Identification of a novel transport system in Borrelia burgdorferi that links the inner and outer membranes

Abstract Borrelia burgdorferi, the spirochete that causes Lyme disease, is a diderm organism that is similar to Gram-negative organisms in that it contains both an inner and outer membrane. Unlike typical Gram-negative organisms, however, B. burgdorferi lacks lipopolysaccharide (LPS). Using computational genome analyses and structural modeling, we identified a transport system containing six proteins in B. burgdorferi that are all orthologs to proteins found in the lipopolysaccharide transport (LPT) system that links the inner and outer membranes of Gram-negative organisms and is responsible for placing LPS on the surface of these organisms. While B. burgdorferi does not contain LPS, it does encode over 100 different surface-exposed lipoproteins and several major glycolipids, which like LPS are also highly amphiphilic molecules, though no system to transport these molecules to the borrelial surface is known. Accordingly, experiments supplemented by molecular modeling were undertaken to determine whether the orthologous LPT system identified in B. burgdorferi could transport lipoproteins and/or glycolipids to the borrelial outer membrane. Our combined observations strongly suggest that the LPT transport system does not transport lipoproteins to the surface. Molecular dynamic modeling, however, suggests that the borrelial LPT system could transport borrelial glycolipids to the outer membrane.


Introduction
Lyme disease is a multisystem disorder that typically results in a combination of cardiac, neurological, dermatological, and rheumatological manifestations. Additionally, Lyme disease is a major public health problem as it is the most common tick-borne illness in both the USA and Europe (Schwartz et al. 2017, van den Wijngaard et al. 2017. Borrelia burgdorferi sensu stricto , the primary causative agent of Lyme disease in North America , can e v ade the host imm une r esponse for years without a ppr opriate antibiotic tr eatment, whic h can r esult in a c hr onic and debilitating disease (Steere et al. 1987, Pachner 1988, Sehgal and Khurana 2015. B . bur gdorferi is a diderm, extracellular pathogen, thus, the outer membrane (OM) is the interface between this spir oc hete and the infected host. Consequently, the OM has garnered much attention over the past several decades.
OM biogenesis studies in B . bur gdorferi ha ve un veiled that the spir oc hete has a quite unique OM as compared to Gramnegativ e or ganisms. Contr ary to the v ery fe w surface-exposed lipoproteins in Gram-negative organisms , B . bur gdorferi encodes over 100 lipoproteins located on its surface (Fraser et al. 1997, Dowdell et al. 2017. Depending on the environmental conditions, the lipopr oteins ar e differ entiall y expr essed with onl y specific subsets being expressed at a given point in the organisms' lifecycle (Ramamoorthy and Philipp 1998, Ojaimi et al. 2003, Hyde et al. 2007, Angel et al. 2010 ). T he B . burgdorferi surface lipoproteins are highly important to both imm une e v asion and ov er all disease pathogenesis (Brightbill et al. 1999, Hefty et al. 2001, Kenedy et al. 2012, Wilson and Bernstein 2016, Coburn et al. 2020 ). The B. burgdorferi OM also lacks the glycolipid lipopolysaccharide (LPS) (Takayama et al. 1987, Fraser et al. 1997, but does produce three other glycolipids (Hossain et al. 2001, Ben-Menachem et al. 2003, Stübs et al. 2011 ). These include c holesteryl 6-O -acyl-β-D-galactofur anoside (BbGL-I), mono-α-galactosyl-diacylgl ycer ol (BbGL-II), and cholesteryl-β-D-galactopyranoside (Hossain et al. 2001, Ben-Menachem et al. 2003, Stübs et al. 2009, Szamosvári et al. 2022. Lyme disease patients de v elop antibodies specific to these glycolipids during infection indicating they are immunogenic during infection and could play an important role in disease pathogenesis (Schröder et al. 2008(Schröder et al. , J ones et al. 2009(Schröder et al. , P ozsga y et al. 2011. While the surface lipopr oteins and gl ycolipids ar e crucial to the structur e and function of the OM and are important with regard to disease pathogenesis, it is still unclear how these molecules are transported to the OM. It seems likely that the lipoprotein and glycolipid constituents of the OM would r equir e the help of specific transport systems for proper localization. Fr eeze-fr actur e electr on micr oscopy has r e v ealed that B. burgdorferi contains 10-fold fewer membrane-spanning OM proteins (OMPs) than Escherichia coli (Lugtenberg andvan Alphen 1983 , Radolf et al. 1994 ). Curr entl y, onl y 10 differ ent OMPs hav e been identified in B . bur gdorferi (Sadziene et al. 1995, Skare et al. 1997, P arv een and Leong 2000, Cluss et al. 2004, Brooks et al. 2006, Antonara et al. 2007, Bunikis et al. 2008, Lenhart and Akins 2010, Kenedy et al. 2012, 2016, Wood et al. 2013, Shrestha et al. 2017 ). Among OMPs with known functions, only the OMP BamA has been determined to play a specific role in borrelial OM biogenesis as it is r equir ed for embedding other pr oteins into the OM (Lenhart and Akins 2010, Dunn et al. 2015, Iqbal et al. 2016. This is analogous to the role of BamA and the gr eater beta-barr el assembl y machinery (BAM) complex in Gram-negative organisms (Kim et al. 2012 ). Another critical system in Gr am-negativ e OM biogenesis is the lipopol ysacc haride tr ansport (LPT) system. The LPT system transports LPS to the surface of Gram-negative organisms. During this process, LPS is extracted from the periplasmic leaflet of the inner membrane and transported through the periplasm before being inserted into the outer leaflet of the OM , Freinkman et al. 2012, Laguri et al. 2017. The LPT complex contains se v en differ ent pr oteins, and they ar e essential for LPS surface localization in Gr am-negativ e or ganisms (Sampson et al. 1989, Sperandeo et al. 2006, Wu et al. 2006, Ruiz et al. 2008 ). The LPT system contains an inner membrane permease unit comprised of an ATP-binding cassette homodimer of LptB and a permease made of LptF and LptG. Together, these units extract LPS from the inner membrane and transport LPS to the LptC protein, which is found in the periplasm but is anc hor ed to the inner membrane by a single transmembrane domain (Simpson et al. 2016 ). The periplasmic bridge pr otein LptA inter acts with LptC and accepts LPS befor e tr ansporting LPS to the OMP LptD, which is the terminal component of the transport system and is r equir ed to localize LPS into the outer leaflet of the OM (Bo wy er et al. , Sperandeo et al. 2011, Dong et al. 2017a, Hicks and Jia 2018. Additionall y, in Gr am-negativ e or ganisms ther e is an LptE protein that interacts with LptD that helps to terminate LPS transport (Wu et al. 2006, Botte et al. 2022. LptE plays se v er al important r oles: it is necessary for proper folding of the OMP LptD, it pr e v ents a ggregation of LPS during the transport and release process, and it also acts as a plug for the beta-barrel pore formed by LptD, which pr e v ents continual and unabated LPS transport , Freinkman et al. 2011, Chng et al. 2012, Maloj či ć et al. 2014. A unique feature of the LPT system is the presence of a beta-taco fold (commonly incorrectly described as a beta-jelly roll fold) in the LPT proteins LptF, LptG, LptC, LptA, and LptD, and this unique structure allows for the transport of amphiphilic molecules such as LPS (Hicks and Jia 2018 ).
B . bur gdorferi has pr e viousl y been reported to contain orthologs for five out of the se v en LPT pr oteins (Putker et al. 2015 ). We have also pr e viousl y r eported that B . bur gdorferi protein BB0838, encoded by open reading frame (ORF) bb0838 , is a surface exposed OMP with a computationall y pr edicted structur e similar to that of LptD . Pukter et al. ( 2015 ) identified BB0465, BB0466, BB0807/BB0808, and BB0838 as orthologs to LptA, LptB, LptF/LptG, and LptD, r espectiv el y. Her e, we describe the identification of B . bur gdorferi protein BB0464, an ortholog to LptC, and provide evidence that BB0807 is the LptF ortholog and that BB0808 is the LptG ortholog. Finally, we have also generated a working model of the LPT system in B . bur gdorferi (consisting of proteins encoded by ORFs bb0464 , bb0465 , bb0466 , bb0807 , bb0808 , and bb0838 ) and empirically determined that these proteins interact as would be expected of an LPT system. While the presence of a putative LPS transport system in an organism that lacks LPS is counterintuitive, we should note that B. burgdorferi does contain an abundance of surface-exposed lipoproteins and multiple glycolipids that are similar in their amphiphilic nature to LPS. Our combined data indicate that the novel borrelial LPT system does not transport lipoproteins to the surface and suggest that the borrelial glycolipids are the cargo instead.

