Heme cross-feeding can augment Staphylococcus aureus and Enterococcus faecalis dual species biofilms

The contribution of biofilms to virulence and as a barrier to treatment is well-established for Staphylococcus aureus and Enterococcus faecalis, both nosocomial pathogens frequently isolated from biofilm-associated infections. Despite frequent co-isolation, their interactions in biofilms have not been well-characterized. We report that in combination, these two species can give rise to augmented biofilms biomass that is dependent on the activation of E. faecalis aerobic respiration. In E. faecalis, respiration requires both exogenous heme to activate the cydAB-encoded heme-dependent cytochrome bd, and the availability of O2. We determined that the ABC transporter encoded by cydDC contributes to heme import. In dual species biofilms, S. aureus provides the heme to activate E. faecalis respiration. S. aureus mutants deficient in heme biosynthesis were unable to augment biofilms whereas heme alone is sufficient to augment E. faecalis mono-species biofilms. Our results demonstrate that S. aureus-derived heme, likely in the form of released hemoproteins, promotes E. faecalis biofilm formation, and that E. faecalis gelatinase activity facilitates heme extraction from hemoproteins. This interspecies interaction and metabolic cross-feeding may explain the frequent co-occurrence of these microbes in biofilm-associated infections.


INTRODUCTION
Biofilms consist of a sessile community of microbes embedded within a matrix of extracellular polymeric substances. Biofilms represent the dominant mode of bacterial life, and biofilm cells exhibit different patterns of behaviour compared to planktonic cells [1]. Characteristically, biofilms confer protection to chemical and physiological stresses [2], sheer forces [3,4], predation [5], and to antibiotic-mediated clearance [6]--rendering biofilm-associated infections difficult to treat. Behaviourally, biofilm-embedded cells are extremely heterogeneous, exhibiting varying levels of metabolic activity (from very active to quiescent) [7,8], transcriptional profiles (depending on the microenvironment) [9,10] and higher frequencies of genetic exchange (via transformation, conjugation and transduction) [11,12].
E. faecalis and S. aureus are both opportunistic pathogens and are among the leading causes of nosocomial infections [43,44]. The biofilm-forming potential of each species is well documented [33,45] and both have been implicated in biofilm-associated infections such as endocarditis [46,47], urinary tract infections [48,49] and chronic wounds [50,51]. Although, E. faecalis and S. aureus are commonly co-isolated in chronic wounds such as diabetic foot ulcers [52], venous leg ulcers [53] and pressure wounds [54], studies of their interactions are largely limited to the transfer of vancomycin resistance genes from E. faecalis to S. aureus in clinical settings [55][56][57]. Therefore, in this study, we explored the molecular interactions between E. faecalis and S. aureus in biofilms.
Aerobic respiration in E. faecalis requires exogenous heme as a cofactor for cytochrome bd [58,59]. Enterococci are unable to synthesize heme due to an absent TCA cycle that prevents the formation of porphyrin precursors. Through a yet unknown importer, heme is taken into the cell and incorporated into cytochrome bd (CydAB), which then converts terminal demethylquinol (DMKH 2 ) to demethylmenaquinone (DMK), consuming O 2 and releasing H 2 O in the process [60]. DMK is reduced by NADH: quinone oxidoreductase back into DMKH 2 consuming NADH. Importantly, Enterococci do not express other membraneembedded electron carriers like ubiquinone or menaquinone [60], but only make DMK which is a modified menaquinone lacking a 2-methyl group [61]. Cytochrome bd is the key respiratory enzyme for E. faecalis and contains two subunits (CydA and CydB) with three cytochromes, b 558 , b 595 and d. The translocation of a proton by cytochrome bd establishes a proton motive force that, when coupled with F0F1-ATP synthase, generates ATP [62,63]. Additionally, cydC and cydD are necessary for cytochrome bd production and have been suggested to be involved in heme transport and/or cytochrome bd assembly [60]. Interestingly, reduction of O 2 by DMK induces extracellular superoxide production, the latter of which was inhibited by exogenous heme [64], while a functional electron transport chain sensitizes E. faecalis to oxidative burst and decreased its survival in human blood [65].
In this study, we show that S. aureus-derived heme is required to activate E. faecalis aerobic respiration, leading to augmented E. faecalis growth and augmented dual-species biofilm production. We speculate that this interspecies cross-feeding of heme, where one species consumes metabolic end-products from another, may affect mixed species infection outcomes in heme-restricted host and environmental niches.

MATERIALS AND METHODS Bacterial strains and growth conditions
Strain, isolate and transposon library details are available in Supplementary Information. Overnight cultures of E. faecalis were grown in Brain Heart Infusion broth (Becton-Dickinson, United States) whereas S. aureus grown in Tryptone Soy Broth (Oxoid, England). Agar Technical Powder No. 3 (Oxoid, England) was used for agar plates. Strains were cultured under static conditions at 37°C. Overnight cultures were spun down and washed once in phosphate-buffered saline prior to normalization. An OD 600 of 1.3 and 3.0 for E. faecalis and S. aureus respectively gave about 1 × 10 9 CFU/ml. MRSA Select II agar (Bio-Rad Laboratories, United States) was used to select S. aureus USA300LAC from mixed-species cultures whereas rifampicin (25 µg/ml) was added to BHI agar to select for OG1RF. Rifampicin (Sigma-Aldrich, United States) was dissolved in methanol to make a stock of 25 mg/ml and stored at −20°C. Hemin (Sigma-Aldrich, United States) was dissolved in DMSO to make a stock of 25 mg/ml whereas human hemoglobin (Sigma-Aldrich) was dissolved in dH 2 0 and filter sterilized to make stock of 10 mg/ml. For hemin and hemoglobin supplementation, a final concentration of 25 μg/ml and 10 μg/ml respectively were used unless otherwise stated.

