Interactions between commensal Enterococcus faecium and Enterococcus lactis and clinical isolates of Enterococcus faecium

Abstract Enterococcus faecium (Efm) is a versatile pathogen, responsible for multidrug-resistant infections, especially in hospitalized immunocompromised patients. Its population structure has been characterized by diverse clades (A1, A2, and B (reclassified as E. lactis (Ela)), adapted to different environments, and distinguished by their resistomes and virulomes. These features only partially explain the predominance of clade A1 strains in nosocomial infections. We investigated in vitro interaction of 50 clinical isolates (clade A1 Efm) against 75 commensal faecal isolates from healthy humans (25 clade A2 Efm and 50 Ela). Only 36% of the commensal isolates inhibited clinical isolates, while 76% of the clinical isolates inhibited commensal isolates. The most apparent overall differences in inhibition patterns were presented between clades. The inhibitory activity was mainly mediated by secreted, proteinaceous, heat-stable compounds, likely indicating an involvement of bacteriocins. A custom-made database targeting 76 Bacillota bacteriocins was used to reveal bacteriocins in the genomes. Our systematic screening of the interactions between nosocomial and commensal Efm and Ela on a large scale suggests that, in a clinical setting, nosocomial strains not only have an advantage over commensal strains due to their possession of AMR genes, virulence factors, and resilience but also inhibit the growth of commensal strains.


