A putative lateral flagella of the cystic fibrosis pathogen Burkholderia dolosa regulates swimming motility and host cytokine production

Burkholderia dolosa caused an outbreak in the cystic fibrosis clinic at Boston Children’s Hospital and was associated with high mortality in these patients. This species is part of a larger complex of opportunistic pathogens known as the Burkholderia cepacia complex (Bcc). Compared to other species in the Bcc, B. dolosa is highly transmissible; thus understanding its virulence mechanisms is important for preventing future outbreaks. The genome of one of the outbreak strains, AU0158, revealed a homolog of the lafA gene encoding a putative lateral flagellin, which, in other non-Bcc species, is used for movement on solid surfaces, attachment to host cells, or movement inside host cells. Here, we analyzed the conservation of the lafA gene and protein sequences, which are distinct from those of the polar flagella, and found lafA homologs to be present in numerous β-proteobacteria but notably absent from most other Bcc species. A lafA deletion mutant in B. dolosa showed a greater swimming motility than wild-type due to an increase in the number of polar flagella, but did not appear to contribute to biofilm formation, host cell invasion, or murine lung colonization or persistence over time. However, the lafA gene was important for cytokine production in human peripheral blood mononuclear cells, suggesting it may have a role in recognition by the human immune response.

Introduction Cystic fibrosis (CF) has a prevalence of~1/3,000 live Caucasian births, making it a commonly inherited disorder leading to a limited lifespan and high morbidity [1]. Mutations within the chloride ion transporter gene, CFTR, result in CF and lead to the production of a thick, sticky, low pH mucus in the lungs of CF patients, creating a breeding ground for bacteria (reviewed in [2]). Common pathogens infecting the lungs of CF patients [3] include Pseudomonas aeruginosa, Staphylococcus aureus, Haemophilus influenzae, Stenotrophomonas maltophilia, and members of the Gram negative Burkholderia cepacia complex (Bcc), which is comprised of over 20 species closely related species including the intracellular opportunist B. dolosa [4].
Members of the Bcc are capable of producing a necrotizing pneumonia in infected patients, characterized by a rapid decline in lung function (1-3 months) with bacteremia and near 100% mortality. This is often called the "cepacia syndrome" [5]. Certain species within the Bcc are capable of facile patient-to-patient transmission and have caused outbreaks in CF clinics worldwide [6][7][8]. The most common outbreak strains capable of spreading between CF patients fall within the B. cenocepacia and B. dolosa species [9]. These outbreaks have resulted in the enforcement of strict regulations and segregation for CF patients. One of the outbreaks occurred at Boston Children's Hospital, in which B. dolosa infected over 40 patients and contributed to the death of many, if not all, of them [10].
The genome of the AU0158 outbreak isolate of B. dolosa was found to contain genes encoding both a polar flagellum (fliC) as well as an additional flagellin gene located at the AK34_RS07895 locus. Preliminary evidence suggests that the protein encoded at AK34_RS07895 serves as the monomer for lateral flagella which are found infrequently in bacteria but contribute to movement on solid surfaces in other pathogens such as Aeromonas hydrophila and Vibrio parahaemolyticus [45,46]. These organisms often express the polar flagellum alone while grown in liquid culture, but express both polar and lateral flagella when grown on a solid medium. The polar flagellum of A. hydrophila has also been shown to act as a sensor that regulates the expression of the lateral flagellin genes [47]. Additionally, it was found that this polar-flagellum-induced increase in lateral flagella gene expression resulted in an increased formation of linkages between bacterial cells that contributed to microcolony and subsequent biofilm formation [48,49]. Our group has previously explored the host immune response to B. dolosa and examined the role of polar flagellum in infection [50] in which we found that wild-type B. dolosa AU0158 is not recognized by the host immune response in vivo in murine models but it is recognized in vitro by cultured cells. Further, the polar flagella may have a role in infection but this may be masked in vivo. The goal of this study was to examine the gene encoding the putative lateral flagellin in B. dolosa in terms of evolutionary conservation and to determine if this putative lateral flagellin gene influenced motility, biofilm formation, host cell invasion, host lung colonization, and host cytokine production in response to B. dolosa.

Conservation bioinformatics and PCR
For each B. dolosa AU0158 flagella-associated gene, BLAST searches of the associated amino acid sequences was conducted in both the burkholderia.com database [15] and GenBank. For the majority of genes, two homologs were present (<1e -10 ) with one homolog being more similar to polar flagella system proteins and the other homolog more similar to lateral flagella homologs. The threshold used to determine a homolog changed dynamically with each gene but in general, there was a wide discrepancy in e-values allowing for assignment as more likely polar flagella-related or lateral flagella-related as these systems are quite distinct from each other at the gene and protein level.
The protein sequence of the B. dolosa AU0158 lafA gene was obtained from http://www. burkholderia.com [15] and used to BLAST [51] two databases: the burkholderia.com collection (for other Burkholderia species); and the nr database at Genbank [52]. The closest homologs in other species were extracted and aligned with Clustal Omega [53]. For visualization, an. aln file of the resultant alignment was uploaded into ESPript [54] using default parameters. For phylogenetic network construction, a.nxs file of the resultant alignment was uploaded into the Splitstree program [55].
To experimentally determine whether the lafA gene is conserved in all strains, genomic DNA was extracted from outbreak and non-outbreak B. dolosa isolates using the method described in [56] except that 0.5 mL of overnight culture was used and 3 μL of RNaseA (10 mg/ml) was added during the lysis step. PCR was performed on 200 ng genomic DNA using GoTaq Green Master Mix (Promega) and 10 μM each Laf_OE_up and Laf_OE_down (Table 1) which should produce a band of 889 bp.
To visualize regions of conservation surrounding the lafA gene in other Burkholderia species, precomputed alignments were analyzed for conservation surround AK34_RS07895 in the VISTA browser located at Integrated Microbial Genomes website: (https://img.jgi.doe.gov/cgibin/w/main.cgi?section=Vista&page=vista). Default parameters for conservation were used. Microbial genomes chosen for comparison included multiple B. cenocepacia, B. multivorans, and B. ambifaria strains as well as select single strains/isolates of other Burkholderia species.

