A carotenoid-deficient mutant of the plant-associated microbe Pantoea sp. YR343 displays an altered membrane proteome

Membrane organization plays an important role in signaling, transport, and defense. In eukaryotes, the stability, organization, and function of membrane proteins are influenced by certain lipids and sterols, such as cholesterol. Bacteria lack cholesterol, but carotenoids and hopanoids are predicted to play a similar role in modulating membrane properties. We have previously shown that the loss of carotenoids in the plant-associated bacteria Pantoea sp. YR343 results in changes to membrane biophysical properties and leads to physiological changes, including increased sensitivity to reactive oxygen species, reduced indole-3-acetic acid secretion, reduced biofilm and pellicle formation, and reduced plant colonization. Here, using whole cell and membrane proteomics, we show that the deletion of carotenoid production in Pantoea sp. YR343 results in altered membrane protein distribution and abundance. Moreover, we observe significant differences in the protein composition of detergent-resistant membrane fractions from wildtype and mutant cells, consistent with the prediction that carotenoids play a role in organizing membrane microdomains. These data provide new insights into the function of carotenoids in bacterial membrane organization and identify cellular functions that are affected by the loss of carotenoids.

Scientific Reports | (2020) 10:14985 | https://doi.org/10.1038/s41598-020-71672-w www.nature.com/scientificreports/ cell, membrane, DRM, or DSM) clustered together. A Student's t-test (paired t-test) was performed and a p value cutoff of ≤ 0.05 and a fold change (FC) of ≥ 2 was used to identify proteins with relative abundances that significantly differed between the wildtype and ΔcrtB samples. In total, 240, 134, 297 and 71 proteins were differentially abundant between wildtype and ΔcrtB cells for the whole cell, MP, DRM and DSM fractions, respectively (Fig. 4). Out of these, 188 (in whole cell), 111 (in MP), 211 (in DRM) and 44 (in DSM), were significantly less abundant in the ΔcrtB mutant in comparison to the wildtype, In total, 21 proteins were found to be differentially abundant across all four fractions in both wild type and ΔcrtB cells. The observed phenotypes found in the ΔcrtB mutant may be a result of the changes in abundance or distribution of these identified proteins.

Cluster of orthologous groups (COG) analysis.
To identify biological processes related to the differentially abundant proteins, functional classification of significant proteins was carried out using the COG database 37 . Significant proteins were grouped into 21 functional classes according to COG classification. Proteins belonging to transcription (K) and carbohydrate transport and metabolism (G) categories were abundant in whole cell samples, whereas cell wall/membrane/envelope biogenesis (M) proteins were abundant in membrane pellet, DRM, and DSM fractions (Fig. 5).