Introduction
Enterococcus faecium colonizes the gastrointestinal tract of healthy individuals and animals but has in recent years emerged as a nosocomial pathogen (Arias and Murray 2012 ).Enterococcus faecium infections are an increasing concern in health care, due to its high intrinsic antimicr obial r esistance, its ca pacity to acquir e nov el r esistance genes, and its ability to withstand harsh conditions, including disinfectants (Arias and Murray 2012, Wagenvoort et al. 2015, Pidot et al. 2018 ).
The clade structure of E. faecium has been characterized by a deep phylogenetic split, with clade B dominating in the comm unity (Lebr eton et al. 2013 ), whic h was r ecentl y r eclassified as Enterococcus lactis (Belloso Daza et al. 2021 ).E. faecium clades are further split into clade A1 and A2, where clade A1 almost ex-clusiv el y accounts for infections and ov erla ps with the former clonal complex 17 (CC17) (Leavis et al. 2007, Guzman Prieto et al. 2016 ).Clade A2 has historically been mainly associated with livestock and domestic animals (Lebreton et al. 2013, Gouliouris et al. 2018 ), and non-hospital-associated human isolates (Arredondo-Alonso et al. 2020 ).The clades differ in their accessory and core genome, enabling adaption to different niches (Gallo w ay-P eña et al. 2012, Palmer et al. 2012, van Hal et al. 2021, AL-Rubaye et al. 2023 ).The accessory genome of clade A1 is enriched in acquir ed antibiotic r esistance determinants (Leavis et al. 2006 ), genomic islands (van Schaik et al. 2010 ), specific insertion sequences (Leavis et al. 2007, Werner et al. 2011 ), and virulence factors (Gao et al. 2018, AL-Rubaye et al. 2023 ).Enterococcus lactis is gener all y more antibiotic susceptible, while clade E. faecium A1 str ains ar e fr equentl y r esistant to m ultiple antibiotics, including ampicillin, due to differences in pbp5 sequence and its expression (Pietta et al. 2014 ), and also acquired resistance to vancomycin, aminoglycosides, and linezolid (Willems et al. 2005, AL-Rubaye et al. 2023 ).Clade A1 strains are predominant in hospitalized patients (Ubeda et al. 2010, Taur et al. 2012 ) and infection is gener all y pr eceded by asymptomatic colonization with antibiotic-resistant strains and within-host evolution (Ubeda et al. 2010, Moradigaravand et al. 2017, Ba yjano v et al. 2019 ).Colonized patients may then contaminate healthcare workers as well as their surr oundings, whic h is acceler ated by E. f aecium's ability to withstand harsh conditions (Wendt et al. 1998, Pidot et al. 2018 ), in turn leading to outbreaks (de Regt et al. 2008 ).Additionally, it has been shown that commensal but not clinical E. faecium strains are susceptible to group IIA-Secreted phospholipase A 2 in human serum, although Grampositiv e bacteria ar e consider ed r esistant to killing by serum (Paganelli et al. 2018 ).
In a murine gastrointestinal colonization model with systemic β-lactam administration, clade A1 strains dominated over clade B/ E. lactis strains (Singh et al. 2022 ).Ho w e v er, in the absence of antibiotics, clade B/ E. lactis strains outnumbered clade A1 strains and persisted longer in the same model, suggesting that clade A1 str ains ar e r eplaced by clade B/ E. lactis str ains once the patient leaves the hospital (Montealegre et al. 2016 ).
While it seems a ppar ent why antibiotic-resistant A1 strains would replace antibiotic susceptible commensal strains, it is an open question whether clade A1 strains can also suppress their growth, or vice versa.
A mode by which bacteria can inhibit the growth of competitors is the release of bacteriocins, and enterococci have been described as one of the most frequent bacteriocin producers (Almeida-Santos et al. 2021 ).Bacteriocins are defined as ribosomally synthesized antimicrobial proteins or peptides, which either remain unaltered (class II bacteriocins) or are post-translationally modified by biosynthetic enzymes (class I bacteriocins).The biosynthetic gene clusters consist of genes encoding the bacteriocin itself and/or its biosynthetic enzymes, export pr oteins, pr oducer imm unity mec hanisms, and, sometimes, regulators of the bacteriocin production (Heilbronner et al. 2021 ).The antibacterial activities of bacteriocins ar e v ery div erse and r ange fr om the degr adation of essential cellular components to the inhibition of specific molecular tar gets, suc h as enzymes r equir ed for the synthesis of the cell wall or cell membr ane, pr oteins or nucleic acids, to the disintegration of bacterial membranes, but much of the mode of action of bacteriocins is not fully understood due to their complexity (Simons et al. 2020 ).Bacteriocin production plays a crucial r ole in sha ping the micr obiome, and they can pr e v ent or pr omote the growth of a bacterial community by another strain or can lead to the redistribution of microbiome members into sub-niches and pr otect a gainst colonization by bacteriocin-susceptible in vaders .This can be utilized in a medical context by using probiotics for pathogen exclusion.In contrast to the majority of current antibiotics, most bacteriocins exhibit a narrow spectrum activity and could thus be attr activ e a gents for pr ecision ther a py (Heilbr onner et al. 2021 ).
Enterococcus -produced bacteriocins, known as enterocins (Brandis and Brandis 1963 ), predominantly fall into class II, c har acterized by their small size ( < 10 kDa) and heat stability.Within class II, further subdivisions include class IIa (e.g.enterocins A and P), class IIb (e.g.enterocin C, 1071, and X), class IIc (e.g.bacteriocin AS-48 and enterocin 4), and class IId (e.g.Enterocin Q and L50) (Almeida-Santos et al. 2021 ).Enter ocin pr oducers hav e been studied in different contexts, especially against food-borne pathogens , and ha v e been pr oposed as pr obiotics, ho w e v er, safety concerns have been raised (Hanchi et al. 2018 ).It has also been discussed that bacteriocins may offer ther a peutic options, either alone or in combination with other antimicrobials (Almeida-Santos et al. 2021 ).Case studies illustrate the feasibility of using bacteriocins a gainst v ancomycin-r esistant enter ococci (VRE) (Phumisantiphong et al. 2017, Bucheli et al. 2022 ).The diversity, classification, potential use, and concerns regarding enterocins have been described in detail in a recent review article (Almeida-Santos et al. 2021 ).
Pr e vious studies on interactions between E. faecium clades used a limited number of isolates to r epr esent the different clades, thus it is unknown how generalizable the observations are .Here , we used a large collection of 125 isolates, r epr esenting the clinical clade A1 and the commensal clade A2 and E. lactis , to study the interactions between clinical and commensal isolates.

Bacterial strain collection
The bacterial strain collection was set up to r epr esent 50 clade A1, 25 clade A2 and 50 E. lactis (former clade B) isolates.E. lactis and A2 isolates were selected from the Tromsø 7 human faecal adult population sample collection (Norwegian National Advisory Unit on Detection of Antimicrobial Resistance (K-res)) and 1 strain from the Netherlands which has previously been shown to be susceptible to human serum (P a ganelli et al. 2018 ).Clade A1 isolates r epr esenting CC17 wer e selected fr om the Norwegian Surv eillance System for Antibiotic Resistance in Microbes (NORM) 2008 and 2014 collections and Norwegian Vancomycin-resistant E. faecium (VRE fm ) collection from 2010 to 2015 (AL-Rubaye et al. 2023 ) as well as VRE isolates from 2019 r eceiv ed at K-res .T he isolates were c hosen to r epr esent a wide r ange of sequence types (ST) and to cover the species phylogeny.Detailed information on the bacterial strain collection is given in Fig. 1 and Supplementary Table S1 .

