Streptococcus agalactiae, or group B streptococci (GBS), is a leading cause of severe neonatal infections (Edmond et al., 2012), while also being a frequent colonizer of the genitourinary and gastrointestinal tracks of a significant proportion of the human population. The onset of GBS infections takes place early in infancy, usually within the first week of life, being designated early-onset disease (EOD). Late-onset disease (LOD) develops after the first week until 3 months of age. Maternal colonization with GBS is the leading risk factor for the development of neonatal disease, but other sources of transmission exist (Manning et al., 2004).

Even though national health authorities formally implemented universal GBS screening of pregnant women only in 2013 (Norma 37/2011, 2013), the neonatology section of the Portuguese Society of Paediatrics issued guidelines for screening and prevention of GBS perinatal disease in 2004. These recommended universal screening of pregnant women at 35–37 weeks gestation and the administration of intrapartum antimicrobial prophylaxis (IAP) to GBS carriers (Pinheiro et al., 2013). Following the publication of these guidelines, a study based on voluntary reporting of GBS neonatal disease in Portugal described reductions in the incidence of EOD of nearly 40%, as well as of the case-fatality rate associated (Neto, 2008), in agreement with what had been observed in other countries where similar measures were implemented (Verani et al., 2010).

Penicillin is the first line agent for prevention and treatment of GBS disease, while clindamycin and vancomycin are considered suitable alternatives for penicillin-allergic patients (Verani et al., 2010). GBS is mostly considered uniformly susceptible to β-lactams. In contrast, increasing resistance to erythromycin and clindamycin in both neonatal and adult invasive infections have been reported worldwide in recent years (Castor et al., 2008; Lamagni et al., 2013), raising concern as to the long-term efficacy of current prophylatic and therapeutic strategies. Development of a universal vaccine is the most promising approach to fight GBS disease, given the potential adverse effects of IAP and the need for effective prevention of both neonatal and adult infections. In recent years, capsular polysaccharide conjugated vaccines have reached phase II clinical trials while a pilus-based vaccine is also being investigated (Nuccitelli et al., 2015).

Serotyping is the classical phenotypic typing method for the characterization of GBS, allowing their classification into capsular polysaccharide-based serotypes Ia to IX. More recently, genotypic tools such as multilocus sequence typing (MLST) have contributed to the recognition of particular genetic lineages within serotypes that were shown to differ in virulence potential and tropism (Jones et al., 2003). The combination of several typing methods, including surface protein gene (Creti et al., 2004) and pilus-islands (Martins et al., 2013) profiling were shown to afford increased discriminatory power. Currently, a genetic lineage of serotype III defined by sequence type 17 (ST17), surface protein rib and the combination of pilus islands 1 and 2b is disseminated worldwide and is associated with an increased ability to cause neonatal invasive disease, particularly meningitis (Manning et al., 2009; Meehan et al., 2014). On the other hand, invasive disease among adults has been mostly attributed to the serotype Ia/ST23/eps and serotype V/ST1/alp3 clones (Lamagni et al., 2013; Meehan et al., 2014; Teatero et al., 2014).

The aim of our study was to perform the characterization of a collection of GBS isolates recovered from neonates with invasive disease in Portugal in 2005–2015, to document the prevalence of serotypes, genetic lineages and antimicrobial resistance patterns.

This study presents the characterization of GBS recovered from invasive infections among neonates in Portugal in 2005–2015. Even though our network comprised most of the hospital microbiology laboratories in Portugal, this study was based on voluntary reporting and is therefore not population based. Notwithstanding, our surveillance network is stable and we consider that any differences observed reflect true variations in cases, still given potentially incomplete reporting we refrained from calculating incidences of GBS disease in Portugal.

