Macrolide Resistance in Streptococcus pneumoniae

Streptococcus pneumoniae is a common commensal and an opportunistic pathogen. Suspected pneumococcal upper respiratory infections and pneumonia are often treated with macrolide antibiotics. Macrolides are bacteriostatic antibiotics and inhibit protein synthesis by binding to the 50S ribosomal subunit. The widespread use of macrolides is associated with increased macrolide resistance in S. pneumoniae, and the treatment of pneumococcal infections with macrolides may be associated with clinical failures. In S. pneumoniae, macrolide resistance is due to ribosomal dimethylation by an enzyme encoded by erm(B), efflux by a two-component efflux pump encoded by mef (E)/mel(msr(D)) and, less commonly, mutations of the ribosomal target site of macrolides. A wide array of genetic elements have emerged that facilitate macrolide resistance in S. pneumoniae; for example erm(B) is found on Tn917, while the mef (E)/mel operon is carried on the 5.4- or 5.5-kb Mega element. The macrolide resistance determinants, erm(B) and mef (E)/mel, are also found on large composite Tn916-like elements most notably Tn6002, Tn2009, and Tn2010. Introductions of 7-valent and 13-valent pneumococcal conjugate vaccines (PCV-7 and PCV-13) have decreased the incidence of macrolide-resistant invasive pneumococcal disease, but serotype replacement and emergence of macrolide resistance remain an important concern.


INTRODUCTION
Streptococcus pneumoniae, the pneumococcus, is a commensal of the human nasopharynx and an opportunistic pathogen that is a leading worldwide cause of death for children under the age of 5 years (Walker et al., 2013). In addition to localized infections such as otitis media and pneumonia, the pneumococcus may cause severe invasive disease (IPD) including bacteremia and meningitis. Development of penicillin resistance in the pneumococcus in the 1980s-1990s shifted antibiotic treatment of suspected pneumococcal upper respiratory infections and pneumonia to macrolides. Widespread macrolide use, however, is associated with increased macrolide resistance in S. pneumoniae (Bergman et al., 2006;Malhotra-Kumar et al., 2007). Clinical failures of macrolide treatment of pneumococcal infections have been reported for lower respiratory tract infections (Klugman, 2002) and bacteremia (Lonks et al., 2002;Schentag et al., 2007). Widespread macrolide use is a strong selective pressure contributing to the expansion of macrolide-resistant S. pneumoniae (Bergman et al., 2006;Keenan et al., 2015). Globally, macrolide resistance among S. pneumoniae is geographically variable but ranges from <10% to >90% of isolates Pan et al., 2015;Xiao et al., 2015).

MACROLIDE ANTIBIOTICS
Macrolides are defined by a complex macrocyclic structure with a 14-, 15-, or 16-membered lactone ring substituted with neutral or amino sugar groups. Macrolides inhibit bacterial protein synthesis by binding to the large 50S ribosomal subunit and disrupting protein elongation by causing the dissociation of the peptidyl-tRNA.
Erythromycin, discovered in 1952, is a 14-membered macrolide produced by Saccharopolyspora erythraeus (formerly Streptomyces erythraeus; McGuire et al., 1952). After the discovery of erythromycin and other naturally-produced macrolides, research focused on the creation of synthetic and semisynthetic macrolides (Kirst, 2010;Seiple et al., 2016). Azithromycin and clarithromycin are semisynthetic macrolides approved for use in the United States, and azithromycin is one of the most prescribed antibiotics in the US (Hicks et al., 2015). Additional macrolides such as roxithromycin and josamycin are approved in other countries worldwide. Macrolides bind reversibly to the 23S rRNA at a site near the peptidyl transferase center of the 50S ribosomal subunit (Kannan and Mankin, 2011). Macrolide binding occurs in pre-structured ribosomal assemblies (Pokkunuri and Champney, 2007). The smaller macrolides (14and 15-membered) partially block the nascent peptide channel to inhibit the elongating peptide chain while larger macrolides (16-membered) fully block the nascent peptide channel and cause ribosomal disassociation that reversibly inhibits protein synthesis (Weisblum, 1995b). Though distinct in chemical structure, the lincosamide and streptogramin class antibiotics have overlapping binding sites with macrolides and have similar mechanisms of action (Kirst, 2010).

