Evolution of Enterococcus faecium in Response to a Combination of Daptomycin and Fosfomycin Reveals Distinct and Diverse Adaptive Strategies

ABSTRACT

Infections caused by vancomycin-resistant Enterococcus faecium (VREfm) are an important public health threat. VREfm isolates have become increasingly resistant to the front-line antibiotic daptomycin (DAP). As such, the use of DAP combination therapies with other antibiotics like fosfomycin (FOS) has received increased attention. Antibiotic combinations could extend the efficacy of currently available antibiotics and potentially delay the onset of further resistance. We investigated the potential for E. faecium HOU503, a clinical VREfm isolate that is DAP and FOS susceptible, to develop resistance to a DAP-FOS combination. Of particular interest was whether the genetic drivers for DAP-FOS resistance might be epistatic and, thus, potentially decrease the efficacy of a combinatorial approach in either inhibiting VREfm or in delaying the onset of resistance. We show that resistance to DAP-FOS could be achieved by independent mutations to proteins responsible for cell wall synthesis for FOS and in altering membrane dynamics for DAP. However, we did not observe genetic drivers that exhibited substantial cross-drug epistasis that could undermine the DAP-FOS combination. Of interest was that FOS resistance in HOU503 was largely mediated by changes in phosphoenolpyruvate (PEP) flux as a result of mutations in pyruvate kinase (pyk). Increasing PEP flux could be a readily accessible mechanism for FOS resistance in many pathogens. Importantly, we show that HOU503 was able to develop DAP resistance through a variety of biochemical mechanisms and was able to employ different adaptive strategies. Finally, we showed that the addition of FOS can prolong the efficacy of DAP and slow down DAP resistance in vitro.

INTRODUCTION

The enormous success of antibiotics is being challenged by the rise of increasingly resistant pathogens that threaten to return us to a preantibiotic era, with potentially devasting consequences (1). The complexity of the antibiotic resistance problem is perhaps best manifested by the increasing number of multidrug-resistant (MDR) pathogens, which include vancomycin-resistant enterococci (VRE) and, increasingly, vancomycin- and daptomycin (DAP)-resistant enterococci (VDRE) (2, 3). Frequently, physicians prescribe daptomycin (DAP) for serious VRE infections (2, 4, 5). DAP acts by inserting itself into the cell membrane of Gram-positive bacteria in a calcium-dependent manner, causing the displacement of critical cell membrane proteins and cell death (6, 7). DAP insertion into the cell membrane is mediated by binding to the anionic phospholipids phosphatidylglycerol (PG) as well as lipid II precursors (7). To date, the two main strategies employed by enterococci in DAP resistance (Dap^r) include (i) an increase in cell envelope charge to decrease the amount of DAP binding (repulsion), which is commonly seen in Staphylococcus aureus (8–10), and (ii) the rearrangement of membrane architecture (redistribution), which is commonly seen in Enterococcus faecalis (11). Genetic changes associated with the redistribution phenotype in E. faecalis occurred in the LiaFSR system, a stress response system involved with cell envelope homeostasis, and cardiolipin synthase (Cls), a protein involved in the synthesis of cardiolipin which is a constituent of the bacterial cell membrane (12–14).

Interestingly, E. faecium exhibits a more diverse ensemble of DAP resistance mechanisms than E. faecalis. While mutations in the LiaFSR pathway remain common, mutations in YycFG, a two-component system pathway involved in cell wall metabolism and restructuring, have also been observed (12–17). Additionally, a previous study also identified mutations in the yvcRS operon, which caused an increase in cell surface charge and decreased DAP binding (13).

Due to the emergence of DAP resistance, new approaches to restore clinical efficacy of DAP have been proposed (2, 13, 14, 18, 19). Combining DAP with other antibiotics is an attractive approach (20–25). Fosfomycin (FOS) has recently joined the list of potential partners with DAP as a DAP-FOS combination (DF), and the combination has already shown promising results against enterococcal infections (26–30). FOS is an antibiotic that acts as an analog of phosphoenolpyruvate (PEP) to inhibit the activity of MurAA, which is an enzyme involved in the first committed step of peptidoglycan synthesis (31). As a drug, FOS has been approved for clinical use orally in the United States and has been used to treat Enterococcus faecium infections. A study by a group in China has discovered that clinical fosfomycin-resistant (Fos^r) E. faecium isolates possess the fosB3 gene, a variant of an enzyme that inactivates FOS through chemical modifications (32, 33). Another study also showed that clinical vancomycin-resistant E. faecium (VREfm) isolates acquire FOS resistance due to a cysteine-to-aspartate mutation in the active site of MurAA (34). These studies show that E. faecium can develop FOS resistance through both mutational resistance and acquisition of exogenous resistance determinants.

In this study, we used in vitro experimental evolution to identify evolutionary trajectories that lead to DF resistance. Using the clinical strain E. faecium HOU503, we showed that resistance to the DF combination was achieved via independent mutations conferring resistance to each drug separately. Moreover, while DF did not prove synergistic against HOU503, this combination did extend the timeline by which E. faecium became resistant to DAP. This delay suggests that DF combinatorial therapy could extend the efficacy of both drugs and delay the rise of resistant variants.

