Role of Staphylococcus aureus Formate Metabolism during Prosthetic Joint Infection

ABSTRACT

Biofilms are bacterial communities characterized by antibiotic tolerance. Staphylococcus aureus is a leading cause of biofilm infections on medical devices, including prosthetic joints, which represent a significant health care burden. The major leukocyte infiltrate associated with S. aureus prosthetic joint infection (PJI) is granulocytic myeloid-derived suppressor cells (G-MDSCs), which produce IL-10 to promote biofilm persistence by inhibiting monocyte and macrophage proinflammatory activity. To determine how S. aureus biofilm responds to G-MDSCs and macrophages, biofilms were cocultured with either leukocyte population followed by RNA sequencing. Several genes involved in fermentative pathways were significantly upregulated in S. aureus biofilm following G-MDSC coculture, including formate acetyltransferase (pflB), which catalyzes the conversion of pyruvate and coenzyme-A into formate and acetyl-CoA. A S. aureus pflB mutant (ΔpflB) did not exhibit growth defects in vitro. However, ΔpflB formed taller and more diffuse biofilm compared to the wild-type strain as revealed by confocal microscopy. In a mouse model of PJI, the bacterial burden was significantly reduced with ΔpflB during later stages of infection, which coincided with decreased G-MDSC influx and increased neutrophil recruitment, and ΔpflB was more susceptible to macrophage killing. Although formate was significantly reduced in the soft tissue surrounding the joint of ΔpflB-infected mice levels were increased in the femur, suggesting that host-derived formate may also influence bacterial survival. This was supported by the finding that a ΔpflBΔfdh strain defective in formate production and catabolism displayed a similar phenotype to ΔpflB. These results revealed that S. aureus formate metabolism is important for promoting biofilm persistence.

INTRODUCTION

Staphylococcus aureus is a leading cause of prosthetic joint infection (PJI) (1–4) that is typified by biofilm development (5, 6). Biofilms are bacterial communities surrounded by a self-produced matrix consisting of extracellular DNA, carbohydrates, and proteins (7). The biofilm state affords tolerance to antibiotics (8, 9), partly due to nutrient and oxygen gradients, which cause metabolic diversity within the biofilm (10, 11). Specifically, the bacterial populations deeper within the biofilm experience lower oxygen levels, resulting in enhanced fermentative metabolism with increased secretion of lactate, acetate, and formate (12–14).

In addition to antibiotic tolerance, S. aureus biofilm biases the immune response toward an anti-inflammatory state to allow for PJI persistence within an immunocompetent host (15–20). This is achieved by the preferential recruitment of granulocytic myeloid-derived suppressor cells (G-MDSCs) to the site of infection (21–24). G-MDSCs are a pathologically activated, immature neutrophil population that suppress monocyte/macrophage proinflammatory activity, in part, by the production of the anti-inflammatory cytokine interleukin (IL)-10 (17, 25, 26). Other studies have identified additional modes of biofilm-leukocyte interaction and how this shapes infection chronicity (15, 27–30).

We have previously shown that S. aureus-derived lactate is involved in biofilm-leukocyte crosstalk by enhancing Il10 transcription during PJI (16). Because of this, we became interested in other biofilm-derived products that influence host immunity. In the current study, RNA-seq was performed on S. aureus biofilm cocultured with G-MDSCs or macrophages in vitro. Several genes involved in fermentative metabolism were significantly upregulated in S. aureus biofilm following leukocyte coculture, including those involved in lactate and formate metabolism. Because lactate influences biofilm-leukocyte crosstalk (16), these results led us to investigate whether S. aureus biofilm-derived formate may represent another bacterial metabolite that contributes to the pathogenesis of S. aureus infection.

S. aureus formate acetyltransferase (PflB) and its activating enzyme (PflA) drive the conversion of pyruvate to formate. The resulting formate can then be oxidized to CO₂, along with the reduction of NAD⁺ to NADH by formate dehydrogenase (Fdh). In S. aureus, all three of these genes (pflB, pflA, and fdh) are highly expressed during biofilm development compared to planktonic growth (31). Pfl has been shown to generate formate for protein and purine synthesis under anoxic conditions, suggesting a critical role in the anaerobic/microaerobic layers of biofilm (32). In Streptococcus pneumoniae, Pfl is also critical for formate production during mixed-acid fermentation, and Pfl inactivation attenuated the onset of bacteremia following intranasal infection in a mouse model (33, 34). However, the role of formate metabolism during S. aureus biofilm infection and how these pathways contribute to biofilm-immune crosstalk has not yet been investigated. Here, we demonstrated that S. aureus formate metabolism was important for intracellular survival in macrophages and establishing chronic infection during PJI.