Recombinant protein and antibody production
Utilizing oligonucleotides BbLptD F and BbLptD R (Table 1 ), the N-terminal domain of BbLptD was amplified from nucleotide 88 to nucleotide 873. The amplicon was digested with restriction enzymes NheI and XhoI and ligated into an NheI/XhoI digested pET23a (EMD Millipore , Billerica, MA). T he resulting construct was transformed into E. coli Overexpress™ C41(DE3) for expression. The recombinant protein was purified using nickel-nitrilotriacetic acid a gar ose (Ni-NTA; Qia gen, Valencia, CA) in His-ta g nativ e and His-tag denaturing purification conditions as described previously (Luthr a et al. 2011(Luthr a et al. , K enedy et al. 2014. A 70% native and 30% denatured mix of the recombinant protein was used for antibody gener ation. Rat pol yclonal antibodies specific for the N-terminal domain of BbLptD (amino acids 29-291) were generated by Envigo (Indiana polis, IN). Rat pol yclonal antibodies specific to OspA, OspC, CspA, or P66, and rabbit polyclonal antibodies specific to FlaB wer e gener ated as pr e viousl y described (Radolf et al. 1994, Cox et al. 1996, Brooks et al. 2005, Kenedy et al. 2014 ). These antibodies were subsequently utilized for immunofluorescence assays and immunoblots.

B. burgdorferi LPT-orthologous system identification and protein modeling
BbLptD was first identified as a putative ortholog to LptD in a pr e vious publication (K enedy et al. 2016 ). The National Center for Biotechnology Information BLAST analysis ( https://blast.ncbi .nlm.nih.gov/Blast.cgi ) was performed using E. coli proteins LptA, LptC, LptB, LptF, and LptG, limiting results to those that belong to the genus Borrelia . Subsequent computational modeling was performed on the following B . bur gdorferi Lpt orthologs: BB0838 (LptD), BB0465 (LptA), BB0464 (LptC), BB0466 (LptB), BB0807 (LptF), and BB0808 (LptG). The pr oteins wer e modeled with AlphaFold2.1 (Jumper et al. 2021, Varadi et al. 2021 with the default settings, and the top model was chosen by pLDDT. Both BbLptA and BbLptD were modeled without their predicted signal peptide sequence, as predicted by PrediSi, Signal-CF, and Signal-P. The working model of the B . bur gdorferi LPT-orthologous system was created utilizing the models of individual proteins obtained from AlphaFold2. BbLptD and BbLptA were inserted into the model as individual proteins. BbLptC, BbLptF, and BbLptG were inserted into the model as a trimer. BbLptB was inserted as a homodimer. The working model was created using BioRender ( https: // biorender.com/ ).
Multimer modeling was performed using AlphaFold2.1 (Evans et al. 2022 ). For each multimer, the proteins of interest were submitted together to determine the likelihood of interaction and at what interface. BbLptA (residues 20-231) and BbLptDNT (residues 30-271) were modeled together, both in the absence of their respecti ve signal pe ptides. BbLptA (residues 20-231) was modeled with BbLptC (residues 19-174), both without their signal peptides. BbLptC was submitted for multimer modeling with BbLptF and BbLptG separ atel y, follo w ed b y BbLptC, BbLptF, and BbLptG together. For these modeling predictions, the full-length proteins were submitted. Of the 25 models generated, one preferred model was selected by visual inspection, ipTM + pTM, and pLDDT scores at the modeled interface.

Modeling of glycolipid cargo in BbLptA
Modeling of BbGL-I and BbGL-II inside BbLptA was performed using a combination of Site Identification by Ligand Competitive Satur ation (SILCS), Ra pid Ov erlay of Chemical Structures (ROCS) b y OpenEy e, and MD using GROMACS (Guv enc h and MacK er ell 2009 , R OCS 3.5.1.1 OpenEy e 2007 , OpenEy e Scientific Softw are 2010 , Hawkins et al. 2007, 2010, Abraham et al. 2015. SILCS simulations were performed on the model of BbLptA, generating 3D maps of functional group affinities patterns (FragMaps), as well as an exclusion ma p wher e functional groups or water are forbidden from interacting. Identification of possible binding sites for BbGL-I was performed by decomposing the molecule into smaller fr a gments that wer e used with SILCS-Hotspots to identify all sites with interaction energies stronger than −2 kcal/mol.
Within the beta-taco fold of BbLptA, there was a hotspot for the sterol of BbGL-I, two hotspots for octane, and two hotspots for 2-methylbutane . T he coordinates for the lowest energy conformation of each probe at its respective hotspot(s) were used as a pharmacophore to place BbGL-I within the beta-taco. A total of 27 551 unique, low-energy conformations of BbGL-I were generated using OMEGA and overlaid on the pharmacophore based on optimization of sha pe/c hemical complementarity using ROCS, both de v eloped b y OpenEy e ( R OCS 3.5.1.1 OpenEy e 2007 , OpenEy e Scientific Software 2010 , Hawkins et al. 2007Hawkins et al. , 2010. The top 10 scoring conformations were evaluated, and one pose was selected based on maximal ov erla p with minimal clashes with the protein backbone . T he conformer was superimposed into the betataco. MD was performed using GROMACS with the CHARMM36 m forcefield (Abraham et al. 2015, Huang et al. 2017. The ligand was parameterized using CGenFF (Vanommeslaeghe et al. 2010 ), and the protein-ligand complex was solvated in a dodecahedron unit cell with a sodium added to neutralize the system. A steepest decent minimization was performed, and the system was equilibrated at 300 K with a 100-ps NVT simulation follo w ed by an additional 100 ps NPT sim ulation. Finall y, a 200-ns NPT pr oduction sim ulation w as performed. BbGL-II w as modeled using a similar protocol. A total of 26 635 conformers were generated using Omega and overlaid to the final BbGL-I pose from the MD simulation. MD was performed using the aforementioned protocol.

Localization immunofluorescence assays
For experiments utilizing B . bur gdorferi strain B31, cells were grown to mid-exponential phase and diluted to 5 × 10 6 organisms/ml. Cell suspensions were coincubated for 1 h with rat-anti-BbLptDNT antibodies at a dilution of 1:100 and rabbit-anti-FlaB antibodies at a dilution of 1:2500. As a contr ol, separ ate cell suspensions were coincubated for 1 h with rat-anti-CspA antibodies at a dilution of 1:100 and rabbit-anti-FlaB antibodies at a dilu-tion of 1:2500. Cells w ere w ashed three times with 1X PBS, and the final pellet was resuspended in 100 μl of 1X PBS. A volume of 10 μl of the final resuspension were spotted onto a microscope slide, dried overnight, and fixed with acetone. Samples were blocked for 30 min with PBS containing 0.2% bovine serum albumin (BSA). Fixed and blocked samples were then incubated for 45 min with Alexa Fluor 488-conjugated goat-anti-rat antibodies (Invitrogen, Waltham, MA) at a dilution of 1:250 and Alexa Fluor 568conjucated goat-anti-rabbit antibodies (Invitrogen) at a dilution of 1:1000. Samples were washed three times with PBS containing 0.2% BSA, mounted with one drop of a 1:1 mixture of Vectashield mounting solution containing 4',6-diamidino-phenylindole (DAPI; Vector Laboratories , Burlingame , C A) and Vectashield mounting solution without DAPI (Vector Laboratories), and sealed with a coverslip. Another set of cell suspensions were incubated without any antibodies, washed, resuspended, spotted, and fixed. The fixed cells were then blocked with PBS containing 0.2% BSA, incubated with the same dilutions and mixtures of primary antibodies for 45 min. The pr e vious description of the protocol postprimary antibody was then performed the same.
Experiments utilizing B . bur gdorferi strain B31-5A4 LK and B31-5A4 LK-flacp::bblptD were performed identically with the exception of antibodies and B . bur gdorferi strains utilized. These strains were coincubated with a combination of rat-anti-CspA antibodies at a dilution 1:100 and rabbit-anti-FlaB antibodies at a dilution of 1:2500 for 1 h at room temperature . T he remaining experimental pr ocedur e was performed exactly as described abo ve . Images of samples were visualized and captured with an Olympus BX-60 fluor escence micr oscope and Ol ympus DP27 camer a (Ol ympus America Inc., Center Valley, PA).

Cloning of B. burgdorferi Lpt-orthologs for copurification assays
The DNA sequences of all putative Lpt-orthologs were amplified from B . bur gdorferi B31 genomic DNA using the primers in Table 1 . Amplicons were cloned into one of two multiple cloning sites of pACYCDuet-1, depending on the desired tag. Amplicons destined for the 6x His tag were digested and cloned into the BamHI and SacI sites of pACYCDuet-1. Amplicons destined for the S tag were digested and cloned into the NdeI and KpnI sites of pACYCDuet-1. Cloning into each multiple cloning site was performed stepwise with complete insertion into the 6x His-tag and then complete insertion into the S-tag. Vectors were transformed into E. coli Over-express™ C41(DE3) for protein expression. Sequences of inserts wer e v erified thr ough DNA sequencing to ensur e the genes of inter est wer e unalter ed during the cloning pr ocess.

