Specialization of independently acquired flagellar FliC proteins in plant-associated Sphingomonas balances swimming and immunogenicity

Summary

Plants monitor their environment for microbial invaders using pattern-recognition receptors that detect microbe-associated molecular patterns (MAMPs). Flagellin, the main component of bacterial flagellum, contains the flg22 epitope recognized by the plant immune receptor FLS2. Immune recognition can create an evolutionary conflict, requiring bacteria to balance flagellar function and immune evasion. Here, we show that plant-associated Sphingomonads resolve this constraint by partitioning two flagellar functions, motility and colonization, across two divergent and independently acquired flagellin genes. Comparative genomics revealed widespread coexistence of FliC proteins expressing either an immunogenic variant (FliC-H) or a non-immunogenic variant (FliC-L). The non-immunogenic FliC-L is necessary and sufficient for full directional swimming, whereas FliC-H is dispensable for swimming, but sufficient for full attachment and colonization. Flagellin expression patterns mirror these functions. Thus, FLS2 recognizes the flagellar variant required for colonization rather than motility, potentially restricting colonizing bacteria from entering internal leaf and root tissues.

Introduction

Plants coexist with diverse microbial communities that inhabit their roots and shoots (1, 2). These microorganisms range from harmful pathogens to beneficial commensals, and the lines between these lifestyles are often blurred by community contexts. Pathogens infect plant tissues to extract nutrients and proliferate, often causing damage to the host in the process. In contrast, beneficial microbes also acquire nutrients and proliferate to survive, but their growth is typically more tightly restricted by the host and constrained to specific tissues or compartments (3–5). Plants can enable microbial colonization but potentially limit beneficial microbial colonization to specific organs and compartments within them. Plants must also maintain the capacity to activate full immune responses to combat pathogens.

To monitor their environment, plants utilize plasma membrane localized extracellular pattern-recognition receptors (PRRs) that recognize conserved microbial epitopes, referred to as microbe-associated molecular patterns (MAMPs), that are typically important for microbial function (6, 7). The recognition of these microbial patterns activates MAMP-triggered immunity (MTI) (8, 9), a fundamental defense strategy that restricts pathogen invasion, but may also limit colonization by beneficial and commensal microbes, thereby fine tuning those interactions. Thus, host-associated microbial communities must reach molecular and ecological détente to ensure host health and productivity while gating the microbial community for potentially beneficial functions and against pathogens (10).

The most thoroughly characterized plant associated MAMP is flg22, a 22-amino acid peptide derived from bacterial flagellin encoded by the FliC gene (6, 11–14). FliC is the major structural protein that forms the bacterial flagellum (15). The FliC protein polymerizes into a helical filament that functions in bacterial motility and plays a crucial role in surface attachment, alginate acquisition, and environmental sensing (16–21). The flg22 peptide represents the minimal active epitope required to trigger immune signaling in Arabidopsis thaliana and many other angiosperms, but longer overlapping peptides containing the flg22 core sequence also retain immunogenicity (22, 23). Thus, immunogenicity depends primarily on the integrity and accessibility of the central flg22 region, rather than strict peptide length. Because flg22 is embedded within the intact flagellin polymer, it is not freely accessible to the plant immune system. Instead, immunogenic peptides are generated through extracellular processing of FliC in the plant apoplast (24). This involves plant glycosidases and proteases that remove glycan shields and trim neighboring peptide regions to expose or release the active MAMP, some even cleave within the flg22 epitope to relax immunity (24–26). The efficiency of active epitope release varies across microbial taxa depending on the FliC sequence and its post-translational modifications (13). The released immunogenic flg22 is then detected by the canonical plant receptor kinase FLS2 (FLAGELLIN-SENSING 2) in A. thaliana and other plant species, triggering a cascade of immune responses, including the production of reactive oxygen species (ROS), MAP kinase activation, and the induction of MTI transcriptional responses (6, 23). It remains unclear what part of the MTI response actually limits bacterial proliferation.

While the overall structure of flagellin is conserved, the flg22 sequence is polymorphic, and genomes from individual bacterial isolates encode diverse versions of flg22 with a range of immunogenic outcomes (11, 12, 14, 27). For example, bacteria like Pseudomonas aeruginosa express highly immunogenic flg22 peptides that strongly activate FLS2, whereas several commensal bacteria, including Rhizobium, produce variants that largely evade immune detection (11, 12, 15, 27). However, immune evasion has a cost: in the canonical flg22 variant from P. aeruginosa, missense mutations that evade receptor detection are also impaired for swimming (12). That trade-off is a classic case of antagonistic pleiotropy and supports a simple model where the plant immune system has evolved to target the precise amino acids and FliC structures that provide the flagellar motility function.

The flg22 antagonistic pleiotropy is further shaped by spatial selection pressures within the plant. FLS2 is not evenly distributed in the plant but rather localizes to the plasma membrane of a variety of cell types that gate access to the leaf apoplast, like stomata, hydathodes and mesophyll cells, as well as sites of lateral root emergence. FLS2 cell-type-specific expression thus positions immune perception at key microbial entry points into leaf and root tissues (28, 29). That spatial organization implies that the intensity of immune surveillance might vary across plant surfaces and organs. As a result, bacteria colonizing distinct microhabitats could experience different selective pressures to retain or modify their flagellar components. This spatial heterogeneity likely influences the evolutionary balance between flagellar functions and immune evasion, setting the stage for evolutionary innovations that might drive functional specialization and diversification of flagellar genes, as we describe below.