RESULTS

Adaptation to DAP-FOS combination successfully delays the onset of resistance.

To study the effect of a DF combination against E. faecium, we evolved E. faecium HOU503 to resistance with each drug individually or in combination (Fig. 1).

FIG 1.

Different antibiotic gradients lead to different resistance timeline. Five populations of HOU503 were adapted daily to increasing either DAP, FOS, or DF concentrations until resistance was achieved. In FT1 (A), the antibiotic selection gradients were less steep than in FT2 (B), resulting in a longer time required to reach DAP or FOS resistance and the smaller difference between the days taken to reach resistance to a single drug versus a combination. Blue, purple, and green lines denote DAP, FOS, and DF adaptation, respectively.

The DAP/FOS ratios to be used for the experimental evolution studies were determined using a checkerboard assay. Since the DF combination did not exhibit synergistic activity when tested against HOU503 (see Fig. S2 in the supplemental material), the DAP/FOS ratio was set to 0.75:64 μg/mL according to the checkerboard assay result. Ten HOU503 populations were evolved to DAP, FOS, and DF in two groups of five (FT1 and FT2) using in vitro experimental evolution (13, 35–37).

FT1 and FT2 used two distinct selection gradients (moderate and strong, respectively) for the initial single-drug evolution studies (File S2). In FT1, we chose conservative DAP or FOS gradients that did not include the highest DAP or FOS concentration the populations could tolerate. In FT2, the DAP and FOS concentrations were adjusted to include higher concentrations each day that were closer to the highest drug concentration that each population would easily tolerate. In other words, FT1 could have achieved resistance in the same 8- to 10-day period seen in FT2, if we had advanced the increase in the single drug closer to the population MIC. As expected, evolution to resistance using a moderate DAP or FOS gradient (Fig. 1A, blue and purple lines) takes longer than using the stronger gradient (Fig. 1B, blue and purple lines) (38, 39). However, the DF combination selection gradient stayed the same for both FT1 and FT2. The changes in single-drug population studies between the selection gradients can explain the contrast between the difference in the resistance timeline of evolution in response to a single drug versus combination in FT1 and FT2 (further discussed in Text S3). Despite these differences, it is evident that HOU503 experienced more difficulty in adapting to DF combination, taking more time to reach DAP resistance when FOS was added in combination.

At the end of adaptation, endpoint isolates were selected randomly for whole-genome sequencing (WGS) to identify mutations acquired and for further characterization (Tables 1 to 3).

TABLE 1.

Whole-genome sequencing data from DAP-evolved HOU503 isolates^a

Isolate	MIC (μg/mL)a		Mutation(s) inb:						Additional mutationsb,c
	DAP	FOS	pgsA	cls	gdpD	CDP-AP	murF	Pyruvate carboxylase
FT1
D1	64	128		H215R and R211L					+4
D2	64	128		H215R and R211L					+4
D3	64	128		H215R and R211L					+4
D4	64	128				−62	S323R		+6
D5	64	128				−62	S323R		+4
D6	64	170.7		A20D					+4
D7	26.7	128	A98D						+3
D8	32	128							+5
D9	16	128		A20D					+2
D10	64	128				−45		Q1073*	+7
D11	32	128		A20D				Q1073*	+6
D12	9.3	128	V85F					Q1073*	+3
D13	32	170.7		R211L					+4
D14	13.3	128				−62			+4
D15	32	213.3		H215R					+3
FT2
D16	32	128	T70P						+2
D17	26.7	128	T70P						+3
D18	13.3	64	T70P						+2
D19	42.7	64		R211Q					+3
D20	32	128		R211Q					+3
D21	32	64		R211Q					+3
D22	>128	128			L93H	−62			+3
D23	>128	128			L93H	−62			+3
D24	128	128			L93H	−62			+4
D25	21.3	128		H215R					+2
D26	16	64		H215R					+2
D27	16	64		H215R					+2
D28	64	128				−62			+3
D29	64	128				−62			+3
D30	64	64				−62			+3

^aThe MICs reported were averaged over three biological replicates. The HOU503 (LiaR^W73C LiaS^T120A) DAP MIC was 3 μg/mL, and the FOS MIC was 128 μg/mL. * indicate a stop codon.

^bThe selected mutations shown do not include those observed in the no-drug control isolates or initial starting ancestor.

^cThe “Additional mutations” column includes occurrences of van operon (vanRSHAXYZ) loss. D5 and D6 retained one van operon copy, while other isolates lost both copies.

TABLE 2.

