A Borrelia burgdorferi LptD Homolog is Required for Flipping of Surface Lipoproteins Through the Spirochetal Outer Membrane

SUMMARY

Borrelia spirochetes are unique among diderm bacteria in their lack of lipopolysaccharide (LPS) in the outer membrane (OM) and their abundance of surface-exposed lipoproteins with major roles in transmission, virulence, and pathogenesis. Despite their importance, little is known about how surface lipoproteins are translocated through the periplasm and the OM. Here, we characterized Borrelia burgdorferi BB0838, a distant homolog of the OM LPS assembly protein LptD. Using a CRISPR interference approach, we showed that BB0838 is required for cell growth and envelope stability. Upon BB0838 knockdown, surface lipoprotein OspA was retained in the inner leaflet of the OM, as determined by its inaccessibility to in situ proteolysis but its presence in OM vesicles. The topology of the OM porin/adhesin P66 remained unaffected. Quantitative mass spectrometry of the B. burgdorferi membrane-associated proteome confirmed the selective periplasmic retention of surface lipoproteins under BB0838 knockdown conditions. Additional analysis identified a single in situ protease-accessible BB0838 peptide that mapped to a predicted β-barrel surface loop. Alphafold Multimer modeled a B. burgdorferi LptB₂FGCAD complex spanning the periplasm. Together, this suggests that BB0838/LptD_Bb facilitates the essential terminal step in spirochetal surface lipoprotein secretion, using an orthologous OM component of a pathway that secretes LPS in proteobacteria.

Graphical Abstarct

CRISPRi knockdown of Borrelia burgdorferi BB0838, an essential outer membrane protein related to proteobacterial LptD lipopolysaccharide transporters, disrupted the translocation of surface lipoproteins through the spirochetal outer membrane but did not affect the topology of outer membrane porins or periplasmic lipoproteins. Thus, BB0838 may function as a surface lipoprotein flippase.

INTRODUCTION

Borrelia spirochetes, the causative agents of vector-borne relapsing fever and Lyme disease, are diderm bacteria with a rather distinct envelope structure. Unlike other diderm bacteria such as the gram-negatives – and even some other spirochetal genera such as Leptospira and Brachyspira – Borrelia are deficient in lipopolysaccharide or lipooligosaccharide (LPS/LOS) biosynthesis pathways and consequently lack LPS/LOS in the surface leaflet of the OM (Takayama et al., 1987, Wanchanthuek et al., 2010, Fraser et al., 1997, Zückert, 2019). Instead, their bacterial surface is dominated by abundant, immunogenic, and serotype-defining surface lipoproteins that are differentially expressed and play major roles throughout subsequent phases of the vector-host transmission cycle (reviewed in (Lopez et al., 2021, Coburn et al., 2021)).

The complexity of the lipoprotein-mediated interface of Borrelia with their vectors and hosts was further uncovered when we showed that two-thirds of the lipoproteome expressed by the Lyme disease pathogen Borrelia burgdorferi localize to the surface (Dowdell et al., 2017). Yet, the molecular mechanisms guiding this diverse cohort of about 90 B. burgdorferi lipoproteins to the spirochetal surface have only slowly come into focus. In other diderm bacteria, lipoproteins destined for the outer membrane (OM), such as the prototypical Braun’s lipoprotein Lpp, are extracted by the Localization of lipoprotein (Lol) pathway and transported through the periplasm to be anchored in the periplasmic leaflet of the OM (Tokuda & Matsuyama, 2004, Okuda & Tokuda, 2011, Zückert, 2014). The B. burgdorferi genome encodes for a partial Lol pathway; the OM lipoprotein receptor/insertase LolB found in γ-proteobacteria is missing, as are any classical Type 2, Type 5, or surface lipoprotein assembly modulator (SLAM) secretion systems that provide surface access to distinct lipoproteins in other diderm bacteria (Fraser et al., 1997, Zückert, 2019, Pugsley et al., 1986, d’Enfert et al., 1987, van Ulsen et al., 2003, Roussel-Jazede et al., 2013, Hooda et al., 2016).

We previously used B. burgdorferi monomeric OspA and dimeric OspC and Borrelia turicatae Vsp1 as model surface lipoproteins in conjunction with a fluorescent mRFPΔ4 localization reporter (Schulze et al., 2010) to show that their sorting determinants are encoded within the disordered N-terminal tether peptides; yet, Lol pathway sorting rules established for gram-negative IM and OM lipoproteins did not apply (Schulze & Zückert, 2006, Schulze et al., 2010, Kumru et al., 2010, Kumru et al., 2011). Mutations within tether peptides generally led to the mislocalization of surface lipoproteins to the periplasmic leaflet of the OM (Schulze & Zückert, 2006, Schulze et al., 2010, Kumru et al., 2010, Kumru et al., 2011). However, this mislocalization could be rescued by introducing mutations or conditions that destabilized the fold of OspA or OspA-calmodulin fusions, respectively (Schulze et al., 2010, Chen & Zückert, 2011). This indicated that surface lipoproteins crossed the OM in an at least partially unfolded conformation. Fittingly, dimeric lipoproteins were shown to assemble on the bacterial surface (Kumru et al., 2011). Tether peptide protease accessibility studies for selected model surface lipoproteins indicated that these surface lipoproteins are indeed anchored in the surface leaflet of the OM (Chen et al., 2011). Together, this suggested the requirement for an OM “lipoprotein flippase” machinery that translocated both the polypeptide through the OM while also facilitating the flipping of the lipid anchor from the periplasmic to the surface leaflet of the OM (Zückert, 2019).

In their studies of the B. burgdorferi OM proteome, Lenhart and Akins (2010) showed that depletion of the B. burgdorferi β-barrel assembly machinery (BAM) protein BamA led to a significant growth defect and also reduced the abundance of surface lipoproteins in the OM (Lenhart & Akins, 2010). This indicated that proper localization of OM lipoproteins is dependent on BAM, most likely due to the BAM-mediated assembly of an essential integral OMP serving as the OM lipoprotein flippase. Here, we used a plasmid-based non-toxic CRISPR interference (CRISPRi) system (Murphy et al., 2023) to show that depletion of BB0838, a distant B. burgdorferi homolog of the proteobacterial OM LPS translocase LptD, leads to selective retention of surface lipoproteins on the periplasmic side of the OM. Assembly of a β-barrel OM protein remained unaffected. This indicates that BB0838/LptD_Bb functions downstream of BAM, enabling OM translocation of B. burgdorferi surface lipoproteins.

RESULTS

Identification of BB0838 as a putative B. burgdorferi OM LptD homolog.

To overcome the biophysical obstacle of moving an amphipathic molecule across a lipid bilayer membrane, an OM lipoprotein flippase must have two separate domains: a hydrophilic transmembrane lumen that accommodates polypeptides of various dimensions, and a hydrophobic cavity that shelters the lipid moiety. To narrow down our list of candidate proteins, we used Phyre2 (Kelley et al., 2015) to generate structural models of 41 B. burgdorferi proteins that had been bioinformatically predicted to localize to the OM and/or assume a β-barrel structure (Kenedy et al., 2016) (Table S1). Among this set were 3 proteins with structural homologs that folded into β-barrels with periplasmic domains: The TolC homolog BB0142/BesC (Bunikis et al., 2008), BB0795/BamA_Bb (Lenhart & Akins, 2010), and BB0838 (Kenedy et al., 2016).