Gene ontology (GO) enrichment analysis.
To gain a deeper understanding of overall changes in protein abundance and distribution between the wildtype and ΔcrtB mutant, functional in silico classification of proteins was achieved via GO analysis using the BLAST2GO tool 38 . All of the proteins that were differentially abundant (p value ≤ 0.05 and a FC ≥ 2) based on the proteomic analyses were organized by GO terms to determine which biological processes were affected by the loss of carotenoids (Table 1). In the whole cell pairwise comparisons, proteins belonging to lipid biosynthesis (GO:0008610), lipid metabolism (GO: 0006629) and oligosaccharide metabolism (GO:0009311) were less abundant in the ΔcrtB mutant. Glycerophospholipids serve as the structural component of biological membranes and their alteration can affect physiology and adaptation 39 . Previously, we reported that the ΔcrtB mutant shows a modest increase in phosphatidylethanolamine (PE) head groups and unsaturated fatty acids when compared to wild type cells 17 . This observation could be the consequence of down regulation of lysophospholipase (PMI39_02976), which is important for glycerophospholipid metabolism 40 , in the ΔcrtB mutant. We also observed that a regulator of protease activity HflC, stomatin/ prohibitin superfamily-ybbK (2,511,379,369) appeared less abundant in the ΔcrtB mutant, although the difference did not meet the criteria to be statistically significant (p value = 0.04 but fold change = 1.4). YbbK, encoded by PMI39_01287, belongs to the reggie (flotillin) superfamily, which includes eukaryotic flotillins and the bacterial homolog FlotP which was identified in Bacillus anthracis membrane microdomains 25,41,42 . The apparent reduction of YbbK in the carotenoid mutant is consistent with changes to microdomain organization, which may affect cellular functions such as protein signaling and transport. It is possible that the observed reduction of indole-3-acetic acid secretion and the decreased pellicle and biofilm formation observed in the ΔcrtB mutant results from changes in membrane domain architecture. Histogram representing the percentage of proteins with predicted transmembrane helix domains (TMHMM) for each sample. Proteins with predicted transmembrane helices were identified using TMHMM software. The membrane fraction samples contain higher relative amounts of proteins with predicted TMHMM domains, with the DSM fraction having the largest enrichment.  (2,511,381,490) were also less abundant in the ΔcrtB mutant whole cell fraction. Notably, four undecaprenyl phosphate proteins (Locus tags-PMI39_03112, PMI39_03113, PMI39_03114, PMI39_03115) in an operon involved in amino sugar and nucleotide sugar metabolism were less abundant in the ΔcrtB mutant 43 . Undecaprenyl phosphate is a 55-carbon polyisoprenoid lipid involved in bacterial cell wall biogenesis by functioning as a lipid carrier, trafficking sugar intermediates across the plasma membrane 44 . There is also growing evidence that polyisoprenoids increase membrane fluidity and ion permeability [45][46][47][48] . The downregulation of this operon may explain, at least in part, the observed decrease in membrane fluidity in the ΔcrtB mutant 17 .
In the membrane fractions, several proteins with functional significance at the membrane were less abundant in the ΔcrtB mutant (p ≤ 0.05 and FC ≥ 2). In particular, proteins belonging to envelope (GO:0030313), cell outer membrane (GO:0009279), membrane biogenesis, and cellular homeostasis were downregulated in the ΔcrtB mutant. Homeostasis is important for living organisms to maintain internal stability and it includes iron and metal homeostasis, membrane lipid homeostasis, and pH homeostasis. For example, TonB (PMI39_04701), an outer membrane receptor for ferrienterochelin and colicins, which are important for iron homeostasis 49 , was less abundant in the ΔcrtB DSM fraction.
Transcriptional analyses of the ΔcrtB mutant. The differences observed in the membrane proteomes between wildtype and the ΔcrtB mutant could be due to many factors, including differences in membrane pro-