Ribosomal Modification
Erythromycin ribosomal methylase (erm) family genes encode adenine-specific N-methyltransferases that methylate the 23S rRNA to prevent antibiotic binding (Weisblum, 1995a). The ribosomal methylase found in S. pneumoniae is primarily encoded by erm(B) whose gene product dimethylates the target site of the 23S rRNA (A2058 in Escherichia coli; Skinner et al., 1983;Johnston et al., 1998). The erm(B) gene is the most common macrolide resistance determinant in S. pneumoniae (Table 1). Erm(A) subclasses erm(A) (Syrogiannopoulos et al., 2001) and erm(TR) (Camilli et al., 2008) are rarely found in S. pneumoniae. Ribosomal methylation by Erm(B) confers resistance to macrolides, lincosamides, and streptogramin B, which is characterized as the MLS B phenotype (Weisblum, 1995a). In addition to the expanded spectrum of resistance, erm(B) provides high-level resistance to macrolides (erythromycin MICs usually ≥256 µg/ml).
The induction of erm(B) allows high-level translation of Erm(B) in the presence of inducers such as erythromycin (Chancey et al., 2011). In the pneumococcus, erm(B) expression may be inducible or constitutively expressed to high levels. As expression of erm genes is repressed in the absence of inducing drugs through a mechanism of translational attenuation; erm(B) expression has been proposed to have a bacterial fitness cost (Min et al., 2008;Chancey et al., 2012;Gupta et al., 2013). A recent study found that a Staphylococcus aureus strain expressing erm(C) was outcompeted by S. aureus expressing catalyticallyinactive erm(C) (Gupta et al., 2013), supporting the need for tight regulation of expression. Interestingly, deletion of the leader sequence of erm(B) in S. pneumoniae was found to confer resistance to telithromycin, the first-generation ketolide, a semi-synthetic macrolide antibiotic, by allowing constitutive expression (Wolter et al., 2008).

Macrolide Efflux
Macrolide efflux in S. pneumoniae has been the most common cause of macrolide resistance in North America, the United Kingdom, and others ( Table 1). Pneumococcal macrolide efflux is encoded by the mef (E)/mel operon and occurs through an as yet incompletely understood mechanism of macrolide binding and membrane targeting for efflux (Chancey et al., 2012). Macrolide resistance in S. pneumoniae requires both mef (E) and mel. These genes are carried on the macrolide efflux genetic assembly (Mega) element and are expressed from a single promoter inducible by 14-and 15-membered macrolides (e.g., erythromycin, clarithromycin, and azithromycin; Gay and Stephens, 2001;Ambrose et al., 2005;Chancey et al., 2015b). Expression of mef (E) and mel is tightly controlled through transcriptional attenuation (Chancey et al., 2015b).
The first gene, mef (E) shares 90% sequence identity with mef (A) from Streptococcus pyogenes (Tait-Kamradt et al., 1997;Roberts et al., 1999). While mef (E) is most common, mef (A) is more common in Germany, Denmark, and Australia (Table 1). Another homolog, mef (I), also shares 91% identity with mef (A), has been found in S. pneumoniae (Cochetti et al., 2005;Wierzbowski et al., 2005b) but is rarely found ( Table 1). Most studies do not distinguish between mef (E) and mef (A) and thus report only mef (E) or mef (A) rather than mef (A/E), which is a more accurate description of the data. In S. pneumoniae, mef (E) encodes a 405 amino acid protein that belongs to the major facilitator superfamily, which utilizes proton motive force-driven efflux to expel molecules from cells (Tait-Kamradt et al., 1997). The second gene, mel (also known as msr(D)) is a homolog of the S. aureus gene mrs(A) , which encodes an ATP-binding cassette (ABC) transporter protein but lacks typical hydrophobic, membrane-binding domains, and is predicted to interact with chromosomally encoded transmembrane complexes (Ambrose et al., 2005). Mef(E) and Mel have been shown to synergistically provide macrolide resistance and operate as a two-component efflux pump in S. pneumoniae (Ambrose et al., 2005;Zhang et al., 2016). A recent E. coli study suggests a physical interaction between Mef(E) and Mel and that Mel may bind macrolides and localize to the membrane (Nunez-Samudio and Chesneau, 2013). In S. pyogenes the presence of msr(D) alone was required for macrolide resistance  and recent evidence suggests antibiotic resistance by ATP-binding cassette proteins may occur through ribosomal protection by displacing ribosome-bound macrolide molecules (Sharkey et al., 2016). Thus, the working Frontiers in Cellular and Infection Microbiology | www.frontiersin.org model for macrolide efflux in S. pneumoniae may be macrolide displacement from ribosomes by mel, which transfers macrolide molecules to mef (E) for efflux. S. pneumoniae with mef (E)/mel have been shown to have an M phenotype, which is resistant to 14-and 15-membered macrolides but susceptible to lincosamides and streptogramin B (Tait-Kamradt et al., 1997). While mef (E)/mel-containing strains display low level resistance (MICs 1-8 µg/ml) to erythromycin, macrolide induction increases expression of mef (E)/mel and results in increased levels of macrolide resistance (Wierzbowski et al., 2005a). Induction of mef (E)/mel by macrolides increases MICs to ≥16 µg/ml (Ambrose et al., 2005;Chancey et al., 2011). The presence of the two-component efflux pump encoded by mef (E)/mel also increases resistance to the human antimicrobial peptide LL-37 (Zähner et al., 2010). LL-37 also induces expression of the efflux pump (Zähner et al., 2010). These data may suggest the efflux pump is induced during nasopharyngeal colonization and primes the mef (E)/mel-containing pneumococci to resist macrolide antibiotics.

