Pseudomonas syringae subpopulations cooperate by coordinating flagellar and type III secretion spatiotemporal dynamics to facilitate plant infection

Nature Microbiology | Volume 10 | April 2025 | 958–972 958 nature microbiology Article https://doi.org/10.1038/s41564-025-01966-0 Pseudomonas syringae subpopulations cooperate by coordinating flagellar and type III secretion spatiotemporal dynamics to facilitate plant infection Nieves López-Pagán1,7, José S. Rufián 1, Julien Luneau2,3, María-Antonia Sánchez-Romero4, Laurent Aussel 5, Simon van Vliet 2,6, Javier Ruiz-Albert 1 & Carmen R. Beuzón 1 Isogenic bacterial populations can display probabilistic cell-to-cell variation in response to challenges. This phenotypic heterogeneity can affect virulence in animals, but its impact on plant pathogens is unknown. Previously, we showed that expression of the type III secretion system (T3SS) of the plant pathogen Pseudomonas syringae displays phenotypic variation in planta. Here we use flow cytometry and microscopy to investigate single-cell flagellar expression in relation to T3SS expression, showing that both systems undergo phenotypic heterogeneity in vitro in apoplast-mimicking medium and within apoplastic microcolonies throughout colonization of Phaseolus vulgaris. Stochastic, spatial and time factors shape the dynamics of a phenotypically diverse pathogen population that displays division of labour during colonization: effectors produced by T3SS-expressing bacteria act as ‘common goods’ to suppress immunity, allowing motile flagella-expressing bacteria to increase and leave infected tissue before necrosis. These results showcase the mechanisms of bacterial specialization during plant colonization in an environmentally and agriculturally relevant system. Flagellar motility is an important trait for environmental adaptation. Members of the genus Pseudomonas can colonize soil, plant or animal tissues. The Pseudomonas syringae species complex includes most of those pathogenic to plants, collectively causing diseases in over 500 species, many agriculturally relevant1. P. syringae can live on the leaf surface, but most favour endophytic colonization, using wounds and stomata to enter the leaf. Flagellar motility confers advantages to epiphytic populations of P. syringae pv. syringae2 and facilitates leaf entry for many P. syringae pathovars3,4. Inside the leaf, P. syringae colonizes the intercellular spaces (apoplast). Once inside, flagellar motility is not required for systemic spread3, while flagellin, the main component of the pilus, triggers immunity 5 . Fittingly, flagellar expression is downregulated in the apoplast6–8. Flagellar biosynthesis is depicted as a deterministic programme with a complex and tiered regulatory hierarchy, where promoters are sequentially activated, maintaining this state through active growth. Received: 19 March 2024 Accepted: 19 February 2025 Published online: 2 April 2025 Check for updates 1Instituto de Hortofruticultura Subtropical y Mediterránea “La Mayora”, Universidad de Málaga-Consejo Superior de Investigaciones Científicas (IHSM-UMA-CSIC), Málaga, Spain. 2Department of Computational Biology, University of Lausanne, Lausanne, Switzerland. 3Department of Fundamental Microbiology, University of Lausanne, Lausanne, Switzerland. 4Departamento de Microbiología y Parasitología, Facultad de Farmacia, Universidad de Sevilla, Sevilla, Spain. 5Aix Marseille Univ, CNRS, LCB UMR7283, IMM, Marseille, France. 6Biozentrum, University of Basel, Basel, Switzerland. 7Present address: Department of Chemical Biology and Genetics, Centre of the Region Haná for Biotechnological and Agricultural Research, Faculty of Science, Palacký University, Olomouc, Czech Republic. e-mail: [email protected] Nature Microbiology | Volume 10 | April 2025 | 958–972 959 Article https://doi.org/10.1038/s41564-025-01966-0 HIM medium, most bacteria express fliC::GFP3 (Flagella ON ) with considerable cell-to-cell variation, but a subpopulation does not (Flagella OFF ) and can be visualized only by membrane staining, overlapping with non-fluorescent bacteria in FC analyses (Fig. 1a). Expression of GFP3 from a constitutive promoter is homogeneous22, ruling out expression noise as the fliC::GFP3 source of heterogeneity. In keeping with flagellar role in facilitating leaf entry2–4, leaf-surface bacteria are mostly Flagella ON . Nonetheless, images of bean (Phaseolus vulgaris) leafs sprayed with fliC::GFP3 bacteria constitutively expressing eCFP show that flagellar expression is also heterogeneous on leaf surfaces (Extended Data Fig. 2). Four days post pressure inoculation (dpi), bacteria recovered from the apoplast display lower average expression than media-grown populations (Fig. 1b), agreeing with transcriptomic data 6–8 . However, most apoplast-extracted bacteria are Flagella ON (Fig. 1a). Cell-to-cell differences in fluorescence were more pronounced among apoplast-extracted bacteria than in media, with a wider distribution revealed by FC (Fig. 1a (right),c). This variability is also apparent in planta, with apoplastic microcolonies of fliC::GFP3 eCFP bacteria showing heterogeneous GFP3 fluorescence overlaid with homogeneous eCFP fluorescence distribution (Fig. 1d). Homogeneous eCFP expression was reported previously23. LB populations were sorted by fluorescence-activated cell sorting based on GFP fluorescence (Fig. 1e), enriching populations with low or high fliC::GFP3 expression (Fig. 1e right) to use in swimming assays. Since sorting causes physical flagellar damage, it can reduce motility differences between sorted subpopulation pairs and increase variation between sorting events. Thus, we compared motility between high/low subpopulations separated in the same sorting event (Fig. 1f): we calculated the ratio between swimming halo diameters of high GFP3and low GFP3-expressing subpopulations (per sorted event). High-to-low motility ratios for sorted LB populations were significantly higher than 1.0, indicating that fliC expression positively correlates with motility (Fig. 1f). Similar results were obtained using HIM populations (Fig. 1g,h). These results indicate that flagella are phenotypically heterogeneous in P. syringae and show that, despite flagellin being an immune elicitor, part of the apoplastic population expresses flagella. Flagella and T3SS show counter-regulation at the cellular level Next, we used a strain carrying transcriptional fusions fliC::tdT (to tdTomato; AAV52169 1) and hrpL::GFP3 (WP_004656240.1) to evaluate T3SS and flagella expression simultaneously. Population profiles for fliC::tdT are similar to those of fliC::GFP3 (Extended Data Fig. 3a,b). Potential interferences between fluorophores (that is, bleedthrough leading to false positives or Förster resonance energy transfer (FRET) leading to false negatives) were ruled out (Extended Data Fig. 3c,d). Both systems