W T 2 _ D S M W T 3 _ D S M W T 1 _ D S M W T 2 _ D R M W T 3 _ D R M W T 1 _ D R M W T 2 _ M P W T 3 _ M P W T 1 _ M P W T 2 W T 3 W T 1 Δ c rt B 1 Δ c rt B 3
Δ c rt B 2 Figure 3. Hierarchical clustering of all proteins identified in Pantoea sp. YR343 wildtype and the ΔcrtB mutant. Heatmap of protein counts in Pantoea sp. YR343 wildtype and the ΔcrtB mutant indicate fraction specific abundance of proteins and differential abundance of proteins between wildtype and the ΔcrtB mutant. Higher protein abundance is indicated by red and lower protein abundance is indicated by blue. The heatmap was generated using gplots in Rstudio and scaled by column.   www.nature.com/scientificreports/ tein insertion and stability, localization, or abundance. To better understand the basis for the observed differences, we performed transcriptional analyses to examine which gene products are transcriptionally regulated. Differentially expressed transcripts between the wildtype and ΔcrtB mutant samples were identified using KBase tools as described in the methods 50 . Only 5 transcripts were significantly upregulated (p value ≤ 0.05 and FC ≥ 2), whereas 879 transcripts were significantly downregulated in the ΔcrtB mutant (Supplemental Table 1). Heat maps representing the differential expression profile of wildtype and ΔcrtB mutant are shown in Fig. 6. Detailed comparisons of the protein abundances and transcriptional regulation were performed for selected functional categories.
Cell wall/membrane/envelope biogenesis. Proteins predicted to be involved in cell wall/membrane/ envelope biogenesis (270 proteins total, COG category M) in Pantoea sp. YR343 were collected from the JGI IMG database. Table 2 lists all the proteins (p ≤ 0.005 and FC ≥ 1) involved in cell membrane biogenesis that were significantly differentially abundant in at least one fraction (56 total). Among these proteins, six undecaprenyl-phosphate (UDP) proteins belonging to peptidoglycan/lipopolysaccharide biosynthesis were found to be significantly less abundant in the ΔcrtB mutant (PMI39_01550, PMI39_02251, PMI39_03115, PMI39_01848, PMI39_03114, PMI39_04793). UDP gene products are involved in exopolysaccharide secretion, cationic antimicrobial peptide resistance, lipid A biogenesis, and peptidoglycan synthesis 44 and most were found in the DRM fractions. Transcript data for two of the UDP proteins, undecaprenyl-phosphate 4-deoxy-4-formamido-Larabinose transferase (PMI39_03114) and UDP-4-amino-4-deoxy-L-arabinose-oxoglutarate aminotransferase (PMI39_03115), showed downregulation at the transcript level in the ΔcrtB mutant ( Table 2). Downregulation of these UDP genes may explain the observed differences in the peptidoglycan layer of the ΔcrtB mutant in comparison to the wildtype 17 . Another protein, UDP-galactose-lipid carrier transferase (PMI39_01848), has a transmembrane domain and was found to be less abundant in the mutant DRM fraction. The gene encoding this protein is the first gene in a large operon that shows significant similarity to operons involved in EPS biosynthesis in the related plant-associated microbes Erwinia amylovora and Pantoea stewartii 51,52 . It is possible that the reduction in this protein decreases EPS production in the ΔcrtB mutant, which may contribute to the defects associated with biofilm formation and plant colonization. Outer membrane proteins (OMP) are important for transport of metabolites and toxins, membrane biogenesis, and for bacterial resistance. The folding and insertion of several OMPs are carried out by BamA along with three lipoproteins: BamB, BamC, and BamE forming the BAM machine (beta-barrel assembly) 53,54 . Lipoproteins are peripherally anchored membrane proteins involved in cell division, chemotaxis, signal transduction and   [55][56][57] . Among the 4 Bam proteins, BamA (PMI39_03681) and BamB (PMI39_03586) were found to be less abundant in the ΔcrtB mutant (Table 2). BamB was identified in both the membrane pellet and the DRM fraction. Studies have shown that BamB contains WD40 repeating units, thereby functioning as a scaffold protein in large multi-protein complexes 58 . It was also shown that the Bam complex increases the efficiency of folding of membrane proteins such as OmpA and EspP 59 . We also found that the Skp protein (PMI39_3680) was more abundant in the ΔcrtB mutant DSM fraction. Skp is a multivalent periplasmic chaperone preventing misfolding and aggregation of OMPs, such as OmpA, during transit from the inner to the outer membranes 60 . The changes in membrane fluidity, lipid content, and the lack of carotenoids in the ΔcrtB mutant may influence assembly of the Bam complex at the outer membrane, leading to misfolding of other OMPs, but not affect localization or function of the periplasmic Skp protein. Thus, the increased abundance of Skp in the ΔcrtB mutant may be a compensatory mechanism to maintain proper folding of proteins to protect the integrity of the cell.