Biofilm assays
Normalized cultures of E. faecalis and S. aureus were mixed (with equal CFU) in the ratio of 1:1 for dual-species biofilms and 8 µl of single-or dualspecies cultures were inoculated in 200 µl of Tryptone Soy Broth in a 96well flat transparent plates (Thermo Fisher Scientific, United States) and incubated under static conditions at 37°C for five days unless otherwise stated. For anoxic experiments, plates were instead incubated in an anaerobic jar (Merck, Singapore) with a gas pack (Becton Dickinson, United States) and incubated at 37°C for five days. Details on Crystal Violet staining and CFU determination is available in Supplementary Information. For biofilm oxygen consumption rate (OCR) assays, 5-day biofilms were grown directly in Seahorse XFe96 FluxPak 96-well plates, with 80 µl of inoculated media added per well. Planktonic cells were removed with three washes of 80 µl PBS using a 96-channel pipettor to prevent crosscontamination and the remaining biofilms were resuspended in 80 µl of PBS. After three baseline measurements were taken, 30 µl of fresh TSB was injected into the wells and OCR measured for 1 h using standard parameters.

Growth kinetics
Normalized cells were diluted 100× in PBS and 8 µl of diluted cells were added to 200 µl TSB in a 96-well plate. The plate was incubated in Tecan Infinite M200 PRO Spectrophotometer at 37°C for 20 h and absorbance (600 nm) recorded for every 15 min after shaking the plate for 3 s.

Transposon library screen and transposon mutants
The E. faecalis Transposon Library was provided by Gary M. Dunny [66] and S. aureus transposon mutants were from Nebraska Transposon Library [67] provided by the Network on Antimicrobial Resistance in Staphylococcus aureus (NARSA) and distributed by BEI resources, NIAID. Details on the screening of the E. faecalis transposon library [66,68] are available in Supplementary Information. The primary screen of dual-species biofilms was performed by adding 5 µl of transposon mutants grown overnight and 3 µl of USA300LAC overnight cultures to 200 µl TSB in the 96-well plate. Each plate had controls of parental strains of OG1RF + USA300LAC. Plates were incubated at 37°C for five days in a humidified incubator and biofilms quantified by crystal violet. E. faecalis transposon mutants that showed reduced staining were selected as hits. Confirmation of primary screen hits was done via a secondary screen (three biological replicates) and secondary screen hits that consistently showed reduced dual-species biofilms were subject to further validation. Transposon mutants without defects in growth kinetics and single-species biofilms, but with reduced dual-species biofilms, were shortlisted as validated hits.
LC-MS for heme quantification from cell pellet E. faecalis strains (OG1RF, cydA::Tn, cydD::Tn and cydC::Tn) were grown overnight in 20 ml of TSB with or without supplementation of 5 µg/ml hemin. Cell pellets were washed twice with PBS before being normalized by OD600. Cell pellets were mixed with 100 µl of lysozyme (10 mg/ml), incubated in a water bath at 37°C for 1 h and then stored at −80°C until LC-MS quantification.
LC-MS details are available in Supplementary Information.

Gelatinase activity assay
Gelatin tubes were made by combining 3 g of gelatin and 3.7 g of BHI powder with 100 ml water that was autoclaved before 3 ml was added into 15 ml tubes. E. faecalis colonies stabbed into the gelatin tubes prior to overnight incubation at 37°C. Tubes were then refrigerated at 4°C for an additional 30 min before the tubes were gently tilted to observe for liquification of gelatin due to gelatinase activity [69]. Any liquification was indicative of positive gelatinase activity. Parental strain OG1RF and OG1RFΔgelE served as positive and negative controls respectively, and assays were repeated four times to confirm gelatinase activity.

Statistics
Statistical analyses used GraphPad Prism 7 software (version 7.03) (GraphPad, United States). The data was analyzed through unpaired students T-test or one-way analysis of variance (ANOVA) with Tukey's posthoc testing for multiple inter-group comparison, or with Dunnett's posthoc testing if comparing to a specific control group. Significance for p < 0.05 are reported.

Ethics
All procedures were performed in accordance with Nanyang Technological University Research Integrity Policy. Anonymized and pure bacterial strains used in this study did not require IRB approval.

