Burkholderia collagen-like protein 8, Bucl8, is a unique outer membrane component of a tetrapartite efflux pump in Burkholderia pseudomallei and Burkholderia mallei

Bacterial efflux pumps are an important pathogenicity trait because they extrude a variety of xenobiotics. Our laboratory previously identified in silico Burkholderia collagen-like protein 8 (Bucl8) in the Tier one select agents Burkholderia pseudomallei and Burkholderia mallei. We hypothesize that Bucl8, which contains two predicted tandem outer membrane efflux pump domains, is a component of a putative efflux pump. Unique to Bucl8, as compared to other outer membrane proteins, is the presence of an extended extracellular region containing a collagen-like (CL) domain and a non-collagenous C-terminus (Ct). Molecular modeling and circular dichroism spectroscopy with a recombinant protein, corresponding to this extracellular CL-Ct portion of Bucl8, demonstrated that it adopts a collagen triple helix, whereas functional assays screening for Bucl8 ligands identified binding to fibrinogen. Bioinformatic analysis of the bucl8 gene locus revealed it resembles a classical efflux-pump operon. The bucl8 gene is co-localized with downstream fusCDE genes encoding fusaric acid (FA) resistance, and with an upstream gene, designated as fusR, encoding a LysR-type transcriptional regulator. Using RT-qPCR, we defined the boundaries and transcriptional organization of the fusR-bucl8-fusCDE operon. We found exogenous FA induced bucl8 transcription over 80-fold in B. pseudomallei, while deletion of the entire bucl8 locus decreased the MIC of FA 4-fold in its isogenic mutant. We furthermore showed that the Bucl8 pump expressed in the heterologous Escherichia coli host confers FA resistance. On the contrary, the Bucl8 pump did not confer resistance to a panel of clinically-relevant antimicrobials in Burkholderia and E. coli. We finally demonstrated that deletion of the bucl8-locus drastically affects the growth of the mutant in L-broth. We determined that Bucl8 is a component of a novel tetrapartite efflux pump, which confers FA resistance, fibrinogen binding, and optimal growth. Author Summary Burkholderia pseudomallei and Burkholderia mallei are highly infectious and multidrug resistant bacteria that are classified by the National Institute of Allergy and Infectious Diseases as Tier one select agents partly due to the intrinsic multidrug resistance associated with expression of the efflux pumps. To date, only few efflux pumps predicted in Burkholderia spp. have been studied in detail. In the current study we introduce Bucl8, an outer membrane component of an unreported putative efflux pump with a unique extended extracellular portion that forms a collagen triple helix and binds fibrinogen. We demonstrate Bucl8’s role in fusaric acid resistance by defining its operon via bioinformatic and transcriptional analyses, as well as by employing loss-of-function and gain-of-function genetic approaches. Our studies also implicate the Bucl8-associated pump in metabolic and physiologic homeostasis. Understanding how Bucl8 efflux pump contributes to Burkholderia pathology will foster development of pump inhibitors targeting transport mechanism or identifying potential surface-exposed vaccine targets.

