Evolution and Global Dissemination of Macrolide-Resistant Group A Streptococci

Abstract

Macrolide-resistant group A streptococci (MRGAS) have been recovered from many countries worldwide. However, the strain typing information that is available has been insufficient for estimating the total number of macrolide-resistant clones, their geographic distributions, and their evolutionary relationships. In this study, sequence-based strain typing was used to characterize 212 MRGAS isolates from 34 countries. Evaluation of clonal complexes, emm type, and resistance gene content [erm(A), erm(B), mef(A), and undefined] indicate that macrolide resistance was acquired by GAS organisms via ≥49 independent genetic events. In contrast to other collections of mostly susceptible GAS, genetic diversification of MRGAS clones has occurred primarily by mutation rather than by recombination. Twenty-two MRGAS clonal complexes were recovered from more than one continent; intercontinental strains represent nearly 80% of the MRGAS isolates under study. The findings suggest that horizontal transfer of macrolide resistance genes to numerous genetic backgrounds and global dissemination of resistant clones and their descendants are both major components of the present-day macrolide resistance problem found within this species.

Penicillin and related β-lactams are usually the antibiotics of choice for the treatment of infections caused by Streptococcus pyogenes (group A streptococci [GAS]). GAS have remained susceptible to penicillin despite decades of exposure (22). Macrolides are preferred for treatment of GAS infections in patients with β-lactam hypersensitivity or chronic, recurrent pharyngitis due to prior treatment failure (26). Clindamycin (a lincosamide) is recommended for patients with life-threatening soft-tissue infections, such as toxic shock syndrome or necrotizing fasciitis, because it halts the exotoxin production that can lead to extensive tissue necrosis, shock, and multiple-organ failure (6).

Macrolide-resistant GAS (MRGAS) is an increasingly recognized problem in many parts of the world. Two mechanisms account for most macrolide resistance in GAS. Target site modification by methylases, namely, Erm(A) subclass TR and Erm(B), leads to ribosomal modification and loss of binding by macrolide, lincosamide, and streptogramin B (MLS) antibiotics (29, 51). Macrolide efflux is mediated by MefA (M phenotype), resulting in resistance to 14- and 15-membered, but not 16-membered, ring macrolides (54). The three major resistance genes (R genes) found in GAS, erm(A), erm(B), and mef(A), are associated with mobile genetic elements (3, 18). Mutations in 23S rRNA and the L4 ribosomal protein also seem to confer macrolide resistance in at least some strains (11, 33, 44).

Studies in Japan, Finland, and elsewhere show a strong correlation between national macrolide consumption and resistance in GAS (14, 15, 17, 35, 41, 50). Between 1998 and 2001, a statistically significant increase in GAS resistant to erythromycin and azithromycin was observed in Spain (2). In Toronto between 1997 and 2001, macrolide resistance in GAS recovered from throat swabs increased nearly sevenfold (27). In a recent 3-year longitudinal surveillance study of GAS throat isolates from children in Pittsburgh, 100% of recovered GAS were macrolide sensitive until the third year of study, when macrolide resistance rose to 48% of all GAS isolates (34). In the Pittsburgh study, the outbreak of MRGAS was due to a single clone. In contrast, a recent multicenter study of GAS pharyngitis in eight communities within the United States showed a macrolide resistance rate of <5%, with some variation in rates between sites or within one site over multiple years; 120 MRGAS isolates representing 14 emm types were identified (56). In another recent analysis, 129 isolates of MRGAS obtained from 45 medical centers within the United States were accounted for by 28 emm types and 44 pulsed-field gel electrophoresis patterns (44).

The extent to which macrolide resistance reflects the emergence of new MRGAS clones from macrolide-sensitive GAS (MSGAS) versus the expansion of widely circulating MRGAS clones is largely unknown. The goals of this study were to begin to identify and catalog the globally disseminated MRGAS clones by using highly portable sequence-based typing tools, estimate the number of times macrolide resistance arose in GAS, and characterize the evolutionary relationships among MRGAS and the mechanisms for their genetic diversification.

MATERIALS AND METHODS

Bacterial isolates.

The 212 isolates of GAS under study are listed in Table S1 in the supplemental material and are deposited at http://www.mlst.net. The year, country, and tissue site of isolation are indicated, as well as the strain provider (7, 10, 58). Isolates were chosen for study primarily based on their country of origin in order to compile a geographically diverse set. The macrolide resistance phenotype and/or R-gene content was provided for many isolates by the contributing investigator; this information guided the final selection of isolates, which was designed to maximize unique combinations of macrolide resistance phenotype or genotype and country of origin. Isolates from Canada (MDS series; n = 25) and Taiwan (TW series; n = 16) had also been partly characterized by pulsed-field gel electrophoresis pattern and/or emm type; this information was considered in selecting isolates in an attempt to increase overall genetic diversity. Previous findings on macrolide resistance phenotype and genotype underwent confirmatory tests as described below.

Antibiotic resistance phenotype.

A double-disc agar diffusion test was used to determine the macrolide resistance phenotype of each isolate. Freshly grown bacteria were suspended in Todd-Hewitt broth to a turbidity comparable to a 0.5 McFarland standard and were inoculated onto Mueller-Hinton agar plates supplemented with 5% defibrinated sheep blood. Erythromycin (15 μg) and clindamycin (2 μg) discs were placed 15 to 20 mm apart on the plates and incubated for 18 h at 37°C in 5% CO₂. Resistance to both antibiotics defined the constitutive MLS (cMLS) phenotype. Resistance to erythromycin and a “D”-shaped zone around the clindamycin disc defined the inducible MLS (iMLS) phenotype. Resistance to erythromycin, with susceptibility to clindamycin, defined the M phenotype. Additional GAS isolates, not included in the set of 212 MRGAS but having related genotypes of interest, were screened for macrolide susceptibility using E-test strips containing erythromycin (AB Biodisk, Piscataway, NJ); MICs ≤ 0.5 μg/ml were scored as susceptible.