His-tag nati v e purifica tion for copurifica tion assays
For the copurification assa ys , pACYCDuet constructs in E. coli stain C41 wer e gr own ov ernight at 37 • C in 35 ml LB cultures . T he 35 ml of starter culture was inoculated into 500 ml of fresh LB and grown at 37 • C to an optical density at 600 nm (OD 600 ) of between 0.55 and 0.75. At this point, pr otein expr ession was induced with 1 mM IPTG follo w ed b y an additional incubation for 3 h at 37 • C. Cells were pelleted at 8200 × g for 20 min at 4 • C. Cell pellets were resuspended in 15 ml of lysis buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 10 mM imidazole, pH 8) with 15 μl of protease inhibitor cocktail. Resuspended cells were lysed by sonication and pelleted at 12 500 × g for 45 min. The resulting supernatant was normalized to an OD 600 of 0.044 and 15 ml of the supernatant was incubated with 2.5 ml bed volume of Ni-NTA a gar ose for 15 min to bind the His-ta gged pr oteins to the a gar ose . T he resin w as w ashed with 100 ml of wash buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 20 mM imidazole, pH 8), follo w ed b y 50 ml of increased imidazole concentration wash buffer (50 mM NaH 2 PO 4 , 300 mM NaCl, 40 mM imidazole, pH 8). Pr oteins wer e not eluted, as solubility of the proteins dr ops dr asticall y upon elution.

SDS-PAGE and immunoblot analysis for copurification assays
Supernatant samples wer e pr epar ed by mixing 1:1 with final sample buffer [FSB; 62.5 μM Tris-HCl (pH 6.8), 10% (v/v) gl ycer ol, 5% (v/v) β-mercaptoethanol, 5% SDS, 0.001% bromophenol blue] and boiled for 10 min. Copurification samples were prepared by taking 200 μl of slurry of each final sample bound to Ni-NTA agarose, removing the wash buffer, resuspending in 160 μl of FSB, and boiling for 10 min. Samples were run by electr ophor esis on SDS-PAGE gels with 2.4% stacking and 12.5% separating. For the anti-His imm unoblots, e v ery normalized supernatant was loaded at 6 ul of the 1:1 sample and 5 μl of e v ery copurification r esin sample was loaded. This was loaded the same for the anti-S immunoblots with the exception of the supernatant samples containing OspC-S. This protein is expressed at significantly higher le v els than all other proteins in the experiment, so to obtain images, the supernatant samples containing this protein for the anti-S immunoblots was diluted 384x. Copurification resin samples, ho w e v er, wer e loaded the same for OspC-S containing samples as all other samples. Gels were transferred electrophoreticall y to PVDF membr ane (pol yin ylidene fluoride; BioRad, Hercules, CA) for immunoblot analysis. All immunoblots were performed as pr e viousl y described (Iqbal et al. 2016 ). To anal yze the pr esence of S-tagged recombinant proteins, immunoblots were incubated with 1:2000 mouse-anti-S antibody (EMD Millipore) follo w ed b y a 45-min r oom temper atur e incubation with horser adish peroxidase-conjugated goat-anti-mouse secondary antibody (Bio-Rad). To analyze the presence of His-tagged recombinant proteins, imm unoblots wer e incubated with 1:2000 HRP-conjugated mouse-anti-His antibody (R&D Systems , Minneapolis , MN) for 1.5 h. Immunoblots were developed with enhanced chemiluminescence (ECL) substrate (Thermo Fisher Scientific, Waltham, MA) and subsequently visualized with the ChemiDoc MP Imaging System (BioRad). For a loading control, 6 μl of all OD 600 normalized supernatants were immunoblotted with 1:4000 mouse-anti-GAPDH antibody (Invitrogen), follo w ed b y 1:4000 horser adish per oxidaseconjugated goat-anti-mouse secondary antibody (BioRad).

Gener a tion of B. burgdorferi strain B31-5A4 LK-flacp::bblptD
The BbLptD IPTG regulatable mutant, designated flacp::bblptD , was generated by inserting the flacp promoter immediately upstream of the start codon of bblptD as pr e viousl y described (Iqbal et al. 2016 ). The primers used to make the construct are listed in Table 1 with the restriction enzymes used for cloning into the pBluescript-II SK + vector (Stratagene , La J olla, C A). Primers " BbLptD up mutant F" and " BbLptD up mutant R" were utilized to amplify 600 bp upstream of the start codon of bblptD and primers " BbLptD down mutant F" and " BbLptD down mutant R" were utilized to amplify 600 bp downstream of the start codon of bblptD . The amplicons were inserted into the multiple cloning site of pBluescript-II SK + vector using the corresponding restriction enzymes/sites. The stre ptom ycin r esistance cassette and the flacp pr omoter wer e digested as a unit from the pTLflacp::795 construct that was described pr e viousl y (Lenhart and Akins 2010 ) and inserted into the XhoI and NdeI sites of the construct. The subsequent construct was electr opor ated into B . bur gdorferi B31-5A4 LK and grown in BSK-II media supplemented with 1 mM IPTG, kanamycin, and stre ptom ycin as described abo ve . Clones were screened by PCR to v erify pr esence of all B . bur gdorferi plasmids and by immunoblot to confirm IPTG-regulation of BbLptD.

Proteinase K (PK) accessibility assays
Proteinase K experiments were performed as previously described (Brooks et al. 2005 ) with B . bur gdorferi strains B31-5A4 LK or B31-5A4 LK-flacp::bblptD supplemented with 0, 0.01, or 1 mM IPTG. Final samples were resuspended in FSB and boiled for 10 min. For immunoblot analysis, samples were subjected to SDS-PAGE electr ophor esis and tr ansferr ed electr ophor eticall y to PVDF membrane . T hese samples were immunoblotted with rat-anti-CspA or rat-anti-P66 antibodies, both at a dilution of 1:4000, follo w ed b y incubation with horseradish peroxidase-conjugated goat-anti-rat antibodies at a dilution of 1:8000. For FlaB immunoblots, samples were incubated with rabbit-anti-FlaB antibody at a dilution of 1:15 000, follo w ed b y secondary incubation with horseradish per oxidase-conjugated goat-anti-r abbit antibodies at a dilution of 1:30 000. All immunoblots were developed and visualized as described abo ve .

Genetic organization of the orthologs that comprise the B. burgdorferi LPT system
T he predicted B . bur gdorferi LPT transport system is comprised of six proteins. We have previously identified B . bur gdorferi protein BB0838 as a 120-kDa, membrane-spanning OMP and determined that it is an ortholog to LptD . Pr e vious studies have also identified BB0838 as an LptD ortholog and proposed additional borrelial ORFs bb0465 , bb0466 , and bb0807 / bb0808 as lptA , lptB , and lptF/G orthologs, r espectiv el y (Putker et al. 2015 ). Scanning the genome and additional computational modeling allo w ed us to identify the protein BB0464 as a predicted LptC ortholog. BLASTP analyses (blast.ncbi.nlm.nih.gov/Blast.cgi) identified the r espectiv e Lpt orthologs as the top query hits for the B . bur gdorferi protein sequences, further confirming the orthology between the r espectiv e B . bur gdorferi pr oteins and the Lpt pr oteins. Consistent with these computer-based analyses, the ov er all genomic or ganization of the genes encoding the borrelial orthologs was found to be identical to the organization of the LPT genes from the enteric pathogens Shigella flexneri and E. coli (Fig. 1 ). Specifically, these pr oteins ar e encoded at thr ee distinct loci on the c hr omosome of eac h bacterium. Pr oteins LptC, LptA, and LptB are encoded in S. flexneri and E. coli at one locus, and the r espectiv e orthologs from B . bur gdorferi, ORFs lptC ( bb0464 ), lptA ( bb0465 ), and lptB ( bb0466 ), are also encoded at one locus in the same order ( Fig. 1 A). LptD is encoded at a second distinct locus in S. flexneri , E. coli , and in B . bur gdorferi (Fig. 1 B). Finall y, the inner membr ane permease units, LptF and LptG, are encoded in S. flexneri and E. coli at a third distinct locus and the same organization is found in B . bur gdorferi ( Fig. 1 C). We have denoted BB0807 as the LptF ortholog and BB0808 as the LptG ortholog, which is most consistent with better computational alignment scores for each pair using PyMOLv2.4.0 (Schrödinger 2020 ) (data not shown).

