The glycerol-3-phosphate dehydrogenases GpsA and GlpD constitute the oxidoreductive metabolic linchpin for Lyme disease spirochete host infectivity and persistence in the tick

We have identified GpsA, a predicted glycerol-3-phosphate dehydrogenase, as a virulence factor in the Lyme disease spirochete Borrelia (Borreliella) burgdorferi: GpsA is essential for murine infection and crucial for persistence of the spirochete in the tick. B. burgdorferi has a limited biosynthetic and metabolic capacity; the linchpin connecting central carbohydrate and lipid metabolism is at the interconversion of glycerol-3-phosphate and dihydroxyacetone phosphate, catalyzed by GpsA and another glycerol-3-phosphate dehydrogenase, GlpD. Using a broad metabolomics approach, we found that GpsA serves as a dominant regulator of NADH and glycerol-3-phosphate levels in vitro, metabolic intermediates that reflect the cellular redox potential and serve as a precursor for lipid and lipoprotein biosynthesis, respectively. Additionally, GpsA was required for survival under nutrient stress, regulated overall reductase activity and controlled B. burgdorferi morphology in vitro. Furthermore, during in vitro nutrient stress, both glycerol and N-acetylglucosamine were bactericidal to B. burgdorferi in a GlpD-dependent manner. This study is also the first to identify a suppressor mutation in B. burgdorferi: a glpD deletion restored the wild-type phenotype to the pleiotropic gpsA mutant, including murine infectivity by needle inoculation at high doses, survival under nutrient stress, morphological changes and the metabolic imbalance of NADH and glycerol-3-phosphate. These results illustrate how basic metabolic functions that are dispensable for in vitro growth can be essential for in vivo infectivity of B. burgdorferi and may serve as attractive therapeutic targets.


Introduction
Lyme disease is the most prevalent arthropod-borne infection in North America with an estimated 476,000 cases annually in the United States [1]. Borrelia (Borreliella) burgdorferi, the enzootic spirochete that causes Lyme disease [2][3][4], is maintained in nature by cycling between Ixodes ticks and a vertebrate host reservoir, primarily white-footed mice; the bacterium is neither free-living nor transovarially transmitted by the female ticks to oocytes. The reduced genome of B. burgdorferi reflects the constraints of host dependence where numerous biosynthetic and energy-producing metabolic pathways have been lost, including amino acid synthesis, nucleotide synthesis, fatty acid synthesis, the citric acid cycle, and the electron transport chain [5,6]. Thus, B. burgdorferi has evolved into an unabashed scavenger of amino acids, nucleosides, peptides and various carbon sources including glucose, N-acetylglucosamine (GlcNAc) and glycerol. The metabolic capacity retained by B. burgdorferi to flourish in the disparate environments of the vertebrate host and tick vector is important to understand as these strategies, in the absence of any identified toxins or secreted effectors, determine survival of the spirochete and thus the pathogenesis of Lyme disease.
Available carbon sources are a dynamic determinant of B. burgdorferi persistence and transit through the enzootic cycle [7,8]. Glucose is likely the preferred carbon source of B. burgdorferi in the vertebrate host and initially during tick feeding, but other carbohydrates can support growth and have a role during the enzootic cycle [9][10][11]. In particular, glycerol becomes important for B. burgdorferi persistence in the tick as glucose levels decrease when the blood meal is consumed by the tick and its microbiome. This importance is exemplified by the finding that B. burgdorferi mutants unable to utilize glycerol for glycolysis are significantly compromised for persistence in the tick, yet remain infectious in the vertebrate host [12][13][14]. Glycerol also supports B. burgdorferi growth in vitro, particularly at 23˚C, the temperature often used to mimic tick-like conditions [12,13,15]. Glycerol enters the cell through the glycerol uptake facilitator GlpF, is converted to glycerol 3-phosphate (G3P) by the glycerol kinase GlpK and either is shuttled to glycolysis via conversion to dihydroxyacetone phosphate (DHAP) by the glycerol-3-phosphate dehydrogenase (G3PDH) GlpD or serves as the threecarbon backbone for lipid and lipoprotein biosynthesis (Fig 1) [5][6][7]. The glp operon, consisting of the glpF, glpK and glpD genes, along with bb0242, is controlled by a diverse repertoire of regulatory mechanisms [16]. Gene expression is induced during nutrient stress by the stringent response mediated by Rel Bbu , (p)ppGpp and the effector protein DksA [15,17,18], by the response regulator Rrp1, which produces c-di-GMP [13,19], at 23˚C [20], by glycerol [13,15], and in the tick [12]. RpoS and BadR both repress levels of glp operon transcripts, although BadR likely exerts its influence through RpoS [21,22]. Additionally, the c-di-GMP effector protein PlzA can either positively or negatively affect glp operon expression depending on its c-di-GMP-binding state [23]. These regulatory pathways targeting the glp operon represent the best understood strategies that B. burgdorferi uses to persist in the tick and illustrate the central role of glycerol regulation in this phase of the enzootic cycle. Other genes involved in carbohydrate utilization, such as malQ and chbC, are not required for B. burgdorferi in either its vertebrate host or tick vector [10,11].
GlpD shuttles G3P towards glycolysis via DHAP to provide energy while the reverse reaction, reducing DHAP to G3P mediated by another predicted G3PDH, GpsA, connects carbohydrates entering glycolysis to lipid and lipoprotein biosynthesis (Fig 1) [5][6][7]. In the absence of glycerol, GpsA, using carbohydrates from glycolysis, serves as the only known pathway to provide G3P for lipid and lipoprotein biosynthesis as B. burgdorferi cannot import this phosphorylated sugar alcohol or salvage glycerolipids [24]. B. burgdorferi must coordinate the activities of GlpD and GpsA to efficiently balance carbon sources and redox cofactors, such as NADH, and respond to the physiological requirements of the spirochete. Thus, the GlpD/ GpsA metabolic node regulates carbon flow between lipid biosynthesis and glycolysis in response to the phase of the enzootic cycle as indicated by the available carbon sources.
In this study we examine the in vivo role of the GpsA/GlpD metabolic node in murine infectivity and tick persistence in an animal model of Lyme disease. Additionally, we molecularly dissect the contributions and interactions in vitro of these two G3PDHs in spirochete survival and morphology as well as broad metabolic regulation using a metabolomics approach.
production is the bidirectional oxidoreductase node mediated by the opposing actions of a pair of G3PDHs GlpD and GpsA (Fig 1) [5][6][7]. Based on sequence homology, GlpD is thought to oxidize G3P to DHAP and concomitantly reduce either NAD + or FAD. GpsA is predicted to catalyze the reverse reaction to reduce DHAP to G3P using the reducing power of NADH or NADPH. Pappas et al., 2011 [12] and He et al., 2011 [13] have shown that GlpD and the glp operon, respectively, are important for B. burgdorferi growth on glycerol and for persistence in the tick, but dispensable for murine infectivity. The function of GpsA either in vitro or in vivo has not previously been evaluated in B. burgdorferi.