R-gene determination.

A multiplex PCR assay was used to determine the R-gene content of each isolate. Thermal cycling conditions used an annealing temperature of 58°C. An ∼348-bp internal fragment of mef(A) was amplified with oligonucleotide primers as described previously (53). An ∼426-bp fragment of erm(A) (subclass ermTR) was amplified with primers 5′-CAT GAG GAT ATT TTG AAG TTT AGC TTT CCT AA-3′ and 5′-CAG TAA CAT TCG CAT GCT TCA GCA CCT GTC TTA ATT G-3′. An ∼623-bp fragment of erm(B) was amplified with primers 5′-CGA GTG AAA AAG TAC TCA ACC AAA TAA TAA AAC AAT TG-3′ and 5′-GCT CAT AAG TAA CGG TAC TTA AAT TGT TTA CTT TGG CGT G-3′. An ∼241-bp fragment of the slo gene, encoding streptolysin O, was amplified as a positive control (primers 5′-AAT ATC AAC ACT ACA CCA GT-3′ and 5′-CTG TTG AAA CAT TGG CAT AG-3′). PCR amplification products targeting the 3′ end of domain V of 23S rRNA and the rplD gene (33) underwent nucleotide sequence determination.

Nucleotide sequence determination of virulence genes.

emm type and emm subtype were ascertained as described previously (31). Different emm-type assignments were given for genes with <92% sequence identity over the first 90 bp encoding the deduced processed M protein of the type reference strain. New emm subtypes, defined as partial alleles spanning the first 150 bp encoding the mature M protein, were deposited in a publicly available database (4). The presence or absence of the prtF1 gene was established using oligonucleotide primers targeting the 5′-end region of the gene (38); the nucleotide sequence was determined for all PCR products. Newly identified partial prtF1 alleles, based on the nucleotide sequence of an aligned and trimmed region of ∼333 bp, were assigned GenBank accession numbers DQ470016 to DQ470037.

Multilocus sequence typing (MLST).

MLST was performed as described previously (8). MLST of GAS is based on nucleotide sequence variation at seven housekeeping genes (gki, gtr, murI, mutS, recP, xpt, and yqiL). Newly described alleles and sequence types (STs), defined as unique combinations of alleles, were deposited in a publicly available database (1).

Diversity indices.

To quantify diversity (D) among the GAS isolates, statistics were used that incorporate the number of clones and their frequency of occurrence, whereby a clone is defined in this report as a unique ST and R-gene combination. n_i and p_i are the number and proportion of isolates of the ith clone, respectively, N is the total number of isolates, and S is the total number of clones. A measure of diversity (Simpson's index) is D = 1 − Σp_i², with the following range: 0, 1 − 1/S. A less biased measure of diversity proposed for use in clinical microbiology is D = 1 − {Σn_i[n_i − 1]/[N(N − 1)]}, with the following range: 0, 1 (23). Both measures estimate the probability that two isolates selected from the sample at random are different clones, whereby a D value of 1.0 indicates that the clonal typing method is able to discriminate between all isolates and a D value of 0.0 indicates that all isolates are the identical clone. Confidence intervals (CI) were determined for the less biased D measure (20).

Two additional statistical measures were made to assist with quantitative comparisons of diversity. The effective number of clones (S_e) is S_e = 1/Σp_i², with the following range: 1, S. This measure has been widely used in the contexts of population genetics (40) and ecology (28) and is the number of equally frequent clones that will produce the observed diversity. The evenness (E) of clones is the ratio of the effective number of clones to the total number of clones (S_e/S), whereby E = (1/Σp_i²) × (1/S), with the following range: 1/S, 1. S_e and E increase as the number of isolates of each clone becomes more equal.

eBURST.

The eBURST clustering algorithm (13) for analyzing relationships between STs was applied using software available at http://eburst.mlst.net. Clonal complexes (CCs) comprised of single-locus variants (SLVs) and double-locus variants (DLVs) were identified with a user-defined setting of ≥5 of 7 shared housekeeping alleles. Identification of the founder ST for a CC was ascertained by bootstrap analysis using 1,000 replicates (13).

Estimates of recombination and mutation events.

Since founders could not be predicted with confidence for most CCs defined by eBURST (13), an alternative strategy was employed for distinguishing recombination and mutation events (38), with the additional assumption that any nucleotide polymorphism that is restricted to a single ST or CC, compared to the entire GAS database (1), is probably the result of mutation, even if there is a >1-bp change in a given allele. Likewise, nucleotide polymorphisms associated with distant STs probably arose as a result of recombination, even if there is only a 1-bp difference between alleles (38), whereby distant STs are defined as sharing ≤2 of the 7 housekeeping alleles. This set of rules provides a minimum estimate for the extent of recombination.

Additional analyses.

Sequence (nucleotide and amino acid) alignments and percent sequence identity calculations were performed using Clustal W, as implemented in MegAlign (DNAStar version 5.05). The average distance between STs was calculated by a distance matrix method (38). For tests for independence, Fisher's two-tailed exact test was used (DnaSP version 4.0).

Nucleotide sequence accession numbers.

Newly identified partial prtF1 alleles, based on the nucleotide sequence of an aligned and trimmed region of ∼333 bp, were assigned GenBank accession numbers DQ470016 to DQ470037. New emm subtypes were deposited in a publicly available database (4). Newly described alleles and STs also were deposited in a publicly available database (1).