Structural models of B. burgdorferi Lpt orthologs
Thr ough structur al modeling of eac h Lpt ortholog in B . bur gdorferi , we hav e pr edicted a working model for the B . bur gdorferi LPT system (Fig. 2 ), that is similar to what has been described for the 72 .10 †: The pLDDT per-residue score is averaged to give the numbers shown for each r espectiv e pr otein. ‡ : BB0838NT indicates the N-terminal domain (residues Q30-I271) was modeled alone. §: BB0838BB indicates the tr ansmembr ane beta-barr el (r esidues F272-K1146) was modeled alone.
LPT system in Gr am-negativ e or ganisms. Taking adv anta ge of advances in computational 3D protein structure prediction, we utilized the AlphaFold2.1 modeling algorithm (Jumper et al. 2021, Varadi et al. 2021 to generate individual models of the B . bur gdorferi Lpt-orthologs (Fig. 2 ). The LptD ortholog, BB0838, is predicted to fold into a large, OM-spanning beta barrel beginning approximatel y at r esidue 294 and continuing thr ough the C-terminus of the amino acid sequence (Fig. 2 ). The beta barrel portion of BB0838 is predicted to be larger than the E. coli LptD and contain 31 beta strands rather than 26. BB0838 also was predicted to include a large, periplasmic loop between beta strands 10 and 11 (Fig. 2 ). Additionally, BB0838 contains a predicted periplasmic, Nterminal domain comprised of residues 30-293 (Fig. 2 ). Residues 76-293 take the conformation of a beta-taco fold motif (Fig. 2 ). The N-terminal beta-taco fold motif has also been identified and characterized in the solved crystal structure of E. coli LptD (Botos et al. 2016 ). Inter estingl y, r esidues 30-75 of the BB0838 N-terminal domain consist of a series of four alpha helices, that are not present in other LptD proteins (Botos et al. 2016 ). The AlphaFold2.1 algorithm provides a predicted local distance difference test (pLDDT) as a per-residue confidence score for each model. Accordingly, an av er a ge pLDDT ≥ 90 indicates very high confidence in the model, a pLDDT from 70 to 89.99 is considered confident, a pLDDT from 50 to 69.99 is considered low confidence, and any pLDDT score lo w er than 50 corresponds to very low confidence (Jumper et al. 2021, Varadi et al. 2021. As shown in Table 2 , the predicted BB0838 model resulted in a pLDDT of 72.70, indicating a confident score for the model shown in Fig. 2 . Inter estingl y, the confidence in the N-terminal domain of BB0838 modeled alone (86.40) is higher than that of the beta barrel of BB0838 modeled alone (72.10) or BB0838 as a whole (Table 2 ). This was not sur prising giv en the highly specialized role the N-terminal beta-taco fold of LptD has in the tr ansport of LPS (Sper andeo et al. 2008, Gu et al. 2015, Laguri et al. 2017. With this data, we conclude that BB0838 is indeed the B. burgdorferi ortholog of LptD. B . bur gdorferi BB0465 (residues 20-231; signal peptide excluded) is predicted to fold into a beta-taco formation similar to the Nterminal domain of BB0838 (Fig. 2 ). When modeled with the Al-phaFold2.1 algorithm, the predicted structure of BB0465 has a confident pLDDT score of 87.96 (Table 2). Further, this model is very similar to that of the known crystal structure of E. coli LptA, which contains the same highly specialized beta-taco fold structur e as pr edicted for BB0465 (Suits et al. 2008 ). These findings combined with the genomic and BLASTP data allow us to confirm that BB0465 is the periplasmic bridge ortholog, LptA, in the B . bur gdorferi LPT system. As shown in Fig. 2 , the AlphaFold2.1 model of BB0464 resembles the known crystal structure of LptC, which is Chec ker ed-filled arr ows indicate genes encoding LptG pr oteins . T he length of arr ows r epr esents the r espectiv e size of the gene compar ed to its ortholog.
Figure 2. Current model of the B. burgdorferi LPT-orthologous system. The B. burgdorferi LPT system deduced from the combined genetic organization and AlphaFold2.1 structural modeling of each B . bur gdorferi Lpt ortholog. BB0838 is shown in cyan, BB0465 is shown in orange, BB0464 is shown in red, BB0807 is shown in blue, BB0808 is shown in green, and BB0466 2 is shown in magenta. These proteins are colisted with their predicted Lpt ortholog. Fatty acid tails of the putative lipoprotein BB0464 are shown in red. Tri-acylated, yellow, globular molecules in the membranes represent borrelial lipoproteins. Di-acylated blue molecules in the membrane represent borrelial phospholipids. Di-acylated purple molecules represent borrelial glycolipids . T his figure was created with BioRender. c har acterized by a periplasmic beta-taco fold with a single alpha helix anchoring LptC into the inner membrane (Tran et al. 2010 ). Contrary to other LptC proteins, ho w ever, the first 14 residues of BB0464 are predicted as a signal II peptide with a cysteine residue dir ectl y after the pr edicted cleav a ge site. Because of this, it seems likely that BB0464 is an inner membrane lipoprotein as suggested by He et al . ( 2023 ) and is illustrated as so in Fig. 2 . This would indicate the mechanism of inner membrane anchoring of the B. burgdorferi LptC ortholog div er ges fr om canonical LptC membr ane anc horing. Ov er all, the model of BB0464 is of very high confidence with a pLDDT of 90.78 (Table 2 ), leading us to predict that BB0464 is the LptC ortholog.
B . bur gdorferi BB0807, BB0808, and BB0466 wer e initiall y modeled with AlphaFold2.1 separ atel y to obtain their individual pLDDT scores ( Table 2 ). The predicted structure of BB0466 (Fig. 2 ) is very similar to the known structure of various Gram-negative LptB homodimers (Owens et al. 2019 ) and contains motifs classified in the ATP-binding cassette protein family. Combined with the very high pLDDT of 92.16 (Table 2), we conclude that BB0466 is the predicted LptB ortholog. The predicted structural models of BB0807 and BB0808 shown in Fig. 2 are also considered confident with r espectiv e pLDDT scor es of 84.71 and 81.80 (Table 2 ). While gener all y, LptF and LptG are similar in their structure (Dong et al. 2017b, Owens et al. 2019, Tang et al. 2019, w e w anted to determine which of the B . bur gdorferi putative orthologs was most similar to either LptF or LptG. To do this, we utilized the solved crystal structure of the Enterobacter cloacae LptB 2 FGC complex (PDB 6MIT) (Owens et al. 2019 ). By aligning BB0807 and BB0808 individually to the LptF and LptG of this model in PyMOLv2.4.0, w e w ere able to determine that the r oot-mean-squar e de viation (RMSD) w as lo w est when BB0807 w as aligned with LptF and BB0808 w as aligned with LptG (data not shown). The lower RMSD score indicates a higher confidence le v el that BB0807 is the LptF ortholog and that BB0808 is the LptG ortholog. BB0807 and BB0808 are predicted to have the same small beta-taco fold domain and larger inner membr ane, tr ansmembr ane domain comprised of se v er al alpha helices similar to what has been observed in Gram-negative LptFG heterodimers (Owens et al. 2019 ). To create a more cohesive and interaction-re presentati ve model, BB0807, BB0808, and BB0464 were also modeled together as a trimer using the Al-phaFold2.1 algorithm, which is what is illustrated in the working model (Fig. 2 ). From this point forw ar d for more clarity, the B. burgdorferi Lpt orthologs BB0464, BB0465, BB0466, BB0807, BB0808, and BB0838 will be r eferr ed to as BbLptC, BbLptA, BbLptB, BbLptF, BbLptG, and BbLptD, r espectiv el y.

The N-terminal domain of BbLptD is periplasmic
The N-terminal beta-taco fold domain of E. coli LptD is known to be periplasmic and soluble when expressed independently , Freinkman et al. 2011. As described above and illustrated in Fig. 2 , BbLptD is predicted to have a similar Nterminal beta-taco fold. To determine if the N-terminal domain of BbLptD was indeed periplasmic, imm unofluor escence assays utilizing an antibody specific to the N-terminal domain of BbLptD (residues 29-291) were performed (Fig. 3 ). Specifically, the BbLptD N-terminal antibody was incubated either with intact B . bur gdorferi cells or with B . bur gdorferi cells that had been fixed through OM permeabilization with acetone (Fig. 3 A). Given that antibodies cannot penetrate an intact OM, the lack of fluorescence on the borrelial surface indicates that the N-terminal domain of BbLptD is not localized on the surface of B . bur gdorferi (Fig. 3 A, upper  left-hand panel). After OM permeabilization, incubation with the BbLptD N-terminal antibody results in fluorescence (Fig. 3 A, lo w er left-hand panel). These results indicate that the N-terminal domain of BbLptD is not surface-exposed, consistent with our prediction that this domain is found in the periplasm. To ensure that B . bur gdorferi cells wer e pr operl y permeabilized in the fixed treatment and that the fr a gile OM r emained intact in the surface treatment, all cells were also incubated with antibodies specific for the periplasmic protein FlaB (Fig. 3 A, middle panels). All spir oc hetes in the microscopic field of view were also visualized through staining with the DNA-binding dye DAPI (Fig. 3 A, righthand panels). As an additional control to confirm proper detection of surface localized proteins by immunofluorescence, we performed the same set of experiments on B . bur gdorferi cells that were incubated with antibodies specific for the abundant surface lipopr otein CspA (Br ooks et al. 2005(Br ooks et al. , K enedy et al. 2009 (Fig. 3 B). As expected, this resulted in fluorescence in both surface and fixed treatments (Fig. 3 B, left-hand panels). Confirmation that the Nterminal domain of BbLptD is not surface localized is consistent with the computational structural modeling.