Whole-genome sequencing data from FOS-evolved HOU503 isolates

Isolate	MIC (μg/mL)a		Mutation(s) inb:		Additional mutationsb,c
	DAP	FOS	pyk	MurAB
FT1
F1	5.3	>1,024	G135S	−73	+4
F2	4	>1,024	K165R	−73	+3
F3	8	>3,584	P330L	−73	+2
F4	4	>1,024	P12S	−73	+4
F5	5.3	>1,024	G162R	−73	+3
F6	4	>1,024	P330L	−73	+5
F7	6.7	>1,024	P266A	Insertion upstream	+2
F8	5.3	>1,024	P266A	Insertion upstream	+2
F9	4	>1,024	P266A	Insertion upstream	+2
F10	6.7	>1,024	M289I	−73	+4
F11	6.7	>1,024	M289I	−73	+4
F12	5.3	>1,024	M289I	−73	+4
F13	12	512		−73	+4
F14	10.7	597.3		−73	+3
F15	5.3	768		−73	+3
FT2
F16	4	>3,584	L264P	−73	+2
F17	4	512		−73	+3
F18	4	>1,024	G257S	−73	+5
F19	4	>1,024	M319I	−73	+4
F20	13.3	>1,024	G235R	−73	+3
F21	13.3	>1,024	G235R	−73	+4
F22	2.7	>1,024	M256I	−73	+3
F23	4	>1,024	R298L	−73	+2
F24	8	>1,024	R298L	−73	+3
F25	8	>1,024	R298L	−73	+2
F26	4	>1,024	M289I	−73	+5
F27	4	>1,024	E303G	−73	+2
F28	5.3	>1,024	P297R	−73	+3

^aThe MICs reported were averaged over three biological replicates. The HOU503 (LiaR^W73C LiaS^T120A) DAP MIC was 3 μg/mL, and the FOS MIC was 128 μg/mL.

^bThe selected mutations shown do not include those observed in the no-drug control isolates or the initial starting ancestor. Two isolates were not included due to contamination issues.

^cThe “Additional mutations” column includes occurrences of van operon (vanRSHAXYZ) loss. All isolates lost both van operon copies.

TABLE 3.

Whole-genome sequencing data of DF-evolved HOU503 isolates^c

Isolate	MIC (μg/mL)a		Mutation(s) inb:								Additional mutationsb,c
	DAP	FOS	cls	pgsA	CDP-AP	MurAB	pyk	Pyruvate carboxylase	murD	efrB
FT1
DF1	13.3	682.7				Insertion upstream and −73	S201T			P17Q	+4
DF2	10.7	>1,024				−73	P12S				+5
DF3	10.7	>1,024				−73	S201T			P17Q	+4
DF4	26.7	>1,024	A20D			Insertion upstream	I251N				+3
DF5	>32	>1,024			−62	−73	H37L			N56K	+7
DF6	>32	>1,024			−62	−73	H37L				+7
DF7	26.7	>1,024	A20D			−73	R401C				+3
DF8	32	>1,024	A20D			−73	F31S				+3
DF9	>32	682.7	A20D			−73					+5
DF10	13.3	554.7				−73		Q1073*	K161K		+6
DF11	10.7	469.3				−73		Q1073*	K161K		+5
DF12	5.3	1,024				−73		Q1073*			+7
DF13	>32	512			−62	−73 and G204R					+7
DF14	32	256			−62	−73 and −60					+8
DF15	>32	314.3			−62	Insertion upstream and −73					+7
FT2
DF16	>32	>1,024			−62	Insertion upstream	P330S				+5
DF17	>32	>1,024			−62	Insertion upstream	P330S				+4
DF18	>32	>1,024			−62	Insertion upstream	P330S				+6
DF19	8	>1,024			−62	−73	S246F				+4
DF20	10.7	>1,024			−62	−73	S246F				+4
DF21	16	>1,024			−62	−73	S246F				+4
DF22	16	341.3		A63T		Insertion upstream and −73		Insertion upstream			+3
DF23	10.7	>1,024		A63T		−73					+5
DF24	16	256		A63T		−73		Insertion upstream			+3
DF25	16	1,024	A20D			−73					+5
DF26	16	256	A20D			Insertion upstream and −73					+6
DF27	13.3	1,024	A20D			−73					+5
DF28	8	1,024				−73	A28P				+5
DF29	16	768				−73					+6
DF30	8	256				−73	A28P				+4

^aThe MICs reported were averaged over three biological replicates. The HOU503 (LiaR^W73C LiaS^T120A) DAP MIC was 3 μg/mL, and the FOS MIC was 128 μg/mL. * indicate a stop codon.

^bThe selected mutations shown do not include those observed in the no-drug control isolates or initial starting ancestor.

^cThe “Additional mutations” column includes occurrences of van operon (vanRSHAXYZ) loss. DF1 and DF17 retained one van operon copy, while other isolates lost both copies.

Increased MurAB expression as an important step in FOS resistance.

FOS has generally been used against Gram-negative bacteria, but it has returned to prominence due to its potential as a “companion” drug against a wide range of Gram-positive MDR pathogens, including enterococci (27, 40).

In contrast to previous studies, all HOU503 Fos^r isolates evolved mutation(s) affecting MurAB, an isozyme of MurAA, an enzyme involved in peptidoglycan synthesis, as a single nucleotide polymorphism (SNP) and/or an insertion of a transposon upstream of the gene (Tables 2 and 3) (28, 32). In S. aureus, under FOS attack, murAB has been shown to be upregulated and or supplement MurAA activity in maintaining cell wall synthesis (41). To verify if a similar mechanism was employed by HOU503, reverse transcription-quantitative PCR (RT-qPCR) was performed. Due to the challenges associated with engineering mutations in E. faecium to create single mutants, Fos^r isolates from FOS-evolved populations, F5 (MurAB⁻⁷³ pyk^G126R) and F8 (MurAB^Isrt pyk^P266A), were used. Indeed, regardless of the type of mutation, murAB expression was significantly upregulated compared to that of the ancestor (Fig. 2A). The effect of an additional second mutation upstream of MurAB, as seen in DF14 (MurAB^{−60, −73} CDP-alcohol phosphotransferase⁻⁶² [CDP-AP⁻⁶²]) is discussed further in Text S6. The ubiquitous presence of mutations upstream of MurAB suggests that MurAB may play a critical role in FOS resistance in E. faecium (Tables 2 and 3).

