C-terminal deletion of RelA protein is suggested as a possible cause of infective endocarditis recurrence with Enterococcus faecium

ABSTRACT

Infective endocarditis (IE) caused by Enterococcus spp. represents the third most common cause of IE, with high rates of relapse compared with other bacteria. Interestingly, late relapses (>6 months) have only been described in Enterococcus faecalis, but here we describe the first reported IE relapse with Enterococcus faecium more than a year (17 months) after the initial endocarditis episode. Firstly, by multi locus sequence typing (MLST), we demonstrated that both isolates (EF646 and EF641) belong to the same sequence type (ST117). Considering that EF641 was able to overcome starvation and antibiotic treatment conditions surviving for a long period of time, we performed bioinformatic analysis in identifying potential genes involved in virulence and stringent response. Our results showed a 13-nucleotide duplication (positions 1638–1650) in the gene relA, resulting in a premature stop codon, with a loss of 167 amino acids from the C-terminal domains of the RelA enzyme. RelA mediates the stringent response in bacteria, modulating levels of the alarmone guanosine tetraphosphate (ppGpp). The relA mutant (EF641) was associated with lower growth capacity, the presence of small colony variants, and higher capacity to produce biofilms (compared with the strain EF646), but without differences in antimicrobial susceptibility patterns according to standard procedures during planktonic growth. Instead, EF641 demonstrated tolerance to high doses of teicoplanin when growing in a biofilm. We conclude that all these events would be closely related to the long-term survival of the E. faecium and the late relapse of the IE. These data represent the first clinical evidence of mutations in the stringent response (relA gene) related with E. faecium IE relapse.

INTRODUCTION

Enterococcus spp. are Gram-positive bacteria that cause serious nosocomial infections, including urinary tract, bloodstream, and endocarditis (1). Enterococci are known for their ability to form biofilms, which are able to attach several biotic and abiotic surfaces (2). The biofilm production not only contributes as a virulence factor enhancing the adherence of the bacteria to several surface but also plays an important role in the bacterial resistance to antibiotics and its pathogenesis efficiency, making their eradication extremely difficult (3–5). Mature biofilm has been reported to increase antibiotic tolerance at concentrations 10–1,000 times higher than the required to kill bacteria without biofilm formation (5). Nevertheless, the resistance phenotype of the Enterococcus spp. is not exclusively related to the virulence factors (biofilm formation). Enterococci are intrinsically resistant to several antibiotics, and they easily mutate and exchange resistance genes with other bacteria. Among Enterococcus spp., Enterococcus faecalis and Enterococcus faecium are the most often associated to multidrug-resistant nosocomial infection (6).

Among the nosocomial infections caused by enterococcal species, infective endocarditis (IE) is the type of infection where Enterococcus spp. is one of the most prevalent species (7). IE is an infection of the endothelium of the heart that even being a rare condition (3–10/100,000 of the population) carries a high risk of morbidity and mortality (30% at 30 days) (8). Interestingly, although enterococci are the third most common cause of IE worldwide, the IE caused by enterococci is characterized by a 5% rate of relapse, which is high relative to that of IE due to other bacterial species (7). Several studies have reported IE relapses by enterococci, but the mechanism underlaying those relapses is not entirely elucidated (9–11). The genetic evolution of this bacterial species and its diversification under selective pressure could play an important role in determining the infection outcome. Here, we aimed to determine the genome modifications that allowed the phenotypic adaptations selected to overcome antibiotic treatment and immune pressure in a patient with E. faecium IE relapse with a gap of time of 17 months between each episode of infection.

RESULTS AND DISCUSSION

Clinical setting

An 85-year-old woman was admitted to our center for abdominal pain, fever, and jaundice. She was diabetic and had been previously diagnosed from gallstone disease, and she underwent surgery for the placement of a biologic prosthesis for aortic stenosis from around 4 years ago. On arrival at the emergency room, she was hemodynamically unstable and fluid supply, support with inotropes, and empiric antibiotherapy with piperacillin-tazobactam were started. Ultrasound and computed tomography (CT) scan revealed a gall bladder with multiple lithiasis, one of them lodged in distal common bile duct, producing a severe dilatation. An echo-guided cholecystostomy was performed and an E. faecium was isolated in the drainage fluid. Once stabilized, an endoscopic retrograde cholangiopancreatography was performed with an endoscopic sphincterotomy and the complete extraction of all gallstones in distal common bile duct. The patient was discharged after completing 3 weeks of piperacillin-tazobactam, but 4 months later, she was readmitted again for abdominal pain and fever. Laparoscopic cholecystectomy was then performed without complications. No blood cultures were taken, and the patient was discharged a few days later receiving oral treatment with ciprofloxacin (7 days).

