Seventy-eight isolates from a collection of 205 AmpR E. faecium (MIC ≥ 16 mg/L) recovered in Portugal from the last decades were analyzed (Novais et al., 2005a,b,c, 2006, 2013). The 78 E. faecium were selected in order to include isolates with different antibiotic resistant phenotypes, from diverse origins (39 from different patients admitted in five hospitals from different cities, 18 from swine feces and piggeries, 4 from retail poultry carcasses, 2 from healthy human feces and 15 from hospital wastewater), isolation date and, whenever possible, clonal relatedness. Susceptibility to ampicillin and 10 other antibiotics of different classes was evaluated by disk diffusion and/or agar dilution method (Clinical and Laboratory Standards Institute, 2012). β-lactamase production was tested in AmpR E. faecium isolates by the nitrocefin test (5 μl were directly placed in bacteria growing around the ampicillin disk) and PCR amplification of blaZ using primers based on the GenBank sequence no. M25257.1 (blaZF-3′-TTGCCTATGCTTCGACTTCA-5′, blaZR- 3′-AGTGAAACCGCCAAGAGTGT-5′). Clonal relatedness was established by Pulsed-Field Gel Electrophoresis (PFGE) and analysis of Multilocus Sequence Typing (MLST) data (Tenover et al., 1995; Homan et al., 2002; Freitas et al., 2013)¹ using the Bayesian Analysis of Population Structure (BAPS) algorithm (Willems et al., 2012; Tedim et al., 2015).

Filter-mating assays were performed in plain Brain Heart Infusion (BHI) not supplemented with antibiotics at 37°C overnight, using a donor/recipient ratio of 1:1 (1-3 experiments per isolate) and E. faecium GE1 as the recipient strain. Further filter mating assays under the same experimental conditions but using E. faecium BM4105RF and E. faecium 64/3 strains as recipients, were performed for those field isolates able to transfer pbp5 to E. faecium GE1. The three laboratory recipient strains used differed in the susceptibility against ampicillin and the presence of pbp5: GE1 [Δpbp5; MIC_Amp < 0.016mg/L; tetracycline (Tet), rifampicin (Rif) and fusidic acid (Fus) resistant (^R); ST515/BAPS 2.3b], BM4105RF (pbp5; Rif^R, Fus^R; MIC_Amp = 0.5mg/L; ST172/BAPS 1.3) and 64/3 (pbp5; Rif^R, Fus^R; MIC_Amp = 1mg/L; ST21/BAPS 2.3a). Recipients acquiring pbp5 are from this point named as transconjugants, based on the most likely transfer mechanism related to pbp5 (conjugation) published during the revision of this article (García-Solache et al., 2016). Transconjugants were selected on BHI agar supplemented with antibiotics (ampicillin-10 mg/L, fusidic acid-25 mg/L; rifampicin-30 mg/L) and incubated for 24 up to 96 h (37°C) to recover potential transconjugants growing at different times. Transconjugants were confirmed by comparison of their antibiotic resistance phenotype and PFGE profile with those of the donors and the recipients. Stability of acquired pbp5 chromosomal regions (from now on designated pbp5 genetic platforms) after serial daily passages (x30) on antibiotic free BHI agar was evaluated in both donor strains and transconjugants. Colonies from each passage were inoculated on plates containing the same agar medium and tested for ampicillin susceptibility by disk diffusion (Clinical and Laboratory Standards Institute, 2012).

Total DNA from a donor strain (AmpR E. faecium clinical isolate HPH2), a recipient strain (AmpS E. faecium GE1) and the resulting GE1 transconjugant (AmpR E. faecium TCGEHPH2.1) was extracted with PureElute^TM Bacterial Genomic Kit (EdgeBio, Gaithersburg, MD, USA). DNA concentration was measured with Nanodrop 2000 (Thermo Scientific, Waltham, MA, USA) and Qubit 2.0 fluorometer (Life Technologies). Genomic DNA (250–350 ng) was sequenced on the Illumina MiSeq system using the MiSeq reagent kit v3 and a read length configuration of 2 × 300 bp (Illumina, San Diego, CA, USA). Sequencing was carried out using a standard 2 × 71 base protocol (300–400 bp insert size) in a Genome Analyzer IIx (Illumina, San Diego, CA, USA) at the sequencing facility of the University of Newcastle (United Kingdom). The main statistics for the three sequence datasets (number of reads and coverage) analyzed are shown in Supplementary Table S1. Assembly of sequence data was done using Newbler software (454 life sciences, Roche, Branford, CT, USA). Supplementary Table S2 shows the final assembly results.

