architectures and tissue-level functions. In single-cell organisms, population-level behavior is achieved by coordination of cells in a community, which may be multi-species, multi-strain or clonal. A classic example of population-level behavior in bacteria is the development of biofilms2,3. Cells in the biofilm mode-of-growth generate a physical structure by coordinating production of the biofilm matrix (extracellular polymeric substances) and benefit from this architecture by becoming more recalcitrant to environmental stressors, such as antibiotics and immune responses. In bacteria such as Streptococcus pneumoniae, population-level behaviors are achieved via cell-cell communication (CCC) in a process that depends on the secretion and sensing of chemical signals4,5. In general, bacterial CCC is framed as a process that is active when the benefits of cooperation outweigh those of isolation. Parameters of CCC include production of the signal, rate of diffusion/degradation of the signal, and sensitivity to the signal, all of which can modify dynamics and sharing of the signal6,7. CCC is usually measured in a population and averaged over cell number, with the underlying assumption that signal is evenly distributed and sensed uniformly throughout the community, producing homogeneous responses. Yet, growing evidence suggests that single cells in a population vary regarding the extent of their response to signal resulting in phenotypic variation, which may contribute to cell survival, adaptability, and virulence8–10. Different exposure to environmental cues, stochastic fluctuations, or epigenetic differences can drive phenotypic heterogeneity within clonal populations9,11–13. Heterogeneity driven by a concentration gradient can be described as a response to an environmental cue. In contrast, heterogeneity, where cells in the same conditions respond differently to shared cues, may reflect population strategies. These strategies can include bet-hedging in a fluctuating environment or establishing a division of labor14. Quantifying heterogeneity and revealing the underlying molecular mechanisms generating diversity requires measurements at the single-cell level. S. pneumoniae is a major human pathogen responsible for almost one million deaths in children annually15,16. Colonization of the upper respiratory tract is asymptomatic and a prerequisite for disease, which develops when S. pneumoniae disseminates from the nasopharynx to other tissues, such as lungs, middle ear, blood, heart, and brain17. Akin to most Gram-positive bacteria, S. pneumoniae performs CCC using secreted peptides. These peptide-mediated signaling systems play roles in horizontal gene transfer, strain competition, biofilm development, and capsule expression among others11,18–25. Studies in animal models of S. pneumoniae carriage and pneumonia demonstrate that CCC systems contribute to colonization and virulence, highlighting the importance of understanding their molecular basis11,26–28. The S. pneumoniae CCC peptides either bind to two component systems on the cell surface or are internalized into the cytosol where they bind to cognate receptors5. The latter set are part of the Rap, Rgg, NprR, PlcR, PrgX, and AimR (RRNPPA) superfamily, and S. pneumoniae strains encode multiple Rgg and PlcR (known as Tpr in S. pneumoniae) systems29. This study focuses on the one Tpr/Phr system that is widely distributed in the S. pneumoniae pangenome and serves as a model for peptide-mediated CCC30. In S. pneumoniae, in addition to extensive genomic diversity31,32, there is also phenotypic variation within clonal populations. The first evidence of such variation came from spontaneous shifts between transparent and opaque colony types on agar plates33. These reflect changes in the capsule and cell wall composition and translate into site-specific fitness differences during mammalian infection34,35. Mechanistically, these phase shifts are driven by a restriction- modification system that causes reversible changes in methylation patterns, thereby altering the transcriptome and influencing virulence12,13 and transformation efficiency36. Heterogeneity has also been observed in CCC. S. pneumoniae cells can organize into chains, and when these are exposed to some antibiotics, the secreted competence stimulating peptide (CSP) is retained by cells within the chain, increasing competence in CSP-producing cells while reducing CSP diffusion as a shared good37. Finally, a study using chromosomal barcoding and high-throughput sequencing demonstrated that when clonal populations are introduced into a murine model of colonization, a small number of cells dominate the population in a process driven by secreted bacteriocins and predation of clonal kin38. Together, these findings illustrate multiple forms of single-cell variability in S. pneumoniae that influence virulence, gene transfer efficiency, and colonization. In this work, we use microdroplet arrays to study the early stages of pneumococcal CCC. We quantify signaling at the single-cell level and use the platform to assess the diversity of signal transduction both within and between small communities, as well as the degree to which signals are uniformly shared across the population. Given the widespread distribution of peptide-mediated CCC systems, our findings are likely relevant to many species of Gram-positive bacteria. Results Generation of microdroplets and encapsulation of pneumococci We used microdroplet arrays to investigate how individual cells contribute to CCC within their local microenvironments. First, we developed a protocol to generate uniform microdroplets and to encapsulate bacteria for growth. Microfluidic devices were used to generate microdroplets using a water-in-oil suspension and the viscosity, flow rate, and dimensions of the orifice were adjusted to generate microdroplets of 55 µm in diameter39. To test for uniformity (Fig. S1A), we conducted four experiments, each one with at least 1000 microdroplets, to calculate the polydispersity index (PDI). The low PDI indicated high uniformity across experiments (Fig. S1B). Second, we encapsulated bacteria in microdroplets and assessed growth (Fig. 1B). To quantify bacterial numbers, we transferred seeded microdroplets into hexagonal chambers for imaging (Fig. S2). To ensure batch-to-batch consistency, we used chemically defined medium supplemented with glucose (CDM-Glu). At time zero, microdroplets contained early log-phase pneumococci in CDM-Glu. Growth measurements revealed a sharp decline in viability after 4 h (Figs. S3A, B), consistent with previous reports attributing this to hydrogen peroxide accumulation40,41. Supplementing CDM-Glu with catalase restored growth, and maintained viability (Figs. S3A, B). Therefore, microdroplets were routinely generated in CDM-Glu supplemented with 300U/ml of catalase. Encapsulation and growth of pneumococci in microdroplets for analysis of PhrA signaling. A Model of TprA/PhrA signaling pathway showing phrA gene (gray arrow) and unprocessed translated peptide (light and dark gray chain), export of PhrA via the Sec pathway (orange circle), processed PhrA (dark gray chain) imported via AmiA-F (blue), and processed peptide binding to the cognate receptor TprA (yellow) releasing inhibition of the phrA promoter driving mCardinal expression (pink). A constitutive promoter (Pc) drives the expression of mNeonGreen (green). B Design of the microfluidic microdroplet platform employed to keep microdroplets stationary while imaging (top). Flow chart outlining steps of the experiments: bacterial growth in tubes, encapsulation of cells into microdroplets, bacterial growth in microdroplets, and imaging of bacteria. C Scatter plot displaying the number of cocci (WT strain) per microdroplet at time of microdroplet generation (0 h), and at 1, 2, 4, and 6 h post-seeding (n = 3, color coded per experiment). The average number of cells is 18 ± 9 (0 h), 33 ± 17 (1 h), 61 ± 30 (2 h), 100 ± 49 (4 h), and 140 ± 61 (6 h) cocci. D 3D reconstitution of confocal images of the pneumococcal chain and coccus (i, ii) and diplococcus (iii, iv) expressing mNeonGreen under a constitutive promoter; panels ii and iv display the outlines generated by the surface feature in IMARIS software that was used to enumerate cocci. E Representative max z-projection of confocal images of WT cells displayed for multiple time points post-seeding, constitutively expressing mNeonGreen. 40x objective, Scale bar: 50 microns. F Representative max z-projections of confocal images of WT and ∆phrA strains at 4 h showing constitutive mNeonGreen, PphrA-driven mCardinal, and merged channels. Transmitted images show the equator of the microdroplet. Scale bar: 50 microns. 