No information about the implementation and effectiveness of screening recommendations and IAP are available in Portugal from official sources. When these measures started being implemented at hospital level in Portugal, a decreasing incidence of neonatal GBS disease in 2002–2004 was found, particularly among EOD cases, suggesting that the recent implementation of IAP was already having a beneficial impact (Neto, 2008). In contrast, when comparing the number of neonatal infections with the number of live births in Portugal in 2005–2015 (Figure 1), we observed significant fluctuations in the number of disease cases in the first years of the study, with a potential increase in infections taking place after 2011 (Figure 1). While this increment did not reach statistical significance (p = 0.08), our observations support the notion that the implementation of universal screening and IAP may have had a limited impact, and is not currently causing further decreases in neonatal invasive infections in Portugal. In agreement with our findings, in England and Wales (Lamagni et al., 2013), the Netherlands (Bekker et al., 2014), and Iceland (Oladottir et al., 2011), neonatal invasive disease has been shown to increase over time in spite of the implementation of obstetric risk-based recommendations for the prevention of GBS disease.

In a previous study in which we compared the GBS populations associated with vaginal carriage in pregnant women and invasive neonatal infections in Portugal, serotypes Ia and III amounted to 69% of all invasive isolates, and were statistically associated with EOD and LOD, respectively (Martins et al., 2007). In the same study, an overrepresentation of serotype III was found in isolates recovered from the CSF, particularly among LOD cases (Martins et al., 2007), in agreement with other reports suggesting an association of this serotype with meningitis (Manning et al., 2009). In this collection, serotypes III and Ia were even more dominant, together accounting for 81% of all isolates, similarly to what has been reported recently worldwide (Ferrieri et al., 2013; Lamagni et al., 2013; Meehan et al., 2014; Teatero et al., 2014). Even though serotype III was the most frequent among meningitis cases (n = 20/26; 77%), this association was not statistically significant.

Only four isolates did not cluster in any of the five most frequent CCs, together responsible for the majority of invasive infections in humans: CC1, CC10, CC17, CC19, and CC23 (Bekker et al., 2014; Meehan et al., 2014; Teatero et al., 2014). We found no significant differences in the relative frequency of CCs over time, indicating that the genetic lineages responsible for neonatal infections in Portugal have been dominant over a significant time-span. Half of our collection was part of the CC17 hypervirulent lineage, sharing the same type III capsular polysaccharide, the surface protein gene rib and the combination of PI-1 and PI-2b, with the exception of a small subset of theses isolates carrying only PI-2b, a finding not previously observed in other CC17 isolates from Portugal (Martins et al., 2013). A recent study from Canada, using whole genome sequencing analysis, revealed the loss of PI-1 in a group of CC17 isolates and integration in the same genomic location of a mobile genetic element carrying multiple antimicrobial resistance genes (Teatero et al., 2016). Another recent publication from China confirmed and extended these findings, further discriminating the CC17/PI-2b isolates into two groups, distinguishable by their antimicrobial resistance profiles, namely lincomycin resistance associated with the presence of lnu(B) gene (Campisi et al., 2016). In our subset of CC17/PI-2b, all but two isolates were simultaneously resistant to erythromycin, clindamycin, tetracycline and streptomycin, and carried the erm(B), tet(O), aph(3′)-II, and ant(6)-Ia genes (Table 2). We found that our CC17/PI-2b also presented resistance to kanamycin and lincomycin, which would place them in the lincomycin resistant sublineage according to Campisi (Campisi et al., 2016). However, none of the our isolates harbored the lnu(B) gene, raising the possibility that our strains represent yet another sublineage of CC17/PI-2b.

Taken together with the fact that all our CC17/PI-2b isolates were recovered after 2010, our data suggests that highly resistant sublineages emerged recently and persist within the hypervirulent clone. To our knowledge this is the first report of a CC17/PI-2b lineage in Europe. However, considering that aminoglycoside susceptibility testing is not routinely performed, and given the lack of specific guidelines for streptococci, these highly resistant isolates may be expanding unnoticed in other countries.