RESULTS
S. aureus augments E. faecalis growth within biofilms and overall biomass accumulation Since S. aureus and E. faecalis are frequently co-isolated during biofilm-associated infections [55], we first examined the consequence of dual species growth on biofilm biomass accumulation. Using a modified polystyrene microtiter assay followed by crystal violet (CV) staining [70], we observed that E. faecalis strain OG1RF (a rifampicin and fusidic acid resistant derivative of a human oral isolate OG1, commonly used in molecular manipulation and virulence studies and used throughout this study) biofilm biomass peaked at day 1 and plateaued for the remaining four days, whereas S. aureus (strain USA300LAC) formed very poor biofilms under these experimental conditions (Fig. 1A). By contrast, following inoculation of E. faecalis together with S. aureus at a ratio of 1:1, dual-species biofilms showed significantly greater biomass from day 1 and peaked at day 4. Since the greatest difference in dual-species biomass accumulation occurred at day 4 and 5, we chose day 5 for subsequent assays.
To determine if increased bacterial growth within dual species biofilm contributed to the augmented biomass, biofilms were manually disrupted and CFU determined for both single and dual species biofilms. In day 5 biofilms, we observed a statistically significant 5-fold increase in E. faecalis CFU within dual-species biofilms compared to mono-species biofilms (Fig. 1B) which correlated with the increased biomass observed by crystal violet staining. E. faecalis out-numbered S. aureus by approximately 60fold in dual-species biofilms. Surprisingly, although S. aureus produced very little biofilm biomass compared to E. faecalis (Fig. 1A), CFU equivalent to that of E. faecalis were recovered, suggesting that E. faecalis may produce more abundant biofilm matrix than S. aureus. Compared to the increased E. faecalis biomass and CFU in the presence of S. aureus, there were no significant differences in CFU sampled from non-adherent (planktonic) volume of the same biofilm wells, suggesting that E. faecalis growth augmentation by S. aureus is specific to biofilms (Fig. 1C).

Dual-species biofilm augmentation is strain dependent
To investigate whether S. aureus augmentation of E. faecalis biofilm is a phenomenon unique to the strains used in the initial studies (E. faecalis strain OG1RF and S. aureus strain USA300LAC), we assayed for biofilm augmentation using five additional commonly used S. aureus laboratory strains, as well as ten clinical wound isolates. We observed that 14 out of the 16 tested S. aureus strains/isolates showed a significant biofilm biomass increase when co-cultured with E. faecalis OG1RF ( Fig. 2A and Supplementary Table 1). Newman strain and isolate C37 did not augment mixed species biofilm biomass, potentially because of antienterococcal activity or biofilm-restricting properties of these strains that are discussed later. Together, the data indicates that S. aureus strain variation influences the nature of the dual-species relationship with E. faecalis vis-à-vis biofilm development.
We next investigated if the E. faecalis strain variation also affects dual-species biofilm augmentation by testing three additional E. faecalis laboratory strains and 28 clinical isolates derived from bloodstream (VRE isolates), wound (TTSHW-EF05 to EF43), urinary tract infections (UTIEF isolates), or the healthy gastrointestinal tract of children (HCG isolates). In contrast to the relative consistent ability of S. aureus strains to augment dual-species biofilms, augmentation observed during co-culture with S. aureus USA300LAC was much more heterogenous with only six out of 32 strains/isolates showing augmented biofilms (Fig. 2B and  Supplementary Table 2). Of these six strains, the degree of augmentation observed varied between 2.06-3.10 fold (Supplementary Table 2). Of the 26 that did not produce augmented biofilms, 17 showed single-species biofilm levels that were lower Fig. 1 Single-and dual-species in vitro biofilms. A E. faecalis (Ef ) and S. aureus (Sa) biofilms were grown in 96-well plates alone or in combination (1:1) over five days before biofilm was quantified by crystal violet (CV). Day 0 refers to 1 h post-inoculation. Data shows mean and SD (N ≥ 3) with ***p < 0.001 and ****p < 0.0001 when compared to E. faecalis-only biofilms of the same time point using 1-way ANOVA with Tukey's post-hoc test. B Day 5 biofilm and (C) planktonic cells of Ef or Sa or both were collected and CFU/well determined using selective agar. Data shows mean and SD (N ≥ 6) with ***p < 0.001 and **** p < 0.0001 using 1-way ANOVA and Tukey's post-hoc test.
than OG1RF under the conditions used (defined arbitrarily as <0.5× the biofilm levels of OG1RF), five were defined as having high biofilm levels (>2× the biofilm levels of OG1RF), and four had similar levels of biofilm to OG1RF (>0.5× but <2× of OG1RF) (Supplementary Table 2). Taken together, we conclude that many E. faecalis isolates are not susceptible to biofilm augmentation by USA300LAC. This is in keeping with other studies showing strain differences (not just species composition) profoundly affects the nature of microbial interactions [71][72][73].
E. faecalis transposon screen identifies menA and cydA to be crucial for dual-species biofilm augmentation To determine the mechanism by which E. faecalis CFU are increased in the presence of S. aureus, leading to dual-species biofilm augmentation, we screened a near-saturated E. faecalis transposon mutant library [66] for mutants that displayed altered dual-species biofilm accumulation, using the same microtiter CV biofilm assay. After secondary screening to eliminate any mutants that were attenuated in biofilm formation, we validated nine mutants in seven unique genes that reduced dual-species biofilms (Fig. 3A).