Infectious Diseases as Tier one select agents partly due to the intrinsic multidrug 51 resistance associated with expression of the efflux pumps. To date, only few efflux 52 pumps predicted in Burkholderia spp. have been studied in detail. In the current study 53 we introduce Bucl8, an outer membrane component of an unreported putative efflux 54 pump with a unique extended extracellular portion that forms a collagen triple helix and 55 binds fibrinogen. We demonstrate Bucl8's role in fusaric acid resistance by defining its 56 operon via bioinformatic and transcriptional analyses, as well as by employing loss-of-57 function and gain-of-function genetic approaches. Our studies also implicate the Bucl8-58 associated pump in metabolic and physiologic homeostasis. Understanding how Bucl8 59 Introduction BurkhFusBCD-1R) and cloned between two unique sites in bucl8, XcmI, and FseI, of 207 pSL525. E. coli JM109 transformants were isolated on a LA medium containing 100 208 μg/mL ampicillin. The plasmid sequence was verified as before. 209

Cloning, expression and purification of Bucl8-derived recombinant proteins 210
Two recombinant polypeptides, derived from the presumed extracellular portions of 211 Bucl8 variant in Bp K96243, were generated for this study: (i) pSL521-encoded rBucl8-212 CL-Ct polypeptide, containing both the collagen-like region and the non-collagen C-213 terminal region and (ii) pSL520-encoded rBucl8-Ct, which only includes the C-terminal 214 region. 215 For cloning, gBlocks (Integrated DNA Technologies) were designed, encoding two 216 recombinant constructs (Table S3), as described [28]. gBlocks were used as templates 217 to produce cloned DNA inserts using primers pSL521-F and pSL521-2R for pSL521 218 construct, and pSL520-F and pSL520-R for pSL520. gBlock DNA fragments were 219 inserted between HindIII and BamHI sites of the pQE-30 vector, resulting in an N-220 terminal 6xHis-tag (Qiagen) for each construct and were then cloned into E. coli JM109. 221 Plasmid constructs pSL520 and pSL521 were confirmed by sequencing (Primers 222 pQE30-F, pQE30-2R). 223 For protein expression, E. coli JM109 with pSL520 or pSL521 constructs were grown 224 in LBM plus 100 μg/mL ampicillin with shaking at 37 C overnight, and then 10 mL Circular dichroism spectroscopy of rBucl8-derived polypeptides was performed as 257 previously described [28]. Briefly, protein samples were dialyzed against 1x Dulbecco's 258 phosphate buffered saline, pH 7.4. CD spectra were taken with a Jasco 810 259 spectropolarimeter, in a thermostatically controlled cuvette, with a path length of 0.5 cm. consisting of glycine, alanine, and serine (GAS) n triplet repeats, here denoted as the CL 326 domain, which is followed by a non-collagen carboxyl-terminal (Ct) region (Fig 1A). 327 Bucl8 is a homotrimeric molecule, which supports triple-helical structure of the 328 extracellular CL-(GAS) n domain, although, the specific (GAS) n sequence has not been 329 studied for triple helix formation. The number of consecutive (GAS) n repeats present 330 fluctuates between Bucl8 variants from different B. pseudomallei isolates. Analysis of 331 ~100 bucl8 alleles showed (GAS) n numbers ranging from 6 to 38 repeats (mode: 20). 332 Notably, 21 consecutive GAS repeats characterize the Bucl8 of B. pseudomallei model 333 strain K96243, while the Bucl8 variants of the strains utilized in this study have fewer 334 (GAS) n numbers, e.g., Bp 1026b/Bp82 has six and B. mallei strain Bm ATCC 335 23344/CLH001 has eight. Following the CL-(GAS) n domain is a Ct region of 72 amino 336 acids that are conserved among B. pseudomallei and B. mallei strains. (File S1) 337 Here, we homology-modelled a representative (GAS) 19 sequence using the structure 338 of the collagen peptide (PPG 10 ) 3 as a template (PDB code 1k6f, seqid 36%) [31] and 339 the software MODELLER (Fig 1B). This structure formed a triple helix of about 163 Å in 340 length. On its C-terminal end, the Ct domain of each chain is predicted by JPRED to be 341 unfolded and was modeled in a random coil conformation (Fig 1B). Consistent with the 342 sequence composition of the (GAS) n repetitive domain, its electrostatic potential surface 343 appears neutral, with only a few positive charges due to the presence of arginine 344 residues in the unstructured Ct regions of the molecule (Fig 1B). 