were expressed heterogeneously in HIM (Fig. 2a,b (left) and Extended Data Fig. 3e). FC showed two subpopulations, formed by bacteria favouring expression of T3SS (T3SSON/FlagellaOFF) or fliC However, single-cell studies revealed stochastic activation pulses and bimodal patterns in the flagellar system of Salmonella enterica, with clonal populations displaying phenotypic heterogeneity 9–13 . In P. syringae, regulation of flagellar synthesis remains mostly uncharacterized, but gene arrangement and promoter conservation with P. putida supports a similar hierarchy of transcriptional regulation 14 , with FleQ at the top15. Apoplastic P. syringae use a type III secretion system (T3SS) to introduce effectors (T3Es) into the host-cell cytosol. T3Es suppress plant immunity including flagellin-triggered responses, allowing bacterial proliferation and disease16. T3SS gene expression requires HrpL17. We showed that a ΔhrpL mutant displays increased swimming motility in apoplast-mimicking Hrp-inducing medium (HIM), while a mutant lacking HrpV, a repressor of hrpL expression18, displays reduced motility19, suggesting a regulatory link between T3SS and flagellar expression. Later, we established that P. syringae T3SS genes display phenotypic heterogeneity during plant colonization and bistable expression with reversible ON and OFF states in HIM20. Here we show that flagellar gene expression displays phenotypic heterogeneity in P. syringae, particularly within the plant apoplast. While average expression levels of flagellin-encoding fliC is downregulated in the apoplastic population, as previously described6–8, a subpopulation expresses high fliC levels during colonization. Analysing flagellar and T3SS expression simultaneously, we found all possible single-cell ON/OFF combinations, revealing the phenotypic complexity arising amid clonal populations. Nonetheless, T3SSON/FlagellaOFF and T3SSOFF/FlagellaON subpopulations were generally more abundant, suggesting counter-regulation between these loci. Expression of these systems impacts bacterial growth at the population and/or cellular level. Analysis of the spatial distribution of expression within apoplastic microcolonies shows that T3SS ON bacteria are more abundant early in the infection and close to the host-cell surface, while FlagellaON bacteria become more frequent later and further away from the host cell. Following bacterial apoplastic multiplication, FlagellaON bacteria actively exit the infected tissue before the onset of necrotic symptoms. Our results support a division of labour model with T3SS ON secreted effectors acting as common goods, allowing FlagellaON bacteria to increase and exit before tissue collapse. Results Flagellar expression in P. syringae is heterogeneous We generated a transcriptional fusion to GFP3 (AAB18957.1) of flagellin-encoding fliC (WP_002554301.1) within its genome location, preserving context and function without affecting fitness or virulence (Extended Data Fig. 1) in model bean pathogen P. syringae pv. phaseolicola 1448A21. Using confocal laser scanning microscopy (CLSM) and flow cytometry (FC), we monitored fliC::GFP3 expression in vitro (Fig. 1a). In rich lysogeny broth (LB) medium and apoplast-mimicking Fig. 1 | Flagella display phenotypic heterogeneity in Pseudomonas syringae. a, Left: CLSM images and FC analysis of bacteria carrying fliC::GFP3 grown in LB or HIM medium, or apoplast extracted from leaf apoplasts at 4 dpi. FM4-64 channel (greyscale) display fluorescence from membrane staining. Scale bars, 2 µm. Right: FC data obtained from fliC::GFP3 bacteria are represented using arbitrary units in logarithmic scale. Non-GFP graphs show autofluorescence levels of wild type as reference for OFF subpopulations (vertical lines leave 99% non-GFP data to the left). Microscopy and FC panels show typical results (n = 14 independent experiments). b, Mean GFP fluorescence intensity of fliC::GFP3-carrying bacterial populations (n = 14) from FC data described in a. c, Robust coefficient of variation (RCV) obtained from FC data described in a. b and c display individual sample values, and sample sets displaying different letters are significantly different as determined using one-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test (P < 0.05); all comparisons showed P < 0.0001. Lines indicate the means of all individual data. Error bars represent s.e.m. d, CLSM images of fliC::GFP3 bacteria expressing a constitutive eCFP gene growing in microcolonies in planta at 3 dpi. GFP channel shows heterogeneous distribution of fliC expression. CFP channel shows homogeneous distribution of CFP (n = 3). Red in all panels corresponds to chloroplast autofluorescence. Scale bars, 5 µm. In a and d, contrast and brightness were adjusted to improve visualization and kept constant across samples and channels. e,g, Histograms showing GFP fluorescence vs cell count from fliC::GFP3 overnight LB-grown cultures (e) or 24 h HIM-grown cultures (g) before (left) and after (right) sorting. Red lines indicate gating used to separate fliC::GFP3 bacteria according to fluorescence intensity. f,h, Relative ratio between halo diameters in swimming plates seeded with cells displaying high GFP expression, divided by the diameters of halos obtained from seeding cells displaying low GFP expression, resulting from the same sorting event (identified with different shades) from cultures grown in LB (f) or HIM (h) (n = 9 as shown). Lines indicate the means of all individual data. Error bars represent s.e.m. Asterisks indicate significant differences established by an unpaired two-tailed t-test (P < 0.05). P values are indicated. Nature Microbiology | Volume 10 | April 2025 | 958–972 960 Article https://doi.org/10.1038/s41564-025-01966-0 (T3SSOFF/FlagellaON) (Fig. 2a,b (left)). Although these combinations were more abundant, a continuous distribution with bacteria expressing both (T3SS ON /Flagella ON ) or none (T3SS OFF /Flagella OFF ) was observed (Fig. 2a,b (left) and Extended Data Fig. 3e). Apoplast-extracted bacteria (4 dpi) also displayed all phenotypic combinations, although T3SS ON /Flagella OFF bacteria were more abundant (Fig. 2a,b (right)). As reported 20 , apoplastic populations did not display bimodal GFP distribution (Fig. 2a,b and Extended Data Fig. 3f). Analysis of T3SS effector gene hopAB1 (ref. 20)(WP_011282445.1) rendered similar results (Fig. 2c,d and Extended Data Fig. 3e,f). High 5 µm High GFP fluorescence intensity Plant ab LB HIM c d GFP FM4-64 Merged GFP CFP Merged fliC::GFP3 eCFP ab c a b c P = 0.0412 * e GFP fluorescence intensity Count HIM Count Low f g GFP fluorescence intensity Low LB Low Cell size fliC::GFP3 Non-GFP GFP fluorescence Cell size Non-GFP Cell size Non-GFP fliC::GFP3 fliC::GFP3 OFF ONOFF OFF ON ON P = 0.0112 * h High Low High GFP fluorescence intensity 101102103104105 1,200 Mean LB HIM Plant LB