FIG 2.

Increased FOS resistance is mediated by upregulation of MurAB expression and increases in PEP flux. (A) qPCR was performed to measure changes in MurAB transcript expression. The assay was conducted in biological triplicates. The housekeeping gene was gyrB. (B) PEP concentration was measured through the PEP colorimetric/fluorometric assay kit (Sigma-Aldrich). Experiments were conducted in biological triplicates with technical duplicates.

Increased PEP level as a mechanism of FOS resistance.

The other mutation facilitating FOS resistance is a mutation in the gene (pyk) coding for pyruvate kinase, an enzyme involved in production of pyruvate from PEP (Tables 2 and 3). As PEP is the substrate for MurAA, and a structural analog of FOS, we postulated that the mutations in pyk decrease the flux of PEP to pyruvate, allowing more free PEP to compete against FOS for the MurAA active site. To address this hypothesis, an assay measuring PEP level was conducted on selected isolates with and without pyk mutations.

Isolates that include a pyk mutation, F3 (MurAB⁻⁷³ pyk^P330L) and F16 (MurAB⁻⁷³ pyk^L264P), showed about a 2-fold increase in PEP concentration. Conversely, F17 (MurAB⁻⁷³), an isolate without pyk mutations, has similar levels of PEP to the ancestor (Fig. 2B). This result suggests that adaptive mutations in pyk probably decrease its activity, resulting in a net increase in free PEP, leading to increased FOS resistance.

Moreover, there is a significant difference in FOS MICs between isolates with and without mutations in pyk. The presence of pyk mutations increased the FOS MIC of isolates by more than 7-fold (from 512 to >3,584 μg/mL) (Table 2). Interestingly, the effect of PEP on FOS resistance seems to be concentration dependent as the isolate with more PEP, F16, exhibited a shorter lag time during growth studies than F3 during FOS stress (Text S5 and Fig. S3). Overall, these results suggest that adaptive mutations in pyk can act as important drivers for E. faecium to achieve higher FOS resistance.

HOU503 achieved DAP resistance through biochemically distinct adaptive strategies.

Previous studies on DAP resistance in enterococci have identified adaptive changes consistent with membrane composition remodeling (11, 42, 43). Indeed, with background mutations in LiaRS, Dap^r HOU503 isolates gained mutations in genes involved in membrane lipid synthesis (CDP-AP, pgsA, cls, and gdpD) (Table 1). Taken together, these mutations suggest that DAP resistance in HOU503 was facilitated by changes to cell membrane lipid flux. However, changes in the homeostasis of a single lipid species can have pleiotropic effects on other lipid pools and membrane fluidity, which have recently been implicated in the antimicrobial action of DAP (6, 10, 44). To provide a more holistic assessment of lipid homeostasis during adaptation to DAP, we quantitated changes in (i) membrane lipid pools (lipidomic analyses), (ii) membrane fluidity (Laurdan dye assay), (iii) anionic lipid displacement (N-nonyl acridine orange [NAO] staining), and (iv) DAP binding (Bodipy-FL–DAP [BDP:DAP] staining).

As expected, phosphatidylglycerol (PG) levels play an important role in DAP resistance. D19 that includes cls^R211Q showed a significant increase in the amount of cardiolipin (CL), but lower PG levels than the ancestor (Fig. 3A and B). This is consistent with a previous study that showed DAP adaptive mutations within cls cause increased cardiolipin synthesis activity (42). Additionally, as cardiolipin is synthesized from PG, its increased activity would also explain the significantly smaller amount of PG (Fig. 3A). These changes in PG and CL levels extended across the range of lipid species, consistent with generalized changes in the PG/CL pools. HOU503 also seems to attain increased DAP resistance through decreased PgsA activity, as seen in D16. As PgsA catalyzes the formation of PG precursor, its decreased activity could explain the smaller amount of PG in D16 (pgsA^T70P) (Fig. 3A). Intriguingly, strains harboring either pgsA^T70P or cls^R11Q correlate with lipid patterning and localization, consistent with a redistribution mechanism in which anionic phospholipids are moved away from the division septa (Fig. 4B).

FIG 3.

DAP resistance in HOU503 is correlated with changes to membrane lipid composition. Abundance of (A) phosphatidylglycerol (PGs), (B) cardiolipins (CLs), (C) diglucosyl-diacylglycerols (DGDGs), and (D) lysl-phosphatidylglycerol (LysylPGs) was normalized to dry bacterial pellet weight and internal standard ion abundance. Individual lipid species are represented as the number of carbons:degree of unsaturation in the fatty acid chains. The figure is presented as a heat map of log₂ fold change over the ancestor. Significance was calculated with Student's t test: *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001; ****, P ≤ 0.0001. For primary data, refer to Fig. S4 in the supplemental material.

FIG 4.