Gene prediction was performed using GeneMark.hmm 3.05 (Besemer et al., 2001). Similarity searches for potential protein-coding regions were carried out against a UniRef100 database² using Best Blast Hit approach (Supplementary Table S3). The sequence of the transferred region containing the pbp5 gene was predicted by a gene-by-gene comparison strategy using blastn (genes present in transconjugant strain TCGEHPH2.1 and donor strain HPH2 but not in recipient strain GE1) (Supplementary Table S4). Sequence of the putative “transferred region” was mapped against the genome of the E. faecium strain DO (RefSeq Accession: NC_017960) using Easyfig 1.2 software (Sullivan et al., 2011). In addition to E. faecium DO, the “transferred region” was also aligned against fully sequenced and closed genomes of E. faecium Aus0004, E. faecium NRRL B-2354 and E. faecium Aus0085 strains (RefSeq accessions numbers: NC_017022, NC_020207 and NC_021994, respectively) with Mauve (Darling et al., 2004). Functional annotation of the query sequences was performed using Blast2GO and Uniprot and KEGG databases³

An 8–10 kb DNA fragment comprising the pbp5 gene and its boundaries was characterized by PCR assays in 15 transconjugants and 21 field isolates (Table 1). PCR conditions were adapted according to the expected amplicon size (<3 kb/>3 kb): 0.5/1 mM of each primer, 2/2.5 mM MgCl₂, 1x of reaction buffer, 0.2/0.4 mM of each deoxynucleoside triphosphate, 1,25U GoTaq^® Flexi DNA Polymerase [Promega Corporation, Madison, WI, USA]/2.5U Takara LA Taq polymerase (Takara^TM Bio Inc., Shiga, Japan). The amplification program was 25 cycles of 30 s at 94°C, 30 s at 55°C, 30 s at 72°C; 1 cycle 10 min at 72°C (for fragments <3 kb) or 35 cycles of 30 s at 96°C, 1 min at 55°C, 7 min at 72°C; 1 cycle 10 min at 72°C (for fragments >3 kb). Amplicons were further discriminated by comparing the corresponding RFLP patterns obtained after digestion with DdeI or ApaI and sequences. Genomic location of pbp5 was identified by hybridization of I-CeuI and SmaI-digested genomic DNA using 23S rDNA and/or pbp5 probes (Liu et al., 1993; Freitas et al., 2013). Labeling and detection were carried out with Gene Images Alkphos direct labeling system kit following the manufacturer’s instructions (Amersham GB/GE Healthcare Life Sciences UK limited). The presence of CTn916, CTn5386 and CTn5382, previously associated with transferable AmpR, was determined by analyzing the presence of specific sequences (integrases, excisionases, non-integrase left region) (Carias et al., 1998; León-Sampedro et al., 2016). The transferred 8–10 kb genetic regions carrying pbp5 characterized in this study were compared with those of all E. faecium available genomes at the National Center for Biotechnology Information (NCBI) database (Supplementary Table S5).

Comparative analysis of all PBP5 protein sequences available in the GenBank database with those identified in this study was performed using the basic local alignment sequence tool (BLAST) (Altschul et al., 1997) and the ClustalW2 program for multiple sequence alignment (Larkin et al., 2007; Goujon et al., 2010), hosted by the National Center for Biotechnology Information (NCBI) and the European Bioinformatic Centre (EBI), respectively. The PBP5 amino acid sequences were designated by a “C” followed by a number in agreement with the nomenclature used in previous works (Supplementary Table S6) (Galloway-Peña et al., 2011). Multiple sequence alignments of all PBP5 sequences were performed through MEGA 7 software⁴ and the topology of the phylogenetic tree was inferred by maximum-likelihood algorithm using PhyML (Guindon et al., 2010) with bootstrap analysis based in 1000 permutations and a cut-off of ≥70%.