HIM Plant Fluorescence intensity mean (A.U.) Robust coefficient of variation High/low ratio High/low ratio 1,000 800 600 400 200 150 100 50 0 0 101102103104105 101102103104105 50 40 30 20 10 0 101102103 6 4 2 0 0 1 2 3 4 High/low LB 101102103 6 4 2 0 101102103 101102103 101102103 101102103 3,0K 400 300 200 100 0 0 0.5 1.0 1.5 300 200 100 0 2,0K 1,0K 0 HIM High/low 5 µm5 µm RCV Nature Microbiology | Volume 10 | April 2025 | 958–972 961 Article https://doi.org/10.1038/s41564-025-01966-0 Expression of T3SS or flagella carries growth penalties Expression of Salmonella pathogenicity island 1 (SPI1) T3SS causes growth delays, which have been proposed to drive phenotypic heterogeneity24,25. We used competitive index (CI) assays (Fig. 3a) to compare growth of a ΔhrpL mutant and wild-type bacteria (Fig. 3b). CIs show that ΔhrpL mutants outgrow wild-type bacteria in HIM (CI = 1.75 ± 0.15), but not in rich medium (CI = 1.07 ± 0.08), where T3SS genes are not expressed. Growth rates of ΔhrpL bacteria are also higher in HIM (Fig. 3c). In planta, wild-type bacteria outgrow a derivative constitutively expressing HrpL (CI = 0.67 ± 0.06) (Fig. 3d). Constitutive expression of HrpL results in overexpression of T3SS genes 20 , supporting that T3SS expression carries a growth penalty. Flagella often carries growth penalties26–28. However, previous CI assays did not show significant growth differences for ΔfliC bacteria in P. syringae29. Fittingly, we did not observe significant growth differences for ΔfliC or ΔfleQ (WP_011169094.1) (lacking flagellar regulator FleQ) mutants based on CI (LB: CI = 1.08 ± 0.22 and 1.00 ± 0.13, respectively; HIM: CI = 1.21 ± 0.22 and 1.31 ± 0.13, respectively), although variation in LB ΔfliC CI was large (Fig. 3b), or growth rates (Extended Data Fig. 4a). However, a ΔhrpL ΔfliC mutant remarkably outgrows wild type, with CI values (LB: CI = 4.195 ± 0.9975: HIM: CI = 25.79 ± 5.156) 4–15 fold those obtained for single ΔhrpL mutant (Fig. 3b). This supports the notion that flagellar expression carries a cost, at least when simultaneously expressed with the T3SS. Constitutive expression of fleQ from pFleQ led to severe growth delays (Extended Data Fig. 4a), although pleiotropic effects could not be ruled out given the severity of the phenotype. In plants treated with flagellin to void differences associated with flagellin-mediated immunity, CI assays revealed growth advantages for ΔfliC mutants (CI = 1.88 ± 0.34) (Fig. 3d). However, we cannot rule out that ΔfliC mutant advantages in flagellin-treated plants may stem from other reasons. Since T3SS and flagellar expression are both highly heterogeneous, population-level assays could be noisy and may lead to underestimating growth penalties. Thus, we analysed these impacts at the single-cell level using HIM-agar pads and time-lapse microscopy. In this setting, hopAB1::GFP3 activation takes place too late in the experiment to allow for confident quantification (Supplementary Videos 1–4). However, activation of fliC::GFP3 takes place earlier (Supplementary Videos 5 and 6, and Fig. 3e). Time-lapse analyses show that fliC expression negatively correlates with bacterial growth (Fig. 3f). Splitting the fliC::GFP3 population into two using the median fluorescence intensity value, we detected a 6.5% growth penalty for the higher-than-median-fluorescence half compared with the lower half (Fig. 3g). Constitutive expression of eGFP does not negatively correlate with growth (Extended Data Fig. 4b–f). Phenotypic specialization during microcolony development Apoplastic FlagellaON/T3SSOFF subpopulations are potentially more vulnerable to plant defences since they trigger but cannot suppress immunity. We used propidium iodide which binds to membrane-compromised (not viable) bacteria to determine any correlation between phenotypic variants and apoplast killing rates 30,31 . No significant differences were observed between dead/live ratios of Flagella or T3SS ON vs OFF subpopulations (Extended Data Fig. 5), supporting that Flagella ON /T3SS OFF cells are protected from plant responses, perhaps by T3SS ON bacteria suppressing defences in trans. We established that T3SS mutant bacteria may proliferate up to wild-type levels when growing within the same microcolony as wild-type bacteria22, while wild-type and T3SS mutant bacteria growing in separate microcolonies develop according to their respective abilities to suppress local immunity. Therefore, trans complementation between T3SS ON and Flagella ON /T3SS OFF bacteria would require phenotypic variants to rise during growth of the microcolony. Fluorescence distribution of fliC:GFP3 in planta shows that for flagella, ON/ OFF variants appear closely located within the microcolony (Fig. 1d). Close examination of fliC::GFP3 distribution (Supplementary Video 7) reveals areas with stronger fluorescence, suggesting a response to local stimuli (or siblings’ inheritance of an expression pattern), but also isolated bacteria displaying GFP fluorescence strikingly different from those of nearby peers, in keeping with stochastic differences. We followed expression of T3SS and flagella genes during the development of apoplastic microcolonies, analysing leaves infiltrated with hopAB1::GFP3 fliC::tdT or hrpL::GFP3 fliC::tdT bacteria at different timepoints (Fig. 4a,b). Stochastic T3SS heterogeneity was detected a b hrpL::GFP3 fliC::tdT HIM hrpL::GFP3 fliC::tdT Plant HIM Plant fliC::tdT tdTomato fluorescence GFP fluorescence hrpL::GFP3 GFP fluorescence hrpL::GFP3 fliC::tdT tdTomato fluorescence T3SSOFF/FlagellaON T3SSON/FlagellaOFF T3SSON/FlagellaON T3SSOFF/FlagellaON T3SSON/FlagellaOFF T3SSON/FlagellaON hopAB1::GFP3 fliC::tdT hopAB1::GFP3 fliC::tdT c d Plant HIM Plant fliC::tdT tdTomato fluorescence GFP fluorescence hopAB1::GFP3 GFP fluorescence hopAB1::GFP3 fliC::tdT tdTomato fluorescence HIM 103 102 101 103 102 101103 102 101 103 102 101 103 102 101 103 102 101103 102 101 103 102 101 Fig. 2 | Flagellar and T3SS expression display independent heterogeneous expression at the single-cell level. a, CLSM images of dual-reporter hrpL::GFP3 fliC::tdT strain grown in HIM or extracted from leaf apoplasts at 4 dpi. Scale bars, 2 µm. Contrast and brightness were adjusted to improve visualization and used throughout. b, FC analysis of bacteria described in a. Graphs show GFP fluorescence intensity vs that of tdTomato. Results shown in the figure are representative of 3 or more independent experiments. c, CLSM images of dual-reporter hopAB1::GFP3 fliC::tdT strain grown in HIM or extracted from leaf apoplasts at 4 dpi. Scale bars, 2 µm. Contrast and brightness were adjusted to improve visualization and used throughout. d, FC analysis of bacteria described in c. Results shown in the figure are representative of at least 3 independent experiments. Nature Microbiology | Volume 10 | April 2025 | 958–972 962 Article