DAP resistance in E. faecium HOU503 displays diverse phenotypes consistent with the identified biochemical mechanisms. (A) Increased CDP-AP transcript expression is observed in mutants with an SNP located upstream of CDP-AP. (B) Visualization of anionic phospholipids in the cell membrane using NAO (250 nM) staining of representative cells of HOU503 and its Dap^r isolates. The top images in the panel show bright-field microscopy, while the bottom row shows fluorescent images (bars, 1 μm). (See Fig. S5 for more images.) (C) Quantification of Bodipy-labeled DAP (BDP:DAP) binding of HOU503 and selected Dap^r derivatives when stained with 8 μg/mL BDP:DAP. (See Fig. S6.) (D) Quantification of the mean baseline fluidity as calculated through the fluorescent Laurdan dye assay for HOU503 and its Dap^r derivatives. Higher GP values indicate more rigid membranes and vice versa. (See Fig. S7.) Significance for panels C and D was calculated using the Mann-Whitney test with post hoc Holm-Bonferroni adjustment and Student's t test, respectively. *, P ≤ 0.05; **, P ≤ 0.01; *****, P ≤ 0.00001.

Mutations were also identified upstream of a gene annotated as a CDP-AP family protein (FGF98_RS11550), but whose function is unknown (Table 1). CDP-AP family proteins include proteins involved in the biosynthesis of essential cell membrane lipids (refer to Text S4 for more information) (45). As the mutation occurred upstream of the putative CDP-AP gene, we carried out RT-qPCR and found a 3- to 5-fold increase in expression among isolates carrying the mutation (D14 and D23) (Fig. 4A). D14 (CDP-AP⁻⁶²) has a significantly greater amount of most species of PG and lysyl-PG compared to the ancestor (Fig. 3A and D, respectively). Interestingly, the NAO staining of isolate D14 resembles that of the ancestor HOU503 and does not suggest a phospholipid redistribution-like mechanism for DAP resistance (Fig. 4B). Instead, the lack of phospholipid redistribution coupled with the drastic increase in Lysyl-PG levels (Fig. 3D) and decreased BDP:DAP binding (Fig. 4C) suggests a resistance mechanism that may be more similar to electrostatic repulsion (11).

Interestingly, when CDP-AP⁻⁶² is paired with adaptive changes in gdpD, DAP resistance markedly increases. D23 (CDP-AP⁻⁶² gdpD^L93H) has a significantly smaller amount of both PG and lysyl-PG (Fig. 3A and D, respectively) and exhibited a more redistribution-like phenotype with a rearrangement in the cell membrane architecture (Fig. 4B). Moreover, the addition of gdpD^L93H also boosted the DAP MIC by >9-fold (13.3 versus >128 μg/mL) (Table 1).

Significantly, despite the variety of mutations these isolates carry, all tested isolates have a significantly more rigid membrane and bound significantly less BDP:DAP than the ancestor, HOU503 (Fig. 4C and D). This decrease in membrane fluidity may influence the way DAP can insert itself into the membrane, as well as the amount it can insert, thus leading to decreased BDP:DAP binding and overall greater resistance to DAP “attack.” In general, these data showed that although E. faecium HOU503 can utilize multiple pathways in becoming DAP resistant, overall changes in PG and a more rigid membrane seem to be consistent phenotypes (6, 7).

HOU503 evolved resistance to DF combination through independent pathways.

As mentioned above, DAP and FOS affect the cell membrane and cell wall, respectively. One of the reasons we chose to study DF combination was to determine if the genetic drivers for DF resistance might be epistatic and thus potentially decrease the efficacy of a combinatorial approach in either inhibiting VREfm or delaying the onset of DAP resistance. As shown in Table 3, we did not observe any evidence of substantial cross-drug epistasis that could undermine a combinatorial approach—i.e., we could not identify any unique mutations in DF-resistant isolates that we believe would give rise to DAP and FOS resistance simultaneously. Changes within populations adapting to the single drugs separately did not markedly increase their resistance to the other. Instead, we saw that resistance to DF was achieved independently with mutations in cell wall synthesis and central metabolism for FOS and in altering cell membrane lipid dynamics for DAP (Fig. 5).

FIG 5.

Representation of enterococcal cell envelope synthesis and the DAP, FOS, and DF resistance pathway. Cell wall (CW) synthesis was started with the peptidoglycan precursor UDP-NAM pentapeptide, which was synthesized through the sequential action of MurA to MurF. UDP-NAM pentapeptide was then transferred to the lipid carrier C₅₅P, forming lipid I, which was then used to form lipid II. Meanwhile, PG and cardiolipin, which are integral parts of cell membrane (CM) lipid, were synthesized through numerous precursors and various proteins, including those in the CDP-AP family. These two pathways were joined through the central metabolism pathway (7, 60). DAP resistance was mainly mediated by mutations to proteins involved in the membrane lipid synthesis pathway (blue dashed box), while FOS resistance was facilitated through mutations in the cell wall synthesis and central metabolism pathway (purple dashed box). Meanwhile, DF resistance was achieved independently through changes to both pathways separately (green dashed boxes). Created with Biorender.