Spot on lawn screening
For the initial inter action scr eening, the indicator was applied to a brain heart infusion (BHI) plate as a lawn and the putative inhibitor was added on top.For this, the indicator was pic ked fr om a blood agar plate and diluted to 0.5 McFarland in 5 ml 0.85% NaCl then diluted 1:10 in 5 ml 0.85% NaCl.This solution was applied with a cotton swab onto a BHI agar plate using a spiral plater for rotating the plate and dried for 5 min before applying the inhibitor.The bacterial solution in NaCl equals 3 × 10 6 ± 7 × 10 5 CFU/ml (determined by CFU count n = 9) and since a cotton swab takes about 150 mg of liquid, 5 ± 1 × 10 5 CFU/plate is applied.The inhibitor w as gro wn ov ernight in 5 ml liquid BHI at 37 • C with 220 r pm shaking, and 10 μl of this culture was applied on top of the lawn.Inhibition was read after 18 ± 2 h of incubation at 37 • C, according to the scoring illustrated in Fig. S1 : 3 -definite inhibition with a wide zone (Inhibition ++ ), 2 -definite inhibition with an intermediate zone (Inhibition + ), 1 -definite inhibition with a narrow zone (Inhibition), 0.5 -non-definite inhibition zone with colonies growing within the zone (Undefined inhibition), 0 -no inhibition.
T he a v er a ge inhibition scor e per str ain was calculated as the sum of inhibition (0 to 3) divided by the number of interactions.

Supernatant on lawn
The supernatant of strains that scored 1 in at least 2 interactions was used for further assa ys .T hr ee r epr esentativ e str ains of each clade were used as indicators and plated out as a lawn.Supernatants were obtained by centrifuging 20 ml ov ernight cultur es at 7000 rpm for 10 min at 4 • C, supernatants were filtrated through a 0.45 μm syringe filter follo w ed b y a 0.2 μm syringe filter (PES, VWR, US) to avoid clogging of the 0.2 μm filter.Sterile supernatants were concentrated 5 times their volume using a 3 MWCO filter (Vivaspin 20, Merc k, German y) at 7000 r pm.Sterile filtr ated concentr ated supernatants wer e tr eated with pr oteinase K (ThermoFisher Scientific, US) (3 μl of 10 mg/ml in 15 μl supernatant for 2 h at 37 • C) or heat-inactiv ated pr oteinase K (100 • C for 10 min) or heat (100 • C for 10 min). 10 μl of untreated and treated supernatant was applied on top of the lawn and inhibition was read after 18 ± 2 h of incubation at 37 • C, according to the scoring described ( Fig. S1 ).The supernatant of strains that sho w ed at least one grade 3 inhibition despite pr oteinase K tr eatment, was inv estigated further.One ml of the indicator at 0.5 McFarland was added in 10 ml top-agar (0.5% a gar, Sigma Aldric h, US, in BHI), whic h was pour ed on a BHI a gar plate and dried for 30 min.Tenfold dilutions of the supernatants in PBS (10 −1 to 10 −8 ) were added on top and the plates were read after 18 ± 2 h of incubation at 37 • C. PHASTER (PHAge Search Tool Enhanced Release) (Arndt et al. 2016 ) was used to identify and annotate pr opha ge sequences within the genomes of the 11 str ains, which sho w ed at least one gr ade one inhibition after pr oteinase K treatment.

Prediction of bacteriocins using a novel database
The whole genome sequences (WGS) of all strains were screened against a bacteriocin database that has recently been published (Tedim et al. 2023 ) using CGE MyDbFinder with a cut-off of 80% identity and 80% cov er a ge.Briefly, this bacteriocin database includes 76 Bacillota bacteriocins (mostly from enterococci) that have either been previously described or detected in the process of creating this proprietary database.