Ribosomal Mutations
Point mutations in 23S rRNA at or near the macrolide binding residue A2058 (E. coli ribosome) have resulted in high-level macrolide resistance (Vester and Douthwaite, 2001;Franceschi et al., 2004). Mutations of ribosomal proteins L4 and L22 confer macrolide resistance in pathogenic and nonpathogenic bacteria including pneumococci. L4 and L22 are ribosomal proteins with domains on the surface of the ribosome as well as tentacles that extend into the exit tunnel in proximity to the macrolide-binding site (Schuwirth et al., 2005). In E. coli, a Lys-63-Glu change in the L4 protein (rplD) as well as a triple amino acid deletion of Met-82, Lys-83, and Glu-84 from L22 (rplV) confer resistance to macrolides (Wittmann et al., 1973;Chittum and Champney, 1994). A variety of additional L4 and L22 mutations have also been found to confer macrolide resistance (Zaman et al., 2007;Diner and Hayes, 2009). While the overall incidence is rare in S. pneumoniae, L4 and L22 mutations have been shown to result in macrolide resistance (Franceschi et al., 2004).

Dual Macrolide Resistance Genotype
S. pneumoniae containing both erm(B) and mef (E)/mel were first reported in the late-1990s (Corso et al., 1998;Nishijima et al., 1999) and are now found worldwide . The dual macrolide resistance genotype occurred in 12% of global isolates collected from 2003 to 2004, which is twice the frequency reported from 1999 to 2000 . In 2004, 18.4% of S. pneumoniae isolates from the US were found to have the dual erm(B) and mef (E)/mel genotype (Jenkins et al., 2008); in a recent study, up to 52% of macrolide-resistant isolates from Arizona were found to contain both macrolide resistance genes (Bowers et al., 2012). Tn2010 has been identified as the major composite mobile element that contains erm(B) and mef (E)/mel (Mega) (Del Grosso et al., 2006). Following introduction of the 7-valent pneumococcal conjugate vaccine (PCV-7) the "replacement" serotype 19A (ST320) with Tn2010 emerged (Del Grosso et al., 2007). ST320 is a multidrug resistant strain that appears to represent a "capsule switch" from serotype 19F and is responsible for a global pandemic in the wake of PCV-7 introduction (Moore et al., 2008;Li et al., 2011). The high-level and broader resistance conferred by erm(B) would predict that mef (E)/mel is functionally redundant in erm(B)-containing S. pneumoniae.
The Mega element lacks genes required for transposition (Gay and Stephens, 2001). Analysis of the Mega insertion sites revealed a putative target sequence of 5 ′ -TTTCCNCAA-3 ′ about six base pairs upstream of the insertion and Tn916like coupling sequences (Chancey et al., 2015a). The genes required for Mega transposition may be present on other conjugative elements of the pneumococcal genome and in non-S. pneumoniae commensal organisms (Gay and Stephens, 2001;Chancey et al., 2015a). While Mega is infrequently transferred through transposition, once stabilized in the genome Mega is widely disseminated through horizontal DNA exchange and homologous recombination.