Phylogenetic congruence of core genomes and pbp5 sequences was established by comparative analysis of their gene tree topologies. Core genomes for the 233 E. faecium strains whose draft genome sequences were available at NCBI Trace Database (last updated on February 2014), were reconstructed independently using an all-against-all reciprocal BLAST approach. Briefly, each CDS was used as a BLAST query against a local database of CDS from all genomes in the sample analyzed. Afterward, we re-annotated the chromosome sequences with GeneMarkS v2.8 (Besemer et al., 2001). The core genome was defined as the non-redundant genes present in all strains using CD-HIT (Li and Godzik, 2006) with the 80% coverage, 80% identity cut-offs, and parsed by homemade Perl script. All core genome genes were concatenated and aligned using MAFFT (Katoh and Toh, 2010). The phylogenetic tree was built by FastTree2 (Price et al., 2010). Tree comparison was carried out by cophyloplot command of APE (Analyses of Phylogenetics and Evolution) (Paradis et al., 2004) using R software.

Twelve isolates (n = 12/78; 15%; Table 1) were able to transfer under our experimental conditions a chromosomal genetic platform containing pbp5 (associated with AmpR) to E. faecium strain GE1. Five of them were also able to transfer pbp5 to E. faecium BM4105RF and two, to E. faecium 64/3. All donor strains belonged to major E. faecium human lineages mostly associated with clinical isolates causing hospital infections, namely BAPS subgroups 3.3a (ST670, ST132, ST280), 3.1 (ST280, ST125), and 2.1a (ST393). They were obtained from samples of hospitalized humans but also from samples of hospital sewage, healthy humans and piggeries. β-lactamase production was not detected in any of the isolates.

All transconjugants had similar PFGE profiles to recipient strains (Figure 1), were resistant to rifampicin and fusidic acid and exhibited a variable susceptibility to ampicillin (MICs = 8–128 mg/L), which was often lower than that of their corresponding donor isolates (8 mg/L or 16 vs. 32 mg/L to >256 mg/L). Besides the AmpR phenotype, some transconjugants also exhibited resistance to vancomycin (n = 8), teicoplanin (n = 8), erythromycin (n = 6), tetracycline (n = 1; in non-GE1 transconjugants; as GE1 strain is tetracycline-resistant), streptomycin (n = 1) and/or gentamicin (n = 1). Transconjugants carrying the pbp5 genetic platform either alone or with plasmids harboring genes encoding resistance to different antibiotics were recovered from selection plates supplemented with ampicillin and not with other antibiotics (data not shown).

Hybridization of pbp5 and 23S rDNA probes with a digested I-CeuI DNA band of high MW revealed a chromosomal location of pbp5. Further hybridization of SmaI-digested genomic DNA with the pbp5 probe showed hybridization to fragments of ∼210 kb in all but one GE1 transconjugants, for which the pbp5 probe hybridized to a band of ∼180 kb (Figure 1). Hybridization of the same probe with fragments of SmaI-digested genomic DNA of E. faecium BM4105RF and E. faecium 64/3 transconjugants of different sizes reflects independent transfer events. Similarity in the PFGE patterns of the transconjugants obtained using different donors indicates that DNA acquisition occurs within a particular region of the genome (Figure 1). The donor strains lack integrases/excisionases of Tn916-like conjugative transposons (Tn916, Tn5386, Tn5382).

