Streptococcus agalactiae (Group B Streptococcus, GBS) is a leading cause of meningitis, sepsis and pneumonia in neonates in the United States. In addition to illness in the first week of life, GBS also causes invasive disease in older infants, pregnant women, children and young adults with underlying medical conditions, and older adults.

Penicillin G and ampicillin are the drugs of choice for the prevention or treatment of S. agalactiae infections, while erythromycin or clindamycin are the recommended alternatives for patients who are allergic to β-lactams.¹ Although reduced susceptibility to penicillin has been reported,² clinical GBS isolates remain susceptible to β-lactams, although the proportion of isolates resistant to erythromycin and clindamycin has increased in recent years.^3–5 Resistance to these antibiotics is usually due to the modification of ribosomal targets, most commonly mediated by an erm methylase, which confers cross-resistance to macrolides (e.g. erythromycin, azithromycin), lincosamides (e.g. clindamycin, lincomycin) and streptogramin B antibiotics (e.g. virginiamycin S1, pristinamycin IA, quinupristin), known as the MLS_B phenotype.⁶

In contrast, resistance specific to lincosamides (L phenotype) is rare in GBS, but has been previously reported in clinical isolates from Canada,⁷ the United States,⁸ Spain,⁹ Argentina,¹⁰ Korea¹¹ and South Africa,¹² due to antibiotic modification mediated by the lnu(B) gene. Members of the lnu (previously lin) gene family encode nucleotidyltransferase enzymes that catalyse the adenylation of lincomycin and clindamycin; lnu(B) was first identified in Enterococcus faecium,¹³ and later in GBS,⁷ Streptococcus uberis,¹⁴ Staphylococcus aureus¹⁵ and Streptococcus lutetiensis.¹⁶ Resistance to lincosamides has also been reported in combination with resistance to streptogramin A compounds (e.g. virginiamycin M1, pristinamycin IIA, dalfopristin) and pleuromutilins (e.g. tiamulin, retapamulin, lefamulin) in GBS isolates from Iceland and New Zealand (LSA and LSAP phenotypes), mediated by the lsa(C) gene, probably by active efflux, but the exact mechanism remains elusive.^17,18 The lsa(E) gene mediates a similar resistance phenotype in S. aureus, in combination with lnu(B).¹⁹

Other genes associated with the LSA phenotype include vga(A) and cfr. The vga genes have been characterized as determinants of streptogramin A resistance; vga(A) also confers low-level resistance to lincomycin in S. aureus and the variant vga(A)_LC has been identified in Staphylococcus haemolyticus strains resistant to lincomycin and clindamycin.²⁰ The cfr gene encodes an rRNA methyltransferase that confers resistance to phenicols, lincosamides, oxazolidinones, pleuromutilins and streptogramin A antibiotics.²¹

Active Bacterial Core surveillance (ABCs) is a core component of the CDC’s Emerging Infections Programs network (EIP). ABCs is an active surveillance system for invasive bacterial pathogens of public health importance, including GBS. During routine ABCs activities, we identified several GBS isolates that expressed resistance to clindamycin in the absence of resistance to erythromycin. In order to determine the nature and frequency of these phenotypes in our GBS collection, we screened the ABCs GBS isolate collection retrospectively for all isolates with erythromycin-susceptible, clindamycin-resistant phenotypes, and performed WGS to identify the genetic determinants that confer these unusual phenotypes.

GBS surveillance areas currently represent over 33 million persons and 400000 live births within 10 states (http://www.cdc.gov/abcs/methodology/index.html). After collection and submission, serotyping of isolates by latex agglutination and antimicrobial testing by broth microdilution was performed at the CDC Streptococcus laboratory. At the time of our study, serotype and antimicrobial resistance data were available for 21186 GBS isolates collected from1998 to 2015.