40x objective G Heatmaps of the z-scores for average intensity of PphrA (left panels) or constitutive promoter, Pc (right panels) in 30 different microdroplets over 0, 1, 2, 4, and 6 h post-encapsulation. The z-scores are calculated relative to the ∆phrA and standardized to each time point. mCardinal expression in the WT is statistically different from the ∆phrA at all time points (p < 0.0001). mNeonGreen expression in the WT is only statistically different from the ∆phrA at 4 h (**, p < 0.001;* <0.01, ***<0001). Confidence intervals for each time point and experiment are in Table S4. Our goal was to seed microdroplets with a low initial inoculum that supports growth while minimizing empty microdroplets. Seeding that contains no more than 1 CFU per microdroplet results in more than 90% of microdroplets being empty due to the Poisson distribution of CFUs. Therefore, we targeted ~8 CFUs per microdroplet, a condition in which more than 99% of microdroplets contain cells (Fig. S3D). To verify inoculation, microdroplets were collected, coalesced using a destabilizer, and plated for CFU enumeration. At time zero (t = 0), the mean inoculum across experiments was 8.5 ± 0.5 CFU per microdroplet, followed a Poisson distribution, and more than 99% of microdroplets contained between 1 and 20 cells (Fig. S3D). For image-based quantification, we constructed strains that constitutively express mNeonGreen (Fig. 1A). Time-series images were analyzed in IMARIS using a seed filter that measured total fluorescence per coccus, independent of whether cells were arranged as diplococci or in chains (Fig. 1D). At t = 0, this approach yielded 18 ± 9 cocci per microdroplet (Fig. 1C). Cell numbers increased markedly up to 4 h, with minimal changes between 4 h and 6 h (Fig. 1C, S3A). Growth curves derived from CFU counts (Fig. S3A) and imaging (Fig. 1C) showed the same pattern: marked increases up to 4 h followed by minimal changes between 4 h and 6 h (Fig. 1C, S3A). Doubling times for the 0–4 h interval were 1.65 ± 0.42 h (CFUs) and 1.66 ± 0.13 h (imaging). These results confirm robust growth within microdroplets, and therefore we focused subsequent analyses on data collected up to 4 h. In summary, we established a reproducible protocol for generating uniform microdroplets, encapsulating and growing S. pneumoniae, and enabling confocal imaging. This workflow combined a newly designed microfluidic device with an image analysis pipeline to quantify growth at the single-cell level across small communities. These procedures were applied in all subsequent experiments to capture pneumococcal signaling. Pneumococcal cell signaling in microdroplets To capture cell-cell signaling at the single-cell level, we focused on a well-characterized peptide-regulator CCC system: TprA/PhrA. The TprA/PhrA system is a virulence determinant composed of an intracellular transcriptional regulator (TprA) and its cognate autoinducing peptide (PhrA)27,28,30. PhrA is processed and secreted from the producing cell, and the active C-terminal form is imported into producing or neighboring cells via the AmiA-F transporter. Once in the cytosol, PhrA binds to TprA relieving TprA-mediated inhibition (Fig. 1A)30,42. We constructed a reporter for TprA/PhrA signaling by cloning the phrA promoter (PphrA) upstream of the fluorescent protein mCardinal43. In this system, induction of PphrA results in mCardinal producing cells (PphrA-hlpA-mCardinal), referred to as WT (Fig. 1A). As a control for the promoter’s basal levels of PhrA-independent activity, we created a strain with a deletion of the phrA gene (ΔphrA PphrA-hlpA-mCardinal) referred to as ΔphrA. We compared the growth of these strains using our image analysis pipeline and determined that the WT and ΔphrA reporter strains display the same doubling time (Fig. 1C and Fig. S4). Importantly, adding exogenous peptide to the ΔphrA strain induces the PphrA promoter in the same manner as in the WT background, demonstrating that deletion of the gene encoding the peptide does not impair the strain’s ability to import or respond to PhrA (Fig. S5A). Throughout the experiments, we employed the WT and ΔphrA reporter strains to measure TprA/PhrA signaling at the single-cell level. To quantify PhrA signaling and enumerate cells, we grew WT and ΔphrA strains in microdroplets (Fig. 1E) and imaged the PphrA-driven mCardinal expression and constitutive mNeonGreen, respectively (Fig. 1F). To assess changes in PphrA signal over time, we enumerated each cell and quantified its fluorescence. For each time point, we calculated the average fluorescence per microdroplet from at least 30 microdroplets and visualized the data as heatmaps (Fig. S6). We also tested whether PhrA peptide diffuses between microdroplets by examining activation of the phrA promoter in ΔphrA cells adjacent to WT microdroplets. No activation was observed, indicating that peptide does not diffuse across microdroplets, or, if it does, the effect on phrA expression is negligible (Fig. S3E). Fluorescence measurements of cell averages within microdroplets revealed that PphrA signal in WT cells increases over time, reaching peak induction by 4 h. To provide statistical context, we displayed the standardized fluorescence values as z-scores (Fig. 1G). To measure the number of standard deviations the average microdroplet intensity differed from the ΔphrA mean, we calculated z-score for both the WT and ΔphrA strains relative to ΔphrA at each time point and applied a threshold at ± 2.57 (CI = 99%). For WT cells, the percentage of microdroplets with PphrA signal above this threshold was 18 ± 16% (1 h), 52 ± 5% (2 h), 64 ± 18% (4 h), and 74 ± 10% (6 h) (Fig. 1G, bottom panels), demonstrating progressive activation of signaling. Moreover, the average cell density at 2 h was 61 ± 31 cells per microdroplet (calculated from 110 microdroplets across 3 experiments using microscopy), suggesting that PhrA signaling is activated with fewer than 100 cells in a microdroplet. As a control, we measured the z-scores for mNeonGreen driven by the constitutive promoter. In contrast to PphrA, the constitutive signal showed only minimal differences between WT and ΔphrA strains: 19 ± 16% were below the threshold of -2.57 at 2 h and 9 ± 8% were above this threshold at 4 h (Fig. 1G, right panels). The slight increase in mNeonGreen signal in both WT and ΔphrA cells over time is likely due to the accumulation of fluorescent protein. We conclude that the PphrA signal increases in a PhrA-dependent manner, whereas constitutive expression shows only minor fluctuations. Overall, these data capture PhrA-mediated cell-cell signaling and establish conditions for single-cell measurements of early TprA/PhrA communication. Heterogeneity in PhrA signaling across single cells Quorum sensing models often assume a uniform response from all cells in a population4. However, when visualizing pneumococcal cells, we observed variability in PhrA signaling across cells within and between microdroplets (Fig. 1F). This qualitative assessment at the PphrA signal led us to the hypothesis that PhrA signaling is heterogeneous across individual cells. Thus, we set out to capture the range of cell-cell signaling levels within and across microdroplets. Our data consist of three independent experiments, each represented in Fig. 2, S7, or S8, where threshold values were calculated with data polled from all three experiments. Additional single-cell data for the constitutive control (Fig. S9) and the log signal intensity at each time point (Fig. S10) are shown for three experiments. Heterogeneous PhrA signaling within and across microdroplets. A–D Data displays one representative experiment; the data from the remaining two experiments is presented in Figures S7 and S8. A Heatmaps depict the ratio of the median signal for the WT to ΔphrA strains at 0, 1, 2, 4, and 6 h post-seeding for PphrA mCardinal signal (Ai) and Pc mNeonGreen signal (Aii and Fig. S9). B Violin plots depict the signaling levels of individual cells for the WT and ΔphrA where the bold dashed lines represent the median and dotted lines denote the quartiles. The number of cells per measurement are: n = 615 (0 h), n = 934 (1 h), n = 1455 (2 h), n = 2634 (4 h), and n = 3423 (6 h) for the WT and n = 738 (0 h), n = 1274 (1 h), n = 1426 (2 h), n = 2823 (4 h), and n = 5185 (6 h) for the ΔphrA strain. For PphrA, the WT (pink) is statistically greater than the ΔphrA (gray) at all time points (***, p < 0.0001), (Bi). The gray shading is the basal threshold defined as the 90th percentile of individual ΔphrA cells with respect to time. For the constitutive promoter, the WT (green) is greater than the ΔphrA (gray) at 0 h, 4 h, and 6 h (CI = 99%), however with a minimum fold change of 1.03 (0 h), 1.02 (4 h), and 1.02 (6 h) (Bii, Aii). C Representative distribution of fluorescent signaling intensity comparing the heterogeneity (shape) and signal intensity (shift) of cells at 4 hours. The number of bins per graph was calculated using the Freedman–Diaconis Rule (Fig. S10). PphrA signal in WT (pink, n = 2634 cells) and