Amongst these was epaOX that encodes a glycosyltransferase involved in the production of cell wall rhamnopolysaccharide enterococcal polysaccharide antigen (Epa), is responsible for E. faecalis biofilm structure and stability in response to antibiotic stress [74], and is a determinant of biofilm-associated antibiotic resistance [75]. Both the transposon mutant and a clean deletion mutant of epaOX showed a reduction in dual-species biofilm levels relative to the single species control which was restored to parental strain levels in the plasmid-complemented strain ( Supplementary Fig. 1a) after correcting for variations in single species biofilm levels ( Supplementary Fig. 1b). As such, it seems likely that epaOX is involved in augmented dual-species biofilms through EPS production.
The roles of four additional validated gene products identified in the transposon screen (an ABC transporter, SufB, RpiR, and a Ser/Thr Phosphatase, Fig. 3A) were not pursued. The essentiality of sufB had been suggested in some studies [76] but deemed otherwise by others [58], and its involvement in respiration, via FeS cluster assembly, merits future attention. Most importantly, we identified three unique menA (OG1RF_11661) and one cydA (OG1RF_11666) transposon mutants that failed to undergo dual species augmentation. Both genes encode proteins participating in oxidative respiration (Fig. 3B) [77] prompting closer investigation into this process.  Table 1). B Four laboratory strains of E. faecalis (OG1RF-V583), together with 28 E. faecalis patient isolates (VRE122-TTSHW-EF43) were grown for five days alone or with S. aureus (USA300LAC) before biofilm was quantified by CV. Results show biofilm levels normalized to OG1RF-only controls, with mean and SD (N ≥ 3). Points colored green are significantly different to respective single species biofilms by p < 0.05 using 1-way ANOVA with Bonferroni's post-hoc test (details in Supplementary Table 2).
E. faecalis OG1RF biofilm augmentation by S. aureus, heme and hemoglobin requires O 2 and cydABCD We hypothesized that E. faecalis respiration is necessary for dualspecies biofilm augmentation, and to test this, biofilms were cultured in oxic or anoxic conditions. As predicted, dual-species biofilm augmentation was not observed in anoxic conditions (Fig. 4A) but was increased by over 2-fold in oxic conditions (Fig. 4B), demonstrating that oxygen is required for dual-species biofilm augmentation.
Since the utilization of oxygen by E. faecalis is dependent cytochrome bd which requires the incorporation of exogenous heme as a cofactor (E. faecalis does not synthesize heme [63,78,79]), we directly investigated the effect of free and conjugated heme (hemin and haemoglobin respectively) on E. faecalis biofilms. Supplementation with either source significantly increased biofilm biomass by over two-fold in oxic conditions (Fig. 4B) and was accompanied by an increase in oxygen consumption rate (Supplementary Fig. 2). Additionally, E. faecalis biofilm growth kinetics with hemin supplementation were measured, with a rapid increase in biofilm staining observed to plateau after day 1 (Supplementary Fig. 3). Under anoxic conditions, hemin supplementation minimally impacted biofilm levels and hemoglobin had no effect (Fig. 4A).
Cytochrome bd complex consists of two subunits that are encoded by cydA and cydB. The same operon (Fig. 5A insert) includes cydC and cydD, which encode an ATP-binding cassette (ABC)-type transporter required for the expression of a functional cytochrome bd complex [63,80]. As a functional cydABCD operon was presumed to be required for aerobic respiration to occur, we tested if disruptions of cydB, cydC and cydD would also attenuate the E. faecalis biofilm response to both S. aureus and hemin. Biofilm augmentation by both hemin and S. aureus was significantly impaired in all the tested transposon mutants of the four genes (Fig. 5A), indicating that that a functional operon was required for augmentation. Deletion mutants of cydB and cydD likewise failed to augment in the presence of hemin and S. aureus, with augmentation completely restored in the complementary strains (Fig. 5A). Additionally, the absence of heme-or Fig. 3 Validation of E. faecalis transposon library screen hits. A The parental strain (OG1RF) and nine transposon mutants were grown alone (circles), or together with S. aureus (Sa, USA300LAC, triangles) for five days before biofilm levels were determined by crystal violet (CV). Data shows mean and SD relative to OG1RF-only control (N ≥ 3). **p < 0.01, ***p < 0.001 and ****p < 0.0001 when compared to respective OG1RFonly or OG1RF + Sa control biofilms using 1-way ANOVA with Dunnett's post-hoc test. B Oxidative respiration of E. faecalis requires MenA, CydA, CydB, heme and O 2 .
hemoglobin-induced augmentation in the menA transposon mutant shows the requirement for demethylmenaquinone for oxidative respiration to occur (Figs. 5A and 3B).
To ensure that the absence of E. faecalis biofilm augmentation by S. aureus in the cydABDC transposon and deletion mutants was not due to their out-competition by S. aureus, the planktonic cells in single-and co-cultured wells were enumerated on selective agar after one day. Of the tested strains, only epaOX::Tn had lower CFUs compared to the OG1RF parental strain, both when grown alone and in the presence of S. aureus, allowing us to exclude outcompetition by S. aureus as a confounding factor ( Supplementary  Fig. 4).