RESULTS

Fermentative genes, including those involved in formate metabolism, are upregulated in S. aureus biofilm following leukocyte coculture.

To investigate biofilm-leukocyte crosstalk, we performed RNA-seq to evaluate how the S. aureus biofilm transcriptome was altered following exposure to G-MDSCs or macrophages in vitro. Interestingly, each immune cell type elicited a distinct transcriptional response. G-MDSCs significantly altered the expression of 118 S. aureus genes compared to biofilm only, whereas only 8 were affected by macrophages (Data Set S1). Many of the genes that were significantly upregulated in S. aureus biofilm following G-MDSC exposure were involved in mixed-acid fermentation (Fig. 1 and Table 1; Data Set S1), including alcohol dehydrogenase (adh) and l-lactate dehydrogenase (ldh) (3.36-and 2.85-fold increase; 0.043 and 0.039 false discovery rate [FDR]-adjusted P value, respectively), the latter of which we recently showed was important for programming epigenetic changes in leukocytes to promote IL-10 production and S. aureus biofilm persistence (16). Genes involved in formate metabolism, including pflA, pflB, and fdh (3.36-, 2.88-, and 2.22-fold change; 0.049, 0.050, and 0.050 FDR-adjusted P value, respectively) were also elevated following G-MDSC coculture (Fig. 1B, Table 1, and Data Set S1). However, no increases in extracellular formate were observed at this early time point (Fig. S1 in Supplemental File 1). This was not unexpected given the short coculture interval required to maintain G-MDSC viability because metabolic alterations are expected to lag transcriptional changes that require translation and enzymatic activity to result in altered metabolite levels. Interestingly, genes involved in formate metabolism were not significantly altered in planktonic S. aureus following G-MDSC exposure (Fig. S2 in Supplemental File 1), suggesting this may be a biofilm-specific response. Together, these results demonstrated substantial crosstalk between S. aureus biofilm and G-MDSCs that promoted a fermentative signature, and that formate metabolism may represent an important response in S. aureus to facilitate biofilm persistence in the face of immune pressure.

FIG 1.

G-MDSCs elicit transcriptional responses in S. aureus biofilm. S. aureus biofilm was cocultured with mouse bone marrow-derived G-MDSCs in vitro for 2 h, whereupon RNA-sequencing was performed. Volcano plot showing log₂ (fold change) and −log₁₀ (P value) for the comparison of biofilm only versus biofilm + G-MDSCs. Differentially expressed genes (false discovery rate [FDR]-adjusted P < 0.05) are depicted in red (enriched in biofilm + G-MDSCs) or blue (enriched in biofilm only). FDR-adjusted P values for pflB and fdh are 0.0502. A volcano plot for biofilm only versus biofilm + macrophages is not shown because none of the FDR-adjusted P values reached significance.

TABLE 1.

Fermentative pathways are upregulated by S. aureus biofilm following G-MDSC coculture

Gene	Full name	Fold-change	P value (FDR-adjusted)
rpmC	50S ribosomal protein L29	3.65	0.039
nirD	Nitrite reductase (NAD[P]H), small subunit	3.55	0.048
adh	Alcohol dehydrogenase	3.36	0.043
pflA	Pyruvate formate-lyase activating enzyme	3.36	0.049
gpmI	NA	3.11	0.043
SAUSA300_1656	Universal stress protein family	3.05	0.046
SAUSA300_0113	Immunoglobulin G binding protein A precursor	3.04	0.039
thrB	Homoserine kinase	2.90	0.039
pflB	Formate acetyltransferase	2.88	0.050
ldh	l-lactate dehydrogenase	2.85	0.039

PflB influences S. aureus biofilm structure.

We first examined the importance of S. aureus formate metabolism under planktonic and biofilm conditions. The growth of a S. aureus formate acetyltransferase mutant (ΔpflB) obtained from the Nebraska Transposon Mutant Library (NTML) (35) was assessed in tryptic soy broth (TSB) as well as RPMI 1640 + 10% fetal bovine serum (FBS) to better model host conditions. PflB loss did not affect planktonic growth in either medium formulation (Fig. 2A and B). Next, biofilm formation was examined because fermentative metabolism is important in hypoxic/anoxic domains within S. aureus biofilm (12, 36). ΔpflB exhibited a slight defect during the first 24 h of biofilm growth but recovered to achieve an equivalent bacterial burden as WT biofilm for the remaining culture period (Fig. 2C). As expected, formate production was minimal in ΔpflB compared to WT biofilm (Fig. 2D). These results indicated that formate biosynthesis did not dramatically affect S. aureus planktonic or biofilm growth, but the metabolite was produced at high levels by WT biofilm.