B. burgdorferi gpsA complements the growth defect of an Escherichia coli gpsA mutant
To genetically confirm the predicted function of B. burgdorferi GpsA, we heterologously complemented the growth phenotype of an E. coli gpsA mutant. The B. burgdorferi gpsA (bb0368) gene was cloned into the E. coli (Ec) isopropyl β-D-1-thiogalactopyranoside (ITPG)-inducible expression vector pUC18. The E. coli gpsA mutant BB20-14 [25], which is a G3P auxotroph and cannot grow on glucose as the sole carbon source, was made competent and transformed with either the empty vector, pUC18, or the vector carrying the B. burgdorferi gpsA gene, pUC18-gpsA Bb . The strains Ec ΔgpsA (BB20-14), Ec ΔgpsA+pUC18 (empty vector) and Ec ΔgpsA+pUC18-gpsA Bb (expressing B. burgdorferi gpsA) were inoculated at 5 × 10 7 cells ml -1 and grown in M9 minimal salts media (Fig 2A) or M9 minimal salts media with glucose ( Fig  2B). Cultures were grown at 37˚C for 9 h with the OD 600 taken every 17 min. In M9 media lacking a carbon source, all three strains grew at approximately the same rate and failed to grow exponentially or reach high cell density (Fig 2A). When glucose was added as the sole carbon source to the M9 media, only the strain expressing B. burgdorferi gpsA (Ec ΔgpsA+-pUC18-gpsA Bb ) had sustained exponential growth and grew to high cell density (>5 × 10 8 cells ml -1 ) ( Fig 2B). These data demonstrate that B. burgdorferi gpsA can heterologously complement the growth phenotype of an E. coli gpsA mutant on glucose as a sole carbon source, providing experimental support for the annotated function of gpsA in B. burgdorferi.

ΔgpsA, ΔglpD and ΔgpsA/ΔglpD mutant strains have wild-type growth in vitro
To investigate the biological role of the metabolic node that interconverts G3P and DHAP regulated by GpsA and GlpD, the genes gpsA or glpD were mutated alone or together by allelic exchange with antibiotic resistance cassettes to generate ΔgpsA, ΔglpD and the double mutant ΔgpsA/ΔglpD strains. All strains were complemented in cis to yield the full array of complemented strains: gpsA complemented (gpsA + ), glpD complemented (glpD + ), gpsA complemented in the ΔgpsA/ΔglpD background (gpsA + /ΔglpD), glpD complemented in the ΔgpsA/ ΔglpD background (ΔgpsA/glpD + ) and both gpsA and glpD complemented in the ΔgpsA/ΔglpD background (gpsA + /glpD + ) (S1 Fig). Immunoblot analyses of the mutants using antibodies against GlpD and GpsA were used to confirm the deletion and complementation of the respective genes (S1H and S1I Fig). To genetically evaluate the role of gpsA and glpD during in vitro growth, cultures were grown in Barbour-Stoenner-Kelly II media containing 6% rabbit serum (BSK + RS) for eight days and spirochetes were enumerated using a Petroff-Hauser cell counter. No statistical difference in growth was observed between wild-type, ΔgpsA, ΔglpD or the double mutant ΔgpsA/ΔglpD strains after day one ( Fig 3A). These data suggest that neither gpsA nor glpD play a significant role during B. burgdorferi growth in nutrient-rich media in vitro.

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick

GpsA is required for survival during nutrient stress
We hypothesized that GpsA functions in survival during nutrient stress, an in vitro condition used to mimic the tick midgut environment between bloodmeals. To test this, B. burgdorferi strains were grown to 7-9 × 10 7 cells ml -1 in normal growth medium, collected, resuspended and incubated in RPMI medium (which contains 2 mg ml -1 glucose) for 24 h, as previously described [18]. Cells were plated in semi-solid BSK, and colonies allowed to grow for approximately two weeks before enumeration. B. burgdorferi survival is represented as the percentage survival after 24 h of nutrient stress compared to cells plated before nutrient stress (0 h). ΔgpsA mutants were almost completely compromised for survival during nutrient stress, while the The E. coli gpsA null mutant, strain BB20-14 (Ec ΔgpsA, white circles), E. coli gpsA null mutant with the inducible pUC18 expression vector (Ec ΔgpsA pUC18, gray circles) or E. coli gpsA null mutant with pUC18 carrying the B. burgdorferi gpsA gene (Ec ΔgpsA pUC18-gpsA Bb ) were grown in M9 minimal media containing 0.1 mM IPTG either without (A) or with 1% glucose (B) at 37˚C. Cell density measurements (OD 600 ) were taken every 17 min. Data are the average from two separate cultures for each strain and error bars represent SEM; the experiment shown is representative of three independent biological replicates. The Ec ΔgpsA pUC18-gpsA Bb strain had significantly higher (p < 0.05) OD 600 values compared to the other two strains from 4 h to 8 h of growth, as determined by one-way ANOVA with a Tukey's post-hoc test. https://doi.org/10.1371/journal.ppat.1010385.g002

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick ΔglpD mutant was not significantly affected compared to wild type ( Fig 3B). This phenotype is restored in the gpsA + and, surprisingly, in the ΔgpsA/ΔglpD double mutant (Fig 3B). Complementing the double mutant with glpD, thus essentially constructing an independent gpsA mutant, also resulted in a B. burgdorferi strain unable to survive nutrient stress, similar to the ΔgpsA mutant (ΔgpsA/glpD + Fig 3B). gpsA complementation of the double mutant and complete complementation of the double mutant significantly increased survival compared to ΔgpsA, but did not fully restore survival to wild-type levels (gpsA + /ΔglpD and gpsA + /glpD + , Fig  3B). These data suggest that gpsA plays a crucial role in survival during nutrient stress in culture while glpD is dispensable. Furthermore, our results suggest we have identified the first . burgdorferi strains were inoculated at 1 × 10 5 cells ml -1 in BSK + RS and grown at 35˚C. Cells were enumerated every 24 h and cell density plotted over eight days. Data are the means from three independent biological replicates and error bars represent SEM. No significant difference in cell density was determined except at day 1 between the wild-type (WT) and ΔgpsA/ΔglpD strains: � indicates p = 0.046 between the mean cell density of WT and ΔgpsA/ΔglpD strains at day 1 as determined by one-way ANOVA with a Tukey's post-hoc test. Strains were grown in BSK + RS at 35˚C (to a cell density of 5-9 × 10 7 cells ml -1 ) before shifting to RPMI alone (B), or RPMI containing 0.4% glycerol (C) or 0.4% N-acetylglucosamine (GlcNAc) (D) and incubated at 35˚C for 24 h. Cultures were plated in semi-solid BSK media and allowed to grow at 35˚C in 5% CO 2 before colony enumeration. Data are presented as the percent survival of each strain before (0 h) shifting to the nutrient stress media. Data are the mean of at least three biological replicates and errors bars represent the SEM. Significance determined by one-way ANOVA with a Tukey's post-hoc test. ( � p < 0.0001; �� p < 0.002; # p < 0.01; † p = 0.0075). https://doi.org/10.1371/journal.ppat.1010385.g003

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick suppressor mutation in B. burgdorferi as deleting the glpD gene in a ΔgpsA mutant background restored viability under nutrient stress.
Because the link from glycolysis to G3P metabolism is severed in the ΔgpsA mutant, we determined if glycerol could restore survival of the gpsA mutant in nutrient stress medium (RPMI + glycerol). Strains were grown and treated as in Fig 3B except that 0.4% glycerol was added to the RPMI medium. Unexpectedly, glycerol in the nutrient stress medium was cytotoxic to wild-type B. burgdorferi (Fig 3C). In fact, glycerol was toxic to all strains except the ΔglpD mutants (Fig 3C), suggesting that metabolism of G3P by GlpD is necessary for the bactericidal activity of glycerol in this restrictive medium. Next, we examined if GlcNAc, a carbohydrate necessary for in vitro growth [9,10,26], could rescue the gpsA survival defect in RPMI medium. The results with GlcNAc were similar to those with glycerol: survival in GlcNAc was significantly compromised in almost all strains compared to those lacking glpD (Fig 3D). The glpD complement of the double ΔgpsA/ΔglpD mutant (ΔgpsA/glpD + ) was significantly compromised for survival compared to the ΔgpsA/ΔglpD mutant (Fig 3D). GlcNAc is not predicted to be metabolized by the action of GlpD, suggesting an unidentified link between GlpD and the metabolism of GlcNAc, possibly involving redox cofactors involved in the GlpD/GpsA oxidoreductase node. Together these results illuminate the importance of the metabolic balance of the intermediates and cofactors involved in the GpsA/GlpD node in B. burgdorferi adaptation to and survival during changing carbohydrate availability.

