Artificial symbiont replacement in a vertically transmitted plant symbiosis reveals a role for microbe–microbe interactions in enforcing specificity

Abstract

Some plants engage in permanent, vertically transmitted symbioses with bacteria. Often, these bacteria are hosted extracellularly within structures on the leaves, where they produce specialized bioactive metabolites that benefit their host. These associations are highly specific, with one plant species associating with a single bacterial species, but little is known about how these symbioses originate and how specificity is maintained. In this study, we show that the symbiotic association between a wild yam and a bacterium can be manipulated experimentally and that bacteria-free plants are open to colonization by environmental bacteria. Through metabolic profiling, we show that the endophytic niche is rich in organic acids and intermediates of the tricarboxylic acid cycle cycle. Environmental bacteria capable of utilizing these acids, such as the soil bacterium Pseudomonas putida, readily colonize aposymbiotic plants. However, successful colonization is contingent upon the absence of the vertically transmitted symbiont or the impairment of its type VI secretion system. Unexpectedly for a vertically transmitted symbiosis, these findings suggest that microbe–microbe interactions, including antagonism, may play a crucial role in maintaining the specificity of an association. However, low transmission rates of synthetic symbionts provide evidence that transmission barriers or bottlenecks may still occur, further enforcing partner fidelity. Together, these results highlight the complexity of mechanisms underlying mutualistic associations, and provide insights into the evolution of bacterial leaf symbiosis.

Introduction

Mutualistic interactions are generally costly to maintain, and a degree of specificity is often important for the maintenance of a beneficial holobiont. Specificity between symbiotic partners may be driven by a combination of three factors: the mode of acquisition of the symbiont, environmental selection (e.g. via restrictive nutrients or physico-chemical properties within the niche), and mechanisms underlying recognition between the host and its symbiont [1]. Partner specificity may be achieved by a combination of an aposymbiotic phase, during which the host exists without the partner, and selective barriers allowing the establishment of a cooperative symbiont. For example, in the well-studied legume-Rhizobium model, plants recruit their symbionts from the environment in a multi-step process. In this case, partner selection relies on the exchange of signaling molecules (flavonoids and Nod factors), and the presence of specific structures on the surface of the bacterial cell [2, 3]. Similarly, the bobtail squid Euprymna scolopes recruits Vibrio fischeri symbionts into ciliated crypts from surrounding sea water. Microbe-associated molecular patterns (MAMPs) derived from the cell surface of V. fischeri trigger morphogenesis of the light organ. Importantly, host tissues contain anti-microbials, which create a selective environment favoring V. fischeri [4].

Microbe–microbe interactions may also play a key role in the establishment of a symbiosis, where a host may foster competition between a preferred symbiont and nonproductive microorganisms [5]. This phenomenon is known as interference competition, and is a mechanism that underlies partner choice in systems in, which hosts acquire bacteria from the environment at the juvenile stage such as the attine ants/Streptomyces and the Riptortus pedestris/Burkholderia symbioses [6, 7]. Finally, some symbionts are also transmitted vertically from mother to offspring. Vertical transmission links the evolutionary fates of the host and symbionts, which promotes the evolution of mutualistic interactions through partner-fidelity feedback (i.e. a mechanism that promotes cooperation between species through mutual dependence) [8]. Moreover, vertical transmission usually occurs at a single stage of the lifecycle (e.g. by infecting the germline), thus sheltering symbionts from competition [9]. Strictly vertically transmitted symbionts do not need to be selected from an environmental population, and sophisticated mechanisms to enforce specificity are thought not to be necessary.

Hereditary symbiosis is common in animals, especially insects, but much less described in plants. However, some plants engage in permanent, strictly specific associations with bacteria within their leaves, a phenomenon known as leaf symbiosis. Leaf symbiosis has been described in ~500 plant species, mostly in the Primulaceae and Rubiaceae families [10, 11]. The bacteria, mainly belonging to the family Burkholderiaceae, are present in leaf nodules that can take various shapes [12]. Several lines of evidence indicate that the leaf symbionts of some species have a defensive role. For example, Candidatus Caballeronia kirkii (Ca. C. kirkii); the symbiont of Psychotria kirkii (Rubiaceae), is linked to the accumulation of bioactive cyclitols in the plant. Of these, kirkamide displays cytotoxic and insecticidal properties [13], and streptol-glucoside is a potent herbicide [14]. The bacterial symbionts are passed on to the next host generation via the seed [11, 15], but how the symbiotic bacteria reliably colonize seedlings in the presence of a complex soil and seed microbiome remains unknown.

The wild yam Dioscorea sansibarensis forms a permanent association with the bacteria Orrella dioscoreae, and offers an experimentally tractable system of heritable leaf symbiosis [16, 17]. The interaction most obviously takes place in a prominent gland at the tip of the leaves. We recently demonstrated that O. dioscoreae is vertically transmitted through vegetative propagules, called bulbils, and therefore present throughout the entire lifecycle of the plant [18]. Co-phylogenetic analyses indicate that the transmission route is not entirely closed, with horizontal transmission or host-switching also occurring in the wild [19]. However, the D. sansibarensis–O. dioscoreae symbiosis is ubiquitous in nature, implying the existence of mechanisms to enforce specificity [19]. In the laboratory, the plant and the bacteria can be grown separately, suggesting that this symbiosis is not obligate despite its high level of specificity in nature. Aposymbiotic plants develop normally in vitro, with fully formed leaf glands devoid of bacteria [20].

We demonstrated in a previous study that inoculation of aposymbiotic plants relies on colonization of the shoot apical bud, which contains the meristematic tissue at the origin of all above-ground organs throughout the life of the plant [18]. In this regard, plants face a drastically different situation from animals regarding vertical transmission of symbionts. Because the meristem links somatic and reproductive organs, plants cannot easily segregate symbionts between a germ line and a somatic population. Moreover, our previous studies showed that the apical bud remains physically open to the environment, creating opportunities for other microorganisms to establish in the plant at virtually any point during the growing phase [20]. What mechanisms evolved to enforce partner specificity in this context is unknown.