To differentiate the involvement of cytochrome bd (CydAB) from the heterodimeric ABC transporter (CydDC) in E. faecalis biofilm augmentation, intracellular heme was measured in the cell pellets of the respective transposon and deletion mutants and complementary strains by mass spectrometry. Consistent with the inability of E. faecalis to synthesize heme, heme was only detected when E. faecalis was cultured in TSB supplemented with hemin, but not in the unsupplemented TSB control (Fig. 5B). Intracellular heme was negligible or not detected in any of the cyd mutants when grown in hemin-supplemented TSB (Fig. 5B), suggesting that this operon is essential for heme import. Complementation of cydD, but not of cydB, partially restored intracellular heme levels (Fig. 5B), potentially due to transcriptional differences arising due to trans chromosomal complementation with a non-native (srtA) promoter. Notwithstanding, partial restoration of intracellular heme levels in the cydD complement indicates that the functional product of the cydDC operon is required for heme import. Taken together, these data support that aerobic respiration in E. faecalis is activated by heme, hemoglobin or S. aureus, leading to augmented biofilm formation, and that import of heme into E. faecalis is mediated by CydDC.
Heme biosynthesis in S. aureus is responsible for dual-species biofilm augmentation Since E. faecalis cannot synthesize heme [63,79], we hypothesized that S. aureus-derived heme might enable the activation of cytochrome bd, giving rise to augmented dual-species biofilms. If so, S. aureus mutants deficient in heme biosynthesis should be unable to augment E. faecalis biofilms. We identified S. aureus transposon mutants in four key heme biosynthesis enzymes (hemA, hemB, hemE and hemL) in the S. aureus Nebraska transposon mutant library [67]. To verify the loss of heme biosynthetic activity, we performed growth assays to query the expected growth defect when exogenous heme was limited [81] and quantified heme in cell pellets by LC-MS. Only hemB::Tn displayed a growth defect (Supplementary Fig. 5a) that was restored upon heme supplementation (Supplementary Fig. 5b). The hemB mutant was also the only one that showed a 40-fold reduction in heme within cell pellets (Supplementary Fig. 5c). These data indicated that the hemB transposon mutant was the only bona fide heme-defective mutant of the four. The hemA, hemE and hemL mutants may have acquired secondary mutations, a documented caveat to this library [82], enabling them to bypass these mutations.
Consistent with our hypothesis, the S. aureus hemB mutant defective in heme biosynthesis did not give rise to augmented dual-species biofilms (Fig. 6). Notably, the single-species parental S. aureus biofilm levels were comparable to the hemB mutant, suggesting that growth restriction of this mutant in the absence of heme supplementation is unlikely to account for the drastic reduction in dual-species biofilms. However, this does not rule out the possibility that the hemB mutant is outcompeted by E. faecalis during biofilm formation. Overall, the data indicate that S. aureus and E. faecalis dual-species biofilm augmentation is dependent on heme biosynthesis in S. aureus. As AtlA is necessary for cell lysis during S. aureus biofilm formation [83], we asked whether autolysis contributed to heme release by S. aureus to augment dual-species biofilms. When E. faecalis and the S. aureus atlA transposon mutant were cocultured, they produced less biofilms than when E. faecalis was cocultured with the S. aureus parental strain, but the biomasses were still significantly augmented compared to the E. faecalis-only control (Fig. 6). This suggests that AtlA contributes to, but is not exclusively required, for heme release from S. aureus.
Recognizing the importance of S. aureus-derived heme in driving dual-species biofilm biomass augmentation, we sought to determine if heme and hemoglobin similarly augmented E. faecalis biofilm CFU as observed in Fig. 1B. As expected, the unaugmented cydA::Tn biofilms had similar CFU to the E. faecalis parental strain single species biofilm (Supplementary Fig. 6). Interestingly, while the E. faecalis parental strain underwent augmented biomass upon supplementation with hemin or hemoglobin (Fig. 4B) this was not reflected in increased CFU ( Supplementary Fig. 6). These data suggest that the mechanism of biofilm augmentation by hemin and hemoglobin may be distinct from that by S. aureus, where hemin and hemoglobin elicit increased extracellular matrix production rather than elevating cell numbers. Further work to understand these differences is worthwhile and could inform the induction of E. faecalis stress responses to heme/hemoglobin.

E. faecalis Gelatinase E (GelE) is involved in using heme synthesized by S. aureus
Having elucidated the role of heme in dual-species biofilm augmentation, we re-examined the S. aureus and E. faecalis strains and isolates to understand combinations that did not give rise to augmented biofilms ( Fig. 2A, B). We assayed intracellular S. aureus heme levels in Newman strain and isolate C37 (both failed to augment dual-species biofilms, Fig. 2A) and found them to have comparable levels as USA300LAC (Supplementary Fig. 7). This suggests that heme biosynthesis in S. aureus, though important, may not be the dominant factor governing dual-species biofilm augmentation. For instance, strain differences in heme or hemoprotein release from S. aureus would affect the efficiency of cross-feeding, whereas production of biofilm or growth inhibitors might curtail E. faecalis biofilm formation to begin with.
