Inducible Mega-Mediated Macrolide Resistance Confers Heteroresistance in Streptococcus pneumoniae

ABSTRACT In Streptococcus pneumoniae (Spn), the 5.4 to 5.5 kb Macrolide Genetic Assembly (Mega) encodes an efflux pump (Mef[E]) and a ribosomal protection protein (Mel) conferring antibiotic resistance to commonly used macrolides in clinical isolates. We found the macrolide-inducible Mega operon provides heteroresistance (more than 8-fold range in MICs) to 14- and 15-membered ring macrolides. Heteroresistance is commonly missed during traditional clinical resistance screens but is highly concerning as resistant subpopulations can persist despite treatment. Spn strains containing the Mega element were screened via Etesting and population analysis profiling (PAP). All Mega-containing Spn strains screened displayed heteroresistance by PAP. The heteroresistance phenotype was linked to the mRNA expression of the mef(E)/mel operon of the Mega element. Macrolide induction uniformly increased Mega operon mRNA expression across the population, and heteroresistance was eliminated. A deletion of the 5′ regulatory region of the Mega operon results in a mutant deficient in induction as well as in heteroresistance. The mef(E)L leader peptide sequence of the 5′ regulatory region was required for induction and heteroresistance. Treatment with a noninducing 16-membered ring macrolide antibiotic did not induce the mef(E)/mel operon or eliminate the heteroresistance phenotype. Thus, inducibility of the Mega element by 14- and 15-membered macrolides and heteroresistance are linked in Spn. The stochastic variation in mef(E)/mel expression in a Spn population containing Mega provides the basis for heteroresistance.


RESULTS
mef(E) and mel mRNA expression levels are correlated with erythromycin MIC values. Gene expression levels of mef(E) and mel in Mega containing Spn strains with a range of baseline erythromycin MICs were examined. Resistant Spn strains GA41317 (erythromycin MIC 3 to 4 mg/mL), GA16242 (MIC 32 to 48 mg/mL), and GA71819 (MIC 64 to 192 mg/mL) were examined. Levels of mef(E) and mel expression correlated with MICs (Fig. 1A). The very high-level resistant strain GA71819 demonstrated very high intrinsic erythromycin MICs and corresponding high mef(E) and mel expression levels. In contrast, the low-level resistant Spn strain GA41317 demonstrated the lower erythromycin MICs and low mef(E) and mel expression levels (Fig. 1A), while strain GA16242 was in between.
In previous studies, induction of a low-level resistant strain, GA17457 (MIC 6 to 12 mg/mL), found a 15-fold increase in mef(E) and mel expression using subinhibitory concentrations of erythromycin, and this increase in expression correlated with increased resistance to erythromycin (11). To determine if similar induction occurred in other strains with various starting MIC values, 1/10th MIC erythromycin was used to induce the strains GA41317, GA16242, and GA71819. Figure 1B demonstrates a 250fold induction over uninduced for mef(E) and mel in strain GA41317, and this correlated with a 6-fold increase in MICs. Similar results were seen with the induction of mef(E) and mel expression in strain GA16242 (Fig. 1B) and a 4.5-fold increase in MICs. The exception was strain GA71819, which was already at very high levels of intrinsic mef(E) and mel expression that did not change significantly with induction and did not result in a significant change in MIC.
Mega-containing Spn strains display heteroresistant colonies with Etesting. In screens of Mega-containing Spn clinical isolates to determine erythromycin MICs by Etests, distinct colonies growing in the zone of inhibition of strains GA41317 ( Fig. 2A), GA16242 (Fig. 2B), and GA71819 (Fig. 2C) were observed, suggesting heteroresistance Expression levels of mef(E) and mel were plotted as a percentage of fab(K). (B) Low-and high-level resistant strains GA41317, GA16242, and GA71819 were induced with subinhibitory concentrations of erythromycin before mRNA expression levels and MICs were determined. mef(E) and mel expression is plotted as fold over uninduced values. Etesting was performed in at least triplicate to obtain erythromycin MIC values. to macrolides. To determine stability of the phenotype, colonies were picked from the Etest zone of inhibition in the high-level resistant strain, GA71819, streaked on blood agar plates without selection overnight, and Etesting was repeated. The repeat MICs were consistent with the parent strain (Table 1), indicating the growth of colonies in the zone of inhibition was not due to a fixed mutation or spontaneous tandem duplication event.