The order or timing of arrival of a species may also influence its success in establishing in a community, a phenomenon known as priority effect [21]. Hereditary symbionts may thus have an ecological advantage as they are already present at the earliest stages of a host’s lifecycle. The presence of O. dioscoreae within shoot buds may prevent strains with similar niche requirements from establishing in the plants via priority effects, for instance by pre-empting access to favored carbon or nitrogen sources. Priority effects may also be paired with other mechanisms such as interference competition, to ensure higher specificity. Type VI secretion systems (T6SSs) are known to have anti-microbial functions through delivery of effectors directly into targeted cells in a contact-dependent manner, and play a key role in microbe–microbe interactions [22, 23]. Indeed, the genome of O. dioscoreae harbors two T6SS gene clusters that are highly expressed in the leaf gland and conserved in all genomes of O. dioscoreae sequenced so far [19, 24]. Both gene clusters code for the 13 core proteins: TssABCEFGJKLM, VgrG, Hcp, and PAAR, as well as for accessory proteins, which usually play a role in regulation and modulation of the system assembling [25]. T6SSs are composed of three main parts, the integral membrane complex (TssJL and M), the baseplate (TssEFG and K), and the tail (Hcp, VgrG, PAAR, TssB and TssC). The tail is constituted of Hcp proteins polymerized into a stack of hexameric rings referred to as the inner tube, topped by the spike VgrG trimer and the PAAR protein [26–28]. The inner tube is surrounded by the sheath (TssBC). Upon contraction of the sheath, the inner tube and the spike are ejected into target cells, delivering the effector proteins [29–33]. In the contracted conformation the ClpV AAA+ ATPase is recruited to depolymerize the TssBC subunits, allowing recycling of the sheath [33–35]. The membrane complex is stable and may be used for multiple injections after reassembly [36]. Inactivation of the ClpV ATPase prevents recycling of the TssB and TssC components, thus precluding re-assembly of the membrane complex after effector delivery and effectively lowering the activity of the T6SS [34, 36, 37]. The role of T6SS-mediated competition in niche monopolization has been demonstrated in various Gram-negative bacteria. In particular, T6SSs were first described in the pathogenic bacterium V. cholerae as essential for both colonization and virulence [22, 38]. Specifically, T6SS-mediated competition is crucial for V. cholerae to outcompete resident microbiota in the human gut during disease [39].

The aim of this study was to acquire a better understanding of the mechanisms underlying specificity in the D. sansibarensis–O. dioscoreae symbiosis. We show that despite a ubiquitous and highly specific association, leaf glands of aposymbiotic D. sansibarensis are open to benign colonization by unrelated bacterial species with similar niche requirements. Co-inoculation assays on aposymbiotic seedlings show that O. dioscoreae is highly competitive in planta, and that this competitiveness is dependent upon a functional T6SS.

Materials and methods

Plant culture and propagation

D. sansibarensis Pax plants were obtained from the greenhouse of the Botanical Garden at the University of Ghent (LM-UGent) in Ghent, Belgium. Chemicals and reagents were purchased from Merck unless otherwise indicated. Plants used throughout in experiments were maintained at the Laboratory of Plants Microbes and Environment Interactions (LIPME) in Castanet-Tolosan, France. Unless otherwise indicated, plants were grown in climate-controlled chambers at 28°C, 70% humidity and a light cycle of 16 h light (210 μmol/m²/s), 8 h dark. Unless otherwise indicated, plants were kept in standard potting soil (PROVEEN SB-2; Bas Van Buuren B.V., Holland).

Bacterial strains, plasmids, and growth conditions

Bacterial strains and plasmids are listed in Table S1. Routine culture of O. dioscoreae strains was done at 28°C in Tryptic Soy Broth (TSB, Sigma) or Tryptic soy agar (TSA, Sigma) supplemented with 30 μg/ml of nalidixic acid, 50 μg/ml of kanamycin, or 20 μg/ml of gentamicin where appropriate. Stenotrophomonas sp., Rhizobium sp., and Sphingomonas sp. strains were grown in TSB medium at 28°C unless otherwise specified. P. putida and Escherichia coli strains were grown in LB medium with appropriate antibiotics, at 28°C and 37°C respectively. To selectively grow O. dioscoreae R-71412 and derivatives, AB minimal medium [40] was supplemented with 0.01% (w/v) yeast extract (ABY medium), 10 mM trisodium citrate (Sigma), nalidixic acid (30 μg/ ml), and gentamicin (20 μg/ ml). Pseudomonas Isolation Agar (PIA, Sigma) supplemented with 2% glycerol was used to grow P. putida selectively.

Production of aposymbiotic plants

Aposymbiotic plants were produced from node cuttings treated with antibiotics as described in [18]. Briefly, nodes were dissected from growth chamber-grown plants, and surface-sterilized for 8 h in a solution of 3× Murashige Skoog medium (MS, Sigma M5524) supplemented with 5% of Plant Preservative Mixture (PPM, Plant Cell Technology, USA). Explants were then aseptically transferred to sterile six-well plates containing MS medium supplemented with sucrose 2% (Merck), myo-inositol 555 μM (Sigma), glycine 26.6 μM (Sigma), cysteine 16.5 μM (Sigma), nicotinic acid 4.06 μM (Sigma), pyridoxine 2.96 μM (Sigma), thiamine 1.88 μM (Sigma), PPM (0.2% w/v), and the antibiotics carbenicillin and cefotaxim (200 μg/ ml each). Explants were incubated in a growth chamber at 28°C under 16 h/8 h day/night cycle. The medium was replaced after 10 days. After 3 weeks of incubation, the explants were aseptically transferred to sterile Magenta boxes (model GA7, Magenta) containing MS medium as above without the carbenicillin and cefotaxim. Effectiveness of the treatment was tested for each explant by collecting the first two leaves, milling in 100 μl of 0.4% w/v NaCl using a Retsch MM400 bead mill (1 m, 30 Hz), and spreading the macerate on TSA medium (Sigma). The absence of visible microbial growth after 2 days of incubation at 28°C was taken as evidence of the aposymbiotic status of the plants.

Isolation and identification of leaf gland bacteria

Aposymbiotic plants were transferred from closed, sterile containers to open pots in the greenhouse (25°C, 16 h light/8 h dark, 60% humidity) for at least 8 weeks. Leaf acumens were dissected and surface-sterilized (5 m in 70% ethanol, sterile distilled water wash, 5 m in 1.6% sodium hypochlorite, three washes in sterile NaCl 0.4%). Samples were homogenized as described above and plated on TSA medium. Bacterial colonies were isolated on TSA. Identification was done by 16S rRNA gene sequencing using Sanger sequencing and PCR primers 27f and 1492r [41].

Inoculations of D. sansibarensis aposymbiotic plant with bacteria

Overnight bacterial cultures were grown in LB or TSB as appropriate, and harvested by centrifugation at 7000 g for 5 m. Cell pellets were washed twice in sterile 0.4% NaCl. Bacterial suspensions were normalized to OD_{600 nm} = 0.2, corresponding to ~0.2 x 10⁹ CFU/ ml for O. dioscoreae and P. putida strains. Prior to inoculation, plants were moved to sterile Microbox containers (SacO2, Belgium) containing 50 ml of MS medium. Plants were inoculated by depositing 2 μl of the bacterial suspension directly onto the apical bud. Inoculated plants were grown at 28°C with a 16 h light/8 h dark cycle for 28 days. When applicable, a second inoculation was done a week later as described above. Colonization was evaluated by spreading serial dilutions of the milled and weighted leaf glands onto appropriate agar medium.