Tn916 is the prototype conjugative transposon that contains the tetracycline resistance gene tet(M), and is found in many Gram-positive bacteria. Tn916 may incorporate additional antibiotic resistance determinants which comprise larger Tn916like composite elements (Roberts and Mullany, 2011). The history and molecular mechanisms of the Tn916 family are beyond the scope of this review, but have been explored previously (Roberts and Mullany, 2009). The most common Tn916-like elements in S. pneumoniae containing erythromycin resistance cassettes include Tn2009, Tn6002, and Tn2010 (Chancey et al., 2015a. Tn2009 is a Tn916-like element with Mega inserted into orf 6 of Tn916 to provide macrolide resistance, the M phenotype (Del Grosso et al., 2004). Tn6002 is also a Tn916-like element with macrolide resistance, with a MLS B phenotype due to the incorporation of an erm(B)containing element into orf20 of Tn916 (Brenciani et al., 2007). The erm(B) gene may also be incorporated into Tn916. Tn917, an erm(B)-containing transposon insertion into orf9 of Tn916 creates Tn3872 (Brenciani et al., 2007). S. pneumoniae with the dual macrolide resistance genotype most often contain Tn2010 or rarely the newly described element Tn2017 (Del Grosso et al., 2009). Tn2010 is a Tn916like element with Mega in orf6 and the erm(B) element in orf20 of Tn916 (Del Grosso et al., 2007). Tn2010 likely arose through the homologous recombination of Tn2009 with Tn6002 (Chancey et al., 2015a). A similar recombination event likely occurred with Tn2009 and Tn3872 to create Tn2017, which is a Tn916-like element with a Mega insertion in orf6 and Tn917 in orf9 of Tn916 (Del Grosso et al., 2009).

Interspecies Exchange of Macrolide Resistance
During the growth cycle, pneumococci develop a natural state of competence and can acquire DNA from the environment. A mechanism of DNA repair allows for integration of new DNA through homologous recombination (Straume et al., 2015). The human nasopharynx is the primary ecological niche for the pneumococcus during asymptomatic carriage (Simell et al., 2012), where S. pneumoniae has the opportunity to acquire DNA from other pneumococci and from commensal bacteria of the upper respiratory tract that may act as a reservoir for antibiotic resistance.
Other bacteria that reside in the human upper respiratory tract carry the macrolide resistance genes. Tn6002 is the most common erm(B)-containing mobile genetic element of S. pyogenes (Brenciani et al., 2007). A recent study found Mega, Tn2009, Tn6002, and Tn2010 in commensal viridans group streptococci isolated from the human oropharynx (Brenciani et al., 2014). In this study, S. mitis was the most commonly isolated streptococcal species with the macrolide resistance elements. Other Gram-positive bacteria have been shown to carry erm(B) and/or mef (E) (Luna et al., 1999;Seppälä et al., 2003;Santoro et al., 2014). The Tn2009 element has been found in commensal, Gram-negative Acinetobacter junii, and there is evidence of this Mega-containing transposon in other Gram-negative species including E. coli, Enterobacter cloacae, Klebsiella sp., Proteus sp., and Pseudomonas sp. (Ojo et al., 2006). Interspecies dissemination of mobile genetic elements containing antibiotic resistance cassettes appears common.
Asymptomatic pneumococcal carriage occurs in children and adults with rates in children ranging from <15% to >90% in developing countries (Shak et al., 2013). Carriage varies based on factors including geography and socioeconomic class (O'Brien and Nohynek, 2003;Simell et al., 2012). During nasopharyngeal carriage, S. pneumoniae forms biofilms that enhance natural transformation (Chao et al., 2014) and genetic exchange during co-colonization by two pneumococcal strains is efficient with transformation efficiencies up to 10 −2 (Marks et al., 2012). This environment may have allowed for the dissemination of macrolide resistance determinants including the assembly and selection of the dual macrolide resistance elements, e.g., Tn2017 and the more common Tn2010 (discussed above).