To look for gene amplifications, GA71819 was grown in the presence of 1/10th MIC and high levels of antibiotic (192 mg/mL) to select for heteroresistant subpopulations. Genomic DNA was isolated from these cultures. Querying this DNA for copy number of mef(E) and mel revealed that there was no increase in gene copy number (Table 2). Similar results were obtained for the lower-level resistant GA17457 strain (data not shown).
Mega-1 and Mega-2 genotypes in five different insertion sites all display heteroresistance. To determine if Spn strains containing the Mega element in the five most common insertion sites displayed heteroresistance, population analysis profiling (PAP) was performed. The Spn strains chosen represented both Mega-1 and Mega-2, five insertion sites, as well as those displaying both high and low levels of macrolide resistance (Table 3). All strains that contained Mega screened by PAP analysis displayed heteroresistance as defined by a greater than 8-fold difference between the lowest concentration of erythromycin exhibiting maximum inhibition and the highest noninhibitory concentration ( Fig. 3A to F). To verify that the expression of mef(E)/mel from the Mega element was necessary for heteroresistance, XB30, a mutant which does not express mef(E)/mel due to a deletion of the transcriptional start site, was screened and found to be completely sensitive to erythromycin (Fig. 3G). To further quantitate heteroresistance, the area under the curve (AUC) and the proportion of the curve falling beyond the MIC breakpoints (HR-AUC), shown in the gray shadings (Fig. 3), was plotted for the heteroresistant strains determined by PAP (Fig. 3, Table 4). HR-AUC values ranged from 21.31 to 34.39 (X = 25.95) and represented 42% to 100% of the total AUC. The XB30 erythromycin sensitive strain AUC was 4.04.
14-and 15-membered ring macrolide induction of mef(E)/mel eliminates heteroresistance, but noninducing macrolides do not affect heteroresistance. To determine the effect of increasing mef(E)/mel expression in a population-wide manner, the inducibility of the mef(E)/mel genes was exploited. As the Mega system is inducible by subinhibitory levels of erythromycin, five Mega-containing strains (MICs 2 to 12 mg/mL) were first induced and subjected to PAP analysis to determine if heteroresistance was maintained upon induction. High-level resistant strains GA16242 and GA71819 were not queried, as induction of these strains leads to MICs approaching 256 mg/mL, which is the highest concentration of erythromycin queried for the PAP analysis. After 1 h of induction with 0.1 mg/mL erythromycin, MICs were markedly elevated (MICs 16 to 64 mg/mL), and none of the strains now displayed heteroresistance, as there was less than a 4-fold difference between the lowest concentration exhibiting maximal inhibition and the highest noninhibitory concentration (Fig. 4a to e). With erythromycin induction, the HR-AUC values (gray shading, Fig. 4, Table 4) falling beyond the MIC breakpoints ranged from 6.57 to 12.48 and were 12 to 24% of the total AUC (Table 4).
While the mef(E)/mel operon is induced by most macrolides containing 14-and 15membered rings, other macrolides such as the 16-membered ring macrolide tylosin do not induce expression (11). To determine the effect of a non-mef(E)/mel-inducing macrolide such as tylosin on the erythromycin resistance phenotype, GA17457 was incubated with subinhibitory concentration of tylosin for 1 h before a PAP analysis was performed. Incubation with tylosin did not alter the heteroresistance patterns seen in GA17457 (Fig. 5), supporting the hypothesis that the induction of mef(E) and mel is required for heteroresistance.
The mef(E)L leader peptide sequence of the 59 regulatory region is required for heteroresistance and induction. To further explore the relationship between inducibility and heteroresistance, a 59 regulatory region deletion mutant XB31, which was unable to be induced, was used (Table 3). This mutant is still able to express mef(E)/mel at low levels and provide resistance at similar levels to the wild-type uninduced parent strain GA17457. When induced with erythromycin, XB31 MICs are increased less than 2-fold, unlike the 2-to-4-fold increase in erythromycin MICs seen in the parent strain (Table 3). To determine if this strain displayed heteroresistance, a PAP analysis was performed. XB31 displayed traditional resistance patterns indicative of a homogeneously resistant population (Fig. 6A).
As previously shown (22), XB38, a deletion mutant of the 59 regulatory region apart from the mef(E)L peptide-encoding sequence, retains the ability to be strongly induced by subinhibitory concentrations of erythromycin (Table 3). To see if the XB38 mutant was also heteroresistant, a PAP analysis was performed. The presence of the mef(E)L sequence restores a heteroresistant phenotype (Fig. 6B). To confirm that the induction of mef(E)/mel at a transcriptional level followed heteroresistance patterns, the transcript levels for mef(E) and mel were queried in wild-type GA17457 and mutants XB31 and XB38. The uninducible XB31 showed no significant increase in either mef(E) or mel transcript levels (Fig. 7). This is in contrast to wild-type GA17457 and XB38. The fold induction in the XB38 mutant was reduced relative to the wild type, likely attributable to the persistent deletion of 235 bp of the operon's promoter. Both GA17457 and XB38, however, showed significant increases in mef(E) and mel transcript levels (Fig. 7).