Detection of bacteria in the second generation of inoculated plants

Previously inoculated plants were transferred from closed containers to open pots in growth chambers (25°C, 16 h light/8 hours dark, 70% humidity) until the end of the growing season. Bulbils were harvested and stored in a dry place until spontaneous sprouting. Sprouting bulbils were laid on a moist sand substrate and transferred to a growth chamber with high hygrometry (25°C, 16 h light/8 h dark, 90% hygrometry). Plantlets with at least two leaves were transferred to lower hygrometry (25°C, 16 h light/8 h dark, 70% hygrometry). Two leaf acumens per plant were tested for the presence of the bacteria. Leaf acumens were dissected and surface-sterilized as described above, and the presence and identity of bacteria were tested as described above.

Observation of bacteria in leaf glands by fluorescence microscopy

Leaf glands were dissected with sterile scissors 8 weeks after inoculation and fixed in 0.3% paraformaldehyde in 0.1 M pH 7 potassium phosphate buffer for two hours at room temperature under vacuum. Samples were rinsed twice with potassium phosphate buffer (0.1 M, pH 7). 100 μm-thick sections were prepared using a vibrating microtome (Leica VT1000S) after embedding in 5% low melting agarose (Nusieve). Sections were observed using an epifluorescence microscope (Zeiss Axioplan2, excitation 475–40 nm and emission 530–50 nm). Images were processed using the ImageJ software version 1.54. Observation of non-fluorescent bacteria was done by collecting leaf glands 5 weeks post inoculation. Samples were prepared by sectioning with a razor blade, followed by staining with SYTO 9 (Thermo Fisher) and visualization using confocal microscope (Leica SP7). Images were processed using Leica LasX software.

Bacterial genetics

T6SS deletion mutants were produced in an R-71412 background by homologous recombination as previously described [42, 43]. Briefly, the flanking regions of the gene to delete were amplified by PCR using specific primers (Table S2). The pSNW2 plasmid DNA was digested by XmaI (New England Biolabs) and gene fragments were assembled using the Pro Ligation-free cloning kit (Applied Biological Materials, Richmond, BC, Canada). All constructs were verified by whole-plasmid sequencing using ONT Nanopore sequencing [44]. Orrella dioscoreae R-71412 was electroporated with plasmid DNA as previously described, and kanamycin-resistant clones were selected [18]. Plasmid pQURE6 harboring a I-SceI nuclease gene was introduced into merodiploid clones by triparental mating. Colonies with gene replacement events were screened for loss of kanamycin resistance and by PCR. The double mutant was produced by deletion of ODI_R0808 in the previously obtained ODI_R3997 deletion mutant. The genome sequences of all mutant strains were obtained by Oxford Nanopore sequencing using R10.4 chemistry on an ONT P2 solo instrument.

For genetic complementation, gene fragments were amplified by PCR from O. dioscoreae R-71412 genomic DNA using specific primers (Table S2). Fragments were cloned into plasmid pSEVA2313 by restriction and ligation. Escherichia coli TOP10 was transformed by electroporation, plasmids were isolated from selected clones and validated by sequencing as above. Plasmid DNA was introduced into O. dioscoreae strains as above. Transformants were selected on TSA medium supplemented with kanamycin (50 mg/L).

Bacterial phenotyping

Phenotyping Microarray (PM) plates were purchased from Biolog (USA). Bacteria were grown on solid media containing appropriate antibiotics and resuspended in sterile distilled water as per the manufacturer’s recommendation. Trisodium citrate was added to wells of the PM03 plate to 0.5% w/v final concentration. NADH oxidation was measured every 15 m for 120 h using an OmniLog instrument. Data were analyzed with the opm R package [45]. For quantitative comparison of O. dioscoreae R-71412 and P. putida KT2440::gfp metabolism on select carbon sources, custom Biolog plates were prepared with ABY medium supplemented with either trisodium citrate, L-malate, sodium succinate, fumarate or sodium pyruvate to 10 mM final concentration. NADH oxidation was measured as above. Data were analyzed on R using the Growthcurver package [46]. Optimal pH range for growth was tested in LB medium supplemented with 10 mM of trisodium citrate adjusted to pH = 5, 6, 7, or 8 and buffered with 100 mM of MES, MOPS or EPPS buffer as appropriate; nonadjusted (pH 7) LB-trisodium citrate medium was used as control. Cultures were inoculated from overnight cultures in TSB, diluted at 1:100. Bacterial growth was monitored by OD₆₀₀ readings every 30 min for 48 h, using BMG FLUOstar Omega microplate reader. The growth rate r was calculated on R using the Growthcurver package [46].

In vitro bacterial competition

LB or ABCY media were inoculated with a suspension of cells of O. dioscoreae mCherry-tagged strain R-71417 and/or with P. putida KT2440::gfp strain suspended in sterile distilled water at OD_600nm = 0.02. Cultures were inoculated with both strains at the same time. As a control for single strain growth, fresh media were inoculated in parallel using each of the same cell suspensions. Growth of O. dioscoreae R-71417 and P. putida KT2440::gfp were estimated by fluorescence measurement, respectively with mCherry (544 nm/590 nm) and GFP (485 nm/520 nm) filters in a BMG FLUOstar Omega microplate reader for 24 h with a measure taken every hour. Growth rate r and carrying capacity K were calculated from fluorescence intensity data using the Growthcurver R package [46]. Statistical analysis was conducted in R v4.1 [47].

In vitro contact-dependent competition assay

Orrella dioscoreae R-71412, T6SS mutants and complemented strains, as well as fluorescent P. putida KT2440::gfp were grown overnight in TSB with appropriate antibiotics at 28°C. Cells were washed in a sterile solution of NaCl 0.4% w/v and density was normalized to OD_{600 nm} = 0.5. Then 10 μl of each sample were spotted onto a 96-well plate (Nunc FluoroNunc) filled with 180 μl of solid growth medium (TSA). Fluorescence was measured for 24 h at 28°C, in a BMG FLUOstar Omega microplate reader (set for spiral top reading, 30 reading points, adjusted with 90% of a 900 arbitrary fluorescence unit gain, filters: excitation 485–12 nm and emission 520–20 nm).

Alternatively, cultures of O. dioscoreae and P. putida or Stenotrophomonas sp. R-67087 strains were prepared as above and concentrated in sterile 0.4% NaCl solution to OD₆₀₀ = 50 for O. dioscoreae strains and OD₆₀₀ = 10 for Stenotrophomonas sp. and P. putida. Then, 5 μl of each sample were spotted onto pre-warmed TSA. After 2, 4, or 6 h at 28°C, cells were resuspended in 1 ml of NaCl 0.4% (w/v), serially diluted and plated on PIA and ABCY with appropriate antibiotics to estimate the respective number of O. dioscoreae and P. putida or Stenotrophomonas sp. colony-forming units (CFU).