FIG 2.

S. aureus PflB inactivation does not affect planktonic or biofilm growth. The growth of S. aureus WT or ΔpflB was measured in liquid culture by (A) OD₆₀₀ in tryptic soy broth (TSB) or (B) RPMI 1640 + 10% FBS (mean ± SD of one representative experiment, n = 29 from 5 independent experiments). (C) CFU counts of in vitro biofilms over 4 days (n = 10 from 3 independent experiments; **, P < 0.01; two-way ANOVA with Sidak’s multiple comparisons analysis). (D) Formate was measured from the supernatant of S. aureus WT or ΔpflB biofilm over the 4-day culture period (n = 3 from one experiment).

Despite the transient decrease in bacterial abundance during acute biofilm formation, ΔpflB displayed a more diffuse biofilm structure (Fig. 3A) with significantly increased average thickness compared to WT (Fig. 3B). Supplementation of the growth medium with 2.5 mM formate (equal to levels produced by WT biofilm [Fig. 2D]) did not significantly reduce the thickness of ΔpflB biofilms. However, the aberrant biofilm morphology was fully rescued by chromosomal complementation of ΔpflB (Fig. 3B), providing direct evidence that formate metabolism was important for dictating biofilm architecture.

FIG 3.

Formate metabolism influences S. aureus biofilm structure. (A) Biofilms were grown for 4 days with GFP-expressing WT, ΔpflB, or a ΔpflB complemented strain (Comp) and imaged by confocal laser scanning microscopy. ΔpflB biofilm was supplemented with 2.5 mM sodium formate at all stages of growth where indicated. Representative three-dimensional (3-D) and orthogonal images are shown. (B) The average biofilm thickness was calculated by Comstat 2 analysis (n = 6 to 14 biological replicates with 2 to 3 Z-stacks acquired from each biofilm from 3 to 6 independent experiments; *, P < 0.05; ***, P < 0.001; one-way ANOVA with Tukey’s multiple-comparison test).

ΔpflB promotes S. aureus survival during chronic prosthetic joint infection (PJI).

Due to the altered structure of ΔpflB biofilm, we sought to determine whether this translated into a diminished ability to sustain a chronic infection. Utilizing a well-established mouse model of PJI, the bacterial burden was significantly reduced as early as day 7 in the femur of ΔpflB-infected mice that persisted throughout the 28-day time course (Fig. 4A). As the infection progressed, ΔpflB abundance became significantly lower than WT in the implant-associated tissue and joint by days 14 and 28 postinfection, respectively (Fig. 4A). Decreased bacterial burden in ΔpflB-infected mice coincided with reduced G-MDSC influx concomitant with increased neutrophil infiltrates, which was not significant until day 28 after infection (Fig. 4B).

FIG 4.

S. aureus ΔpflB is critical for bacterial persistence and leukocyte influx during chronic prosthetic joint infection. C57BL/6NCrl mice were infected with 10³ CFU of S. aureus WT or ΔpflB and sacrificed at the indicated time points postinfection. (A) Bacterial burden was measured in the implant-associated tissue, joint, and femur and (B) leukocyte infiltrates were quantified in the implant-associated tissue by flow cytometry. Significant differences are denoted by asterisks (n = 14 to 24 mice/group from 3 to 5 independent experiments; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; two-way ANOVA with Sidak’s multiple comparisons analysis). The absence of asterisks indicates a lack of statistical significance.

Based on the reduction of anti-inflammatory G-MDSCs during PJI with ΔpflB, we next examined whether this coincided with a shift in the inflammatory milieu to facilitate biofilm clearance. Even in the context of lower bacterial burden in ΔpflB-infected mice, a heightened growth factor/inflammatory response was observed in the implant-associated tissue at day 14 postinfection, including significant increases in macrophage colony-stimulating factor (M-CSF) and macrophage inflammatory protein-1α (CCL3/MIP-1α) (Fig. 5). Other inflammatory mediators, including interleukin-1 beta (IL-1β), monocyte chemoattractant protein-1 (CCL2/MCP-1), and macrophage inflammatory protein-1β (CCL4/MIP-1β) were also elevated upon ΔpflB infection but did not reach statistical significance (Fig. 5, Fig. S3 in Supplemental File 1). Although IL-10 production was also increased in ΔpflB-infected animals, this was not sufficient to impact bacterial clearance. The differences in most inflammatory mediators elicited by ΔpflB were complementable (Fig. 5).

FIG 5.