https://doi.org/10.1038/s41564-025-01966-0 a b c d Wild type Mutant 1:1 mix Incubate o/n 28 °C No Ab Ab Inoculate plant leaf 4 dpi In vitro CI In planta CI CI = c.f.u. mutant/c.f.u. wild type (output) c.f.u. mutant/c.f.u. wild type (input) Competitive index value No difference Mutant increase Mutant decrease 1 0.1 10 Plate dilutions onto g e f 0 h 12 h8 h4 h 2 h fliC::GFP3 N = 1,263 cells 500 250 0 750 1,000 1,250 1,500 0 1,750 Growth rate (µm h−1) 0.1 0.2 0.3 0.4 0.5 R = –0.17 P = 1.03 × 10–9 Fluorescence intensity (a.u.) 2,000 fliC::GFP3 Low GFP High GFP 0.2 0.1 0 0.3 0.4 0.5 **** Growth rate (µm h−1) 0.6 0 h 8 h8 h4 h2 h ∆hrpL vs WT ∆fliC vs WT ∆fleQ vs WT ∆hrpL P = 0.0006 pHrpL vs WT ∆fliC vs WT P < 0.0001 P = 0.0161 P = 0.0054 ∆hrpL∆fliC vs WT P = 0.0086 P = 0.0329 HIM LB HIM LB HIM LB HIM LB Pph 0.1 1 10 100 0.15 0.10 0.05 0 Competitive index value *** ** * Plant Plant (+flg22) 0.1 1 10 Competitive index value Growth rate **** ** * 5 µm 5 µm 5 µm5 µm 5 µm 5 µm 5 µm 5 µm 5 µm 5 µm Nature Microbiology | Volume 10 | April 2025 | 958–972 963 Article https://doi.org/10.1038/s41564-025-01966-0 throughout, while zonal distribution of flagellar and T3SS fluorescence within the microcolonies changed over time. One-dpi microcolonies displayed heterogeneous expression of T3SS, but no flagellar expression, supporting the idea that flagella is turned off in early infection stages6–8. Two-dpi microcolonies displayed orange and yellow tones consistent with expression of both systems. By 3–4 dpi, predominantly red (fliC::tdT) or green (hopAB1::GFP3 or hrpL::GFP3) zones with thoroughly overlaid heterogeneity becomes the norm (Fig. 4a,b). Zones with higher expression of hopAB1::GFP3 generally display lower expression of fliC::tdT and vice versa, although there are also orange and yellow spots indicative of T3SSON/FlagellaON bacteria (Fig. 4a,b). Most remarkably, this zonal distribution shows a consistent spatial pattern: bacteria closest to the lower epidermis (Fig. 4c–e), or to a spongy mesophyll host cell (Fig. 4a (right),f,g), are predominantly green, that is, T3SSON, whereas those further away from the host-cell surface, that is, the innermost or furthermost parts of the microcolonies, are predominantly red, that is, FlagellaON (Fig. 4, and Supplementary Videos 8 and 9). Although the T3SSON zone in the proximal side varies in width, it is consistently wider than a single bacterial layer, supporting the idea that the signal(s) activating T3SS expression include diffusible molecules. FlagellaON bacteria actively exit infected tissues The increase of Flagella ON bacteria in older microcolonies suggests a function at later infection stages. We took leaves at 1 dpi, 2 dpi (asymptomatic), 5 dpi (onset of macroscopic symptoms) and 7 dpi (fully symptomatic) and submerged them into a sterile MgCl 2 solution for 30 min. We used CLSM to evaluate hopAB1::GFP3 and fliC::tdT expression in those bacteria found in the solution. Bacteria exited even intact 1 dpi leaves. Shorter incubations also allowed bacterial recovery but not enough for reliable quantification. We compared T3SS or fliC expression between bacterial populations exiting the leaf and whole-leaf populations obtained by mechanical extraction from replicate leaves without incubation (Fig. 5 and Extended Data Fig. 6). Exiting and total apoplast-extracted populations showed similar hopAB1::GFP3 expression profiles at 1 dpi, 2 or 7 dpi (Fig. 5a,b and Extended Data Fig. 6), while significantly fewer T3SS ON bacteria were detected in the exiting fraction by 5 dpi (Extended Data Fig. 6). In contrast, bacteria exiting 1 dpi leaves were significantly enriched in FlagellaON (Fig. 5a,b), even though total apoplast-extracted bacterial populations showed almost no FlagellaON bacteria (Figs. 4 and 5a, and Extended Data Fig. 6). This supports the notion of Flagella ON bacteria being more efficient in actively exiting the leaf in humid conditions. We confirmed this notion using an experimental setting mimicking rain or heavy dew, by spraying the infected leaves every 5 min for 30 min and collecting the dripping solution for microscopic examination (Fig. 5a bottom panels). The bias towards Flagella ON bacteria in the exiting populations detected at 2 and 5 dpi and progressively decreasing until they are no longer detected at 7 dpi (Fig. 5a,b and Extended Data Fig. 6), when leaf integrity has been lost, suggests that flagellar motility is not only important for bacterial entry into the leaf but also for fast and active exit of apoplast-grown bacteria before necrosis. Discussion Bacterial encounters with environmental changes lead to gene activation often assumed to happen in a non-fluctuating homogeneous manner. However, studies averaging population data lose cell-to-cell variation, and an increasing number of traits have been shown to manifest differently across clonal populations. Such phenotypic variation may reflect microenvironmental differences but may also be caused Fig. 4 | Stochastic and spatially structured time-course distribution of flagellar and T3SS expression within apoplastic microcolonies. a, CLSM Z-stack compilation at different dpi showing apoplastic microcolonies of hopAB1::GFP3 fliC::tdT bacteria. All images examined (n = 5 or more) per biological replicate (n = 2) and independent experiment (n = 2 for 1 and 2 dpi; n = 4 for 3 and 4 dpi). b, Images from a including orthogonal projections. The colored lines within the central images indicate the planes that generate each orthogonal section: green lines indicate the section that is displayed in the green-boxed images (images above per time point), red lines indicate the section that is displayed in the red-boxed images (images on the right per time point). In greenor red-boxed images, the blue line correspondes to the section displayed in the central image. Scale bars, 5 µm. c, Schematic leaf 3D structure with close-up showing location of apoplastic microcolonies. Position and orientation of Z-stack acquisition are indicated; 1–3 represent three planes of a bacterial microcolony growing from the lower epidermis inner surface inwards (1 closest to cell surface and 3 furthest) as shown in d and Supplementary Video 8. Plane 4 cuts through a bacterial microcolony growing wrapped around a spongy mesophyll cell (SMC) (as microcolony and frame shown in f and Z-stack shown in Supplementary Video 9). d, Frames of a Z-stack acquisition taken from an apoplastic microcolony of hrpL::GFP3 fliC::tdT bacteria at 4 dpi. Contrast and brightness were adjusted to improve visualization and kept constant throughout. Scale bars, 10 µm. Lower epidermis and mesophyll indicate relative position of frames within the leaf. e, Schematic distribution of fluorescence for fliC::tdT and T3SS::GFP3 within the type of apoplastic microcolonies indicated in c (1–3) and shown in d: planes closest to lower epidermis (abaxial