The gpsA mutant displays increased round body formation and decreased reductase activity
To assess the redox potential of the G3PDH mutants, overall reductase activity in individual cells was monitored by microscopy using a fluorescent reporter. Wild-type and mutant strains were grown to~7-9 × 10 7 cells ml -1 , collected and incubated in RPMI medium (nutrient stress) for 16 h. Cultures were then incubated with BacLight RedoxSensor Green and propidium iodide for 10 min before wet-mounting live cells to be imaged by fluorescence microscopy. RedoxSensor Green stain fluoresces (shown as cyan in Fig 4) when reduced, indicating intracellular reductase activity and cell viability. Almost all (90%) wild-type cells in nutrient stress media for 16 h show reductase activity/viability while only about 30% of ΔgpsA cells fluoresce (cyan staining, Fig 4A, 4B, and 4F). Strikingly, ΔgpsA cells undergo a dramatic morphological change from flat-wave to a condensed spherical form called round bodies (RBs) during nutrient stress that is rarely seen in wild-type B. burgdorferi (Fig 4A, 4B, and 4F). The physiological role of RBs remains unknown but the transition is triggered by environmental stress and may be related to persistence, as this form has been observed in vivo in ticks [18,[27][28][29]. Both the decrease in reductase activity and the increase in RB formation in ΔgpsA mutant cells are restored not only in the gpsA + strain but also in the ΔgpsA/ΔglpD double mutant (Figs 4C, 4D, and 4F), which is further evidence of glpD functioning as a suppressor mutation of the pleiotropic gpsA phenotypes. Deletion of the glpD gene alone affected neither reductase activity nor morphology during nutrient stress (Fig 4E and 4F). These data further support the findings that GpsA is important for cell survival under nutrient stress, a result likely reflected in reduced reductase activity. Additionally, the dramatic increase in RB formation of the ΔgpsA mutant is evidence that GpsA is a key modulator of morphological changes during nutrient stress.
A recent study has implicated GpsA in resistance of Streptococcus pneumoniae to oxidative stress [30]. To examine if gpsA also protects B. burgdorferi from oxidative stress, we measured the viability of wild-type, ΔgpsA and gpsA + strains following exposure to H 2 O 2 and found no significant differences between the strains, at least under the conditions tested (S2 Fig).

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick

GpsA and GlpD influence the B. burgdorferi metabolome
To better understand how GpsA and GlpD affect the global physiology of B. burgdorferi, we performed semi-targeted metabolomics by liquid chromatography-tandem mass spectroscopy on the ΔgpsA and ΔglpD mutants in culture. Strains were grown in BSK + RS at 35˚C to a density of~3 × 10 7 cells ml -1 before processing and analysis. Comparing the mutant metabolomes to wild type by unbiased principal component analysis, separate axes were readily observed that defined the metabolic effects of ΔgpsA versus the metabolic effects of ΔglpD. Complements of both mutants resulted in a return toward wild type and the ΔgpsA/ΔglpD double mutant comigrated with ΔglpD in agreement with the increased survival phenotype of the ΔgpsA/ ΔglpD double mutant ( Fig 5A). The first principal component, which defines the separation of ΔgpsA from wild type, is heavily loaded by the opposing behavior of ATP and AMP, positively loaded for NADH, and negatively loaded with G3P and three-carbon glycolytic intermediates ( Fig 5B). By contrast, ΔglpD separation along principal component two is positively loaded for G3P and negatively loaded for NADH in support of an opposing reaction directionality between GpsA and GlpD. Interestingly, both primary principal components are positively loaded for AMP and negatively loaded for ATP suggesting that proper functioning of the G3P arm of B. burgdorferi metabolism is essential for energy metabolism (Fig 5B).
In further support of the anticipated enzymatic function of GpsA, univariate analysis of the metabolite datasets showed that the largest metabolic changes in the ΔgpsA mutant compared to wild type were localized to the putative redox cofactor (NADH, 23.7-fold increase) and the anticipated product (G3P, 20.5-fold decrease) of GpsA ( Fig 5C). Despite its predicted role as a substrate of GpsA, DHAP levels were only 1.7-fold higher in the ΔgpsA mutant. This smaller effect compared to G3P and NADH is likely due to triose phosphate isomerase converting

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick DHAP to glyceraldehyde-3-phosphate (GAP) for glycolysis ( Fig 5D). With the exceptions of DHAP, GAP, and fructose-1,6-bisphosphate, glycolytic intermediates, particularly pyruvate (6.6-fold decrease), were decreased in the ΔgpsA mutant. Concomitantly, ATP levels decreased (2.1-fold) and AMP levels increased (2.6-fold). These changes in glycolysis likely drive the decreased energy levels in the cell and may account for the susceptibility of the ΔgpsA mutant strain to nutrient stress (Fig 3). The complete list of metabolite levels in all the mutants is included (S2 Table).
The dramatic 23.7-fold increase in NADH levels in the gpsA mutant compared to wild type suggests that GpsA is a dominant regulator of NADH levels in B. burgdorferi. The dysregulation of glycolysis may be a consequence of these elevated levels of NADH, which would inhibit glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and thus glycolysis in the ΔgpsA mutant [31] (Fig 5D). The changes in metabolite levels in the ΔgpsA mutant were largely

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick restored to wild-type levels in the gpsA + , but energy stress was still apparent in elevated AMP levels ( Fig 5C). Metabolite levels in the ΔgpsA/ΔglpD double mutant shifted toward wild-type levels, although not as completely as in the gpsA + . Univariate analysis further supported the action of GlpD as metabolically opposed to GpsA. While the metabolic phenotype in the ΔglpD mutant was of a smaller magnitude compared to the ΔgpsA mutant, DHAP levels in the ΔglpD mutant were 4.9-fold lower and G3P levels 4.2-fold higher than in the wild type (S3A Fig), supporting the predicted GlpD function. NADH levels were lower in the ΔglpD mutant, suggesting that either NADH is the reduced cofactor as G3P is oxidized to DHAP or that the effects of glpD deletion on GAP limit the recovery of NADH by GAPDH ( S3B Fig). The effects on glycolysis in the ΔglpD mutant were localized to DHAP, GAP, phosphoenolpyruvate, and pyruvate (S3A Fig). Complementation of the ΔglpD mutant overcorrected the elevated levels of G3P and somewhat restored levels of NADH associated with glpD deletion suggesting the reintroduction of active enzyme (S3A Fig). However, dysregulation of glycolysis and energy metabolism was not restored between the ΔglpD mutant and the glpD + , indicating that there was only partial restoration of the wild-type phenotype even though glpD was complemented in cis with the native glp promoter.

GpsA regulates the NADH/NAD + ratio
To confirm and quantify the GpsA-mediated regulation of nicotinamide cofactor levels observed in the metabolomics analysis, the NADH/NAD + molar ratios in the ΔgpsA mutant were measured via an in vitro luminescence assay. Strains were grown and samples prepared as described for the metabolomics studies before measuring the NADH and NAD + levels. The NADH/NAD + molar ratio was approximately fourfold higher in the ΔgpsA mutant strain than in wild-type cells (Fig 6). This difference was fully restored in the gpsA + , but, curiously, not in the double ΔgpsA/ΔglpD mutant strain (Fig 6). These data support the finding that gpsA highly regulates NADH levels in B. burgdorferi (Fig 5).