We modeled the structure of BB0838 using i-Tasser and AlphaFold 2 (Yang & Zhang, 2015, Jumper & Hassabis, 2022) (Fig. 1). i-Tasser built the top model on the Shigella flexneri complex of LptD and LptE (C-score −2.22, T-score 0.45 ± 0.15, RMSD 14.9 ± 3.6 Å on PDB accession number 4Q35) despite less than 20% amino acid identity between them (Fig. 1A and Supplemental Fig. S1). In gram-negative bacteria, LptD together with lipoprotein LptE form the OM translocon for transporting LPS across the OM and displaying it on the cell surface (reviewed in (Konovalova et al., 2017)). The proteobacterial LptD N terminus folds into a β-jellyroll domain with 20 antiparallel β-strands forming a hydrophobic groove extending into the periplasm while the LptD C terminus forms a β-barrel transmembrane domain with 26 antiparallel β-strands (Dong et al., 2014). The LPS lipid A moiety is directly inserted into the OM via LptD’s N-terminal hydrophobic core while the hydrophilic oligosaccharide core and polysaccharide O antigen moiety of LPS enter the OM via the hydrophilic, laterally gated transmembrane lumen of the LptD β-barrel. LptE is anchored in the periplasmic leaflet of the OM and inserts into the LptD β-barrel lumen to facilitate the re-orientation of LPS to the bacterial surface (Chimalakonda et al., 2011, Chng et al., 2010, Freinkman et al., 2011, Ruiz et al., 2010, Wu et al., 2006).

FIG. 1. Structural Modeling of BB0838. (A) i-Tasser and AlphaFold models of BB0838.

Both algorithms predict a BB0838 fold similar to LptD. In both models, the N-terminal periplasmic β-jellyroll domain is colored in yellow, the central β-barrel domain in orange, and the C-terminal extensions form either a coiled “plug” or β-barrel extension domain in purple. The i-Tasser model is shown in a side view emphasizing the predicted plug domain. The Alphafold model is shown in a front view emphasizing its separate N-terminal α-bundle and β-jellyroll domains. Both top views are rotated 90° along the vertical axis. The amino and carboxy termini of the protein are indicated by N and C, respectively, in both side/front and top models. Green arrows point to a proteinase K-accessible surface loop (see Supplemental Table S5 and Fig. S3). Stippled brackets indicate the approximate position of the OM lipid bilayer. (B) Comparison of predicted domain structures of spirochetal LptD homologs to gram-negative LptD. Color coding and domain labeling is as in panel A. Sf, S. flexneri; Bb, B. burgdorferi; Tp, T. pallidum; and Li, L. interrogans. ORF numbers for each homolog are indicated below each domain bar, and numbers indicate the total preprotein length in amino acids. Note that only LPS-containing L. interrogans is predicted to have a LptE homolog.

The i-Tasser BB0838 model predicted a slightly shortened N-terminal β-jelly roll domain composed of 18 antiparallel β-strands preceded by a coiled-coil. Compared to S. flexneri LptD, the β-barrel domain expanded to 28 antiparallel β-strands due to 94 additional amino acids, suggesting an expanded transmembrane lumen that may accommodate larger cargo. In addition to a predicted larger β-barrel, BB0838 has an additional 362 amino acid C-terminal extension that i-Tasser modeled as a coiled-coil on LptE within the LptD lumen (Fig. 1). The AlphaFold model of BB0838 (Fig. 1A and Supplemental Table S2) similarly predicted a β-jelly roll domain of 18 antiparallel β-strands, but the extreme N terminus assumed an additional domain consisting of a cluster of four short α-helices connected to the β-jelly roll via a flexible linker. In another contrast to the i-Tasser model, BB0838’s C-terminal extension was incorporated into the β-barrel as additional β-strands, leading to a 32-stranded β-barrel that was partially restricted by a large periplasmic loop. In both models, the β-barrel had a lateral opening aligned with the periplasmic β-jelly roll domain. We, therefore, concluded that BB0838 indeed represents a structural LptD homolog that–in the absence of lipidated polysaccharides within the system–may have evolved to translocate other amphipathic envelope components such as lipoproteins instead. Intriguingly, even distant and smaller LptD homologs of other spirochetal model organisms, such as Treponema pallidum and Leptospira interrogans, show a similar domain structure with a C-terminal extension, but only the LPS-containing L. interrogans also encodes for a LptE homolog (Hawley et al., 2021) (Fig. 1B and Supplemental Table S3).

Generation of a conditional BB0838 knockdown strain using CRISPR interference.

A B. burgdorferi Tn mutagenesis screen suggested that BB0838 is likely essential for cell viability (Lin et al., 2012, Lin et al., 2014). BB0838 is the last gene in a 3-gene operon downstream of uvrB and uvrA (Fig. 2A), which appear non-essential for growth based on the analysis of a ΔuvrA strain (Sambir et al., 2011). We, therefore, generated a conditional BB0838 knockdown strain using a fully inducible, non-toxic CRISPR interference (CRISPRi) system encoded by a single recombinant E. coli/B. burgdorferi shuttle vector, pJJW101 (Murphy et al., 2023). pJJW101 carries a B. burgdorferi codon-optimized version of dCas9-myc under standard P_QE30 (T5/lac hybrid promoter) control, which avoids toxicity of the original Streptococcus pyogenes dCas9 in B. burgdorferi under higher induction conditions (Takacs et al., 2020). Stringency is further maximized by placing the single guide RNA (sgRNA) module under trc (trp/lac hybrid) promoter control. Gene-specific target sgRNA sequences were designed with the web-based CRISPy-web interface on an artificial contig assembly of the B. burgdorferi B31 chromosome and plasmids to eliminate potential off-target effects as described (Murphy et al., 2023, Blin et al., 2016). Four target sgRNAs, complementary to either the preferred non-template (NT) strand or the non-preferred template (T) strand (Fig. 2B and Table 1), were then ligated into the pJJW101 sgRNA module. NT1 and NT2 sgRNAs targeted an overlapping sequence close to BB0838’s 5’ end, whereas T1 and T2 sgRNAs targeted separate sequences further downstream. This resulted in 5 different strain constructs: a mock control strain transformed with a sgRNA-deficient “empty” pJJW101 plasmid, and four BB0838 knockdown (KD) strains transformed with the respective sgRNA-containing pJJW101 vectors.

FIG. 2. BB0838 operon structure and sgRNAs selected for CRISPRi.

(A) BB0838 is the last gene in a 3-gene operon with uvrB and uvrA. uvrA and BB0838 coding sequences overlap by 46 bps, i.e., solely within the signal I leader peptide of BB0838. Arrowheads show the position of all designed sgRNAs within the coding sequence for BB0838’s N-terminal β-jellyroll domain. (B) NT1, NT2, T1, and T2 sgRNA sequences are shown in bolded letters, with their start and end nucleotide positions within BB0838 indicated. Associated PAM sequences are underlined.

Table 1.

Oligonucleotide primers used in this study

Primer	Sequence (5’ to 3’)†	Description
BorFlaLeo-R-ok	GCTGGTGTGTTAATTTTTGCAG	qRT-PCR primer for flaB (de Leeuw et al., 2014)
FlaW-sense4	AGCAACTTACAGACGAAATTAATAG	qRT-PCR primer for flaB (de Leeuw et al., 2014)
Bor-0838-F1	TCTTTCGTTTGGAGGAAAGGT	qRT-PCR primer for BB0838
Bor-0838-R1	ACTGGAACTCTTCCCATGTA	qRT-PCR primer for BB0838
sgRNA T1 F	TAGTACTTGAGCTTTCAACCGATG	T1 sgRNA oligonucleotide
sgRNA T1 R	AAACCATCGGTTGAAAGCTCAAGT	T1 sgRNA oligonucleotide
sgRNA T2 F	TAGTAAAAAAATGGGCCTTAAAAG	T2 sgRNA oligonucleotide
sgRNA T2 R	AAACCTTTTAAGGCCCATTTTTTT	T2 sgRNA oligonucleotide
sgRNA NT1 F	TAGTAATTTTCATCATCTATAGTC	NT1 sgRNA oligonucleotide
sgRNA NT1 R	AAACGACTATAGATGATGAAAATT	NT1 sgRNA oligonucleotide
sgRNA NT2 F	TAGTATTTTCATCATCTATAGTCT	NT2 sgRNA oligonucleotide
sgRNA NT2 R	AAACAGACTATAGATGATGAAAAT	NT2 sgRNA oligonucleotide

^†nucleotide extensions for cloning into pJWW101 BsaI site are underlined.