In this work, we surveyed a published dataset of experimentally measured immune activation of flg22 sequences from 627 plant associated bacterial genomes (11). We identified rare isolates that encode diverging FliC genes. Specifically, we identified six isolates of the ubiquitously plant-associated beneficial commensal genus Sphingomonas (30) that encode two independently acquired and divergent FliC-containing gene clusters. These divergent FliC genes encode either an immunogenic or a non-immunogenic flg22 epitope. We determined that this occurrence of duplicate FliC arrangement is widespread across the Sphingomonas sp. genomes. We focused on a single experimentally amenable isolate, Sphingomonas sp. MF220, and deleted each of the FliC genes and tested the functions of the remaining one. We found that the FliC protein carrying the non-immunogenic flg22 (FliC-L) is necessary for full swimming capacity, while the copy carrying the immunogenic flg22 (FliC-H) is sufficient for full attachment to the root and colonization of both roots and shoot. These results indicate that the plant immune receptor FLS2 targets Sphingomonas sp. flagella specialized in attachment and colonization, and, at least in this genus, not the classical flagellar function of motility. These results reinforce the concept that FLS2-dependent immune surveillance does not interfere with surface colonization of leaves or roots by Sphingomonas sp. MF220 and strains with similar FliC architectures but can influence their subsequent migration into intimate contact with both the leaf and the root. This restriction may contribute to partitioning of the bacteria into specific niches that allow them to thrive (and provide plant-beneficial functions) minimizing immune responses.

Results

Arabidopsis-Colonizing Commensal Bacterial Genomes Harbor Both Immunogenic and Non-Immunogenic Flagellin Gene Clusters

We previously presented a detailed analysis of immunogenic and non-immunogenic flg22 peptides derived from diverse FliC proteins (11, 12). We noted that some strains encoded multiple FliC genes. We reanalyzed this dataset that quantified the plant reactive oxygen species (ROS) burst induced by 97 synthetic flg22 epitopes from 627 Arabidopsis-colonizing bacterial genomes (11) to search for isolates carrying two diverging flg22 peptides: one that elicits high immune response from the plant and another that does not (Fig. 1A). We calculated the difference in immunogenicity between the most and least immunogenic flg22 epitopes for each bacterial isolate possessing multiple copies of the FliC gene (Fig. 1B). Most bacteria with multiple FliC genes, typically generated via flagellar gene cluster duplication, harbored copies with similar flg22 immunogenicity. However, eight isolates displayed flg22 copies that triggered significantly divergent ROS levels (SI Appendix, Table S1). These eight isolates contained FliC genes encoding both an immunogenic flg22 epitope (FliC-H; high ROS burst, red throughout) and a non-immunogenic flg22 epitope (FliC-L; low ROS burst, blue throughout) (Fig. 1B and 1C; SI Appendix, Fig S1). Six of these eight isolates belong to the genus Sphingomonas, a prevalent and abundant plant colonizer that expresses, among other traits, the ability to suppress pathogen infection via weak induction of host immune response (30–32).

Figure 1. Arabidopsis-associated Sphingomonas encode independently evolved and acquired flagella-encoding genomic islands.

(A) An illustration of plant-colonizing bacteria. To study the function of flagella that elicit flg22 epitope-induced immune responses (11), we defined strains that encoded two divergent flagella: (1) those expressing immunogenic flg22 epitopes (FliC-H, red, High, inducing ROS burst; right); (2) those expressing non-immunogenic flg22 epitopes (FliC-L, blue, Low, do not induce ROS burst; left). (B) Differential plant ROS burst between the most (max) and least (min) immunogenic flg22 variants in each Arabidopsis-associated bacterial isolate that encodes multiple FliC genes. Eight isolates carry FliC genes encoding functionally diverse flg22 epitopes (purple). (C) The host ROS burst elicited by the 16 flg22 peptides encoded by the eight isolates defined in B. ROS burst is presented as Z-score values (11). (D) Synteny plot of the genomic islands of the immunogenic (FliC-H, x-axis) and the non-immunogenic (FliC-L, y-axis) flagellar genes in Sphingomonas MF220, the reference strain for this study. Genes are colored on the axes by function. Black squares represent flagellum-related orthologous genes in the two islands. (E) All-by-all synteny analysis of FliC-anchored islands in six Sphingomonas isolates that encode FliC genes expressing both immunogenic and non-immunogenic flg22 epitopes. Black squares represent flagellum-related orthologous genes across the two islands. Background color represents the island-similarity index, accounting for shared gene content and gene order (Methods).

We performed a synteny analysis on the genomic islands surrounding the FliC genes from the six Sphingomonas sp. strains to determine whether the two copies resulted from duplication and diversification or independent acquisition from distinct sources (Methods). We found that while the genomic islands flanking the immunogenic FliC-H and non-immunogenic FliC-L genes in Sphingomonas isolates contain similar flagellar biosynthetic genes, they share little similarity in gene order (Fig. 1D). Although the two islands share similar gene content (Jaccard index Ո/Ս=28/45=0.62), their distinct gene order and chromosomal locations suggest independent evolution and acquisition events. Furthermore, we observed that while the genomic islands flanking FliC-H are highly conserved across isolates, containing 40–42 genes, those flanking FliC-L are highly divergent, ranging from degenerated islands of six genes to full islands of 63 genes and defining three distinct islands (Fig. 1E; SI Appendix, Fig S2). Notably, the FliC-L Islands often include chemotactic machinery that is absent in all FliC-H islands (Fig. 1D; SI Appendix, Fig S2). The distinct gene order and accessory gene content in the genomic islands suggest that the immunogenic and non-immunogenic flagella have followed independent evolutionary trajectories and were likely acquired and maintained under distinct selective pressures.