RESULTS

Phenotypes and genotypes of MRGAS isolates.

Disc agar diffusion tests were used to verify the macrolide resistance phenotype of the GAS isolates under study. The M phenotype was present in 85 isolates, and the MLS phenotype was present in 127 isolates (Table S1 in the supplemental material). Multiplex PCR amplifications were used to establish the R-gene content of 212 MRGAS isolates (Table S1 and Fig. S1 in the supplemental material). mef(A) was present in 77 isolates, whereby each isolate had the M phenotype. erm(A) was present in 76 isolates, in which 56 and 20 isolates had the iMLS and cMLS phenotypes, respectively. erm(B) was present in 4 isolates with the iMLS phenotype and in 44 isolates with the cMLS phenotype. Additionally, one isolate with both erm(A) and erm(B) was of the cMLS phenotype.

A defined R gene could not be identified in 10 isolates, even though repeated susceptibility testing showed that 8 and 2 isolates had the M and cMLS phenotypes, respectively. In order to determine whether they had ribosomal mutations similar to those previously reported for macrolide resistance in GAS (33), a portion of the 23S rRNA gene and the complete rplD gene were sequenced for the 10 isolates. None of the known ribosomal mutations were evident, and only synonymous nucleotide substitutions within rplD were noted compared to strain MGAS10394 (data not shown).

MLST was used to define the genetic background based on the allelic forms of seven housekeeping genes that are assumed to be neutral in their genetic variation. The combination of housekeeping alleles (i.e., allelic profile) is used to define the sequence type (ST). The set of 212 MRGAS isolates was comprised of 52 distinct STs (Table S1 in the supplemental material). Fifteen new STs were discovered: seven were the result of new housekeeping alleles, and eight were the result of new combinations of previously recognized alleles. Two of the new STs, ST403 and ST404, lacked a yqiL gene.

emm typing is a sequence-based scheme that assesses genetic variation in emm, encoding the M protein surface fibril. Thirty-three emm types were present among the 212 MRGAS isolates, and 45 distinct partial emm alleles (i.e., emm subtypes) were identified; included are 8 newly recognized emm alleles. For two isolates, an emm type could not be established, although the presence of group A carbohydrate was confirmed by latex bead agglutination (24); additional GAS isolates sharing the same ST were also identified (Table S1 in the supplemental material) (8).

The 212 MRGAS isolates represent 16.2% (52 of 321) and 20.1% (33 of 164) of the known STs and emm types, respectively, based on GAS data available in public databases (1, 4) as of December 2005. These values provide a minimum bound for the extent of genetic diversity among GAS organisms that have acquired macrolide resistance. All isolates sharing the same ST also shared the same emm type; however, several emm types were associated with more than one ST.

For this study, a clone is defined by the combination of ST and R-gene content. Among the 202 isolates for which a specific R gene was identified, 67 unique combinations of ST and R gene, or clones, were recognized. An additional five STs were represented among the 10 MRGAS isolates for which an R gene remains unidentified; MRGAS isolates with an unknown R gene are classified as representing five distinct clones. Of the 72 total MRGAS clones found among the 212 MRGAS isolates, 28% of the clones had mef(A), whereas 42% and 22% harbored either erm(A) or erm(B), respectively (Table 1).

TABLE 1.

Diversity among defined sets of MRGAS isolates

R gene of MRGAS clones	No. of isolates	No. of ST-R gene combinations (clones)	Index of diversity, D (95% CI)	Effective no. of clones (Se)	Evenness of clones (E)
mef(A)	77	20	0.825 (0.764, 0.886)	5.4	0.269
erm(A) only	76	30	0.926 (0.897, 0.956)	11.6	0.388
erm(B) only	48	16	0.912 (0.885, 0.940)	9.4	0.585
erm(A) and/or erm(B)	125	47	0.960 (0.948, 0.972)	21.0	0.447
Totala	212	72	0.963 (0.952, 0.974)	24.1	0.334

^aTotal includes isolates with undefined R genes.

Assessment of sample diversity.

Simpson's index of diversity (D) can be used to assess the discriminatory power of a typing scheme (23). The combination of ST and R gene, defining 72 MRGAS clones, provides a D statistic of 0.963 (Table 1), indicative of high discriminatory power for the complete set of 212 MRGAS isolates. For the R-gene-defined subsets of MRGAS clones, D was lowest for the 77 mef(A)-positive isolates and highest for the 76 erm(A)-positive isolates. Calculation of 95% CI allows one to compare the diversity index of different subsets of isolates from the same sample (20). The difference in diversity of the mef(A)-positive versus erm(A)-positive groups of GAS isolates was statistically significant (P < 0.05; Table 1). The lower diversity and evenness for the mef(A)-positive clones may be explained by the finding that a large proportion (57%) of the isolates are represented by only two clones, ST36 and ST39, found in association with emm12 and emm4, respectively. The difference in diversity of the mef(A)-positive versus erm(B)-positive groups of GAS isolates falls just outside the range for statistical significance (Table 1). In contrast, the 95% confidence intervals for erm(A)- and erm(B)-positive isolates display extensive overlapping, signifying that their clones are similarly diverse.

Measurement of the effective number of clones (S_e) and evenness (E) of the clones may provide additional biological insights. The effective number of clones was lowest for the mef(A)-containing subset of isolates (Table 1). Evenness is equivalent to the ratio of S_e to the actual number of clones (S); E values approaching 1 indicate that the actual number of clones are nearly equal in their frequencies. The erm(B)-positive isolates display the highest level of evenness of the R-gene-defined groups of isolates (Table 1).