Isolation of a ptsD Tn mutant
To study the involvement of ptsD, we used a mariner Tn mutant library in the strain E8202, as described in (de Maat 2022 ).A tr ansposon m utant with an insertion in the gene encoding ptsD ( ptsD locus NZ_LR135344.1 711045..711866; E1162 locus tag EfmE1162_1918) was isolated from an E. faecium E8202 (hospitalized patient isolate , T he Netherlands , 2015, E745) (Top et al. 2020 ) Tn mutant library ( mariner Tn cassette carrying a gentamicin (GM) resistance gene) as described pr e viousl y (Zhang et al. 2017 ), using the ptsD -specific primers ptsd_Efm_Tn_Fw 5 -CGGAAGATGTTTTGGCGCTC-3 and ptsd_Efm_Tn_Rv 5 -TCCCAA GA CGA CCATTCCAAA-3 as well as the bidir ectional primer, whic h is complementary to the r epeats flanking the mariner Tn sequence, ftp_tn_both_ends_MmeI 5 -CGGGGACTT A TCA TCCAACC-3 .The mutant E. faecium E8202 ptsD ::GM was first confirmed by in-house Sanger sequencing of the ptsd_Efm_Tn_Rv-ptsd_Efm_Tn_Rv PCR product, and then genome sequenced using the Illumina MiSeq paired-end platform.Genomic comparison of the Tn mutant and wild-type was performed by using Nullarbor v.2.0.20191013 pipeline ( https://github.com/ tseemann/ nullarbor), resulting in one synonymous core SNP between the strains (pos 855 132, CGC to CGA, silent (both arginine)).Abricate v.1.0.1 with a custom database and a minimum identity and cov er a ge of 60% was used in confirming the presence of the insertion only in the mutant.The genomes were further compar ed with pgv-mauv e v.0.3.2 to exclude an y other genomic differences.

Characterization of the ptsD Tn mutant
Ov ernight cultur es of the wildtype strain E8202 and its mutant E8202 ptsD ::GM were diluted 1:100 in BHI, BHI with 2 g/L mannose, Müller Hinton (MH) broth, MH with 2 g/L mannose (Sigma-Aldrich, US), Lysogen y br oth (LB) or LB with 2 g/L mannose and growth was measured in an Epoch 2 Spectrophotometer with Gen5 Software (BioTek Instruments Inc., Winooski, Vermont US) at 37 • C, shaking at 425 rpm, with OD 600 measurement every 10th min for 18 h (n = 6, biological triplicates, technical duplicates).E8202 and its mutant E8202 ptsD ::GM were subjected to the spot on lawn assay as described above and in addition to BHI, the assay was also conducted on BHI with 2 g/L mannose, LB, and MH agar plates.

Selection of representati v e strains
The str ains wer e c hosen to r epr esent the div ersity of clades A1, A2 ( E. faecium ), and B ( E. lactis ), as shown in Fig. 1 .All A1 strains originate from clinical samples, while all A2 and B strains are from faeces of non-hospitalized individuals.Detailed strain characteristics ar e giv en in Table S1 .A1 is the least diverse clade, while E. lactis shows the highest br anc hing.

Clinical A1 strains inhibit commensal strains
Spot on lawn assays were conducted with overnight cultures of one clade and lawns of another clade .T hese spot on lawn assays sho w ed that clade A1 strains inhibited the growth of clade A2 and B ( E. lactis ) strains to a higher degree than vice versa (Fig. 2 , 3 ).While only 36% of commensal isolates (48% of A2 and 30% of B/ E. lactis ) could inhibit A1 strains, 76% of the clinical A1 strains could inhibit commensal strains .T he sum of inhibition values (giving eac h inter action a v alue fr om 0 to 3) for A1 strains was 4742, while it was 2333 for commensal strains .T he strains which sho w ed the highest inhibition score belonged to ST5, ST165, ST69, ST2027, and ST1940 in A2, to ST2016, ST60, ST361 and ST96 for E. lactis and to ST787, ST78, ST117, ST80, ST192, and ST203 for A1.Phylogeneticall y r elated str ains show a similar inhibition pattern.The highest inhibition score of clade A1 strains was mediated by the E. lactis (clade B) 51003823 ST94 and 51015840 ST2016 as well as clade A2 strains 50980395 ST165 and 51021120 ST5.The highest inhibition score of commensal strains was mediated by the clade A1 strains 51269070 ST117 and 51273089 ST192 (Fig. 2 , Table S1 ).
Av er a ge inhibition scor es wer e calculated as the av er a ge inhibition per interaction.Clade A1 inhibited commensal strains at a significantly higher level (Fig. 3 ).Strains that sho w ed at least one grade 2 inhibition were used for further investigation, that is 35 out of 50 A1 strains, 10 out of 25 A2 strains, 16 out of 50 B/ E. lactis .