B. dolosa mutant generation
An in-frame deletion mutation in the B. dolosa lafA gene was generated via gene splicing by overlap extension [57] as previously described {Roux, 2017 #7359}. Briefly, 500 bp upstream and 500 bp downstream of AK34_RS07895 (sequences obtained at http://www.burkholderia.com [15]) was amplified by PCR using Bd_lafA_Dl-5'up and Bd_lafA_Dl-3'up or Bd_lafA_Dl-5'down and Bd_lafA_Dl-3'down (Table 1) respectively. These templates were purified and used in a second PCR as templates with Bd_lafA_Dl-5'up and Bd_lafA_Dl-3'down to generate fused amplicons. These amplicons were cut using the XmaI and EcoRI restriction enzymes and ligated into in the XmaI and EcoRI sites of pEXKm5 [58] to generate the pEXKm5Tet-lafAdel plasmid. They were then transformed into DH5α competent cells in LB agar supplemented with 50 μg/ml kanamycin, followed by Sanger sequencing. Cells of the conjugative donor E. coli strain SM10λpir were transformed with pEXKm5Tet-lafAdel and selected on LB agar supplemented with 50 μg/ml kanamycin + 10 μg/ml tetracycline. Plasmids were then transferred into B. dolosa by conjugation and insertion of the plasmid into the chromosome was verified by PCR. Counter-selection was performed by plating the verified transconjugants on LB medium + 10% sucrose for B. dolosa and checked for loss of the plasmid backbone by blue-white screening on X-Gluc (25 μg/ml)containing LB medium and checked for lafA gene deletion by PCR. The growth rates of mutant strains were not significantly different than the wild-type in LB liquid medium.
To generate the lafA complemented strain, PCR was used to amplify 253 bp upstream of BDAG_04366 through the coding region with the LafAexp-UP and LafAexp_DOWN primers. The amplicon was cloned into pUCP18T-miniTn7-Tp as a XmaI/EcoRI fragment and verified by Sanger sequencing to generate pUCP18T-miniTn7-Tp-lafA. B. dolosa AU1058 wild-type, ΔfliC, and ΔlafA were conjugated with E. coli SM10 λpir containing pUCP18T-miniTn7-Tp empty vector and transconjugants were selected on LB + 1000 μg/ml trimethoprim + 10 μg/ml gentamicin. In parallel, B. dolosa ΔlafA was also conjugated with pUCP18T-miniTn7-Tp-lafA. Transconjugants were verified by PCR for transposon insertion downstream of BDAG_04221 and to verify the original flagellin deletions where appropriate.

Microscopy
For examination of flagella by transmission electron microscopy (TEM) B. dolosa AU0158 wild-type, ΔfliC, and ΔlafA mutant strains were grown for 18 hours on trypticase soy agar (TSA) plates or in LB broth. Bacteria on plates were fixed by placing a drop of 1% glutaraldehyde onto colonies then a Formvar-coated copper grid was floated on top for~1 minute. A drop of bacteria grown in suspension was placed on a section of parafilm and a Formvarcoated copper grid was floated on top for~30 seconds. The grid was washed by floating on top of a drop of DI water and then fixed with 1% glutaraldehyde~1 minute. Grids from both bacteria grown on plates and in suspension were washed in DI water and stained with 0.75% uranyl acetate for 1 minute before visualization on the electron microscope. At least 10 separate images of each strain grown in each condition where taken and a representative image is shown.
To visualize movement of the B. dolosa ΔfliC mutant, agar concentrations of 0.3%, 0.4%, and 0.5% and differing temperatures of 23˚C, 30˚C, 37˚C, and 42˚C were used to grow B. dolosa wild-type and ΔfliC strains. Wet mounts were prepared and visualized using bright field microscopy.

Internalization assays
Mid-log phase cultures of B. dolosa AU0158 wild-type, AU0158 ΔfliC, and AU1058 ΔlafA strains were mixed with RAW264.7 murine macrophages in 96-well plates at a multiplicity of infection of~100 bacteria:eukaryotic cell. As a negative control, medium alone was added to eukaryotic cells. The plate was centrifuged at 800 x g for 4 minutes to ensure contact between the bacteria and the eukaryotic cells. After a 2 hour incubation, the DMEM medium was removed and the cells washed with 1X PBS. DMEM containing an antibiotic cocktail [final concentrations: 1 mg/ml kanamycin, 1 mg/ml ceftazidime, 1X Pen-Strep (Sigma Aldrich)] was added to all cells to kill extracellular bacteria. After an additional 2 hour incubation, eukaryotic cells were lysed with 1% Triton X-100 in DMEM, serially diluted and plated on LB agar. Bacterial counts were given in CFU/ml and compared for significance by one-way ANOVA analysis using the Tukey multiple comparison post-test.

Biofilm assays
B. dolosa strain AU0158 wild-type, ΔfliC, and ΔlafA mutants were grown overnight in LB broth, then diluted 1:100 into fresh medium. One hundred microliters of each culture was pipetted into replicate wells of a 96-well PVC plate and incubated for 3 days at 37˚C. Biofilm formation was quantitated as described in [59] except: 1X TSB with 1% (w/v) glucose was used for the medium in the wells; biofilms were quantitated after 3 days; and plates were incubated at 37˚C in a humidifying chamber (plates were stored on wet paper towels inside a closed container to prevent drying). Biofilm formation was compared between each strain and tested for significance by one-way ANOVA analysis using the Tukey multiple comparison post-test.