It is important to emphasize that the D, S_e, and E measures are dependent on the definition of “clone.” If a clone were to be redefined as the unique combination of R gene and groups of closely related clones having ≥6 of 7 shared housekeeping alleles instead of ST, the D values decrease to 0.783, 0.910, and 0.911 for the mef(A)-, erm(A)-, and erm(B)-defined subsets of isolates, respectively, and both erm(A)- and erm(B)-positive isolates become significantly more diverse than mef(A)-positive isolates (P < 0.05) (data not shown). However, if the definition of “clone” were to be more highly discriminatory and include unique combinations of three genotypes, ST, R gene, and emm subtype, the D statistic increases to 0.863 and 0.918 for the mef(A)- and erm(B)-positive isolates, respectively, and there are no longer any statistically significant differences in diversity between the R-gene-defined subsets of isolates. Nonetheless, the relative diversity of mef(A)-positive versus erm-positive isolates remains the same regardless of how “clone” is defined, whereby mef(A)-positive isolates are always less diverse.

Acquisition of macrolide resistance by horizontal gene transfer.

A central goal of this study was to determine how many times macrolide resistance arose in GAS and how widely individual MRGAS clones and their descendants have spread throughout the world. In order to establish a framework for analysis, a simple evolutionary model is put forth, whereby an MSGAS isolate acquires a single R gene by horizontal gene transfer (HGT) to generate an MRGAS clone, which in turn reproduces and may undergo genetic diversification at any locus. An additional assumption is that once an R gene is acquired by GAS, the R gene becomes fixed and is not subsequently lost by its progeny.

In considering only a single representative of each of the 52 STs comprising the complete set of 212 MRGAS isolates, the mean distance between STs was calculated. The mean distance between STs, measured by an all-ways pairwise comparison of the 52 STs, was 6.14 alleles; an ST differed in allelic profile from its most similar ST (nearest neighbor) at an average of 2.46 housekeeping alleles. Thus, many of the MRGAS isolates under study exhibit large genetic distances, consistent with the idea that macrolide resistance arose independently on numerous genetic backgrounds.

eBURST is a clustering algorithm that can depict the relationships between very closely related genotypes (13). A user-defined setting of ≥5 of 7 shared housekeeping alleles was used for clustering closely related genotypes. Each cluster signifies a clonal complex (CC) that includes single-locus variants (SLVs) and/or double-locus variants (DLVs). The three major R-gene-defined subsets of MRGAS isolates were separately analyzed by eBURST. The 77 mef(A)-containing isolates, represented by 20 STs, are comprised of four CCs and 10 unrelated clones or singletons differing from the other isolates at more than two housekeeping loci (Table 2). The 76 erm(A)-containing isolates, represented by 30 STs, are comprised of eight CCs and 13 unrelated singletons, whereas the 48 erm(B)-containing isolates, represented by 16 STs, are comprised of two CCs and 12 unrelated singletons.

TABLE 2.

Genetic relatedness among STs comprising each R-gene-defined group of MRGAS isolates

R gene	No. of isolatesa	Total no. of STsb	No. of CCsc	No. of singleton clones	Minimum no. of horizontal transfers of R gene
mef(A)	77	20	4	10	14
erm(A)	76	30	8	13	21
erm(B)	48	16	2	12	14
Total	201	66	14	35	49

^aExcludes 1 isolate with both erm genes and 10 isolates with an undefined R gene.

^bFourteen STs are associated with two or three different R genes (see Table 3).

^cCC is defined as STs sharing ≥5 of 7 housekeeping alleles; 4 of the 14 CCs are comprised of a DLV pair.

Assuming that each CC and singleton clone corresponds to an independent HGT event, the findings suggest that mef(A) was acquired by GAS at least 14 times. The data also suggest that erm(A) and erm(B) were acquired 21 and 14 times, respectively. Taken together, the genotype findings suggest that macrolide resistance was independently acquired by GAS, via an HGT event, at least 49 times. This calculation does not include the 10 isolates comprising five distant STs (Table S1 in the supplemental material), whereby an R gene remains undefined.

The relationship between emm type and each of the R-gene-defined CCs was also examined. All isolates belonging to 12 of the 14 R-gene-defined CCs (Table 2), which range in content from 2 to 30 isolates, share the same emm type (Table S1). One exception lies with the erm(A)-containing DLV of ST4 and ST395, which are associated with emm43 and emm33, respectively. The second exception is found in the erm(B)-containing DLV of ST101 and ST253, which harbor emm89 and emm78, respectively. These two DLV pairs of STs, each comprising a CC, were scored as a single acquisition of an R gene (Table 2). However, to evolve either of these CCs following a single R-gene acquisition, whereby each ST of the DLV pair differs in emm type, requires diversification at two housekeeping gene loci plus a recombinational replacement with a new emm type. It is possible that each emm-ST combination within the DLV pairs acquired macrolide resistance by an independent HGT event. Hence, the estimate of 49 independent acquisitions of macrolide resistance by S. pyogenes may be conservative.

The 202 MRGAS isolates harboring mef(A), erm(A), and/or erm(B) are represented by 50 distinct STs (Table S1 in the supplemental material). Several STs were found in association with more than one type of R gene (Table 3). Two STs, ST36 (emm12) and ST46 (emm22), were found in association with all three R-gene forms. Twelve STs were found in association with two R-gene forms, whereby four of the STs constitute SLV pairs. One SLV pair is comprised of ST38 and ST39, and MRGAS isolates of both STs harbor mef(A) or erm(A); all isolates of ST38 and ST39 have emm4. The MRGAS isolates of the second SLV pair are uniformly emm28; this pair is comprised of ST52 and ST244, whereby isolates of ST52 have either erm(A) or erm(B) and isolates of ST244 have either erm(A) only or both erm(A) and erm(B) genes. The single ST244 isolate harboring both erm genes is the only isolate in this study having two identifiable R genes. The other STs associated with more than one R gene are represented by emm types 1, 6, 11, 58, 73, 75, 76, and 89 (Table 3). Since macrolide resistance appears to have arisen more than once on the same genetic background, these 14 STs may represent particularly problematic clones.