IMPACT OF PNEUMOCOCCAL CONJUGATE VACCINES ON MACROLIDE RESISTANCE
Between 1994 and 1999, macrolide-resistant invasive pneumococcal disease (MR-IPD) rapidly emerged in the US largely due infections caused by isolates containing mef (E)/mel (Gay et al., 2000;Stephens et al., 2005). Introduction of PCV-7 in 2000 significantly reduced the incidence of MR-IPD in the US as the highest rates of macrolide resistance were present in PCV-7 vaccine serotypes Rudolph et al., 2013;Hawkins et al., 2015). Similar vaccine specific reductions were observed worldwide, which was observed in Germany through the reduction of mef (A)-containing serotype 14 (ST9) isolates (Bley et al., 2011;Imöhl et al., 2015). The incidence of MR-IPD from 2002 through 2009 stabilized while macrolide-resistant PCV-7 serotypes continued to decline; this decline was offset by the rapid emergence of macrolide-resistant serotypes not covered by PCV-7, specifically serotype 19A, ST320 (formerly CC271; Del Grosso et al., 2007;Bowers et al., 2012;Chancey et al., 2015a).
The incidence of MR-IPD caused by serotype 19A isolates with the dual macrolide resistance phenotype (both erm(B) and mef (E)/mel) rapidly increased from 2003 through 2010 in the US and worldwide (Li et al., 2011;Quintero et al., 2011;Bowers et al., 2012;Sharma et al., 2013;Pan et al., 2015;Lyu et al., 2016). Selective pressure by PCV-7 and the continued high-level use of macrolides provided an opportunity for this 19A clone to expand worldwide. The introduction of PCV-13 in the later-2010, which contains serotype 19A, was successful in reducing carriage and IPD caused by vaccine serotypes including macrolideresistant serotype 19A isolates (Desai et al., 2015;Imöhl et al., 2015). Overall, pneumococcal conjugate vaccination has yielded sustained reductions in pneumococcal disease (Pilishvili et al., 2010). Despite challenges with serotype replacement, PCVs are an effective intervention in reducing the incidence of disease caused by macrolide-resistant pneumococcal serotypes contained in the vaccine. Continued expansion of pediatric pneumococcal vaccination into developing countries is predicted to greatly reduce the global burden of pneumococcal disease and antibiotic resistant pneumococci (Rodgers and Klugman, 2011).

CONCLUSIONS
Macrolide resistance rapidly emerged in S. pneumoniae in the early-1990s. The introduction and widespread use of semisynthetic macrolides including azithromycin and clarithromycin were important drivers of macrolide resistance in pneumococci. Macrolide resistance in S. pneumoniae is predominantly due to ribosomal methylation by the gene product encoded by erm(B) and macrolide efflux by a two-component efflux pump encoded by mef (E)/mel on the transformable genetic element Mega. Both of these macrolide resistance determinants are associated with larger composite elements (i.e., Tn6002 and Tn2009) and can be found on the same element. PCVs are effective in reducing macrolide resistance caused by vaccine serotypes and thus have been effective in the reduction of MR-IPD caused by vaccine strains. But "serotype replacement" has been an issue (e.g., 19A) and emergence of macrolide resistance in new serotypes is a concern. Continued research is needed to better understand the mechanism of macrolide efflux by Mef(E)/Mel, the emergence of genetic elements containing both erm(B) and mef (E)/mel, and to continue surveillance to monitor new changes in macrolide resistance in pneumococci.

AUTHOR CONTRIBUTIONS
MS wrote the paper and MS and DS developed and edited the paper.

FUNDING
The work was supported by Emory University (Ph. D. thesis).

ACKNOWLEDGMENTS
Thanks to Yih-Ling Tzeng and Sarah Satola for discussions of macrolide resistance and Valaria Cantos and Jeffery Collins for clinical insight.