Murine models
All mouse work was performed according to protocol 02791 approved by the Harvard Medical School Institutional Animal Care and Use Committee which is accredited by the Association for the Assessment and Accreditation of Laboratory Animal Care, International (AAALAC). Bacterial inocula of B. dolosa AU0158 wild-type, B. dolosa AU0158 ΔlafA, and P. aeruginosa PAO1 wild-type (which served as a well-studied positive control for cytokine production and bacterial persistence) were prepared from dilutions of mid-log phase cultures grown in LB to obtain a solution of~2.5 x 10 8 CFU/ml. Six to seven week-old C57BL/6 female mice (Charles River) were anesthetized by intraperitoneal injection of ketamine/xylazine and 10 μl of bacterial inoculum was instilled into each nare (i.e. 20 μL final volume resulting in an inoculum of 5 x 10 6 CFU/mouse in sterile PBS) as described previously [50]. PBS alone was also instilled in 4 mice as negative controls. Mice were observed twice daily until day 6 as part of routine care.
At 1, 6, 24 and 144 h post-infection, 4-5 mice inoculated with each strain were sacrificed using CO 2 asphyxiation followed by cervical dislocation, and the lungs harvested and weighed. A bronchoalveolar lavage (BAL) was performed with 1 mL cold PBS and lungs were then homogenized in 1 ml sterile PBS. BAL fluid was separated for immediate serial dilutions for bacterial counts on LB agar plates and then centrifuged. The supernatant was frozen for subsequent cytokine measurement. One-way ANOVAs using the Tukey multiple comparison posttest were used to assess for significance differences between all treatments at each time point.

Cytokine assays
Measurements of cytokines were performed on supernatants of centrifuged BAL fluid from each mouse by Luminex magnetic assay with a mouse cytokine 15-plex panel (Life Technologies, Grand Island, NY). The following murine cytokines were evaluated: IL-1α, IL-1β, IL-2, IL-4, IL-5, IL-6, IL-10, IL-12 p40, IL-12 p70, IL-17, IL-33, TNF, IFNγ, MCP-1 and GM-CSF. Of these, MCP-1 and GM-CSF failed to produce detectable levels of cytokines in at least 3 biological replicates for either our positive control or test strains and thus were eliminated from analyses.
For in vitro ELISAs with cultured murine cells, RAW264.7 macrophage cells were seeded in a 96 well plate and incubated for~2 hours. B. dolosa AU0158 wild-type and lafA mutant strains and P. aeruginosa PAO1 wild-type (which served as a positive control) were grown overnight in LB medium, diluted 1:10 in fresh medium, harvested at mid-log phase growth, washed PBS, and resuspended in warmed DMEM medium to the desired concentration. Quadruplicate samples of each bacterial strain or medium alone (which served as a negative control) were added at designated M.O.I.s and incubated for six hours at 37˚C and 5% CO 2 . At 6 hours postinfection (h.p.i.), cells were lysed and TNF and MIP-2 were assessed by the Ready-Set-Go! TNF alpha ELISA kit [eBioscience; limits of detection 8 pg/mL] and Mouse MIP-2/CXCL2 ELISA Kit [Boster, limit of detection = 15.6 pg/mL], respectively, according to the manufacturer's instructions.
For ELISAs on human PBMCs, the buffy coat was collected from an Lymphocyte Separation Medium (Fisher Scientific, Fair Lawn, NJ) gradient, washed in PBS (10 mM, pH 7.2; Gibco, Grand Island, NY) and the PBMCs were plated in RPMI-1640 (Gibco) containing 10% FBS (Atlanta Biologicals, Norcross, GA). PBMCs were incubated overnight at 37˚C in 5% CO 2 . They were stimulated the following day with either live or heat-killed (30 min, 65˚C) bacteria at an M.O.I. of 10 bacteria:1 PBMC for 20 hours. IL-1β, IL-8, and TNF released into 20 hr cell-free supernatants were measured using eBioscience ELISA Ready-SET-Go! kits (San Diego, CA), according to the manufacturer's instructions. For all ELISAs, the positive and negative controls performed as expected.
For all cytokine analyses, t-tests were used to compare wild-type and mutant strain induction of cytokines in host cells.

Transposon library screening
The plasmid pSAM_DTc was created by replacing the erthryomycin cassette from pSAM_-DYH [60] with a tetracycline resistance cassette from pmini-CTX1 [61] inserted as a MfeI/ XbaI fragment. The resulting plasmid, pSAM_DTc was verified by PCR and sequencing and subsequently transformed into E. coli SM10λpir to generate a donor strain for conjugation.
A transposon mutant library of B. dolosa AU0158 was created using the pSAM_DTc plasmid and~200,000 individual transconjugants were collected and grown in LB broth overnight. Twenty colonies from the library were screened by semi-random PCR to ensure random insertion. The library was grown in 25 mL LB overnight and aliquoted into 1 ml freezer stocks containing 20% glycerol. One freezer stock was inoculated into 25 mL LB + 75 μg/ml tetracycline and grown overnight. One mL of this culture was diluted into 25 mL, grown overnight and was then subjected to genomic DNA purification (input). The remaining cells were washed twice with sterile 0.9% NaCl, concentrated to 4.1 X 10 10 CFU/ml, and 10 μL of this solution was instilled into each nare of 5 C57Bl/6 female mice (6-8 weeks old; Charles River) that had been anesthetized with ketamine and xylazine resulting in the instillation of 8.2 X 10 8 CFU/ mouse. After 48 hours, all mice looked healthy and the mouse lungs were harvested following CO 2 euthanasia, weighed, homogenized, and inoculated into 25 mL LB broth. Cells were grown overnight in LB liquid, plated on plain LB agar to ensure only the expected colony morphology (i.e. no contamination occurred from gut microbiota), and 1 mL of this was subjected to genomic DNA purification as in [56].
Tn-seq libraries were prepared from output and input libraries according to published procedures [60,[62][63][64]. Briefly, 50 μg of genomic DNA was restricted using 50U MmeI (New England Biolabs) overnight at 37 degrees followed by gel extraction (Qiagen Gel Extraction Kit) to excise a band 1.7-2.1 kb in size. These products were ligated to double-stranded barcoded DNA adaptors using T4 DNA ligase (New England Biolabs) and purified (Qiaquick PCR Purification Kit). Ligation products were used a templates in replicate PCR reactions with high-fidelity polymerase (KAPA Biosystems) and the LIB_PCR_5 and LIB_PCR_3 primers [65]. Pooled replicate reactions were sequenced by Illumina HiSeq 2500 at the Harvard Biopolymers Sequencing Facility. Reads were separated by barcode, trimmed to remove adaptor and transposon sequences, and mapped to the B. dolosa AU0158 genome using the CLC Genomics Workbench RNA-seq module and default settings as in [60,62,63,66]. The fold change between normalized input and output sample RPKMs for each annotated gene was calculated using the Set Up Experiment function of CLC Genomics Workbench. Statistical analysis of the Tn-seq data was done using the On Proportions function of CLC and the Baggerley's post-test was used to generate adjusted q-values for each gene. Heat maps representing the input/output fold change for flagella-associated genes generated by MeV software (http:// www.tm4.org/mev.html).
A similar protocol was employed to identify genes essential for survival in an intraperitoneal model of infection. The B. dolosa transposon mutant library was grown overnight in 25 mL of TSB + 75 μg/ml tetracycline, then washed twice in sterile PBS. Two aliquots of 1 mL of this culture were used in genomic DNA extractions for the input samples. Four female 6-8 week old FVB mice were inoculated intraperitoneally with 4.2 x 10 8 CFU/mouse. After 48 hours, all mice appeared healthy and a peritoneal lavage, the heart, and the spleen were collected from each mouse. Samples were homogenized if necessary (for tissues) and used to inoculate 25 mL LB medium + 75 μg/ml tetracycline. After growth overnight, 1 ml of culture was used for genomic preparation, library prep, and sequencing as described above. The output samples from the spleen did not produce the threshold number of mapped reads (>1 million) and were not included in the analyses.