ΔphrA (gray, n = 2823 cells) (***, p < 0.0001 (Ci). Pc signal in WT (green, n = 2361 cells) and ΔphrA (gray, n = 2548 cells) (***, p < 0.0001) (Cii). The basal threshold (2940 arbitrary units (AU); log = 3.47) is shown by the black vertical line. D Graphs display the signal intensity of PphrA-driven expression (Di) and Pc-driven expression (Dii) at 4 h for individual cells in each sampled microdroplet. Open circles are ΔphrA strains and closed circles are WT strains. Red and green dashes represent the median. PphrA-driven expression between the WT (pink) and ΔphrA (gray) strains, values are statistically significant (***, p < 0.0001). Microdroplets containing WT cells (WT microdroplets) were divided into three categories for the 37 microdroplets analyzed for PphrA: (i) 3 microdroplets (#’s 10, 15, 24) have a median value of expression less than 3 microdroplets that contain ΔphrA cells (CI = 99%), (ii) 22 microdroplets (#’s 1, 6, 7, 8, 9, 10, 12, 13, 15, 16, 17, 18, 20, 21, 22, 23, 24, 25, 26, 28, 34, 36) are not statistically different (ns) from most ΔphrA-containing microdroplets (CI = 99%), (iii) 15 microdroplets (#’s 2, 3, 4, 5, 11, 14, 19, 27, 29, 30, 31, 32, 33, 35, 37) have a median value of expression statistically greater than most ΔphrA-containing microdroplets (CI = 99%) (Dii). For constitutively-driven expression between the WT (green) and ΔphrA (gray) strains, values are not statistically significant from ΔphrA-containing microdroplets (CI = 99%, ns). Specifically: (i) No WT microdroplets have a median value of expression less than any ΔphrA-containing microdroplets (CI = 99%), (ii) all 37 WT microdroplets are ns from most ΔphrA-containing microdroplets (CI = 99%), (iii) no WT microdroplets have a higher median than most ΔphrA-containing microdroplets (CI = 99%). Most is defined as more than half. (Di). We first collected single-cell fluorescence data from all microdroplets in each experiment and plotted the values (Fig. 2B-C, S7B-C, S8B-C). As expected, the WT cells consistently displayed higher PphrA signaling than the ΔphrA baseline. To quantify this difference, we calculated the ratio of the median WT signal to the median ΔphrA signal at multiple time points (Fig. 2A, S7A, S8A) and defined the threshold as 90% of the ΔphrA reporter signal. Across all experiments, the ratio of PphrA signal (mCardinal) relative to the basal level ranged from 1.37 to 1.99 across the sampled time points (Figs. 2Ai, S7Ai, S8Ai). In contrast, the ratio of the median constitutive signal (mNeonGreen) remained near 1, ranging from 0.94 to 1.05 (Figs. 2Aii, S7Aii, S8Aii). These results suggest that the PphrA reporter, but not the constitutive promoter (Pc) reporter, is induced in the WT cells. To capture signaling dynamics, we plotted the log-transformed frequency distribution of single-cell fluorescence at the four-hour time point (Fig. 2C, S7C, S8C). PphrA activity showed a broad range: in the experiment illustrated in Fig. 2, the log values range from 0.845 to 4.76, such that 39% of WT cells exceed the ΔphrA threshold (Fig. 2Ci, pink). The replicate experiments showed a similar pattern, with 42% and 41% of WT signal above the ΔphrA threshold (Figs. S7Ci & S8Ci). In contrast, this pattern is not observed in the constitutive promoter control, where signal for WT and ΔphrA overlapped extensively (Figs. 2Cii, S7Cii, S8Cii). Across all experiments, approximately half (49 ± 7.7%) of the PphrA signal from the WT cells are statistically different from those of most of the ΔphrA cells (Figs. 2Di, S7Di, S8Di) consistent with PphrA induction. Only a small portion (2.8 ± 4.8%) differed for the constitutive promoter (Pc) (Figs. 2Dii, S7Dii, S8Dii). Together, the wide range and distribution of the PphrA values across single cells suggests heterogeneous signaling. We next examined whether signaling variability is observed across, within microdroplets, or both. To this end, we plotted the single-cell fluorescence values grouped by microdroplet (Fig. 2D, S7D, S8D). As expected, microdroplets containing ΔphrA cells display uniformly low phrA activity (Figs. 2Di, S7Di, S8Di). In contrast, microdroplets containing WT cells displayed multiple distinct patterns. For example, in the microdroplets displayed in Fig. 2D, we observed at least three patterns. First, microdroplets with almost no detectable signal (examples microdroplets #10, 15, 24); second, microdroplets resembling most ΔphrA cells (examples microdroplet #1, 6, 28); third, microdroplets with overall higher signal than ΔphrA cells (examples microdroplet #2, 14, 31). These data capture the variability in the brightness of cells in individual microdroplets. Finally, we recorded the variability of each cell’s brightness within the microdroplets across three experiments. For ΔphrA cells, the range within microdroplets was 2128 ± 556 (Q1 = 809 ± 361, Q3 = 1503 ± 674) for the five microdroplets with the tightest distribution of fluorescence and 12,029 ± 2267 (Q1 = 1184 ± 241, Q3 = 2593 ± 557) for the five microdroplets with the widest distribution of fluorescence. In contrast, for the WT cells, values were 2848 ± 1,206 (Q1 = 769 ± 215, Q3 = 1810 ± 788) and 31,454 ± 20,000 (Q1 = 1590 ± 796, Q3 = 4941 ± 1654), respectively. The substantial variation observed within the microdroplets indicates heterogeneity of PhrA expression in cells within the same community. Additionally, differences in variability across microdroplets suggest heterogeneity between distinct communities. Together, these data reveal single-cell PhrA signaling heterogeneity both within and across small communities (qualitatively observed in Figs. 2Di, S7Di, S8Di). Exogenously added synthetic PhrA increases overall induction without decreasing the heterogeneity of the signal If heterogeneity is an intrinsic property of the TprA/PhrA signaling pathway, then the diversity of signaling across cells should not decrease when overall signal intensity increases. In contrast, if heterogeneity is a consequence of low signaling, then increasing signal intensity towards saturation should drive the response towards uniformity and decrease the diversity of signal across single cells. To test this, we added PhrA peptide to the media to simulate a high PhrA signaling environment, as has been previously observed in planktonic cultures30,42 (Fig. 3A, S11A, S12A). To this end, we used 1 μM of synthetic peptide corresponding to the C-terminal 10 amino-acids of PhrA (SNGLDVGKAD)30,42. This experiment was performed in triplicate and each iteration is illustrated separately in Fig. 3, S11 and S12. Extracellular PhrA induces signaling but does not decrease the degree of heterogeneity in signaling. A–G Data displays one representative experiment; the data from the remaining two experiments is presented in Figs. S11 and S12. A Representative max z-projection of confocal images of WT cells at 4 hours with (WT + peptide) and without (WT) the addition of exogenous PhrA. Panels show Pc-driven mNeonGreen, PphrA-driven mCardinal, merged fluorescent channels, and transmitted images focused on the equator of the microdroplet. Scale bar: 50 microns (objective 40x). B Graph of the mean fluorescent intensity of PphrA (Bi) and Pc (Bii) per microdroplets over time. The number of microdroplets analyzed for the WT are: n = 37 (0 h), 35 (1 h), 37 (2 h), 37 (4 h), 32 (6 h) and for the WT + peptide condition: n = 35 (0 h), 37 (1 h), 32 (2 h), 35 (4 h), 38 (6 h). The values for PphrA-driven fluorescence between WT (gray) and WT + peptide (pink) are statistically different by 2 hours post-seeding (CI = 99%) (Bi). Pc-driven fluorescence is not statistically different between the WT (gray) and WT + peptide (green) strains at any time point (CI = 99%). C Heatmaps display the z-scores for average intensity of PphrA (Ci) and Pc (Cii) in 30 different microdroplets over time. The z-scores were calculated relative to the untreated WT at the same time point and panels represent the WT (left panel) and the WT treated with peptide (right panel). PphrA driven expression in the treated (WT + peptide) condition is statistically different from the WT after 2 h (CI = 99%) (Ci). Pc-driven expression of the treated and untreated cells show no statistical difference across experiments (CI = 99%) (Cii). Confidence intervals for C are presented in Table S4. D Heatmaps display the quantification of signaling in WT and WT + peptide conditions. Heatmaps depict the ratio of the median signal for WT + peptide versus WT at 0, 1, 2, 4, and 6 h for PphrA activity (Di) and constitutive activity (Diii). Percentage of signaling cells, where the basal threshold for signaling was set to the 90th percentile of ΔphrA signaling across the three experiments presented in Fig. 1Gi. E Violin plots present the brightness of all cells for WT and WT + peptide conditions, in a manner that does not consider in which microdroplet the cell is localized. Bold dashed