GpsA and GlpD levels are independent of each other
Because GpsA and GlpD catalyze the interconversion of G3P and DHAP, B. burgdorferi may respond to the mutation of one enzyme by altering the levels of the other G3PDH to compensate. To examine this possibility, levels of GpsA protein were examined in the ΔglpD mutant and levels of GlpD protein were examined in the ΔgpsA mutant by immunoblot analyses. Cell lysates analyzed from wild-type, ΔglpD mutant and glpD + strains grown in normal growth medium (BSK + RS) at 35˚C or incubated in nutrient stress medium (RPMI) showed no appreciable difference in GpsA protein levels ( Fig 7A). FlaB was used as a loading control ( Fig  7A and 7B). Similarly, GlpD levels did not change in the ΔgpsA mutant compared to the wild type or the gpsA + under similar conditions ( Fig 7B). Thus, GpsA and GlpD protein levels are independent of each other, at least during in vitro culture.

GpsA is required for murine infectivity
Because GpsA and GlpD dramatically affect B. burgdorferi metabolism, particularly NADH levels, we examined the role of this oxidoreductase cycle in vivo using the mouse-tick model of Lyme disease. To genetically assay the function of GpsA and GlpD in murine infection, 10 4 cells of wild-type, ΔgpsA, gpsA + , ΔglpD, glpD + and ΔgpsA/ΔglpD strains were intradermally needle-inoculated in C3H-HeJ mice. Infection and dissemination were assessed via ear biopsies three weeks post-inoculation and tissues were cultured for B. burgdorferi. None of the mice injected with the ΔgpsA mutant were infected (Table 1).

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick Infectivity was restored in the gpsA + but not in the ΔgpsA/ΔglpD double mutant. Ear, ankle and bladder tissues were collected from the same mice five weeks post inoculation and cultured for B. burgdorferi to follow infectivity and dissemination. The results at five weeks were consistent with those at three weeks: the ΔgpsA mutant and ΔgpsA/ΔglpD double mutant were still unable to establish infection (Table 1). Because GpsA is potentially involved in providing G3P for lipoprotein biosynthesis, we examined the levels of two lipoproteins, outer surface proteins A and C (OspA and OspC), in the ΔgpsA and ΔgpsA/ΔglpD mutants. In strains grown in vitro, under the same conditions used for mouse infectivity studies, there was no appreciable difference in the levels of OspC or OspA (S4 Fig). The absence of glpD did not affect murine infectivity by needle inoculation, results agreeing with previous studies [12,13].
To determine if the infectivity defect of the ΔgpsA and ΔgpsA/ΔglpD mutants could be overcome by increasing the challenge dose, mice were inoculated with 10 5 or 10 6 cells of each strain, and infectivity and dissemination monitored as described above at three and five weeks

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick after inoculation. The ΔgpsA mutant was noninfectious even with a dose of 10 6 cells (Table 1). At three weeks post challenge, B. burgdorferi could not be isolated from any mice injected with either 10 5 or 10 6 of ΔgpsA/ΔglpD double mutants (Table 1). Surprisingly, five weeks after challenge with the higher doses of the ΔgpsA/ΔglpD double mutant, 20/24 tissues were positive for B. burgdorferi re-isolation, indicating that deletion of glpD suppresses the non-infectious phenotype of the ΔgpsA mutant, thus partially restoring infectivity, albeit with delayed kinetics. These data provide additional in vivo evidence that glpD functions as a suppressor mutation of the non-infectious phenotype of the ΔgpsA mutant.

GpsA is crucial for persistence in the tick and transmission to mice
gpsA is essential for survival during nutrient stress during in vitro cultivation (Fig 3B, 3C, and  3D); therefore, we hypothesized that gpsA functions in vivo during persistence in the tick, when the spirochetes experience nutrient limitation after the tick absorbs the blood meal. To

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick test this hypothesis, we artificially infected naïve larvae with wild-type, ΔgpsA, gpsA + and ΔgpsA/ΔglpD strains as previously described [32]. B. burgdorferi loads were measured in ticks through maturation and quantified by qPCR using primers to the B. burgdorferi flaB gene on DNA isolated from infected ticks. The ΔgpsA mutant was unable to establish a robust colonization in larvae compared to wild type and this deficiency was restored in the gpsA + (Fig 8A). The ΔgpsA/ΔglpD double mutant was also compromised for colonization of larvae, indicating that this ΔgpsA mutant phenotype was not suppressed by the deletion of glpD (Fig 8A). Persistence through the molt from fed larvae to flat nymphs continued to be compromised in both the ΔgpsA mutant and the ΔgpsA/ΔglpD double mutant compared to wild type and the gpsA + , as B. burgdorferi loads fell to an average of less than 10 per tick (Fig 8B). Following feeding of infected nymphs on naïve mice, B. burgdorferi levels in ΔgpsA and ΔgpsA/ΔglpD-infected ticks increased but remained greater than three orders of magnitude below those found in wild type and gpsA + -infected fed nymphs (Fig 8C). These in vivo data demonstrate a crucial, although not essential, role for gpsA to colonize and persist in ticks in the animal model of Lyme disease.
Transmission by tick bite is the natural route of B. burgdorferi infection of vertebrates. To examine the role of gpsA in transmission of B. burgdorferi from nymphs to naïve mice, three nymphs per mouse were allowed to feed until repletion. Murine infections were monitored three and five weeks post-feeding by culturing collected murine tissues as described above. The ΔgpsA mutant was not transmitted to mice (0/3 mice infected) and this defect was restored in the gpsA + strain (2/3 mice infected) ( Table 2). The ΔgpsA/ΔglpD double mutant was also compromised for transmission (0/3 mice infected) ( Table 2), suggesting that glpD does not suppress the gpsA mutation in tick transmission.

Discussion
We have provided experimental evidence that the B. burgdorferi gpsA encodes a G3PDH, as predicted [5], by heterologously complementing an E. coli gpsA mutant (Fig 2). Our metabolomics analyses suggest that GpsA converts DHAP to G3P, as the ΔgpsA mutant showed reduced levels of G3P and increased levels of DHAP and NADH (Fig 5), providing additional evidence that gpsA encodes a G3PDH (or, based on the directionality of the reaction, a DHAP reductase). We also identified NADH as the likely cofactor of the redox reaction. Previous studies have shown that GlpD plays a role in growth on glycerol, but the substrates and products of GlpD have not been experimentally examined [12,13,15]. Here, our metabolomics analyses of the glpD mutant demonstrate a significant increase of G3P and decrease of DHAP levels compared to wild type, experimental evidence suggesting GlpD oxidizes G3P to DHAP. The lower

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick NADH levels seen in the ΔglpD strain suggest that NAD + serves as the cofactor for oxidation of G3P by GlpD. An alternative possibility is that NAD + is reduced downstream by a different cofactor, such as FADH 2 , used by GlpD.