BB0838 is essential for cell viability.

To test for the expected role of BB0838 in cell viability, we monitored cell growth and phenotypic changes of B. burgdorferi BB0838 KD cells over time, using non-depleted and wild-type cells as a control. 1 × 10⁵ cells/ml were inoculated in complete BSK-II medium with or without 0.25 mM IPTG and followed for 3 days post-inoculation by phase contrast microscopy. As shown by the growth curves in Fig. 3A, a marked growth defect in the induced BB0838 KD cultures emerged at day 2 post-inoculation for NT1 and NT2 sgRNA constructs; strains expressing T1 and T2 sgRNAs grew like the mock control strains. Higher inducer concentrations did not lead to more severe growth defects (data not shown). We next evaluated sgRNA efficiency on the transcript level by quantitative reverse transcription polymerase chain reaction (qRT-PCR) assays, comparing transcription levels of BB0838 and a flaB normalization control for the mock control and four sgRNA-expressing constructs. IPTG-driven induction of dCas9 and sgRNAs NT1 and NT led to an 84- and 12-fold knockdown of BB0838 mRNA transcript compared to the mock control, respectively, in total RNA samples collected at day 2. Consistent with the growth curve data, T1 and T2 sgRNA led to less efficient 3- and 2-fold transcriptional knockdowns (Fig. 3B). Due to its highest knockdown efficiency, we selected the BB0838 KD NT1 construct for further study. Phase contrast micrographs of cells immobilized on an agarose pad (Jutras et al., 2016) showed that depletion of BB0838 led to envelope disturbances and blebbing in about two-thirds of the cells (Figs. 3C and 3D).

Fig. 3. Phenotypes of B. burgdorferi BB0838 CRISPRi knockdowns.

(A) Growth curves of wild-type (WT), mock control (with “empty” pJJW101), and BB0838 KD cells harboring pJJW101 with respective NT1, NT2, T1 or T2 sgRNAs, grown under non-inducing or inducing (+0.25 mM IPTG) conditions. Cells were counted under phase contrast in a Petroff-Hausser counting chamber. Growth curves are from 3 parallel biological replicates; error bars indicate mean ± SD. (B) Total RNA samples obtained from B. burgdorferi cells harvested at 2 days post-inoculation were assayed by qRT-PCR using specific primers for BB0838 and flaB, which was used as a normalization control. Conditions and labels are as in panel A. Error bars indicate mean ± SD. (C) Representative phase micrographs of WT and B. burgdorferi BB0838 NT1 KD cells at day 3 post-inoculation. Conditions and labels are as in panels A and B. White arrowheads indicate visible disturbances in the spirochetal envelope. Size bars are 5 μm. (D) Percentage of cells with membrane disturbances/blebbing visible under phase contrast microscopy. Conditions and labels are as in panel C. Error bars indicate mean ± SD.

BB0838 depletion affects the localization of a major outer surface lipoprotein.

To test our hypothesis that BB0838 plays a crucial role in surface lipoprotein translocation, we used proteolytic shaving to assess any changes in surface exposure of lipoproteins in both mock control and BB0838 KD NT1 cells. Cultures inoculated with freshly cultured stationary phase cells at 1×10⁶ cells/ml final concentration were grown in selective BSK-II medium with or without 0.25 mM IPTG for 2 days. Harvested and washed intact cells were then treated in situ with proteinase K to remove surface-exposed proteins or peptides as described (Bunikis & Barbour, 1999, Zückert et al., 2004, Schulze & Zückert, 2006, Chen et al., 2011). As shown by Western immunoblotting, surface lipoprotein OspA was accessible to proteinase K and degraded in both mock and uninduced BB0838 KD cells, while it remained largely protected in the induced BB0838 KD cells. At the same time, the topology of the integral OM protein (OMP) and adhesin P66, as assessed by protease accessibility of a surface-exposed loop (Bunikis & Barbour, 1999, Bunikis et al., 1998, Bunikis et al., 1995, Bunikis et al., 1996, Coburn & Cugini, 2003, Kenedy et al., 2014, Skare et al., 1997) remained unaffected. Periplasmic flagellar subunit protein FlaB and cytosolic protein dCas9-Myc served as internal cell integrity controls (Fig. 4A).

Fig. 4. Assessment of BB0838-depended protein localization defects.

(A) Western immunoblots of total cellular proteins of B. burgdorferi cells harboring “empty” pJJW101 (mock) or pJJW101 with the two BB0838-specific sgRNAs (NT1 and NT2), under non-inducing (− IPTG) or inducing (+ IPTG; 0.25mM) conditions, before (– pK) or after (+ pK) in situ proteolysis with proteinase K. CRISPRi induction and expression of the C-terminally c-Myc-tagged BbdCas9 was verified using an anti-c-Myc antibody. Surface lipoprotein OspA served as the model surface lipoprotein. OMP P66 served as an OMP topology control, and periplasmic flagellar protein FlaB was used as a cell integrity and constitutively expressed loading control. (B) Western immunoblots of B. burgdorferi outer membrane vesicle (OM) and protoplasmic cylinder (PC) fractions from B. burgdorferi cells harboring “empty” pJJW101 (mock) or pJJW101 with the BB0838-specific NT1 sgRNA. IM lipoprotein OppAIV and conditionally expressed dCas9-c-Myc served as OMV purity controls. Note that the PC fraction also contains OM proteins such as OspA due to the partial separation of OMVs from protoplasmic cylinders by treatment of Borrelia cells with hypotonic citrate buffer (Skare et al., 1995).

To assess any defects in surface lipoprotein transport from the IM to the OM under BB0838-depleting conditions, we purified and analyzed the protein content of outer membrane vesicles (OMVs). B. burgdorferi cells cultured and harvested as described above on day 2 were shocked osmotically by resuspension in hypotonic citrate buffer, and the released OMVs were purified from remaining cell material (intact cells and protoplasmic cylinders) by ultracentrifugation on a discontinuous sucrose gradient as described (Skare et al., 1995, Radolf et al., 1995, Dowdell et al., 2017). Western immunoblots showed that OspA localized to OMVs under both mock control and BB0838-depleting conditions. OMV purity was assessed by the absence of IM lipoprotein OppAIV (Fig. 4B). Together, this set of experiments indicated that depletion of BB0838 primarily blocks the translocation of surface lipoproteins through the OM, but does neither significantly affect their transport through the periplasm, nor disturb the proper topology of integral OMPs such as P66.

BB0838/LptD_Bb depletion leads to a specific localization defect of surface lipoproteins.

To evaluate if the observed BB0838-dependent localization defect of OspA extends to the surface lipoproteome in general, we used quantitative multidimensional protein identification technology (MudPIT) mass spectrometry. B. burgdorferi BB0838 KD NT1 cells were cultured with and without IPTG induction, harvested, and subjected to proteolytic shaving as described above. Samples were enriched for membrane-associated proteins by Triton X-114 detergent extraction followed by a 2-step precipitation with acetone and trichloroacetic acid as described (Dowdell et al., 2017). Two biological replicates were submitted for MudPIT analysis and processed in 5 or 6 technical replicates.