lines: median and regular dotted lines: quartiles. The number of cells analyzed for the WT are: n = 732 (0 h), n = 1173 (1 h), n = 1766 (2 h), n = 3213 (4 h), and n = 4011 (6 h)), and for WT + peptide are: n = 1070 (0 h), n = 2034 (1 h), n = 2585 (2 h), n = 4107 (4 h), and n = 6686 (6 h). For the PphrA driven signal the WT + peptide (pink) is statistically greater than the WT (gray) after 1 h (CI = 99%) (Ei). The gray shading is the basal threshold with respect to time. For the constitutively driven signal the WT + peptide condition (green) was statistically different from the WT (gray) at 4 and 6 h, yet for a fold change of 1.03 (4 h) and 1.02 (6 h) (CI = 99%) (Eii) (Fig. S14). F Distribution plot of fluorescent signaling intensity to visualize the heterogeneity (shape) and signal intensity (shift) of WT and WT + peptide cells at 4 h. Width mean ratio is 3.28 (WT) and 3.34 (WT + peptide). The number of bins per graph was calculated using the Freedman–Diaconis Rule. PphrA signal in WT + peptide (pink, n = 4108) is highly significantly different than WT (gray, n = 3213), (CI = 99%, p < 0.0001) (Fi). Constitutive signal in the WT + peptide condition (green, n = 4071) is significantly different than in the WT (gray, n = 3195) (***, p < 0.0001). The basal threshold is shown by the black vertical line (Fii). G Graphs display the signal intensity of PphrA-driven expression (Gi) and constitutive expression (Gii) at 4 h for each sampled microdroplet (‘P’ denotes microdroplets seeded with peptide, and red and green dashes represent the median). PphrA expression is statistically different between WT and WT + peptide collectively (***, p < 0.0001) (Gi). The 35 microdroplets are divided into three categories as described in Fig. 2D: (i) 3 (#’s 18, 25, 33) have a median value of expression less than 4 microdroplets that contain untreated cells (WT microdroplets) (CI = 99%), (ii) 27 (#’s 1, 3, 4, 6, 7, 8, 10, 11, 12, 13, 14, 16, 18, 19, 20, 21, 22, 23, 24, 25, 26, 28, 29, 30, 32, 33, and 35) are not statistically different (ns) from most WT microdroplets (CI = 99%), (iii) 7 (#’s 2, 5, 9, 15, 17, 31, 34) have a median value of expression statistically higher than most WT microdroplets (CI = 99%). Constitutive expression of mNeonGreen is statistically different between the treated and untreated conditions collectively (CI = 99%, p < 0.0001) (Gii). After dividing into three categories we found: (i) 1 (# 16) has a median value of expression less than 1 WT microdroplet (CI = 99%), (ii) All P microdroplets are ns from most WT microdroplets (CI = 99%), in fact 15 (#’s 1, 4, 5, 11, 18, 19, 20, 22, 23, 25, 26, 27, 28, 31, 34) are ns from all WT microdroplets (CI = 99%), (iii) No P microdroplets have a higher median than most WT microdroplets (CI = 99%). Most is defined as more than half. To ensure that pneumococcal cells were not exposed to peptide before encapsulation, we adapted a flow-focusing device with two inlet streams to combine the cells and the peptide immediately upon formation of microdroplets44. We first analyzed the data by averaging the fluorescence intensity values per microdroplet. At 0 hours, average PphrA activity was the same between WT and WT + peptide conditions, but by 4 hours, all replicates displayed a statistically significant difference between both conditions (Figs. 3Bi, S11Bi, S12Bi). In contrast, the reporter for the constitutive promoter displayed similar levels between the WT and WT + peptide conditions (Figs. 3Bii, S11Bii, S12Bii, S13, S14). To interpret these averaged intensity values statistically, we presented them as z-scores, representing the number of standard deviations each value deviates from the population mean (Fig. 3C, S11C, S12C). To measure the effect of exogenous peptides on promoter induction, the z-score for the treated condition (WT + peptide) was standardized to that of the untreated WT condition for each time point. Results showed the exogenous peptide induces the phrA promoter after 2 and 4 hours of incubation compared to the WT control (Fig. 3Ci, S11Ci, S12Ci). To evaluate the proportion of microdroplets where the z-score of cells was above the threshold, we calculated the percentage of microdroplets that contained treated cells with PphrA activity that was statistically significant from the population mean of the WT cells. Of the treated conditions, 2.2 ± 1.9% (0 h), 2.2 ± 1.9% (1 h), 13 ± 13% (2 h), 24 ± 6.9% (4 h), and 25 ± 15% (6 h) (n = 3) of microdroplets signal above the threshold while untreated microdroplets do not exceed 2.2 ± 1.9%. The same analysis performed for Pc activity revealed that the average for both treated and untreated cells did not exceed 4% for all time points; where the average z-score at 4 hours was 0.67 ± 1.20 for Pc. Overall, we conclude that the Pc-driven mNeonGreen control is relatively stable considering the fluctuations in the ratio of the median signal for WT + peptide versus WT ranging between 1.02 to 1.07 (Fig. 3Diii, S11Diii, S12Diii, S14). At 4 h post-seeding, the ratio for PphrA signal in all cells increased to 1.57 ± 0.10 fold (n = 3) (Fig. 3Di, S11Di, S12Di) while the constitutive signal increased by 1.04 ± 0.03 (Figs. 3Diii, S11Diii, S12Diii). Further, the percentage of cells signaling above the ΔphrA threshold (90th percentile of cells with ΔphrA signal) increased from 33 ± 7% to 55 ± 4% (n = 3) (*, p < 0.01) (Fig. 3Dii) from the untreated to the treated condition. Together, these data demonstrate that exogenously added PhrA peptide induces PphrA activity. We next analyzed the data to display all the single cells for each experiment, independent of the microdroplet, in which the cell is encapsulated (Figs. 3E-F, S11E-F, S12E, F). In agreement with the analysis of average signal per microdroplet, the combined data from the three experiments shows that PphrA activity is higher in the presence of exogenous peptide, and that this difference becomes statistically significant after 2 h and increases up to 4 h (Figs. 3Ei, S11Ei, S12Ei). To visualize signaling intensity and heterogeneity, we plotted them in a histogram using a log scale (Figs. 3Fi, S11Fi, S12Fi, and S15). Relative to WT, the curve for WT + peptide is shifted to the right towards higher intensity, such that 35 ± 15% of peptide-treated cells are above the averaged 90th percentile for untreated single cells, demonstrating that the PhrA peptide promotes PphrA activity. Further, the WT and WT + peptide curves have the same width/mean ratio, showing that the increase in signaling did not decrease the heterogeneity of the response. As expected for the control, the mNeonGreen signal has a similar intensity range in the presence or absence of peptide (Figs. 3Fii, S11Fii, S12Fii), suggesting that the addition of peptide did not influence the activity of the constitutive promoter. Finally, we analyzed the data from the perspective of single cells within the same microdroplet (Fig. 3G, S11G, S12G). We find that 31 ± 17% (n = 3) of microdroplets expressing PphrA are statistically higher than most untreated cells in microdroplets, whereas no microdroplets containing treated cells expressing constitutive mNeonGreen were higher than most untreated microdroplets. Overall, we conclude that treatment with PhrA does increase PphrA activity, but it does not change the heterogeneity of the signal. These data suggest that, in these conditions, the heterogeneity in the TprA/PhrA signaling may be an intrinsic property of the signaling network. PhrA signaling displays a preference toward self-signaling The TprA-PhrA circuit can be defined as a ‘secrete-and-sense circuit’45, in which cells can communicate with self (autocrine signaling), with neighboring cells (paracrine signaling), or with both, since the same cell encodes the signal and the receptor. In the Tpr/Phr systems, the peptide is absorbing; that is, the active signal is sequestered by the sensing cell, such that each molecule activates only a single target cell and generally mediates short-range signaling. This contrasts with non-absorbing signals, which can act on multiple cells through reversible interactions with surface receptors and therefore support longer-range communication6. Our microdroplet setup captures cells at the early stages of TprA/PhrA signaling. Importantly, in vitro experiments comparing WT to a deletion mutant of the importer (ΔamiA-F), demonstrate that PhrA signaling is negligible in the absence of the peptide importer, suggesting that in WT cells, PhrA is active only after export45. To test whether PhrA-mediated signaling tends towards self-signaling (import into producing cells) or signaling to neighboring cells, we mixed WT and ΔphrA cells and measured the extent of signaling in each population. We reasoned that if most of the secreted signal is evenly distributed in the population, over time the signal in the phrA deletion mutants should approach that of the WT, given