Metabolic virulence
The intersection of metabolism and virulence in microbial pathogens is gaining attention in the quest to define the molecular mechanisms responsible for causing disease [33]. In particular, bacterial glycerol metabolism and the associated physiological landscape can influence disease pathogenesis [34]. In the limited metabolic capacity of B. burgdorferi, glycerol metabolism plays an important role in the ability of the spirochete to traverse its enzootic cycle and, thus, the glycerol utilization genes are subject to complex regulation [5,7,8,16]. In this work we have presented evidence identifying a physiological linchpin connecting glycerol metabolism/lipid biosynthesis and glycolysis/energy generation that regulates stress survival in vitro and is crucial in vivo for both murine infectivity and persistence in the tick vector. This oxidoreductive node consists of two opposed G3PDHs, GpsA and GlpD, which catalyze the interconversion of G3P and DHAP, and likely communicates closely with the adjacent GAPDH. We found that gpsA is essential for disseminated infectivity in mice, by needle inoculation of up to 10 6 cells (Table 1), and is crucial for persistence in ticks (Fig 8). ΔgpsA mutants and ΔgpsA/ΔglpD double mutants were also compromised for transmission by nymphs, but the tick transmission phenotype is likely due, at least in part, to reduced spirochete loads in the tick. The infectivity phenotype may result from the inability of these mutant strains to establish infection at the inoculation (or tick bite) site or may reflect a defect in dissemination or persistence, but uncovering these details awaits further investigation. Additionally, a more detailed time course of ΔgpsA mutant and ΔgpsA/ΔglpD double mutant replication during tick feeding may shed light on how these strains are unable to complete the vector-host lifecycle in the laboratory model of Lyme disease. Recent studies have suggested that gpsA may be a virulence factor in other pathogens. Green et al., 2021 [30] found using a forward genetic screen that GpsA is important for S. pneumoniae to establish nasopharyngeal colonization and plays a role in resistance to oxidative stress. A G3PDH mutant strain of the plant pathogen Acidovorax citrulli showed reduced Fig 8. Persistence of the gpsA and glpD mutants in ticks. B. burgdorferi strains were introduced to larval ticks by immersion infection and fed on naïve mice. Acquisition and persistence of wild type (WT, black circles), ΔgpsA mutant (white circles), gpsA complement (gpsA + ) (dark gray circles) and ΔgpsA/ΔglpD double mutant (light gray circles) was assessed in ticks one week after larvae fed to repletion (A), following the molt to nymphs, nine weeks later (B) and one week after nymphs fed to repletion on naïve mice (C). DNA from infected ticks was isolated at each stage and B. burgdorferi load in ticks measured by TaqMan qPCR using primers/probe to flaB. Bars represent the mean of data points from a single experiment. The means of both the ΔgpsA and ΔgpsA/ΔglpD mutants were significantly different ( � p < 0.0007) from both the wild type and the gpsA + determined by one-way ANOVA with a Tukey's posthoc test in (B). The mean of the gpsA + was significantly different ( �� p � 0.03) from both ΔgpsA and ΔgpsA/ΔglpD mutants as determined by one-way ANOVA with a Tukey's post-hoc test in (C). https://doi.org/10.1371/journal.ppat.1010385.g008

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick virulence and increased resistance to antibiotics [35]. Additionally, a genome-wide mutagenesis screen identified a gpsA mutant in Bartonella that was avirulent, but the transposon insertion is upstream of the gpsA gene, complicating interpretations of the results [36]. Thus, GpsA activity, converting DHAP to G3P using a redox cofactor such as NADH, while not important for viability and growth under optimized culture conditions, represents a metabolic virulence factor important during the challenges of establishing host infection and resistance to immune defenses.

GlpD as a suppressor mutant
We have identified the glpD mutant as a suppressor of several gpsA mutant phenotypes, including murine infectivity by needle inoculation at high doses, restoration of viability under nutrient stress, and morphological changes. Much of the metabolic imbalance observed in our metabolomic studies of the ΔgpsA mutant were restored in the ΔgpsA/ΔglpD double mutant, particularly the dramatic increase in NADH and decrease in G3P levels. Alleviating this imbalance by the absence of GlpD activity, which results in a decrease of NADH levels and an increase in G3P levels, may explain the suppression of the defects in the ΔgpsA mutant. There is precedent for the interdependence of gpsA and glpD in E. coli. The gpsA gene was thought to be essential in E. coli as mutants were not generated in a genome-wide, single gene deletion library known as the Keio collection [37]. However, deletion of the gpsA gene previously had been accomplished in a glpD mutant background in E. coli [38], supporting our assertion that glpD suppresses aspects of the pleiotropic gpsA mutant phenotype. Together, these findings suggest this node may be a common stress point, although bacteria with a more diverse metabolic repertoire may be better able to adjust for the metabolic imbalances created by genetic perturbation of either of these G3PDHs. The glpD mutant suppressed some, but not all, of the defects of the gpsA mutant in the animal model of Lyme disease: infectivity by needle inoculation (10 5 and 10 6 cells) was restored, but persistence in ticks and transmission from nymphs to mice were not. These results suggest that certain requirements and/or signals in ticks cannot be overcome by complete disruption of the GpsA/GlpD node and likely require efficient carbon flow between glycolysis and lipid/ lipoprotein biosynthesis. Even though levels of the outer membrane lipoproteins OspC and OspA were not affected in vitro in the ΔgpsA and ΔgpsA/ΔglpD mutants, there may be differences in lipoprotein production in vivo that contribute to the defect in persistence in the tick.

The GpsA and GlpD metabolomes
Although the host metabolome in response to Lyme disease has been analyzed [39,40], here we report the first broad metabolomic datasets of B. burgdorferi, consisting of 129 metabolites. Our results identified GpsA as a dominant regulator of NADH levels and redox potential in the spirochete. The 23-fold increase in NADH in the ΔgpsA mutant compared to wild type was partially dependent on GlpD activity as levels were largely restored in the ΔgpsA/ΔglpD double mutant (5.5-fold increase over wild type; Fig 5C). The redox state of the cell, reflected at least in part by the NADH/NAD + ratio, is intimately entwined with glycolysis and secondary carbon metabolism [6,7]. Therefore, changes in the levels of glycolytic intermediates seen in the ΔgpsA mutant were likely a secondary effect of extremely high levels of NADH, a cofactor known to inhibit GAPDH activity [31]. This NADH-mediated inhibition could dramatically affect carbon flow in glycolysis. In fact, we found evidence to support this conclusion as levels of the two metabolites upstream of GAPDH, fructose-1,6-bisphosphate and GAP, were elevated and levels of all metabolites downstream were decreased. Again, in the ΔgpsA/ΔglpD double mutant where levels of NADH were closer to those in wild-type cells, the levels of all

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick intermediates in glycolysis trended back toward wild-type levels, with the majority no longer significantly different. The same trend was observed in energy metabolism as represented by the levels of ATP, which could explain the differing viability in vitro under nutrient stress between the ΔgpsA mutant and the ΔgpsA/ΔglpD double mutant (Fig 3B, 3C, and 3D).
The B. burgdorferi genome lacks the genes encoding the enzymes required for de novo NAD + synthesis, but contains homologs of enzymes that convert nicotinamide to NAD + (pncA, pncB, nadD and nadE), although a nicotinamide transporter has not yet been identified [5,6]. In the absence of an electron transport chain and reductive biosynthetic processes such as fatty acid synthesis, many routes of NADH oxidation are absent in B. burgdorferi [7]. The ability of B. burgdorferi to recycle NADH is further limited by the absence of genes encoding pyruvate dehydrogenase, lactate oxidase and pyruvate formate lyase. NAD + can be recovered by lactate dehydrogenase converting pyruvate, formed in glycolysis, to lactate, which is then excreted from the cell through lactate permease [5]. The reverse reaction of converting lactate to pyruvate would regenerate NADH from NAD + . The constraints placed on the NAD + / NADH balance in B. burgdorferi likely result in increased sensitivity to disruption of this ratio and a greater need to recycle both the oxidized and reduced cofactors, particularly in glycolysis where NAD + is used as an electron acceptor. GpsA represents one of the few enzymes identified in B. burgdorferi able to recycle NADH to NAD + , while at the same time supplying G3P, but at a cost to energy production from glycolysis.
NADH-dependent oxidative stress enzymes, such as coenzyme A disulfide reductase (CoADR) and superoxide dismutase (SodA) also can affect the redox state of the cell. CoADR is involved in reducing the disulfide form of coenzyme A occurring in an oxidative environment, using exclusively NADH [41]. The response of CoA-metabolites to gpsA deletion suggests that the glycerol oxidoreductase node and the CoADR systems are in metabolic communication, likely via NADH (Fig 5C). CoADR protects B. burgdorferi from lipid peroxidation, is essential for host infectivity and plays a role in survival in fed nymphs [41,42]. SodA in B. burgdorferi is required for murine infectivity, like gpsA and cdr, but the role in ticks has not been investigated [43]. A sodA deletion mutant has altered levels of NADH and NAD + , as well as ATP, but all were approximately twofold different from wild type [44]. Other NADHdependent enzymes are present in B. burgdorferi, but many have not been studied or are presumed to be essential, and none have been shown to influence metabolite (NADH and G3P) levels to the degree observed in the ΔgpsA mutant.