Of the E. faecalis strains and isolates, it was notable that OG1X, a sister clone of OG1RF lacking in gelatinase activity [84,85], was not augmented by USA300LAC (Fig. 2B). We therefore investigated whether gelE was responsible for use of USA300LAC-derived heme. Significant biofilm augmentation of E. faecalis OG1RFΔgelE only occurred during supplementation with hemin, but not S. aureus or hemoglobin (Fig. 7). This result demonstrates that free heme (in the form of hemin), but not conjugated heme (in the form of hemoglobin) could be used to augment biofilms in the gelE mutant. Moreover, the inability of the gelE mutant to augment biofilms in the presence of USA300LAC (Fig. 7) suggests Fig. 5 The E. faecalis cydABCD operon is required for biofilm augmentation by hemin and S. aureus, and for heme uptake. A Transposon mutants for cydA, cydB, cydC, cydD and menA, deletion mutants of cydB and cydD, their chromosomally complemented strains, along with parental wild-type OG1RF, were grown alone, in the presence of hemin (25 μg/ml), or with S. aureus (Sa, USA300LAC) for five days before biofilm was quantified with crystal violet (CV). Insert shows cydABDC operon and arrows indicate the direction of transcription. B Transposon and deletion mutants for cydA, cydB, cydC and cydD, along with complementary strains and parental wild-type OG1RF, were grown overnight in the presence of hemin (5 μg/ml) before cells were pelleted, lysed and analyzed by LC-MS for intracellular heme. Data shows mean values and SD. *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001 when compared to respective OG1RF control (first of each group) using Tukey's post-hoc tests (N ≥ 3). Fig. 6 Effect of heme-deficient S. aureus mutant on dual-species biofilm augmentation. Parental strain S. aureus USA300LAC, the hemB transposon mutant and the atlA transposon mutant were grown alone or in combination with E. faecalis (Ef ) for five days before biofilm was quantified with crystal violet (CV). Data shows mean and SD (N = 6). *p < 0.05, **p < 0.01, ***p < 0.001 and ****p < 0.0001 when compared to Ef alone, or as otherwise indicated by brackets, using 1-way ANOVA with Tukey's post-hoc tests.
that heme produced by USA300LAC is also in a conjugated hemoprotein form. Additional post-hoc testing also revealed that hemin-augmented biofilms of E. faecalis ΔgelE were lower than those of WT, likely due to the established role of GelE in E. faecalis biofilm [33] that extend beyond its hemoproteolytic activity.
We then selected a subset of E. faecalis strains and isolates (20 out of the original 32, and including OG1RFΔgelE as control) to assay for biofilm augmentation by hemin and hemoglobin. Hemin supplementation augmented biofilms in 17 of the 21 tested strains/isolates ( Supplementary Fig. 8 and Supplementary Table 3). The four that showed little to no augmentation by hemin (HCG5A4, HCG9A2, TTSHW-EF12 and TTSHW-EF30) also did not show augmentation with hemoglobin and S. aureus, consistent with the requirement for a functional respiratory chain in E. faecalis biofilm augmentation by hemoglobin and S. aureus. For TTSHW-EF30, intrinsically high biofilm levels (≥ four-fold that of OG1RF) may have been less amenable to further augmentation. Of the 17 that were augmented by hemin, 10 were also augmented by hemoglobin ( Supplementary Fig. 8 and Supplementary Table 3). Of these ten, eight (80%) were gelatinase-positive whereas two (HCG8E9 and TTSHW-EF16) were gelatinase-negative (Supplementary Table 3). In contrast, out of the seven that were augmented by hemin but not by hemoglobin, only V583 was gelatinase-positive. Together these data suggest that gelatinase activity is important for hemoglobin digestion (heme acquisition) but may not be the exclusive enzyme for this activity and may not be sufficient for heme acquisition.
Of the eight gelatinase-positive strains and isolates that were augmented by both hemin and hemoglobin, five were also augmented by S. aureus whereas three were not (Supplementary Fig. 8 and Supplementary Table 3). The three that were not augmented by S. aureus (TTSHW-EF9, TTSHW-EF18 and TTSHW-EF28) all produced higher levels of biofilms in the unsupplemented E. faecalis-only controls (2.6-2.9× more biofilms than OG1RF). This implies that the ability to extract heme from hemoglobin does not necessarily result in dual-species biofilm augmentation. Though there appear to be additional factors that contribute to the response of E. faecalis clinical isolates to S. aureus, it remains clear that gelE facilitates heme acquisition by E. faecalis from hemoglobin and S. aureus hemoproteins, highlighting a previously unappreciated role for this established virulence factor in E. faecalis polymicrobial biofilm formation.

DISCUSSION
Despite their co-isolation in biofilm associated infections, the only studies characterizing the molecular interaction between S. aureus and E. faecalis pertain to vancomycin resistance transfer [55][56][57]. In this study, we show that S. aureus and E. faecalis synergize to produce more biofilm and demonstrate that this is dependent on the synthesis of heme by S. aureus and cross-feeding by E. faecalis to exploit this resource for growth in an oxygen-dependent manner.