that the ΔphrA cells respond to exogenous peptide in the same manner as the WT cells (Fig. S5A) and the distances between WT-WT and WT-ΔphrA cells are indistinguishable (Fig. S5B). In contrast, if the signal is preferentially taken up by the producing cells, then most of the signal should be concentrated in the WT population. We loaded microdroplets with seed culture adjusted to a 1:1 ratio of WT:ΔphrA cells. Both strains encoded the PphrA-mCardinal reporter for quantification of TprA/PhrA signaling, and to distinguish the populations, only the ΔphrA cells expressed mNeonGreen. At t = 0 the WT:ΔphrA ratio was 0.61, indicating that most microdroplets contained an excess of ΔphrA cells. As a control for baseline signaling in the ΔphrA cells, we seeded monocultures of ΔphrA cells in microdroplets. At 4 h post-seeding, we quantified the total signal associated with WT cells and ΔphrA cells (Fig. 4). In mixed populations, most signal was associated with the WT cells (Fig. 4Ai, Aii). Specifically, WT median signal was 5× and 3.6× baseline (n = 2), whereas ΔphrA signal was 1.07× and 1.16× baseline (ΔphrA monoculture). Thus, ΔphrA cells exhibited only a slight increase in signal when grown in the presence of PhrA-producing cells versus alone, suggesting limited intercellular signaling. As expected for the control, the mNeonGreen signal was unchanged for ΔphrA, whether grown alone or with WT (Fig. 4Bi, Bii). Thus, we conclude that in our model of early stages of infection, PhrA signaling is preferentially autocrine, with minimal sharing between neighboring cells. Pneumococcal cells tend to autocrine (PhrA) signal at 4 h. A, B Violin plots show ΔphrA in monoculture (green, n = 5590 and 5356 cells), ΔphrA in coculture with the WT strain (yellow, n = 2414 and 2155 cells), and the WT in co-culture with the ΔphrA strain (purple, n = 1516 and 1265 cells) (data from two experiments). Panels represent PphrA-driven (A) and constitutive-driven signal (B). The ratio of ΔphrA and WT cells at time 0 in the co-cultures was 1:1 in the first experiment (4 ± 2 to 4 ± 2) and 1.25:1 in the second experiment (5 ± 3 to 4 ± 3) P-values are displayed (CI = 99%). Discussion We employed microdroplets to measure CCC in S. pneumoniae, focusing on the widely distributed PhrA-mediated signaling system. We developed a protocol to encapsulate and cultivate S. pneumoniae in microdroplets, as well as quantify the total number of cells and the number of PhrA-signaling cells. As observed previously in batch culture, PhrA signaling is activated by the addition of exogenous PhrA peptides. Notably, signaling is heterogeneous both within individual microdroplets and across different microdroplets, regardless of whether exogenous peptide is added. Further, when PhrA-producing cells are mixed with cells that can sense but not produce PhrA, most signaling is not shared across the two populations, consistent with preferential signaling back to producing cells rather than use of the peptide as a common good. Overall, our study provides a platform to capture single-cell signaling intensity and measure the extent of signal sharing within a population. This platform can be applied to a wide range of signaling systems and species. This work shows the inherent heterogeneity of individual cells that is overlooked when measuring averages. Although we do not investigate the molecular basis of this heterogeneity (epigenetic, stochastic, and other sources), previous studies suggest that a major contributor to bacterial heterogeneity is translation bursting, often associated with a small number of expressing cells, during which cells cycle between higher and lower expression states46. In this manner, the observed heterogeneity could be explained by PhrA-signaling undergoing stochastic translation bursts. Further, the TprA/PhrA CCC network includes a positive feedback loop because PhrA peptide is autoinducing. This positive feedback may further amplify variability in intercellular signal47. Our data suggest that, under the conditions tested, the PhrA signal is taken up primarily by the producing cells (Fig. 4). These data are consistent with a study using the PlcR/PapR system in Bacillus6 (TprA and PlcR are part of the same family of receptors, yet there is no similarity between the functional portions of their cognate peptides PapR (ADLPFEF) and PhrA (LDVGKAD)). This study found that PapR communicates over a distance of 13 μm within a densely packed cell population, consistent with short-range CCC signaling. In our system, the average distance between WT cells exceeds 20 μm (Fig. S5B). Importantly, these experiments examined a tightly packed biofilm with constant cell density, whereas our model involves dispersed cells within microdroplets. Thus, although one might expect equal sharing of PhrA between WT and ΔphrA cells, our findings indicate that autocrine signaling predominates in this microculture environment. Autocrine signaling occurs when a cell produces signaling molecules that bind to receptors on its own surface, triggering changes in the producing cell. Our work in microdroplets captures the initial stages of TprA/PhrA signaling and primarily measures signaling to the producing cell population. We propose that the early preference for autocrine signaling at this stage serves to efficiently trigger autoinduction. Recent studies, including mathematical models, suggest that signaling exclusively to self (autocrine) or exclusively to neighbors (paracrine) corresponds to the opposite ends of a cell-cell signaling spectrum, and that biological systems with secreted molecules span this spectrum. In this framework, systems that are exclusively autocrine may have evolved from modulating an ancestral circuit that originally mediated cell-to-cell signaling. The position of a molecular circuit in this signaling spectrum is shaped by molecular tuning factors: the expression of the receptor, the rate of peptide secretion, diffusion, the degradation of the signal outside the cell, and the extent of positive feedback. Very recent studies in our lab have revealed that PhrA secretion levels are tightly regulated by an ABC transporter/protease pair, PptAB/Eep. This inhibition was measured in vitro and at early time points of pneumococcal seeding on epithelial cells, but inhibition was released 2 hours after pneumococci were seeded onto host cells42. When integrated with our single-cell data, these data lead us to propose that PptAB-mediated modulation of PhrA may serve to fine-tune the circuit, such that release of this inhibition may drive higher levels of PhrA and promote an increase in communication between neighboring cells. These conditions are not observed in our microdroplet system, where the signal is mostly captured by producing cells. Thus, we propose that the network architecture is adaptable, that is, by modulating the concentration of the extracellular signal, the bacteria fine-tune the extent of signaling to self vs signaling to neighboring cells. A major consequence of the heterogeneity is to increase the diversity in a regulatory signal, where the population includes cells with low and higher expression of a given target. This diversity serves as a pool for fitness differences, such that heterogeneity and factors that increase heterogeneity may represent evolutionary strategies for S. pneumoniae during colonization. Multiple explanations have been proposed for heterogeneity14. The most common are bet-hedging and division of labor; further group-level phenotypes and variation in cellular vigor are also possible explanations48. Heterogeneity has been observed with bacterial CCC, where subpopulations commit to quorum sensing at different times. For example, in Pseudomonas quorum sensing systems, one fraction of cells activates quorum sensing immediately, while another delays activation until a higher cell density is reached49. This difference in timing may allow the population to turn off quorum sensing if conditions change. It is possible that a related phenomenon occurs in Streptococcus, where variability in signal response at intermediate densities could represent a bet-hedging strategy. By staggering activation, the population may balance the benefits of cooperation with the flexibility to adapt to stresses and environmental fluctuations. Alternatively, the observed heterogeneity may arise from other factors, such as differences in cellular vigor48. The molecular mechanism driving these response changes may include variation in the basal expression levels of key pathway components, differences in degradation or binding of the peptide, or variability in stimulation cues such as local nutrient availability. Further studies are required to understand the physiological significance of variation in early TprA/PhrA signaling. Microdroplet arrays offer a powerful platform for dissecting bacterial cell-cell communication at