Increased round body formation
B. burgdorferi morphology can change from the signature flat wave to RBs in vitro in response to environmental conditions, including nutrient stress [18,27,28,45] and in vivo in the tick [29] to presumably aid in B. burgdorferi persistence. RBs are a condensed spherical form of B. burgdorferi within an outer membrane that are viable and transitory [18,[27][28][29]. The stringent response intracellular second messenger (p)ppGpp and the alternative sigma factor RpoS have been shown to control RB formation, as disrupting these global transcriptional regulators increased RB formation [18,29]. Cellular physiology also appears to be important as mutation of CoADR increases RB formation [42] and levels of acetyl-CoA and CoA are elevated in the ΔgpsA mutant (Fig 5C). Our results showing a massive transition of the ΔgpsA mutant to RBs during nutrient stress identify another metabolic enzyme involved in this morphologic change. Disruption of the redox potential of the cell may be a crucial signal for RB formation as both GpsA and CoADR use NADH as a cofactor and NADH levels dramatically increase and viability/reductase activity decrease in the ΔgpsA mutant (Fig 4, 5, and 6). Notably, many of the genes that regulate RB formation, rel Bbu , cdr and gpsA, are also important for persistence in the PLOS PATHOGENS B. burgdorferi GpsA is crucial for host infection and persistence in the tick tick. The RB is not simply an aberrant non-viable form as many RBs remain viable (Fig 4) and can revert back to flat wave morphological forms [18,29,45]. Thus, environmental stresses that significantly alter the redox potential of the cell could trigger RB formation to aid survival under harsh conditions, such as in the tick between blood meals, but the mechanism and benefits of this morphological transformation for B. burgdorferi remain to be determined.

Compromised survival in vitro and in vivo during persistence in ticks
B. burgdorferi must adapt to the stress of depleted nutrients in the tick midgut as the blood meal is consumed following larval and nymphal feeding. We and others have used RPMI in vitro to mimic the nutrient stress experienced in the tick midgut between blood meals and have identified the Rel Bbu -mediated stringent response and its effector DksA as important for survival under these conditions [17,18]. Here we found that gpsA, while not necessary for growth in nutrient-rich medium, was crucial for survival during nutrient stress in vitro, suggesting that NADH levels and redox potential have a key role in adapting to nutrient limitation. Completely severing the G3PDH node in the ΔgpsA/ΔglpD double mutant partially restored NADH levels and rescued both the survival defect and reductase activity, indicating that the inability to shuttle DHAP to G3P for lipid biosynthesis was not the cause of cell death in vitro. Also, the metabolome of the ΔglpD mutant showed altered levels of both G3P and DHAP (S3 Fig) but this mutant was neither compromised for survival nor reductase activity, suggesting the changes in these metabolites were not involved in adapting to nutrient stress. Furthermore, the increased levels of NADH in the ΔgpsA mutant could have theoretically increased the resistance to peroxide, but that was not observed under our conditions (S2 Fig). Other redox enzymes using NAD + /NADH as a cofactor play a role in B. burgdorferi survival in response to environmental stresses, particularly oxidative stress. The CoADR mutant had a growth defect in vitro and was more susceptible to lipid peroxidation, but not reactive oxygen species, and a sodA mutant was also more susceptible to superoxide radicals [41][42][43]46]. Thus, oxidoreductase enzymes using NAD + /NADH as a cofactor likely play a variety of roles in adapting to external stresses encountered throughout the enzootic environment.
Because decreased levels of G3P, needed for lipid and lipoprotein biosynthesis, in the ΔgpsA mutant could account for the survival defect observed under nutrient stress, we added glycerol to RPMI in an attempt to chemically rescue the phenotype. Surprisingly, even in wild-type B. burgdorferi, glycerol was cytotoxic in nutrient stress media (Fig 3C). Furthermore, glycerol was cytotoxic to all strains except those lacking glpD (ΔglpD, ΔgpsA/ΔglpD and gpsA + /ΔglpD; Fig 3C), indicating that GlpD was necessary for glycerol-mediated cell death. The cytotoxicity of glycerol may be due the continued action of GlpD oxidizing G3P to feed glycolysis, which amplifies the deficit of NAD + . Even more unexpected was the finding that added GlcNAc produced the same glpD-dependent cell death phenotype (Fig 3D). Because there is no known shared metabolic pathway between transport or metabolism of GlcNAc and transport or metabolism of glycerol (until glycolysis), we hypothesize that cofactor dysregulation in nutrient-depleted media may cause cell death. Perhaps the addition of another nutrient to the media could rescue B. burgdorferi viability by restoring the metabolic balance, but these studies remain to be done.
gpsA is also crucial for persistence in vivo in the tick: gpsA mutants seem to have difficulty colonizing larvae, although the difference from wild type-infected larvae was not significant, and the mutants were severely compromised (1000-10,000-fold) for persistence in unfed nymphs and replication in fed nymphs (Fig 8). Again, the dysregulation of the redox potential and NADH levels may explain the persistence defect as similar pathways have been identified as important in other studies. CoADR and Dps/NapA/BicA involved in the oxidative stress

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick response were both shown to function during persistence in feeding nymphs [42,47]. Glycerol metabolism has been shown to be important for persistence in ticks [12,13], as glpD mutants have a threefold survival defect in fed larvae [14] and ten-to fifteenfold survival defect following nymph feeding [13]. The degree of dysregulation of NADH/NAD + levels correlates with the severity of the persistence defect, suggesting this could be driving the tick phenotype. Unlike our in vitro results, the persistence defect in ticks was not restored in the ΔgpsA/ΔglpD double mutant indicating that there are additional challenges in the tick environment that the double mutant is unable to overcome. The defect in persistence in the tick could also be explained by compromised lipid and lipoprotein production in the gpsA mutant strains. The biosynthetic lipid and lipoprotein pathways, including GlpF, GlpK, 1-acyl-G3P acyltransferase (bb0037), and two fatty acid CoA ligases (bb0137 and bb0593), are induced in ticks compared to mammalian hosts, suggesting an important role in the vector [48]. Thus, disruption of G3P production in the gpsA mutants may stress the lipid/lipoprotein biosynthetic machinery resulting in decreased viability and persistence. Taken together, our studies and previous work highlight the importance of this metabolic node for persistence in the tick vector and infectivity in the vertebrate host.

Ethics statement
All animal experiments were approved by the University of Montana Institutional Animal Care and Use committee and followed the Guidelines for the Care and Use of Laboratory Animals from the National Institutes of Health.