S. aureus ΔpflB elicits a heightened proinflammatory response in vivo. C57BL/6NCrl mice were infected with 10³ CFU of S. aureus WT, ΔpflB, or a ΔpflB complemented strain (Comp) and implant-associated tissue was collected at day 14 postinfection, whereupon M-CSF, IL-1β, CCL2, CCL3, CCL4, and IL-10 were quantified using a multianalyte bead array. Results were normalized to the amount of protein in each sample to correct for differences in tissue sampling size (n = 4 to 5 from one experiment; *, P < 0.05; one-way ANOVA with Tukey’s multiple comparisons analysis). The absence of asterisks indicates a lack of statistical significance.

S. aureus pflB is critical for intracellular survival in macrophages.

The heightened proinflammatory response and increased neutrophil influx in ΔpflB-infected mice implied better immune-mediated killing of ΔpflB during PJI. In macrophages, intracellular survival of ΔpflB and WT S. aureus was similar at early time points. However, ΔpflB was significantly more sensitive to macrophage bactericidal activity by 24 h (Fig. 6). The late-stage susceptibility of ΔpflB to macrophage killing was reminiscent of the delayed clearance of ΔpflB biofilm in vivo (Fig. 4). The impaired survival of ΔpflB in macrophages was complementable (Fig. 6), demonstrating the importance of formate metabolism for S. aureus intracellular survival. Interestingly, neutrophils killed ΔpflB and WT S. aureus to similar extents (Fig. S4 in Supplemental File 1). However, the increased abundance of neutrophils infiltrating ΔpflB-infected mice could explain the reduction in biofilm burden together with increased macrophage bactericidal activity. We next examined whether formate directly suppressed macrophage proinflammatory activity by treating cells with S. aureus-relevant pathogen-associated molecular patterns (PAMPs), either a synthetic lipopeptide (Pam3CSK4) or S. aureus peptidoglycan (PGN), in the presence of exogenous sodium formate. Sodium formate did not affect cytokine production (Fig. S5 in Supplemental File 1), suggesting that the increased cytokine expression observed in vivo with ΔpflB may result from other formate-dependent pathways.

FIG 6.

PflB is important for S. aureus intracellular survival in macrophages. Mouse bone marrow-derived macrophages were infected with WT, ΔpflB, or a pflB complemented strain (Comp) at an MOI of 10:1 (bacteria:macrophage) for 1 h. Extracellular bacteria were then killed with a high dose of gentamicin (100 μg/mL) for 30 min, which was then replaced with low-dose gentamicin (1 μg/mL) for the remainder of the assay. At the indicated time points following high dose gentamicin treatment, macrophages were washed, lysed with sterile water, and viable intracellular bacteria were quantified (one representative experiment shown; n = 24/time point from 6 independent experiments; *, P < 0.05; ****, P < 0.0001; two-way ANOVA with Tukey’s multiple-comparison test). The absence of asterisks indicates a lack of statistical significance.

Host-derived formate influences biofilm persistence.

Although formate was significantly lower in the implant-associated tissue of ΔpflB compared to WT mice during PJI, surprisingly, high levels of formate were still detected in the femur of ΔpflB-infected animals (Fig. 7A). ΔpflB did not produce formate in vitro (Fig. 2D), which suggested host-derived formate as a source. Because ΔpflB could still metabolize exogenous formate via formate dehydrogenase (Fdh), we created a ΔpflBΔfdh strain that could not synthesize as well as catabolize host-derived formate. An Δfdh single mutant was not examined because it would still be capable of formate production. Like ΔpflB, the ΔpflBΔfdh strain had no significant changes in planktonic or biofilm growth. However, ΔpflBΔfdh also resulted in a significantly thicker and more diffuse biofilm architecture compared to WT (Fig. S6 in Supplemental File 1). Bacterial burdens were again reduced with ΔpflB in all locations but interestingly, S. aureus burden was only significantly decreased in the femur with ΔpflBΔfdh (Fig. 7B), in agreement with the increased levels of formate detected in the femur of ΔpflB-infected mice (Fig. 7A). This suggested a unique action of formate for S. aureus survival in bone. In summary, our results indicated that S. aureus formate metabolism was important for biofilm structure and immune modulation, and utilization of host-derived formate likely contributed to biofilm persistence in bone.

FIG 7.

Host-derived formate influences S. aureus persistence in the femur. (A) Formate levels were quantified in implant-associated tissue and femur homogenates at day 28 postinfection. (B) Bacterial burden in the implant-associated tissue, joint, and femur of mice infected with S. aureus WT, ΔpflB, ΔpflB complemented strain (Comp), or ΔpflBΔfdh at day 28 postinfection (n = 10 mice/group from one experiment; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; (A) unpaired Student's t test or (B) one-way ANOVA with Tukey’s multiple-comparison test). The absence of asterisks indicates a lack of statistical significance.

