An increase in erythromycin resistance in methicillin-susceptible Staphylococcus aureus from blood correlates with the use of macrolide/lincosamide/streptogramin antibiotics. EARS-Net Spain (2004–2020)

Objectives To describe and analyse erythromycin resistance trends in blood isolates of Staphylococcus aureus (EARS-Net Spain, 2004–2020) and the association of these trends with the consumption of macrolide, lincosamide, and streptogramin B (MLSB) antibiotics. To assess molecular changes that could be involved in erythromycin resistance trends by whole genome analysis of representative isolates. Materials and methods We collected antibiotic susceptibility data for all first-blood S. aureus isolates in patients from 47 Spanish hospitals according to EARS-Net criteria. MLSB antibiotic consumption was obtained from the Spanish Agency for Medicines and Medical Devices (2008–2020). We sequenced 137 representative isolates for core genome multilocus sequence typing, resistome and virulome analysis. Results For the 36,612 invasive S. aureus isolates, methicillin resistance decreased from 26.4% in 2004 to 22.4% in 2020. Erythromycin resistance in methicillin-susceptible S. aureus (MSSA) increased from 13.6% in 2004 to 28.9% in 2020 (p < 0.001); however, it decreased from 68.7 to 61.8% (p < 0.0001) in methicillin-resistant S. aureus (MRSA). Total consumption of MLSB antibiotics increased from 2.72 defined daily doses per 1,000 inhabitants per day (DID) in 2014 to 3.24 DID in 2016. By WGS, the macrolide resistance genes detected were erm (59.8%), msrA (46%), and mphC (45.2%). The erm genes were more prevalent in MSSA (44/57, 77.2%) than in MRSA (38/80, 47.5%). Most of the erm genes identified in MSSA after 2013 differed from the predominant ermC gene (17/22, 77.3%), largely because ermT was significantly associated with MSSA after 2013 (11/29, 37.9%). All 13 ermT isolates in this study, except one, belonged to ST398 and came from 10 hospitals and six Spanish provinces. Conclusion The significant increase in erythromycin resistance in blood MSSA correlated with the consumption of the MLSB antibiotics in Spain. These preliminary data seem support the hypothesis that the human ST398 MSSA clade with ermT-mediated resistance to erythromycin may be involved in this trend.


Introduction
The ever-increasing prevalence of antibiotic resistance in bacteria is a serious concern that requires an international approach to management.Hence, the World Health Organization (WHO) and the European Commission both recognize the importance of understanding the emergence and determinants of resistance and the need for control strategies.In Europe, the European Antimicrobial Resistance Surveillance Network (EARS-Net) has collected antimicrobial susceptibility data for isolates from routine blood and cerebrospinal fluid cultures since 1988 (Gagliotti et al., 2011).Funded by the European Commission, EARS-Net is a European network of national surveillance systems coordinated by the European Centre for Disease Prevention and Control (ECDC), whose goal is to collect comparable, reliable data to identify variations in antimicrobial resistance over time and space, providing a basis for infection prevention and control programs (ECDC, 2022).
An EARS-Net indicator organism is Staphylococcus aureus, which is a pathogen of major clinical importance for nosocomial infections.Since the introduction of antibiotics in clinical practice, S. aureus has progressively developed resistance to the most frequently used antibiotics.It is noteworthy that the first clinical isolate of methicillinresistant S. aureus (MRSA) was reported in 1961, just 1 year after the launch of methicillin (Jevons, 1961).WHO considers MRSA the most important Gram-positive bacterial strain (Priority 2/High) for research and the development of new antibiotics (Tacconelli et al., 2018).
Macrolides, lincosamides, and streptogramin B (MLS B ) are alternative antibiotics used to treat severe staphylococcal infections, mainly in penicillin-allergic patients, along with vancomycin and the combination of linezolid and rifampicin (Ji and Xu, 2021).Macrolide resistance emerges quickly and persists, even after a short course of therapy (Malhotra-Kumar et al., 2009;Van Heirstraeten et al., 2012).Total MLS B consumption did not change significantly in Europe from 2007 to 2017(Adriaenssens et al., 2021)); however, the consumption of long-acting macrolides increased, and seasonal variation was high, suggesting that MLS B antibiotics were prescribed inappropriately in many countries (Adriaenssens et al., 2021).
Staphylococcal resistance to macrolides includes target-site modification by methylation, such as erythromycin ribosome methylase (erm) genes; antibiotic efflux pumps, such as ABC-F proteins (msr genes) and major facilitator superfamily transporters (mef genes); and drug inactivation, such as phosphotransferases (mph genes) and esterases (ere genes) (Feßler et al., 2018;Lade et al., 2022).The erm gene products methylate specific targets in the 23S rRNA, preventing the antibiotic from binding to its ribosomal target (di Bonaventura et al., 2019).This is the most widespread mechanism of resistance to macrolides, lincosamides and streptogramin B cross-resistance (MLS B phenotype).The most common erm genes in S. aureus are erm(C) and erm(A), which can be either constitutive or inducible (Fyfe et al., 2016;di Bonaventura et al., 2019).
The goals of this study were (i) to describe and analyse erythromycin resistance trends in the blood isolates of S. aureus collected by EARS-Net Spain (2004-2020), (ii) to compare erythromycin resistance with trends in the consumption of the MLS B antibiotic family (J01F family), and (iii) to characterize antibiotic resistance genes and the prevalence of resistant clones by whole genome sequencing (WGS) in a representative sample of S. aureus isolates.