TABLE 3.

STs that are associated with different R genes

R genes	Associated STs (emm type)
mef(A), erm(A), and erm(B)	ST36 (emm12), ST46 (emm22)
mef(A) and erm(A)	ST38 (emm4), ST39 (emm4), ST176 (emm58), ST49 (emm75)
mef(A) and erm(B)	ST28 (emm1), ST37 (emm6)
erm(A) and erm(B)	ST52 (emm28), ST244 (emm28), ST50 (emm76), ST101 (emm89), ST331 (emm73), ST403 (emm11)

Identification of ancestral clones among MSGAS.

Of the 50 STs associated with one or more of the defined R genes, MSGAS of the same ST was identified for 23 STs based on susceptibility to erythromycin using the E-test (Table S2 in the supplemental material). Isolates of an additional six STs are reported to be susceptible to erythromycin by disk diffusion (37), and data for an MSGAS isolate of another ST are posted at http://spyogenes.mlst.net (1) (Table S2 in the supplemental material). The 30 MRGAS-associated STs which are also found among MSGAS likely represent STs of the recipients of the horizontally transferred R genes.

The remaining 20 MRGAS-associated STs were found among isolates that are uniformly resistant to macrolides (n = 20). At least some of the MRGAS-associated STs that have not yet been detected among MSGAS may be direct descendants of other MRGAS clones. Indeed, 9 of the 20 MRGAS isolates lacking a known MSGAS ancestral genotype are shown to be SLVs of an MRGAS isolate (Table S2 in the supplemental material).

Genetic diversification of MRGAS at housekeeping loci.

The extent of diversification at housekeeping loci was examined for the MRGAS isolates. Since variation in housekeeping genes is presumably neutral, it was also of interest to determine the dominant mechanism for genetic change in MRGAS, either recombination or mutation. In addition, by incorporating findings on the relationships between SLV pairs of MRGAS isolates plus MLST data on MSGAS, estimates for the number of acquisitions of macrolide resistance can be further refined.

Vertical transmission of an R gene is characterized by an MRGAS clone that undergoes change at a housekeeping gene to yield an SLV. Among the R-gene-defined CCs identified among the MRGAS isolates (Table 2), there were 13 distinct SLV pairs of STs sharing the same R gene but differing at a single housekeeping locus (Table 4). erm(B) was present in only one SLV pair, whereas erm(A) and mef(A) were each present in six SLV pairs. Two SLV pairs were associated with two different R-gene forms (ST37-ST382 and ST38-ST39; discussed below), whereas nine SLV pairs had only one of the defined R genes.

TABLE 4.

Epidemiological features and genetic diversification within SLV pairs of MRGAS

emm type	R-gene content	ST-A of SLV	ST-B of SLV	Oldest known ST-A isolateb	Oldest known ST-B isolate	No. of ST-A isolatesc	No. of ST-B isolates	MSGAS of ST-A detectedd	MSGAS of ST-B detected	ST of predicted founder (bootstrap support)a	Total no. of mutation eventsf	Total no. of recombination events
11	erm(A)	ST403	ST404	1986	1999	0	0	No	No	404 (33%)	0	1
28	erm(A)	ST52	ST244	1986	1997	18	1	Yes	No	None	2g	0
77	erm(A)	ST63	ST369	1986	1999	29	1	Yes	No	None	1	0
75	mef(A)	ST49	ST388	1986	1999	5	0	Yes	No	None	1	0
6	mef(A) or erm(B)	ST37	ST382	1937	1999	7	1	Yes	Yese	37 (33%)	1	0
4	mef(A)	ST39	ST373	1986	1988	40	1	Yes	No	38 or 39 (28 or 29%)	1	0
4	mef(A) or erm(A)	ST39	ST38	1952	1986	40	4	Yes	Yes	38 or 39 (28 or 29%)	1	0
4	mef(A)	ST38	ST391	1952	1990s	4	0	Yes	No	38 or 39 (28 or 29%)	1	0
22	erm(A)	ST46	ST387	1986	1994	12	0	Yes	No	46 (98%)	1	0
22	mef(A)	ST46	ST389	1986	1999	12	0	Yes	No	46 (98%)	1	0
22	erm(A)	ST45	ST45	1947	1986	1	12	Yes	Yes	46 (98%)	0	1

^aBased on eBURST using single representative isolates of all 321 STs (December 2005).

^bBased on www.mlst.net database (December 2005) plus 212 MRGAS of this report.

^cBased on www.mlst.net database (December 2005) only, excluding 212 MRGAS of this report.

^dBased on E-test and data from Table S2 in the supplemental material.

^eMSGAS data deposited at www.mlst.net but unpublished.

^fDetailed analysis of nucleotide polymorphisms and supporting data are presented in Table S3 in the supplemental material.