DISCUSSION
The Mega element in Spn results in a broad range of observed MICs to macrolides and has been well characterized by genetic structure, gene products, genomic locations, and mechanism of inducibility by macrolides (8,9,11). This study links mef(E) and mel expression levels and MIC and documents heteroresistance associated with the Mega-mediated macrolide resistance phenotype in Spn. The heteroresistance phenotype spans Mega-type, Mega-insertion class, and high and low levels of macrolide resistance, and is dependent upon a Spn strain's ability to express mef(E)and mel.
A range of Mega-containing Spn strains were queried in this study; all were found to contain subpopulations that can persist in the face of high-level macrolides (Fig. 8).
All examined strains are characterized as resistant to macrolides (MIC .1 mg/mL); however, Fig. 8 demonstrates that all strains reach or surpass the breakpoint of 16 mg/mL used as a cutoff for high-level macrolide resistance (23). As Spn is a naturally transformable organism, even more so in the native environment of the NP (24), this is concerning clinically for the spread of resistance mechanisms that could lead to high-level resistance, as could occur in instances of Spn strain cocolonization of the NP (25).
Macrolide heteroresistance and the inducibility of the mef(E)/mel operon are linked in Spn containing the Mega element. More broadly, it is hypothesized that inducible systems are likely to demonstrate antibiotic heteroresistance, which will require future studies. As previously described, the mechanism of macrolide inducibility is dependent   Table 3. The lower limit of detection is indicated on each graph with a dashed dark gray line. Shaded gray areas indicate the region from the highest noninhibitory concentration to the lowest concentration exhibiting maximal inhibition.
Pneumococcal Mega-Mediated Macrolide Heteroresistance Antimicrobial Agents and Chemotherapy on ribosome stalling at the mef(E)L peptide sequence, facilitating stability of the mRNA (22). For mutant XB31, in which the leader peptide sequence has been deleted, there is no operon induction, and in strains in which expression is upregulated via induction, heteroresistance is eliminated. The stochastic induction of mef(E) and mel in a subpopulation leads to increased resistance to macrolides in the inducible Mega-containing strains. This mechanism resembles the stochastic expression in the antibiotic resistance activator mar(A) in E. coli, resulting in heteroresistance (26). The heteroresistant colonies isolated from Mega-containing strains display the same resistance phenotype as the parent strains, and are thus not a result of a gene mutation or gene-amplification. Highly resistant subpopulations resulting from gene mutations or gene amplifications have stable or semistable expression of higher levels of resistance than their parental strains (18,27). Selection for heteroresistant colonies using high concentrations of antibiotic also confirmed that there were no alterations in gene copy number of either mef(E) or mel. Heterogeneity in gene expression is advantageous when populations of bacteria are exposed to environmental fluctuations, including antibiotic selection (28,29). Heterogeneity is typically ascribed to stochastic gene expression, or cell-to-cell fluctuations in expression levels (30). Stochastic expression of an antibiotic resistance gene activator provides transient resistance in single cells in the mar(A) system (26). Transiently resistant phoenix  colonies of Pseudomonas aeruginosa grow within the zone of inhibition of aminoglycoside antibiotics, but numbers of colonies which emerge remain stable from generation to generation, indicating no heritability of high levels of resistance (31) (20). Half of the M-phenotype strains for which MIC data were available were associated with high-level macrolide resistant MICs, and the other half were not (20). Other studies have also shown treatment failures resulting from strains demonstrating the M phenotype (21) or the presence of mef more specifically (33). This is further supported by data presented in a population-based surveillance study of cases of pneumococcal bacteremia, which found that low-level resistance conferred by mef(A) was found to be overrepresented among macrolide failures (36). Heteroresistance associated with the Mega element may contribute to these treatment failures observed with macrolides.
Heteroresistant populations can be successfully treated by combination therapies, as subpopulations which are resistant to one antibiotic are generally independent from those that are resistant to a second antibiotic (37). Combination therapy is already used in the treatment of community-acquired pneumonia in the cases of presence of comorbidities or  risk factors for drug-resistant Spn infection, and treatment with a beta-lactam or tetracycline plus a macrolide is often recommended in these cases (5,38). Our data suggest any Megacontaining Spn strains are capable of heteroresistance and should be considered high level when making clinical treatment decisions.