Statistics
Unless otherwise stated, statistical analyses were conducted in GraphPad Prism 5.1 using either one-way ANOVAs or t-tests as indicated.

Identification and conservation of the lafA gene
There are currently two genes in the B. dolosa AU0158 genome annotated as flagellin. One of these (encoded at the AK34_RS27500 locus) encodes the flagellin for the polar flagellum that has orthology with other polar flagellin genes in and outside the Bcc cluster. The amino acid sequence corresponding to the gene encoding the second flagellin protein, found at locus AK34_RS07895, bears resemblance to the lateral flagellin in other species and we have designated this gene lafA for lateral flagellin gene A, consistent with field nomenclature. This gene is bordered on one side by a gene encoding a CheY-like response regulator and on the other by other lateral flagella-associated genes. A further inspection of the genome revealed a number of other flagella-associated genes located in three clusters elsewhere in the genome that are similar to genes encoding lateral flagella biosynthetic proteins in other bacteria species (Fig 1  and S1 Table) and share a low similarity with genes encoding the traditional and more common polar flagella. The organization of the genomic loci in B. dolosa is distinct from that of A. hydrophila and V. parahaemolyticus though there are some similarities. For instance, the flgBDEFGHIJKL genes are common located together in polar and lateral flagella systems ( Fig  1); however, the location of the B. dolosa lafA gene is found next to the fliM gene rather than next to the lafB gene as in A. hydrophila or the fliD gene (a homolog of lafB) in V. parahaemolyticus. The B. dolosa lateral flagella system also seems to lack lateral flagella-specific genes encoding FliL, FliO, FlhF, FlhG, FliT, FlhC, and FlhD (S1 Table) based on the current annotation and BLAST searches of the nucleotides. FliL is required for flagella rotation [67] while FliO is not required for flagellar biosynthesis [68]. The FlhC, FlhD, FlhF, and FlhG proteins are all transcription factors. It is possible that the polar flagella homologs to these proteins can work with both the lateral and polar flagella systems to lead to motility. Alternatively, the proteins involved in any of the three Type 3 Secretion Systems might also serve roles in the B. dolosa lateral flagellin system.
To determine the conservation of presence of the lafA gene in the outbreak strains and other B. dolosa strains, genomic DNA was prepared from 32 outbreak B. dolosa isolates from 11 patients (some patients were sampled multiple times) [70] and 9 additional non-outbreak B. dolosa isolates and used as a template for PCR with primers specific for the lafA gene (Table 1). In parallel, we also prepared and used genomic DNA from B. cenocepacia strain MC0-3 which has a homologous lafA gene based on the genome sequence (even though similar homologs are not found in the other sequenced B. cenocepacia strains). The lafA gene was found in almost all outbreak B. dolosa isolates (Fig 2). It appears to be divergent/absent in 2 outbreak isolates (AU13412 and AU4083) which were obtained from two different patients at two different times during the outbreak. AU4083 was isolated from a patient who later produced an isolate containing the gene which may suggest that there is heterogeneity in the chronic population. The lafA gene is also found in most non-outbreak clinical isolates and interestingly, this gene is also found in the sole environmental B. dolosa strain LMG 21443  [45,69] and V. parahaemolyticus [46] are shown. Gene names are given above the groupings and polar flagella genes begin with "fl" while lateral flagella genes begin with "lf" or "laf". The first B. dolosa gene in each region has its AK34_RS designation shown. For example, the first gene in the lateral flagella gene set is shown as 27345 for AK34_RS27345. Conservation was determined dynamically using polar flagella homologs as thresholds for e-values. Groups of genes are color coded based on the A. hydrophila clusters. Asterisks indicate currently unannotated genes in the published B. dolosa genome sequence. https://doi.org/10.1371/journal.pone.0189810.g001 The putative lateral flagella of the cystic fibrosis pathogen Burkholderia dolosa which was isolated from the rhizosphere of a maize plant in Senegal. These results together suggest that this gene is found frequently in the B. dolosa species and is not just a feature of the outbreak strain.