DISCUSSION

MDR pathogens like VRE are increasingly responsible for difficult-to-treat infections that require the use of drugs of last resort, such as DAP. Due to the occurrence of DAP-resistant E. faecium infections, strategies to prolong or augment the activity of this antibiotic, such as combination therapy, are needed (5, 46). FOS has been shown to exhibit synergy with DAP against other Gram-positive bacterial infections (40, 47). We investigated a DAP-FOS drug combination against a clinical strain of E. faecium to discern whether the combination may delay the onset of DAP resistance or, in the worst-case scenario, facilitate the quick ascent of strains with resistance to both.

Although the DF combination did not exhibit synergistic activity against HOU503, the onset of DAP resistance was successfully delayed. Our in vitro evolution studies showed that DF resistance in HOU503 occurred independently, with no clear evidence of mutations that could act as epistatic genetic drivers providing resistance to both drugs simultaneously. The lack of cross-drug epistasis also supports the potential efficacy of a DF combinatorial approach in delaying the onset of DAP resistance.

Previous studies have shown that DAP resistance in enterococci can be achieved through “repulsion” of the antibiotic molecule or the rearrangement of the cell membrane architecture (8, 9, 11). Interestingly, E. faecium has shown its capability to utilize either mechanism, which is influenced by several factors, including genetic background and adaptive environment (13). In this article, we showed that DAP-resistant E. faecium isolates derived from the same ancestor and evolved in the same environment were able to utilize both resistance strategies.

Isolates that exhibited a “repulsion”-like mechanism acquired a mutation upstream of a CDP-AP family protein (FGF98_RS11550), which includes those involved in the biosynthesis of cell membrane lipids (45). As the isolate with an increase in CDP-AP expression, D14, also had an increase in its lysyl-PG levels, we initially hypothesized that CDP-AP may have similar functions to the product of mprF, which catalyzes the addition of lysyl group to PG (10, 13). However, owing to its low homology (less than 8%) to either mprF1 or mprF2 and an increase in D14’s PG levels as well, it is very likely that CDP-AP has a different biochemical role and is in a different protein family from the protein coded for by mprF. Additionally, as D14 has an overall increase in both PG and lysyl-PG levels, it is more likely that CDP-AP may play a role upstream of PG synthesis and thus the downstream lysyl-PG. Interestingly, an extra mutation in gdpD in D23 (CDP-AP⁻⁶² gdpD^L93H) reverses the phenotype of the resistance mechanism to be more redistribution-like, with a rearrangement in the cell membrane architecture. Moreover, when paired with mutation to gdpD, the DAP MIC increases by at least 9-fold to over 128 μg/mL. We speculate that the mutation observed in gdpD led to a decrease in its enzymatic activity and subsequently glycerol-3-phosphate, which may explain the decline in D23’s PG and lysyl-PG levels.

Furthermore, HOU503 also exhibited cell membrane architecture rearrangement when it acquired mutations in either cls or pgsA. The observed cls mutation is consistent with our earlier study, suggesting that adaptive changes in cls in response to DAP are correlated with an increase in Cls activity (42). Although insufficient by itself, adaptive mutations in cls when combined with preexisting mutations in the LiaFSR regulon seem to be sufficient for DAP resistance (48). On the other hand, mutations in pgsA seem to decrease its enzymatic activity, leading to a decrease in PG content. This decrease in available PG likely may contribute to resistance by providing less opportunity for the critical PG-dependent DAP binding event that is part of the initial DAP interaction with the cell membrane (7).

Remarkably, despite the different adaptative mechanisms seen in DAP-resistant E. faecium, each of the endpoint isolates bound less BDP:DAP than the ancestor. Additionally, although the Dap^r isolates have different membrane lipid compositions, they all show evidence for a more rigid cell membrane than the ancestor. Both strains that included upregulation of CDP-AP (D14 and D23) showed larger changes in lipids with longer chain lengths, and this may be a clue about the potential role or substrate preference of CDP-AP (Fig. 3). As membrane fluidity has been implicated previously in DAP binding, it is possible that these changes in membrane rigidity decrease DAP binding to the cell, leading to less BDP:DAP being bound (6).

In contrast, mutations observed during adaptation to FOS were less diverse. FOS acts by competing with PEP for the MurAA active site, and the mutations observed were focused on either MurAB, a homolog of MurAA, or pyruvate kinase (pyk). The mutations observed in pyk are likely decrease-of-function mutations as intrinsic PEP levels in isolates with pyk mutations were higher than in the ancestor. Conversely, mutations upstream of MurAB resulted in a significant upregulation of MurAB expression. A study in S. aureus showed that MurAB was able to compensate for, or complement, MurAA function (41). Our qPCR data suggest that HOU503 may have employed a similar mechanism. However, this hypothesis requires further studies to be validated, especially regarding whether MurAB can catalyze the same reaction as MurAA. Mutations upstream of MurAB were ubiquitous across all our populations, suggesting that it is a critical contributor to FOS resistance. Interestingly, while HOU503 possesses the fosB3 gene (FGF98_01880), which was implicated in E. faecium FOS resistance previously, we did not observe any mutations affecting this gene (28, 32). Additionally, we also did not observe mutations affecting MurAA, which had been documented previously in Fos^r VREfm isolates (34). The lack of mutation within MurAA in DF-resistant isolates was surprising. MurAA is a promising target to achieve FOS resistance (34). Additionally, mutations in MurAA had been seen previously in a Dap^r E. faecium isolate lacking the liaR response regulator, suggesting it may contribute to cellular fitness during DAP exposure (35). This absence implies that resistance strategies rely heavily on the genetic background of the strain.