Plant-Associated Sphingomonas sp. Genomes Often Encode FliC Proteins Containing Functionally Divergent flg22 Epitopes

To characterize the prevalence of Sphingomonas sp. encoding FliC proteins containing functionally divergent flg22 epitopes, we analyzed a collection of 400 genome-sequenced Sphingomonas isolates, predominantly from geographically diverse Arabidopsis plants (30). We identified all FliC genes and the corresponding encoded flg22 epitopes (Methods). Based on sequence similarity to previously tested peptides, we defined 416 immunogenic (flg22-H, red throughout) and 143 non-immunogenic (flg22-L, blue throughout) epitopes (Fig. 2A, SI Appendix, Fig S3). Our analysis revealed that non-immunogenic flg22 epitopes exhibited greater sequence diversity than immunogenic flg22 epitopes (Fig. 2B). The sequence-diverse amino acid positions altered in the non-immunogenic flg22 epitopes include residues required for interaction with the FLS2 receptor complex (33) (Fig. 2B).

Figure 2. Prevalence and distribution of functionally divergent flg22 epitopes among wild Sphingomonas isolates.

(A) Maximum-likelihood phylogenetic tree of 559 flg22 epitopes derived from a collection of 400 Sphingomonas isolate genomes (30). The epitopes cluster into two distinct clades: those with high sequence similarity to the non-immunogenic Sphingomonas flg22-L epitope (blue) and those with high similarity to the immunogenic flg22-H epitope (red). (B) Consensus sequence motifs for the flg22-H and flg22-L epitope clades. The canonical flg22 sequence from P. aeruginosa appears on the top for reference (asterisks denote amino acids that contact the FLS2 receptor). Note the sequence divergence within the non-immunogenic flg22-L clade compared to flg22-H, particularly FLS2 contact residues. (C) Multilocus sequence typing (MLST) phylogenetic tree depicting the evolutionary relationships of all Sphingomonas isolates included in this analysis (MLST based on 48 conserved genes, Methods). The inner circle indicates the presence of a FliC-embedded immunogenic flg22-H epitope (red). The middle circle indicates the presence of a FliC-embedded non-immunogenic flg22-L (blue). The outer circle denotes the isolation source: A. thaliana (green) or other plant hosts (gray). Isolates harboring both flg22-L and flg22-H epitopes are prevalent across multiple clades and are found in bacteria isolated from both A. thaliana and other plant hosts.

We projected the immunogenic flg22-H and non-immunogenic flg22-L epitopes onto a multilocus sequence typing (MLST) phylogenetic tree of all isolates to characterize the evolutionary trajectories of the genomes encoding functionally divergent flg22 epitopes (Fig. 2C; Methods). The MLST tree defined the genomic diversity within this Sphingomonas genome collection and showed that most isolates possessed FliC-H genes. Isolates carrying both FliC-H and FliC-L genes were broadly distributed across different clades of the tree. A large clade (bottom-left quarter of the tree) contained many isolates with both genes; conversely, other distributed clades also contained isolates with both FliC-H and FliC-L. These results demonstrate that strains carrying FliC genes encoding divergent flg22 epitopes are prevalent among environmental isolates of Sphingomonas. As a whole, the divergent flg22-L epitope, the variability of the FliC-L island, and the sparse presence-absence of FliC-L across the Sphingomonas lineages suggest that FliC-L genes and islands were acquired later than FliC-H and multiple times from different sources. The widespread occurrence of isolates carrying both FliC-H and FliC-L across different clades and derived from different plant hosts suggests that this dual-flagellum strategy is a conserved feature across diverse Sphingomonas lineages, potentially conferring adaptive advantages in heterogeneous environments.

The Two Diverging Flagella are Differentially Regulated and Display Distinct Morphologies

Given that the two FliC genomic islands likely originated from different sources, we tested whether FliC-H and FliC-L expression were co-regulated. We tested the growth of several isolates presented in Figure 2. Most did not grow well in either liquid LB or LB agar plates. Very few exhibited dispersed planktonic growth. We therefore focused on the fast-growing plant-associated isolate Sphingomonas MF220A (hereafter MF220). We inoculated wild-type MF220 onto 0.3% swimming agar and sampled bacteria from both the slow-swimming center of a swimming population and the fast-swimming front after three days of incubation. We measured the expression of each FliC gene in swimming bacteria (Methods). We found that while cells in the center of a swimming population express similar levels of FliC-H and FliC-L, bacteria at the front of the swimming halo exhibited distinctly higher expression of the non-immunogenic FliC-L (ratio of FliC-L/FliC-H = 189; Fig. 3A; SI Appendix, Fig. S4).

Figure 3. FliC-L, encoding a non-immunogenic flg22 epitope, is required for full swimming motility.

(A) Expression ratio of the two FliC genes in MF220 from a bacterial colony on a normal 1.5% agar plate and from 0.3% swimming agar plate populations. Fast-swimming bacteria at the front of a swimming plate express 189-fold more FliC-L than FliC-H while non-swimming bacteria in a colony and from slow-swimming bacteria in the center of a swimming population express both genes to similar levels (n = 9 biologically independent samples from 3 independent experiments, denoted by shape). Boxplots show the median (black line), 1^st and 3^rd quartiles (box), and one standard deviation (whiskers) of treatment. (B) Illustration of the FliC gene deletions in Sphingomonas MF220. The two flagellar genes, residing in different loci on the chromosome, were deleted to generate four strains: MF220_WT (gray, wild-type, carrying both FliC genes); MF220_L (blue, carrying only FliC-L); MF220_H (red, carrying only FliC-H); and DD (purple, a double deletion lacking both FliC genes). (C) Scanning electron microscopy (SEM) of cells grown in rich media expressing either of two distinct morphologies: a straight, spear-like flagellum (observed in wild-type MF220_WT and in the MF220_L derivative) or a wobbly, twisted flagellum (observed in wild-type MF220_WT and in the MF220_H derivative). As expected, the double deletion (MF220_DD) derivative exhibits no flagellum. (D) Quantification of swimming motility for MF220 derivatives on semi-solid agar. The radius of swimming populations is shown over time (Top panel). A representative final image with the identified population area is presented (bottom panel). The MF220_L derivative (blue), encoding only the non-immunogenic FliC-L gene, swims as effectively as the wild-type (MF220_WT, gray), whereas derivatives lacking FliC-L (MF220_H, red; and MF220_DD, purple) exhibit severely compromised swimming.