345 To experimentally validate this homology-modelled structure, two recombinant 346 proteins, derived from the extracellular portion of Bucl8 variant in strain Bp K96243, 347 were designed and expressed in E. coli. The construct rBucl8-CL-Ct includes the CL-348 (GAS) 19 domain and adjacent unstructured C-terminus (Ct), while construct rBucl8-Ct 349 encompasses the Ct region only. Both Bucl8-derived polypeptides migrate aberrantly in 350 SDS-PAGE in relation to molecular weight standards, e.g., rBucl8-CL-Ct of expected 351 11.7 kDa and rBucl8-Ct of 7.8 kDa (Fig 1C). Structural analysis of rBucl8-CL-Ct 352 rendered at 25 o C, using circular dichroism spectroscopy, confirmed a triple helical 353 structure, demonstrated by a shallow peak at 220 nm (Fig 1D). As a control, denatured 354 rBucl8-CL-Ct (50 o C line) displayed a further-depressed peak at 220 nm that no longer 355 held a triple-helical collagen structure. The 220 nm peak in rBucl8-CL-Ct is less 356 pronounced when compared to typical triple helices formed by perfect GPP collagen 357 repeats. This feature suggests the coexistence of both triple helix and random coil 358 structures and/or the contribution of the non-collagen Ct region to the spectrum; such 359 effects on CD spectra were previously reported for streptococcal collagen-like rScl 360 constructs [41]. Additionally, the rBucl8-Ct structure was also analyzed by circular 361 dichroism spectroscopy. The absence of ellipticity maxima and/or minima of known 362 structures, e.g., α-helices or β-strands [42], indicates an unstructured protein (Fig 1D). 363 Altogether, using in silico modeling and experimental CD spectroscopic analyses of the 364 representative recombinant protein, we demonstrated that repeating (GAS) n of the 365 predicted Bucl8-CL region from B. pseudomallei and B. mallei can form a stable 366 collagen triple helix; to our knowledge, this is the first such demonstration obtained for 367 the unusual repeating (GAS) n collagen-like sequence. 368 Bacterial proteins harboring CL domains from diverse genera have been 369 demonstrated to bind ligands, including extracellular matrix proteins (ECM), and have 370 been shown to participate in pathogenesis [43][44][45]. Here, we screened several human 371 compounds by ELISA to ascertain a potential ligand binding function of Bucl8's 372 extracellular region, rBucl8-CL-Ct; ligands included fibrinogen, collagen-I and IV, elastin, 373 plasma and cellular fibronectin, and vitronectin. Of the ligands tested, rBucl8-CL-Ct 374 construct showed significant binding to fibrinogen, but not to collagen I and elastin ( Fig  375   2A), while binding to other ligands tested was also not significant (not shown). rBucl8-376 CL-Ct binding to fibrinogen-coated wells was concentration-dependent in contrast to 377 control BSA-coated wells. In addition, rBucl8-Ct construct showed limited level of 378 binding to fibrinogen in this assay (Fig 2B). 379 LysR-type regulators; therefore, we hypothesized bucl8 transcription to be regulated by 403 the fusR product. Using predictive software and analysis of transcriptome data, the 404 promoters, transcription initiation sites (TIS), and FusR binding sites were identified in 405 the intergenic region between fusR and bucl8 (Fig 3B).   Sub-inhibitory concentrations of FA, e.g., 1000 µM for Bp82 and 60 µM for CLH001 464 that did not inhibit the growth rates were used in subsequent induction experiments (Fig  465   4A). Total RNA was isolated from the cultures of Bp82 and CLH001 that were either 466 non-treated or treated with FA (1000 µM or 60 µM, accordingly) at OD 600 ~0.4 for one 467 hour. Both fusR and bucl8 genes were expressed in non-treated cultures at basal 468 levels, but transcription of bucl8 in Bp82 was significantly induced with FA by an 469 average 82-fold change in relative expression and a 20-fold change of fusR, using 2 Δ Δ Ct 470 calculations (Fig 4B). CLH001 also demonstrated about a four-fold increase for fusR 471 and bucl8 when induced with 60 µM FA (Fig 4C), although this change is comparatively 472 lower than that recorded in FA-induced Bp82.