In summary, we show that E. faecium evolved resistance to DF combination independently, without evidence of epistatic mutations that affect both drugs simultaneously (Fig. 5). DAP resistance in HOU503 was mediated by changes to the cell membrane lipid profiles and fluidity. Despite the different mutations observed, overall changes to PG content and higher membrane rigidity were shared among all the DAP-resistant isolates tested. This suggests that these may be the consistent phenotypes for DAP-resistant E. faecium HOU503. On the other hand, the increases in MurAB expression and PEP levels were fundamental to FOS resistance in HOU503. Despite lacking synergistic activity, the DF combination was able to extend the timeline in HOU503 to achieve DAP resistance, signifying its potential for the treatment of E. faecium infections clinically.

MATERIALS AND METHODS

Bacterial strain and growth conditions.

E. faecium HOU503 is a VRE clinical strain that harbors common LiaRS substitutions (LiaR^W73C and LiaS^T120A) with DAP and FOS susceptibility (DAP MIC, 3 μg/mL; FOS MIC, 128 μg/mL) (49, 50). The FOS MIC breakpoint was extrapolated from the CLSI breakpoint for E. faecalis. All cultures were grown in brain heart infusion (BHI) at 37°C with shaking at 225 rpm unless noted otherwise. CaCl₂ was used at 50 mg/L unless noted otherwise. Glucose-6-phosphate was not added as it did not create noticeable differences in the FOS MIC values (see Text S1 and S2 and Fig. S1 in the supplemental material).

In vitro adaptation of E. faecium HOU503 to DAP and FOS.

E. faecium HOU503 was adapted to DAP and FOS individually as well as in combination with five populations per condition. The evolution was conducted in the following manner. Single colonies were used to start the adaptation. Each day, a 1% dilution of the culture with the best growth was passaged to a new tube with an increasing antibiotic concentration. Serial passage was continued until resistant colonies (MIC of ≥2× CLSI breakpoint) were obtained. One population served as a no-drug control. At the end of adaptation, each population was serially diluted onto nonselective BHI plates. Three isolates from each population were selected at random for further analysis. The evolution was conducted in duplicate. Two different antibiotic selection gradients were used in FT1 and FT2. In the first replicate (FT1), a slower increase in the antibiotic concentration for the single-drug evolutions was employed than for the second replicate (FT2). Incremental increases in the antibiotic concentrations used in FT1 and FT2 can be found in File S2.

Determination of DAP/FOS ratio using a checkerboard assay.

A 96-well plate containing the first antibiotic (DAP) along the ordinate and the second antibiotic (FOS) along the abscissa was prepared. Five microliters of inoculum that was standardized to an optical density at 600 nm (OD₆₀₀) of 0.05 were used. The plates were incubated overnight. The fractional inhibitory concentration (FIC) index value was calculated according to Chen et al. (32).

Determination of endpoint isolates’ fitness and MIC.

Fitness of each endpoint isolate was tested by comparing the growth curve of each resistant isolate to HOU503. Briefly, overnight cultures were normalized to an OD₆₀₀ of 0.05, and 5 μL was used to inoculate a microplate with fresh BHI. The OD₆₀₀ was measured every 5 min at 37°C for 24 h in a BioTek EpochII microplate reader. The DAP and FOS MICs of all endpoint isolates were tested using an agar dilution assay in BHI agar with CaCl₂ (51). Both experiments were conducted in biological triplicates.

Determination of PEP concentration.

PEP quantification was performed with the PEP colorimetric/fluorometric assay kit (Sigma-Aldrich) according to the manufacturer’s instructions, with modifications. Briefly, mid-log-phase cultures were pelleted and lysed using the bead mill method. The resulting supernatant was used for both the PEP and Bradford assays. Subsequent steps were done as directed by the manufacturer’s protocol. The assay was conducted in half-area 96-well plates, and fluorescence was measured in a BioTek EpochII microplate reader at 570 nm. The data were normalized to the protein concentration obtained from the Bradford assay. The assay was conducted in biological triplicates and technical duplicates.

Determination of mRNA expression by qPCR.

Total RNA was extracted using the Qiagen RNeasy minikit, with additional incubation of samples with 5 μL of 5 U/mL mutanolysin and 12.5 μL of 200 mg/mL lysozyme at 37°C for 30 min. Samples were then treated with Turbo DNase according to the manufacturer’s protocol, and the Thermo Fisher SuperScript III kit was used to synthesize cDNA. qPCR was performed according to the manufacturer’s protocol, using the Bio-Rad iQ SYBR green master mix on a Bio-Rad CFX Connect real-time system. gyrB was used as the internal reference. Relative expression was calculated using the threshold cycle (2^−ΔΔCT) method. This assay was conducted in biological and technical triplicates. Primer sequences are shown in Table S1.

Membrane phospholipid distribution and BODIPY-DAP binding.