We screened MIC results for clindamycin and erythromycin, in order to identify isolates susceptible to erythromycin (MIC ≤0.25mg/L) and non-susceptible to clindamycin (MIC ≥0.5mg/L). Sixty-five (0.31%) isolates with this resistance phenotype were found. Genomic DNA was extracted from these 65 isolates and WGS was performed using an Illumina MiSeq. The presence of 10 genes previously associated with LSA resistance was investigated by mapping to the full structural gene sequences using the SRST2 bioinformatics tool:²² erm(A) (X03216), erm(B) (M11180), cfr(NG_047631), lnu(A) (J03947), lnu(B) (AJ238249), lnu(C) (NG_047924), lnu(D) (NG_047925), vga(A)_LC (DQ823382), lsa(C) (NG_047934) and lsa(E) (JQ861959). The maximum divergence and minimum coverage thresholds were both set at 80% initially, then lowered to 50% for isolates against which none of the genes mapped at 80%. SRST2 was also used to determine MLST profiles. SplitsTree4²³ was used to build a phylogenetic network with the concatenated sequences of the seven MLST loci.

In order to determine if these isolates were also cross-resistant to streptogramin A or pleuromutilins, we determined the MICs of virginiamycin M1 (Cayman Chemical, Ann Arbor, MI) and tiamulin (Wako Chemicals, Richmond, VA) by broth microdilution, following CLSI guidelines.²⁴ Since interpretive standards had not been established for these compounds against GBS, we used previously reported values along with MIC distributions of control isolates to determine cut-off values.^25–27 Briefly, we determined broth dilution MIC values of virginiamycin M1 and tiamulin in 45 GBS control isolates displaying three different phenotypes (15 of each): EryS/CliS, EryR/CliS and EryR/CliR; and compared these values with those obtained from the 65 EryS/CliR test isolates.

The EryS/CliR isolates are described in Table 1. The most common serotype among these isolates was serotype III (n = 24, 37%), followed by serotype V (n = 18, 28%), and serotype Ia (n = 16, 25%). Of the 60 isolates for which age group data were available, 11 (18.3%) were from infants, 32 (53.3%) from adults 17–59 years old, and 17 (28.4%) from adults 60+ years old; the most common diagnosis among the 60+ group was cellulitis (41%), while the largest proportion of adults 17–59 years old (14, 44%) and all infants presented with bacteraemia/sepsis. The majority of infants (7, 64%) were under 7 days old (early-onset disease) and of those, almost all (6/7) carried isolates belonging to serotypes Ia, III or V, which have been previously associated with early-onset disease.³

Thirteen sequence types (STs) were identified among the isolates. The most common ST was ST19 (24, 37%), followed by ST23 (13, 20%). The majority of the isolates (81.5%) belong to one of the three previously reported clonal complexes, i.e. CC19, CC23 and CC12,²⁸ but there was no apparent correlation between ST and serotype or resistance determinant. This is consistent with extensive horizontal transfer and limited clonal expansion (Figure 1).

Of the 65 isolates, 49 (75%) were positive for lsa(C), 12 (18%) were positive for both lnu(B) and lsa(E), and 4 (6%) were negative for all accessory MLS determinants investigated. The sequences from these four isolates were subjected to further analysis, including variant calling, genomic island prediction (IslandViewer3²⁹) and gene detection using several resistance databases (Resfinder,³⁰ RGI/CARD³¹ and ARG-annot³²), resulting in no determinants being identified.

Among these isolates, lnu(B) and lsa(E) were carried on a 75.5 kb mobile element, containing genes from two other elements: one described in a GBS isolate (SGB76)³³ and one described in a Streptococcus suis isolate (SC070731);³⁴ this element was invariably integrated into the chromosome at rum(A) [23S rRNA(uracil-5-)methyltransferase], in the same position as previously described for van(G)-containing elements in GBS and Streptococcus anginosus³⁵ (Figure 2a).