474
In a following experiment we confirmed the boundaries of the fusR-bucl8 operon by 475 RT-qPCR. Results show that transcription levels of fusR-bucl8-fusC-fusE were all 476 significantly upregulated in samples treated with FA, compared to non-treated controls 477 (fusR = 20-fold ± 1.37; bucl8 = 82-fold ± 8.73; fusC = 40-fold ± 2.84; fusE = 86-fold ± 478 10.65; Fig 4B). In contrast, the expression change of tar was significantly lower than 479 genes from the fusR-bucl8-fusC-fusE operon and the associated regulatory gene fusR 480  Table 1) was constructed in the E. coli vector pMo130, which is 506 suicidal in Burkholderia, to generate an unmarked deletion mutant (Fig 5A). Construct 507 pSL524, carrying upstream and downstream sequences flanking bucl8 locus was 508 transferred to B. pseudomallei Bp82 via biparental mating. Deletion was achieved in a 509 two-step insertion/excision process, as detailed in Materials and Methods section. 510 Successful deletion of the bucl8-fusCD-fusE segment from the chromosome was 511 confirmed by PCR (Fig 5B) and sequencing. We did not delete the fusR gene on 512 purpose to avoid a possible global regulatory effect associated with unknown FusR 513 function. 514 To exhibit gain-of function, we complemented a heterologous E. coli host in-trans 515 with a plasmid construct pSL525 ( Table 1)  by PCR (Fig 5D) and sequencing. Since Bp82 represents the 1026b strain harboring 519 Bucl8 variant with (GAS) 6 repeats in the CL region, we made an additional construct, 520 pSL529, that contains (GAS) 21 repeats, to represent the majority of B. pseudomallei 521 strains, by extending the number of GAS triplets in pSL525. 522 MICs were determined for bacterial growth on LA plates containing FA or pHBA 523 chemicals, ranging from 0 to 800 μ g/mL FA and 500-2,500 μ g/mL pHBA (Fig 6A). There 524 was a 4-fold decrease in MIC to FA from 400 µg/mL to 100 Although deletion of the Bucl8 pump resulted in a drastically decreased MIC, 532 Bp82Δbucl8-fusE mutant still maintained residual level of FA resistance (100 533 μ g/mL).Therefore, we hypothesized that additional proteins annotated as FusC are 534 contributing to the remaining FA resistance recorded in the Bp82Δbucl8-fusE mutant. 535 Within Bp 1026b and K96243 genomes, there are six genes present that are annotated 536 as FusC-type proteins (Pfam #PF04632), including the protein arbitrarily designated as 537 FusC, which is associated with Bucl8, whereas remaining five were designated FusC 2 538 thru FusC 6 ( Table S2). These protein sequences ranged roughly in three different lengths: ~200 amino acids for FusC 3, ~350 for FusC 4 and 6, and ~750 amino acids for 540 FusC, FusC 2 and FusC 5. Upon examination, the loci around FusC genes 2 thru 6 541 were not arranged in as discernable tripartite-pump operons, like FusC, although some 542 were adjacent to either a MFS transporter protein or genes encoding amino acid 543 permeases. To test whether these genes are regulated by FA addition, we performed 544 RT-qPCR on RNA isolated from Bp82 cultures induced with 1000 μ M FA and without 545 treatment. The transcription of fusC 2-6 genes showed little to no fold-change (0-2-fold; 546 Fig 6B) when compared to non-treated samples, which contrasts with ~40-fold 547 difference in fusC transcription (Fig. 4B). Thus, we conclude that these fusC genes are 548 not inducible by FA. 549 In the clinical laboratory setting, the Burkholderia failed to grow in commercial 558 medium, and therefore only the E. coli data were generated. Overall, there was not a 559 significant increase in resistance to any of the antibiotics tested; JM109::525/529 560 showed only increased resistance to the β-lactam antibiotics, which was associated with 561 the resistance gene present on the inserted plasmid. A disc diffusion test, including 562 ampicillin, ciprofloxacin, levofloxacin, tobramycin, gentamicin, tetracycline, doxycycline, 563 and trimethoprim-sulfamethoxazole, resulted in similar zones of inhibition for both Bp82 564 and Bp82Δbucl8-fusE cultures, as well as E. coli JM109 and JM109::525/529, again 565 with the exception of the plasmid-derived β-lactam resistance determinate. 566

Bucl8-pump does not contribute to the multidrug resistance (MDR)
Microarray data comparing the effect of 84 growth conditions on B. pseudomallei 567 transcriptome showed that chloramphenicol (CHL), which contains an aromatic ring in 568 its structure, induced bucl8 expression, thus, implying CHL might be a substrate for 569 Bucl8-associated pump [52]. Here, we determined the CHL-MICs of our B. 570 pseudomallei and E. coli strains using a growth assay on the LA medium; however, the 571 MIC for all the strains was the same (8 µg/mL; Fig 6A). In addition, the exogenous CHL 572 at 8 µg/mL or 4 µg/mL concentrations did not significantly induced the transcription of 573 bucl8-associated genes (not shown). Thus, our results indicate the Bucl8-associated 574 pump is not needed for CHL resistance in B. pseudomallei [49]. 575