BbLptA specifically interacts with the N-terminal domain of BbLptD
Pr e vious studies in E. coli have demonstrated that the C-terminus of LptA specificall y inter acts with the N-terminal beta-taco domain of LptD (Freinkman et al. 2012 ). Since prior studies in our lab have identified BbLptD as an integral OMP , the finding that the N-terminal domain is not surface localized indicates it is likely localized to the periplasm. For these r easons, we c hose to utilize this N-terminal domain rather than the entir e BbLptD pr otein for these pr otein-pr otein inter action studies. To determine if the r espectiv e B. burgdorferi orthologs similarl y inter act, we utilized a copurification assa y for the B . burgdorferi recombinant proteins. Specifically, the pACYCDuet-1 construct was utilized to coexpress BbLptA with an N-terminal His-tag and the N-terminal region of BbLptD (BbLptDNT) with a C-terminal S-tag in E. coli . This was done to ensure that the termini of each protein that were predicted to interact were free fr om an y obstruction fr om the ta gs . For these experiments , supernatants were examined of the coexpressed proteins to illustrate the ov er all expr ession of the two proteins of interest. The copurification lanes display the His-tagged protein that was purified and whether it interacts with the coexpressed S-tagged protein of interest.
Using this system, we observed that BbLptDNT-S copurifies with BbLptA-His, indicating that BbLptA and the periplasmic region of BbLptD interact (Fig. 4 A). Both of these proteins were well-expressed, as seen in the supernatant lanes of Fig. 4 (A). To confirm that the interaction between BbLptDNT-S and BbLptA-His was specific, a copurification experiment was also performed with BbLptA-His and OspC-S utilizing coexpression of BbLptA-His and OspC-S in the pACYCDuet construct (Fig. 4 B). OspC is an abundant B . bur gdorferi pr otein, so this contr ol ensur es that ther e are no nonspecific interactions between BbLptA-His and other B. burgdorferi proteins or the S-tag itself. As seen in Fig. 4 (B), OspC-S does not copurify with BbLptA-His. Additionally , BbLptDNT -S was expressed alone in pACYCDuet to ensure that the recombinant protein does not purify on the Ni-NTA resin in the absence of BbLptA-His. As shown in Fig. 4 (C), BbLptDNT-S, despite being expressed at high levels in the supernatant, does not purify on the Ni-NTA resin alone . T he combined controls in Fig. 4 (B) and (C) indicate that the interaction between BbLptA-His and BbLptDNT-S is specific. A loading control using antibodies specific for the housek ee ping protein GAPDH was also included (Fig. 4 D) to ensure all of the supernatants were loaded equally.

BbLptC specifically interacts with BbLptA
LptA and LptC hav e been found to inter act as part of the LPT system in E. coli (Bo wy er et al. 2011, Sperandeo et al. 2011, Schultz et al. 2013. Ther efor e, the pACYCDuet copurification system was also used to investigate the interaction between BbLptA and BbLptC. Similar to the experiments for BbLptA and BbLptDNT abo ve , the pACYCDuet system was used to coexpress and copurify BbLptC-His and BbLptA-S in E. coli (Fig. 5 A). BbLptC was expressed on the N-terminal His-tag to k ee p the C-termin us of the protein available for binding, as the C-terminus of LptC is necessary for interaction with LptA (Sperandeo et al. 2011 ). Both BbLptC-His and BbLptA-S were expressed in the supernatant (Fig. 5 A). After purification of BbLptC-His, BbLptA-S was found to copurify with BbLptC-His (Fig. 5 A), indicating an interaction between the two proteins. To confirm that the BbLptC-His and BbLptA-S interaction was not the result of nonspecific interactions, OspC-S was coexpressed with BbLptC-His in pACYCDuet and, separ atel y, BbLptA-S was expressed alone in the pACYCDuet construct. As expected, there was no copurification observed in either control despite expression of each protein in the supernatant (Fig. 5 B and C). The combined controls confirmed that the interaction between BbLptC-His and BbLptA-S is specific. The supernatants used in these experiments were normalized as illustrated by expression of GAPDH (Fig. 5 D).

Structural modeling of B. burgdorferi Lpt-ortholog interactions
To further supplement the pr otein-pr otein inter action studies between BbLptA/BbLptDNT and BbLptC/BbLptA, each protein pair was subjected to AlphaFold2.1 multimer modeling (Fig. 6 ). BbLptA without the predicted signal peptide was submitted as a multimer with the N-terminal domain of BbLptD (Fig. 6 A). These two pr oteins ar e pr edicted to inter act at an a ppr oximate interface comprised of N-terminal residues R105-N118 of BbLptDNT and C-terminal residues Y209-Q221 of BbLptA to create a continuous beta-taco fold. AlphaFold2.1 provides a confidence score of this pr edicted inter action composed of the pr edicted template modeling (pTM) plus the interface predicted template modeling (ipTM). A weighted sum of these two numbers is used to provide an overall model confidence where a score closer to 1 indicates a more confident model and interaction interface (Evans et al. 2022 ). The confidence score for the interaction interface between BbLptA and BbLptDNT is 0.84 (Table 3 ), indicating that there is a high likelihood that this interface is where the interaction between BbLptA and BbLptDNT occurs. Additionally, the ov er all pLDDT of 88.22 of the two proteins indicates overall confidence in their structures (Table 3 ). Multimer modeling was also performed to examine the BbLptA and BbLptC interaction, both in the absence of their signal peptides . T he model predicted an interaction at the C-terminal end of BbLptC and the N-terminal end of BbLptA (Fig. 6 B). This interaction forms a cohesive beta-taco fold structure, similar as to the interaction between BbLptDNT and BbLptA shown in Fig. 6 (A). The combined pLDDT of the BbLptC/BbLptA interaction was 87.62, indicating confidence in the ov er all pr edicted structur e. Furthermore, the ipTM + pTM value was 0.84 (Table 3 ), which suggests high confidence in the pr edicted inter action interface at residues N162-N174 of BbLptC and residues F41-V52 of BbLptA.
We also investigated the interaction of BbLptC and the putativ e inner membr ane permease. Multimer modeling pr edicted a higher likelihood that the N-terminal region of the beta-taco fold of BbLptC interacts with BbLptF than BbLptG. Both models, howe v er, do r epr esent ov er all confident tertiary structur es with pLD-DTs of 85.70 and 82.40 for BbLptC/BbLptF and BbLptC/BbLptG, r espectiv el y (Table 3 ). Ov er all, the inter action model between BbLptC and BbLptF illustr ates a m uc h mor e cohesiv e structur e, particularly in the connections of the beta-taco folds from each pr otein r ather than that of the disjointed interaction shown between BbLptC and BbLptG (Fig. 6 C and D, r espectiv el y). The pr edicted interaction interface was predicted to be between residues S34-V46 of BbLptC and residues Y224-Y234 of BbLptF. In addition to the structural models, the ipTM + pTM score of 0.79 for BbLptC/BbLptF indicates a m uc h mor e likel y inter action than the 0.67 score of the BbLptC/BbLptG multimer (Table 3 ). This observation also is consistent with pr e vious studies suggesting that LptC inter acts specificall y with LptF r ather than LptG in E. coli (Benedet et al. 2016 ). To support this finding, BbLptC, BbLptF, and BbLptG were modeled together as a trimer using AlphaFold2.1, and the result was consistent with BbLptC interacting preferentially with BbLptF ( Fig. 6 E). Additionall y, the ov er all pLDDT was consider ed confident with a score of 81.98, and the predicted interface of the interaction was close to the BbLptC/BbLptF ipTM + pTM at 0.74 (Table 3 ).

Gener a tion of an IPTG regulatable BbLptD mutant
Pr e vious genome-wide transposon mutagenesis studies in B. burgdorferi yielded no mutants with transposons in bb0838 (re- ferred to as bblptD forw ar d) (Lin et al. 2012 ). This suggests that bblptD and the protein it encodes are essential in B . bur gdorferi , which is consistent with observations that LptD is essential in Gr am-negativ e or ganisms (Br aun andSilhavy 2002 , Wernebur g et al. 2012 ). Given these prior observations, it was not surprising that we were unable to obtain a bblptD deletion mutant in B . bur gdorferi after repeated attempts . T herefore , we generated an IPTG-r egulatable bblptD m utant to utilize in further studies. As illustrated in Fig. 7 (A), the IPTG-inducible flacp promoter was inserted upstream of bblptD . We confirmed all plasmids found in the parental B31-5A4 LK strain were also present in the mutant str ain, designated flacp::bblptD . Anal ysis of whole cell l ysates of uninduced (0 mM IPTG) flacp::bblptD vs. B31-5A4 LK wild-type organisms illustrates a greatly reduced level of BbLptD expression (Fig. 7 B). Expression of BbLptD is restored almost back to wild-type le v els with only 0.01 mM IPTG added to the growth media and is ov er expr essed with 1 mM IPTG added to the media (Fig. 7 B). As expected, FlaB le v els r emained consistent r egardless of the amount of supplemented IPTG added to the growth media (Fig. 7 B).