^gAppears to involve two mutations, whereby the intermediate form has not been found; details are presented in Table S3 in the supplemental material.

eBURST can be used to predict the primary founder ST of a CC and to calculate the level of bootstrap support for that assignment (13). However, for the MRGAS data set, most bootstrap confidence intervals were <70%, which might be below the limit of what is often considered to be biologically significant. Therefore, in order to determine the ancestral and descendant STs for each of the nine SLV pairs harboring a single R-gene form, epidemiologic and MLST findings for all characterized GAS isolates were considered (1); as of December 2005, the online database contained information on >800 GAS isolates and 321 STs. For one SLV pair (ST403-ST404), no MSGAS isolate sharing either ST was identified among the broader GAS population (8, 36-38, and unpublished data), whereas for the ST45-ST46 pair, MSGAS isolates corresponding to both STs were found. However, for both the ST403-ST404 and ST45-ST46 SLV pairs, the mismatched housekeeping alleles were widely distributed throughout the GAS population on genetically distant STs (Table S3 in the supplemental material). Therefore, genetic diversification at housekeeping loci within these SLV pairs was probably the result of recombination following HGT of one of the mismatched alleles from an unrelated donor strain.

For the other seven SLV pairs harboring a single R-gene form, an MSGAS isolate was found for only one ST of the SLV pair. Furthermore, for each SLV pair, isolates of the MSGAS-associated ST were recovered at an earlier date, ranging from 1952 to 1986 (Table 4). The MSGAS-associated ST probably corresponds to the ST of the MRGAS ancestor. This finding is further supported by bootstrap analysis (98% confidence) for two SLV pairs showing that ST46-erm(A) is the likely ancestor of ST387-erm(A) and that ST46-mef(A) is the likely ancestor of ST389-mef(A). All seven of these diversification events appear to involve mutations (Table S3 in the supplemental material). This includes the SLV pair ST52-ST244, whereby the mismatched allele differs by two nucleotide polymorphisms that are unique to these STs.

Two SLV pairs, ST37-ST382 and ST38-ST39, were each found in association with two different R-gene forms. Comparison of the nucleotide sequence of the mismatched housekeeping alleles in the ST37-ST382 pair suggests that diversification occurred by a single point mutation in the gtr locus of ST382 (A to G at nucleotide position 199), because both gtr52 and the nucleotide polymorphism are restricted to ST382 compared to all known STs of this bacterial species (Table S3 in the supplemental material). Since the probability is low that the same point mutation arose on more than one occasion and assuming that R genes are not lost, the most parsimonious model of evolution is that MSGAS of ST37 and ST382 each acquired their two R genes independently via four HGT events. This scenario is further supported by the identification of MSGAS isolates for both STs (Table 4 and Table S2 in the supplemental material). Similar explanations apply to the evolution of the ST38-ST39 SLV pair (Table 4).

Estimates for the minimum number of horizontal acquisitions of R genes, as presented in Table 2, may be conservative, because only one HGT event was scored for every R-gene-defined SLV pair, rather than two HGT events for some special cases. If MRGAS clones of ST37, ST382, ST38, and ST39 are separately scored as independent HGT events for each R-gene form, then the total number of horizontal acquisitions increases by 4 (from 49 to 53) and the number of vertical transmissions decreases by 4 (from 13 to 9). The number of horizontal acquisitions increases further to 57 if CCs comprised solely of DLV pairs are each scored as two independent HGT events instead of one event (Table S1 in the supplemental material). The number of HGT events can increase even further if other scenarios are considered, such as attributing the acquisition of erm(A) by ST45 and ST46 to two separate events, based on the observation that each ST has an MSGAS counterpart. This adjustment eliminates one vertical transmission event as well (Table 4).

Using the most conservative estimate for HGT, with 49 independent events, horizontal acquisition of R genes among GAS exceeds their vertical transmission by >3.75-fold. The ratio of horizontal to vertical transmission events is further increased by applying the relaxed criteria for estimating HGT. Of the vertical transmission events of R genes, only two are characterized by recombination among housekeeping genes (Table 4). Although the number of genetic diversification events at housekeeping genes is small, mutation appears to be the dominant mechanism.

Genetic diversification of MRGAS at virulence loci.

Unlike housekeeping genes, emm type is subject to selection via host immune pressures. Of the 20 emm types associated with the 76 erm(A)-positive isolates, all were represented by a single emm allele (Table S1 in the supplemental material). Similarly, of the 15 emm types found among the 48 erm(B)-positive isolates, only 1 emm type had multiple alleles, emm89.0 and emm89.9; the two emm89 alleles were both associated with ST101 (Table 5). In sharp contrast, emm types present among the 77 mef(A)-positive isolates accounted for most of the observed diversification of emm. Four emm types within mef(A)-positive isolates, emm4, emm6, emm12, and emm22, together accounted for 13 allelic forms.

TABLE 5.

MRGAS clones associated with multiple emm alleles

R gene	ST	No. of emm alleles	emm alleles	No. of polymorphic sites	No. of synonymous substitutions	No. of nonsynonymous substitutions
erm(B)	101	2	89.0, 89.9	1	0	1
mef(A)	36	3	12.0, 12.33, 12.34	2	0	2
mef(A)	37	2	6.1, 6.5	4	1	3
mef(A)	39	4	4.0, 4.4, 4.7, 4.8	4	1	3
mef(A)	46	2	22.0, 22.2	Indel (9 bp)	NAa	NA
mef(A)	382	2	6.0, 6.4	4	1	3

^aNA, not applicable.

Sequence alignments of emm alleles of the same emm type, and sharing the same ST background, revealed a total of 15 nucleotide polymorphisms plus a 9-bp indel (Table 5). Only 3 of the 15 nucleotide substitutions were synonymous, whereas 12 nucleotide substitutions were predicted to result in an amino acid change. Eleven of the 12 nonsynonymous substitutions, plus the 9-bp indel, were associated with mef(A)-containing clones. The findings are consistent with the idea that MRGAS harboring mef(A) are subject to stronger host immune pressures, leading to diversifying selection of emm, compared to organisms harboring either of the erm methylase genes.