Ultra-high-performance liquid chromatography-high-resolution mass spectrometry analysis

Leaf acumens from aposymbiotic plants inoculated with O. dioscoreae R-71412 (6 plants, Sym1-Sym6) or with a solution of NaCl 0.4% (w/v; 4 plants, Apo1-Apo4) were dissected, and immediately frozen in liquid nitrogen. One leaf acumen was randomly chosen for each individual. Samples were milled using a Retsch MM400 bead mill (30 s, 30 Hz), and stored at −80°C until extraction. Samples were extracted with a mixture of methanol: water 80: 20, with a proportion of 1 ml of solvent per 100 mg of sample. Samples were dried and extracts were accurately weighed prior to diluting in solvent. Ultra-high-performance liquid chromatography-high-resolution MS (UHPLC-HRMS) analyses were performed on a Q Exactive Plus quadrupole (Orbitrap) mass spectrometer, equipped with a heated electrospray probe (HESI II) coupled to a U-HPLC Ultimate 3000 RSLC system (Thermo Fisher Scientific, Hemel Hempstead, UK). Separation was done on a Luna Omega Polar C18 column (150 mm × 2.1 mm i.d., 1.6 μm, Phenomenex, Sartrouville, France) equipped with a guard column. The mobile phase A (MPA) was water with 0.05% formic acid (FA), and the mobile phase B (MPB) was acetonitrile with 0.05% FA. The solvent gradient was 0 min, 100% MPA; 1 min, 100% MPA; 22 min, 100% MPB; 25 min, 100% MPB; 25.5 min, 100% MPA; and 28 min, 100% MPA. The flow rate was 0.3 ml/min, the column temperature was set to 40°C, the autosampler temperature was set to 5°C, and the injection volume was fixed to 5 μl. Mass detection was performed in positive ionization (PI) mode at resolution 35 000 power [full width at half-maximum (FWHM) at 400 m/z] for MS1 and 17 500 for MS2 with an automatic gain control (AGC) target of 1 × 106 for full scan MS1 and 1 × 105 for MS2. Ionization spray voltages were set to 3.5 kV, and the capillary temperature was kept at 256°C. The mass scanning range was m/z 100–1500. Each full MS scan was followed by data-dependent acquisition of MS/MS spectra for the six most intense ions using stepped normalized collision energy of 20, 40, and 60 eV. Raw data were processed with MS-DIAL version 4.70 for mass signal extraction between 100 and 1500 Da [48]. MS1 and MS2 tolerance were set to 0.01 and 0.025 Da in the centroid mode. The optimized detection threshold was set to 5 × 10⁵ concerning MS1 and 10 for MS2. Peaks were aligned on a QC reference file with a retention time tolerance of 0.15 m and a mass tolerance of 0.015 Da. Peak annotation was performed with an in-house database built on an MS-FINDER model [49]. MS-DIAL data were then cleaned with the MS-CleanR workflow by selecting all filters with a minimum blank ratio set to 0.8, a maximum relative standard deviation (RSD) set to 30, and a relative mass defect (RMD) ranging from 50 to 3.000. The maximum mass difference for feature relationships detection was set to 0.005 Da and the maximum RT difference to 0.025 min. Pearson correlation links were considered with correlation ≥0.8 and statistically significant with α = 0.05. Two peaks were kept in each cluster, viz., the most intense, and the most connected. The kept features (m/z × RT pairs) were annotated with MS-FINDER version 3.52. The MS1 and MS2 tolerances were, respectively, set to 10 and 20 ppm. Formula finders were only processed with C, H, O, N, and S atoms. Databases (DBs) based on Dioscorea (genus), Alcaligenaceae (family) were constituted with the dictionary of natural products (DNP, CRC press, DNP on DVD v. 28.2). The internal generic DBs from MS-FINDER used were KNApSAcK, PlantCyc, NANPDB, UNPD, COCONUT, and CheBI. Data were normalized per sample based on the sum of all peak areas. Statistical analyses were done using the R software and standard packages.

Results

Aposymbiotic D. sansibarensis are open to colonization by bacteria other than O. dioscoreae

We reported previously that aposymbiotic D. sansibarensis plants were amenable to colonization by exogenously applied O. dioscoreae, suggesting that aposymbiotic plants remained receptive to bacteria during vegetative growth. To test whether this applied to bacteria other than O. dioscoreae, we produced bacteria-free plants by antibiotic treatment of explants. After ~2 months in sterile cultures, plants were moved to open pots for the rest of the growing season. The leaf glands of several of those aposymbiotic plants became spontaneously colonized with bacteria unrelated taxonomically to O. dioscoreae (Table 1). We selected three isolates belonging to the genera Stenotrophomonas sp. (R-67087), Rhizobium sp. (R-71694), and Sphingomonas sp. (R-71695) for further study. Artificial inoculations of aposymbiotic plants showed that these three strains occupy the lumen of the leaf glands, similar to O. dioscoreae R-71412 (Fig. 1A–H). Strains Stenotrophomonas sp. (R-67087) and Rhizobium sp. (R-71694) reached similar population densities within the leaf gland as O. dioscoreae (averages ranging from 0.4 to 10.3 x 10⁸ cfu/leaf acumen, P > .05) (Fig. 1). Sphingomonas sp. (R-71695) also colonized leaf glands, but yielded much lower average densities at 0.03 x 10⁸ CFU/leaf gland (Fig. 1).

Table 1.

Identification of isolates from glands of aposymbiotic Dioscorea sansibarensis.

Plants colonized	Best BLAST hit		Identity (%)
5	Rhizobium radiobacter ATCC 19358		99.7
1	Cupriavidus plantarum ASC-64		99.4
1	Variovorax sp. PMC12		100.0
1	Pantoea sp. ND03		99.9
1	Pseudomonas atacamensis M7D1		99.8
1	Enterobacter soli ATCC BAA-2102		99.7
1	Agrobacterium sp. K599		100.0
2	Stenotrophomonas indicatrix WS40		100.0
1	Pseudomonas fortuita GMI12077		100.0
2	Pseudomonas lactis DSM 29167		100.0
1	Pseudomonas citronellolis NBRC 103043		100.0
1	Pseudomonas neuropathica P155		99.9
1	Pseudomonas paralactis DSM 29164		99.5
1	Cupriavidus campinensis WS2		99.7
1	Sphingomonas sp. 1429(T)		97.6

Figure 1.

Nonsymbiotic bacteria effectively colonize D. sansibarensis leaf glands. Confocal images of cross sections of D. sansibarensis leaf glands colonized by taxonomically diverse bacterial strains after artificial inoculation. D. sansibarensis leaf glands colonized by (A) O. dioscoreae R-71412, (C) Rhizobium sp. R-71694, (E) Stenotrophomonas sp. R-67087, and (G) Sphingomonas sp. R-71695. Details of trichomes surrounded by mucus and green foci corresponding to stained bacteria (B) O. dioscoreae, (D) Rhizobium sp., (F) Stenotrophomonas sp., and (H) Sphingomonas sp. Bacteria were stained with SYTO 9 and autofluorescence was used to visualize plant cells. I. Bacterial titers inside surface-sterilized leaf glands of D. sansibarensis inoculated with the same strains as above (CFU/leaf gland). Bars indicate statistically significant differences between groups according to a Kruskal-Wallis rank sum with Dunn’s post-hoc test and Bonferroni correction for multiple testing. For clarity, only pairwise comparisons with P-value adjusted <0.05 are shown.