Cell motility (N).
Beyond the defects in IAA secretion, biofilm formation, and root colonization previously reported 29 , we also observed that the ΔcrtB mutant appeared to be less motile than wildtype on swimming motility plates (Fig. 7a). To further characterize this defect, we compared motility patterns of wildtype and mutant cells by microscopy. We found that the average mean speed was 3.5 μm/s for the ΔcrtB mutant which was significantly reduced compared to 4.9 μm/s for wild type cells (Fig. 7b). Moreover, flagella staining and quantification using ImageJ indicated that the ΔcrtB mutant had significantly shorter flagella when compared to wildtype (Fig. 7c). The average flagellar length for the wildtype cells were 7.4 μm whereas the ΔcrtB mutant flagella measured 2.8 μm (Fig. 7d). To help explain the motility defect, we examined the 92 proteins predicted to be involved in cell motility, of which 18 were differentially abundant in the mutant compared to wildtype (Table 3). Among the 28 proteins that form the flagellar complex 61 , only 3 proteins: flagellar FliL (PMI39_02182) and two flagellar hookassociated protein 2 (PMI39_02605 and PMI39_02159) were significantly less abundant in the ΔcrtB mutant and only one of these, PMI39_02605, was significantly downregulated based on the transcriptomic data (  62,63 . It is possible that the changes in membrane lipid composition and/or organization affect assembly or function of the flagellar motor apparatus, leading to the observed motility defects. In addition to the proteins involved in the flagellar motor apparatus, we also found differences in protein abundance and transcriptional regulation of several methyl-accepting chemotaxis proteins (MCP) in the ΔcrtB mutant compared to wild type (Table 3). MCPs undergo reversible methylation in response to changes in the concentration of attractants or repellents in their environment 64 . Interestingly, two MCP proteins, encoded by PMI39_02297 and PMI39_01148, were found to be more abundant in the ΔcrtB whole cell fraction, whereas the MCP encoded by PMI39_02163 was found to be more abundant in the ΔcrtB DRM fraction. This increased abundance in the ΔcrtB mutant did not appear to be due to transcriptional upregulation (Table 3). Additional experiments are needed to distinguish whether these proteins are differentially localized, more stable, or perhaps more easily extracted from the ΔcrtB mutant.

Lipid transport and metabolism (I).
In Pantoea sp. YR343, 150 proteins are found in the lipid transport and metabolism COG category I, of which 26 proteins were found to be significantly abundant in at least one fraction (Table 4). Two choline dehydrogenases (encoded by PMI39_02890 and PMI39_00318) were significantly more abundant in all or most fractions of the ΔcrtB mutant. Surprisingly, however, these genes were transcriptionally downregulated (Table 4). Choline dehydrogenase catalyzes the first step in glycine betaine synthesis to produce the final compound betaine, an effective osmoprotectant 65,66 . It is possible that the lipid composition or membrane organization in the carotenoid mutant promotes choline dehydrogenase protein stability or, alternatively, promotes its efficient extraction. Other proteins such as lysophospholipase (PMI39_01261, PMI39_04916) and NAD(P) dependent dehydrogenasese (PMI39_04227, PMI39_04693, PMI39_04133) were also more abundant in the ΔcrtB mutant.

Signal transduction mechanism (T). Bacterial signal transduction networks regulate sensing and
responses to environmental and intracellular parameters. In Pantoea sp. YR343, 235 proteins are predicted to be involved in signal transduction based on the COG category T (Table 5). Among these proteins, only 32 proteins were found to be differentially abundant in at least one fraction. In our data, we observed an abundance of OmpR family proteins 68 , including the phosphate regulon response regulator OmpR (PMI39_03347) in the ΔcrtB DRM fractions. OmpR, along with its histidine kinase partner EnvZ, are important for osmotic tolerance, virulence and motility in Acinetobacter baumanii [69][70][71] . In E. coli, OmpR and EnvZ regulate OmpF and OmpC proteins that are essential for responding to environmental signals.