The prtF1 gene, encoding a fibronectin-binding protein and located within the highly recombinatorial FCT region (5) of the GAS chromosome, is reported by other investigators to be present more often among MRGAS than among MSGAS, and it may be associated with persistent infection or carriage (9, 39). Indeed, a prtF1 gene was detected in a high proportion, 86.3% (183/212), of the MRGAS isolates under study (Table S4 in the supplemental material). Among the 77 mef(A)-positive isolates, a prtF1 gene was detected in 73 (95%). The percentage of mef(A)-containing isolates with a prtF1 gene was significantly higher than that for either the erm(A)- or erm(B)-containing isolates (P < 0.05; Fisher's exact test). However, comparison of the number of distinct clones, as defined by ST and R gene, indicates that there is no significant difference in the distribution of the prtF1 gene among mef(A)- and erm(A)-positive clones (Table S4 in the supplemental material and data not shown).

The nucleotide sequence was determined for all MRGAS isolates that yielded a PCR product corresponding to the coding region for the N terminus of mature Protein F1. Among the 183 partial prtF1 genes, 17 allelic forms were identified (Table S1 in the supplemental material). Thus, for the 212 MRGAS isolates, the targeted ∼333-bp region near the 5′ end of prtF1 displayed less allelic diversity than the emm type-specific region, which gave rise to 45 partial emm alleles over 150 nucleotides.

Each of the 67 ST and R-gene-defined clones (Table 1) were uniform in their prtF1 allele content, with 3 exceptions. One isolate corresponding to each of the ST46-erm(B) and ST331-erm(B) genotypes had lost its prtF1 gene, whereas all other isolates of each of these two clones shared a prtF1 allele (Table S1 in the supplemental material). All isolates of ST176-erm(A) shared the prtF1.7 allele, except for MDS2573, which had a single nucleotide deletion leading to premature termination of translation (prtF1.22). Based on dates of isolation of several prtF1-containing GAS isolates sharing the same ST (Table S1 in the supplemental material and data not shown), there is no supporting evidence that the isolates lacking prtF1 or having a defective gene are ancestral.

In summary, for isolates sharing the same ST and R-gene form, genetic diversification of emm typically results in nonsynonymous nucleotide substitutions and amino acid changes, whereas genetic change in prtF1 resulted in loss of the prtF1 gene or its product in the two instances noted.

Worldwide geographic distribution of MRGAS clones.

The country of origin is known for all 212 MRGAS isolates; the organisms originate from 34 countries and all six major continents (Table S5 in the supplemental material). Their numbers range from 5 MRGAS isolates collected in Africa to 87 isolates from Europe. mef(A) and erm(A) were recovered from 29 and 27 countries, respectively, and all six continents (Table S1 in the supplemental material). Similarly, erm(B) was recovered from 27 countries and five continents. Prior knowledge of geographic origin and the macrolide resistance phenotype and/or genotype for many isolates facilitated the assembly of three R-gene-defined subsets of isolates that are also genetically diverse (Table 1).

The geographic origin was examined for the isolates comprising each of the R-gene-defined CCs and genetically distant singleton clones representing the 49 independent HGT events for acquisition of macrolide resistance by GAS (Table 2). Twenty-two of the 49 CC/singleton clones corresponded to isolates obtained from more than one continent (Table S6 in the supplemental material). Of the 22 intercontinental CC/clones, 6, 9, and 7 had mef(A), erm(A), and erm(B), respectively. The most widely distributed CC/clone, recovered from all six continents, was a singleton (ST89) bearing emm94 and erm(A). Three CC/clones were recovered from five continents: emm58-erm(A), emm4-mef(A), and emm12-mef(A). Intercontinental erm(B) clones, although numerous, appear to be somewhat less widely dispersed on a global scale.

Taken together, the 22 intercontinental MRGAS CC/clones accounted for 79% (167/212) of the MRGAS isolates. Thus, in addition to numerous independent acquisitions of R genes by GAS, the broad geographic distribution of individual clones and their descendants may contribute to the problem of antibiotic resistance in this pathogen.

DISCUSSION

An important feature of sequence-based strain typing methods is that data can be easily reproduced and shared by investigators via the internet (1, 4). An additional advantage of MLST is its high discriminatory power, which can distinguish between many isolates of GAS (8). Prior reports on MLST analysis of MRGAS have been limited to isolates obtained from a single country, including Germany (43), Poland (55), Italy (18), the United Kingdom (37), The Netherlands (19), and Norway (46). The aims of this report are to provide a global picture of MRGAS genotypes and to characterize their evolutionary relationships.

Macrolide resistance mediated by mef(A), erm(A), and erm(B) is acquired by GAS via HGT rather than by spontaneous mutation of a core chromosomal gene. mef(A), erm(A), and erm(B) can be found within GAS in association with transposons inserted into prophage (3, 18, 47) and plasmids (49). R genes that are identical or nearly identical in nucleotide sequence to the MRGAS R genes are also present in other bacterial species (3, 48, 51, 52, 57), indicating that the reservoir for these genes may be very large. The therapeutic use of erythromycin began in the mid-1950s, and shortly thereafter macrolide resistance was detected in GAS (32). The R genes of the first MRGAS clone and its bacterial donor source are unknown. The earliest recorded date for isolation of MRGAS within the set of 212 isolates is 1976, for an erm(B)-containing strain (emm12-ST36) from Canada. Only 5 of the 212 MRGAS isolates are known to be recovered prior to 1990, and therefore the earliest stages in the evolutionary history of MRGAS cannot be adequately addressed with this data set.