BB0838/LptD_Bb peptides were detected in 9 of the 10 uninduced control samples but were absent from all induced knockdown samples, confirming an efficient CRISPRi-mediated knockdown on the protein level (Supplemental Table S4). We detected 44 of the previously localized 79 lipoproteins encoded by our B. burgdorferi clone B31-e2 parent strain (Dowdell et al., 2017, Murphy et al., 2023). Nine of the undetected lipoproteins (BB0456, BB0475, BB0735, BB0844, BBA33, BBA65, BBB08, BBP39, BBS41) were previously shown to be not transcribed in mid-exponential phase (Arnold et al., 2016), and the expression level of the remaining 26 lipoproteins was most likely below the MudPIT detection limit. 13 periplasmic IM, 5 periplasmic OM, and 10 surface lipoproteins were detected in at least two replicates of the uninduced control (BB0838 KD NT1 −IPTG)) and analyzed further (Supplemental Table S4). Based on the normalized spectral abundance factor (dNSAF) as a measure of protein abundance, we calculated the average lipoprotein surface exposure by dividing each protein’s dNSAF value in the untreated (−pK) samples by the dNSAF value in the corresponding protease-treated (+pK) samples, resulting in a −pK/+pK dNSAF ratio for each protein. Lipoproteins that localize to the periplasm, i.e., are protected from proteinase K, have close to identical dNSAF values in both the −pK and +pK samples and thus dNSAF ratios close to 1, whereas protease-accessible surface lipoproteins show a dNSAF ratio larger than 1 due to lower dNSAF values in the +pK sample (Dowdell et al., 2017). Conditional retention of a surface lipoprotein in the periplasm could therefore be followed by a drop of its −pK/+pK dNSAF ratio towards a value of 1.

As expected from our earlier study (Dowdell et al., 2017), the dNSAF ratios for surface and periplasmic lipoproteins showed a clear separation (Fig. 5, Supplemental Table S4 and Fig. S2). dNSAF ratios for all 28 lipoproteins in the control uninduced (BB0838 KD NT1 −IPTG) samples generally tracked the values obtained with the proteomically more complex B. burgdorferi B31 clone B31-A3 (Dowdell et al., 2017). The 18 periplasmic IM and OM lipoproteins had dNSAF ratios around 1 (range 0.38 to 2.160. Calculated ratios for surface lipoprotein ranged between 7.12 and 97.63, with three proteins having infinite (∞) ratios due to undetectable peptides in the protease-treated samples; BBD10 was already known to be a surface lipoprotein outlier with a low dNSAF ratio in B31-A3 (1.83; Dowdell et al., 2017; Supplemental Table S4). In the induced BB0838 knockdown (BB0838 KD NT1 +IPTG) sample, the +pK/−pK dNSAF ratios for the periplasmic IM and OM lipoproteins remained stably around 1 (range 0.20 to 1.83), indicating their continued periplasmic localization. Surface lipoprotein ratios, however, collapsed from their high values in the controls to a range of 0.00 to 3.98, indicating that they were now substantially retained within the periplasm. For comparison to the immunoblot data shown in Fig. 4A, OspA showed a drop in the +pK/−pK ratio from 45.83 to 1.81. Together, these data indicate that depletion of BB0838 leads to a protein localization defect in B. burgdorferi cells that is specific to surface lipoproteins.

Fig. 5. Quantitative proteomic analysis of BB0838-dependent lipoprotein localization defects.

Multidimensional protein identification technology (MudPIT) quantitative proteomics analysis of membrane-associated proteins in the recombinant B. burgdorferi BB0838 NT1 KD strain. The plot shows the mean normalized spectral abundance factor (dNSAF) counts after vs. before proteinase K treatment (+pK/−pK dNSAF ratio) for a subset of previously localized IM, periplasmic OM, and surface lipoproteins in B. burgdorferi B31 clone B31-A3 (Dowdell et al. 2017). Data are shown for uninduced (NT1–IPTG, clear squares) and induced (NT1 +IPTG, clear inverted triangles) conditions. Data are from duplicate biological replicates that were processed in 5–6 technical replicates. Centered dots mark data points of infinite (∞) dNSAF ratios that were capped at the highest calculated ratio for OspB. Note that the y-axis is split to help visualize the lower dNSAF ratios from low-expression proteins. See also supplementary data Table S4 and Fig. S2).

A comparison of dNSAF values in the unproteolyzed BB0838 control and knockdown samples showed generally moderate protein level changes of the 28 analyzed lipoproteins. The BB0838 KD NT1 −IPTG/+IPTG dNSAF ratios hovered around 1 (mean ±SD=0.88±0.38; range 0.29–1.42). BB0689/ChpAI was excluded as it was undetectable in the knockdown sample and thus had an infinite ratio. Intriguingly, the periplasmic OM lipoprotein BB0323 showed the highest measurable change, being about 3.4× more abundant in the BB0838 knockdown sample. BB0323 was previously shown to mediate OM stability and cell separation, with two proteolytically processed, interacting domains thought to anchor the OM to the peptidoglycan via a C-terminal LysM domain (Stewart et al., 2004, Kariu et al., 2013, Takacs et al., 2018). If confirmed by other experimental means, an upregulation of BB0323 under BB838 knockdown conditions could represent an attempt by B. burgdorferi to re-stabilize its OM.

Proteomic peptide analysis identifies a BB0838/LptD_Bb surface-exposed loop.

Since MudPIT provides quantitative and qualitative information on a peptide level, we investigated if a comparison of proteomic samples from untreated vs. in situ proteolyzed cells could provide some initial topology information that may begin to validate the in silico structural models of BB0838/LptD_Bb shown in Fig. 1. We, therefore, contrasted the list of BB0838 peptides identified in the B. burgdorferi B31-A3 (Dowdell et al., 2017) and uninduced control B31-e2 BB0838 KD NT1 samples before and after proteinase K treatment. Of the 48 sets of peptides with identical N-terminal sequences, only one peptide (NSGISAVQSPLEPQKPSSPYK, residues 860–880) was detected in all seven untreated samples, but absent from all proteinase K-treated samples (Supplemental Table S5 and Fig. S3). This peptide mapped to a predicted surface loop within BB0838’s β-barrel domain in both the i-Tasser and AlphaFold 2 models (Fig. 1). In the AlphaFold model, the predicted surface loop is situated opposite of the periplasmic β-jelly-roll domain. Validating these data, a peptide corresponding to a known surface-exposed P66 loop epitope (Bunikis et al., 1998, Bunikis et al., 1996, Kenedy et al., 2014) was efficiently removed by in situ proteinase K treatment as well (Supplemental Table S5).

Modeling of the B. burgdorferi Lpt pathway.

In gram-negative bacteria, the LPS-transporting Lpt pathway forms a continuous periplasmic bridge between the IM and OM. A dimer of the cytoplasmic ATPase LptB forms an ABC transporter-like IM complex with integral membrane proteins LptF and LptG. This IM complex interacts via LptF with LptC, whose periplasmic domain is anchored in the IM via a single N-terminal transmembrane domain. LptC’s C-terminus itself interacts with the N-terminus of periplasmic LptA, which connects to the N-terminal periplasmic domain of LptD at the OM. LptC, LptA, and the N-terminus of LptD create a hydrophobic “greasy slide” within a continuous β-jelly-roll structure that allows for the periplasmic transport of LPS’s fatty acid membrane anchors. Having shown that BB0838/LptD_Bb plays a role in surface lipoprotein translocation without an apparent LptE homolog, we wondered whether B. burgdorferi harbors Lpt homologs upstream of LptD that would complete the pathway.

A prior search of the NCBI Protein and CDD databases by Putker et al. (Putker et al., 2015) using COG database categories (Tatusov et al., 2000) identified B. burgdorferi homologs for LptA (BB0465), LptB (BB0466), LptF (BB0807) and LptG (BB0808), but no clear LptC homolog. In other bacterial systems, LptC is often encoded within the same locus as LptA and LptB in a lptCAB operon (Putker et al., 2015). Based on a likely synteny in B. burgdorferi, we speculated that ORF BB0464 upstream of lptA (BB0465) could encode for the B. burgdorferi LptC homolog. Like LptD_Bb, all five predicted Lpt homologs appeared essential based on a lack of Tn insertions (Lin et al., 2012, Lin et al., 2014).