To appreciate these findings and provide context to the current literature, it is helpful to understand how E. faecalis and S. aureus interact with other species. E. faecalis produces bacteriocin which reduces growth of C. perfringens [21] and also hinders the secretion of botulinum neurotoxin of C. botulinum [86]. In oropharyngeal candidiasis, E. faecalis bacteriocin, EntV, prevents hyphal morphogenesis of C. albicans to suppress hyphaldependent cytotoxicity [87]. Under iron-limiting conditions, E. faecalis secretes L-ornithine which promote siderophore synthesis in E. coli [25,88]. Furthermore, E. faecalis suppresses host NF-κBdependent immune activity to promote the virulence of uropathogenic E. coli [89], and in early biofilm stages, AI-2 of E. faecalis acts as a chemo-attractant and augments E. coli biofilms [90]. E. faecalis also promotes biofilm matrix production in P. aeruginosa [91] through increasing production of exopolysaccharide Pel. In these studies, the effect of E. faecalis on other species is highlighted, yet details on how other species influence E. faecalis is conspicuously lacking.
The literature on S. aureus interactions with other co-colonizing species is more abundant and has been reviewed [92]. In biofilmassociated infections with Candida albicans, S. aureus exhibits increased resistance to vancomycin [93], increased mucosal adhesion and pathogenicity [94][95][96], and increased binding to hyphae via hyphal protein Als3p [97]. By contrast, S. aureus interactions with other bacteria appear competitive: Staphylococcus epidermidis and Bacillus subtilis interferes with S. aureus quorum sensing by inhibiting the accessory gene regulatory (agr) system, resulting in suppression of virulence genes [98][99][100]. Additionally, Pseudomonas aeruginosa, Streptococcus pneumoniae and lactic acid bacteria (LAB) produces toxic products such as pyocyanin, LasA and hydrogen peroxide (H 2 O 2 ) to inhibit the growth of S. aureus [101][102][103][104]. LAB also inhibit the colonization of S. aureus by competing for host adhesion sites [105]. As countermeasures, S. aureus forms small-colony variants to counter P. aeruginosa toxins and produces bacteriocins to inhibit LAB [106,107]. By contrast, S. aureus promotes colonization of Haemophilus influenzae within the nasal cavity through hemolysis of erythrocytes and the subsequent release of heme and NAD that drives H. influenza growth [108,109]. Notably, S. aureus responds to hemoglobin in nasal secretions to increase colonization by dampening agr expression [110].
One challenge in microbial ecology is in understanding how strain differences contribute to mixed species interactions. Here we show that most S. aureus strains augment E. faecalis biofilm formation but that the converse is not true. Of the 16 tested S. aureus strains grown with E. faecalis, only Newman and C37 did not show any biofilm augmentation. In the case of Newman, supernatants inhibit biofilm formation by other S. aureus strains via a proteinase-sensitive, heat-tolerant and soluble protein [111]. It is possible that this same biofilm-inhibitory protein could also interfere with E. faecalis biofilm formation giving rise to antagonism such that dual-species biofilm levels were lower than E. faecalis-only biofilms. Of 31 E. faecalis strains tested, only five were amenable to biofilm augmentation in dual-species biofilms. Fig. 7 Effects of gelatinase E (gelE) deletion on heme-source utilization. Parental strain E. faecalis (OG1RF) and respective gelE deletion mutant (ΔgelE), were grown alone or in combination with hemin (25 μg/ml), hemoglobin (Hb, 10 μg/ml) or S. aureus (Sa, USA300LAC) for five days before biofilm was quantified with crystal violet (CV), Data shows mean and SD (N ≥ 4). ****p < 0.0001 when compared to respective Ef controls using 1-way ANOVA with Bonferroni's post-hoc tests.
For several strains, this may be due to their intrinsically lower biofilm producing capacity in the first place. However, the majority appeared to be able to form biofilms yet behaved differently from OG1RF under the conditions tested and restricting the generalizability of dual-species augmentation. This emphasizes the importance of strain differences and raises questions about intrinsic differences in heme uptake and/or aerobic respiration capacity in different E. faecalis strains.
The majority of transposon insertion mutants identified as deficient in undergoing biofilm augmentation in the presence of S. aureus were in menA (1,4-dihydroxy-2-naphthoate isoprenyltransferase, which converts 1,4-dihydroxy-2-naphthoate to demethylmenaquinone) and cydA (cytochrome d ubiquinol oxidase subunit I), both of which encode proteins involved in oxidative respiration [60]. For respiration to occur in E. faecalis, oxygen and heme are required [60,112], the latter serving as an essential cofactor for cytochrome bd (made up of CydA, CydB and three moieties of cofactor heme) and which acts as the terminal demethylquinol (DMKH 2 ) oxidase, generating H 2 O from O 2 and establishing a proton motive force to generate ATP [113]. Since E. faecalis is unable to produce porphyrins required to synthesize heme [78], heme must be supplied exogenously. In our experiments, heme was supplied in the form of hemin (free heme) or hemoglobin (protein-conjugated heme). Both hemin and hemoglobin augmented E. faecalis biofilms only under oxic conditions, leading to the conclusion that heme augments E. faecalis biofilms by enabling aerobic respiration. Clinically, this could translate to dual-species biofilm being more prominent in sites where oxygen is available, such as superficial wounds, and less of a concern in anoxic regions like the bladder [114,115].