Amino acid conservation of LafA
To further demonstrate that the LafA protein encoded at the AK34_RS07895 locus was distinct from the polar flagellin, FliC, we analyzed the protein sequence by comparing it to others in the GenBank and burkholderia.com databases. First, the amino acid sequences for B. dolosa AU0158 FliC and LafA were aligned and they appear distinct at many residues through there are some similarities, including sites known to be important for TLR5-dependent host immune recognition in other species [71][72][73][74][75] (Fig 3). Additional lateral flagellin proteins were identified in other species and compared to the sequence of the predicted LafA protein using a multi-species alignment (S1 Fig). The B. dolosa AU0158 LafA protein shows high similarity to a protein in another sequenced B. dolosa isolate PC543 and a homolog in the related Bcc species Burkholderia ubonensis. LafA homologs are also observed in: one Burkholderia cenocepacia strain MC0-3 which was isolated from a maize rhizosphere in Michigan; B. cepacia isolates; B. territorii which has been found in waters [76]; and B. diffusa which has been found in both The putative lateral flagella of the cystic fibrosis pathogen Burkholderia dolosa human and natural environments [77]. In addition, proteins homologous to B. dolosa LafA are found in a number of other β-proteobacterial species including Burkholderia thailandensis, Polyangium brachyosporum, Chromobacterium violaceum, Pseudoxanthomonas sp., Duganella sp., Janthinobacterium lividium, and Roseateles depolymerans. It also has a limited amino acid similarity to lateral flagellins in the γ-proteobacteria species Pseudomonas stutzeri, Vibrio The putative lateral flagella of the cystic fibrosis pathogen Burkholderia dolosa parahemolyticus, and Aeromonas hydrophila (S1 Fig). The B. dolosa LafA proteins are distinct from other LafA-containing strains with at least 16 amino acids that are unique to B. dolosa (S1 Fig). We used this protein alignment to assess the phylogeny of this protein using a phylogenetic network. The inclusion of FliC, the polar flagellin monomer, from B. dolosa AU0158 and PC543, Burkholderia pseudomallei and thailandensis, and A. hydrophila served as outgroups. The LafA protein from B. dolosa was most similar to LafA proteins in other Burkholderia strains followed by the rest of the β-proteobacteria (Fig 3). The LafA proteins from γ-proteobacteria clustered separately as did the polar flagella FliC proteins. This suggests that the LafA from B. dolosa is, in fact, distinct from the polar flagella and may play a role in non-host environments since it is also found in related, environmental, non-pathogenic species.
To further analyze the lateral flagellin conservation, a whole genome alignment was analyzed through the Integrated Microbial Genomes VISTA software. B. dolosa appears to contain a large region not found anywhere in most other Burkholderia genomes (S2 Fig). This region includes three of the lateral flagellin loci as well as genes encoding hypothetical proteins and parts of a Type III Secretion System similar to those found in mammalian pathogens such as Salmonella. There is little but some homology of fragments from the B. dolosa AU0158 genome in this region with other sequenced Burkholderia species, especially B. cenocepacia MC0-3 which has the greatest conservation with B. dolosa AU0158 of all the species examined. One of the regions surrounding the lafA locus is orthologous to regions located on chromosome 2 in other Bcc species while the other border region is more similar to regions found on chromosome 1 in those genomes suggesting that B. dolosa may have obtained this island from another source and integrated it during a genomic rearrangement event. The GC content of these unique region/island is consistent (64% to 67%) with the genome overall (67%- [11]).

The effect of mutations in the lateral flagellin gene in vitro
To further assess the function of the protein encoded by the lafA gene, markerless in-frame mutations were created in AK34_RS27500 (fliC; [50]) and AK34_RS07895 (lafA). In addition, a lafA complemented strain was also created. We tested the effect of these mutations in four agar concentrations representing a gradient of swimming and swarming conditions. The B. dolosa fliC mutant did not swim past the diameter of the original spotted culture in all agar concentrations and thus was completely non-motile under these conditions ( Table 2). The lafA mutant, however, swam significantly farther than the wild-type strain in all agar conditions suggesting that the presence of LafA inhibits swimming motility by the polar flagella. The The putative lateral flagella of the cystic fibrosis pathogen Burkholderia dolosa lafA complemented strain was not significantly different than the wild-type strain under any condition and was significantly lower than the uncomplemented lafA mutant suggesting that the swimming phenotypes and all subsequent phenotypes can be attributed solely to the loss of the lafA gene. At the highest concentration of agar, the motility defects of the ΔfliC strain were not significantly different than the wild-type but the ΔlafA uncomplemented strain still showed a higher degree of motility than wild-type.
The contribution of LafA to host cell internalization was analyzed by infecting RAW264.7 murine macrophages with wild-type B. dolosa AU0158 or the B. dolosa lafA deletion mutant. There were no significant differences between the wild-type and mutant strain in four replicate experiments (S3 Fig). Likewise, there were no significant differences in biofilm formation between the wild-type and lafA mutant strains based on crystal violet staining (S4 Fig). These data suggest the putative lateral flagella do not play a role in macrophage invasion/phagocytosis or in biofilm formation.
The production of flagella was also examined using transmission electron microscopy. When grown to stationary phase in a nutrient rich liquid medium, wild-type B. dolosa has few attached flagella but when observed, there is a single flagellum extending from one end (Fig 4). Because there are a number of flagella observed in the background of these images (S5 Fig), it is likely they were shed upon entering stationary phase. On an agar medium appears to have a few polar flagella extending from a pole in a manner more similar to a lophotrichous arrangement rather than a peritrichous arrangement (Fig 4). The higher number of flagella found on solid surfaces has also been observed in Burkholderia glumae, a plant pathogen that lacks a homolog to the lafA gene, and the number and localization of the flagella in this organism has been shown to be condition-dependent [78].
The B. dolosa ΔfliC mutant lacks a detectable flagellum in liquid or on solid surfaces (Fig 4). We note here that under low agar conditions (concentrations of 0.3-0.5%) and growth at 30˚C, the B. dolosa ΔfliC mutant is able to observably move both on agar plates and using wet mounts under bright field microscopy. Whether this is due to Brownian motion, Type IV pilimediated twitching motility (for which B. dolosa AU0158 has the genes), or the function of lateral flagella remains to be determined.
Interestingly, the B. dolosa ΔlafA mutant appears to have a 1-2 flagella when grown in liquid (Fig 4) but shows 2-8 flagella extending from a single pole when grown on an agar medium (Fig 4). This strain, however, should not produce lateral flagellin monomers which suggest that these may be extra polar flagella. Together the data shown in Fig 4 suggest that the ΔfliC mutant is not motile due to a lack of lateral and polar flagella under these conditions while the extended swimming phenotype of the ΔlafA mutant appears to be due to the overproduction of flagella in these cells. The number of flagella extending from each cell was quantitated from 4-12 images taken in replicate trials (examples shown in S5 Fig) and this data is presented in Table 3. B. dolosa AU0158 has the ability to persist in mouse lungs for more than 6 days without causing symptoms even at high doses [38]{Roux, 2017 #7359}. To determine if the lateral flagella plays a role in host association and persistence in vivo, the wild-type and lateral flagella deletion mutant strains were inoculated intranasally into 4-5 C57BL/6 mice and monitored over the course of 6 days. At 1, 6, 24, and 144 hours post-infection, bronchoalveolar lavages (BALs) were collected for bacterial survival analyses and cytokine expression and then the lungs were removed and homogenized for bacterial counts. P. aeruginosa PAO1 served as a positive control because its behavior and cytokine recruitment has been well-described in the published literature. As a negative control, mice inoculated with the PBS vehicle alone were run in parallel.