To evaluate the five additional Lpt pathway candidates, we generated structural models using AlphaFold (Jumper & Hassabis, 2022) (Fig. 6A). Signal peptides predicted by SignalP 6.0 (Teufel et al., 2022) were excluded. pLDDT (predicted Local Difference Distance Test) values were used as model confidence metrics. The top model of BB0465 (LptA_Bb) without its signal I peptide showed the expected β-jellyroll fold (pLDDT 86.4). Compared to E. coli LptA, BB0465 is 45 amino acids larger, which in the model extended the fold by four β-strands to 12 total. Both N and C termini appeared disordered, with the potential for some short N-terminal α-helical structure. BB0464 (LptC) intriguingly was predicted to be a lipoprotein with a relatively short 14-amino acid N-terminal signal II peptide (SignalP6.0 probability = 0.99). Thus, LptC would be anchored in the IM via a different mechanism than gram-negative LptC, which uses an N-terminal α-helix membrane anchor (Wilson & Ruiz, 2022). The top model showed the expected overall β-jellyroll fold (pLDDT 90.3), with a short N-terminal α-helix following four disordered tether residues after the predicted N-terminal cysteine. The LptB (BB0466) model predictably assumed the structural fold of an ABC transporter ATPase (pLDDT 92.8). The LptF (BB0807) model showed six transmembrane helixes, with the third helix uniquely extending into the periplasm as a two-α-helix stalk before transitioning into a typical small periplasmic β-jelly roll domain (pLDDT 85.7). LptG (BB0808) modeled similarly to LptF, but lacked the periplasmic α-helical stalk (pLDDT 78.1). This analysis suggested that the B. burgdorferi Lpt pathway consisted of a multimeric complex of LptB₂FGCAD, i.e., was missing only a recognizable LptE homolog found in other diderm bacterial systems.

Fig. 6. Model of B. burgdorferi lipoprotein transport. (A) Composite AlphaFold Multimer model of the B. burgdorferi Lpt pathway.

Overlapping Alphafold and Alphafold Multimer models of LptD, LptD_NA, LptD_NAC, LptCFG, and LptFGB₂ were aligned in PyMol (see also text). The location of the subunit N and C termini are indicated in the same color as the subunits. The periplasmic span of the modeled complex was measured in PyMOL as the distance between the top of the membrane-spanning LptGF helices and the end of the periplasmic β-jelly roll domain of LptD. Note that the kink of the complex at the IM is likely an artifact of the model. (B) Proposed role of the B. burgdorferi Lpt and Lol pathways in lipoprotein sorting and secretion. B. burgdorferi homologs of known components are indicated by their TIGR ORF number. Canonical pathway components that are apparently missing in Borrelia are indicated by red strikethroughs. Model OMPs and surface, periplasmic OM, and IM lipoproteins are indicated. Our model proposes that after complete posttranslational modification by the lipoprotein processing machinery in the IM, (i) Borrelia surface lipoproteins such as OspA are recognized and extracted by the Lpt pathway and fast-tracked along a periplasmic bridge and through the OM to the bacterial surface; (ii) the Lol pathway is solely responsible for ensuring proper localization of periplasmic OM lipoproteins such as the abundant Lp6.6 or the BAM complex associated lipoproteins BamB and BamD, and (iii) IM lipoproteins such as OppAIV avoid interaction with both the Lpt or Lol pathway to remain in the IM. It is currently unknown if the proposed Lpt pathway is involved in the transport of other lipidated components of the B. burgdorferi envelope such as the glycolipids found in lipid rafts in both the IM and OM.

To predict protein-protein interactions and gain some initial dimensional insights, we used AlphaFold-Multimer (Evans et al., 2022) to generate a first model of the B. burgdorferi Lpt pathway. In addition to overall pLDDT, the predicted TM (pTM) and interface predicted TM (ipTM) scores were used as model confidence metrics. For multimer modeling purposes, we used only the N-terminal periplasmic domain of LptD (LptD_N). The overlapping multimer models of LptD_NA (pLDDT 85.6, pTM 0.79, piTM 0.79), LptD_NAC (76.9, 0.70, 0.62), LptAC (78.4, 0.76, 0.71), LptCFG (76.1, 0.69, 0.64) and LptFGB₂ (82.4, 0.78, 0.77) were assembled with the original LptD model by alignment in PyMol, resulting in the overall structural model of the B. burgdorferi Lpt pathway shown in Fig. 6A. Modeling of the entire LptB₂FGCAD_N complex resulted in the same overall structure (pLDDT 75.5, 0.8×piTM+0.2×pTM=0.545; not shown). Three key features of the predicted complex are notable: First, LptC, LptA, and the N terminus of LptD form an approximately 185 Å-long periplasmic bridge in a continual head-to-tail (N- to C-terminal) orientation. AlphaFold rejected the introduction of any additional LptA subunits into this bridge, although we were able to model LptA as a head-to-head (N-to-N terminus) homodimer and tail-to-head (C-to-N terminus) homotrimers (Supplemental Table S2). Second, the small α-helical domain predicted by AlphaFold at the N terminus of LptD_Bb moved from the monomer model orientation to now laterally interact with LptA. This suggests that this domain might function as a “clasp” to stabilize the LptA-LptD protein junction. Third, compared to LptG and other homologs, LptF contains an insertion that is predicted to protrude as an α-helical “stalk” domain into the periplasm. While this could be an artefact of the model, the DeepTMHMM transmembrane protein topology algorithm (Hallgren et al., 2022) equally predicted this domain to be “outside”, i.e., not part of an extended transmembrane domain.

DISCUSSION

Envelope homeostasis is a fundamental process that ensures the continued growth and persistence of bacterial cells in a variety of sometimes hostile environments. In the arthropod-borne Lyme disease spirochete B. burgdorferi, this process includes the mechanisms that properly segregate over 130 lipoproteins to be either retained in the IM, transported to the periplasmic side of the OM, or secreted to the bacterial surface. Our earlier lipoproteome compartmentalization analysis showed that two-thirds, i.e. more than 87 of these lipoproteins reach the borrelial surface (Dowdell et al., 2017). This suggested an efficient pathway with broad cargo specificity that acts downstream of the posttranslational lipoprotein modification machinery in the IM and guides lipoproteins through the periplasm and the OM.

This study provided additional insights into this pathway by characterizing the biological role of B. burgdorferi BB0838. Both experimental evidence and structural modeling indicate that BB0838, already shown to be an integral OM protein (Kenedy et al., 2016), is required for the terminal step of surface lipoprotein secretion at the B. burgdorferi OM. First, the depletion of BB0838 specifically prevented the translocation of surface lipoproteins through the OM. This specific surface lipoprotein mislocalization phenotype went hand-in-hand with a significant growth defect. Second, high confidence in silico protein modeling indicated that BB0838 is a structural homolog of the gram-negative OM protein LptD, which, analogous to the secretion of amphipathic lipopolysaccharide molecules by proteobacteria, would allow for the secretion of amphipathic lipoproteins in B. burgdorferi. The two top models consistently showed a periplasmic β-jelly roll linked to a relatively large, laterally open β-barrel pore domain that would allow for the passage of both hydrophobic and hydrophilic lipoprotein moieties and insertion into the surface leaflet of the OM lipid bilayer. However, the two models differed in their integration of BB0838’s C-terminal extension which contributes to its significantly larger size (1146 amino acids, 120 kDa) compared to a typical protobacterial LptD (784 amino acids, 87 kDa). The first model, built on a gram-negative LptDE complex by i-Tasser, inserted a LptE-like C-terminal plug domain from the outside into the β-barrel pore. The second model built by AlphaFold appeared less constrained and integrated the extension into a larger β-barrel with an estimated pore diameter of 3.7 nm. For comparison, B. burgdorferi outer membrane porin/adhesin P66 forms a pore with a deduced entry diameter of 1.9 nm and an internal constriction of 0.8 nm (Barcena-Uribarri et al., 2013). Thus, it is likely that BB0838 either assumes a structure that is a hybrid of the two models, or that the larger pore is obstructed by other means, e.g., by a yet-to-be-identified interacting protein or by transiting pathway cargo. Notably, both models are compatible with current topology data showing at least one protease-accessible surface loop as well as a protease-protected C terminus (Kenedy et al., 2016). Interestingly, LptD_Bb does not contain any Cys residues, indicating that, unlike proteobacterial LptD, it does not use disulfide bridges to coordinate conformational changes that lead to interaction with periplasmic binding partners (Ruiz et al., 2010, Chng et al., 2012).