side) show predominantly green bacteria and those furthest predominantly red, with intermediate plane displaying yellow. Heterogeneity is observed within each plane. f, Z-stack frame taken from an apoplastic microcolony of hopAB1::GFP3 fliC::tdT bacteria (Z-stack in Supplementary Video 9) from experiments described in a. Scale bars, 20 µm. g, Schematic fluorescence distribution for fliC::tdT T3SS::GFP3 bacteria within apoplastic microcolonies of the type indicated in c (4) and shown in f: bacteria closest to the cell are predominantly green, turning to yellow and red as they get further away. Heterogeneity is displayed throughout. In e and g, legend indicates green as T3SSON/ FlagellaOFF bacteria, red as T3SSOFF/ FlagellaON and yellow/orange as bacteria expressing both to different levels. Fig. 3 | Flagellar and T3SS expression impact on bacterial growth. a, Left: competitive index assay overview: inocula containing 1:1 mixes of wild-type and mutant (or plasmid-carrying) LB-grown cultures are used to inoculate leaves (in planta CI) or to start fresh cultures (in vitro CI). Resulting mixed cultures and apoplast-extracted bacteria (4 dpi) are serially diluted and plated with or without antibiotics for c.f.u. determination (output). Right: CI is calculated as mutant-to-wild-type c.f.u. ratio in output divided by ratio in input. CIs different from 1.0 indicate strains growing significantly differently: mutant outgrowing wild type (CI > 1.0) or vice versa (<1). b, CIs for mutants vs wild-type strain (WT) in HIM or LB (n = 3–9 as shown). c, HIM growth rates for ΔhrpL mutant and wild type. d, In planta CIs for wild type constitutively expressing hrpL (pHrpL) or ΔfliC, vs WT at 4 dpi (n = 10 as shown). Plants for ΔfliC CI pretreated with flg22. In b–d, data show mean ± s.e.m. Individual data for each biological replicate shown; different shades identify independent experiments, and results with asterisks are significantly different from 1.0 (non-parametric two-sided Student’s t-test). Significant P values indicated. e, Time-lapse images of fliC::GFP3 bacteria growing on agar pads: phase-contrast channel (top), GFP fluorescence (bottom). Contrast and brightness adjusted to improve visualization and kept constant across frames. f, Correlation between growth rates and fluorescence intensity of individual fliC::GFP3 cells. Each point represents a single cell, with fliC expression and growth rate averaged over cell lifetime. Grey-shaded area shows 95% confidence interval around linear regression model fit. Correlation tested with Pearson’s correlation coefficient using two-sided test (R = −0.17; P = 1.03 × 10−9). g, Growth rates for high-GFPand low-GFP-signal cells (determined by splitting the population using median fluorescence intensity) show significantly slower growth in high-GFP cells (two-sided Mann–Whitney U-test, P = 3.6 × 10−8). Data presented as boxplots with first and third quartiles as bottom and top bounds, median shown as horizontal line in the box, and 1.5 interquartile range shown as whiskers. Outliers not shown. Grey violin plot shows distribution of all data points summarized in boxplot. White diamonds show mean growth rate for all cells considered. Nature Microbiology | Volume 10 | April 2025 | 958–972 964 Article https://doi.org/10.1038/s41564-025-01966-0 by stochastic cellular changes (gene expression noise) and particular regulatory circuits (enhancing noise)32. P. syringae produce polar flagella 33,34 upon activation of FleQ 15 under a regulatory cascade that is mostly uncharacterized. Gene arrangement and promoter motifs conservation with P. putida14, which also produces polar flagella, support a similar organization. Nonetheless, we show that P. syringae flagellar regulation shares an important feature with that of peritrichous Salmonella: stochastic cell-to-cell variation (Fig. 1 and Extended Data Figs. 1–3)9–13,35,36. Bistable activation of Salmonella flagella is established by two mechanisms 9–13 : (1) a double-negative feedback loop and (2) a developmental checkpoint. A dual mechanism is also involved in turning P. syringae c e Lower epidermis MesophyllZ-stack d a 1 dpi 5 2 dpi 5 4 3 SMC SMC Lower epidermis Lower epidermis T3SS::GFP3 fliC::tdT T3SS::GFP3 fliC::tdT 123 4 2 1 3 dpi 5 f g 10 1 dpi 10 2 dpi 10 3 dpi 10 4 dpi b Nature Microbiology | Volume 10 | April 2025 | 958–972 965 Article https://doi.org/10.1038/s41564-025-01966-0 T3SS noisy expression into bimodality20: (1) a double-negative18,37 and (2) a positive feedback loop38. Research into the regulation of flagellar synthesis in P. syringae will help identify elements involved in phenotypic heterogeneity. Flagellar motility is often costly 26–28 . Flagellar expression also has a cost in P. syringae (Fig. 3 and Extended Data Fig. 4), although this cost is clearer at the single-cell level (Fig. 3e–g). In Salmonella, costs linked to expression of SPI1 T3SS have been proposed as a factor favouring both counter-regulation with flagella and maintenance of heterogeneity 24,25 . T3SS expression also carries a growth penalty in P. syringae (Fig. 3), and the strong growth advantage of a ΔfliC ΔhrpL double mutant indicates an even greater fitness cost associated to expression of both. Phenotypic heterogeneity is considered particularly advantageous when affecting loci producing immunogenic and/or energetically costly goods, with cost to the producers and benefits to the population defining a cooperative behaviour39–41. The growth penalties detected for flagellar and T3SS expression in P. syringae provide selective advantages to phenotypically heterogeneous populations and may drive selection of genetic and/or epigenetic mechanism(s) underlying such heterogeneities. For bacteria engaging in host interactions, heterogeneous expression lowers the overall amount of flagellin displayed by the population, or by a microcolony at a local level, and could be also advantageous by reducing defence elicitation and facilitating immune suppression. Cooperative virulence can also prevent the rise of non-producing mutant variants at the expense of the producers 42–45 . Phenotypically OFF bacteria can outcompete less frequent mutant OFF variants and safely revert to ON, thus preventing loss of function and stabilizing the cooperative behavior43. Cheater mutant variants affected in type III secretion have been described in P. syringae populations colonizing Arabidopsis, although at frequencies lower than expected considering the potential for exploitation of T3SS effectors as public goods 22,46 . Spray Extraction Natural 0 50 100 fliC::tdT <0.0001 **** MergedtdTomatoGFP Total30-min incubation hopAB1::GFP3 fliC::tdT b a 1 dpi 7 dpi Extraction Natural 0 50 100 fliC::tdT 0.237 Extraction Natural 0 50 100 150 200 250 Fluorescence intensity (a.u.) hopAB1::GFP3 0.349 Extraction Natural 0 50 100 150 Fluorescence intensity (a.u.) hopAB1::GFP3 0.361 Fig. 5 | Active exit from infected tissues is carried out by FlagellaON bacteria. a, Selected images of apoplast-extracted bacteria (using negative pressure; top panels), bacteria exiting the tissue on their own (during 30 min leaf incubation within an MgCl2 solution; middle panels) or bacteria recovered from excess solution dripping from leaves sprayed every 5 min to mimic rain (bottom panels). Contrast and brightness were adjusted to improve visualization but were kept constant across the different conditions. b, Fluorescence quantification of the images obtained in a (left images) at 1 dpi (left graphs) and at 7 dpi (right graphs) of the same time-course experiment, using Fiji software. Graph shows arbitrary units of GFP fluorescence corresponding to the expression of the T3SS gene fusion hopAB1::GFP3, or tdTomato fluorescence corresponding to the expression of fliC::tdT. Each dot corresponds to an individual bacterium as analysed from the image (n = 44). Data are presented as mean ± s.e.m. Graphs show results representative of 5 independent experiments (n = 2 for strain carrying hrpL::GFP3 fliC::tdT and n = 3 for strain carrying hopAB1::GFP3 fliC::tdT). Comparisons between apoplast extraction and natural exit were carried out per sample using an unpaired t-test (****P < 0.0001) and results shown for each pair. Nature Microbiology | Volume 10 | April 2025 | 958–972 966 Article https://doi.org/10.1038/s41564-025-01966-0 This role of T3SS effectors as public goods has been reported using mutants 32 and metapopulations, where mixed isogenic P. syringae derivatives lacking effectors, and each overexpressing a single effector, cooperatively restore virulence46. Phenotypic heterogeneity can also allow bacterial populations to cope with rapid environmental changes, since preexisting subpopulations preadapted to incoming stresses can overcome these changes faster39,40. This is known as ‘bet-hedging’41. Phenotypic heterogeneity can also lead to ‘division of labour’ when a population diversifies, allow - ing the distribution of tasks among phenotypically different variants. While in bet-hedging, one subpopulation is fitter than the other in each environment, division of labour increases fitness for the entire population 41 . We found no evidence of any T3SS/flagellar phenotypic variant benefiting during plant colonization, as expected in a bet-hedging scenario. However, different defence elicitation contexts (for example, resistant plants) might differentially benefit certain phenotypes, as shown for genome-reorganized variants 47,48 . Our results support a ‘division of labour’ scenario, with T3SS ON bacteria complementing T3SSOFF probably through suppressing immunity by ‘common goods’ T3SS effectors22 (Fig. 4 and Extended Data Fig. 5). The percentage of T3SS ON bacteria required for effective trans complementation of T3SSOFF, since a subpopulation is expressing immunogenic flagellin, is likely to be balanced by growth trade-offs. As eukaryotic cell types work within tissues, bacterial cooperation emerges within spatially structured populations49. However, structured cooperation has been rarely shown within the context of disease 50 . Indeed, although phenotypic heterogeneity is well documented in planktonic cultures, how these processes occur in spatially structured communities is less understood51. The regulatory loops involved in establishing bistable expression of the T3SS in planktonic HIM20 cultures are in place during plant colonization 19,52 . However, apoplastic bacteria encounter different stimuli specific to their microenvironment, which changes over time, potentially leading to cell-to-cell variation orthogonal to that generated through T3SS and flagellar stochastic heterogeneity. Time-course analysis of T3SS and flagellar expression during plant colonization shows heterogeneity throughout time and locations, but also an overlapping structured pattern (Fig. 4). On the leaf surface, most bacteria are T3SS OFF regardless of the single-cell status of flagellar expression, changing to a majority of T3SSON/ FlagellaOFF during early growth in the apoplast (Fig. 4a,b, 1 dpi). From 2 dpi onwards, heterogeneity is the norm for both loci, but predominantly T3SSON or FlagellaON areas appear. These areas show a three-dimensional (3D) pattern: those closely associated to host-cell surfaces are enriched in T3SS ON bacteria, while distal areas, or the innermost parts of the microcolony, are predominantly FlagellaON (Figs. 4 and Extended Data Fig. 7, and Supplementary Videos 4 and 5). This is consistent with the notion of T3SS ON bacteria predominantly located where T3SS activity is relevant to suppress cellular immunity and explains the lack of selective killing of T3SS OFF bacteria 26 (Extended Data Fig. 5). Such phenotypic differentiation into spatially distributed subpopulations cooperating in a complex environment (that is, in planta vs in vitro) has rarely been investigated, particularly for more than a trait 53 , but it is reasonable to expect it to take place in the context of host colonization processes. Spatially structured cooperation in the context of disease has been demonstrated for cholera toxin and toxin-coregulated pilus in Vibrio cholerae in mice intestinal microcolonies, and for nitric oxide (NO)-detoxifying hmp gene in Yersinia pseudotuberculosis in spleen microcolonies49,53. Spatial heterogeneity has also been described through single-cell omics within in vitro biofilms of human pathogen P. aeruginosa, including variation in flagellar expression54. Host-cell signals, cell-to-cell interactions or lineage history55 could be involved in T3SS activation in host-proximal areas. How environmental signals combine with stochastic processes to generate phenotypic variation is a current topic of interest 55 , one particularly challenging to study beyond in vitro systems if positional information is to be obtained55. As apoplastic microcolonies develop, FlagellaON bacteria increase in numbers. Motile bacteria are expected to exit the leaf more efficiently (Extended Data Fig. 7), and our results experimentally support this notion by showing that mostly FlagellaON bacteria exit colonized tissue before necrosis (Fig. 5). The final stages of the infection and host exit is an understudied domain, even for model pathogen P. syringae, so whether this active early exit is more biologically relevant than passive exit from necrotic tissue remains to be explored, but it appears a safer strategy since necrosis leads to the release of toxic compounds and to the activation of additional defence responses56,57. Only a handful of studies have analysed expression of two bistable heterogeneous loci simultaneously49,53,58. Ours studies two loci relevant for bacterial–host interaction (Supplementary Video 10) and does it in the context of disease within a spatially