Antibiotic resistance
Forty-seven Spanish EARS-Net hospitals collected antibiotic susceptibility data for all S. aureus isolated from the first blood collected from each patient from 2004 to 2020.Using EARS-Net criteria, we selected hospitals that were distributed evenly across the country (TESSy-The European Surveillance System, 2022).These hospitals serve ~13.5 million people, ~32% of the total Spanish population, and are representative of 15 of the 17 Spanish autonomous communities.Each laboratory identified the isolates and tested their susceptibilities according to standard microbiological procedures using commercial microdilution broth assays (Gagliotti et al., 2011(Gagliotti et al., , 2021)).Results were interpreted according to EUCAST criteria. 1A quality assurance exercise (UK National External Quality Assessment Scheme) was performed annually to ensure comparable results among the hospital laboratories.All the S. aureus isolates included in EARS-Net have information on susceptibility to methicillin (the only mandatory indicator), but the number of isolates tested for other antibiotics may vary (Table 1).

Antibiotic consumption
Community consumption of the MLS B antibiotic family (WHO code J01F) according to the public health prescriptions for the period 2012-2020 was provided by the Spanish National Action Plan on Antimicrobial Resistance (PRAN), coordinated by the Spanish Agency for Medicines and Medical Devices (AEMPS) from the Ministry of Health, and was obtained from the database of retail pharmacy sales from National Health System prescriptions (covering near 100% of the Spanish population).In addition, the community use of MLS B antibiotics from private health prescriptions was 1 https://www.eucast.org/clinical_breakpointsobtained from market research companies provided by the PRAN for the period 2012-2020.MLS B hospital dispensing data were also available from public hospital pharmacies and market research companies for private hospital pharmacies, both provided by the PRAN.
The consumption data were tabulated, and the number of units was converted into defined daily doses (DDD) of active drug ingredients according to WHO methodology (WHO Collaborating Centre for Drug Statistics Methodology, 2023).The number of DDD per 1,000 inhabitants per day (DID) was calculated for each active drug ingredient.

WGS of Staphylococcus aureus isolates
To study the population structure and macrolide resistance genes in S. aureus by WGS, we selected a total of 137 isolates sent to the S. aureus Reference Laboratory of the Spanish National Microbiology Centre according to the following characteristics: 46 MRSA isolated before 2013, 34 MRSA isolated after 2013, 28 MSSA isolated before 2013, and 29 MSSA isolates after 2013.These isolates were selected to have a broad geographic representation, and to represent MSSA and MRSA from periods before and after the start of the trend change in erythromycin resistance detected by EARS-Net at MSSA in 2013.