DISCUSSION

RNA-seq was used in this study to identify transcriptional changes in S. aureus biofilm in response to G-MDSCs or macrophages as a model to identify how biofilm adapts to immune pressure. Although we previously examined transcriptional changes in S. aureus biofilm following macrophage exposure (37), this was performed using Affymetrix gene arrays that did not interrogate transcriptional responses in an unbiased manner, and G-MDSCs were not examined. Here, we reported that several fermentative genes were significantly increased in S. aureus biofilm following G-MDSC exposure, including those involved in ethanol, formate, and lactate metabolism. We recently reported a critical role for S. aureus biofilm-derived lactate in promoting leukocyte IL-10 production (16), and a previous study demonstrated that the genes involved in formate metabolism (pflA, pflB, and fdh) are more highly expressed in S. aureus biofilm compared to planktonic cells (31), which was confirmed by proteomic analysis (38). Our results advanced these findings by demonstrating that formate metabolism was a target of biofilm-leukocyte crosstalk because the expression of formate genes was increased in biofilm following G-MDSC exposure.

Interestingly, while G-MDSCs induced a robust transcriptional response in S. aureus biofilm in this study, macrophages had little effect. We previously reported that macrophages globally suppressed S. aureus biofilm transcription following a 1 h coculture period (37). The differences in macrophage effects may be explained by the increased coculture time used in the current study (2 h), which was employed to allow for direct comparisons with G-MDSC-induced changes in the biofilm transcriptome. Macrophages are exquisitely sensitive to S. aureus biofilm and the additional 1 h coculture period likely resulted in fewer viable macrophages to modulate biofilm responses. This is supported by our previous finding that incubation of macrophages with S. aureus biofilm (24 h) resulted in few transcriptional changes (37). Nevertheless, RNA-seq detected increased pflB and pflA expression in S. aureus biofilm following macrophage exposure (1.5-fold), although this did not reach statistical significance (Data Set S1). It is likely that S. aureus shifts to fermentative metabolism as an adaptation to G-MDSC exposure rather than G-MDSCs consuming oxygen because a similar fermentative profile would have been expected to occur in response to macrophages that also utilize oxygen for cellular respiration. Activated G-MDSCs also produce large amounts of ROS by the action of NOX2 (39), which could induce S. aureus to favor formate metabolism to maintain redox balance during this oxidative stress.

Using a mouse model of S. aureus PJI, ΔpflB elicited decreased G-MDSC recruitment over time, whereas monocyte infiltrates remained low. This supports the interplay between formate metabolism in S. aureus biofilm and G-MDSCs, which is reminiscent of the selective effects of G-MDSCs but not macrophages on formate gene expression in S. aureus biofilm by RNA-seq. A prior study showed that S. aureus ΔpflB produced significantly fewer formylated peptides than WT bacteria under anaerobic growth conditions (32). Formylated peptides induce leukocyte chemotaxis upon recognition by the formyl-peptide receptor 1 (FPR1) (40), which is most highly expressed on mature neutrophils (41). Because current literature considers G-MDSCs to represent pathologically activated neutrophils (42), the reduction in formate in tissues of ΔpflB-infected mice could result in less formylated peptide production to account for decreased G-MDSC recruitment. However, a corresponding increase in neutrophil influx was also observed, which makes this scenario less likely. Instead, the decrease in G-MDSCs concomitant with enhanced neutrophil recruitment with ΔpflB may be explained by reduced bacterial burden, a relationship that we have observed in other immune modulation studies in the PJI model (16–18, 43–45). The chemokines responsible for G-MDSC recruitment in the context of PJI remain to be identified. However, based on studies in cancer models, possible candidates include CXCL1, CXCL2, and/or CCL5 (46, 47).

One interesting observation from this study was the altered biofilm structure of ΔpflB compared to WT because both had equivalent numbers of viable bacteria and no defects in planktonic growth were observed. ΔpflB biofilm had a more diffuse morphology similar to our prior report with ΔatpA biofilm (43), supporting a link between S. aureus metabolism and biofilm structure. The mechanism responsible for the altered architecture of ΔpflB biofilm is unknown, but possibilities include changes in matrix deposition or bacterial lysis that may be affected by impairments in one-carbon metabolism. Alternatively, the loss of formate production may lead to intrinsic changes in S. aureus redox balance that impacts biofilm structure. For example, aconitase is sensitive to redox inactivation (48, 49). Therefore, changes in S. aureus TCA cycle activity may be affected leading to biofilm stress. The diffuse structure of ΔpflB biofilm suggests that a subpopulation of bacteria may adopt a planktonic phenotype that would render it more susceptible to leukocyte phagocytosis and killing (20). This was supported by our finding that macrophages were better able to kill ΔpflB, and the less compact structure of ΔpflB biofilm may permit increased leukocyte invasion to diminish the biofilm burden. Indeed, mice with ΔpflB infection displayed significant reductions in biofilm abundance in the PJI model concomitant with increased neutrophil recruitment. Besides the intrinsic effects of formate metabolism on S. aureus biofilm organization, it is important to recognize the potential role of bacterial-derived formate on leukocyte metabolism because metabolism is intimately linked to inflammatory function (50). Further studies examining how S. aureus formate impacts leukocyte metabolism are warranted.