However, 1 month later, she went to the hospital again for fever and chills, without an evident infectious focus. Repeated blood cultures were obtained with the isolation of E. faecium in all of them. A CT scan revealed two splenic infarcts and another one in the left kidney. A transoesophageal echocardiography revealed a calcified mitral valve, and the aortic prosthesis showed thickened valves and a mobile vegetation (5 × 2 mm) in the non-coronary cusp. A diagnosis of late prosthetic endocarditis was established, and she received 6 weeks of IV teicoplanin (10 mg/kg) with complete clinical recovery and follow-up negative blood cultures at 2, 4, and 6 months after the treatment finalization.

At 11 months after the final visit, the patient came again with fever, low cognitive status, acute renal failure, and left ventricular failure, dying within 72 hours of admission. An ultrasound scan revealed a new big splenic infarct and E. faecium was again isolated in serial blood cultures. Transesophageal echocardiography was not performed, and necropsy was denied.

This case is the first reported IE relapse with E. faecium more than a year (17 months) after the initial endocarditis episode. Relapses after treatment of enterococcal IE are frequent and can occur more than 1 year after the initial IE episode (9). However, late relapses (>6 months) have only been described in E. faecalis (9, 12). In order to demonstrate that the second episode of IE was caused by a relapse, we performed a genetic characterization of the strains.

Genetic characterization

To elucidate whether both E. faecium isolates were clonally related and mutations with clinical relevance were present, allowing the isolates to overcome all treatment used and to survive a long-term in a latent phase, both isolates (EF646 and EF641) were sequenced and assembled to obtain their genome. Multi locus sequence typing (MLST) showed both isolates to belong to the same sequence type (ST117). Analysis of the functional roles of predicted coding regions showed multiple antibiotic resistance genes, including msr(C), aac(6′)-li, tet(M), lnu(C), gyrA(S83Y), and parC(S80I) genes for both isolates, and aac(6′)-aph(2″) gene for EF641 (Fig. S1). Considering that EF641 has been able to overcome starvation and antibiotic treatment conditions surviving for a long period of time, the mechanism underlaying the relapse could be related to certain mutations in the virulence genes of the isolates. Thus, E. faecium virulence gene sequences [adherence (acm, ebpA, ebpB, ebpC, srtC, ecbA, efaA, esp, scm, and sgrA), antiphagocitosis (cpsA/uppS and cpsB/cdsA), and biofilm formation (bopD)] were compared between EF641 and EF646. No insertion, deletion, or substitution was detected when EF641 and EF646 virulence genes were compared (Table S1). Regarding the lack of mutations in the virulence genes, the sequence of the regulatory gene of the stringent response (relA) was analyzed. The stringent response is a near-universal stress response modulated by hyper-phosphorylated derivatives [guanosine tetraphosphate (ppGpp) and guanosine pentaphosphate (pppGpp)], also known as alarmone (13, 14). Stress conditions could induce the accumulation of alarmone [(p)ppGpp] within the cell, which is the main effector molecule of the stringent response, leading to transcriptional modification that make the cell move from active growth state to semi-dormant state (15). Several studies have associated the production of (p)ppGpp with the expression of virulence traits, which includes adhesion, biofilm formation, toxin production, motility, sporulation, and antibiotic tolerance. Noteworthy, several of those studies have indicated that the basal level of (p)ppGpp might be the cause of bacterial persistence and virulence but not the stringent response (16–19).

The modulation of the (p)ppGpp basal levels through its synthesis and degradation is mediated by RelA enzymes that are universally composed of two functional regions: the catalytic region (CR) and the regulatory region (RR). The CR comprises the (p)ppGpp hydrolase domain and the (p)ppGpp synthetase domain. Both domains are allosterically regulated by the RR. The RR contains two regulatory domains [Thr-tRNA synthetase, GTPase and SpoT domain (TGS) and Aspartokinase, Chorismate mutase and TyrA (ACT)] that regulate RelA enzymatic activities (synthesis and hydrolysis, respectively). In addition, TGS and ACT are separated by two other structural elements: an α-helical domain (AH) and a zinc-finger domain (ZFD) (Fig. 1) (18, 20, 21).