For whole genome sequence a representative donor:recipient pair and its transconjugant were analyzed. The donor was the clinical HPH2 strain; the recipient, E. faecium GE1 which did not originally carry pbp5, and thus limiting the possibilities of recombination events between the acquired and natural pbp5 genetic platform in the recipient strain; and the size of the SmaI digested genomic DNA fragment, the one that most commonly hybridizing with the pbp5 probe (∼210 kb). Sequencing of the recipient (E. faecium GE1), the donor (E. faecium HPH2) and the transconjugant (E. faecium TCGEHPH2.1) allowed us to identify 7 contigs with genes present in the donor and its transconjugant but absent in the recipient strain. The fragments which comprised the pbp5 gene, represent a genetic platform of approximately 280 kb that is slightly larger than that inferred from the SmaI-PFGE gels. A more detailed analysis of the contigs revealed that two of them (contig00158 and to a lesser extent contig00068) also contained genes present in the recipient and transconjugants but not in the donor strain, which could be explained either by the partial transfer of the pbp5 genetic platform and/or post-transfer recombination events (Figure 2). The comparative analysis of the genomes sequenced in this work with the four closed E. faecium genomes available at the NCBI database at the time of writing this manuscript [DO (NC_017960.1), NRRL B-2354 (CP004063.1), Aus0004 (NC_017022.1), Aus0085 (NC_021994.1) strains], revealed a common region of 153kb for our donor/transconjugant strains and the four genomes analyzed, which was absent in the recipient E. faecium GE1 genome (Figure 3). The remaining sequence of the 280 kb “transferred chromosomal region” was variably present and located in different genomes (Figure 3).

Figure 2 shows a detailed characterization of the 280 kb transferred genetic platform, with ORFs of a G + C content ranging from 24 to 41%. Besides the pbp5-related resistance to β-lactams, this chromosomal transferable genetic platform contains genes involved in different cellular functions including amino acids and carbohydrate transport, redox processes, survival under stressful conditions in the intestinal environment (e.g., acid and bile tolerance) and also other genes related to β-lactam resistance. Among the last category is a gene (named ponA) encoding a bifunctional class A PBP with transpeptidase and transglycosylase activity, which is involved in the synthesis of peptidoglycan, allowing the survival in the presence of cell wall inhibitors (Rice et al., 2009). A copy of the CiaRH operon, coding for a two-component signal-transduction system responsive to cell envelope lysis-stress response and restricted to Streptococcus spp. to date, was also detected. The 280 kb genetic platform containing pbp5 also had genes with presumed influence on microbe-host interactions. These include three phosphotransferase systems (PTS, namely glucitol/sorbitol, L-ascorbate and mannose/fructose/sorbose), the ABC transport and transformation system of maltodextrines (malACDX) and the N-acetylmannosamine-6-phosphate epimerase (NanE; part of a pathway that allows the usage of sialic acids, major components of glycoproteins, gangliosides, and other sialoglyco-conjugates). Other metabolic genes in the pbp5 genetic platform could be involved in tolerance to the intestinal acidic environment, including glyceraldehyde-3-phosphate dehydrogenase, ATP synthase subunit α, NADH dehydrogenase, glutaminase and genes encoding enzymes involved in the production of ammonia from glutamine and deamination or transport of branched/nitrogenated amino acids. Finally, five stress response proteins including two belonging to the Csp (cold shock proteins) system (CspA and CspC) and the small chaperone Hsp20, and Gls33 (also present in E. faecium genomes available in the GenBank databases) were also identified. They are involved in stress responses to salts, pH and ethanol exposure in Clostridium. Hsp20 is a small chaperone protein involved in survival of different abiotic stress conditions including heat (55°C) and salt (5 mM) in Bifidobacterium longum (Khaskheli et al., 2015).

To date, only two E. faecium strains lack pbp5. These are strains GE1 and D344SRF, which are AmpS isolates. The D344SRF strain is susceptible to cephalosporin and ampicillin due to the spontaneous deletion of a 170 kb genome fragment that includes pbp5 and other pbp genes. This deletion occurred by the interaction of CTn5386 (a 60 kb element that comprises pbp5 and vanB2) with Tn916 (Rice et al., 2007). The GE1 laboratory strain does only harbor the integrase of Tn5801. The causes of the loss of the pbp5 genetic platform are still unknown.