Inhibition is mediated by secreted, heat-stable, proteinaceous compounds
Whether supernatants were responsible for the inhibition seen in the initial spot on lawn assay screening was evaluated.Five times concentrated supernatants of all strains, which sho w ed at least one grade 2 inhibition in the assay using ov ernight cultur e wer e used on lawns of three strains representing A1, A2, and B (Fig. 4 ).
The concentrated supernatants exhibited a similar inhibitory activity as the bacterial overnight culture.Most inhibition was observed mediated by the supernatant of clade A1 strains compared to the other clades.As a control, the flow through from concentr ation (pr oteins < 3 kDa) was also tested on the lawn, yielding no observable inhibition.Among the concentrated supernatants, strains 51271945 ST78, K59-17 ST22, and K60-7 ST578 from clade A1 exhibited the highest inhibition scor es.Notabl y, the concentr ated supernatants wer e mostl y heat stabile but susceptible to pr oteinase K tr eatment (2 mg/ml for 2 h at 37 • C).It is noteworthy that a gentler pr oteinase K tr eatment (1 mg/ml for 1 h at 37 • C, see Fig. S2 ) did not eliminate the inhibitory effect.Upon heatinactivation of proteinase K (10 min at 100 • C), the inhibitory activity was r estor ed (see Fig. S2 ).
The inhibition mediated by concentrated supernatants of clade A1 str ains a gainst clade A2 and B str ains could not be full y abolished by proteinase K treatment, indicating the potential involve-ment of non-proteinaceous compounds .T he strains 51276488 ST117, K59-36 ST575, K60-7 ST578 and K60-14 ST192 sho w ed grade two inhibition even after proteinase K treatment (2 mg/ml for 2 h at 37 • C).Hence, we suspect the involvement of phages or other agents that are not sensitive to serine proteases.In silico analysis in Phaster sho w ed that these strains carry one intact pr opha ge r egion eac h ( Fig. S3 ).Out of the other nine str ains whic h sho w ed at least one grade one inhibition after proteinase K tr eatment, thr ee str ains did not contain pr opha ge r egions (51269068, K59-55 and K60-14), four contained one pr opha ge ).Since we suspected that phages could be involved in the phenotype of the supernatants of 51276488, K59-36 and K60-7, we expected that single plaques would be visible upon dilution of the native concentrated supernatants.Ho w ever, 10 −1 dilution sho w ed grade 1 inhibition, while dilutions 10 −2 to 10 −8 sho w ed no inhibition (data not sho wn).Still, w e cannot exclude the involvement of phages and further investigation would be needed.
Strains that sho w ed high inhibition and contained many bacteriocin genes could be good candidates for antibiotic de v elopment, specificall y a gainst A1 or as pr obiotic str ains.Commensal str ains carrying entB, enxA and enxB or entL50A and entL50B all inhibited clade A1 strains, and entL50AB, bacAS4 and duracin are exclusive to commensal strains in this strain collection.The commensal A2 strain 51010478 ( entA,bacAS32,bacAS3,entP,entL50AB ,entB ,enxAB ) stands out as a good probiotic candidate since it sho w ed high inhibition to clinical strains and the combination of entL50AB with entB and enxAB is unique to this strain.The A2 strain 50976613 also sho w ed high inhibition and high bacteriocin count, ho w e v er, its bacteriocin gene profile is the same as 51024665, which does not show any inhibition, thus the predicted bacteriocins of 50976613 are most likely not responsible alone for the observed inhibition phenotype.
Clinical strains that show high inhibition and contain many bacteriocin genes can indicate which bacteriocins are clinically r ele v ant.The bacteriocin genes bacAS11 , bacAS5, entQ , and bac43 ar e exclusiv e to clade A1 and corr elate with an inhibition phenotype (clade B/ E. lactis strain 50966714 has a 98.67% identical variant of bac43 but does not show inhibition in supernatant assa y).T he gene encoding bacteriocin 43 is often co-present with the plasmid replicase rep 18b (10 bac43 and rep18b /14 bac43 , strains with bac43 but without rep 18b are 51273106, 51269059, K60-15 and KresVRE0016).Clades A1 and A2 share entB and enxAB , while A2 and B/ E. lactis share bacAS4 and duracin , but none of the bacteriocin genes found in A1 and B/ E. lactis were absent in A2.The genes bacAS3, bac32 and bac51 also correlated with an inhibition phenotype and were found in clinical and commensal strains.BacAS31 was unique to A2, while bacAS8 , entSE-K4 and GM-1 were unique to clade B/ E. lactis .
The two strains, clade B/ E. lactis strains 51001117 and 50981525, sho w ed inhibition in the spot on lawn interaction screening, but no bacteriocin genes were found.Of note, neither of the strains sho w ed inhibition in the supernatant assa y.T hus , certain cues might be necessary for their inhibition phenotype or other factors might be involved.