Genomic library preparation and sequence analysis
Genomic library preparation and sequence analysis were conducted as described (Pérez-Vázquez et al., 2019).Raw sequence data were submitted to the European Nucleotide Archive (PRJEB61102).The quality of the short reads was assessed using FASTQC, and they were assembled into contigs with Unicycler 0.4.8 (Wick et al., 2017).The quality of the assembly was assessed with QUAST. 2 Prokka v1.14-beta (Seemann, 2014) was used for automatic de novo assembly annotation.

Phylogenetic analyses
Sequence types (STs) were calculated according to the multilocus sequence typing (MLST) scheme of the Public databases for molecular typing and microbial genome diversity (PubMLST) 3 using Ariba v2.6.2 (Hunt et al., 2017).A simple diversity index (SDI; Gastmeier et al., 2006) was applied to analyze population diversity.Core genome MLST (cgMLST), consisting of 1861 targets for S. aureus provided by SeqSphere+3.5.0 (Ridom, Münster, Germany), was performed with the 137 sequenced isolates.Additionally, ST398 isolates from this study were analyzed using cgMLST together with a collection of 239 S. aureus of this ST downloaded from the NCBI database using "chromosome" and "complete" as filtering criteria for assembly level and the absence of mec genes as the genotypic criterion.

Analysis of antimicrobial resistance and virulence genes
Antibiotic resistance genes were analyzed by Ariba v2-6.2 using the CARD database 4 and ResFinder (CGE server 5 ).Virulence genes were analyzed with the previous methodology using the database Virulencefinder_db. 6

Statistical analysis
The significance of the trends in macrolide resistance was calculated by the χ 2 test for trend.Trends in J01F antibiotic family consumption were examined by simple linear regression analysis.The strength of the association between MLS B antibiotic use and erythromycin resistance was determined by linear regression analysis (Cuevas et al., 2011).The resistance proportion was transformed to the natural logarithm of the odds of resistance.The log of the odds of resistance (as the dependent variable) was expressed as a simple linear function of the independent variable (antibiotic use) (Cuevas et al., 2011).Macrolide resistance for 2013-2020 was correlated with 2 http://quast.bioinf.spbau.ru/,accessed on 3 March 2023.antibiotic use in the prior year.Fisher's exact test was used to compare the prevalence of resistance or virulence genes by groups.p values < 0.05 were considered statistically significant.Statistical analyses were performed using GraphPad Prism software v.7.02 (GraphPad Software Inc., San Diego, CA, United States).
For the MSSA isolates collected after 2013, the group showing the most EARS-Net-based increase in erythromycin resistance, erm gene prevalence was 75.9% of (22/29), similar to MSSA strains isolated before 2013 (23/28, 82.1%) (Table 3).However, after 2013, the erm genes in MSSA were mostly different from the predominant erm(C) and included 11 strains with erm(T) and six with erm(A) (17/22, 77.3%) (Table 3; Supplementary Table S1).In fact, the erm(T) gene was significantly associated with MSSA isolated after 2013 (11/29, 37.9%), whereas it was identified in only two of the other isolates (2/108, 1.8%) (p < 0.0001).All 13 isolates with erm(T), except one, belonged to ST398 and were collected from 10 hospitals in six different Spanish provinces.The sequence of the genetic environment of erm(T) genes identified the rep13 gene, which is involved in plasmid replication.
Nine isolates (6.6%) carried resistance genes for lincosamides (Supplementary Table S1), including lincosamide nucleotidyltransferases (lnu) genes in six isolates [five lnu(A) and one lnu(B)], genes encoding the ABC-F proteins vga and lsa in one isolate each, and the cfr methylase gene in two isolates.One isolate had both lnu(B) and lsa (Table 3).
All MRSA had the mecA gene, and 80% (64/80) of MRSA and 29.8% (17/57) of MSSA had resistance genes to at least one of the three main aminoglycosides (gentamicin, tobramycin, or amikacin)  S1).Fifteen isolates more than two mutations in topoisomerase II/IV genes.
The comparative analysis of ST398/erm(T) isolates included in this study with a collection S. aureus ST398 isolates from NCBI database showed that the most frequent mechanism of macrolide resistance was erm gene production, being erm(C) most frequent one followed by erm(A), erm(B) and erm(T).All MSSA isolates with erm(T) are grouped, MSSA erm(T) isolates from this study are grouped in a cluster together with animal-independent ST398 MSSA isolates reported in New York (Uhlemann et al., 2012) (Figure 4).
Other important virulence genes in S. aureus, such as Panton-Valentine leucocidin (PVL) genes lukS and lukF, and exfoliative toxin A gene eta were detected in eight (5.8%; 75% of them belonging to  ) and six (4.4,66.7% belonging to ST15) isolates, respectively (Supplementary Table S1).We found no pattern of virulence genes associated with the different groups, except for the presence of the lukF/luKS genes in the MRSA isolates.Of the genes mentioned above, the ST398/erm(T) isolates only presented homogeneously the hla (100%), chp (91.7%) and scn (91.7%) genes.