The mechanistic underpinnings of the growth defect observed in the BB0838/LptD_Bb knockdown are likely multifactorial and need to be further explored. At this point, we cannot entirely exclude that BB0838/LptD_Bbonly plays an indirect role inflipping lipoproteins through the OM. An earlier finding that depletion of the B. burgdorferi BB0795/BamA_Bb led to a decrease of both OMPs and OM lipoproteins (Kenedy et al., 2016) is compatible with the expected dependence of BB0838/LptD_Bb insertion and folding on the BAM complex. Our finding that OMP P66 maintains its wild-type topology corroborates that BB0838/LptD_Bb acts downstream of BAM. Together with the maintained protection of periplasmic proteins from exogenous protease, it also indicates that depletion does not lead to a more generalized OM disturbance. High-passage B. burgdorferi strain B31 clones that lost most of the endogenous plasmids and thus most of their surface lipoproteome have occasionally shown growth defects on solid medium, and some clones seemingly compensated for the loss of some major Osp lipoproteins by deregulated expression of other Osps (Sadziene et al., 1993). Thus, it is conceivable that the OM’s surface leaflet is stabilized by the presence of surface lipoproteins. At the same time, the accumulation of abundant surface lipoproteins within the periplasm is likely to be toxic to the cell, as some E. coli OM lipoproteins are when mislocalized to the IM (Grabowicz & Silhavy, 2017). Whether this triggers an envelope stress response remains to be determined. The apparent compensatory upregulation of BB0323, a periplasmic OM lipoprotein involved in providing OM stability (Stewart et al., 2004), may provide a first hint towards the involved mechanisms.

A potentially more canonical use of the B. burgdorferi Lpt pathway would be the transport of glycolipids, as recently suggested for the predicted T. pallidum Lpt pathway with an intermediately sized LptD (Fig. 1B) (Hawley et al., 2021). B. burgdorferi membranes contain three major glycolipids, cholesteryl 6-O-acyl-β-D-galactopyranoside or cholesteryl 6-O-palmitoyl-β-D-galactopyranoside (ACGal), cholesteryl-β-D-galacto-pyranoside (CGal), and mono-α-galactosyl-diacylglycerol (MGalD), and the phospholipids phosphatidyl glycerol (PG) and phosphatidylcholine (PC) (Belisle et al., 1994, Toledo et al., 2018). The two cholesterol glycolipids, ACGal and CGal, appear more abundant in the OM, while MGalD is more abundant in the IM (Toledo et al., 2018). It is currently unclear whether these glycolipids are asymmetrically distributed like LPS in the gram-negative OM (Kamio & Nikaido, 1976), but proteomic analyses of OM and IM lipid rafts formed by the two cholesterol glycolipids showed enrichment of cytoplasmic, periplasmic and surface proteins (Toledo et al., 2018, Toledo et al., 2015). The OM glycolipid rafts contained a subset of surface lipoproteins including OspA and OspB, as well as the OMP P66 (Toledo et al., 2014). It is equally unknown if these glycolipids are essential for B. burgdorferi. Cells are able to recover from the partial chemical removal of cholesterol glycolipids from membranes via brief treatment with methyl-β-cyclodextrin, with no changes in OspA, OspB, and P66 levels (LaRocca et al., 2010). If the B. burgdorferi Lpt pathway were involved in glycolipid transport, depletion of BB0838/LptD_Bb would lead to a misdistribution of glycolipids in the two OM leaflets. This, in turn, would likely disrupt the overall charge balance of the bilayer leaflets, with the potential of broadly disturbing the OMP topology (Bogdanov et al., 2014). As stated above, our assays did not detect such pleiotropic effects on OM protein topology beyond surface lipoproteins. Thus, while we cannot currently exclude an involvement of the predicted B. burgdorferi Lpt pathway in glycolipid transport, our current dataset leads us to favor a more specific and direct role in surface lipoprotein secretion.

Multimer modeling of the predicted B. burgdorferi Lpt pathway complex suggests a continuous periplasmic bridge formed by head-to-tail monomers of LptC, LptA, and the N-terminus of LptD that is estimated to be about 18 nm long (Fig. 6A). This distance corresponds to the average distance between the B. burgdorferi IM and OM as determined by cryo-ET (Charon et al., 2009). Yet, periplasmic widths vary depending on the presence of flagella (Kudryashev et al., 2009), and the longitudinal dimensions of the periplasm-spanning B. burgdorferi Tol-like BesABC complex (Bunikis et al., 2008, Greene et al., 2013) and the BAM complex-associated, IM-anchored TamB (Iqbal et al., 2016) are modeled to be slightly larger. Therefore, stoichiometric studies will have to determine whether additional LptA_Bb subunits, each extending the periplasmic bridge by about 6 nm, are needed to form functional multiprotein Lpt complexes.

The observed accumulation of surface lipoproteins in the periplasmic leaflet of the OM under BB0838/LptD_Bb-depleting conditions is reminiscent of the phenotypes we obtained in our previous lipoprotein localization studies. There, mutations in the N-terminal disordered tether peptides of various Borrelia surface lipoproteins similarly led to their mislocalization to the periplasmic leaflet of the OM (Schulze & Zückert, 2006, Schulze et al., 2010, Kumru et al., 2010, Kumru et al., 2011). The ability to redirect these tether mutants to the surface by destabilizing their tertiary structures (Schulze et al., 2010, Chen et al., 2011) suggested that these mutants were equivalent to secretion intermediates in the periplasmic leaflet of the OM. It remains to be determined how surface lipoproteins stalled at the periplasmic side of the OM would be able to re-engage with the secretion machinery. Recent studies in E. coli have posited that one of the roles of the Lol pathway is to alleviate toxic mislocalization of periplasmic lipoproteins(Grabowicz & Silhavy, 2017). It is therefore possible that at least some of the lipoprotein tether mutants are rejected by the Lpt pathway and rerouted through the Lol pathway.

Our identification of a surface-exposed loop within the BB0838/LptD_Bb β-barrel domain is in line with data obtained by Kenedy et al. (2016). Their studies showed the release of a C-terminal BB0838 fragment migrating at about 34 kDa upon in situ proteolysis with trypsin, while proteinase K treatment produced a slightly smaller C-terminal fragment migrating at about 32 kDa. These fragments could be explained by the trypsin accessibility of K819 at the beginning of the identified surface loop peptide and the removal of the entire tryptic peptide by proteinase K, which would yield 34.0 and 31.8 kDa C-terminal fragments, respectively. The fact that BB0838/LptD_Bb is an essential, partially surface-exposed OMP with variants in other pathogenic spirochetes makes it an attractive therapeutic target. LptD proteins have been successfully evaluated as potential vaccinogens for Neisseria gonorrhoeae and Vibrio parahaemolyticus (Zielke et al., 2016, Zha et al., 2016). Moreover, two novel antimicrobials targeting Pseudomonas aeruginosa and E. coli LptD have been identified. The β-hairpin-like peptidomimetic L27–11 targeting P. aeruginosa LptD showed potent antimicrobial activity in a mouse septicemia model (Srinivas et al., 2010) while the JB-95 β-hairpin macrocyclic peptide bound to E. coli LptD and BamA and showed potent antimicrobial activity against a panel of clinical strains (Urfer et al., 2016). The β-hairpin structure of these peptidomimetics might interact with the N terminus β-jellyroll of LptD (Andolina et al., 2018).