Residual formate was detected in ΔpflB-infected mice that was most prominent in the femur. Because ΔpflB is unable to produce formate, this suggested the contribution of host-derived formate to the total formate pool during S. aureus infection. Formate in mammalian cells arises from many sources such as serine, glycine, and methanol for use in one-carbon metabolism (51). Because of the essential nature and multiple routes for formate biosynthesis, it was not feasible to inhibit host-derived formate to specifically interrogate its role during S. aureus PJI. However, we generated a ΔpflBΔfdh strain to determine whether consumption of host-derived formate was important for biofilm persistence. Loss of both PflB and Fdh reversed the lower biofilm burden observed with ΔpflB in the tissue and joint of infected mice, but not in the femur where ΔpflBΔfdh and ΔpflB remained lower than WT bacteria. The femur of ΔpflB-infected mice contained higher levels of formate than the tissue, demonstrating a spatial niche for host formate consumption on biofilm virulence. Bone has a low oxygen tension under homeostatic conditions, which is exacerbated during S. aureus infection (52), supporting the importance of fermentative metabolism in bacterial survival. Collectively, our results demonstrated a previously unappreciated role for S. aureus biofilm formate metabolism in influencing the host immune response.

MATERIALS AND METHODS

Bacterial strains.

The WT S. aureus LAC-13C strain used in this study was derived from a USA300 clinical isolate recovered from a skin and soft tissue infection (53), which was cured of a plasmid conferring resistance to erythromycin (erm) (35, 54). S. aureus transposon mutants with erm resistance markers ΔpflB::erm and Δfdh::erm were obtained from the NTML (35). Mutants were confirmed by PCR using pflB_F and pflB_R primers flanking the transposon insertion sites (Table S1 in Supplemental File 1) and then moved into the S. aureus LAC-13C background using φ11 transduction. To generate the S. aureus ΔpflBΔfdh strain, the erm cassette in ΔpflB::erm was swapped for kanamycin (kan), as previously described (55). ΔpflB::kan was then moved into the Δfdh::erm strain using φ11 transduction. To visualize biofilm development, strains were transduced with a GFP reporter plasmid (pCM29) (56), where chloramphenicol (10 μg/mL) was used for selection and to ensure plasmid maintenance during overnight culture and biofilm growth.

Chromosomal complementation of ΔpflB was performed by insertion of the pflBA operon into a neutral locus using the pJC1111 suicide vector, which integrates at S. aureus pathogenicity island 1 (SapI1), as described previously (57, 58). Briefly, the pflBA operon was amplified using pflB_compl_fwd and pflB_compl_rev primers, and the shuttle vector was amplified using pJC1111_fwd and pJC1111_rev primers (Table S1 in Supplemental File 1). Fragments were assembled using the NEBuilder HiFi DNA assembly cloning kit (New England Biolabs) to create the pSC45 complementation plasmid, which was electroporated into E. coli E10B and then electroporated into S. aureus RN4220 and transduced into ΔpflB using φ11.

S. aureus growth.

Bacterial glycerol stocks were freshly streaked on Trypticase soy agar (TSA) enriched with 5% sheep blood before experiments. For in vitro studies, a single colony was inoculated into TSB or biofilm medium (RPMI 1640 supplemented with 10% heat-inactivated FBS, 1% l-glutamine, and 1% HEPES) and grown overnight (16 to 18 h) at 37°C, 250 rpm. To quantify planktonic growth under aerobic conditions, overnight cultures were diluted to an optical density at 600 nm (OD₆₀₀) of 0.05 in a 96-well clear, flat-bottom plate and OD readings were taken every 30 min using an Infinite Pro 200 (TECAN). For biofilm growth, 96-well plates or 8-well chamber slides were coated with 20% human plasma in carbonate-bicarbonate buffer overnight at 4°C. Overnight bacterial cultures were diluted to an OD₆₀₀ of 0.05 in the plasma-coated plates or slides and grown under static conditions at 37°C in room air for 4 days with the medium replenished every 24 h. For metabolic complementation studies, the biofilm medium was supplemented with 2.5 mM sodium formate during the overnight culture and daily medium replenishment. Biofilms grown in chamber slides were visualized using confocal laser scanning microscopy (Zeiss 710) with a 40× oil lens. Z-stack images were acquired (1 μm sections) and used to construct 3-D images, which were analyzed by Comstat2 (Image J) (59–61).