B. burgdorferi strain construction
The parental strain B31-5A4, referred to as wild type, was used to construct all mutant strains. E. coli TOP10F 0 grown in lysogeny broth [51] were used for cloning. To construct the gpsA null mutant strain the regions upstream and downstream of the gpsA gene were amplified by PCR using KOD polymerase (Novagen) and the primers gpsA_U886F/gpsA_41R+AatII+AgeI or gpsA_1051F+AatII/gpsA_D1924R+AgeI (S1 Table), respectively, and B31-5A4 genomic DNA as a template. PCR products were cloned into pCR2.1-TOPO (Invitrogen) and verified by DNA-sequencing. Using the artificially engineered AgeI and AatII restriction sites, the two cloned PCR products were digested with AatII and AgeI and ligated together at the junction of the AatII site. The streptomycin/spectinomycin resistant cassette with the flgB promoter from B. burgdorferi [52] and trpL terminator from Bacillus subtilis [53] (flgBp-aadA-trpLt) containing flanking AatII sites was placed between the upstream and downstream DNA segments. The plasmid containing the aadA cassette replacing gpsA was purified and linearized using the restriction enzyme AhdI before electroporating competent B31-5A4 cells. 24 h after transformation, cells were plated in liquid BSK + RS in 96-well plates and transformants selected with 50 μg ml -1 streptomycin [54]. Colonies resistant to streptomycin were screened for the absence of gpsA by PCR analysis using the primers gpsA_U87F and gpsA_D1174R (S1B Fig). To

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick generate the gpsA cis-complemented strain, the entire gpsA ORF and promoter were amplified by PCR using the primers gpsA_U886F and gpsA_D1125R+AatII+AgeI to yield the upstream segment. PCR amplification using the primers gpsA_1051F+AatII and gpsA_D1924+AgeI were used to produce the downstream fragment. As described above, these segments were ligated together and a gene conferring resistance to kanamycin (flgBp-aphI-trpLt) inserted downstream of the gpsA gene. The resulting plasmid was purified, linearized, and electroporated into competent ΔgpsA mutant cells, which were plated as described above except that transformants were selected in 200 μg ml -1 kanamycin. Restoration of the gpsA gene in isolated clones was confirmed by PCR analyses using the primers gpsA_U87F and gpsA_D1174R (S1B Fig). The glpD null mutant strain was constructed as described above for the gpsA null mutant except that the primers glpD_U1016F and glpD_10R+AatII+AgeI were used to make the upstream segment and glpD_1568F+AatII and glpD_D2523R+AgeI were used to make the downstream segment. The gene conferring gentamicin resistance (flgBp-aacC1-trpLt) was inserted and transformants selected in 50 μg ml -1 gentamicin. Clones resistant to gentamicin were screened by PCR for the absence of the glpD gene using the primers glpD_U56F and glpD_D1669R (S1D Fig). The glpD cis-complemented strain was constructed as described for the gpsA complement except the upstream and downstream segments were amplified by PCR using the primers glpD_U719F/glpD_D1613R+AatII+AgeI and glpD_D1611F+AatII and glpD_D2523R+AgeI, respectively (S1 Table). The flgBp-aphI-trpLt I gene was inserted downstream of the glpD gene and transformants selected in kanamycin and clones screened by PCR using the primers glpD_U56F and glpD_D1669R (S1C and S1D Fig and S1 Table). Competent ΔgpsA mutant cells were transformed with the construct described above and selected in gentamicin to generate the ΔgpsA/ΔglpD double mutant strain. The same strategy used to make the gpsA complemented strain was used to make the gpsA complement of the double ΔgpsA/ ΔglpD mutant to yield the gpsA + /ΔglpD strain. Likewise, the glpD complement of the double ΔgpsA/ΔglpD mutant to generate the ΔgpsA/glpD + strain was as described above. The gpsA and glpD double complement of the ΔgpsA/ΔglpD double mutant was made by transforming competent gpsA + /ΔglpD cells described above with a glpD complement construct described above except that a streptomycin resistance cassette (flgBp-aadA-trpLt) was used and transformants selected in streptomycin. Absence of the gpsA gene in the double ΔgpsA/ΔglpD strain and restoration of the gpsA gene in the gpsA + /ΔglpD and double complemented gpsA + /glpD + isolated clones was confirmed by PCR analyses using the primers gpsA_385F and gpsA_493R (S1F Fig). Absence of the glpD gene in the ΔgpsA/ΔglpD double mutant and restoration of the glpD gene in the ΔgpsA/glpD + single complement and gpsA + /glpD + double complement was confirmed by PCR using the primers glpD_1F+SacI and glpD_1267R+AatII (S1G Fig). All strains used for infectivity experiments were screened for the presence of plasmids important for murine infectivity and persistence in the tick [55].

Heterologous expression of B. burgdorferi gpsA in E. coli and growth analyses
The B. burgdorferi gpsA gene was amplified by PCR using KOD polymerase and the primers gpsA_1F+SacI and gpsA_1092R+PstI (S1 Table). The PCR product was cloned into pCR2.1-TOPO (Invitrogen) and verified by DNA sequencing. The gpsA gene was inserted into the multiple cloning site of the expression vector pUC18 by digesting both with the restriction enzymes PstI and SacI followed by ligation. The resulting pUC18-gpsA Bb or the empty vector pUC18 were transformed into a gpsA null mutant E. coli strain (CGSC # 5424, E. coli Genetic Stock Center, Yale University) [25]. The E. coli strains ΔgpsA, ΔgpsA + pUC18 and ΔgpsA

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick + pUC18-gpsA Bb were grown in M9 minimal salts media containing 1% casamino acids with or without 1% glucose. To induce gpsA Bb expression, 0.1 mM IPTG was added to all cultures. Cultures were grown in 96-well plates at 37˚C with constant shaking in a BioTek Synergy HT plate reader (Agilent) where the OD 600 was measured every 17 min.

Nutrient stress and semi-solid BSK plating assay
B. burgdorferi strains were grown to 4-9 × 10 7 cells ml -1 , a portion diluted in BSK (approximately 200 spirochetes), and plated in semi-solid BSK (0 h time point). The remaining cultures were collected by centrifugation at 8000 x g for 10 min at 4˚C. Cells were resuspended in RPMI 1640 without L-glutamine (Corning) either without or with 0.4% glycerol (Fisher Scientific) or 0.4% N-acetylglucosamine (MP Biomedicals) for nutrient stress assays as previously described [18]. After 24 h at 35˚C cells, were diluted in BSK and plated in semi-solid BSK (24 h time points) and all plates (0 h and 24 h) allowed to grow at 35˚C for about 14 days in a 5% CO 2 incubator before colony enumeration.

RedoxSensor staining and microscopy
B. burgdorferi strains were grown in BSK + RS at 35˚C to~7 × 10 7 cells ml -1 before collection at 8,000 x g for 10 min at 4˚C. Cell pellets were resuspended in warm RPMI and incubated at 35˚C for 16 h. Cultures were then stained with BacLight RedoxSensor Green and propidium iodide according to the manufacturer's instructions (1 μl RedoxSensor Green and 1 μl propidium iodide in 1.0 ml culture) from the BacLight RedoxSensor Green Vitality kit (Invitrogen). Cells were stained for 10 min at 35˚C, collected by centrifugation (13,000 x g, 5 min, 4˚C) and cell pellets resuspended in 1.0 ml PBS. 10 μl of cells were wet-mounted on a slide and live cells imaged using an Olympus BX51 fluorescence microscope with 100x/1.30 NA objective. Images were processed using ImageJ (National Institutes of Health; http://rsbweb.nih.gov/ij/) and Pixelmator (Pixelmator Team, Ltd). The mean percent cells stained with RedoxSensor Green and mean percent round body (RB) cells were determined from three independent biological replicates where at least 100 cells were enumerated for each strain. Significant difference between the means was determined by one-way ANOVA with a Tukey's post-hoc test.