The effect of mutations in the lateral flagellin gene in vivo
As expected, the wild-type B. dolosa strain persisted for up to 24 hours in the BAL and 144 hours in the lung tissue, presumably due to their intracellular nature in host cells (Fig 5). The lafA mutant was found at similar levels such that there were no significant differences in bacterial viable counts between the wild-type and lafA deletion mutant strain at most time points other than 1 hour after infection, when counts of the lafA deletion mutant were about 5-fold lower than the wild-type strain in the lung tissue. Similarly, the expression of 13 different cytokines was not significantly different between wild-type and lafA deletion mutant strains in  The putative lateral flagella of the cystic fibrosis pathogen Burkholderia dolosa The putative lateral flagella of the cystic fibrosis pathogen Burkholderia dolosa almost all cases (Fig 6) suggesting that lafA is not playing a significant role in host response survival in vivo.
It has been noted in prior studies that wild-type B. dolosa does not produce robust cytokine expression in the murine lung possibly due to a suppression of the immune response; however, it can produce a pro-inflammatory cytokine responses far above the limit of detection in both human and murine cell lines [50]. To test whether the B. dolosa lafA deletion mutant is altered in its ability to induce a cytokine response in cultured or primary cells, we stimulated RAW264.7 murine macrophages with B. dolosa AU0158 wild-type and the lafA deletion mutant strains and measure the levels of the pro-inflammatory cytokines TNF and MIP-2 (an IL-8 homolog). The production of cytokines from this cell line was moderate but detectable after bacterial infection and there were no significant differences between the wild-type and mutant strains (Fig 7). This agrees with previous studies in which B. dolosa produces a moderate cytokine production in these cells that is not significantly different from that produced by a mutant lacking the polar flagellum [50].
We also tested the effect of the lafA mutant on eliciting cytokines from human peripheral blood mononuclear cells (PBMCs). Interestingly, the TNF and IL-8 cytokine responses in PBMCs were robust, and significant differences were observed between the wild-type and lafA deletion mutant strains (Fig 7), with the lafA deletion mutant inducing 1.4-fold less TNF and 1.7-fold less IL-8 production from host cells, suggesting that LafA plays a role in recognition by the host immune response. Taken together with previous data, this suggests that the role of LafA in host immune recognition may be masked in vivo by an unknown mechanism as was proposed previously for the polar flagellin [50].