In isolates where tet(M) was present along with lsa(C), both genes were located in a Tn916-like genetic element 23.5 kb long inserted into the chromosome (Figure 2b), while in isolates where lsa(C) was present by itself, it was inserted in the same position as previously described for a GBS isolate (UCN70).¹⁷

The phenotypes and MIC ranges associated with each genotype are shown in Table 2. Based on the bimodal distribution of MIC values, the cut-offs were determined as follows: isolates with MIC values against virginiamycin M1 ≥4mg/L were considered non-susceptible to streptogramins A (Figure 3), while isolates with MIC values against tiamulin ≥1mg/L were considered non-susceptible to pleuromutilins (Figure 4). These values are consistent with reported values for related compounds, dalfopristin (streptogramin A) and retapamulin (pleuromutilin), against GBS.^17,36

In order to determine whether the frequency of these phenotypes among the GBS isolates in our collection has increased in recent years, the proportion of isolates with any of these three phenotypes (L/LSA/LSAP) for each year was plotted for the 2006–15 period (Figure 5), and a logistic regression model was fitted. The resulting log likelihood ratio (χ²) confirmed that this upward trend was significant (P value <0.001).

Resistance to clindamycin in the absence of erythromycin resistance has previously been reported in GBS, both on its own or in combination with resistance to streptogramins A and pleuromutilins. While they still remain relatively rare, the frequency of these resistance phenotypes appears to have increased in the last decade, both in the ABCs collection and in other reports in the literature.^{7–12,17,18,37}

The L phenotype mediated by the nucleotidyltransferase expressed by lnu(B) has been reported in several species in a number of countries.^7–12,37 Among our isolates, lnu(B) is invariably present in combination with lsa(E) on the same transposable element, as previously described in S. aureus¹⁵ and GBS,³³ and in association with cross-resistance to clindamycin, virginiamycin M1 and tiamulin (LSAP). The MIC values against virginiamycin M1 and tiamulin for these isolates positive for both lnu(B) and lsa(E) were up to 4-fold and 16-fold higher, respectively. It appears likely that these two genes act synergistically to produce this phenotype.

The first report of the LSAP phenotype described a GBS isolate from New Zealand carrying the lsa(C) gene. This isolate was originally identified as expressing the LSA phenotype³⁶ and later found to be resistant to tiamulin as well.¹⁷ Since then, only 10 additional lsa(C)-positive GBS strains have been reported.^18,37 To the best of our knowledge, this study is the first time the LSAP phenotype has been reported among GBS isolates recovered in the USA. As the use of WGS becomes more prevalent in the study of antimicrobial resistance, the number of isolates reported to carry these and other potentially novel determinants will likely increase.

Of the four isolates where no determinant was identified, the three that expressed the LSA phenotype had clindamycin MIC values up to 32-fold higher than those of the lsa(C)/lnu(B)+lsa(E)-positive isolates and the one isolate expressing the L phenotype, which suggests that different mechanisms may be involved. Further investigation is being conducted by our group into the genetic bases for both of these phenotypes, to be presented in a future publication.

Although streptogramins are not as widely used as clindamycin, the presence of genes that confer resistance to streptogramin A in conjunction with widespread MLS determinants that confer resistance to streptogramin B (such as erm genes), could lead to the ineffectiveness of the quinupristin (streptogramin A) plus dalfopristin (streptogramin B) drug combination, an important agent for the treatment of complicated skin infections caused by S. aureus or Streptococcus pyogenes.

Pleuromutilins, on the other hand, are almost exclusively used for veterinary applications (retapamulin is the only pleuromutilin approved for topical use in humans), though new drugs in this class are under investigation for human use. In fact, the semi-synthetic pleuromutilin lefamulin is one of the only 11 antimicrobial drugs currently in Phase 3 of the antimicrobial development pipeline.³⁸ The emergence of resistance to antibiotics that are yet to be approved for human use is concerning and stresses the importance of limiting the use of antibiotics in agriculture, as well as the critical need to develop new agents.