In summary, we used CRISPRi-mediated protein depletion to establish BB0838 as an LptD homolog that impacts the terminal step of lipoprotein secretion to the B. burgdorferi surface. This suggests that the substrate specificity of Lpt pathways in diderm bacteria extends beyond LPS. Supporting this is the notion that LptD orthologs were also found in T. pallidum and other diderm LPS-deficient bacteria such as Novosphingobium aromaticivorans, Deinococcus radiodurans, Thermus thermophilus, and Thermotoga maritima (Putker et al., 2015). Yet, as in other diderm bacterial systems, BB0838/LptD_Bb would be involved in the display of lipidated immunodominant virulence factors and thus be part of an elemental envelope homeostasis pathway that is also indispensable for microbial pathogenesis. Our current working model proposes an emerging dichotomy between Lol-mediated periplasmic lipoprotein sorting and Lpt-mediated surface lipoprotein secretion (Fig. 6B). In this model, sorting of the diverse B. burgdorferi lipoproteome to its three possible destinations occurs at the IM: (i) surface lipoproteins such as OspA are fast-tracked after being recognized by LptBFG and pushed along the LptCAD periplasmic bridge and through the LptD lumen and lateral opening to the bacterial surface; (ii) periplasmic OM lipoproteins such as Lp6.6 or the essential BAM associated lipoproteins BamB and BamD avoid interaction with the LptBFG complex to be recognized by LolCDE and transported and inserted into the periplasmic leaflet of the OM by LolA; and (iii) IM lipoproteins such as the oligopeptide-binding OppAIV would avoid either pathway to remain in the IM (Fig. 6B). Proper insertion and topology of LptD as the final component in the Lpt complex would be extrinsically linked to the Sec- and Lol-dependent assembly of the BAM complex in the OM. Yet, a requirement for LptC lipidation may represent an additional early and intrinsic quality control checkpoint at the IM that prevents assembly of a functional Lpt complex if its own maturation – and that of its lipoprotein cargo – is disturbed. At the same time, a lipidated LptC would lack the recently discovered rate-modulating function of the anchoring transmembrane α-helix of gram-negative LptC (Wilson & Ruiz, 2022). Our future studies will test these hypotheses and determine structure-function relationships of the B. burgdorferi Lpt and Lol pathway components, define their individual cargo specificities and potential interactions, and reassess previously identified Borrelia lipoprotein sorting determinants.

EXPERIMENTAL PROCEDURES

Bacterial strains and culture conditions.

E. coli NEB 5-alpha cells (New England Biolabs) were used for plasmid construction and propagation, and grown in LB broth or on LB agar supplemented with selective antibiotic (40 μg/ml kanamycin; Sigma) as indicated. Our stock of Borrelia burgdorferi B31-e2, a non-infectious clonal derivative of type strain B31 (Babb et al., 2004) encodes for 79 fully localized lipoproteins (Dowdell et al.) due to its reduced plasmid content (cp26, cp32–1, cp32–3, cp32–4, lp17, and lp54; i.e., lacking lp38 in comparison to the original B31-e2 clone) (Murphy et al., 2023). B. burgdorferi was cultured in liquid or solid BSK-II medium at 34°C under a 5% CO₂ atmosphere (Barbour, 1984, Zückert, 2007). The parent and all recombinant daughter strains were isogenic as determined by multiplex PCR (Bunikis et al., 2011)(data not shown). Selective BSK-II medium contained 200 μg/ml kanamycin. Protein expression from hybrid lac promoters was induced by the addition of isopropyl-β-D-thiogalactopyranoside (IPTG 0.25 mM; Sigma) where indicated.

Recombinant plasmid construction and sgRNA design for CRISPRi.

pJJW101 (Murphy et al., 2023) is a derivative of the E. coli-B. burgdorferi shuttle vector pJSB142 and the Mobile-CRISPRi plasmid pTn7C107 (Blevins et al., 2007, Peters et al., 2019). The plasmid encodes for an IPTG-inducible, codon-optimized non-toxic variant of SpydCas9 (BbdCas9) as well as an IPTG-inducible single guide RNA (sgRNA) cassette that allows for easy insertion of sgRNA spacer sequences. BB0838-specific sgRNA spacer sequences were designed using the web-based CRISPy-web interface (Blin et al., 2016) on an artificial genomic contig of B. burgdorferi B31 as described (Murphy et al., 2023). Four target sgRNAs, two complementary to the non-template (NT) strand, and two complementary to the template (T) strand were selected (Table 2), synthesized as complementary pairs of single-stranded oligonucleotides with BsaI site overhangs (Table 1), annealed, and directionally ligated into the cut BsaI site array within the sgRNA cassette of pJJW101 as described (Murphy et al., 2023).

Table 2.

sgRNA spacer sequences used in this study.

sgRNA	Start†	End†	Strand	ORF	PAM	Sequence (5’ to 3’)	Core %GCPAM	Mismatched Core Off-Targets
								0 bp	1 bp	2 bp
T1	T1158	181	T	BB0838	AGG	ACTTGAGCTTTCAACCGATG	46.2	0	0	0
T2	185	208	T	BB0838	AGG	AAAAAAATGGGCCTTAAAAG	46.2	0	0	0
NT1	86	109	NT	BB0838	TGG	TGTTTGATAACATGTGCATT	30.8	0	0	0
NT2	85	108	NT	BB0838	GGG	TGAGACCCTGAAAGTGATGC	46.2	0	0	0

^†Coordinates of the first and last nucleotide of the spacer sequence relative to the 5’ end of the ORF. See text for explanation of other parameters.

Growth curve assays and phase contrast microscopy.

B. burgdorferi cells were grown from frozen stocks in a selective liquid BSK-II liquid medium containing 200 μg/ml kanamycin to mid-exponential phase (about 2 days). Bacteria were then seeded at final concentrations of 1×10⁵ organisms/ml into liquid medium containing kanamycin (200 μg/ml) without or with IPTG (final concentration of 0.25 mM) and incubated at 34°C for 3 days. Spirochete numbers were determined by counting in a Petroff-Hausser counting chamber under phase contrast microscopy (Nikon Eclipse E400, Nikon Plan 40×/0.65 Ph2 DL objective). For phenotypic analysis, cells were collected by centrifugation at 5000 × g for 10 min at room temperature, washed once with Dulbecco’s phosphate-buffered saline (dPBS, pH 7.4), spotted on 2% (wt/vol) agarose/dPBS pads as described (Jutras et al., 2016), covered with a No. 1.5 coverslip, and imaged under phase contrast using a Nikon Eclipse E600 microscope (Nikon Plan Fluor 100×/1.30 Oil Ph3 DLL objective) connected to an INFINITY 3 digital camera (Teledyne Lumenera). Images were captured using Infinity Capture version 6.3.2 software (Teledyne Lumenera) and processed using Adobe Photoshop and Illustrator 2021.

Total RNA extraction and qRT-PCR analysis.