The essentiality of the lateral flagella in vivo
To determine if the lateral flagella was essential for colonization in vivo, we created a transposon mutant library of B. dolosa AU0158 and used this library as an input for mouse lung colonization or septic dissemination via intraperitoneal injection. After 48 hours inside the mouse, the mutant libraries were cultivated from the tissues and the frequency of all transposon mutants was measured by counting the number of transposon-junction fragments in the input and output pools. Those genes with reads that were significantly reduced (fold changes < -2) in the outputs are deemed important under these conditions as mutations in these genes did not survive as well as bacteria containing mutations in other genes. Those mutants found at higher levels in the outputs (fold changes > 2) are usually thought to contain transposon insertions in genes encoding proteins or regulatory RNAs involved in immune recognition; thus mutations in these genes will survive better than the general population as they will be less well recognized by the host.
We observed that most genes involved in lateral flagella synthesis (34 total) were not required for colonization of the mouse lung to a high degree (Fig 8) with most fold changes between -1 and -1.55. For only 11 genes was this decrease significant (q values <0.05) suggesting that the lateral flagella may only play a weak or indirect role, if any, in murine lung colonization. One exception to the general downward trend of these mutants is that mutations in AK34_RS07560, which encodes putative lateral flagella homolog to the basal body rod protein (FlgB in polar flagella), that showed a 2.42-fold increase between output and input in the lung and a 2.99-fold increase in the peritoneum, suggesting that interruption of this gene may improve survivability in the murine lung and peritoneum.
In the peritoneal model, two outputs were evaluated: a peritoneal lavage and heart tissue. For almost no gene was there a significant decrease based on adjusted q values even though there fold changes between -5.85 and 13.92. This was due to large variability in these data, so it is difficult to make conclusions whether lateral flagella are playing a role in lung colonization or septic dissemination under these conditions. This is in contrast to the data for the polar flagella system of B. dolosa. There is a small but significant effect of polar flagella mutations in the murine lung with 24 of the 38 associated genes showing a 2-fold increase or more after colonization, all of which are significant (Fig 8).
In the septic dissemination model, there are 20 and 18 genes that show an increase of 2-fold or more in the peritoneum and heart outputs respectively though very few of these are significant. These data are consistent with expectations that polar flagella mutants may fair better in lungs and the blood stream as they will be less recognized by the TLR5 host immune receptor and thus cleared less rapidly. It also suggests that mutations in the gene encoding the polar flagella biosynthetic proteins are fundamentally different than those encoding the genes encoding the lateral flagella biosynthetic components.
We further looked at additional genes that were required for murine lung colonization that were putatively not involved in flagella production. Of the 293 genes that showed a fold change greater than 2 and had q-values <0.05, there are a number of genes encoding proteins involved in phenylacetate and anthranilate metabolism, pilus assembly, and sensory or transcriptional signaling systems (S2 Table). Mutations in the gene encoding the LexA transcriptional repressor seemed to be particularly unfit in the lung colonization model as this gene was under-represented 55-fold in output compared to input samples.
Also, the 6 genes that encode the TolQRAB-Pal outer membrane complex (BDAG_02395-2400) are all highly required for mouse lung colonization. In other species, this complex is known to play a role in membrane integrity, antibiotic resistance, and cell division. However, the role of this complex in virulence or in host colonization has not been described to date so whether the requirement for this complex is direct (truly deficient in lung colonization) or indirect (generally sick) remains to be explored.
Genes required for septic dissemination were also examined and we found only 182 genes that showed a 2-fold change and q-values of <0.05 between out and input samples (S3 Table). Within this group, there are genes required for peritoneal colonization and septic dissemination into the heart including components of the Flp pilus, transcriptional regulators, and The putative lateral flagella of the cystic fibrosis pathogen Burkholderia dolosa several major facilitator family proteins, which are usually involved in membrane transport of small molecules. Conversely, mutated genes that were over-represented in the outputs seemed to have no patterns in terms of functional groupings or proximity of genetic loci outside of the genes encoding flagella-related proteins. Together, these in vivo results suggest that there are a number of genes required for lung colonization and these genes are different than those required for a septic dissemination/intraperitoneal colonization model. These genes will need to be explored in future studies. The putative lateral flagella of the cystic fibrosis pathogen Burkholderia dolosa