576
Efflux pumps extrude a variety of compounds that are toxic to the cells and play 577 physiological functions linked to pathogenesis [12]. We observed the growth of the 578 BpΔbucl8-fusE mutant was considerably reduced than that of the parent Bp82 and did 579 not reach the same OD 600 in the stationary phase (Fig 6C). CFU for Bp82 increased by 580 approximately four logs, while the mutant increased by two logs from hour 0 to 12. (Fig  581   6D). These results suggest that the pump components are needed for normal growth 582 physiology under laboratory conditions in rich medium. 583

584
The protein Bucl8 was previously predicted to be the outer membrane in B.   sensitive to FA with MIC <50 µg/mL. We also observed that our Bp82Δbucl8-fusE 689 mutant retained 100 μg/mL residual resistance to FA. Through transcriptional analysis, 690 we found that the five fusC/FusC genes/proteins outside of the Bucl8-operon showed 691 little to no induction, indicating that the Bucl8 pump is the main contributor to FA 692 resistance in B. pseudomallei. 693 The multidrug resistance in B. pseudomallei is substantially attributed to previously 694 studied RND efflux pumps BpeAB-OprB, AmrAB-OprA, and BpeEF-OprC. At the same 695 time, little is known about the role of FA pumps in resistance against clinically used 696 drugs. In our studies, we assessed the Bucl8-pump's role in multidrug resistance in two 697 ways: (i) we compared the spectrum of resistance between parental strain Bp82 and 698 Bucl8-pump deletion mutant, and (ii) we expressed the bucl8-operon in a sensitive E. 699 coli strain. Although MICs for FA changed as predicted, deletion of the Bucl8-pump did 700 not affect MIC values for the clinically-used drugs. This result is comparative to an FA 701 pump in S. maltophilia, which did not determine the resistance to a large panel of 702 therapeutics tested [50]. At the same time, a different study in the same organism 703 showed that deletion of the pcm-tolCsm operon, encoding a different efflux pump, 704 resulted in decreased MICs for several antimicrobials of diverse classes (β-lactams, 705 chloramphenicol, quinolone, tetracycline, aminoglycosides, macrolides), and also 706 decreased FA resistance [69]. Microarray data suggested bucl8 expression was 707 upregulated in the presence of chloramphenicol [52] and deletion of the tolCsm in S. 708 maltophilia resulted in decreased resistance to both CHL and FA [69]. Both CHL and FA 709 harbor aromatic rings in their structures, however, our investigations did not detect 710 bucl8-operon induction by CHL nor changes in CHL resistance levels in Bp82Δbucl8-711 fusE mutant or complemented E. coli. 712 The decrease in bacterial growth of the Bp82Δbucl8-fusE mutant suggests that the 713 Bucl8-pump may be involved in modulating essential cellular stresses, both in the 714 environment and in infected human host [12]. Limited studies show that FA repressed 715 quorum sensing genes, expression of stress factors, secretion of siderophores, 716 production of anti-fungal metabolites, and iron uptake [70][71][72][73]. Additionally, Bucl8 pump 717 may be involved in a transport of aromatic carboxylic acid compounds and act as a 718 pHBA-metabolic efflux valve [38]. Further investigation will be needed to determine what 719 cellular processes are associated with the Bucl8-pump. 720 In summary, we conclude that Bucl8 is a component of a previously unreported 721 tetrapartite efflux system that is involved in FA resistance and cell physiology. We have 722 demonstrated that the extracellular Bucl8-CL domain forms the prototypic collagen 723 triple-helix, while the extracellular Bucl8-CL-Ct portion is capable of binding to 724 fibrinogen. Further studies will investigate what role fibrinogen binding plays in 725 pathogenesis. While the Bucl8-pump is likely not be involved in the MDR phenotype of 726 Burkholderia, we have identified FA and pHBA as inducible substrates of the pump and 727 will continue to investigate metabolite analogs that may affect cell function. Importantly, 728 the growth of the Bucl8-pump deletion mutant was significantly affected even in the 729 absence of FA and pHBA. By characterizing the Bucl8-associated efflux system, we can 730 advance therapies and strategies for combating these pathogens, including developing 731 pump inhibitors, targeting transport mechanisms, or identifying potential surface-732 exposed vaccine targets derived from Bucl8. 733