Fig 1.

Predicted pathways of stringent response without (A) and with (B) the relA mutation. (A). Under high availability of resources (HAR), RelA is responsible for the degradation of alarmone [(p)ppGpp]. On the other hand, under low availability of resources (LAR), RelA is responsible for the production of alarmone. (B) In the relA mutation, a duplication of 13 nucleotides changed the reading frame and, consequently, a premature stop codon is generated in the relA gene sequence, losing most of the hydrolase regulatory part of the RelA protein. This modification in RelA could led to increased resting levels of (p)ppGpp, allowing the bacteria to survive for long term in the patient through biofilm formation and reducing the growth rate.

The comparison between the relA genes from EF641 and EF646 strains showed a 13-nucleotide duplication (positions 1638–1650) that resulted in a premature stop codon, predicted to encode an RelA (p. Glu551_Gly718del). Thus, in our RelA mutant (from EF641), 167 amino acids from the C-terminal domains of the RelA enzyme corresponding to the ACT and ZFD domains of the enzyme were lost (Fig. 1). Previous studies have showed that removal of the C-terminal domains of RelA increases the rate of (p)ppGpp production by RelA enzymes, proposing the C-terminal domains as the inhibitor of the synthetase activity (22–26). In E. faecium, a single missense mutation in the RelA (Leu152Phe) was described in a bacteremia case. This mutation raised the baseline levels of alarmone leading to a high level of antibiotic tolerance (27). Thus, based on previous results and our own evidence, we suggest that the mutation found in EF641 isolate provides a high basal level of alarmone allowing this isolate to adapt and respond to stress conditions (antibiotic treatment, starvation, etc.) through the expression of several virulence factors, biofilm formation, and reduction of growth rate entering in a semi-dormant state for long-term survival.

Finally, in order to confirm that the mutation in relA was the only gene responsible for the observed phenotype, the single nucleotide polymorphism (SNP) calling was performed, and we only observed 18 SNPs. Among them, seven were found in non-coding region, two were found in pseudogenes, three corresponded to silent mutations, and only 6 out of the 18 SNPs were missense mutation leading to amino acid substitution in four coding regions. Those coding regions belong to the following proteins: IIHAD hydrolase family (N159D); LCP family protein, which is a cell wall polymer synthetase (R312S); HelD, which is a DNA helicase IV (D152Y); and GntR transcriptional regulator (T105R and V106I) (Table S2). Thus, we can conclude that EF646 and EF641 are near-isogenic strains.

Variations in fitness, susceptibility, biofilm formation, and tolerance

Since a possible mis-regulation of (p)ppGpp due to deletion of the regulatory part of the RelA enzyme would be implicated in alterations of growth rates, susceptibility, biofilm formation, and tolerance, we aimed to determine the effect of this mutation in both bacterial processes. Figure 2A shows how EF646 was able to prosper in tryptic soy broth (TSB). However, we observed a defect in growth capacity in EF641 isolate that may be related to the mutation observed in RelA, suggesting a modification of nutritional competence balance. High (p)ppGpp levels lead to slow growth or growth arrest in bacteria (28), and this, probably, could reinforced the non-inherent tolerance of the strain EF641 to raise concentrations of teicoplanin (29). Regarding the antimicrobial susceptibility patterns, there were no changes between the strains EF646 and EF641. Both strains were susceptible to teicoplanin, linezolid, and vancomycin (Table S3). The strain EF641 was resistant to aminoglicosides, most likely due to the presence of the aac(6′)-aph(2″) gene. In the time-kill assay, we did not observe differences between the strains EF646 and EF641 independently of the teicoplanin concentration (Fig. 2A).

Fig 2.

(A) Time-kill curves for EF646 and EF641 exposed to different (0, 4, and 30 µg/mL) teicoplanin concentrations. (B) Biofilm biomass quantification by OD₆₀₀ meauserment of solubilized crystal violet. (C) The relA mutation contributes to tolerance to teicoplanin in biofilms. (B and C) Data represent the mean of three replicates; error bars represent standard deviation. Different conditions were compared by Student’s t test: *P < 0.05; **P < 0.01.