Characterization of the metabolic niche of O. dioscoreae

The diverse taxonomic range isolates from aposymbiotic leaf glands prompted us to test whether bacteria with similar physiological properties to O. dioscoreae might be able to colonize aposymbiotic D. sansibarensis. O. dioscoreae grows aerobically, requires a range of temperatures for growth of 15–40°C, and pH slightly acidic to neutral (Fig. S1). We screened a library of compounds to get a comprehensive overview of potential nutrients available for growth of O. dioscoreae. Of the 190 carbon sources tested, only 29 were oxidized by cultures of O. dioscoreae after 72 h (Table S3). The oxidation of citrate, L-malate, succinate, and fumarate indicated that assimilation pathways converge towards the tricarboxylic acid cycle (TCA). Hexose sugars were not utilized by O. dioscoreae, and gluconate supported growth for some, but not all, strains [17]. We also screened 95 compounds as potential sources of nitrogen (Table S3). In particular, ammonia, urea, and the amino acids L-Ala, L-Asn, L-Asp, L-Glu, L-Gln, Gly, L-His, L-Ile, L-Leu, L-Pro, L-Ser, L-Met, D-Ala, D-Asp, D-Glu, and D-Ser supported cellular activity as nitrogen sources. Metabolomics analysis of symbiotic vs. aposymbiotic glands show distinct metabolic profiles (Figs. S2 and S3), with differentially abundant features dominated by fatty acids. These potentially correspond to plant vs bacterial membrane lipids that are expected to vary as symbiotic sample types contain large amounts of bacteria absent in aposymbiotic samples. The data also confirmed that malate, citrate, aspartate, and glutamate were abundant in whole leaf glands, with citrate and malate being the fourth and 14th most abundant ions overall (Table 2). Malate was also significantly enriched by nearly two-fold in aposymbiotic vs. symbiotic leaf glands (Student t-test P = .012). Concentrations of citrate were not significantly different, but tended to be more abundant in extracts of aposymbiotic leaf glands, (1.82× P = .087) (Table 2). We were unable to detect other intermediates of the TCA in our samples, but the fact that citric acid and malic acid were abundant and slightly depleted within symbiotic leaf glands suggests that these compounds may be metabolized by the bacteria within the leaf gland.

Table 2.

Detection of metabolites supporting the growth of O. dioscoreae in symbiotic and aposymbiotic leaf glands of Dioscorea sansibarensis. Average peak intensity of selected metabolites detected in the symbiotic and aposymbiotic leaf glands of D. sansibarensis. Statistical significance of the fold change was calculated by Student’s T-test. Standard deviations are indicated for each average intensity value. ^a Feature ID corresponds to the field “Alignment ID” of Table S4 and S5. ^b Peak intensities are reported normalized and with background subtracted (mean value from blank samples). Negative values correspond to features that have higher average peak intensities in injection blanks. ^c Mean peak intensity across all samples (blanks excluded). Rank is based on mean peak intensity of all features, with lower ranks indicating higher average abundances.

Metabolite	Feature IDa	Peak Intensity			Fold change	Student’s T-test P value
		Mean (rank)c	Aposymbiotic	Symbiotic
Citrate	162_C18neg	5442.9 (4/1367)	7310.6 ± 3646.7	4022.1 ± 2342.1	1.82	0.087
Gluconate	175_C18neg	294.1 (102/1367)	330.2 ± 124.0	233.7 ± 140.2	1.42	0.133
Aspartic acid	63_C18neg	389.8 (84/1367)	487.1 ± 169.6	299.3 ± 123.0	1.63	0.056
L-pyroglutamic acidb	51_C18neg	29.5 (812/1367)	−34.0 ± 3.2	−39.6 ± 9.6	0.86	0.091
Malate	66_C18neg	2471.5 (14/1367)	3520.5 ± 909.2	1760.7 ± 534.4	2.00	0.012
L-Glutamine	83_C18neg	226.7 (130/1367)	145.2 ± 71.42	249.6 ± 76.6	0.58	0.041
L-glutamic acid	84_C18neg	461.4 (76/1367)	391.4 ± 54.89	427.4 ± 219.5	0.92	0.355
Galactonic acid gamma-lactoneb	323_C18pos	23.8 (927/1367)	−35.9 ± 0.91	−3.7 ± 19.3	9.69	0.002
L-Serine	4_C18neg	483.2 (72/1367)	875.2 ± 370.06	206.8 ± 140.1	4.23	0.016

Metabolic niche overlap between P. putida and O. dioscoreae

We hypothesized that the ability to utilize metabolites found in the leaf gland might explain the ability of bacteria to colonize D. sansibarensis. However, strains from our collection of isolates (Table 1), might share underlying adaptations necessary for an endophytic lifestyle inside D. sansibarensis, such as the ability to circumvent host immunity or to avoid competition with O. dioscoreae. We therefore searched the available literature for bacterial strains capable of utilizing the identified carbon sources of O. dioscoreae. Pseudomonas putida KT2440 is a well-described model organism capable of growth on many organic acids, and is derived from P. putida mt-2 originally isolated from soil [50, 51]. The metabolic requirements of strain KT2440 are well-documented and comprise those of O. dioscoreae [52–56]. First, we confirmed that a GFP-tagged derivative of P. putida KT2440 could utilize a majority of the carbon (22/29) and nitrogen sources (41/41) that support growth of O. dioscoreae using Biolog phenotype microarrays (Table S3). We next measured the respiration of O. dioscoreae and P. putida KT2440::gfp on five substrates likely to occur within the plant environment: citrate, L-malate, succinate, fumarate, and pyruvate. Cultures of P. putida KT2440::gfp showed significantly higher yield (Fig. 2A) and respiration rate (Fig. 2B) with L-malate, citrate, and succinate than O. dioscoreae. We detected no differences between the two strains regarding the oxidation of fumarate or pyruvate. Generally faster respiration rates and higher final yields indicate that P. putida is a strong competitor against O. dioscoreae for all potential substrates found in the leaf gland.

Figure 2.