The potential role of ptsD in susceptibility to inhibitors
Given that some mannose-specific phosphotransferase systems (PTS) have been documented as a receptors for class II bacteriocins and bacteriophages in various bacterial species (Jeckelmann and Erni 2020 ), and considering that a mannose PTS encoded by ptsD and exclusive to clinical strains (referred to as pts Clin ) of E.
has been identified (Zhang et al. 2013 ), we proceeded to explor e the involv ement of ptsD in the observed inhibition.All A1 strains, except K59-44 and K59-46, carried ptsD , while none of the Clade A2 or B strains carried this gene ( Fig. S5 ).Ther efor e the Tn mutant E8202 ptsD ::GM was isolated from an E8202 Tn mutant library and confirmed by WGS.E8202 and its mutant E8202 ptsD ::GM did not show significant growth differences in BHI, BHI with 2 g/L mannose, LB, LB with 2 g/L mannose, MH or MH with 2 g/L mannose ( Fig. S6 ).They also did not show a difference in susceptibility to inhibition by other A1, A2 or B strains in spot on la wn assa ys ( Fig. S7 ).

Discussion
Our results show that clade A1 isolates can suppress the growth of commensal isolates .T his ma y help to understand the dynamics of clade A1 carriage preceding infection in a nosocomial setting.Our findings are in line with the results from a murine gastrointestinal colonization model with systemic β-lactam administr ation wher e clade A1 str ains dominated ov er E. lactis (clade B) strains (Singh et al. 2022 ) and in the absence of antibiotics, E. lactis (clade B) outnumbered A1 strains (Montealegre et al. 2016 ).Of note, the strength of these two studies is the use of a murine gastrointestinal colonization model, ho w ever, they only used a total of 12 strains and the use of antibiotics may limit the comparability to our findings.While many previous studies were biased to w ar ds either clinical m ultidrug-r esistant (MDR) E. f aecium str ains or nonhuman envir onmental E. f aecium str ains or wer e limited in strain number, the strength of our study is the balance of human clinical isolates and human commensal isolates and the high number of isolates.
In the spot on lawn screening the cell density of the spot is m uc h higher than the lawn, which represents the conditions of ov er gr owth.Her e, we found that nosocomial clade A1 strains inhibit commensal strains significantly more than the other way ar ound.With onl y 3 exceptions, all other 72 commensal strains are susceptible to inhibition by multiple nosocomial clade A1 str ains, whic h can partiall y explain the predominance of these in infections .T hr ee of the nosocomial ST192 str ains whic h show high inhibition belong to the core genome MLST cluster type 3 which has been described as a successful clone in Norway (AL-Rubaye et al. 2023 ).The A1 STs showing inhibition of commensal str ains ar e known as inv asiv e in Eur ope (Werner et al. 2020, AL-Rubaye et al. 2023 ).Inter estingl y, we also found se v er al commensal str ains, whic h can inhibit all clinical str ains, including VRE and other MDR strains .T hese strains ha ve high potential for use as pr obiotics.Mor eov er, since we also show that many commensal strains lack the ability to inhibit nosocomial str ains, pr obiotic strains should be chosen carefully after phenotypic characterization to fully harness their potential.It is a limitation of this study that cell-to-cell contact or other cues (Gonzalez and Mavridou 2019 ) which might be r equir ed to trigger the phenotype of some interactions was not considered.Pilot experiments where both strains were grown together, and the supernatant from this co-culture was isolated, did not show inhibition.The presumable reason is that the supernatant represents a mixture of secretions of both strains and thus also putative immunity proteins, as well as a dilution of putative effectors.Future studies could investigate whether certain stressors can enhance interactions.
Since the interactions were mostly mediated by heat-stable pr oteinaceous secr eted compounds, the pr esence of bacteriocins in the genome sequences was predicted using a novel database (Tedim et al. 2023 ).We observed that clades A1 and A2 encoded significantly higher numbers of bacteriocins compared to E. lactis .The number of bacteriocins is in line with the results from our inter action scr eening, wher e clinical str ains inhibit commensal strains significantly more often than the other way around.Many of the strains showing a high inhibition phenotype also encoded a high number of bacteriocins, such as entA (Aymerich et al. 1996, Fugaban et al. 2021), bac43 (Todok or o et al. 2006 ), entP (Cintas et al. 1997 ), bac32 (Inoue et al. 2006 ), bac51 (Yamashita et al. 2011 ) and entL50A and B (Cintas et al. 1998 ).All known predicted bacteriocins were of Class II, meaning ribosomally synthesized antimicrobial proteins or peptides, which do not undergo posttranslational modification.In addition, nine genes encoding no vel bacteriocins , called bacAS# , were detected in the strain sequences .T he bacteriocin encoding genes entB, enxA and B , entL50A and B were all found in strains able to inhibit clade A1 strains, and entL50A and B ar e exclusiv e to commensal str ains .T hese bacteriocins would thus be good candidates for antimicrobial development or as adjuvants of antibiotics .T he bacteriocin encoding genes bac43, bacAS11 and bacAS5 were exclusive to clade A1 and co-occurred with an inhibition phenotype .T hey thus stand out as clinicall y r ele v ant bacteriocins.Bacteriocin 43 has originally been discov er ed in a clinical VR E. faecium strain from the US and was described as a Class IIa bacteriocin, active against several enterococcal species and located on mobilizable plasmids (Todokoro et al. 2006 ).It was later also discov er ed on small theta-replicating plasmids ( rep 18b ) of differ ent VR E. f aecium str ains fr om hospitalized patients from Germany and Canada (Freitas et al. 2016 ).In our strain collection, the bac43 gene was found in VR E. faecium ( v anA or v anB ) and in VSE str ains and bac43 was mostly co-present with rep 18b (10 bac43 and rep18b /14 bac43 ) which is in line with the previous findings.Of note, some bacteriocins, such as Enterocin Q (Cintas et al. 2000, Criado et al. 2006a ,b ), ar e optimall y pr oduced at higher temper atur es and might thus not have been responsible for the phenotype we see in the interaction screening.