We generated three MF220 derivatives by deleting either or both FliC genes to investigate their functions (Fig. 3B): the MF220_H derivative (red) is a deletion of FliC-L and thus encodes only FliC-H and thus an immunogenic flg22; the MF220_L derivative (blue) is a deletion of FliC-H and encodes only FliC-L and thus a non-immunogenic flg22; the double deletion derivative (MF220_DD, purple), lacks both FliC genes (Methods). We identified two distinct wildtype MF220_WT polar flagellar morphologies under a scanning electron microscope (SEM) that did not obviously co-occur in the same cell: the majority of the cells carry a straight, spear like flagellum (Fig 3C, MF220_WT-1; SI Appendix Fig. S5) while others carry a wobbly and twisted flagellum (Fig. 3C, MF220_WT-2; SI Appendix Fig. S5). As expected, the double FliC deletion derivative MF220_DD expresses no flagella, but rather only the initial hook structure (Fig. 3C; SI Appendix Fig. S5). Looking at the single mutant, we found that while most MF220_L bacteria carry the straight, spear-like flagellum, most MF220_H bacteria carried the twisted and wobbly flagellum (Fig. 3C; SI Appendix Fig. S5). These distinct flagellar morphologies may underline functional differences in propulsion efficiency navigation and/or host interaction, further supporting a potential division of labor between FliC-H and FliC-L in MF220.

FliC-L is Required for Directional Swimming.

We next tested the role of the two flagellar types in directional swimming along a nutrient gradient using MF220_WT and the three FliC deletion derivatives. We quantified the radius of each directionally swimming population over 72 hours (Fig. 3D; Methods; Video S1; SI Appendix, Fig. S6). We found that while the MF220_L derivative (blue) swam as effectively as the wild-type MF220_WT (black), the MF220_H derivative (red) exhibited severely compromised swimming. The double deletion MF220_DD (purple) swam more poorly than the MF220_H derivative (Fig. 3D). The sufficiency of the FliC-L flagellum for full directional swimming is consistent with our expression data (Fig. 3A). These results suggest that the immunogenic FliC-H flagellum is not required for directional swimming and cannot substitute FliC-L, likely because it fails to effectively engage the chemotactic machinery optimized for FliC-L mediated motility (SI Appendix, Fig. S2).

FliC-H, Encoding an Immunogenic flg22 Epitope, is Required for Full Colonization and Attachment to A. thaliana

We next investigated the contribution of each of the two flagella types to plant colonization. We began by testing the first step in plant colonization, attachment. We performed short-term attachment assays (washing to eliminate loosely adherent bacteria and then grinding tissue to free tightly adherent and endophytic compartment bacteria for cfu counting; Methods) using the four MF220 derivatives, quantifying bacterial load after four hours of attachment to A. thaliana shoots and roots in water (Fig. 4A; Methods). In the shoot (Fig. 4A, top), the double-deletion derivative (MF220_DD, purple) exhibited compromised attachment. By contrast, either of the two FliC-containing flagellar types was sufficient to provide wild type levels of attachment. That result suggests that while a flagellum is required for optimal attachment, the two flagellar types are redundant for shoot attachment. On roots (Fig. 4A, bottom), on the other hand, both the double-deletion derivative (MF220_DD, purple) and MF220_L (blue) also exhibited compromised levels of attachment. Although our data do not resolve the individual steps of surface attachment, these results demonstrate that FliC-H (red) is sufficient for full attachment to the root, while FliC-L is not.

Figure 4. The immunogenic FliC-H is essential for robust colonization and attachment to Arabidopsis thaliana roots.

(A) Quantification of short-term attachment of MF220 derivatives at four hours post-colonization to A. thaliana shoots (top) and roots (bottom) in liquid. MF220_DD (purple) bacteria were compromised in short-term attachment to A. thaliana shoots compared to MF220_WT (gray). By contrast, both the MF220_H (red) and MF220_L (blue) derivatives exhibited attachment levels similar to MF220_WT. In the root, MF220_DD (purple) bacteria were again compromised in short-term attachment. Additionally, MF220_L also exhibited compromised attachment (linear model, P < 0.05, n = 12 biologically independent samples from 3 independent experiments, denoted by shape). (B) Quantification of MF220 derivatives colonizing the shoots (top, SI Appendix, Fig S6) and roots (bottom) of A. thaliana for 12 days. In both plant organs, the double deletion MF220_DD (purple) and the MF220_L (blue) derivatives exhibited significantly reduced colonization compared to the wild-type MF220_WT bacteria (gray). Conversely, the MF220_H derivative (red) colonized at MF220_WT levels (linear model, P < 0.05, n = 15 biologically independent samples from 4 independent experiments, denoted by shape). (C) Expression ratio of the FliC genes in Sphingomonas MF220 bacteria colonizing the shoots (top) and roots (bottom) of A. thaliana for 12 days analyzed by RT-qPCR. Plant-associated bacteria (right, comprising strongly attached bacteria and endophytes) exhibit significant shift in the FliC-L to FliC-H expression ratio towards FliC-H compared to the non-attached bacteria in the phyllosphere (top, 4.9 fold) and rhizosphere (bottom, 2.8 fold)(left; linear model, P < 0.05, n = 9 biologically independent samples from 3 independent experiments, denoted by shape). In all panels, boxplots show the median (black line), 1^st and 3^rd quartiles (box), and one standard deviation (whiskers) of treatment.