Discussion
In addition to methicillin resistance, the cross-resistance to MLS B antibiotics among S. aureus strains, as well as the rapid transmission of resistance genes, is a major concern for the future efficacy of antibiotic therapy.Specifically, the extensive use of MLS B antibiotics against Gram-positive bacteria is of concern because macrolideresistant MRSA strains are believed to be a major cause of clinical infections (Miklasińska-Majdanik, 2021), and they are associated with increased mortality rates (Bishr et al., 2021).In our study, erythromycin resistance in MRSA was approximately three times more common than in MSSA.However, our major concern that led to this research was the significant increase in erythromycin-resistant MSSA blood isolates that were not observed in MRSA.
The main strengths of this study were the analysis of bacteraemia caused by S. aureus, representing a broad national caseload over 17 years, in conjunction with antibiotic consumption data.Our molecular analyses represent a pilot study of a representative sample of S. aureus isolates to develop hypotheses about the EARS-Net findings, which will then require further studies.EARS-Net is coordinated by ECDC to collect, analyse, and report data on antimicrobial resistance through a network of national surveillance systems across Europe and to take actions to address antimicrobial resistance.During its more than 20 years of operation, EARS-Net has been effective in detecting changes in trends in antibiotic resistance in S. aureus, both at a national and a European level (Gagliotti et al., 2021).
The increase in erythromycin resistance in MSSA correlated temporally with an increase in the consumption of MLS B in Spain; however, in 2016, outlier erythromycin resistance data reduced the statistical significance of the association.The cause of this single year of atypical data is unknown.The overall rise in MLS B antibiotics consumption in this study was probably mostly due to the increased prescription of these antibiotics for community-acquired respiratory infections, and specifically the prescriptions of 3-day azithromycin courses.The association between macrolide consumption and increased resistance to MLS B antibiotics has been described previously, especially in the context of specific pathologies such as cystic fibrosis (Tramper-Stranders et al., 2007) or trachoma (Bojang et al., 2017).Although the use of antibiotics (even appropriate usage) entails inevitable selection of resistance, this can be mitigated by the implementation of stewardship programs, including the diverse and combined use of antibiotics.
Although erm(C) is the most common macrolide resistance gene in S. aureus worldwide, it varies by geographical region and by the phenotype of susceptibility or resistance to methicillin (Miklasińska-Majdanik, 2021).Although we confirmed the general dominance of erm(C), we found a high prevalence of msr(A), which was most common in MRSA.A study in Spain (2006Spain ( -2007) ) showed 23.1% prevalence for erm genes and 15.8% for msr genes in invasive S. aureus (Pérez-Vázquez et al., 2009).
The present study showed a strong association between msr(A) and mph(C) genes; only 2/63 isolates with msr(A) lacked mph(C).Matsuoka et al. (2003) found the mph(C) phosphotransferase gene on plasmid pMS97, 342 bp downstream of the msr(A) gene.These authors suggested, based on their findings, that a region of the msr(A) gene is required for the full expression of mph(C).However, that plasmid also carried erm(Y) (15), a gene that we did not find in our study.
One of the main findings of this study was the prevalence of erm (T)-bearing ST398 isolates in MSSA collected during the period in which EARS-Net found a significant increase in erythromycin resistance in MSSA isolates that produced bacteraemia.Although this pilot study included a limited number of representative isolates, our finding suggests that the increased resistance to macrolides in these MSSA identified by EARS-Net could be due, at least in part, to the spread of ST398 in this group of isolates.ST398 is relevant to public health because methicillin-resistant strains are associated with livestock capable of infecting humans (Silva et al., 2023).However, previous studies suggested two subpopulations in clonal complex 398 S. aureus-a human-adapted clade mostly with the erm(T) gene, and an animalassociated clade with mecA, tet(M), and erm(C) genes (Price et al., 2012;Argudín et al., 2018).A recent Belgian study identified different ST398 subpopulations, including typical human and animal clades, as well as new emerging mixed subpopulations that underlie the ability of this lineage to acquire resistance and virulence genes (Argudín et al., 2018).The proximity of erm(T) and rep13 genes in MSSA ST398 isolates suggests that erm(T) is on a plasmid, as has been described recently (Salgueiro et al., 2023).The MSSA ST398 human clade was reported mainly in China and France (Bouiller et al., 2020), and was frequently implicated in severe infections, whereas the MRSA ST398 animal clade was reported mainly in the skin and soft tissue (Bouiller et al., 2020).
Most well-known staphylococcal virulence genes, such as enterotoxins, toxic shock syndrome toxin, or PVL, were absent in ST398 isolates from this study, as previously communicated (Bouiller et al., 2020).However, previous studies have suggested that MSSA ST398, which is prevalent in bacteraemia, may be a more virulent subtype (Bouiller et al., 2020) with higher mortality (Bouiller et al., 2016).Among the virulence factors previously associated with ST398 MSSA, only scn and chp genes were identified in erm(T) ST398 isolates of our study, but the immune evasion cluster (IEC) sak and sea genes were not found.
A limitation of this research could be the reduced number of sequenced isolates by group, which, although they were strictly chosen to be representative of the issue to study, not all belonged to the same collection in which the increase in resistance was primarily detected.