Utilization of substrates found in leaf glands by P. putida and O. dioscoreae. Substrate oxidation by O. dioscoreae R-71412 and P. putida KT2440::gfp was monitored on ABY minimal medium supplemented with trisodium citrate (ABCY), fumarate, L-malate, sodium pyruvate or sodium succinate for 48 h. (A) Maximum Curve height (absorbance) and (B) Respiration rates were computed from curves modeled with R package Growthcurver. Data corresponding to P. putida are shown in dark grey, and O. dioscoreae in light grey. Statistical significance between each strain for each carbon source was calculated using a one-way ANOVA (significance levels: Ns P > .05; ^*P <= .05; ^**P <= .01; ^***P value <= .001; ^****P <= .0001).

To test whether both strains also compete for the same resources in a more complex environment, we monitored growth of O. dioscoreae R-71417 and P. putida KT2440::gfp in single and co-cultures. The carrying capacity of P. putida was significantly decreased in the presence of O. dioscoreae in LB medium, although by a mere 5% (Fig. 3A). However, the carrying capacity of O. dioscoreae R-71417 was severely impacted when co-cultured with P. putida KT2440::gfp, with a decrease of 62% in LB (Fig. 3B) and 77% in ABCY medium (Fig. S4A). In contrast, the carrying capacity of P. putida was significantly increased by 26% when grown in the presence of O. dioscoreae in ABCY medium (Fig. S4A). This correlates with a 42% decrease in growth rate (Fig. S4B), suggesting a metabolic shift towards more efficient resource utilization. Together, these results indicate that P. putida and O. dioscoreae compete for resources in liquid cultures. In addition, this suggests that P. putida KT2440::gfp adapts its growth strategy depending on the quality of the environment to consistently outgrow O. dioscoreae.

Figure 3.

In vitro competition between P. putida and O. dioscoreae. Cultures in LB medium in microtiter plates were inoculated with strains P. putida KT2440::gfp and mCherrry–tagged O. dioscoreae R-71417 together or separately. GFP- and mCherry-specific fluorescence was used to monitor the growth of each strain and derive growth characteristics. Values were normalized to the average of values in single culture. (A) Carrying capacity of strain P. putida KT2440::gfp inferred from GFP-specific fluorescence in single (left) or in co-culture with O. dioscoreae R-71417. (B) Carrying capacity of strain O. dioscoreae R-71417 inferred from mCherry-specific fluorescence in single (left) or co-culture with P. putida KT2440::gfp. Horizontal bars indicate P-values (Wilcoxon rank sum test). Data shown were collected from three independent experiments.

Effective colonization of D. sansibarensis leaf glands by P. putida

To test whether P. putida KT2440 could establish stable colonization in D. sansibarensis, we inoculated aposymbiotic plantlets with ~4 × 10⁵ CFU of P. putida KT2440::gfp. We detected P. putida KT2440::gfp within the glands of all plants colonized with the strain, but in none of the mock-inoculated controls. Epifluorescence imaging of leaf tip sections harvested 2 months after inoculation further confirmed that cells of P. putida KT2440::gfp occupied the lumen of the leaf glands (Fig. 4A–C). Leaf glands of plants inoculated with P. putida KT2440::gfp contained an average of 10⁹ CFU/g of fresh tissue (Fig. 4D), whereas macerates of lamina of surface-sterilized leaves did not yield colonies upon plating on nonselective TSA growth medium. This is similar to the colonization levels of wild-type O. dioscoreae in leaf glands of D. sansibarensis (Fig. 1).

Figure 4.

P. putida colonization of D. sansibarensis leaf glands. (A) Cross-section of a D. sansibarensis acumen colonized by P. putida KT2440::gfp after artificial inoculation. Merged image of light microscopy (B) and epifluorescence (C) observations. Scale bar = 500 μm. (D) Quantification of P. putida KT2440::gfp in leaf glands of D. sansibarensis after single inoculation or co-inoculation with O. dioscoreae R-71412 in equal ratios. Student’s t-test significance levels: Ns P > .05; ^*  P < .05; ^**P < .01; ^***P < .001; ^****P < .0001. (E) Competitive index of O. dioscoreae R-71412 in successive co-inoculations with P. putida KT2440::gfp. Aposymbiotic plants were inoculated first with either O. dioscoreae R-71412 or P. putida KT2440::gfp, or with the same strain twice. Competitive index was calculated as the log10 of the leaf gland colonization (CFU per g) by O. dioscoreae R-71412 over that of P. putida KT2440::gfp. Horizontal lines represent the median. Different letters indicate statistically significant differences between groups according to ANOVA with Tukey post-hoc test.

Competitive advantage of O. dioscoreae in the leaf gland

Because colonization of leaf glands did not seem to be specific to O. dioscoreae, we wondered whether better niche adaptation of O. dioscoreae, or competitiveness, might account for specificity in nature. To test this, we co-inoculated aposymbiotic plants with cell suspensions of P. putida KT2440::gfp alone or mixed with O. dioscoreae. The presence of O. dioscoreae reduced the average titer of P. putida KT2440::gfp in the leaf gland from 9.15×10¹⁰ to 1.67×10⁷ CFU/g (Fig. 4D). Although co-culture experiments in liquid media may allow for the detection of nutrient-based competition, they fail to account for several factors that could be relevant inside the host. Niche conditioning by O. dioscoreae might explain the lower rates of growth of commensal strains in the leaf gland in presence of O. dioscoreae (e.g. by depleting nutrients, saturating favored spatial niches, secreting anti-microbial compounds, or inducing specific plant defenses). Manipulating the order of inoculation between the strains would thus be expected to exacerbate the differences. To test this, we inoculated the two strains in varying order: O. dioscoreae first, P. putida KT2440::gfp first or both simultaneously. Co-inoculations always resulted in a significant competitive advantage of O. dioscoreae (Fig. 4E). However, we did not observe a significant effect of the order of arrival on the competitive index of either strain, indicating that niche conditioning or priority effects do not account for the majority of the competitive advantage of O. dioscoreae in the leaf gland.

Competitive advantage of O. dioscoreae is mediated by T6SSs

The genome of O. dioscoreae encodes two T6SSs that belong to two different phylogenetic subclasses, and possess different arsenals of effectors (Fig. S5). Both are expressed in planta [24], and we reasoned that anti-microbial effectors secreted by these T6SS may give a competitive advantage to O. dioscoreae. We tested contact-dependent growth inhibition of P. putida KT2440::gfp by O. dioscoreae in a microtiter plate assay. We observed a significant decrease of GFP-specific fluorescence by 29% in the presence of O. dioscoreae R-71412 (Fig. 5). This effect was reduced (6–11% reduction, P > .05) in the presence of strains harboring mutations in either or both of the T6SS clpV genes, although O. dioscoreae mutant strains had similar growth characteristics (Fig. S6). Complementation of the single and double mutant with copies of clpV1 or clpV2 in trans restored wild-type levels of competition against P. putida (Fig. 5).

Figure 5.