Despite the lack of an extended temporal scale for the 212 MRGAS isolates, several important characteristics of the emergence and genetic diversification of MRGAS can be inferred. By sampling primarily for maximal geographic diversity to include 34 countries and all six major continents, 72 unique combinations of ST and R-gene content were identified. Using a simple evolutionary model as a framework for data analysis, it is estimated that macrolide resistance emerged in GAS at least 49 times via separate HGT events. Unless GAS can acquire more than one macrolide R-gene form through a single genetic step, the estimate of 49 HGT events provides a lower boundary for the emergence of macrolide resistance in this species.

The number of independent acquisitions of macrolide resistance by GAS may be higher if R genes achieve fixation in newly emerged MRGAS clones. This is because MSGAS corresponding to both STs of an SLV pair of MRGAS (Table 4) suggests a history of two distinct HGT events rather than the conservative estimate of one event. However, if R genes are not fixed, it is possible that a new MRGAS clone diversifies at one housekeeping gene to yield an SLV that is also macrolide resistant, followed by R-gene loss from a descendant to yield a new MSGAS clone, which is an SLV of the MSGAS progenitor. R-gene loss might be facilitated in a microenvironment devoid of the antibiotic, particularly if the R gene or its mobile genetic element exacts a biological cost on the bacterial cell (16, 30, 42).

The approach used to estimate the minimum number of R-gene acquisitions may be well suited for GAS but less ideal for other bacterial species that are more clonal in their population genetic structure. For GAS which are highly recombinogenic and nonclonal (12, 21, 25), separate eBURST analyses were performed for each of the three R-gene-defined subsets of organisms. However, for more clonal species such as Staphylococcus aureus, a better approach may be to consider all known STs, irrespective of resistance phenotype or genotype, in a single phylogenetic analysis (45). When this latter approach is applied to all 321 STs that are currently defined for GAS, only 46 HGT events are predicted (data not shown). However, eBURST with 321 STs yields several CCs that contain a large number of STs, whereby highly divergent genotypes are assigned to the same CC. An example is the erm(A)-containing ST4 and ST101 strains, which have only two housekeeping alleles in common and probably arose via two independent HGT events, rather than by a single HGT event followed by diversification through multiple intermediate genotypes which have yet to be sampled (data not shown). Thus, for GAS, consideration of all known STs results in a further underestimation of the number of HGT events.

Twelve emm types were found in association with more than one R gene. This finding suggests that it is not uncommon for organisms of the same emm type and ST (or CC) combination to serve as the recipient of an R gene via lateral transfer more than once. Examination of the online database of >800 GAS isolates (1) reveals that GAS sharing STs with these particular MRGAS CC/clones are numerous and have a wide geographic distribution. Thus, it seems plausible that multiple organisms sharing an ST may have acquired the same R gene at different times and in different places, although only a single HGT event is scored.

The high number of acquisitions of macrolide resistance by GAS, even by conservative estimates, may be a sign that the mobile genetic elements are highly promiscuous and/or selection pressures are very strong. Irrespective of macrolide resistance status, GAS isolates sharing most of the 52 STs represented in the MRGAS data set have been recovered from multiple countries and/or continents (1, 8, 36, 38). Thus, in general terms, there is ample evidence that many GAS strains have been transmitted over wide geographic distances. Because the same R gene may have been acquired multiple times by the same genetic background but in a multitude of locales, it is difficult to completely disentangle the relative contribution of HGT events versus strain migration to the present-day distribution of genetically diverse MRGAS.

Using MLST data, it was previously established that the population genetic structure of GAS is highly recombinogenic over the long term (12, 21, 25). Genetic changes in SLVs can be used to empirically determine the relative contributions of recombination and mutation to genetic diversification on a short-term evolutionary scale (12, 13). Previous analysis of 495 GAS isolates representing 238 unique combinations of emm type and ST, and whose macrolide resistance status is not known but is presumed to be mostly susceptible, yielded a ratio of recombination to mutation events of ≥1.4 (38). This ratio may not be precise, because the number of SLVs was not large (n = 48). For the set of 212 MRGAS isolates, even fewer SLVs were apparent (n = 11). Nonetheless, the MRGAS data are striking, in that the number of mutational events far exceeds those attributable to recombination. Based on these rough estimates, the relative rate of recombination may be approximately fivefold lower among MRGAS than among the (largely) susceptible GAS data set.

Host tissue site of infection may a factor in the opportunities that are available for lateral gene exchange. However, the predicted tissue site preferences among MRGAS strains fail to explain the inversion in the ratio of recombination to mutation. In the sample of 495 GAS isolates, recombination was restricted to strains of the emm pattern genotype-defined skin specialist and generalist subpopulations (38), whereas for MRGAS strains, eight of the nine SLVs that diversified by mutation were also of the generalist genotype (38).

Recombination and mutation rates are controlled at multiple levels by the machinery of the bacterial cell. Because they emerged following an HGT event, MRGAS ancestral strains are not lacking in an intrinsic ability to incorporate DNA from another organism. However, the mef(A)-containing mobile genetic element of one MRGAS strain (MGAS10392) also harbors genes encoding a type II restriction-modification system (3). Although the DNA target site specificity of this restriction enzyme is not entirely clear, it is conceivable that it may render the MRGAS host cell refractory to homologous recombination following intraspecific HGT with double-stranded DNA and thereby depress the ratio of recombinational to mutational change. Whatever the underlying mechanism, the higher relative rate of mutation among MRGAS may give rise to a population genetic structure that is distinct from that of ancestral MSGAS, and furthermore it may impact evolution in response to other selective pressures, such as those imposed by host immunity.