Through a very different mechanism, the shift from anaerobic to aerobic respiration because of interspecies interactions has been described for Aggregatibacter actinomycetemcomitans. Though primarily fermentative, the opportunistic oral pathogen A. actinomycetemcomitans switched to respiratory metabolism when grown in the presence of Streptococcus gordonii in response to enhanced O 2 bioavailability during coinfection [116]. Although it was not determined how O 2 levels were augmented, the assumed provision of electron-acceptors (O 2 ) from S. gordonii (which requires O 2 to produce H 2 O 2 ) resulted in increased A. actinomycetemcomitans growth and persistence, giving rise to the term "cross-respiration". Opposite to this, S. aureus has been shown to increase biofilms when grown in anoxic conditions via a SrrABdependent programmed cell lysis, whereas the inactivation of the heme production (hemB) permitted greater biofilm growth under oxic conditions [83]. Differences between the previous S. aureus study and ours could be due to differences in biofilm assay conditions, in particular the duration of incubation (22 hr used previously compared to 5 d used in our study).
Although exogenous heme utilization has been established for both oxidative respiration and catalase (KatA) stability and activity in E. faecalis [117,118], no heme import machinery has been defined. Incidental to our main findings is the discovery that cydDC is essential for heme import. Previously shown to be essential for cytochrome bd assembly in E. coli, the CydDC proteins are ABC transporters that export redox-active thiol compounds such as cysteine and glutathione [119][120][121][122]. In E. coli, it has also been suggested that CydDC may bind heme to enhance ATP hydrolysis and thiol export [78,[123][124][125]. Given that no intracellular heme could be detected in E. faecalis cydD or cydC mutants cultured in heme-supplemented media, we propose that the CydDC heterodimer may also function as a key, if not the sole, heme importer in this species. However, others have suggested that CydDC is not involved in heme import or cytochrome assembly based on everted E. coli membrane vesicles involving a cydD1 point mutant that showed similar patterns of hemin uptake as those in the parental strain and may instead be important for maintenance of a suitable redox environment in the periplasm for conversion to heme d [126,127]. Further work is needed to rule out the possibility that cytochrome bd assembly require CydDC, which could also impact heme levels in the membrane, or that other import mechanisms may exist since heme-dependent KatA activity has been described in cydABDC deletion mutants [78]. Moreover, we only performed heme quantification on whole-cell lysates, and fractionation experiments could ascertain if heme is internalized into the cytosol or directly incorporated into the cell membrane.
Loss-of-function mutations to the cydABDC operon are enriched in E. faecalis isolates resistant to killing in the hemoglobin-rich environment of whole blood [128] and this has been attributed to the induction of respiration in parental strain of E. faecalis in the presence of heme which increased vulnerability to neutrophil killing [65]. It is therefore possible, in the context of host infection, that only within the immune-shielded environment of biofilms can heme utilization by E. faecalis safely take place without concomitant increasing their susceptibility to immune clearance -hence the importance of E. faecalis augmenting biofilms in response to heme.
In vitro, it is unclear why E. faecalis augmentation by S. aureus occurs in biofilms but not planktonic cells. It has been reported by others (and our own studies agree, data not shown) that E. faecalis planktonic growth is unaffected by heme supplementation [118]. We speculate that the spatially confined microenvironment of biofilms, or the unique transcriptional profile of biofilm cells, results in sensitivity to heme-based growth augmentation.
Biofilm augmentation of many E. faecalis strains and isolates by hemin but not by hemoglobin or heme-producing USA300LAC suggests that an important secondary E. faecalis determinant is involved in this mixed species interaction. Comparative genomics led us to identify GelE as required for E. faecalis OG1RF biofilm augmentation in the presence of hemoglobin and USA300LAC. Free heme could augment biofilm formation in the gelE deletion mutant, but this strain was unable to extract heme from hemoglobin or S. aureus hemoproteins. Together with its established role in biofilms, GelE has previously been shown to hydrolyse hemoglobin [129] which would result in heme release and subsequent import, potentially by CydDC, and incorporation into cytochrome bd of E. faecalis to allow aerobic respiration to augment biofilms. We speculate that S. aureus hemoproteins, whether secreted or released during cell death, may likewise need to be hydrolysed for heme release and subsequent uptake by E. faecalis.
In summary, we show that augmentation of S. aureus and E. faecalis dual-species biofilms requires the biosynthesis of heme by S. aureus which facilitates oxidative respiration in E. faecalis. In most cases, successful cross-feeding of heme (likely in the form of secreted hemoproteins) will require gelatinase-mediated heme acquisition by E. faecalis. It is conceivable that the increased energy derived from aerobic respiration enables production of complex and "expensive" extracellular biofilm-associated proteins like EpaOX or surface-anchored pili (EbpABC) [130]. Since epaOX (along with fsrA, gelE and epaI) correlate with biofilm-associated antibiotic resistance [75], the potential impact of this inter-species interaction on antibiotic efficacy is worth considering, especially since this has been demonstrated for other mixed species interactions [131,132]. The findings of this study highlight the importance of understanding inter-species interactions in biofilms and underscore the usefulness of identifying potentiating determinants, like heme, to develop interventions relevant in a complex host setting.