structured dynamic microcolony. It provides an example of division of labour and cooperative virulence in a plant host, validating the importance of phenotypic heterogeneity in nature and shifting the emphasis from its mechanistic understanding to its ecological and biological relevance. Finally, the study establishes phenotypic heterogeneity and cooperative virulence as a conserved strategy by which bacterial pathogens cope with the fluctuating challenges encountered in plant and animal hosts. Methods Bacterial strains and growth conditions Bacterial strains used and generated in this work are detailed in Supple - mentary Table 1. E. coli and P. syringae strains were grown with aeration in LB medium59 at 37 °C for E. coli or 28 °C for P. syringae. Antibiotics were used, when necessary, at the following concentrations: ampicillin, 100 µg ml −1 for E. coli and 500 µg ml −1 for P. syringae; kanamycin, 50 µg ml−1 for E. coli and 15 µg ml−1 for P. syringae derivative strains; gentamycin, 10 µg ml−1; nitrofurantoin, 40 µg ml−1; and cycloheximide, 2 µg ml−1. To induce the expression of the hrp/hrc genes, bacteria were initially cultured overnight in LB at 28 °C, supplemented with the appropriate antibiotic, then washed twice in 10 mM MgCl2 before being cultured in hrp-inducing minimal medium (HIM), containing 10 mM fructose60. For this study, the pH of HIM was adjusted to 7.0 with 10 N NaOH. The initial optical density (OD) was adjusted to 0.13 and cultures were incubated at 28 °C with agitation. Fluorescence labelling of bacterial strains Bacterial strains carrying a chromosome-located transcriptional fusion of fliC gene to a promoterless tdTomato gene were generated using an adaptation of the method previously described in ref. 61. The plasmids used and generated for this purpose are detailed in Supplementary Table 2. The primers used are described in Supplementary Table 3. For the generation of the allelic exchange plasmid, two fragments of ~500 pb were amplified from Pph 1448A genomic DNA using Q5 High-Fidelity DNA Polymerase (New England Biolabs); one of these fragments (A) encompasses the 3′ end of the open reading frame (ORF), including the STOP codon, while the other fragment (B) covers the sequence immediately downstream of the STOP codon. All primers used are described in Supplementary Information. Each reaction was carried out at 98 °C for 1 min for the initial denaturation step, followed by 30 cycles at 98 °C for 30 s, annealing at 62 °C for 30 s and extension at 72 °C for 30 s, followed by 5 min at 72 °C for the final extension step. The reaction mixture for each PCR included 0.64 mM deoxynucleotide triphosphate (dNTP) mix, 0.4 ng of each primer, 1 ng of genomic DNA, the appropriate enzyme buffer and commercial ultrapure water (Nalgene). A volume of 2 µl of each gel-purified PCR product was employed as template for the subsequent fusion PCR, employing primers A1 and B2 in a PCR reaction conducted under the conditions described, with an extended elongation time of 1 min. The resulting bands, comprising the end of each ORF and its downstream sequence separated by an EcoRV restriction site, were A/T cloned into Nature Microbiology Article https://doi.org/10.1038/s41564-025-01966-0 Extended Data Fig. 1 | See next page for caption. Nature Microbiology Article https://doi.org/10.1038/s41564-025-01966-0 Extended Data Fig. 1 | Transcriptional fusions to tdTomato do not affect gene function. (a) A strain carrying a fliC::tdT transcriptional fusion displays wild type motility in the photographs (a) and quantifications (b) of swimming plates (n = 4). No statistical differences were observed as indicated by an unpaired two-tailed t test (P < 0.05). Scatter dot graph shown mean +/− SEM. The fliC::tdT construct is integrated in the native chromosome location of the fliC gene as established by Southern blot analysis (using the nptII gene conferring resistance to Km as a probe) of a Alw21I (Thermo Fisher Scientific, USA) digestion of genomic DNA from the wild type (Pph) and fliC::tdT derivative strains (c) showing the expected band sizes (d). (e) Representative images of the symptoms caused by the wild type strains and their relative hopAB1::GFP3 fliC::tdT, hrpL::GFP3 fliC::tdT and hrcU::GFP3 fliC::tdT strains at 5 and 7 dpi. (f) Representative images of the colonies obtained after plating serial dilutions of the indicated strains previously adjusted to 0.1 optical density (Abs600). (g) Quantification of the data obtained in f. All dilutions reach similar values indicating no differences in CFU/ ml occasioned by expression of fluorescence reporter genes. f and g illustrate one experiment (>3 technical replicates), data from 3 additional independent experiments carried out in the same manner are showed in h. (h) Mean number of cfu/ml obtained after plating a dilution of the culture previously adjusted to 0.1 optical density (Abs600). Data obtained for the different strains (n = 3 biological replicates; technical replicates not used for statistical analysis) did not show significative differences according to a one-way ANOVA followed by Tukey’s multiple comparison test (P < 0.05). Graph shows individual values (scatter plot) and mean +/− SEM is also indicated. Nature Microbiology Article https://doi.org/10.1038/s41564-025-01966-0 tdTomatoeCFP Bright field hopAB1::GFP3 fliC::tdT eCFP Extended Data Fig. 2 | See next page for caption. Nature Microbiology Article https://doi.org/10.1038/s41564-025-01966-0 Extended Data Fig. 2 | Flagella expression is heterogeneous on plant surface. Selected images of the hopAB1::GFP3 fliC::tdT eCFP strain on the plant surface at 6 h post inoculation (hpi) by dipping the leaf into a 5 × 107 CFU/ml bacterial suspension. tdTomato panel shows the fluorescence of tdTomato associated to fliC expression, and eCFP panel shows the fluorescence of eCFP as constitutive expression reporter, using grey scale in both cases to improve contrast. The GFP panel, corresponding to hopAB1 expression, is not shown since no fluorescence was detected. Scale bars correspond to 20 µm. Contrast and brightness were adjusted to improve visualization but were kept constant across panels. Circles highlight bacteria detected in the eCFP panel without expression in the tdTomato channel (FlagellaOFF). Right panels correspond to the magnification of the area selected in rectangles. Images are representative of 3 independent experiments. Nature Microbiology Article https://doi.org/10.1038/s41564-025-01966-0 Extended Data Fig. 3 | See next page for caption. Nature Microbiology Article https://doi.org/10.1038/s41564-025-01966-0 Extended Data Fig. 3 | Flagella display heterogenous expression in Pseudomonas syringae. (a) CSLM of a strain carrying fliC::tdT grown O/N in LB (upper), or 24 h in HIM (central), or extracted from bean leaf apoplasts 4 dpi (bottom). tdTomato panels show tdTomato