Metabolomic analyses
For all liquid chromatography-mass spectrometry (LCMS) methods, LCMS grade solvents were used. Tributylamine and all synthetic molecular references were purchased from Millipore Sigma. LCMS grade water, methanol, isopropanol and acetic acid were purchased through Fisher Scientific.
B. burgdorferi strains were grown in BSK + RS at 35˚C to~3 × 10 7 cells ml -1 before collection at 8,000 x g for 10 min at 4˚C. Cells were washed twice with HEPES-NaCl buffer and pellets were flash frozen in dry-ice and ethanol and stored at -80˚C until processing. To process cells for metabolomic analysis, cell pellets were thawed on ice, resuspended in 150 μL ice-cold methanol (Sigma) and then incubated at room temperature for 10 min. Following incubation an equal volume of LCMS grade water was added and the samples were vigorously vortexed, then centrifuged at 13,000 x g for 15 min. Supernatants were collected, filtered in a 0.2 μM nitrocellulose syringe filter (GE Healthcare), and then diluted 1:3 before analysis. All samples were separated using a SCIEX ExionLC AC system and measured using a SCIEX 5500 QTRAP mass spectrometer. Polar metabolites were analyzed using a previously established ion pairing method with modification [56,57]. Quality control samples were injected after every 10 injections to control for signal stability. Samples were separated with a Waters Atlantis T3 column (100Å, 3 μm, 3 mm X 100 mm) using a binary gradient from 5 mM tributylamine, 5 mM acetic

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick acid in 2% isopropanol, 5% methanol, 93% water (v/v) to 100% isopropanol over 15 minutes. Each metabolite was identified and measured with two ion fragmentation pairs and a defined retention time.
All signals were integrated using MultiQuant Software 3.0.3. Signals with greater than 50% missing values were discarded and remaining missing values were replaced with the lowest registered signal value. All signals with a QC coefficient of variance greater than 30% were discarded. Metabolites with multiple ion pairs were quantified with the signature that displayed the highest signal to noise. All filtered datasets normalized against the total signal sum for the injection prior to analysis. Single and multi-variate analyses were performed in MarkerView Software 1.3.1. All univariate comparisons were subjected to a Benjamini-Hochberg cut-off at a false discovery rate of 5%. NAD + /NADH assays B. burgdorferi strains were grown at 35˚C to~5 × 10 7 cells ml -1 before collection at 14,000 x g for 10 min at 4˚C. Cell pellets were washed twice with 1.0 ml cold H-N buffer (50 mM HEPES, 50 mM NaCl, pH 7.6), 15,800 x g, 5 min at 4˚C, in cryovials. The supernatant was discarded and cell pellet flash frozen in liquid N 2 and stored at -80˚C. Samples were solubilized and NAD + and NADH levels measured separately using the NAD/NADH-Glo Assay Kit (Promega) according to the manufacturer's instructions. Luminescence was measured using a BMG Labtech/CLARIOstar plate reader. The levels of NAD + and NADH were normalized to cell number from three independent biological replicates. Significant difference between the means was determined by one-way ANOVA with a Tukey's post-hoc test.

Mouse infectivity by needle inoculation and tick transmission
C3H-HeJ mice (Charles River Laboratories) were intradermal injected with 1 × 10 4 to 1 × 10 6 B. burgdorferi cells grown at 35˚C to cell densities of 5-9 × 10 7 cells ml -1 . Infection was determined by culturing mouse ear biopsies in BSK II containing 50 μg ml -1 rifampicin, 20 μg ml -1 phosphomycin and 2.5 μg ml -1 amphotericin B and examining cultures by dark-field microscopy for the presence of spirochetes for 14 days. Five weeks post-infection, mice were sacrificed and ear, ankle and bladder tissues were collected, cultured and examined for spirochetes. Uninfected Ixodes scapularis larvae (National Tick Research and Education Resource, Oklahoma State University) were maintained in a 98% humidified chamber. To allow ticks to acquire B. burgdorferi, approximately 100 larvae per mouse were allowed to feed to repletion. Spirochete loads per tick were quantified by qPCR as described below. After infected larvae molted into nymphs (about 8 weeks), three mice were infested with three nymphs each and

PLOS PATHOGENS
B. burgdorferi GpsA is crucial for host infection and persistence in the tick allowed to feed to repletion. Murine transmission was monitored by culturing ear, ankle and bladder tissues as described above. Persistence of B. burgdorferi in ticks was followed by qPCR as described below.

Quantification of B. burgdorferi in ticks
Spirochete burdens in artificially infected ticks were assessed in fed larvae (one week, in groups of five), unfed nymphs (in groups of five) and fed nymphs (one week post feeding to repletion). DNA was extracted from ticks using the DNeasy Blood/Tissue kit (Qiagen) and the number of B. burgdorferi genomes in each tick determined by TaqMan qPCR with primers flaB_425F and flaB_542R and the flaB probe to the flaB gene (S1 Table) were 1 copy of flaB = one genome as previously described [18].
Supporting information S1  Strain construction and analyses. (A) Illustration of the gpsA mutant (ΔgpsA) and gpsA complemented (gpsA + ) strains constructed by homologous recombination. The gpsA gene (bb0368) was replaced with the B. burgdorferi promoter from flgB fused to aadA, a gene conferring resistance to streptomycin (strep R ). A wild-type copy of gpsA was reintroduced in cis using the B. burgdorferi promoter from flgB fused to aphI, a gene conferring resistance to kanamycin (kan R ) to yield the gpsA + strain. (B) Genomic DNA isolated from wild-type (WT), ΔgpsA and gpsA + strains, and a no template control (NTC) was analyzed by PCR using primers 1 (gpsA_U87F) and 2 (gpsA_D1174R). (C) Illustration of the glpD mutant (ΔglpD) and glpD complemented (glpD + ) strains constructed by homologous recombination. The glpD gene (bb0243) was replaced with the B. burgdorferi promoter from flgB fused to aacC1, a gene conferring resistance to gentamicin (gent R ). A wild-type copy of glpD was reintroduced in cis using the B. burgdorferi promoter from flgB fused to aadA, a gene conferring resistance to streptomycin (strep R ) to yield the glpD + strain. (D) Genomic DNA isolated from WT, ΔglpD and glpD + strains, and a NTC was analyzed by PCR using primers 3 (glpD_U56F) and 4 (glpD_D1669R). (E) Illustration of the double gpsA/glpD mutant (ΔgpsA/ΔglpD) and single and double complemented strains constructed by homologous recombination as described above. The gpsA complement of the double mutant (gpsA + /ΔglpD) was used to construct the gpsA-glpD double complement (gpsA + /glpD + ). (F) Genomic DNA isolated from WT, the double gpsA/glpD mutant (ΔgpsA/ΔglpD), the gpsA complement of the double mutant (gpsA + / ΔglpD) and the gpsA and glpD double complement of the double mutant (gpsA + /glpD + ) strains, and a NTC was analyzed by PCR using primers 5 (gpsA_385F) and 6 (gpsA_493R). (G) Genomic DNA isolated from WT, the double gpsA/glpD mutant (ΔgpsA/ΔglpD), the glpD complement of the double mutant (ΔgpsA/glpD + ) and the gpsA and glpD double complement of the double mutant (gpsA + /glpD + ) strains, and a NTC was analyzed by PCR using primers 7 (glpD_1F+SacI) and 8 (glpD_1267R+AatII). B. burgdorferi strains were grown in BSK+RS at 35˚C, total cell lysates were collected and equal amounts separated by SDS-PAGE; proteins were transferred to membranes and analyzed by immunoblot with antibodies against