To assess the roles of FliC genes during long-term plant colonization, we tested the colonization of the four MF220 derivatives for 12 days in two systems. To test shoot colonization, we spray-inoculated A. thaliana leaves on plants grown in potting soil with each of the four derivatives separately (Methods). We quantified shoot colonization over 12 days by grinding washed A. thaliana leaves (Methods) and plating the colonizing bacteria on MF220-selective Rifampicin plates. Thus, this assay captures tightly adherent surface bacteria and those that have colonized the leaf apoplast. We found that the abundance of all four derivatives decreases over time (SI Appendix, Fig S7). Nevertheless, MF220_WT (black) and MF220_H (red) persist, and hence are more abundant than MF220_DD (purple) and MF220_L (blue) after 12 days (Fig. 4B, top). To test root colonization, we used a gnotobiotic system. We inoculated 0.5x MS plates independently with each MF220 derivative before transferring seven-day-old A. thaliana seedlings onto the pre-inoculated plates. After 12 additional days, we quantified the root-associated bacteria (operationally defined as tightly attached bacteria at the root rhizoplane and the root endophytes; Methods). We found that both the double deletion MF220_DD (purple) and the MF220_L derivative (blue) exhibited compromised root colonization (Fig. 4B, bottom). In contrast, the MF220_H derivative (red) colonized roots at levels comparable to wild-type MF220_WT (black). These results highlight the specific role of FliC-H in the colonization of both the roots and the shoot of A. thaliana. We additionally determined that these phenotypes are independent of the flg22 receptor FLS2 (SI Appendix, Fig S8).

The requirement of FliC-H for colonization led us to test flagellar expression during colonization (Methods). We compared Sphingomonas MF220_WT bacteria loosely attached to the plant tissue (root rhizosphere and leaf phyllosphere; see Methods for the methodological definitions of these compartments) (34) with those colonizing the ‘root’ (defined operationally as the combined firmly attached rhizoplane and endophytic bacteria) or the ‘shoot’ (defined operationally as ethanol resistant shoot endophytes). We observed that the ratio of FliC-L to FliC-H expression shifts towards greater FliC-H expression in both organs in those fractions associated with more intimate contact with the host (Fig. 4C; SI Appendix, Fig S4B). The shift towards higher FliC-H expression during plant advanced colonization is in alignment with the requirement for FliC-H in colonization (Fig. 4B).

Discussion

Our study reveals a previously unrecognized strategy by which the prevalent and plant-enriched Sphingomonas sp. resolve the evolutionary conflict between flagellar function and immune evasion. We identified isolates of this genus that encode two independently acquired, divergent flagellin genes: an immunogenic (FliC-H) and a non-immunogenic (FliC-L). Further characterizing the function of each flagellar variant in an experimentally tractable isolate revealed a striking division of labor: the immunogenic variant (FliC-H) is specialized for host attachment and colonization, whereas the non-immunogenic variant (FliC-L) is specialized for directional swimming. Both contribute to short term bacterial attachment to shoots. Our data does not rule out the possibility that FliC-H promotes colonization by improving the bacterial attachment to the plant or that FliC-H is required for host surface sensing, and hence MF220_L bacteria are locked in planktonic phase. These two flagella differ not only in immunogenicity and function but also in their expression patterns, genomic context, and morphology, suggesting functional specialization into motility and colonization roles. Thus, rather than diversification of a single flagellin to balance motility, attachment, and immune evasion, many Sphingomonas may partition these functions across two distinct loci. Most importantly, our findings suggest that the plant immune receptor FLS2 recognizes the flg22 epitope derived from the Sphingomonas sp. flagellar variant required for intimate plant colonization rather than the flagellar variant required for motility toward and on plant surfaces (Fig. 5).

Figure 5. Model: The plant immune receptor FLS2 specifically targets Sphingomonas sp. FliC variants required for full plant colonization.

Flagella of Sphingomonas MF220 contribute to swimming motility, plant colonization, and attachment to plant roots. While flagella composed of the non-immunogenic FliC-L protein (blue) facilitates more efficient swimming up a nutrient gradient, flagella containing the immunogenic FliC-H protein (red) are more effective at colonizing both the shoot and the root, and at attaching to plant roots. To effectively monitor bacteria capable of efficient root attachment and plant colonization, the plant FLS2 immune receptor has evolved to specifically recognize flg22-H epitopes (red triangles, derived from the immunogenic FliC-H required for attachment), rather than the flg22-L epitope (blue squares, derived from the non-immunogenic FliC-L essential for directional swimming).

Flagellin-FLS2 interactions are typically framed through the lens of antagonistic pleiotropy that balances immune recognition and motility (12). This model looks at mutational variations in a single flagellin gene performing multiple functions and predicts that immune evasion cannot occur without fitness costs. We examined attachment and colonization in Sphingomonas, where independently derived and regulated flagella provide disparate functions. Our findings reveal an alternative evolutionary trajectory for flagellar diversification. Instead of optimizing a single flagellin for immune evasion, motility, and colonization, Sphingomonas resolves conflict with the host plant via a division of labor. In certain members of this genus, a highly divergent collection of motility-flagella are immune-evasive while the flagella required for full colonization inside plant organs are immunogenic. In that case, even if the more ancient FliC-H was responsible for both motility and attachment, the acquisition of the motility-only FliC-L could enable a possible dual function flagella to specialize in attachment, and lose it’s motility function. Thus, Sphingomonas sp. avoids the pleiotropic constraint identified by Parys et al. (2021) (12) by separating motility and attachment to different flagellar system rather than by altering a constrained epitope.

Spatial receptor distribution provides a mechanistic context for the flagellar specialization we describe in Sphingomonas sp.. In Arabidopsis, FLS2 is not uniformly present across the plant body. Instead, FLS2 accumulates in potential microbial entry or attachment points inside the shoot and in compartments protected by the endodermal barrier in roots. (28, 29). This spatial organization suggests that immune surveillance is most intense where microbes transition from swimming and exploring to intimate host contact. Our data showing that FliC-H is required for colonization and expressed to a higher degree in colonizing bacteria, compared to host organ surface bacteria, suggest that the spatial organization of FLS2 is aimed at detecting FliC-H where it is most needed. Thus, receptor localization on the host and flagellin deployment by the bacterium converge to structure the outcome of immune perception.