Contact-dependent competition between O. dioscoreae and P. putida. Fluorescence intensity of cultures of P. putida KT2440::gfp after 4 h of growth alone or in contact with O. dioscoreae strains. O. dioscoreae T6SS mutants ∆clpV1, ∆clpV2, and ∆clpV1∆clpV2 and complemented strains ∆clpV1/clpV1+, ∆clpV2/clpV2+, ∆clpV1∆clpV2/clpV1+, and ∆clpV1∆clpV2/clpV2+ were tested. Different letters indicate statistically significant differences between groups according to ANOVA with Tukey post-hoc test (significance threshold P = 0.05). Results shown here are from one of three independent experiments showing similar results.

To confirm that contact-dependent killing is responsible for the reduction in GFP-specific fluorescence in the microtiter plate assay and to gain more quantitative insights, we counted the number of CFU by dilution plating of both P. putida KT2440::gfp and O. dioscoreae strains after direct contact for 2, 4 or 6 h. We observed a reduction by an average of 6429-fold in the number of P. putida KT2440::gfp colonies upon co-culture with O. dioscoreae R-71412, but not with the ΔclpV1ΔclpV2 mutant (1.2-fold decrease, Fig. S7). The effect was also pronounced upon contact with the ΔclpV1 strain (9520-fold average reduction), but not with the ΔclpV2 strain (nine-fold reduction, P > .05). This indicates that the contribution of T6SS-2 to the killing of P. putida is larger than that of T6SS-1. However, the significant reduction of P. putida CFU counts in competition with the ΔclpV1ΔclpV2 vs. the ΔclpV2 strain (7.58-fold, P < .05) indicates that, although minute, the contribution of T6SS-1 is still detectable in this assay.

We performed a co-inoculation experiment to test the contribution of T6SS to the competitive advantage of O. dioscoreae in planta. The T6SSs of O. dioscoreae did not affect the ability to colonize the leaf gland in single inoculations (Fig. S8). The presence of the wild-type strain of O. dioscoreae negatively impacted the colonization of P. putida KT2440::gfp by three orders of magnitude (1703-fold fewer CFU on average, P = 1.44x 10⁻³) (Fig. 6). In contrast, this negative effect was much alleviated when co-inoculating plants with the O. dioscoreae ΔclpV1ΔclpV2 strain, with only a 2.39-fold reduction in the number of P. putida CFU (P = .24). Microscopy observations of leaf glands by epifluorescence further confirmed that wild-type O. dioscoreae R-71412 effectively blocked the colonization of leaf glands by P. putida KT2440::gfp, but O. dioscoreae ΔclpV1ΔclpV2 did not (Fig. S9). Colonization of D. sansibarensis by Stenotrophomonas sp. R-67087 was also attenuated (23-fold, P = .04) when co-inoculated with O. dioscoreae R-71412, but not with O. dioscoreae ΔclpV1ΔclpV2, confirming the role of T6SS in antagonism against diverse taxa capable of colonizing the leaf gland (Fig. S10).

Figure 6.

In planta competition between P. putida and strains of O. dioscoreae affected in type VI secretion. Aposymbiotic plants were inoculated with P. putida KT2440::gfp alone or in 1:1 ratio with O. dioscoreae strain R-71412, or strain ∆clpV1∆clpV2. Freshly grown acumens were weighed and macerated. Serial dilutions were plated on selective media to quantify bacterial load. Horizontal bars indicate P-values between groups according to Dunn’s test. P-values adjusted for multiple testing with the Bonferroni method are shown, with unadjusted P-values in parentheses. Results shown here are from one of two independent experiments showing similar results.

Evidence of species-selective transmission

We tested whether P. putida could also be transmitted to the next generation of plants. To test this, we inoculated plants with P. putida KT2440::gfp. We collected bulbils produced by the plants at the end of the growing season and let them germinate after a storage period of 10 to 16 months. The plants produced on average 5.9 bulbils per plant, with a bulbil germination rate of 9.1% (Table 3). None of the daughter plants contained detectable P. putida in the leaf glands. As a control, we also inoculated plants with a 1:1 cell mixture of P. putida KT2440::gfp and O. dioscoreae R-71412. These produced 4.4 bulbils per plant on average, for a bulbil germination rate of 15.7%. Of the 109 daughter leaf glands tested only one contained both P. putida KT2440::gfp and O. dioscoreae (0.9% of the leaf glands), whereas we detected O. dioscoreae R-71412 in 69 samples (63.3%). The plants inoculated with P. putida KT2440::gfp and O. dioscoreae ΔclpV1ΔclpV2 produced 11.86 bulbils by plant on average. None of the 26 leaf glands of the progeny contained P. putida but 18 leaf glands contained O. dioscoreae (69.2%). This confirmed that the T6SS is not relevant for the vertical transmission, even in presence of nonsymbiotic bacteria.

Table 3.

Detection of bacteria in offspring of inoculated Dioscorea sansibarensis.

			Bacteria detected in leaf glands of offspring
Inoculum	Average number of bulbils produced by plant	Sprouting efficacy (%)	P. putida only	O. dioscoreae only	O. dioscoreae and P. putida	No bacteria
P. putida	5.9	9.1	0	0	0	35
P. putida + O. dioscoreae R-71412	4.4	15.7	0	68 (62.3%)	1 (0.9%)	40 (36.7%)
P. putida + O. dioscoreae ΔclpV1ΔclpV2	5.6	11.7	0	18 (69.2%)	0	8 (30.8%)

Discussion

Heritable leaf symbioses are highly specific in nature: one plant species associates with only one bacterial species. At least in the case of the D. sansibarensis leaf symbiosis, this specificity is not due to an obligate interaction: aposymbiotic plants develop normally in a controlled environment, and bacteria can grow axenically as well [18]. Several recent studies have shown that free-living bacteria could effectively replace vertically transmitted insect symbionts, under some conditions. For example, the obligate symbiont Sodalis pierantonius of the grain weevil Sitophilus zeamais could be replaced by a free-living strain of Sodalis praecaptivus engineered to secrete aromatic amino acids [57]. Similarly, a free-living E. coli strain was successfully evolved in the laboratory and provided an effective replacement for the vertically transmitted Pantoea symbiont of Plautia stali stinkbugs upon acquisition of a mutation in global regulators controlling carbon utilization pathways [58]. In both of these examples, stable symbiont replacement occurred with closely related strains (same genus or family) after some genetic modification. Here, we show that a vertically transmitted plant symbiosis is promiscuous, with taxonomically diverse strains capable of stable colonization. This is in stark contrast to the 100% specificity for O. dioscoreae observed in field samples [19, 24].