Formate quantification.

Formate production by WT or ΔpflB biofilm was quantified during growth in 96-well plates in vitro. Supernatants were collected daily over 4 days and frozen at −80°C until analysis. Implant-associated tissue and femur homogenates from in vivo studies were frozen at −80°C. Formate concentrations were determined using a Formate Colorimetric assay kit (BioVision) following the manufacturer’s instructions.

RNA-Sequencing.

(i) Cell recovery and RNA isolation. Macrophages and G-MDSCs were expanded from the bone marrow of WT C57BL/6NCrl mice for 7 or 4 days, respectively, using the supernatant from L929 cells as a source of M-CSF for macrophages or recombinant mouse G-CSF, GM-CSF, and IL-6 for G-MDSCs (45, 62). Macrophages or G-MDSCs were then cocultured with 4-day-old S. aureus biofilm for 2 h, whereupon RNA was isolated from each condition using a bead-beater and TRIzol^TM Reagent (Invitrogen) according to the manufacturer’s instructions. A total of two biofilms only, three biofilm-macrophage coculture, and three biofilm-G-MDSC coculture biological replicates were collected for RNA-sequencing.

(ii) NGS library preparation. RNA sequencing libraries were prepared beginning with 250 ng of total RNA using a ScriptSeq Complete kit (Bacterial) – Low Input protocol (catalog no. SCL6B, Epicentre Technologies, Madison, WI). The protocol was performed as recommended by the manufacturer with the following modification in step 3.B.1. To remove both eukaryotic and prokaryotic rRNA from the total preparation, RiboZero human/mouse/rat removal solution was mixed with the bacterial rRNA removal solution. In addition to the above, AMPure bead purification of the rRNA-depleted sample was used in step 3.D.2 of the protocol.

(iii) Sequencing. The RNAseq libraries were denatured with 0.2N NaOH and 1.2 pM of the libraries were sequenced on a NextSeq500 instrument (Illumina). The libraries were subjected to 75 bp paired-end sequencing at a depth of 20 million paired reads per sample.

(iv) Differential expression. The Illumina reads were checked for quality and subsequently trimmed. The resulting paired fastq files were sorted and mapped against the EBI Ensembl USA300_FRP3757 assembly cDNA library using Salmon (63) with default parameters. DESeq2 (64) was then used to determine differential expression estimates and controls for false discovery rate (FDR). Genes with a P < 0.05 were considered differentially expressed. The complete RNA-seq data set has been deposited in the GEO database (accession number GSE213381).

Gene expression analysis of planktonic S. aureus.

Overnight broth cultures of S. aureus were diluted to 10⁹ CFU/well in a 12-well plate (equivalent to the number of organisms in biofilms grown in a 12-well plate) and cocultured with 5x10⁵ mouse bone marrow-derived G-MDSCs for 2 h. RNAprotect Bacteria Reagent (Qiagen) was then added to each sample and S. aureus was lysed using a combination of lysostaphin (50 μg/mL) and bead beating. RNA was immediately isolated using a RNeasy minikit (Qiagen). cDNA was synthesized using an iScript cDNA Synthesis kit (Bio-Rad) and qPCR was performed using iTaq Universal SYBR green Supermix for fdh, pflA, pflB, and gyrB with primer sets provided in Table S1 in Supplemental File 1. Gene expression levels were normalized to gyrB and are presented as the fold-induction (2^-ΔΔCt) in planktonic bacteria + G-MDSCs relative to bacteria alone.

Mice.

C57BL/6NCrl mice (Research Resource Identifier [RRID], IMSR_CRL:27) were bred in-house at the University of Nebraska Medical Center (UNMC). Mice of the same sex were randomized into standard density ventilated microisolator cages upon weaning (n = 5 mice per cage) in a restricted-access biosafety level 2 (BSL2) room maintained at 21°C under a 12-h light:12-h dark cycle with ad libitum access to water (Hydropac; Lab Products, Seaford, DE) and Teklad rodent chow (Harlan, Indianapolis, IN) with nestlets provided for enrichment. This study was conducted in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was approved by the UNMC Institutional Animal Care and Use Committee (18-013-03).

Mouse model of S. aureus prosthetic joint infection (PJI).