The occurrence of different insertion sequences (ISs), (Figure 4) in the boundaries of the pbp5 gene and the presence of a >3027 bp psr-pbp5 fragment in E. hirae (99.9% identical to that found on E. faecium) suggest the presence of hot spots that could facilitate recombination of regions containing the pbp5. To test this possibility, we further analyzed the 8–10 Kb genetic environment of pbp5 from 21 wild type and 15 transconjugant isolates included in this study as well as available enterococcal genomes. Identical RFLP patterns of these 8–10 kb fragment were observed for each pair of donor and its corresponding transconjugants, but were different from those naturally occurring in the pbp5 carrying recipient strains E. faecium BM4105RF and 64/3 (Figure 4). Sequencing of fragments representing distinct RFLP patterns and comparative analysis with similar pbp5 genetic environment in GenBank available genomes, revealed 21 variants of such 8–10 kb chromosomal fragments (designated by roman numerals), which differed in the number and type of ISs (Figure 4). Three variants (types I, II, III) corresponded to pbp5 transferable platforms described in this study (Figure 4; Table 1).

The predominant 8–10 kb fragment identified in both AmpR and AmpS E. faecium analyzed in this study did not contain indels and is considered here as the pbp5 genetic environment prototype; arbitrarily named Type I (Supplementary Table S5, Figure 4 and 5). The pbp5 or other genes (e.g., psr. ftsw, ion efflux genes) were flanked by one or two insertions sequences of the IS256, ISL3 or IS982 families (ISEf1, ISEfm1, ISEfm2, IS1542, IS256-like, ISEfa11, or IS1251-like) in isolates carrying types II to XXI, with the exception of type VIII that had CTn5382 (vanB2) inserted downstream of pbp5. The identifiable boundaries of ISs detected within the 8–10 kb boundaries of pbp5, as well as the common nucleotide positions at which insertions occurred, suggested recent acquisition events at some hot-spots on an ancestral sequence (Figure 4). Despite of predominance of type I, some types seem to be more associated with specific hosts as the case of type V, predominant in strains from pigs and type II/II-like, III or type XIV-like which are frequent among clinical isolates. Epidemiological distribution of isolates appears in Supplementary Table S5.

We identified 75 PBP5 protein variants (Supplementary Tables S5 and S6) corresponding both to AmpR and AmpS strains which comprise 20 of the previously described variants C4, C7, C9, C11, C15-C18, C21, C24, C46-C48, C50, C51, C61, C63, C65, C67, C71 (Rybkine et al., 1998; López et al., 2009; Galloway-Peña et al., 2011). The other 55 variants were firstly detected in this study either in strains from Portugal (n = 11) or from available genomes at NCBI database (n = 44). The C4, C7, C8, C9, C12, C19, and C20 sequences were identified in isolates able to transfer pbp5 (Table 1, Supplementary Table S6; Figure 5), which belonged mostly to ST18, its SLV ST125, ST280 and ST670 (BAPS 3.3a, 3.1 and 2.1a, respectively) (Table 1).

Figure 5 shows the phylogenetic tree constructed with all PBP5 protein sequences from this study and those available at GenBank databases. The tree is split into two major clades arbitrarily named B and A, mirroring the clades associated with populations of non-hospitalized persons and hospital isolates respectively, which were previously inferred from phylogenomic studies of E. faecium (Lebreton et al., 2013). Clade B comprises PBP5 of AmpS isolates (PBP5-S), which mainly correspond to strains of BAPS subgroups 1.2 and 1.5. Some are similar to the prototype PBP5-S C46 sequence of E. faecium BM4107 strain (Sifaoui et al., 2001) but most of them also exhibit mutations at positions T25A, S39T and D644N, which are also common to PBP5 sequences of clade A isolates (Supplementary Table S6). Two strains, isolated in 1964 and 2006, further showed changes at S27G + T324A and S27G respectively, such mutations corresponding to the PBP5-R consensus sequence (Pietta et al., 2014). The type I chromosomal region mentioned above was observed in all available isolates from this group (Figure 5; Supplementary Table S5).