The evolutionary strategy we define may lead to immune-mediated détente between the host and strains of Sphingomonas carrying divergent FliC loci: some Sphingomonas sp. are allowed to colonize plant organ surfaces via FliC-L driven flagellar motility but are gated out of shoot and root endophytic compartments by the requirement of FliC-H for attachment and colonization. This model supports the concept that the plant benefits provided by various Sphingomonas sp. might be largely limited to non-endodermal compartments by the FLS2-mediated host immune response in endodermal compartments. The weak MTI response generated by at least one Sphingomonas sp. might reflect this gating in action as the bacteria are continually probing the plant’s inner niches and thus triggering weak MTI (31, 32).

Sphingomonas are not the only bacteria encoding more than one flagellum, yet the specialization and division of labor we identified for Sphingomonas is unique. In biphasic Salmonella enterica, phase variation between two distinctly immunogenic flagellins (FliC and FljB) facilitates immune evasion in animal hosts (35), but the two proteins are functionally redundant for motility. Recent work in plant-associated Pseudomonas identified strains with duplicated FliC genes that differ in immunogenicity (36), though only one copy supports flagellar biogenesis and motility while the second is poorly expressed and non-functional. Importantly, Sphingomonas sp. strain A1 also possesses multiple distinct flagellin genes that assemble into a single heteromeric flagellum (37, 38). In this strain, the flagellum is a bifunctional appendage required for both motility and alginate uptake (39, 40). The alginate binding properties of Sphingomonas A1 flagella provide a plausible precedent for our observations in Sphingomonas MF220, where the FliC-H variant may have evolved a specialized role in promoting attachment to plant tissues, independent of the flagellum’s primary role in motility. Other genera, including Vibrio and Aeromonas, use physically distinct polar and lateral flagella for different types of motility (41). In contrast, our results show functional specialization across two flagellins, suggesting division of labor strategy, where one enables full directional swimming and immune evasion, and the other supports colonization despite being immunogenic.

In summary, we propose a refined model of MAMP-PRR coevolution: instead of targeting traits required for generic microbial viability, like swimming, plant immune receptors preferentially target bacterial functions directly associated with intimate host interaction, like attachment and colonization. This turns the plant immune system into a strategic gatekeeper in commensal interactions, enabling microbial arrival and spatially limited proliferation but restricting the transition from surface exploration to persistent colonization and damaging bacterial population expansion.

Material and Methods

Synteny analysis of the genomic island surrounding the FliC genes.

To examine genomic context and conservation of flagellum-associated regions, we performed a synteny analysis centered around each FliC gene. For every genome, we extracted all genes located upstream and downstream of both FliC loci and annotated them using KEGG Orthology (KO) identifiers. Genes assigned to flagellum-related KOs (e.g., motor, basal body, hook, export apparatus) were used to define a “flagellum island,” operationally defined as the continuous genomic segment containing FliC and its adjacent flagellar genes. We then compared the flagellum islands surrounding the 12 focal FliC alleles at two levels: (i) gene content similarity, quantified using the Jaccard index based on shared KO identifiers, and (ii) gene order conservation, quantified as the fraction of conserved adjacent gene pairs (i.e., the proportion of neighboring genes in island A that remain neighbors in island B). For each pairwise comparison, we calculated an island similarity index by multiplying the gene-content Jaccard index and the adjacency index, providing a composite measure of synteny that captures both presence/absence of genes and conservation of local genomic architecture.

Identification of FliC genes and cognate flg22 epitopes in wild Sphingomonas

To identify FliC homologs in environmental Sphingomonas isolates, we constructed a hidden Markov model (HMM) profile from the 12 experimentally flg22-validated FliC proteins in our strain collection and searched each genome using HMMER (42). Hits with high statistical confidence (E-value < 1×10⁻⁵⁰) were retained as bona fide FliC candidates.

Next, to extract the corresponding flg22 epitope from each identified FliC gene, we developed a guided positional alignment strategy. We started by identifying alignment consensus sequences built exclusively from the six strains for which flg22 immunogenicity had been experimentally characterized. We aligned each feature (FliC-H, FliC-L, flg22-H, flg22-L) independently. To construct four alignment consensus sequences. Then, for every newly identified FliC sequence, we performed a pairwise alignment to the FliC-H, FliC-L, flg22-H, and flg22-L consensus sequences to identify the correct region of the protein. The 22-amino-acid segment mapping to the flg22 position was extracted as the predicted flg22 sequence. This two-step positional alignment allowed us to reliably identify the flg22 epitope from highly divergent FliC and flg22 sequences, ensuring accurate recovery of the correct region across hundreds of wild Sphingomonas genomes.

Constructing a Sphingomonas MLST tree.

To infer phylogenetic relationships among Sphingomonas isolates, we constructed a multilocus sequence typing (MLST) tree based on conserved core genes. We identified 48 single-copy housekeeping genes present across all genomes in the dataset. Each gene was aligned independently using MAFFT (43). The resulting alignments were concatenated into a single “core-genes-genome”. To reduce noise from poorly aligned or rapidly evolving regions, we applied ZORRO (44) to compute highly divergent regions in the alignment. Finally, a maximum-likelihood phylogenetic tree was then generated using FastTree2 (45).

Swimming assay.