Lipoprotein CspA is transported to the surface despite BbLptD do wn-regula tion
While it is known that B . bur gdorferi encodes numerous surfaceexposed lipoproteins (Fraser et al. 1997, Brightbill et al. 1999, the pr ocess involv ed in surface localization of these lipoproteins has long remained a question. It has pr e viousl y been pr oposed that all surface-exposed lipoproteins in B . bur gdorferi ar e tr ansported to the surface through the same general " flippase" (Zückert 2014 ). Because lipoproteins have a similar amphiphilic nature to LPS and serve as a component of OM asymmetry in B . bur gdorferi , we sought to investigate these lipoproteins as the potential cargo of the LPT-orthologous system in B . bur gdorferi . We utilized the IPTG regulatable BbLptD mutant to determine if lipoproteins can be localized to the surface of B . bur gdorferi when BbLptD is down-regulated. To examine this possibility, we performed surface pr oteol ysis assays combined with imm unofluor escence on whole cells of B31-5A4 LK wild-type organisms or on flacp::bblptD cultures supplemented with either 0, 0.01, or 1 mM IPTG.
For surface pr oteol ysis assa ys , or ganisms wer e incubated either with or without addition of proteinase K (PK) (Fig. 8 A). Since PK cannot penetrate the OM, only surface-exposed proteins are degraded when cells are treated with PK. Following surface proteol ysis, eac h PK-tr eated sample and the corr esponding contr ol was immunoblotted with antibodies specific for CspA, an abundant surface-exposed lipoprotein in B . bur gdorferi (Cordes et al. 2004 ). As shown in Fig. 8 (A), when B31-5A4 LK wild-type cells ar e tr eated with PK, ther e is full degr adation of CspA, indicating it is surface exposed as expected. Similarly, when the BbLptD r egulatable str ain was supplemented with 0.01 or 1 mM IPTG, we found that CspA is fully degraded and, therefore, surfaceexposed (Fig. 8 A). Inter estingl y, we also observ ed that CspA was full y degr aded fr om the borr elial surface when the BbLptD r egulatable strain was cultured without IPTG (Fig. 8 A). Full degradation of CspA when BbLptD is not induced for expression indicates CspA is not dependent on BbLptD for transport to the B . bur gdorferi surface . As a control, we also examined PK-treated and untreated cells by immunoblot with P66-specific antibodies, a known membrane-spanning, surface exposed OMP. P66 was degraded and resulted in a smaller, 50 kDa band as shown pr e viousl y (Curtis et al. 2022 ). Lastly, the periplasmic FlaB protein was also  The pLDDT per r esidue scor e is av er a ged to give the numbers shown for each r espectiv e pr otein. ‡ : The ipTM + pTM score is a measure of confidence in the predicted interaction. This is weighted 80% ipTM + 20% pTM.
used for immunoblots on the same whole-cell lysates to confirm the organisms' OMs were not disrupted and that all degradation was indicative of only surface protein degradation (Fig. 8 A).
To more closely examine whether surface-exposure of CspA is dependent on BbLptD expression, we also performed surface localization imm unofluor escence assays on B31-5A4 LK wild-type cells and the regulatable mutant grown in 0, 0.01, or 1 mM IPTG (Fig. 8 B). Cells were incubated with CspA antibodies either before or after permeabilization of the OM. Since antibodies do not penetrate the OM of B . bur gdorferi , immunofluorescence observed on intact cells indicates surface localization. As shown in Fig. 8 (B), wild-type and all mutant organisms strongly fluoresced with the CspA antibody, indicating that CspA is surface exposed regardless of BbLptD expression (Fig. 8 B, left-hand panels). To ensure that OMs were not disrupted prior to antibody incubation, cells were also incubated with FlaB-specific antibodies. In the unfixed, intact cells (Fig. 8 B, center panels), there is no fluorescence of the FlaB antibodies, indicating the OMs were intact. As a control, the same experiment was performed on fixed cells and assayed with CspA ( Fig. 8 C, left-hand panels) or FlaB (Fig. 8 C, center panels) antibodies and, as expected, both CspA and FlaB were observed by fluorescence. To visualize all spirochetes in each microscopic field, all slides were stained with DAPI ( Fig. 8 B and C, right-hand panels). Taken together, the surface pr oteol ysis and imm unofluor escence assays illustr ate that the down-r egulation of BbLptD in the flacp::bblptD 0 mM IPTG mutant does not affect the surface localization of CspA, and str ongl y suggests that BbLptD and the B. burgdorferi LPT system are not r equir ed for lipopr otein tr ansport to the surface.

Structural modeling of the B. burgdorferi BbLptA with borrelial glycolipids
Since the data indicated surface lipoproteins do not r equir e BbLptD to be transported to the surface, we next investigated the possibility that the amphiphilic glycolipids could be the cargo. To begin these studies, we used computational structural modeling. We focused these studies on interactions between BbLptA and the two major B . bur gdorferi glycolipids , BbGL-I and BbGL-II (Figure S1, Supporting Information). Using AlphaFold2.1, BbLptA was modeled similarly to the solved structures of the LptA proteins of Gr am-negativ e or ganisms. Specificall y, it has a tightl y pac ked interior composed entir el y of hydr ophobic r esidues and a surface highl y enric hed in hydr ophilic and polar r esidues. Since the pr eviousl y solv ed structur es of LptA pr oteins hav e been done without bound cargo, we sought to model BbGL-I and BbGL-II in the hydrophobic interior of BbLptA to explore the potential interaction between glycolipids and BbLptA. The beta-taco of BbLptA was closed, which excludes traditional docking approaches from placing the glycolipids in the hydrophobic interior, so we used a combination of tools to explore the flexibility of BbLptA with BbGL-I and BbGL-II.
First, a SILCS (Site Identification by Ligand Competitive Satur ation) sim ulation (Guv enc h and MacK er ell 2009 ) was performed. This utilizes combined grand canonical Monte Carlo (GCMC) with molecular dynamics (MD) in a mixed-solv ent envir onment containing small probes of a variety of biochemical characteristics suc h as pr opane , benzene , methanol, acetate , and methylammonium, among others (Lakkaraju et al. 2015 ). This mix of small molecules can aid in the identification of binding regions of larger molecules based on shar ed c har acteristics between the small molecules and large molecules (MacKerell et al. 2020 ). Regions with higher probe residency time during the simulations suggest favor able inter actions, and inter action fr ee ener gy ma ps of v arious functional gr oups, Fr a gMa ps, can be generated based on these pr obe r esidency times (Guv enc h and MacK er ell 2009 ). Because of The flacp IPTG-regulatable promoter and the stre ptom ycin resistance cassette were inserted upstream of bblptD through homologous recombination. Homologous recombination was ac hie v ed by first generating a construct in which the regions 600 bp upstream and downstream of the insertion site were cloned into pBluescript SK + . The stre ptom ycin cassette as w ell as the flacp promoter w ere then cloned into the construct. The final construct w as electr opor ated into B . bur gdorferi B31-5A4 LK, and a stre ptom ycin resistant clone was selected and designated flacp :: bblptD . A full plasmid analysis was also performed on the mutant strain (data not shown). (B) IPTG dose-dependent expression of BbLptD in the flacp::bblptD mutant strain. The B. burgdorferi flacp::bblptD mutant strain was cultivated in 0, 0.01, or 1.0 mM IPTG, and whole-cell lysates of the mutant and wildtype strains were subjected to immunoblot with BbLptD antibodies. Whole-cell lysates immunoblotted with FlaB antibodies are also shown as controls for equal loading.
the flexibility that MD affords, SILCS has been used to identify occluded or cryptic pockets not present in experimental or Alphafold apo protein structures (Lakkaraju et al. 2015 ).
Using the Fr a gMas fr om the SILCS sim ulations, we combined SILCS-Hotspots and shape o verla y tools to place the glycolipids within the interior pocket of the structural model of BbLptA. SILCS-Hotspots performs Monte Carlo-based "docking" of fragments into the Fr a gMa ps acr oss the entir e pr otein to identify binding sites for each fragment (MacKerell et al. 2020 ). Because of its large size, BbGL-I was decomposed into smaller fragments, and binding conformations for each fragment were predicted at eac h inter action site. To place the entir e gl ycolipid into BbLptA, we sought to find a low-energy conformation of BbGL-I that also resembled the same favorable interactions in the interior of the beta-taco that wer e cur ated fr om the SILCS-Hotspots r esults. A set of over 20 000 low energy conformations was generated by OMEGA, and ROCS (Ra pid Ov erlay of Chemical Structur es) was used to identify conformations of BbGL-I that overlaid with the hotspot BbGL-I fr a gments ( R OCS 3.5.1.1 OpenEy e 2007 , OpenEy e Scientific Software 2010 , Hawkins et al. 2007Hawkins et al. , 2010. The pose with the highest o verla y score that did not pr otrude thr ough the beta sheets was selected. Finally, because the closed model of BbLptA clashed with the glycolipid pose, we employed a MD simulation to allow the protein to undergo a conformational change to accommodate the glycolipid. The complex was solv ated, neutr alized, minimized, and then equilibrated in the presence of restraints for 100 ps . T he restraints were then removed and a 200 ns MD simulation was performed using GROMACS (Groningen Machine for Chemical Simulations) (Abraham et al. 2015 ). BbGL-II was modeled into the structure of BbLptA in final frame of the MD simulation of the BbGL-I complex using the same protocol, except the final BbGL-I conformation was used as the query for ROCS.
During the MD simulations, both the protein and ligand had a r a pid shift a wa y from the starting structure upon the release of the r estr aints as the pr otein and ligand both moved into lo w er energy conformations; ho w ever, in each simulation the protein and ligand seemed to conv er ge to w ar d a lo w er-energy conformation at around 100 ns ( Fig. 9 A/B and D/E). The final frame of each trajectory provides a snapshot of how each glycolipid potentially interacts with the protein as it slides through BbLptA ( Fig. 9 C/F). Given that the gl ycolipids ar e adjacent to where BbLptC would be located, these models r epr esent how the glycolipid may bind after they are passed from BbLptC to BbLptA. In both models, the sugar is oriented a wa y fr om wher e BbLptC would be located and would theor eticall y tr av el to w ar d the OM. The sugar of the glycolipid is pointed upw ar d to inter act with the hydr ophilic surface r esidues and solvent, while the lipid tails are oriented to w ar d the inner membrane ( Fig. 9 C/F). Further empirical studies will be needed to better define this mechanism.