The membrane dye 10 N-nonyl acridine orange (NAO) was used to visualize anionic phospholipids according to a previously published protocol, with several modifications (11). Briefly, cells were grown to mid-log phase in tryptic soy broth (TSB) and then incubated with 250 nM NAO at 37°C with shaking in the dark for 3.5 h. Cells were then washed in saline three times and immobilized on 1% agarose pad.

Conjugation of Bodipy-FL (BDP) to DAP was carried out as described previously (52). Briefly, BDP and DAP were incubated in 0.2 M sodium carbonate buffer (pH 8.5) with shaking at room temperature for 1 h. This was followed by dialysis against distilled H₂O at 4°C overnight. The resulting BDP:DAP was used as a proxy to visualize binding of DAP to cells according to the previously published protocol (11, 13, 53). Cells were grown until mid-log phase in BHI with CaCl₂. Then, cells were incubated with BDP:DAP in the dark with shaking at 37°C for 20 min. After staining, cells were washed once with HEPES (20 mM [pH 7.0]) and immobilized on a 1% agarose pad.

Fluorescence images of both assays were taken on a Keyence BZ-Z710 microscope with a 100× objective using a fluorescein isothiocyanate (FITC) filter (excitation, 490 nm; emission, 528 nm). Both assays were conducted in triplicate.

Membrane fluidity analysis by in vivo Laurdan generalized polarization spectroscopy.

Laurdan (6-dodecanoyl-2-dimethylaminonaphthalene) is a membrane dye used to determine alterations of membrane fluidity (54). Laurdan generalized polarization (GP) spectroscopy was performed according to a previously published protocol (55). Briefly, mid-log-phase culture in BHI supplemented with 0.1% glucose was incubated with 10 μM Laurdan in a light-protected box with shaking at 37°C for 10 min. The sample was then washed four times with phosphate-buffered saline (PBS) supplemented with 0.1% glucose and CaCl₂. After the final wash, the sample was resuspended in PBS-glucose-calcium solution. The assay was conducted with four technical replicates of each condition (untreated and ethanol) and two replicates of the control Laurdan background. Baseline fluorescence intensity was measured with excitation at 350 nm and emission at 435 nm and 500 nm for 5 min at 37°C on a Synergy H1 spectrophotometer (BioTek). Cells were then treated with 10% ethanol as a membrane fluidizer. Measurements were taken every 2 min, as described above, for the next 15 min. The Laurdan GP index was calculated with the formula (I₄₃₅ – I₅₀₀)/(I₄₃₅ + I₅₀₀), where I is the fluorescent intensity at the specified wavelength.

Lipid extraction for lipidomics.

Lipid extraction was carried out as described previously, with minor modifications (56). Briefly, cells were grown without the presence of antibiotics before they were pelleted and processed. Additionally, dried extracts were reconstituted in 500 μL of 1:1 chloroform-methanol solution mix. For analysis, 10 μL of the lipid extract per 10 mg of original pellets was transferred to a liquid chromatography (LC) vial, dried under argon or nitrogen, and reconstituted to 100 μL with 2:1 acetonitrile-methanol solution mix.

Membrane lipid compositions by hydrophilic-interaction liquid chromatography (HILIC) ion mobility-mass spectrometry (IM-MS).

For liquid chromatography, bacterial lipids were separated by a Waters ultraperformance liquid chromatography (UPLC) device (Waters Corp., Milford, MA, USA) as described previously (44, 56, 57).

The Waters Synapt XS platform was used for lipidomics analysis, with similar parameters and columns to those described by Zhang et al. (56).

Data analysis for HILIC-IM-IS data.

Data alignment, chromatographic peak detection, and normalization were performed in Progenesis QI (Nonlinear Dynamics). A pooled quality control sample was used as the alignment reference. The default “All Compounds” method of normalization was used to correct for variation in the total ion current among samples. Lipid identifications were made based on m/z (within 10 ppm mass accuracy), retention time, and collisional cross section (CCS) with an in-house version of LipidPioneer, modified to contain the major lipid species, including free fatty acids (FFAs), diglucosyl-diacylglycerols (DGDGs), PGs, CLs, and lysyl-PGs, with fatty acyl compositions ranging from 25:0 to 38:0 (ratio of total carbons to total degree of unsaturation), and LiPydomics (44, 58, 59). The quantitation of FFAs was performed based on their ion mobility retained chromatographic peaks using Waters DriftScope 2.5 and TargetLynx (Waters Corp., Milford, MA, USA).

Whole-genome sequencing and analysis.

Genomic DNAs of HOU503 and endpoint isolates were isolated using the Qiagen DNeasy UltraClean microbial kit following the manufacturer’s instructions, with additional incubation at 37°C in the microbead solution with 5 μL of 5 U/mL mutanolysin and 12.5 μL of 200 mg/mL lysozyme for 1 h. Paired-end libraries were generated using the Plexwell 384 library generation kit. Samples were sequenced by a commercial facility with 2× 150-bp reads on Hi-seq. Endpoint isolates were sequenced with a minimum coverage of 100×, while populations were sequenced with at least 300× coverage. Illumina sequences were compared to the closed ancestor genomes (503F_del_LiaR under BioProject no. PRJNA544687) using the Breseq genomic pipeline. Population samples were processed using the polymorphism (-p) mode.

Data availability.

All genomic sequences are available under BioProject no. PRJNA789547.