Conclusion
The deletion of carotenoids in the ΔcrtB mutant leads not only to increased sensitivity to oxidative stress, but also to defects in IAA secretion, pellicle and biofilm formation, motility, and root colonization. In addition to the differences in lipid composition and membrane fluidity previously reported 17 , the loss of carotenoids also results in changes to the proteome of the ΔcrtB mutant compared to wildtype. We report a detailed proteome analysis comparing the wildtype and ΔcrtB mutant focusing on changes in membrane protein distribution and abundance. Consistent with the observed phenotypes in the mutant, we found that several classes of proteins belonging to membrane biogenesis, signal transduction, and cell motility were affected in the ΔcrtB mutant. The most dramatic changes to the proteome were observed in the DRM fraction, which is consistent with the idea that the DRM fraction represents membrane microdomains and the presence of cholesterol (in eukaryotes) or carotenoids and hopanoids (in prokaryotes) is vital to the organization of these domains 72 . In the absence of carotenoids, these microdomains may be unstable or have a change in membrane thickness which, in turn, www.nature.com/scientificreports/ affects protein insertion, stability, or recruitment. These data underscore the importance of bacterial membrane organization for cellular functions such as secretion, motility, and signaling.

Methods
Bacterial strains and growth conditions. Pantoea sp. YR343 and ΔcrtB cells were grown in Luria-Bertani broth (per 1 L, 10 g Bacto-tryptone, 10 g NaCl, 5 g yeast extract) medium at 28 °C with shaking to OD 600 of 1 (stationary phase). The ΔcrtB mutant was constructed as described 29 . , and DNase I was added to the washed cells. Cells were disrupted using French press followed by a short centrifugation to eliminate cell debris. The membrane fraction was precipitated by ultracentrifugation (100,000×g for 1 h at 4 °C). The resulting cell pellet was resuspended in Buffer H with 10% glycerol. At this stage, a fraction of the membrane pellet was collected. Table 3. List of significantly differentially abundant proteins involved in cell motility. Protein list from JGI for each COG category was matched with the proteomics dataset and only proteins that were significantly different in at least in one fraction in the wildtype or ΔcrtB mutant are reported. TM/SP-proteins with transmembrane helices or signal peptide. Red: proteins that are significantly less abundant in the ΔcrtB mutant, green: proteins that are significantly more abundant in the ΔcrtB mutant, grey: non-significant proteins and white: proteins that are not detected. www.nature.com/scientificreports/ Table 4. List of significantly differentially abundant proteins involved in lipid transport and metabolism. Protein list from JGI for each COG category was matched with the proteomics dataset and only proteins that were significantly different in at least in one fraction in the wildtype or ΔcrtB mutant are reported. TM/SPproteins with transmembrane helices or signal peptide. Red: proteins that are significantly less abundant in the ΔcrtB mutant, green: proteins that are significantly more abundant in the ΔcrtB mutant, grey: non-significant proteins and white: proteins that are not detected. To isolate DRM and DSM fractions, the membrane pellet was incubated for 30 min at 4 °C with lysis and separation buffer (CelLytic MEM protein extraction kit from Sigma Aldrich). After incubation, the membrane pellet was mixed 1:1 with 80% sucrose and carefully overlaid with 20% sucrose. Using a swinging bucket rotor, separation was carried out at 100,000×g at 4 °C for 16 h. The DRM and DSM fractions were collected and stored at − 20 °C for proteomic analysis.