Discussion
Pr e viousl y, BB0807, BB0808, BB0465, BB0466, and BB0838 have been identified as Lpt-related proteins in B . bur gdorferi (Putker et al. 2015. We further examined the similarity between the borrelial proteins and known Lpt orthologs using a combination of genetic and structural data. Given that the genetic locus encoding lptC in Gr am-negativ e or ganisms is al ways found to consist of lptC dir ectl y upstr eam of lptA and lptB , we hypothesized that bb0464 was the lptC ortholog in B . bur gdorferi . Subsequent structural modeling confirmed BB0464 as the LptC ortholog FlaB antibodies were included in imm unofluor escence assays to ensure the OM remained intact during the surface assa ys , and DAPI counter-staining was included to identify all organisms in a given field in (B) and (C). as the structure was remarkably similar to other LptC proteins. Additionally, LptC has been shown to preferentially interact with LptF in E. coli (Benedet et al. 2016 ). The multimer modeling performed her e pr edicted that BB0464 pr efer entiall y inter acts with BB0807. Combined with the better RMSD scores for BB0807 aligned to LptF and BB0808 aligned to LptG, r espectiv el y, we hav e further c har acterized BB0807 as the LptF ortholog and the BB0808 as the LptG ortholog. The genetic and structur al anal yses also allo w ed us to confirm BB0465, BB0466, and BB0838 as LptA, LptB, and LptD orthologs, r espectiv el y. Accordingl y, the B . bur gdorferi proteins BB0464, BB0465, BB0466, BB0807, BB0808, and BB0838 should be referr ed to, r espectiv el y, as BbLptC, BbLptA, BbLptB, BbLptF, BbLptG, and BbLptD moving forw ar d.
Inter estingl y, the structur al models of BbLptD and BbLptC differ ed slightl y fr om their Gr am-negativ e counter parts. Specificall y, these differences included the alpha helices at the N-terminal domain of BbLptD and the prediction that BbLptC is a lipoprotein. Differences in overall protein structure are not surprising considering the function of the B . bur gdorferi LPT system clearly div er ges fr om LPS tr ansport. Additionall y, while Fig. 2 illustrates a single BbLptA molecule linking BbLptC to BbLptD, it is important to note that LptA oligomerizes to span the length of the periplasm in other organisms (Santambrogio et al. 2013 ). The periplasmic space of B. burgdorferi is ∼160 Å (Charon et al. 2009 ), and the r espectiv e lengths of eac h beta-taco fold (measur ed thr ough PyMOLv2.4.0; Sc hrödinger 2020 ) of BbLptC, BbLptA, and the N-terminal domain of BbLptD are 41.9, 55.2, and 52.9 Å . The combined a ppr oximation of 150 Å indicates that it is entir el y possible that BbLptA functions as a monomer in this system. Ho w e v er, giv en the substantiall y lar ger periplasmic space pr esent wher e the endofla gella ar e located (Char on et al. 2009 ), it is also possible that multiple BbLptA proteins are needed in these regions. Further studies would be needed to resolve these questions.
While the genetic and structural data strongly indicated B. burgdorferi encoded a nov el LPT tr ansport system, we sought to determine empirically that these orthologs do indeed inter act, whic h would be r equir ed for them to form a bridge between the inner and OMs. We examined the interactions between the N-terminal periplasmic domain of BbLptD with BbLptA as well as BbLptC with BbLptA. These three components of the LPT system are known to be the periplasmic bridge proteins in other or ganisms (La guri et al. 2017 ), and w e w anted to confirm the borrelial orthologs also inter acted similarl y, whic h would indicate they could link the inner Figur e 9. MD simulations con verged to a final model of BbLptA with two glycolipids. (A) Plot of C-alpha RMSD vs. time of BbLptA in the simulation with BbGL-I r elativ e to its position following equilibr ation. Upon r elease of the r estr aints, ther e is a r a pid 4 Å c hange in RMSD, while after 100 ns, it conv er ges to around 2.5 Å from the starting structure. (B) Plot of ligand RMSD vs. time of BbGL-I r elativ e to its position following equilibration. Upon release of the restraints, there is a rapid 4 Å change in RMSD, while after 50 ns, BbGL-I conv er ges to around 12 Å from the starting structure . T his corresponds to the acyl chain folding under the carbohydrate and sterol. (C) The beta taco fold of BbLptA (green) with BbGL-I (salmon) following 200 ns MD simulation. The beta taco has opened to allow the sterol group and acyl chain to insert into the hydrophobic interior, while the carbohydrate is oriented outw ar ds to inter act with the hydr ophilic surface of BbLptA or solv ent. The r ed atoms on BbGL-I indicate oxygens pr esent in the sugar gr oup or gl ycer ol. The first arr ow r epr esents a 90 • r otation along the x -axis . T he second arr ow r epr esents an additional 90 • r otation along the y -axis. (D) Plot of C-alpha RMSD vs. time of BbLptA in the simulation with BbGL-II relative to its position following equilibration. Upon release of the restraints, there is a r a pid 2 Å change in RMSD, while after 170 ns, it conv er ges to around 3 Å from the starting structure. (E) Plot of ligand RMSD vs. time of BbGL-II r elativ e to its position following equilibration. Upon release of the restraints, there is a rapid 4 Å change in RMSD, while after 100 ns, it conv er ges to around 7 Å from the starting structure . T his corresponds to the carbohydrate orienting itself in the direction of the OM. (F) The taco fold of BbLptA (c y an) with BbGL-II (purple) following 200 ns MD simulation. As with BbGL-I, the beta-taco has opened to allow the acyl chains to insert into the hydrophobic interior, while the carbohydrate is oriented outw ar ds to interact with the hydrophilic surface of BbLptA or solvent. The red atoms on BbGL-I indicate oxygens present in the sugar group or glycerol. The first arrow represents a 90 • rotation along the x -axis . T he second arrow represents an additional 90 • rotation along the y -axis. and OMs in B . bur gdorferi . Utilizing the pACYCDuet coexpression and copurification system, we observed that BbLptC interacts with BbLptA and that BbLptA interacts with the periplasmic domain of the OMP BbLptD. The combined findings are most consistent with the BbLptC/BbLptA/BbLptD proteins interacting and provid-ing the periplasmic bridge for the LPT system in B . bur gdorferi . Future studies examining the sites of interaction and the residues involved in the interactions between these proteins will be necessary to fully understand which residues are most relevant to these pr otein-pr otein inter actions.
B . bur gdorferi contains six of the se v en orthologs to Gramnegative Lpt proteins, but it lacks an LptE ortholog. LptE is essential in Gamma pr oteobacteria and is known to form a stable complex with LptD with LptE residing inside the beta-barrel of LptD , Botos et al. 2016. LptE has been shown to play se v er al r oles in the LPT system including, but not limited to, acting as a " plug" to the beta-barrel in LptD to inhibit continuous LPS transport (Grabowicz et al. 2013 ), binding to and inhibiting a ggr egation of LPS to allow pr oper tr ansport (Maloj či ć et al. 2014 ), and being essential to the folding of LptD through stabilization and proper oxidation of the two di-sulfide bonds in LptD , Chimalakonda et al. 2011, Lo Sciuto et al. 2018. LptE is known to be the least conserved protein in the LPT system, and pr e vious studies hav e found difficulties in identifying LptE orthologs in other organisms outside of the Gammaproteobacteria group (Putker et al. 2015 ). Giv en the gener al lac k of sequence conservation seen in the B . bur gdorferi LPT-system orthologs, this is one possible explanation for the lack of identification of an LptE ortholog in B . bur gdorferi . Ho w e v er, it is also possible that there is no LptE ortholog in B . bur gdorferi because it is not essential to the function of the LPT-orthologous system. BbLptD does not contain cysteines, so one of the known roles of LptE in assisting di-sulfide bond formation would not be a r equir ed function in B . bur gdorferi . Additionally, it may also be possible that an LptE ortholog is not essential for the transport of the yet-to-be-identified cargo of this system, as the other roles of LptE all involve direct interaction with LPS.