B. burgdorferi cells grown as described above were harvested by centrifugation at 2844 × g for 20 min at room temperature, washed once by resuspension in sterile room-temperature PBS containing 5 mM MgCl₂ (PBS+Mg), and repelleted. Total RNA was extracted with TRIzol (Invitrogen) according to the manufacturer’s instructions. To remove residual DNA, the RNA samples were treated with 0.1 U DNase I (Invitrogen) for 1h at 37°C, followed by phenol-chloroform extraction (Ambion) and ethanol precipitation overnight at −20°C (Sambrook & Russell, 2001). The purified RNA was quantified using a Nanodrop ND1000 spectrophotometer. RNA samples were then subjected to qRT-PCR on an ABI Prism 7500 system (Applied Biosystems) using oligonucleotide primers listed in Table 1 and the Luna Universal One-Step RT-qPCR kit (NEB). BB0838 transcript levels were validated and normalized against flaB mRNA, and fold changes were calculated using the comparative CT (2^-ΔΔCT) method for quantification.

SDS-PAGE and immunoblotting.

B. burgdorferi cells were harvested by centrifugation and washed with PBS+Mg as described above. Bacterial pellets were solubilized in 1× sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) sample buffer containing 50 mM dithiothreitol (DTT), boiled for 5 min, and stored at −20 °C. Whole-cell proteins were separated on a 12% polyacrylamide SDS-PAGE gel (Sambrook & Russell, 2001) and visualized by Coomassie blue staining (Fisher BioReagents^™ EZ-Run^™ Protein Gel Staining Solution, #BP3620-1). For immunoblots, proteins were electrophoretically transferred to nitrocellulose membranes (Millipore) using a Transblot semi-dry transfer cell (Bio-Rad). The membranes were rinsed in Tris-buffered saline (TBS) (20 mM Tris, 500 mM NaCl, pH 8.0). TBS with 0.05% Tween 20 (TBST) containing 5% dry milk was used for membrane blocking and subsequent incubation with primary and secondary antibodies; TBST alone was used for the intervening washes (Sambrook & Russell, 2001). Antibodies used were anti-OspA mouse monoclonal antibody H5332 (1:100 dilution) (Barbour et al., 1983), anti-c-Myc mouse monoclonal antibody (1:1000, Thermo Fisher, 9E10), anti-P66 rabbit polyclonal antibody (1:100) (Bunikis et al., 1995), and anti-OppAIV rabbit polyclonal antibody (1:100) (Bono et al., 1998). Secondary antibodies were Alkaline Phosphatase (AP)-conjugated anti-mouse (1:30,000, A3562, Sigma), and anti-rabbit antibodies (1:30,000, A3687, Sigma). Blots were developed using AP substrate CDP-Star (BioRad) for chemiluminescent detection, and signals were detected and captured using a Fujifilm LAS-4000 CCD imaging system.

Surface proteolysis of intact B. burgdorferi spirochetes.

Proteolytic shaving of intact spirochetes with proteinase K was performed as described (Bunikis & Barbour, 1999, Zückert et al., 2004, Schulze & Zückert, 2006). Briefly, B. burgdorferi cells harvested and washed as described above were resuspended in PBS+Mg without or with proteinase K (Invitrogen, 200 μg/ml final concentration). Proteinase K-containing and control samples were incubated for 1 h at room temperature, and reactions were stopped after 1 h by adding phenylmethylsulfonyl fluoride (PMSF) to a final concentration of 5 mM. Subsequently, cells were pelleted by centrifugation, resuspended in 1× SDS-PAGE sample buffer, boiled for 5 min, and stored at −20°C for further analysis.

Membrane fractionations.

B. burgdorferi outer membrane vesicles (OMVs) and protoplasmic cylinders (PCs) were isolated as described (Skare et al., 1995, Lenhart & Akins, 2010). Briefly, cells were harvested at room temperature by centrifugation for 20 min at 2844 × g, washed in PBS with 0.1 % bovine serum albumin (BSA), and repelleted. The pellet was then resuspended in 38 ml ice-cold 25 mM citrate buffer (pH 3.2) with 0.1% BSA and incubated at room temperature for 2 hours with agitation and a 1-min vortexing step every 30 min. Next, samples were pelleted by centrifugation at 20,000 × g for 30 min and resuspended in 6 ml ice-cold 25 mM citrate buffer (pH 3.2) with 0.1% BSA. This sample was then layered on a discontinuous 56% (wt/wt in 4 ml), 42% (wt/wt in 15.5 ml), and 25% (wt/wt in 12.5ml) sucrose gradient in 25 mM citrate buffer in 38 ml Ultraclear tubes (Beckman-Coulter, 344058). The gradient was centrifuged at 100,000 × g for 18 h at 4°C (Beckman-Coulter XPN-80 Ultracentrifuge, SW32 Ti swinging-bucket rotor) to separate the OMV (upper band) and PC (lower band) fractions. The PC fraction was collected, diluted with PBS, repelleted at 20,000 × g for 20 min, and resuspended in 1 ml PBS with 1mM PMSF for storage at −20°C. The OMV fraction was collected, diluted with PBS, repelleted at 100,000 × g for 4 h at 4°C, and resuspended in 100 μl PBS for storage at −20°C.

Quantitative mass spectrometry (MudPIT).

B. burgdorferi cells were subjected to surface proteolysis with proteinase K as described above and harvested. Membrane-associated proteins were then enriched by overnight extraction with Triton X-114, as described (Carroll, 2010, Dowdell et al., 2017). The washed detergent extracts were then precipitated at −20 °C overnight in final 80 % (vol/vol) acetone, resuspended in 0.1 M Tris-HCl (pH 8.5), and precipitated again overnight using 20% trichloroacetic acid (TCA). The addition of acetone precipitation in the protocol was necessary to effectively remove detergent prior to analysis by MudPIT (Dowdell et al., 2017). Desiccated frozen protein samples from three biological replicates were then submitted for MudPIT analysis (Proteomics Center, Stowers Institute for Medical Research, Kansas City, MO). Resuspended protein samples were digested with endoproteinases Lys-C (Roche) and trypsin (Promega) at 0.1 μg/μl final concentration each. The proteinase-digested samples were then analyzed by MudPIT on an LTQ linear ion trap (Thermo Scientific) coupled to a Quaternary Agilent 1100 series high-performance liquid chromatograph (HPLC) (Florens & Washburn, 2006). Protein content in mock control versus proteinase K-treated whole-cell protein preparations was analyzed by comparison of the average distributed normalized spectral abundance factor (dNSAF) for each unique protein, which correlates directly with the relative abundance of a particular protein in the sample (Zhang et al., 2010). A mean dNSAF ratio of untreated control to protease-treated sample (dNSAF −pK/+pK ratio) was calculated for each protein. All MudPIT raw datasets have been deposited in the MassIVE Repository (ftp://MSV000089990@massive.ucsd.edu, with password: Zueckert_1732) and will also be available after publication from the Stowers Original Data Repository at https://www.stowers.org/research/publications/libpb-1732.

Bioinformatics and molecular modeling.

NCBI BLASTP (Johnson et al., 2008) was used to identify protein homologs among pathogenic spirochetes. The original structural model for BB0838 was generated by homology modeling using i-Tasser and the Shigella flexneri LPS-assembly protein LptD (PDB accession number 4Q35) (Qiao et al., 2014) as a template. Pairwise amino acid sequence alignment was performed using ClustalW2 (Madeira et al., 2022). Monomer models were generated using AlphaFold-monomer (Jumper & Hassabis, 2022) and assessed based on pLDDT scores and visual inspection. Protein complex models were generated using AlphaFold-Multimer (Evans et al., 2022) and assessed based on the weighted sum of pTM and piTM values (the AlphaFold-Multimer ranking metric), pLDDT scores, and visual inspection. Model PDB files were visualized using PyMOL (The PyMOL Molecular Graphics System, Version 2.0 Schrödinger, LLC).