Discussion
The discovery of the lateral flagellin system in B. dolosa is surprising in that it has yet to be reported in any other Bcc species, despite the plethora of genomes available for this group. Our analysis also shows that the genetic loci that comprise the presumptive lateral flagella system in B. dolosa is unique compared to that in the well-characterized pathogen A. hydrophila in that it is not found in one single chromosomal locus but rather is spread over three regions on the second chromosome (Fig 1). A survey of additional literature also shows that it is distinct from other organisms bearing lateral flagella systems. For example, Vibrio parahemolyticus has two regions containing all genetic loci required for lateral flagella (Fig 1), one on the major chromosome and one on the minor chromosome [46,79], and the order and composition is different from what is observed in B. dolosa. For example, there does not appear to be a motY L gene in B. dolosa. Similarly, differences are observed in the order and composition of lateral flagella genes in other A. hydrophila and A. caviae strains [45,49], and in the pathogenic 042 strain of E. coli, which also bears a lateral flagella system [80].
The bioinformatic finding that there is a B. dolosa LafA homolog in the soil-, water-, and plant-associated β-proteobacteria Duganella, Janthinobacterium, Pseudoxanthomonas, Roseatales, Polyangium, and Chromobacterium suggests that the presence of this system is perhaps more ancestral than previously recognized. This observation is also supported by the fact that the B. dolosa LafA is similar to other Burkholderia LafA homologs and distinct from those found in γ-proteobacterial strains (Fig 3) suggesting it was not recently acquired in this species.
The lafA gene is notably absent from most Bcc isolates despite the plethora of genomes available for organisms within this complex. This could suggest that perhaps B. dolosa is more ancestral than other members of the complex which may have lost these genes over time. Since the resolution of the phylogeny of the Bcc is hampered by their high genetic relatedness, the true phylogeny of B. dolosa cannot be inferred accurately yet; thus, it is difficult to determine if it is less derived than other Bcc species. Conversely, it could be that B. dolosa acquired these genes from a related, high GC species shortly after diverging from the rest of the Bcc. Homologous genes are found in at least one isolate of B. cenocepacia (MCO-3) and in only a few other Burkholderia species, including B. thailandensis and B. pseudomallei, which would support the hypothesis that this gene was lost from many strains over time as it is less likely to be acquired by many phylogenetic lines through horizontal gene transfer.
There have been a number of studies that examine the hierarchical transcriptional regulation of the lateral flagella systems in A. hydrophila, V. parahemolyticus, and E. coli 042. In these species, the σ 54 and σ 28 proteins (encoded by the rpoN and fliA genes), along with the transcriptional regulators FlhD, FlhC, and FleQ, coordinate the expression of the late genes involved in flagella maturation [46,81]. B. dolosa has multiple homologs of rpoN (AK34_RS14460 and AK34_RS10110), fliA (AK34_RS27345 and AK34_RS07625), and fleQ (AK34_RS04955 and AK34_RS04950) but only one homolog each that shows similarity to flhD and flhC. This suggests that perhaps B. dolosa shares some regulators between the polar and lateral flagella-a phenomenon also observed in V. parahemolyticus and Azospirillum brailense [82,83] but not in Aeromonas species [47,81]. The change in the production of the polar flagella observed by transmission electron microscopy supports the notion that the regulation of expression for these two flagellar systems is intertwined as a deletion in the lafA gene leads to the overproduction of the polar flagella (Fig 4) and an increase in the swimming ability of this strain compared to wild-type (Table 2).
Using transmission electron microscopy, the polar flagella of B. dolosa is observed in the wild-type and mutant strain lacking the lateral flagellin gene though the localization and/or number of flagella differs in the latter. The lack of observable lateral flagella in the polar flagellin mutant strain is surprising given the putative role for lateral flagella in movement over semi-solid surfaces and could suggest that either the conditions for our assays will not allow mature lateral flagella production, that the lateral flagella of B. dolosa cannot be stained by uranyl acetate, or the lateral flagella in B. dolosa are used for other purposes beside swarming, such as movement inside host cells. The overproduction of flagella in the B. dolosa ΔlafA mutant is also somewhat surprising as this strain lacks the gene for the monomer that comprises the putative lateral flagella. This suggests that the additional flagella we observe are either mis-regulated polar flagella or that the FliC flagellin monomer for the polar flagella can be assembled on the lateral flagella machinery; however, the latter option seems less likely given the number of amino acid differences in these two flagellin proteins (Fig 2).
The role of the lateral flagella is thought to be beneficial for human pathogens because it may allow for greater attachment to host cells [48,49,[84][85][86][87], better biofilm formation [48,49,88], and thus better colonization of hosts. In most cases, it is assumed that the polar flagella will allow the bacterium to travel greater distances to find new hosts. Once bound, the lateral flagella may facilitate spread of the bacterium on the surface of the host tissues. Our assays showed that the B. dolosa lafA gene is required neither for biofilm formation (S4 Fig) nor internalization by murine macrophages (S3 Fig). This is inconsistent with the published studies in Aeromonas species in which it was shown that the lateral flagella are required for biofilm formation [48,49]. The reason for this discrepancy is not clear though we speculate that perhaps the long divergence between the γ-proteobacteria Aeromonas species and the β-proteobacteria B. dolosa has also contributed to a divergence in the function of the lateral flagella.
While the function of the B. dolosa lateral flagella is not known, a study from French et al. has shown that the proposed lateral flagella of the closely related B. thailandensis contributes to intracellular swimming but not swarming on soft agar [89]. In this previous study, time lapse images show the effect of mutation in the motor component of this system and their striking inability to move inside cultured cells in contrast to the wild-type. The LafA protein located at locus BTH_II0151 in B. thailandensis E264, is~54% identical to the predicted lateral flagellin proteins located at the AK34_RS07895 locus in B. dolosa (lafA), the Bcenmc03_4651 locus of B. cenocepacia MC0-3, and has even higher homology to proteins encoded in some, but not all, B. pseudomallei strains (>96% identity). So, even though the select agent, BSL-3 pathogen B. pseudomallei has been extensively studied for virulence factors, the lateral flagella have yet to be implicated in human pathogenesis. In our study, a strain lacking the lateral flagellin gene was not defective in murine lung colonization (Fig 5), macrophage internalization (S3 Fig), or septic dissemination (Fig 8) which we speculate could be because the function of this particular lateral flagella system is not for host cell adherence, invasion, or swimming in liquids such as blood by this pathogen.
Very little is known about B. dolosa specifically, and even less is known about its ability to cause disease or why it caused an outbreak at Boston Children's Hospital. Our study showed an effect for the B. dolosa lateral flagellin deletion mutant that was not observed for a B. dolosa polar flagellin deletion strain-in primary human PBMCs, we observed a modest but significant decrease in the cytokine response to the lateral flagellin deletion mutant compared to the wild-type strain. We suspect this may be due to the ability of host receptors to recognize polar flagellins, which bear some homology to the lateral flagellin. However, it was previously shown that B. dolosa fliC mutants were not significantly different from wild-type in terms of cytokine production [50], suggesting that they have either evolved a polar flagellin that is incapable of being recognized by the TLR5 or there are other factors that participate in the host sensing of lateral flagella. It should be noted that the use of wild-type mice to simulate cystic fibrosis-like conditions is not ideal even for examining virulence or immune responses in CF pathogens such as B. dolosa. It was previously shown that macrophages from CF patients are deficient in autophagy, fail to efficiently clear CF pathogens including Burkholderia cepacia complex isolates [90][91][92][93][94], and exhibit a heightened cytokine production associated with defects in the CFTR gene [95]. This may contribute to the lack of functional cohesion in the genes required for lung and peritoneal colonization in the Tn-seq screen (S2 and S3 Tables) as many of the genes had no seemingly obvious role in host cell attachment or the stress response as might be expected in a CFTR deficient murine model. The basis of future experiments will focus on the ability of B. dolosa AU0158 strain to invade and replicate inside human CFTR-deficient macrophages and mice, to better understand the role of the TolQRAB complex in lung colonization, and the potential production, uses, and regulation of the lateral flagellin system of B. dolosa in vitro and in vivo.  Table. Genetic loci for the polar and putative lateral flagella of B. dolosa AU0158. Gene annotations for putative polar and lateral flagella are given in both the BDAG and AK34 designations. Genes are grouped by function based on other polar flagella systems. Genes not homologous to known flagellar proteins that were not found in the current B. dolosa annotations are listed as not found. Genes that appear to be present but are not currently annotated are listed as unannotated. (PDF) S2 Table. Genes required for murine lung colonization based on Tn-seq analysis. A B. dolosa transposon mutant library was inoculated into the lungs of C57Bl/6 mice. After 48 hours, the library was harvested and junction fragments from input and output libraries were sequenced, mapped to the B. dolosa AU10158 genome, and normalized levels compared. Those with ! 2-fold changes and q values 0.05 are shown. (PDF) S3 Table. Genes important for septic dissemination in a murine peritoneal model of infection. A B. dolosa transposon mutant library was inoculated into the peritoneum of C57Bl/6 mice. The library was harvested from the peritoneal cavity and the heart and junction fragments from input and output libraries were sequenced, mapped to the B. dolosa AU10158 genome, and normalized levels compared. Those with ! 2-fold changes and q values 0.05 are shown.