With the presence of an inner membrane permease BbLptF/BbLptG and the ATP-binding cassette, consisting of a homodimer of BbLptB, it is very likely that the novel LPT system in B . bur gdorferi transports an OM constituent to the surface of this spir oc hete. What the car go for the borr elial LPT system might be is still unclear. Possible candidates include two major OM localized constituents that contain fatty acids and have a similar amphiphilic nature to LPS. These are the surface lipoproteins and the unique borr elial gl ycolipids. While both the lipoproteins and the glycolipids of B . bur gdorferi are found to be immunogenic and contribute to disease pathogenesis (Sc hr oder et al. 2008, Stübs et al. 2009, Kenedy et al. 2012, it is unkno wn ho w these major OM constituents ar e tr ansported to the surface of this spir oc hete. Giv en that the B . bur gdorferi genome does not likely encode specific transporters for each of the more than 100 known surface lipoproteins (Schulze and Zuck ert 2006 , Zück ert 2014 ), it has been proposed that there is a general lipoprotein flippase that can tr ansport lipopr oteins fr om the periplasm to the surface (Zückert et al. 2004 ). It is tempting to speculate that this LPT-orthologous system in B . bur gdorferi ma y serve as a route for lipoprotein surface localization, particularly due to the amphiphilic nature of the lipoproteins and the membrane asymmetry they provide in B . bur gdorferi. We in vestigated this possibility using the r egulatable BbLptD m utant gr own in 0 mM IPTG, whic h r esults in down-r egulation of BbLptD expr ession at a le v el wher e it is not e v en observ ed by imm unoblot anal ysis. Even with the lack of observable BbLptD, we observed no change in the surface localization of the major surface lipoprotein CspA. The observation that CspA is still localized to the surface of B. burgdorferi , e v en when BbLptD is not expressed or, at a minimum, down-regulated to an extremely low level, suggests that BbLptD and the LPT system do not function as a lipopr otein tr ansport system in B . bur gdorferi .
A r ecent r eport has emer ged fr om He et al. ( 2023 ) addr essing similar questions proposed here . T he in vestigators had similar computational findings about the presence of a B . bur gdorferi LPT system, and we have provided further empirical evidence that this system exists in this spir oc hete and that the proteins within the system interact. The conclusions from their study, ho w ever, are differ ent fr om our observ ations. He et al . ( 2023 ) determined that surface lipoproteins, including CspA, are not properly localized to the surface when BbLptD is down-regulated, while we found that CspA is surface localized independent of BbLptD expression. It should be noted, ho w e v er, that the methodology to create a BbLptD knockdown system between the two studies differed significantl y. Her e, we use a well-established method (Caimano et al. 2007, Gilbert et al. 2007, Lenhart and Akins 2010, Dunn et al. 2015, Iqbal et al. 2016 , Dr ec ktr ah and Sam uels 2018 ) to r egulate the expression of BbLptD through an inducible promoter. He et al . ( 2023 ) used a newer CRISPR interference B . bur gdorferi system to knoc kdown BbLptD (Mur phy et al. 2023 ). Ov er all, our data led us to conclude that, rather than lipoproteins, the glycolipids are the likel y car go of the B . bur gdorferi LPT system-a possibility that He et al . ( 2023 ) do not presume, but also do not exclude.
The conjecture that glycolipids are the most likely cargo of the B . bur gdorferi LPT system is consistent with the fact that Treponema pallidum , another spir oc hete that lacks LPS, also contains a predicted LPT system (Putker et al. 2015, Hawley et al. 2021 ). Unlike B . bur gdorferi , ho w e v er, T. pallidum does not have an abundance of surface-exposed lipoproteins and, in fact, only contains three potential surface-exposed lipoproteins (Radolf and Kumar 2018 ). T. pallidum is similar to B . bur gdorferi in that it also contains specific glycolipids (Radolf et al. 1995 ). Ther efor e, it seems m uc h mor e likel y that the r ole of the LPT system in these spir ochetes would be to transport glycolipids. Considering the dearth of surface-exposed lipoproteins in T. pallidum , we would also propose that glycolipids are the most likely cargo of the treponemal LPT system, which was also previously proposed by Hawley et al. ( 2021 ). Inter estingl y, other LPS-lac king diderms suc h as Thermus thermophilus and Thermotoga maritima also contain Lpt-orthologs (Putker et al. 2015 ) and possess their own unique membrane glycolipids (Manca et al. 1992, Leone et al. 2006 ) but do not encode surface lipoproteins . T his is again consistent with our suggestion that gl ycolipids ar e the car go for these LPT systems in organisms that lack LPS.
The BbLptA/BbGL-I and BbLptA/BbGL-II computational modeling provides insight into the possibility that the borrelial glycolipids are indeed the cargo of the LPT system in B . bur gdorferi . This was illustrated based on the identification of putative binding sites for glycolipids inside the beta-taco fold using the SILCS tec hnology. Subsequent doc king model MD sim ulations pr edicted a 3D model of the binding of the glycolipids to the proteins . T he buried lipids make nonspecific hydrophobic interactions in the interior of the beta-taco while the carbohydrate interacts with solvent and the hydrophilic surface side chains . T he predicted binding poses are consistent with the ATP ase, BbLptB, pr oviding the energy to slide the glycolipid through the channel by pushing another glycolipid behind it. Based on the favorable nature of the gl ycolipid-pr otein inter action, the models pr opose a r easonable method of transport of glycolipids through BbLptA. Future studies utilizing regulatable mutants of the B . bur gdorferi Lpt-orthologs will be necessary to definitiv el y determine if glycolipids are transported by this system.
With so few trans-envelope transport/export systems having been identified in B . bur gdorferi , m uc h is still unknown about this spir oc hete's OM biogenesis . T he identification of a novel LPT system will have high impact on future B . bur gdorferi OM biogene-sis studies and may also provide a novel target for antimicrobials. While doxycycline and other broad spectrum antibiotics are very effective in treating Lyme disease, the USA has seen a large increase in Lyme disease without any indication of a future decline in cases (Stone et al. 2017 ). This has led to discussion of control and/or elimination of B . bur gdorferi in the tic k r eservoir (Dolan et al. 2011, Richer et al. 2011, Bernard et al. 2020 ). Doxycycline containing bait targeting mice in Lyme disease endemic regions have pr ov ed effectiv e in dr asticall y r educing the number of small mammals and ticks infected with B . bur gdorferi , but on a large scale, this strategy causes concern on the development of antibiotic resistance to the most effective Lyme disease drug (Dolan et al. 2011, Bernard et al. 2020. The identification of this novel and essential transport system in B . bur gdorferi could provide an additional target to combat this issue. Additionally, the identification of an LPTorthologous system in LPS-lacking B . bur gdorferi pro vides no vel insight into the function of the LPT system in gener al. Perha ps, the sole and specific role of this system is not to transport LPS, but rather to transport a variety of fatty acid-containing molecules to the surface of different diderm species. We hav e pr ovided computational and empirical evidence of the presence of such a system in B . bur gdorferi . We ha ve identified an LPT-orthologous system in this spir oc hete and hav e thus far pr ovided e vidence that lipopr oteins are not transported by this novel LPT system in B . bur gdorferi . While we hypothesize here that the borrelial glycolipids are the likel y car go of the LPT-orthologous system in B . bur gdorferi , further mechanistic studies will be required to examine this issue for confirmation.