Heteroresistance is often undetected due to the low frequency of resistant subpopulations, on the order of 1 in 10,000 cells (39). Whole-genome sequencing approaches are enhancing antibiotic resistance predications (40), and whole-genome sequencing is used in monitoring of invasive Spn isolates (41). Understanding the mechanisms underlying heteroresistance allows more accurate predictions of antibiotic resistance based on genomic data. This study adds to the understanding of heteroresistance in Mega-containing Spn isolates.

MATERIALS AND METHODS
Strains and growth conditions. Representative strains were chosen to encompass the full range of mef(E)/mel associated MICs and are listed in Table 1. XB30, XB31, and XB38 were mutants made in the GA17457 background and detailed in Chancey et al., 2015 (22). Spn strains were grown on Trypticase soy agar (TSA) II containing 5% sheep's blood (blood agar) or in Todd-Hewitt broth containing 0.5% yeast extract broth (THY). Plate cultures were grown at 37°C with 5% CO 2 , and broth cultures were grown in a 37°C water bath.
Antibiotic susceptibility. Spn strains were streaked from frozen onto TSAII blood agar plates with or without 0.1 mg/mL erythromycin and grown for 18 to 20 h (11,32). Bacteria from these plates were resuspended in cation-adjusted Mueller-Hinton broth to an optical density of 0.5 McFarland standards. These suspensions were streaked onto Mueller-Hinton agar containing 5% sheep's blood, and erythromycin Etest strips (bioMérieux) were applied. The plates were incubated for 20 to 24 h before reading results.
Population analysis profiling. A modified protocol of the microdilution plating method for population analysis profiling (42) was performed as detailed below. Spn strains were streaked from frozen onto blood agar plates and grown overnight. Primary THY cultures were inoculated to an OD600 of approximately 0.2, and grown for 90 to 120 min. Secondary cultures were inoculated at an OD600 of 0.05 and allowed to grow to early midlog phase (OD600 ;0.3) before half the culture was treated with 1/10th MIC of erythromycin (or tylosin, as indicated) for an hour before both untreated and treated cultures were serially diluted in cation-adjusted Mueller-Hinton broth. Thirty microliters of 21, 23, and 25 dilutions were spotted in triplicate on Mueller-Hinton agar plates with 5% sheep's blood and 2-fold increasing concentrations of antibiotics. Plates were incubated for 20 to 24 h before colonies were counted. The following concentrations of antibiotic were used for erythromycin and tylosin: 0, qPCR. A modified protocol for qPCR was performed based on previously published methods (43). Spn strains were streaked and grown through secondary culture as above, grown in triplicate. One culture remained untreated, one was treated with 1/10th MIC, and the final culture was treated with 192 mg/mL erythromycin. Cultures were grown for 1 h before the bacteria were pelleted at 4K rpm for 5 min and frozen at 280°C. Genomic DNA was isolated from the pellets using the Zymo Quick-DNA Fungal/Bacterial Miniprep kit (Zymo Research) according to the manufacturer's protocol. Genomic DNA concentration was determined using a NanoDrop 8000 spectrophotometer. The samples were diluted to 15 ng/mL, and qPCR was performed with standard curves using 10-fold serial dilutions of untreated genomic DNA. Gene copy numbers were determined relative to the fab(K) control using the DCt method.
qRT-PCR. Spn strains were streaked from frozen onto blood agar plates and grown overnight. Primary THY cultures were inoculated to an OD600 of approximately 0.2 and grown for 90 to 120 min. Secondary cultures were inoculated at an OD600 of 0.05 and allowed to grow to midlog phase. Half of the culture was treated with 1/10th MIC of erythromycin for an hour before both untreated and treated cells were collected and treated with RNAprotect Bacterial Reagent (Qiagen). RNA was isolated using the RNeasy minikit (Qiagen) and subjected to DNA-free DNase treatment (Invitrogen) before cleanup with an RNA clean and concentrator kit (Zymo Research). RNA was quantitated, and equal amounts of RNA from each sample (150 ng) were used to create cDNA libraries using the iScript Reverse Transcription Supermix kit (Bio-Rad). cDNA was diluted 1.5Â before qRT-PCR. qRT-PCR was performed using iQ SYBR green Supermix (Bio-Rad) with a CFX96 real-time PCR detection system. qRT-PCR primers used are listed in Table A1 in the supplemental material. Expression values were calculated using the DDC T method.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.1 MB.