It is a limitation of this study, that the regulation and expression of the individual enterocins were not studied.Production, release, and activity of bacteriocins, are tightly controlled by a complex network of genetic elements.For example, the expression of cytolysin is tightly controlled by a two-component regulatory or a quorum-sensing system (Haas et al. 2002 ).Also, some Class IIa bacteriocins with a double glycine leader, which also includes a fe w enter ocins, ar e r egulated by a thr ee-component r egulatory system that encompasses a peptide pheromone, a membranebound histidine protein kinase that serves as receptor for the peptide pheromone, and finally, a response regulator protein that activates the operons participating in the bacteriocin biosynthesis upon phosphorylation (Nes et al. 2014 ).Future studies could also study the bacteriocin-operons' immunity genes.
The presence or absence of bacteriocin genes could not explain all phenotypes that we observed in the spot on lawn interaction screening.For example, some strains share the same bacteriocin genes, but show a different inhibition pattern, while others share an inhibition pattern but hav e differ ences in their bacteriocin genes.We thus suspect that there might be other bacteriocins or yet other factors than bacteriocins involved.Since some of the inhibition mediated by clade A1 supernatants was not abolished by proteinase K treatment, we suspect the involvement of non-pr oteinaceous a gents, suc h as lipopeptides or pha ges, and in silico analysis in Phaster sho w ed that these strains carry an intact pr opha ge r egion.
A receptor for bacteriocins and bacteriophages across bacterial species is the mannose-specific phosphotr ansfer ase system (PTS) which also acts as a mannose transporter (Jeckelmann and Erni 2020 ).In E. faecalis , it has been shown that the mpt operon, encoding a mannose phosphotr ansfer ase system, is involv ed in bacteriocin susceptibility (Héchard et al. 2001, Opsata et al. 2010 ).Bacteriocin resistance of E. faecalis was linked to reduced expression of the mpt operon, and an mptD mutant was bacteriocin (pediocin PA-1 and mesentericin Y105) r esistant (Héc hard et al. 2001, Opsata et al. 2010 ).Mor eov er, mannose-induced PTS expr ession leads to enhanced sensitivity of E. faecalis JH2-2 to the bacteriocin mesentericin Y105 (Héchard et al. 2001 ).In E. faecium , a mannose PTS, encoded by ptsD exclusive to clinical strains, is involved in colonization by clinical strains during antibiotic treatment (Zhang et al. 2013, AL-Rubaye et al. 2023 ) and it was suggested that the mannose PTS system of E. faecium could be a mean to control E. f aecium (Somar ajan and Murr ay 2013 ).It was also speculated that ptsD could be targeted by bacteriocins (Zhang et al. 2013 ), since mannose PTSs are common targets of bacteriocins (Kjos et al. 2010, Opsata et al. 2010 ).In our study, we did not observe a difference in susceptibility to inhibition between E8202 and its mutant E8202 ptsD ::GM.This might be because the secretions of the inhibitors contain multiple compounds and e v en if one lac ks its target the other compounds could still exert the phenotype.In addition, E8202 could compensate for the lack of ptsD by the use of another mannose PTS.Potentially, the effect of a compound on ptsD can only be seen for isolated compounds at higher concentrations and the spot on la wn assa y might ha v e lac k ed the sensiti vity to detect this .Furthermore , it could be that the whole pts operon or a different component than ptsD is targeted since an extracellular loop of the membrane-located protein MptC was responsible for specific target recognition by the class IIa bacteriocins in Listeria monocytogenes (Kjos et al. 2010 ).
In summary, our study offers critical insights into the dynamics of E. faecium colonization within both hospital and community en vironments .Specifically, in a hospital setting, we predict the displacement of commensal strains by nosocomial clade A1 strains.This replacement is driven by the pronounced ability of clade A1 strains to inhibit clade A2 and B strains and will be propelled in the pr esence of selectiv e pr essur e by antimicr obials, wher e the AMR profile will give clade A1 strains an ad ditional ad vantage over sus-ceptible clade A2 and B str ains.In contr ast, in a comm unity setting, we anticipate the replacement of clinical clade A1 strains by commensal strains, as these are also able to inhibit the growth of clade A1 strains .T his is likely to be spurred by the carriage of AMR genes in A1, which will pose a fitness cost to the carrier in the absence of antibiotic pr essur e.
Ov er all, this study provides insights into the complex interplay of E. faecium clades, which has important implications for clinical and public health str ategies.Firstl y, the implications include the sensible use of antimicrobials to minimize the selective pressure driving the ov er gr owth and spr ead of clade A1 strains as well as the need for antimicrobial stew ar dship programs.Secondly, the predicted displacement of clinical strains by commensal strains in the community underscores the potential benefits of promoting commensal strains to prevent the overgrowth of clinical strains, which could include interventions promoting commensal colonization, such as probiotics or faecal microbiota transplantation.