single-cell resolution. Our system confines bacterial populations within uniformly sized microdroplets arranged in a monolayer, enabling real-time imaging of gene expression dynamics. Unlike bulk assays, which report population averages, microdroplets allow quantification of fluorescent reporter activity from individual cells, revealing heterogeneity within the population. In this study, for example, we observed variability in the magnitude of PhrA-mediated signaling even when the activating peptide was supplied exogenously. Beyond characterizing single-cell variability, this platform provides a foundation for diverse experimental designs. The setup can be adapted to track the same microdroplets over time, which would enable longitudinal studies of signaling dynamics. In addition, the system can be adjusted to introduce gradients or molecules, allowing quantification of responses to variable levels of nutrients, metals, peptides, or other compounds. For example, a logical next step would be to examine how independent and combined changes in carbohydrates and pH influence signaling. The approach could also incorporate multiple strains or species and measure signaling across the population. Ultimately, integrating microdroplet technology with cell-sorting methods may enable enrichment of specific subsets for downstream phenotypic analysis, including virulence. Altogether, this technology has the potential to reveal how single-cell features and their responses to environmental cues drive bacterial communication. Materials and Methods Bacterial strains and culturing The strains in this study were derived from Streptococcus pneumoniae D3950. The reporter strains and the phrA deletion were constructed as described below. Bacterial cultures were grown from frozen stocks conditioned to chemically defined media supplemented with 55 mM glucose pH 6.5 (CDM-Glu)11. To prepare these stocks, strains were streaked from frozen glycerol stocks onto Trypticase™ Soy Agar (TSA-II) supplemented with 5% sheep blood (BD, BBL, New Jersey, USA) and incubated overnight at 37°C and 5% CO2. Colonies were inoculated into 10 mL of CDM-Glu and incubated (without shaking) at 37°C at 5% CO2 until OD600 reached 1 ± 0.2. CDM-Glu used to prepare frozen pre-cultures was used within 2 days of preparation51. Aliquots of 400 µL cell culture and 200 µL bacterial freezing media were transferred to sterile 1.7 mL microcentrifuge tubes. These pre-cultures were frozen at -80°C and used within 30 days of preparation. All strains used in this study are listed in Table S1. Seeding cell culture for microdroplet generation Cells in microdroplets were seeded from cultures treated as follows. Frozen pre-cultures were thawed at room temperature, centrifuged at 8000 rpm to pellet cells and remove supernatant, resuspended in 10 mL CDM-Glu with 300 U/mL catalase (MP BiomedicalsTM, CA, USA) in a 50 mL tube, and adjusted to an initial OD600 of 0.025. Seed cultures in 50 mL tubes were incubated at 37 °C without shaking until the OD600 reached 0.055 ± 0.005. This procedure yields over 97% viable cells (based on >30,000 cells counted by FACS), where non-viable cells are identified by the cell-impermeable dye SYTOX Green. When grown in microdroplets, cells exhibited an estimated doubling time of 1.66 ± 0.13 h during a 4-hour observation period (Fig. 1C). Bacterial strain generation The PphrA and constitutive reporter and the ΔphrA strains were created using homologous recombination. The reporters were generated from Gibson assemblies that were transformed into the D39 strain. Table S2 lists all primers used to construct the reporters and ΔphrA strains. The PphrA reporter was generated using the phrA promoter and tetracycline marker (amplified from an Escherichia coli plasmid Ppp3), and hlpA-mCardinal (amplified from a synthetic gene fragment)43,52. The PphrA-hlpA-mCardinal reporter was integrated into the genome, replacing bgaA. The constitutive reporter was provided by the Veening lab and can also be constructed using an E. coli strain containing plasmid EPY Plac-linker-mNeonGreen (#122640, Addgene, MA, USA). The Pami-hlpA-mNeonGreen construct was inserted downstream of the bgaA gene52,53. The phrA deletion mutant (ΔphrA) was generated by replacing the phrA gene with a spectinomycin cassette. Transformations were conducted by growing the strains in Columbia broth to an OD600 of 0.05 and mixing with the transforming DNA and 125 ng CSP-1 peptide (EMRLSKFFRDFILQRKK, GenScript, USA). After a 2-hour incubation period, cultures were plated onto Columbia agar with the relevant antibiotic (final concentrations: spectinomycin 100 μg/mL, tetracycline 1 μg/mL, or kanamycin 150 μg/mL, Table S2). The EVOS™ Imaging System (M5000, Thermo Fisher Scientific, MA, USA) was used to select for colonies that were expressing mNeonGreen and mCardinal. Finally, the PphrA reporter was validated by testing for induction in CDM supplemented with 55 mM galactose pH 6.5, and synthetic C-terminal PhrA (C10 sequence: SNGLDVGKAD). Stocks were prepared by growing resistant colonies in 3 mL of Columbia broth supplemented with antibiotics. All transformants were verified by Sanger sequencing of PCR amplicons of the transformed regions. Plating CFUs from microdroplets The microdroplet suspension (~12 μL emulsion and ~40 μL carrier oil) was withdrawn using polyethylene (PE) tubing (#BB31695-PE/6, Scientific Commodities, Inc., AZ, USA) connected to a 1 mL plastic syringe (BD, NJ, USA) using a 20 gauge precision tip (Nordson, Ohio, USA) that was cut and sanded with a rotary tool, cleaned with a wire bristle brush and was transferred to a clean PCR tube. Microdroplets in the PCR tube (~106 per mL) were destabilized by adding 10 μL 1H,1H,2H,2H-Perfluorooctanol (Alfa Aesar, MA, USA) and waiting for ten minutes. The cell suspension located on top of the oil in the amount of 7.5 μL was diluted with 1x PBS, spot plated onto TSA II agar supplemented with 5% sheep blood, and incubated at 37 °C overnight. Colonies were counted the next morning. Microdroplet generation Microdroplets were generated using cell suspension (OD600 = 0.055) and saturated carrier oil using a microfluidic device with a planar flow focusing geometry54. The orifice of the device is 50 μm wide and the depth of the device is 200 μm. We set the initial flow rates for carrier oil and cell culture to 10 μl/min to generate microdroplets with an average diameter of 55 μm, as this size enables a close packed geometry within the hexagonal chambers of the microfluidic device used for imaging cells (top of Fig. 1B, S2). The carrier oil is composed of 2% 008 fluorosurfactant (RAN Biotechnologies, MA, USA) and saturated 3 M™ Novec 7500 HFE fluid. Water-saturated oil was prepared as previously described55. Carrier oil was sterile filtered using a 0.22 μM nylon syringe filter (Simsii, Inc., WA, USA). Uniformity of microdroplets within and across experiments was established by generating microdroplets for 10 minutes preceding collection to establish equilibrium within the device. The standard for assessing uniformity in a set of microdroplets is the polydispersity index (PDI) (Eq. 1, Fig. S1, where σ and X correspond to the standard deviation and mean, respectively. To ensure the microdroplets have the desired diameter, the microdroplets were measured in real time. Microdroplets were deposited into a C-chip (SKC, Inc, GA, USA), a disposable hemocytometer-like counting slide that was prefilled with 7.5 μL carrier oil after the microfluidic device reached equilibrium. After microdroplets were deposited, all openings were sealed with epoxy to prevent them from evaporating. The C-chip was mounted onto the stage of an EVOS™ Imaging System and measured using a 20x air objective. If necessary, the flow rates were adjusted to achieve comparable microdroplet sizes at 55 μm across conditions and within experiments and returned to the EVOS™ to be measured again. Once microdroplets reached the desired size and uniformity, no more than 200 μL of emulsion was collected into a clean 1.5 mL microcentrifuge tube. After collection, microdroplets were withdrawn into a 1 mL plastic BD™ syringe prefilled with 100 μL carrier oil, using PE tubing. Microdroplets were incubated at 34°C, to mimic the temperature of the human upper respiratory tract. At each respective time point, these microdroplets were loaded into a separate microfluidic device (top of Fig. 1B) for imaging and analysis. Fluid flow for microdroplet generation (Fig. 1. 