To determine the role of formate metabolism during S. aureus PJI, a mouse model was used as previously described (65). Briefly, equal numbers of male and female C57BL/6NCrl mice (8 to 10-week-old) were anesthetized with a ketamine/xylazine cocktail, whereupon a parapatellar arthrotomy was performed to expose the knee joint. A 26-gauge needle was used to bore a hole in the femoral intercondylar notch where an orthopedic-grade K-wire (0.6 mm diameter, nitinol [nickel-titanium]; Custom Wire Technologies) was inserted. Approximately 1 mm of the wire was left protruding into the joint space, whereupon 10³ CFU of S. aureus WT, ΔpflB, or ΔpflB/Δfdh in PBS was used to inoculate the implant tip. The surgical site was sutured closed and Buprenex slow release (Buprenex SR) was administered for pain relief. Mice were closely monitored following surgery and until sacrifice when they exhibited normal ambulation and no discernible pain behaviors.

Quantification of leukocyte infiltrates and biofilm burden.

The implant and implant-associated knee tissue, joint, and femur were recovered from mice at days 7, 14, or 28 postinfection. Samples were weighed, homogenized, and serial 10-fold dilutions were performed and plated on TSA with 5% sheep blood to quantify bacterial burdens expressed as log-transformed CFU per gram of tissue. To quantify leukocyte infiltrates, homogenized implant-associated tissue was passed through a 70 μm cell strainer and red blood cells were lysed using a red blood cell lysis buffer (BioLegend). Cells were then stained with CD11b-FITC, CD45-APC, Ly6G-PE, Ly6C-PerCP-Cy5.5, and F4/80-PE-Cy7 antibodies (BioLegend and BD Biosciences) as well as a Live/Dead Fixable Blue Dead Cell Stain kit (Invitrogen). Stained cells were acquired on an LSR II Flow Cytometer System (BD Biosciences) and analyzed using BD FACS DIVA software. Live CD45⁺ singlets were then further identified as G-MDSCs (CD11b^highLy6G⁺Ly6C⁺F4/80⁻), neutrophils (CD11b^lowLy6G⁺Ly6C⁺F4/80⁻), and monocytes (Ly6G-Ly6C⁺F4/80⁻) as previously described (23).

Multianalyte microbead array.

To quantify inflammatory mediator expression during S. aureus PJI, homogenates prepared from the soft tissue surrounding the infected joint were analyzed using a custom 19-plex Milliplex MAP mouse cytokine/chemokine magnetic bead panel (MilliporeSigma, Billerica, MA). The analytes in this panel included: granulocyte colony-stimulating factor (G-CSF), macrophage colony-stimulating factor (M-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), interferon-gamma (IFN-γ), tumor necrosis factor-alpha (TNF-α), interleukin-1 beta (IL-1β), IL-6, IL-10, IL-12p70, IL-17, monocyte chemoattractant protein-2 (CCL2/MCP-2), macrophage inflammatory protein-1 alpha (CCL3/MIP-1α), macrophage inflammatory protein-1 beta (CCL4/MIP-1β), regulated upon activation T cell expressed and secreted (CCL5/RANTES), keratinocyte chemoattractant (CXCL1/KC), macrophage inflammatory protein-2 (CXCL2/MIP-2), monokine induced by IFN-γ (CXCL9/MIG), IFN-γ-induced protein 10 kDa (CXCL10/IP-10), and leukemia inhibitory factor (LIF). Results were normalized to the total protein concentration per sample to adjust for any differences in tissue sampling size. Signals were observed in PJI tissues for all mediators except GM-CSF and IL-12p70, which fell below the limit of detection and are not reported.

Gentamicin protection assay.

Bone marrow-derived macrophages were suspended in biofilm medium and transferred to a 96-well tissue culture-treated plate at a density of 5x10⁴ cells/well and allowed to adhere overnight. The following day, overnight bacterial cultures were diluted to achieve a multiplicity of infection (MOI) of 10:1 and added to macrophages or thioglycolate-elicited peritoneal neutrophils for 1 h to allow for phagocytosis. Next, macrophages and neutrophils were treated with a high dose of gentamicin (100 μg/mL) for 30 min to kill extracellular bacteria, then switched to a low dose of gentamicin (1 μg/mL) for the remainder of the assay to prevent intracellular flux of the antibiotic that can occur with prolonged incubations with high doses (66). At each time point (0, 2, 4, 6, and 24 h), leukocytes were washed with PBS to remove gentamicin and lysed with sterile water. The lysate was then serially diluted and plated on TSA with 5% sheep blood to enumerate intracellular bacterial burden.

Statistics.

Significant differences were determined using a two-way analysis of variance (ANOVA) with Tukey’s multiple-comparison test, one-way ANOVA with Sidak’s multiple-comparison test, or unpaired Student's t test using GraphPad Prism version 9.0.2. P < 0.05 was considered statistically significant.