High (p)ppGpp production has also been related with small colony variant (SCV) in Staphylococcus aureus (30). During bacterial growth studies, we observed this phenomenon, with the presence of SCVs in EF641 on blood agar and auto-aggregation processes in LB broth (Fig. S2). SCVs obtain a survival advantage by slow replication and, thereby, reduced susceptibility to antibiotics (31).

As mutations in relA have been implicated in alterations in biofilm formation, we next determined the capacity of these strains to produce biofilms. As can be observed in Fig. 2B, EF641 exhibited a significant higher capacity to produce biofilms compared with the strain EF646. Previous studies have demonstrated that (p)ppGpp levels have a profound effect on the ability of Enterococcus to produce, develop, and maintain stable biofilms (32). Thus, E. faecalis mutant strains lacking relA gene showed less biovolume accumulation during biofilm growth experiments, suggesting that ΔrelA cannot use (p)ppGpp to sense and respond to stresses (32, 33). In addition, it has been demonstrated that the presence (p)ppGpp is crucial for valve colonization during enterococci IE in mice (34) and promotes biofilm formation on urinary catheters (35).

In contrast to the data presented here, Honsa et al. (27) described a persistent bacteremia episode caused by E. faecium with a single missense mutation in the RelA (Leu152Phe) that produced raised levels of alarmone, which conferred a defect in biofilm formation and bacterial survival within a biofilm. This difference in biofilm formation may be due to several factors. First, there are substantial differences in growth conditions. Thus, the mutant RelA (Leu152Phe) was cultured in Todd-Hewitt broth supplemented with 3% yeast extract with 0.2% of glucose, whereas our strains were cultured in TSB (0.25% glucose) and TSB supplemented with 0.5% and 1% glucose, as previously performed (36). It has been demonstrated that higher glucose concentration in culture media increased biofilm formation by E. faecalis (37). Our results also showed that biofilm formation is increased in TSB supplemented with 0.5% and 1% in comparison with TSB, but only for the strain EF641 (Fig. 2B). Furthermore, Honsa et al. (27) used well plates coated with bovine plasma fibronectin [which has been previously demonstrated to induce biofilm production (38)], to test biofilm formation. This could have distorted the results of the investigation. Finally, the mutation analyzed by Hones et al. produced the altered residue (L152F), adjacent to residues essential that could reduce the hydrolase activity of RelA. In our case, a fraction of the regulatory C-terminal part of the relA gene was lost, so the hydrolase activity might have been suppressed.

Biofilm associated enterococcal infections not only are difficult to eradicate but also serve as a focus for bacterial dissemination and as a reservoir for antibiotic resistance (39). Here, the question is whether the mutation found in RelA could be related to antibiotic tolerance without conferring antibiotic resistance. Previous studies have demonstrated that planktonic enterococci respond differently to antibiotics in comparison with biofilm-embedded enterococci. To answer this question, we performed tolerance assays with teicoplanin, the antibiotic used for the IE treatment (Fig. 2C). We observed that the biofilm generated by the strain EF641 was capable to survive to higher teicoplanin concentrations unlike the strain EF646. In corcordance with our results, a previous study investigating bacterial biofilm formation from IE patients demonstrated how the standard antibiotic susceptibility testing differed from the susceptibility of bacteria growing as biofilms (40). The study performed by Honsa et al. (27) also showed that high concentrations of different antibiotics, such as vancomycin or daptomycin, failed to eradicate the E. feacium harboring the relA mutation in a biofilm model, although this strain was susceptible according to standard MIC determinations. Despite these findings, there is not a standardized method to evaluate biofilm inhibition and/or eradication capacity of antibiotics, so the real impact during IE treatment remains to be elucidated.

Considering that the two strains are near-isogenic and that the only significant difference is the 13-nucleotide duplication found in EF641 genome, we can conclude that any phenotypic characteristic showed by EF641 and not by EF646 would be due to the loss of the C-terminal region in RelA.