2) was driven by two 1 mL syringes mounted onto two separate syringe pumps (Pico Plus Elite Dual, Harvard Apparatus, MA, USA) that were each injected into a flow focusing device with one inlet for cell culture. Cells in Fig. 3 were seeded with peptide immediately upon encapsulation by employing three 1 mL syringes that each contain: a 2x cell concentration and 2 μM C10 peptide mounted onto a single syringe pump and carrier oil mounted onto a second syringe pump. All fluid was driven into a flow focusing device with two inlets for the aqueous phase, where the net aqueous flow rate was 10 μl/min and the flow rates for cells and peptide were each 5 μl/min. The final concentrations of C10 peptide and cells in the microdroplets were 1 μM and 1x, respectively. In Fig. 4, the cell ratio was adjusted to 1:1 for the ΔphrA and the WT strains. Design of microfluidic microdroplet storage device for imaging microdroplets arrays We designed a microfluidic device to image microdroplets in a monolayer for high-resolution imaging using a glass coverslip bottom coated with a PDMS membrane (Top of Fig. 1B, S2)55. The shape of each chamber in a microchannel enables microdroplets to form a close-packed geometry. The dimensions of each hexagon are selected to permit ~1200 55 µm microdroplets to form a close-packed geometry within each hexagon. Each microchannel is used to image a separate condition or individual time point. Channels are positioned 3.5 mm apart to prevent photobleaching between neighboring reservoirs during fluorescence imaging. Fabrication of microfluidic device for imaging microdroplets arrays The microfluidic device is composed of top and bottom layers of silicone rubber bonded together, enclosing microchannels that are 45 µm in height. The top layer is composed of polydimethylsiloxane (Sylgard™ 184 Elastomer Kit, Dow, MI, USA) (PDMS). For the bottom layer, a 40 mm diameter glass #1.5 coverslip was cleaned with Windex® rinsed with dH2O, and set on a hot plate lined with lens paper at 120 °C for 20 min located in a laminar flow hood. To coat coverslips with PDMS, they were centered on the mount of a spin coater (Laurell Technologies Corporation, PA, USA) and ~0.3 g 10:1 (base to curing agent) PDMS was applied to the center of each coverslip and spun at 5000 rpm for 5 min to achieve a film thickness of ~12 µm56. To create a wafer for the microfluidic device, SU-8 3050, an epoxy-based photoresist (Kayaku Advanced Materials, Inc, MA, USA) was spun onto a silicon wafer (Silicon Sense, Inc., NH, USA) at 500 rpm for 20 s at a ramp rate of 85 rpm/s on the spread cycle (step 1), and spun at 5500 rpm for 30 s at a ramp rate of 340 rpm/s for the spin cycle (step 2). Steps for pouring and curing the PDMS followed standard procedure57. The top and bottom layers were bonded together using a plasma cleaner (Harrick Plasma, NY, USA) with an exposure time of 56 s. The assembled devices were heated to 180 °C for 1 h to increase hydrophobicity of the microchannels and soaked in saturated 3 M™ Novec 7500 HFE fluid that was sterilized with a 0.22 µm nylon filter. Operating the microfluidic device for imaging microdroplets The microfluidic device was used to keep microdroplets stationary during 3D imaging. The microfluidic device was sealed into a Sykes-Moore Chamber (Bellco Glass, NJ, USA) using two Teflon gaskets between the bottom ring and the coverslip, and one O-ring gasket between the top of the coverslip and the upper chamber and was tightened with a customized wrench. Approximately 20 μL microdroplets were dispensed from the 1 mL syringe to increase the proportion of monodisperse microdroplets entering the chamber. Then, carrier oil was injected into the microchannel followed by the microdroplets of the desired time point, condition, and/or strain. To prevent microdroplets from evaporating, the inlets were covered with additional carrier oil, and the top layer of PDMS was submerged in 1x PBS. Water-saturated HFE oil was used to generate microdroplets to avoid microdroplet shrinkage downstream. The Sykes-Moore chamber was held on lens paper for two minutes and if no leaking occurred, it was submitted for imaging. Confocal microscopy of cells in microdroplets A Carl Zeiss LSM-880 microscope was used to image all cells in microdroplets. Smart Setup (ZEN, Carl Zeiss Microscopy, NY, USA) was used to adjust laser settings. Laser lines 633 nm and 488 nm were used simultaneously for mCardinal and mNeonGreen, respectively. Averaging was set to 1 for a 16-bit image, pixel dwell time was 1.52 μsec, scan time was 930.01 msec, and the image size was set to 212.5 x 212.5 μm. The micron per pixel ratio was 0.415 μm/pixel. The depth of the z slice was 1 micron and the stack began and ended after cells were out of focus. At least 30 stationary microdroplets were imaged per time point. We used a 40x objective (EC Plan-Neofluar 40x, Carl Zeiss Microscopy, NY, USA) with water immersion. Processing confocal micrographs in Fiji Z-stacks were reconstructed in Fiji58 to enable IMARIS (v9.9.0, Oxford Instruments, Oxfordshire, UK) to recognize microdroplets individually and assign cells to their respective microdroplet. To enable recognition of microdroplets, the slice of the transmitted z-stack corresponding to the equator of the microdroplet was selected, and the transmitted image was converted to a binary image. The oval tool was used to fill each microdroplet with white while leaving room for boundaries between neighboring microdroplets. The boundaries were reinforced by applying Watershed or placing a black line between microdroplets. A binary image was created so that microdroplets appeared red; the image that resulted contained white microdroplets and a black background. The image was converted to a 16-bit image, duplicated to the total number of slices in the original transmitted z-stack, and a stack was created from each binary image. The binary stack was merged with the stacks of fluorescent channels and was saved as a TIFF file and uploaded to IMARIS. Automating measurements of microdroplets To automate the process of measuring the area of microdroplets to calculate the diameter, we assumed the microdroplets were shaped as spheres inside the C-chip because the height of the chamber was 100 μm. Brightfield images were taken using a 20x air objective on the EVOS™ Imaging System. The focus was set on the thin dark boundary or equator of each microdroplet to increase contrast to ameliorate the process of thresholding in Fiji. The area of microdroplets was measured after uploading images into Fiji software (n > 1000 microdroplets). The micron-to-pixel ratio was adjusted to 0.309 μm/pixel (Analyze > Set Scale). A threshold was applied to each image to highlight the boundaries of microdroplets. The result is an 8-bit image containing white microdroplets surrounded by black boundaries. To compute the area (Analyze > Analyze particles), the settings were adjusted as follows: minimum size: 5000-pixel units, circularity: 0.50–1.00, and edges were excluded to avoid artifacts and incomplete microdroplets. This feature provides a list of areas that were used to calculate the diameter of >1000 microdroplets. Cell enumeration and distance from microdroplets To enumerate cells to fit a Poisson distribution as in Fig. S3D, microdroplets were injected into a microfluidic device to image cells and were imaged on an EVOS™ Imaging System using a 40x multi-immersion objective in water (40x/0.90 NA Multi-Immersion Ph3, Carl Zeiss Microscopy, NY, USA). Z-stacks were collected for >100 complete microdroplets. Images were uploaded into Fiji and the cell number was counted manually, accounting for each slice in the stack. To determine the number of reporter cells (WT) in Fig. 4, transmitted images were used to count the total number of cells, and the number of mNeonGreen (ΔphrA) cells was subtracted from each microdroplet. To measure the distance between cells we calculated Euclidean distances between every combinatorial pair of cells within three microdroplets using 3D spatial coordinates of cells extracted from IMARIS software and imported into MATLAB (Fig. S5B). Testing PhrA diffusion across microdroplets PhrA diffusion was assessed by comparing the brightness of mCardinal in ΔphrA cells in microdroplets for three conditions: microdroplets containing ΔphrA cells, a 1:1 mixture of microdroplets containing ΔphrA cells with microdroplets containing WT cells, and ΔphrA cells in microdroplets with 1 µM PhrA peptide. Cells were imaged using confocal microscopy after 4 hours of incubation. Brightness of mCardinal was compared by measuring the brightness-to-surface-area ratio of cells from max-z projections of ten microdroplets per condition. The initial OD600 for each condition was 0.055 ± 0.005. Reconstructing Fiji-derived images in IMARIS to quantify fluorescence IMARIS 9 (v9.9.0, Oxford Instruments, Oxfordshire, UK) was used to extract fluorescent intensities of bacterial cells and enumerate them, while mapping each cell to a select microdroplet. To do this, TIFF images were uploaded to IMARIS and a functional surface was created around each microdroplet and cell using the surface rendering method. The following parameters were used during surface rendering on microdroplets: (i) object-object statistics was enabled to apply