Figure 1 .
Figure 1.Phylogenetic relation and characteristics of strains used in this study.The strains are sorted according to their phylogenetic relationships as indicated by the phylogenetic tree on the left.Clades A1, A2, B ( E. lactis ) are indicated in front of the strain name.Characteristics (source, country of origin (NOR: Norway, NL: The Netherland), isolation site, resistance to vancomycin, ampicillin, and gentamicin (blue: presence, white: absence), and year of isolation) are color-coded.STs are also color-coded as indicated on the rainbow scale to the right.

Figure 2 .Figure 3 .
Figure 2. Inhibitory properties of the clades.(A) Clade A1 used as target lawn (x-axis) for investigating inhibition by clade A2 and B ( E. lactis ) strains (y-axis).(B) Clade A2 and B ( E. lactis ) strains used as target lawn (y-axis) for investigating inhibition by A1 strains (x-axis).Interaction is rated as indicated on the right.

Figure 4 .
Figure 4. Inhibition mediated by supernatant.(A) Clade A1 strains were used as target lawns (x-axis) and supernatants of clade A2 and B ( E. lactis ) str ains (y-axis) wer e placed on top to investigate their ability to inhibit the target.(B) Clade A2 and B ( E. lactis ) str ains (x-axis) wer e used as target lawns and supernatants of clade A1 strains (y-axis) were placed on top.Supernatants were 5x concentrated, or 5x concentrated and exposed to heat or 5x concentrated and treated with proteinase K. Inhibition is rated as indicated.

Figure 5 .
Figure 5. Presence of bacteriocins in genomes of clade A1, A2 and B ( E. lactis ) strains.(A) Bacteriocins in clade A1, (B) Bacteriocins in clade A2 and B ( E. lactis ).Bacteriocin presence was predicted using a novel database.Bacteriocin classes are indicated on the right side of the panel.Total number of bacteriocins is given below, where dark green is the highest number of different bacteriocins (n = 9).The inhibition score from the initial spot on lawn assay is shown at the bottom of the gr a ph, wher e dark blue is the highest av er a ge inhibition score.STs are indicated for clade A1 strains below the panel.
Futur e r esearc h should continue to explor e these inter actions and the potential interventions our study has suggested.Medical Researc h Pr ogr amme pr oject number HNF1474-19 and internal funding from the Institute for Medical Biology at UiT. Carolin K ornelia F enzel and Daniel Engi were supported as master's students by internal funding from the Institute for Medical Biology at UiT. Ana C. Almeida-Santos was supported by a PhD fellowship from Fundação para a Ciência e a Tecnologia (FCT) (UI/BD/151 317/2021) granted by FCT/MCTES through national funds.Part of this work was supported by the Portuguese funds from FCT-I.P., in the scope of project UIDP/04 378/2020 and UIDB/04 378/2020 of the Research Unit on Applied Molecular Biosciences-UCIBIO, project LA/P/0140/2020 of the Associate Laboratory Institute for Health and Bioeconomy-i4HB, and the explor atory pr oject EXPL/SAU-INF/0261/2021.