Conclusion

It is known that surgery associated with antibiotic therapy is the best way to prevent Enterococcus spp. IE relapses (9). However, in cases in which surgery is not indicated or patients decline surgical intervention, the efficiency of antibiotic activity into deep-seeded biofilms and into bacterial vegetations may be compromised, even if a prolonged course of antibiotics is used. In our case, teicoplanin was used to treat the E. faecium IE instead of vancomycin to avoid nephrotoxicity, and to enable continuation of therapy at home with a single daily dose. Previous studies have reported that teicoplanin is a safe alternative therapeutic agent for treating enterococcal IE (41–44). Nevertheless, the study performed by De Nadaï et al. (45) showed that all relapses occurred during teicoplanin treatment were in patients presenting prosthetic valve endocarditis (46). The presence of an implant alters innate immune function and also the mechanisms by which systemic co-morbidities influence biofilm infection (47), and it seems to be related with more endocarditis relapses in patients with prosthetic valves (48–50). Given the above, the question that arises is whether it would be necessary to explore new strategies for the treatment of Enterococcus spp. IE in patients with prosthetic valves and when surgery is not indicated, such as anti-biofilm treatments (51, 52), or to consider the RelA-alarmone system as a new potential target for new treatment development.

To our knowledge, this is the first described case of IE relapse caused by E. faecium more than a year (17 months) after the initial endocarditis episode. In addition, this is the first time that an in vivo mutation occurring in RelA that results in a deletion of the C-terminal domain of the protein is described, possibly causing increased levels of alarmone, with a significant impact in biofilm production, bacterial growth, and increased tolerance to antibiotics. All these events would be closely related to the long-term survival of the E. faecium strain and, consequently, the IE relapse. In future studies, anti-biofilm treatments as well as the RelA-alarmone system could be considered as new targets to avoid Enterococcus spp. relapses in IE.

MATERIALS AND METHODS

Bacterial isolates

Two E. faecium isolates (named EF646 and EF641), one from each IE episode (blood culture), were received from the clinical microbiology laboratory at University Hospital Virgen del Rocio. Both isolates were identified by mass spectrometry and frozen for later analysis.

Susceptibility profile

The susceptibility profile of EF646 and EF641 isolates was performed using gradient strips (Liofilchem, Italy), due to the lack of growth of EF641 in broth liquid mediums. Clinical breakpoints were established according to the European Committee on Antimicrobial Susceptibility Testing (53).

Genomic DNA extraction, sequencing, and genomic analysis

Bacteria were grown in blood agar medium, and then genomic DNA was extracted by using the QIAamp DNA Mini kit (QIAGEN, Hilden, Germany) following the manufacturer’s instructions. In order to determine the mechanisms that lead to an IE relapse, whole-genome sequencing was performed using short-read Illumina sequencing (Illumina, USA) and long-read Nanopore (Oxford Nanopore Technologies, UK). A DNA library was prepared either (i) using a NEB Ultra II FS DNA library prep kit for Illumina (NEB, Evry, France) and then run on the MiSeq sequencer to generate paired-end 150 bp reads and (ii) using the rapid barcoding kit (SQK-RBK004) and a R9.4 flow cell on a MinION Mk1c sequencer (Oxford Nanopore Technologies). The basecalling of Nanopore reads was performed using Guppy Basecaller v.6.5.7 (54) in high-accuracy mode. Quality control and filtering of Illumina and Nanopore reads were conducted using Fastp v.0.23.4 (55). We employed high-quality Nanopore reads and the Flye v.2.9.2 assembler (56) for genome assembly. This was followed by long-read polishing with Medaka v.1.11.1 (https://github.com/nanoporetech/medaka) and four rounds of Illumina short-read polishing using Pilon v.1.24 (57). Finally, SNP calling was performed with Snippy v.4.6.0 (https://github.com/tseemann/snippy) using the reference genome NZ_CP038996.1.

The assembled genomes were also annotated using the Rapid Annotations using Subsystems Technology server (58, 59). MLST was obtained by using MLST v.2.0 (Center for Genomic Epidemiology). The acquired and intrinsic antimicrobial resistance and virulence factor genes were identified using ResFinder v.3.0, PlasmidFinder v.2.1, and VirulenceFinder v.2.0 (Center for Genomic Epidemiology) (https://www.genomicepidemiology.org/). Additionally, virulence factors were searched and analyzed by an automatic and comprehensive platform for accurate bacterial virulence factor identification, named VFanalyzer (60).