Sphingomonas strains MF220_WT, MF220_DD, MF220_L and MF220_H were grown overnight at 28 °C in 2×YT medium supplemented with rifampicin (50 μg/mL). The following day, 1 mL of each culture was centrifuged at 6,000 × g for 5 minutes, washed twice with fresh 2×YT medium, and resuspended in 500 μL of 2×YT. For motility assays, 2 μL of each bacterial suspension was spotted onto 0.3% LB soft agar plates and allowed to dry. Plates were incubated at 28 °C for 48 hours, after which swimming halos were imaged, and the radius of the swimming zone was used to quantify motility. At the end of swimming experiments, bacteria were sampled from the outermost swimming halo and from the center of the swimming MF220_WT population for expression analysis (Methods). In addition, bacteria were also collected from a 2-days old bacterial colony grown on a standard 1.5% LB agar plate as reference.

Deleting FliC-L and FliC-H from Sphingomonas MF220.

We generated three Sphingomonas MF220 deletion derivatives lacking either FliC-L (MF220_H derivative), FliCH (MF220_L derivative), or both genes (MF220_DD derivative) using homologous recombination. To construct the deletion plasmids, we amplified ~1,000 bp regions upstream and downstream of each FliC gene from MF220 genomic DNA. For MF220_L derivative, we used primers GD03_flkU_FliCH_F/GD04_flkU_FliCH_R and GD05_flkD_FliCH_F/ GD06_flkD_FliCH_R; for MF220_H derivative, we used GD11_flkU_FliCL_F/GD12_flkU_FliCL_R and GD13_flkD_FliCL_F/GD14_flkD_FliCL_R. The internal primers (GD04_flkU_FliCH_R and GD05_flkD_FliCH_F, and their MF220_H derivative counterparts) included 5’ overhangs complementary to one another, while the outer primers contained 5’ overhangs homologous to the linearized pMo130 vector backbone.

We assembled the flanking fragments and pMo130 by Gibson assembly to generate pMo130_dFliCH and pMo130_dFliCL. We transformed the constructs into E. coli Top10 via heat shock and selected transformants on LB agar containing 50 μg/mL kanamycin. We validated colonies by PCR using flanking primers GD07_Vo_FliCH_F/GD08_Vo_FliCH_R and GD15_Vo_FliCL_F/ GD16_Vo_FliCL_R, propagated the plasmids in 2×YT medium with 50 μg/mL kanamycin, and purified them using the QIAprep Spin Miniprep Kit (QIAGEN).

We electroporated pMo130_dFliCH and pMo130_dFliCL into E. coli WM3064, a diaminopimelate (DAP)-auxotrophic conjugation donor strain. We selected transformants on LB agar supplemented with 0.3 mM DAP, 50 μg/mL kanamycin, and 50 μg/mL erythromycin. We mated the resulting donor strains with MF220 wild-type cells on LB agar with 0.3 mM DAP. After overnight incubation, we selected transconjugants on LB agar with 50 μg/mL kanamycin.

We purified single crossover transconjugants by re-streaking and counter selected for double recombinants on LB agar containing 10% sucrose. We screened sucrose-resistant colonies by PCR: primers GD07_Vo_FliCH_F/GD08_Vo_FliCH_R (for MF220_L derivative) or GD15_Vo_FliCL_F/ GD16_Vo_FliCL_R (for MF220_H derivative) amplified across the flanking regions and yielded a product only upon deletion, while primers GD09_Vi_FliCH_F/GD10_Vi_FliCH_R (for MF220_L derivative) or GD17_Vi_FliCL_F/GD18_Vi_FliCL_R (for MF220_H derivative) within the deleted gene yielded no product. We confirmed deletion in MF220_H derivative and MF220_L derivative and further generated the MF220_DD strain by sequential deletion in the MF220_H derivative background using the same strategy.

All oligonucleotide sequences used are listed in the SI Appendix, Table S2.

Scanning Electron Microscopy (SEM).

Sphingomonas strains used in this study included MF220_WT, MF220_DD, MF220_L and MF220_H. Cultures were grown overnight at 28 °C in 2×YT medium supplemented with rifampicin (50 μg/mL), then diluted 1:100 into fresh 2×YT medium (without antibiotics) and incubated at 28 °C with shaking until reaching mid-log phase (OD₆₀₀ ≈ 0.4–0.5). To preserve intact flagella, cells were harvested by gentle centrifugation at 1,000 × g for 2 minutes at room temperature, then immediately resuspended in fixative solution (2% paraformaldehyde/2.5% glutaraldehyde in 0.15 M sodium phosphate buffer, pH 7.4) for subsequent processing and imaging by scanning electron microscopy. A droplet of each fixed sample was plated on 12 mm round glass coverslips and allowed to adhere at 4°C overnight. Coverslips were washed three times with 0.15 M sodium phosphate buffer, then incubated with 1% buffered osmium tetroxide for 30 minutes at room temperature. Coverslips were then washed three times in deionized water and dehydrated through an ascending series of ethanol (30%, 50%, 75%, 100%, 100% 100%). Coverslips were transferred in 100% ethanol to a Samdri-795 critical point dryer and dried using liquid carbon dioxide as the transitional solvent (Tousimis Research Corporation, Rockville, MD). Once dry, coverslips were mounted on 13 mm diameter aluminum stubs using carbon adhesive tabs and sputter coated with 5 nm of a 60:40 gold/palladium alloy using a Cressington 208HR Sputter Coater (Ted Pella Inc., Redding, CA). Images were obtained using a Zeiss Supra 25 FESEM operating at 5 kV using an SE2 detector, 20 μm aperture, and approximate working distances of 5 to 8 mm (Carl Zeiss Microscopy, LLC, Peabody, MA).

Characterizing short-term attachment to the plant.