Using metabolomics data from D. sansibarensis leaf glands and metabolic profiling of O. dioscoreae, we established the conditions likely encountered by the bacteria inside the plant. Conditions in the D. sansibarensis leaf glands are micro-oxic (De Meyer et al. 2019) O. dioscoreae grows best at neutral to slightly acidic pH. Additionally, O. dioscoreae putative high-affinity iron-acquisition systems are overexpressed in planta compared to axenic cultures [24]. This is perhaps indicative of depletion of iron in the leaf gland, which may be associated with plant immunity [59]. Furthermore, genes involved in oxidative stress responses (e.g. putative catalase and peroxidases) are overexpressed by the bacteria in planta and functions linked to resistance to oxidative stress are highly conserved in the core genomes of leaf symbionts [24, 60]. TCA cycle substrates could be commonly used as major carbon sources by leaf endosymbionts. Taken together, these data indicate that the conditions in the leaf glands may contribute to selecting bacterial specialists adapted for utilization of acids and tolerance to oxidative stress, but may not be otherwise particularly stringent.

P. putida KT2440 presents an overlapping metabolic profile to O. dioscoreae and is able to colonize the leaf glands of D. sansibarensis to high densities. Pseudomonas putida KT2440 derives from a soil isolate and is unlikely to be pre-adapted to an endophytic lifestyle. This suggests that once the plant is deprived of its symbiont the leaf gland becomes open to colonization by other bacterial strains. Moreover, P. putida is metabolically well adapted to inter-species competition [61, 62]. Our results show that P. putida KT2440::gfp consistently outcompeted O. dioscoreae in liquid cultures, adopting seemingly distinct strategies to maximize resource utilization. This is illustrated by the fact that upon entering co-culturing with O. dioscoreae in rich medium (LB), which offers a broad metabolic niche, P. putida increases growth rate at the cost of a lower carrying capacity (Fig. S4C). In contrast, ABCY medium offers a narrower metabolic niche, with citrate as the only abundant carbon and energy source. Upon competition with O. dioscoreae, the growth rate of P. putida in ABCY medium slows down, perhaps allowing for greater efficiency in resource utilization and maximizing carrying capacity (Fig. S4A and B). Despite this adaptive metabolic potential, P. putida is unable to outcompete O. dioscoreae in plant co-inoculation assays (Fig. 4).

In D. sansibarensis glands, bacteria are present to high titer up to 1×10¹⁰ CFU/g of tissue, at least an order of magnitude higher than what is achievable in liquid cultures. High cell-density and proximity may therefore allow for other mechanisms of competition, for example for contact-dependent competition. Experiments on solid media indeed showed that O. dioscoreae kills P. putida KT2440::gfp cells in a contact-dependent manner, and that this antagonistic activity is mediated by two T6SS gene clusters encoded in the genome of O. dioscoreae. Both T6SS-1 and T6SS-2 of O. dioscoreae LMG29303^T are overexpressed in planta and are conserved in all O. dioscoreae genomes, although the repertoire of predicted effectors vary between strains [19, 24]. The anti-microbial activity of T6SS and associated effectors is well-described, and contrasting T6SS effector sets may reflect varying pathogen or competitor pressures in natural populations of O. dioscoreae (reviewed in [35, 63]). Both O. dioscoreae T6SS clusters encode effector proteins with putative anti-microbial activities. The T6SS-1 cluster encodes a putative colicin (ODI_R3996) and a VgrG-family protein with a C-terminal Tle1 phospholipase domain (ODI_R3993). T6SS-2 encodes effectors containing Tle4 (ODI_R794), M15 metallopeptidase (ODI_R0790), and Tle1 (ODI_R0793) domains. Aside from a role in protecting the symbiotic niche, it is unclear whether O. dioscoreae also provides protection against bacterial or fungal pathogens, as has been demonstrated in other systems [64, 65]. Complete T6SS were only detected in the genomes of leaf symbionts of Fadogia homblei and Vangueria pygmaea, but not in other leaf symbionts of the Caballeronia clade [12]. How specificity is maintained in leaf symbioses in the absence of a T6SS remains unknown, but other secreted toxins might play a role.

Alone, antagonistic microbe–microbe interactions are unlikely to fully explain the absolute specificity between D. sansibarensis and O. dioscoreae observed in the wild. Type III secretion systems (T3SS) are often involved in associations with eukaryotic hosts through delivery of effectors into host cells [66]. These effectors may modulate immune pathways and play a key role in symbiotic partner recognition and specificity [67]. Genomes or metagenome-assembled genomes (MAGs) of O. dioscoreae isolated from leaf glands of D. sansibarensis lack a conserved T3SS, suggesting that this pathway is not essential for the association [19]. Although the genome of O. dioscoreae LMG 29303^T (the parental strain for all strains used in this study) does encode a minimal T3SS gene cluster, previous studies did not find evidence of expression in vitro or in planta [24]. The fact that P. putida readily colonized D. sansibarensis in the absence of a functional T3SS is further evidence that secretion of T3 effectors is not required for colonization. The absence of a phenotype in aposymbiotic plants suggests that symbiotic functions may be important for fitness in response to environmental or herbivory pressures. We are currently assessing the phenotypes of plants under an array of conditions. If the symbiosis is indeed essential for survival under natural conditions, vertical transmission, and partner-fidelity feedback may also contribute to specificity by culling plants with ineffective symbionts from the population [68, 69].

Despite D. sansibarensis being amenable to artificial inoculations, transmission of symbionts to bulbils appears more stringent. Transmission appears to be imperfect in our experiments, with symbiotic bacteria detected in only 62% of bulbils. This indicates that barriers to transmission exist. These barriers appear somewhat selective, with plants inoculated with P. putida alone unable to pass on the bacterium to their offspring. Only when we co-inoculated plants with O. dioscoreae did we detect P. putida in the offspring of plants, albeit at anecdotal frequencies. It is unclear what the mechanisms enforcing specificity of transmission might be, but we speculate that the ability to withstand a quiescent phase of several months before bulbils germinate might be an important factor in transmission success.

This study emphasizes the complexity underlying the regulation of specificity in the Dioscorea-Orrella symbiosis. The absence of a strong phenotype of aposymbiotic plants, the permissiveness of the plant for a broad range of bacteria in artificial inoculations, and the fact that specificity seems enforced (at least partially) by microbe–microbe interactions suggest that the barriers to evolving vertically transmitted plant symbioses may be unexpectedly low. This plasticity may explain why leaf symbiosis seems to have evolved independently several times in distinct plant lineages. What adaptations underlie the evolution of vertical transmission in these plant-bacteria associations remain to be discovered.

Conflicts of interest

None declared.

Funding

L.N. and A.C. acknowledge support by the French Laboratory of Excellence project “TULIP” (ANR-10-LABX-41; ANR-11-IDEX-0002-02). This study is set within the framework of the “École Universitaire de Recherche (EUR)” TULIP-GS (ANR-18-EURE-0019). AC also wishes to acknowledge funding from the French National Research Agency under grant agreements ANR-22-CE92–0042 and ANR-23-CE11–0015, and the Région Occitanie through the “RAMSY” project. Financial support was received from the French National Infrastructure for Metabolomics and Fluxomics, Grant MetaboHUB-ANR-11-INBS-0010.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary information files.