Bacterial growth and time-kill kinetics

The strains EF646 and EF641 were grown overnight in TSB at 37°C under shaking conditions (250 rpm). Cultures were diluted to produce an approximate starting culture of 10⁵ CFU/mL. Replicates of each culture (in triplicate) were separately analyzed over time, containing increasing concentrations of teicoplanine at 4 µg/mL (resistant MIC breakpoint) and 30 µg/mL (target Cmin for teicoplanin was previously determined to 15–30 µg/mL during the treatment of endocarditis) (61). Growth controls were performed without teicoplanin. Cultures were incubated at 37°C under shaking conditions (250 rpm), and bacterial growth was determined at 0, 2, 4, 6, and 24 hours. Each sample was analyzed by performing serial dilutions to determine the number of CFU per milliliter over time.

Biofilm assay

Biofilm formation was performed by the Crystal Violet assay as described previously (62). Briefly, overnight cultures were inoculated in TSB and TSB supplemented with 0.5% and 1% glucose (Oxoid, Hampshire, England). The inoculum was spectrophotometrically adjusted to approximately 0.5 McFarland standard. Each inoculum was diluted 1:100 in fresh TSB, and 200 µL of each diluted inoculum was distributed in three wells of a 96-well plate and incubated 48 hours at 37°C. Negative control wells were included. Planktonic cells were removed, and wells were washed with sterile phosphate buffer for removal non-adherent cells, decanted, and dried. Adherent cells were fixed with ethanol and then they were stained with 0.5% (wt/vol) crystal violet for 15 minutes. Excess stain was gently rinsed off with water. Plates were allowed to air-dry. The dye bund to the adherent cells was re-solubilized with 200 µL of ethanol (100%). The optical density was measured at 600 nm in a plate reader (Thermo Scientific Multiskan FC). Readings from triplicate wells were averaged.

Biofilm antibiotic tolerance assays

Minimal biofilm eradication concentrations were determined by a modification of the Calgary Biofilm Device method (63). A 200 µl aliquot of TSB supplemented with 1% glucose (36) with an approximate starting culture of 10⁵ CFU/mL was added into single wells of the 96-well plates. Each condition was tested in triplicate. The plate was then fitted with a 96-peg lid (Dojindo Laboratories, Mashiki, Japan). After 24 hours of incubation at 37°C (permitting the formation of biofilm on the pegs), the lid was washed with sterile saline and placed onto a fresh 96-well microtiter plate (Nunclon surface; Nunc A/S, Roskilde, Denmark) with each well containing TSB with different concentrations of teicoplanin: (i) without antibiotic; (ii) 0.125 (sub-inhibitory concentrations); (iii) 1 mg/mL (MIC); (iv) 4 µg/mL (resistant MIC breakpoint); (v) 32 µg/mL (target Cmin for endocarditis treatment); (vi) and 1,024 µg/mL (high teicoplanin concentration). The plate was incubated at 37°C overnight. A second wash in sterile saline was performed, and the peg lid was placed into a new Nunc plate with 100 µL TSB and put in an ultrasonic bath (Bandelin-BactoSonic, Berlin, Germany) and sonicated at 60% amplitude for 5 minutes. After that, plate was centrifugated at 800 × g for 20 minutes to remove biofilms from the pegs and the plate was incubated at 37°C overnight in a plate spectrophotometer reader (Thermo Scientific Multiskan FC) recording absorbance at 600 nm after 24 hours, allowing the quantification of viable cells from the biofilms. In addition, serial dilutions were performed to determine the number of CFU per milliliter in each condition.

DATA AVAILABILITY

The corresponding sequences of both isolates were deposited in NCBI database under the Illumina accession numbers SRR25234121 (EF646) and SRR25234122 (EF641), and Nanopore accession numbers SRR27293075 (EF646) and SRR27293074 (EF641).

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/aac.01083-23.

Supplementary Figure S1. aac.01083-23-s0001.tiff.

Lineal illustration of the composite transposon encompassing the bifunctional enzyme aac(6')-aph(2″).

Supplementary Figure S2. aac.01083-23-s0002.tiff.

Macroscopic analysis of colony growth.

Supplementary tables. aac.01083-23-s0003.docx.

Tables S1–S3 and legends of Fig. S1 and S2.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.