surface rendering to all microdroplets and cells collectively, (ii) the source channel was set to the binary image while surface detail was set to 1.00 μm for microdroplets, (iii) absolute intensity was used to threshold microdroplets semi-automatically (see Thresholding below), (iv) split touching objects was enabled while setting the seed points diameter to 40 μm based on intensity, (v) the filter type was set to ‘quality’ and adjusted to recognize all microdroplets, (vi) each microdroplet with a functional surface was duplicated and renamed. Parameters used for surface rendering of pneumococcal cells (except reporter cells in Fig. 4) were as previously described above except: (ii) the source channel was set to mNeonGreen with no smoothing or surface detail, (iv) the seed points diameter was set to 0.5 μm instead of 40 μm and still based on intensity, (v) the filter type was changed from quality to volume and set to 0.0650 μm3 to resemble the volume of a small coccus, (vi) each sub-community of cells were assigned to the prospective microdroplet using the ‘Filter’ feature followed by ‘Shortest Distance to Surfaces’ (inactivate the lower threshold while setting the upper threshold value to zero) and then newly assigned cells were renamed to reflect the microdroplet that they were assigned. For thresholding, the value of the automated threshold was recorded for the strain and/or condition with the highest intensity for PphrA-mCardinal. For consistency, these values were averaged across different images of cells for ~30 microdroplets and applied manually to the corresponding pair (WT & ΔphrA) or conditions (WT vs WT + peptide) at each time point. Based on the automatic threshold values that were generated, averaged threshold values were manually applied at 0 h, 1 h and 2–6 h to account for the increasing PphrA signal over time. This guideline was followed for Figs. 1–3 and the ΔphrA strain in Fig. 4. To easily differentiate between the reporter strain and the ΔphrA strain in Fig. 4, we employed a PphrA-mCardinal reporter cell that does not express mNeonGreen. This approach required a modified protocol in IMARIS to quantify PphrA-mCardinal signal from this reporter strain. The parameters used for surface rendering of the PphrA-mCardinal reporter cell were as previously described above except: (ii) the source channel was set to mCardinal instead of mNeonGreen. Careful consideration was taken when creating consistent 3D functional surfaces for this reporter in each condition. Consequently, 0 to 1 h and 2 h use independently averaged threshold values as described above. For consistency, at each time point, preference was given to visible coverage of the 3D functional surface instead of thresholding at 4 and 6 hours to be commensurate with the size of cocci across conditions while applying surface rendering. Finally, to estimate the extent of mCardinal accumulation over time, we compared signal values at 3 and 4 hours for the constitutive promoter (Pc) and the PphrA promoter (PphrA). The average increase from 3 to 4 hours was 194 ± 810 for Pc versus 17,117 ± 2,832 for PphrA. These results demonstrate that the dramatic increase observed for PphrA-driven mCardinal is not an artifact of signal accumulation. Using quantitative data from IMARIS to construct graphs To create heatmaps, 30 microdroplets were randomly selected using the Randperm function in MATLAB. When z-scores were calculated for the WT and ΔphrA strains, the mean and standard deviation were calculated relative to the ΔphrA for each time point. Similarly, when z-scores were calculated for the WT and WT + peptide conditions, the mean and standard deviation were calculated relative to the WT for each time point. The ratio of the median signal was calculated using the median value at each time point and calculating the ratio of WT to ΔphrA or the ratio of WT + peptide to WT. The percentage of signaling cells was defined as the proportion of cells having a PphrA-mCardinal signal that is above the averaged 90th percentile of the ΔphrA strain for three experiments (referred to as the ΔphrA threshold for signaling), relative to the total number of cells for the given condition and time. Histograms that plot signal intensity versus frequency of single cells were plotted in MATLAB (R2021a, MathWorks). The Freedman-Diaconis Rule was used to accurately calculate the number of bins for each histogram that contains a non-normal dataset. Intensity values from cell-signaling were transformed using log10(x) for easy visualization. Figure 1D was generated from IMARIS. Figure 1E, F were created using PowerPoint. Max z-projections were curated in Fiji. All other graphs were produced using GraphPad Prism 10.2.3. Statistics and Reproducibility A two-tailed t-test was used to detect statistical differences between the net growth rates of the WT and ΔphrA strains using 0 to 4 hours (α = 0.05). A Chi-squared test (CI = 95%) was used to detect any significance of the proportion of single cells above and below the limit of detection in WT and ΔphrA strains and the WT and WT + peptide conditions for the constitutive promoter. The test reported no reproducible statistical differences between strains in Fig. 2 and between conditions in Fig. 3 (Table S3). A Mann-Whitney test was used to measure statistical differences of single cells (CI = 99%) which are represented in violin plots and histograms. Cell averages in microdroplets that are plotted as heatmaps (Table S4) were tested for statistical differences between strains and conditions using the Mann-Whitney test (CI = 99%). Individual single cells assigned to microdroplets were measured for statistical differences using the Kruskal-Wallis test followed by Dunn’s multiple comparisons test (CI = 99%) (tests conducted in GraphPad Prism 10.2.3). Diffusion across microdroplets was tested for statistical differences using a Nested One-way ANOVA. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary Acknowledgements We would like to thank Dr. Karina Mueller-Brown for her guidance in generation of gene constructs and primer design, as well as help in collecting CFU data. Further, we are grateful for Dr. Drew Bridges’ constructive feedback on manuscript preparation, and productive conversations with Dr. Hasan Yesilkaya. We would like to thank Dr. Ajit Ray for his assistance in developing the IMARIS protocols for cell capture and enumeration and Xinyao (Kaitlyn) Li for helping in the development of protocols for this project. We would also like to thank the Veening lab for providing reagents. Figure 1A-1B were curated using BioRender. Author contributions C.J., N.L.H., and S.A. designed experiments, analyzed and interpreted the data, acquired funds, and wrote the manuscript. C.J. performed the experiments, and M.H., A.L., L.E., and R.E. assisted in experiments and interpretation. L.E. and F.L. provided resources, participated in methodology development, and data acquisition and interpretation. J.G. contributed to data analysis. All authors edited and approved the manuscript. Peer review Peer review information Communications Biology thanks Connie Chang and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Primary Handling Editor: Tobias Goris. A peer review file is available. Funding N.L.H. discloses support for the research of this work from the National Institute of Allergy and Infectious Diseases, NIH (R01AI139077 and R01AI139077-S1). N.L.H., S.L.A., and C.J.B. disclose support for the publication of this work from the Carnegie Institute of Technology, Carnegie Mellon University through the Catalyst Award. N.L.H. and C.J.B. also disclose support for the publication of this work from the Mellon College of Science, Carnegie Mellon University through the Margaret Carver Research Enrichment Award. C.J.B. also discloses support from the Burroughs Wellcome Fund (G-1021662) and the Ford Foundation Postdoctoral Fellowship. Data availability The source data that support the findings of this study are available in Supplementary Datasets 1-3 and other relevant data for the findings are included in the Supplementary Figs. and Tables. All other data are available from the corresponding author upon reasonable request. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Contributor Information Shelley L. Anna, Email: sanna@andrew.cmu.edu N. Luisa Hiller, Email: lhiller@andrew.cmu.edu. Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s42003-026-10451-1. References Associated Data Supplementary Materials Transparent Peer Review file Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Reporting Summary Data Availability Statement The source data that support the findings of this study are available in Supplementary Datasets 1-3 and other relevant data for the findings are included in the Supplementary Figs. and Tables. All other data are available from the corresponding author upon reasonable request.