Three-week-old A. thaliana Col-0 plants grown on 0.5× MS agar plates were used for bacterial attachment assays. Shoots and roots were carefully separated and transferred into individual wells of a sterile 12-well plate. Plant tissues were submerged in 2 mL of bacterial suspension (10⁶ CFU/mL in sterile distilled water) containing Sphingomonas strains: MF220_WT, MF220_DD, MF220_L and MF220_H. Inoculated plates were incubated with subtle shaking at 50 rpm for 4 hours. After incubation, tissues were washed three times with sterile water and vortexed briefly between washes to remove loosely attached bacteria. Washed tissues were weighed, transferred to microcentrifuge tubes containing three 4-mm beads and 250 μL of sterile water, and then homogenized using a bead mill. Serial dilutions were spotted onto LB agar supplemented with rifampicin (50 μg/mL), and colony-forming units (CFUs) were counted after 48 hours of incubation at 28°C. Attachment was quantified as CFUs per mg of fresh tissue. Experiments were performed with at least three independent biological replicates per strain and tissue type.

Characterizing long-term plant colonization.

Root Colonization

Surface-sterilized A. thaliana seeds (Col-0 background, fls2 mutant) were stratified at 4°C in the dark for 2 days and then transferred to 0.5× MS agar plates to germinate under long-day conditions (16 h light / 8 h dark) at 22°C. Seven-day-old seedlings were carefully transferred to new 0.5× MS plates supplemented with Sphingomonas strains (MF220_WT, MF220_DD, MF220_L and MF220_H) adjusted to ~10⁷ CFU/plate. After 12 days of co-cultivation, roots were harvested and washed three times with sterile distilled water to remove loosely attached bacteria. Roots were then weighed, ground in water, and serial dilutions were plated on 2X YT agar plates containing Rifampicin (50 μg/ml). CFUs were enumerated after 48 hours of incubation at 28°C. Colonization data were normalized to root weight. Each experiment was performed with at least three biological replicates.

Shoot Colonization

Four- to six-week-old A. thaliana plants (Col-0 and fls2) grown in pots were spray-inoculated with bacterial suspensions (10⁹ CFU/mL) of Sphingomonas MF220 strains ( MF220_WT, MF220_DD, MF220_L and MF220_H). Plants were maintained under short-day conditions for up to 12 days post-inoculation (dpi). Leaves were harvested at 0, 3, 6, 9, and 12 dpi, and the fresh weight of each sample was recorded prior to bacterial recovery. For each strain, three leaves were pooled per tube, with three independent tubes used per experimental set. To recover bacteria, leaves were washed three times in 1 mL sterile water for 30 s each. The first wash fraction was retained and designated as the loosely attached epiphytic population. Washed leaves were then transferred to 2 mL tubes containing three 4-mm glass beads and 500 μL sterile water and homogenized using a bead mill to release tightly attached and endophytic bacteria. Serial dilutions of all fractions were prepared in 96-well plates and spotted on selective medium for CFU enumeration after 2 days of incubation at 28°C.

Tissue Harvest for In Planta Characterization of FliC expression.

For expression analysis, shoot and root tissues were harvested separately from a colonization assay. Surface-sterilized A. thaliana Col-0 seeds were stratified at 4°C in the dark for 2 days, then transferred to 0.5× MS agar plates to germinate under long-day conditions (16 h light / 8 h dark) at 22°C. Seven-day-old seedlings were carefully transferred to new 0.5× MS plates supplemented with 10⁷ CFU/plate MF220_WT Sphingomonas. After 12 days of co-cultivation under long-day conditions, roots were harvested and washed three times with sterile distilled water to remove loosely attached bacteria. The first wash was processed as loosely attached bacteria (root rhizosphere). After three washes, roots were frozen in liquid nitrogen to process firmly attached rhizoplane and endophytic bacteria. Similarly, for shoots, the first water wash was processed as loosely attached bacteria (leaf phyllosphere). Subsequently, surface-sterilized shoots by immersion in 70% ethanol for 60 seconds, followed by two rinses with sterile distilled water. Finally, shoots were frozen in liquid nitrogen for endophytic bacterial gene expression analysis. Each experiment was performed in three biological replicates.

RNA Isolation and qRT-PCR.

Total RNA was extracted using the ZymoBIOMICS DNA/RNA Miniprep Kit (Zymo Research, Cat. R2002) following the manufacturer’s instructions. The isolated RNA was treated with DNase I to remove any DNA contamination using the Turbo DNA Free Kit (Invitrogen, Cat. AM1907). cDNA synthesis was performed using SuperScript IV Reverse Transcriptase (Invitrogen). Quantitative real-time PCR (qRT-PCR) was carried out using PowerUp SYBR Green Master Mix (Applied Biosystems) on a ViiA^™ 7 Real-Time PCR System (Applied Biosystems). qRT-PCR was performed with an initial denaturation at 95 °C for 10 min, followed by 40 cycles of 95 °C for 20 s, 58 °C for 20 s, and 72 °C for 30 s, with fluorescence acquisition during the extension step. Melt-curve analysis was performed using the instrument’s default settings to confirm amplification specificity. Expression levels of FliC-H and FliC-L were calculated based on –ΔCT values normalized to recA as the internal control. No amplification was detected in the no-bacteria negative control.

All oligonucleotide sequences used are listed in the SI Appendix, Table S2.

Significance Statement.

The bacterial flagellum controls microbial motility and surface attachment. Some flagella contain a flg22 peptide epitope that is recognized by the plant immune receptor FLS2. This immune surveillance creates a trade-off between flagellar function and immune evasion. A strain of plant-associated bacterium Sphingomonas resolves this conflict through the acquisition of two evolutionarily independent flagellin-encoding systems and a division-of-labor strategy: it employs a non-immunogenic flagellum for motility and a second, immunogenic flagellum for root attachment and colonization. Thus, the plant immune system monitors the attachment-associated flagellin rather than the motility associated flagellin. This division-of-labor could exist in other plant-associated Sphingomonas strains that carry two diverging flagellar loci. Our data illustrate microbial adaptation to plant associated lifestyle can intersect with plant immunity.