Isolation of whole cell
Protein extraction and digestion. Cell pellets were suspended in sodium dodecyl sulfate (SDS) lysis buffer (2% in 100 mM of NH 4 HCO 3 , 10 mM DTT). Samples were physically disrupted by bead beating (0.15 mm) at 8,000 rpm for 5 min. Crude lysates were boiled for 5 min at 90 °C. Cysteines were blocked to avoid disulfide bridge reformation by adjusting each sample to 30 mM IAA and incubating in the dark for 15 min at room temperature. Proteins were precipitated using a chloroform/methanol/water extraction. Dried protein pellets were resuspended in 2% sodium deoxycholate (SDC) (100 mM NH 4 HCO 3 ) and protein amounts were estimated by performing a BCA assay (Pierce Biotechnology). In general, membrane fractions from the mutant showed a reduced protein concentration compared to wild type (MP: 3.7 mg/mL (wt) and 2.8 mg/mL (ΔcrtB); DSM: 858 µg/mL (wt) and 845 µg/mL (ΔcrtB); DRM: 165 µg/mL (wt) and 138 µg/mL (ΔcrtB)). For each sample, an aliquot of approximately 500 µg of protein was digested via two aliquots of sequencing-grade trypsin (Promega, 1:75 [w:w]) at two different sample dilutions, (overnight) followed by incubating 3 h at 37 °C. The peptide mixture was adjusted to 0.5% formaldehyde (FA) to precipitate SDC. Hydrated ethyl acetate was added to each sample at a 1:1 [v:v] ratio three times to effectively remove SDC. Samples were then placed in a SpeedVac Concentrator (Thermo Fischer Scientific) to remove ethyl acetate and further concentrate the sample. The peptideenriched flow through was quantified using the BCA assay, desalted on RP-C18 stage tips (Pierce Biotechnology) and then stored at − 80 °C prior to LC-MS/MS analysis.

LC-MS/MS. All samples were analyzed on a Q Exactive Plus mass spectrometer (Thermo Fisher Scientific)
coupled with a Proxeon EASY-nLC 1200 liquid chromatography (LC) pump (Thermo Fisher Scientific) as previously described 73 . In brief, peptide mixtures were separated on a 75 μm inner diameter microcapillary column packed with 30 cm of Kinetex C18 resin (1.7 μm, 100 Å, Phenomenex). For each peptide mixture, a 2 μg aliquot was loaded in buffer A (0.1% formic acid, 2% acetonitrile) and eluted with a linear 150 min gradient of 2-20% of buffer B (0.1% formic acid, 80% acetonitrile), followed by an increase in buffer B to 30% for 10 min, another increase to 50% buffer for 10 min and concluding with a 10 min wash at 98% buffer A. The flow rate was kept at 200 nL/min. Mass spectra data was acquired with the Thermo Xcalibur software version 2.2, and a topN method where N could be up to 15 was employed for data-dependent acquisition 73 .
Peptide identification and protein inference. MS raw data files were searched against the Pantoea sp.
YR343 FASTA database to which common contaminate proteins had been added. A decoy database, consisting of the reversed sequences of the target database, was appended to discern the false-discovery rate (FDR) at the spectral level. For standard database searching, the peptide fragmentation spectra (MS/MS) were analyzed by the Crux pipeline v3.0 74 . The MS/MS were searched using the Tide algorithm 75 and was configured to derive fully-tryptic peptides using default settings except for the following parameters: allowed clip nterm-methionine, a precursor mass tolerance of 10 parts per million (ppm), a static modification on cysteines (iodoacetamide; + 57.0214 Da), and dynamic modifications on methionine (oxidation; 15.9949). The results were processed by Percolator 76 to estimate q values. Peptide spectrum matches (PSMs) and peptides were considered identified at a q value < 0.01. Across the entire experimental dataset, proteins were required to have at least 2 distinct peptide sequences and 2 minimum spectra per protein.
Protein quantification. For label-free quantification, MS1-level precursor intensities were derived from MOFF 77 using the following parameters: 10 ppm mass tolerance, retention time window for extracted ion chro-Scientific Reports | (2020) 10:14985 | https://doi.org/10.1038/s41598-020-71672-w www.nature.com/scientificreports/ Statistical analysis for differential abundances. For this study, we performed ANOVA with post-hoc Tukey's test to identify differential protein abundances across the wildtype Pantoea sp. YR343 dataset comparisons or ΔcrtB mutant dataset comparisons and protein abundances were considered to have a significant change in abundance for p values < 0.05 and at least one absolute value of log2 fold-change differences > 1. To identify differential protein abundances between wildtype and ΔcrtB fractions, we performed a Student's t-test for the pairwise comparisons. A protein was categorized as having a significant abundance difference if it passed a significance threshold requiring a p value < 0.05 and absolute value of log2 fold-change difference > 1. Hierarchical clustering (one-way; Fast Ward method) was performed to identify differential abundance patterns.
Gene ontology enrichment. Gene ontology (GO) term annotation was performed using Blast2GO 38 with a blastp E-value hit filter of 1 × 10 -5 , an annotation cutoff value of 55 and a GO weight of 5. Using the Cytoscape 79 plugin ClueGO 80 , observed GO biological processes were subjected to the right-sided hypergeometric enrichment test at medium network specificity selection and p-value correction was performed using the Holm-Bonferroni step-down method 81 . For each cluster, we required a minimum of 3 and a maximum of 8 selected GO tree levels, and each cluster was set to include a minimum of 3-4% of genes associated with each term. The GO terms at adjusted p < 0.05 were considered significantly enriched.
RNA extraction, sequencing and analysis. Wild type and ΔcrtB cells were grown to stationary phase (OD 600 = 1). RNA was extracted using RNeasy mini kit (QIAGEN, Valencia, CA) following manufacturer's instructions and quantified using Nanodrop (Thermo Scientific). Sequencing was carried out by GENEWIZ Next Generation Sequencing Services. Transcript analysis was carried out using KBase 50 (https:// kbase. us/). KBase and its tools were used to generate the sample set, align and assemble reads to the genome, and identify differentially abundant genes between wild type and ΔcrtB.