Conclusion
The highly significant increase in resistance to macrolides in MSSA causing bacteraemia, in contrast to the absence of this trend in MRSA, correlated with the consumption of MLS B antibiotics in Spain.Our data seem to support the hypothesis that the human ST398 MSSA clade with erm(T)-mediated resistance to erythromycin may be involved in this trend.An epidemiological surveillance system for MRSA and MSSA is important to monitor the emergence of dangerous new S. aureus subpopulations.Further WGS research is needed to identify the emergence of these subpopulations in the clinical setting, as well as their correlation with changes in the patterns of antibiotic susceptibility.
FIGURE 1Antibiotic resistance trends in MRSA (A) and MSSA (B) blood isolates.(A) Shows significant decrease in resistance to ciprofloxacin, gentamicin, erythromycin and clindamycin; while (B) displays significant increase in resistance to erythromycin, clindamycin and ciprofloxacin.

FIGURE 3
FIGURE 3 Population structure of Staphylococcus aureus isolates from this study: minimum-spanning tree.Distances shown are based on cgMLST of 1861 genes using the parameter "pairwise ignoring missing values."Fill colors in each circle indicate MSSA and MRSA and the year of isolation, color of the dashed line in circles indicates macrolide resistance mechanism type.

FIGURE 4
FIGURE 4 Population structure of ST398 Staphylococcus aureus in comparison with other publicly available whole genomes of the same ST.Minimum spanning tree, distance based in a cgMLST scheme of 1861 genes.Fill colors in each circle indicate MSSA and MRSA from this study or from other origins.Blue dashed line around circles remark isolates with erm(T) genes and green dashed line represent isolate NC_017631.1 (Uhlemann et al., 2012).Gray shadows represent groups of strains; a threshold of 5 alleles was applied.

TABLE 2
Community and nosocomial consumption of macrolide, lincosamine, and streptogramin antibiotics group (WHO code J01F).
Expressed in DDD per 1,000 inhabitants per day (DIDs).A B FIGURE 2Occurrence of erythromycin in S. aureus causing blood infections (years 2013-2020), plotted against total use of family J01F antibiotics (2012-2019) in Spain with 95% confidence intervals (A).In (B), the outlier erythromycin resistance data of 2016 (consumption of 2015) is excluded.Consumption is expressed in DIDs, DDDs/1000 inhabitants/day.Log odds is the natural logarithm of the OR.

TABLE 3
Distribution of macrolide and lincosamide resistance genes and main sequence types in S. aureus according to the groups studied.