Clade A includes PBP5 variants grouped in two main clusters arbitrarily designed here A0 and A1, with most isolates sharing mutations at positions V24A, S27G, E100Q, K144Q, T172A, T324A, N496K, A499T and E525D (7 of the 21 positions used to establish sequence diversity of this protein) (Pietta et al., 2014). The Clade A0, represented by C30, C31, C37, C52, C53, and C54 PBP5 sequences, corresponds to AmpS isolates of different BAPS groups (BAPS 1.2, BAPS 2.3a and BAPS 3.3b), recovered from different hosts (animals and humans), different countries and collected from 1995 to 2001 (Figure 5, Supplementary Table S6). Most isolates within clade A0 carried pbp5 within a type I fragment. Sequences of Clade A1 were classified in three subclusters that parallels E. faecium populations of BAPS groups 2 and 3 (Willems et al., 2012; Tedim et al., 2015). They were designated as A1S, comprising a subset of PBP5-AmpS variants (mainly associated with human and animal isolates of BAPS subgroup 2.1b); A1M, comprising a subset of PBP5-AmpS and PBP5-AmpR (mostly associated with animal isolates of diverse BAPS subgroups 2.1b, 2.3a, 2.3b, 3.1, 3.2, and 3.3b); and A1R, including almost all PBP5-AmpR (mostly associated with the clinical setting and BAPS subgroups 3.1, 3.3a and 2.1a, the latter two only observed in this group).

All isolates from Clades A1S, A1M and A1R share mutations R34Q, G66E, L177I and A216S but also present some differences. Mutations S39N and A401S plus A499I were specific for subclades A1S and A1M, respectively. The last two polymorphisms were previously documented in two AmpR strains (Pietta et al., 2014) but this study suggests that they were fixed in certain populations despite having been previously discarded as relevant changes. One of the A1M strains had also variations at specific positions linked to the A1R group (A68T, E85D, M485T, V586L, E629V and P667S). Of note is the PBP5 variant C23 within this cluster, which was overrepresented (n = 44).

PBP5 variants within the A1R group exhibited six mutations (A68T, E85D, S204G, 466′S/D, M485A/T, E629V, P667S) predominant in this group, some of them located in the active site of the protein (466′S/466′D, M485A/T) and in the end of a turn between the β1 and β2 strands (E629V; P667S) (Fontana et al., 1996; Rice et al., 2004). It is of note that some AmpS isolates (BAPS subgroups 3.1, 2.3a and 3.2) with PBP5 sequences that cluster in the A1R subgroup (including the PBP5 of the recipient E. faecium 64/3) lack the mutation M485A and E629V, which confirmed that such mutations might be necessary for AmpR phenotype, as reported (Rice et al., 2004). Changes A68T, E85D and S204G were shared by AmpS and AmpR isolates of A1R group (Supplementary Table S6). Some strains exhibited the particular pattern of mutations Q408H, A558T, G582S, K632Q and, eventually, V462A, N546T and P642L. They had not previously been associated with AmpR. Distribution of different variants of the 8–10 kb boundaries of pbp5 gene is shown in Figure 5 and Supplementary Table S5. Of note, is the detection of type I in isolates of different groups of clade A1 as well as the association of types V and XI with isolates of group A1M and of the similar types II, III, XIV, XV, and XVII with isolates of group A1R.

We also analyzed the diversity of other genes previously associated with AmpR (ddcP, ddcY, ldtEfm, pgtA, lytG) in available genomes, (protein sequence per UniRef100 available at Uniprot using as reference E. faecium strain Aus0004). Single-locus phylogenetic tree of PBP5 was congruent with those of proteins codified by ldtEfm and ddcY but was non-congruent with those of ddcP, pgtA and lytG (data not shown).

Analysis of gene tree topologies (E. faecium core genomes vs. E. faecium pbp5 genes) revealed discrepancies that indicate transfer of pbp5 between E. faecium populations (Figure 6) and explains the occurrence of both the same PBP5 variants in isolates of different clonal lineages, and different PBP5 sequences in strains of the same phylogenomic groups (Figure 5, Supplementary Table S5).

AmpR was maintained in the wild-type strains and respective E. faecium GE1 transconjugants over 200 generations in antibiotic free BHI indicating that the acquired genetic platforms containing pbp5 may persist in absence of selective pressure in different genetic backgrounds, suggesting no-fitness cost imposed by the acquired pbp5 platforms. The morphology and growth rate of all evolved transconjugants was similar among each other and to the normal growth of the E. faecium GE1 recipient strain.