Serotype Ia was represented by two sublineages within CC23: ST23/eps/PI-2a and its DLV ST24/bca/PI-2a. While ST23 and SLVs were dominant among EOD, ST24 and SLVs were associated with LOD cases (p < 0.001), indicating that within the same clonal complex, particular sublineages may be better adapted to cause specific disease presentations. Furthermore, the proportion of ST24 and SLVs within CC23 appears to be increasing when compared to our previous studies. While in this collection the ratio ST24/CC23 was 34% (n = 17/50), in a previous study including both invasive, non-invasive and colonization strains it was 26% (n = 18/69) (Martins et al., 2013). Taken together, these data suggest that not only is ST24 well established in the Mediterranean region, but its frequency may be increasing together with a higher propensity to cause LOD.

A small number of isolates was represented by ST1/alp3/PI-1+PI-2a. This genetic lineage is dominant among serotype V isolates and has been mostly found among invasive disease cases in non-pregnant adults (Martins et al., 2012; Meehan et al., 2014). An interesting finding in our study was a subset of serotype Ib isolates identified as ST1/alp3/PI-1+PI-2a. Considering that we have not found other serotype Ib isolates with these characteristics in any of our previous studies, as well as in the literature, it is likely that these isolates are the result of a recent capsular switching event (Martins et al., 2010).

Penicillin and ampicillin are first line drugs for prevention and treatment of GBS disease. Penicillin-allergic patients should be given clindamycin, and vancomycin is the option for clindamycin-resistant isolates (Verani et al., 2010). In our study, macrolide and lincosamide resistance rates increased significantly over time (p < 0.001) (Figure 2), similarly to what has been reported elsewhere (Castor et al., 2008; Lamagni et al., 2013). Even though overall macrolide resistance among neonates (16.1%) was higher than previously documented among non-pregnant adults (12.9%) in Portugal (Martins et al., 2012), these difference hide fluctuations in time. In fact, in the years common to both studies (2005–2008), macrolide resistance among neonates ranged between 3.9 and 9.1% (Figure 2) whereas among adults it ranged between 14.0 and 20.5% (Martins et al., 2012). These data show that in the same period, macrolide resistance among non-pregnant adults was much higher than in neonates and that resistance in the neonatal isolates has increased significantly only in recent years. Furthermore, in the present study, the cMLSB phenotype was overrepresented and associated with the erm(B) gene, in contrast to that observed previously among non-pregnant adults, in which macrolide resistance was evenly distributed among the cMLSB and iMLSB phenotypes, with the erm(A) [erm(TR) subclass] gene being the most frequent (Martins et al., 2012). In both studies resistance was mostly found within the same genetic background, represented by ST1/alp3/PI-1+PI-2a. However, this lineage expressed serotype V in non-pregnant adults, but in this study presented both serotypes V and Ib, raising the possibility that antimicrobial use has been contributing to the selection and expansion of resistant clones with new serotype/genotype combinations.

GBS causing invasive disease among neonates in Portugal are characterized by a limited number of serotypes and genetic lineages. While the dominance of serotype III and higher prominence of serotype Ia have been described worldwide, diversification within these serotypes appears to be ongoing, as illustrated by the increasing frequency of ST24 among serotype Ia and the recent emergence of highly resistant lineages within CC17. Antimicrobial resistance has a significant impact in clinical practice, not only in preventive strategies such as IAP but also in the empirical therapy for treatment of GBS infections. The increasing burden of GBS disease and dynamic nature of this pathogen substantiate the need for detailed characterization of the genetic lineages causing infections as well as for continued monitoring of antimicrobial resistance.

EM, JM, and MR contributed to the work design; EM, CP, and PGSSI performed the acquisition and analysis of the data. EM, JM, and MR participated in the interpretation of the work, and co-wrote the paper. All authors read and approved the final paper.

EM was supported by a grant from Fundação para a Ciência e a Tecnologia (SFRH/BPD/80038/2011). This work was partially funded by a grant from the governments of Iceland, Lichtenstein and Norway (EEA-PT06). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.