fluorescence as reporter of fliC expression and BF panel corresponds to bright field. Scale bars correspond to 2 µm. Images are representative of 3 independent experiments (b) FC analysis of fliC::tdT strain in conditions described in a. Dot plots show cell size versus tdTomato fluorescence intensity, represented as arbitrary units in logarithmic scale. Data collected for 100,000 events per sample. The non-tdT graph show autofluorescence levels displayed by wild type not carrying any fluorescent gene. Vertical lines leave 99% of data of non-tdT strain to the left and is used as reference to differentiate for OFF cells. a and b show representative results of at least three independent replicates. (c) FC analysis shows GFP3 and TdTomato fluorescence detection is not bias by bleed through. Left graphs correspond to fliC::GFP3 strain and right to the fliC::tdT strain. Upper graphs show tdTomato fluorescence versus cell count and bottom graphs show GFP fluorescence versus cell count. In all cases the non-fluorescent strain is included as negative control and data were treated as in b. (d) Quantification of GFP lifetime (τ) in the presence or absence of tdTomato. Bacteria expressing similar levels of hrpL::GFP3 (determined as GFP fluorescence intensity) and differing in the expressing of fliC::tdT expression (as shown by different tdTomato fluorescence intensity) were chosen for the analysis (upper panel). GFP lifetime was obtained for each pixel of the image (lower-left) and mean was calculated for each single bacterium. Average of four bacteria in each condition is represented (lower-right) and statistical significance determined with a t-test. (e and f) Dot plot graphs display fluorescence intensity of GFP or tdTomato versus cell size in non-fluorescent bacteria (wild type), or strains carrying hrpL::GFP3 fliC::tdT or hopAB1::GFP3 fliC::tdT corresponding to data shown in Fig. 4 as GFP fluorescence versus that of tdTomato. Figure show representative results of at least three independent experiments. In (e), bacteria grown 24 h in HIM after diluting an O/N LB culture, display reproducible bistable expression of both hrpL::GFP3 and hopAB1::GFP3 and heterogeneous (occasionally bistable) expression of fliC::tdT. In the case of hopAB1::GFP3 fliC::tdT strain in HIM, an additional vertical line (dashed) was added to separate the subpopulations differing in hopAB1 expression, which is higher due to the high basal levels of this gene. In (f), bacteria extracted from bean leaf apoplasts 4 dpi display typical heterogeneous (never bistable) expression of hrpL::GFP3, hopAB1::GFP3 and fliC::tdT. Nature Microbiology Article https://doi.org/10.1038/s41564-025-01966-0 Extended Data Fig. 4 | See next page for caption. Nature Microbiology Article https://doi.org/10.1038/s41564-025-01966-0 Extended Data Fig. 4 | T3SS and flagella expression impact on bacterial growth and growth cost associated with single cell levels of flagella production is not due to GFP. (a) Growth rates of 1448 A wild type, a derivative carrying a plasmid that determines constitutive expression of hrpL (pHrpL), or of fleQ (pFleQ), or a fleQ deletion (ΔfleQ), in HIM. Data correspond to 4 biological replicates. Data are presented as mean values +/− SEM. Significative differences are observed between WT and pFleQ strains according to a one-way ANOVA followed by Tukey’s multiple comparison test (P < 0.05). P value is indicated in the graph. (b) Selected timelapse images of wild type constitutively expressing GFP (eGFP) during the microcolony development on agar pads in phase contrast (top) and GFP fluorescence (bottom) channels. Contrast and brightness were adjusted and maintained throughout. (c) Comparisons of mean growth rates for eGFP strain and fliC::GFP3 strain. Note that GFP in the eGFP strain has a much higher fluorescence intensity compared to fliC::GFP3. eGFP shows a significantly lower average growth rate compared to fliC::GFP (Mann-Whitney U test, P = 7.7 × 10−15). (d) Comparison of growth rate of cells displaying high or low GFP signal (determined by splitting the population into two groups using the median fluorescence intensity value) for eGFP strain shows no significant differences between growth rates of cells expressing high or low GFP levels (Mann-Whitney U test, P < 0.05). (e) Correlation between growth rates and fluorescence intensity of individual fliC::GFP3 cells indicates that higher fliC expression is associated with slower growth. The shaded area shows the 95% confidence interval. (f) Correlation between growth rates and fluorescence intensity of individual eGFP cells indicates that higher GFP expression is not associated with slower growth. The shaded area shows the 95% confidence interval. Nature Microbiology Article https://doi.org/10.1038/s41564-025-01966-0 Extended Data Fig. 5 | Dead-live staining shows neither bias towards FlagellaON cells nor for T3SSOFF cells during plant growth. Graph shows the dead/live ratio for bacteria either expressing or not fliC::GFP3, hrpL::GFP3, hopAB1:GFP3 or hrcU::GFP3 during plant growth. Apoplast-extracted bacteria at 4 dpi in bean leaves were stained with a solution of Propidium iodide (PI). Data were obtained by flow cytometry analysis (100,000 events/ per replicate) and GFP levels were used to differentiate between ON and OFF cells using the non-fluorescent wild type strain as reference, as indicated before. Fluorescence of PI was used to differentiate dead and alive bacteria comparing to the non-fluorescent wild type. Each dot represents biological replicate in two different experiments (Rep 1 and Rep 2). Nature Microbiology Article https://doi.org/10.1038/s41564-025-01966-0 Extended Data Fig. 6 | Active exit from infected tissues is preferentially carried out by FlagellaON bacteria. (a) Selected images of apoplast-extracted bacteria (using negative pressure; upper panels) or bacteria exiting the tissue on their own (during leaf incubation within a MgCl2 solution; natural exit, lower panels) at 1 dpi and 2 dpi (asymptomatic tissue), or 5 dpi (just before visible symptoms appear) and 7 dpi (fully symptomatic necrotic tissue). Contrast and brightness were adjusted to improve visualization but were kept constant across the different conditions. (b and c) Fluorescence quantification of the images obtained in a for 2 dpi and 5 dpi using Fiji software. Graph shows arbitrary units of GFP fluorescence, corresponding to the expression of the T3SS gene fusion hopAB1::GFP3 in the samples, or tdTomato fluorescence corresponding to the expression of fliC::tdT. Each dot corresponds to an individual bacterium (n = 44) as analyzed from the image. Comparisons between apoplast extraction and natural exit were carried out per each sample using an unpaired two-tailed t-test. Data are presented as mean values +/− SEM. Graphs show results representative of 3 independent experiments.