Motility assays.
To compare the swimming motility function of Pantoea sp. YR343 and ΔcrtB cells, cells were grown overnight with shaking (250 rpm) in LB medium at 28 °C. Swimming motility was examined on LB containing 0.3% w/v agar. A 5 μL aliquot of cells were inoculated in the center of the plate and incubated at 28 °C for 18 h. Live cell imaging of bacterial motility was measured using a Nikon Eclipse Ti-U inverted microscope. Cells from motility plates were inoculated in R2A media overnight at 28 °C with shaking (250 rpm). Next day, cells were reinoculated in R2A media and grown to an OD 600 of 0.5. A 20 µL aliquot of cells were placed on a coverslip and 10 s videos were captured using NIS-Elements imaging software. Trajectories and velocities (pixels/frame) of Pantoea sp. YR343 and ΔcrtB cells were calculated with the "TrackMate" plugin (https:// imagej. net/ Track Mate).
Flagella staining. Flagella staining was carried out using a protocol adapted from Turner et al. 82 . Briefly, Pantoea sp. YR343 and ΔcrtB cells from swimming plates were inoculated overnight in R2A medium at 28 °C with shaking (250 rpm). Next day, cells were diluted 1:10 in fresh R2A medium and grown to OD 600 of 0.5. Motility of the culture was confirmed using a confocal microscope. Cells were collected by centrifugation (2000*g, 3 min) and washed three times in buffer (0.01 M KPO 4 , 0.067 M NaCl, 10 −4 M (Ethylenediaminetetraacetic acid (EDTA) [pH 7.0]). Alexa Fluor 594 carboxylic acid succinimidyl ester (ThermoFisher Scientific) was added to the concentrated bacterial suspension and incubated in the dark for 1 h. Cells were then washed three times with buffer containing Brij 35 (10 −4 %) and 0.4% glucose. Concentrated cells were then placed on an agarose pad (1% agarose in phosphate buffered saline) and imaged using a Zeiss LSM 710 confocal microscope. Flagellar length of 30 wildtype and ΔcrtB cells were measured using ImageJ. The data are represented as the mean flagellar length in µm ± SE calculated using unpaired t-test.

Data availability
All proteomics mass spectrometry data collected in this study was deposited at the ProteomeXchange Consortium via the MASSIVE repository under the project identifier MSV000085068.