C. elegans strains SS104 [glp-4(bn2)], N2 wild type, and pmk-1(km25) were maintained on standard nematode growth medium (NGM) (Stiernagle, 2006) seeded with Escherichia coli strain OP50 as a food source at 15°C, unless otherwise noted (Stiernagle, 2006). C. elegans – E. faecium experiments were performed at 25°C to induce sterility of temperature-sensitive bn2 allele, similarly to other glp-4(bn2)-based assays (Moy et al., 2009; Kirienko et al., 2014; Kim et al., 2020). For experiment with N2 and pmk-1(km25), sterility was induced by feeding worms with cdc-25.1(RNAi)-expressing E. coli. Absence of progeny facilitates analysis of assay outcomes.

E. faecium strains used in this study were collected over the past 30 years. The goal of our study was to compare the virulence of E. faecium belonging to different clades in C. elegans model. Therefore the strains representing clinical strains (mostly clade A1), commensal strains from healthy volunteers (mostly clade B), and isolates from animal sources or published clade A2 strains were selected. These isolates included 75 health care-derived isolates (mostly clade A1) from USA and Colombia (Arias et al., 2009; Sillanpää et al., 2009; Tran et al., 2013; Diaz et al., 2014), 21 human commensal strains (mostly clade B) (Sillanpää et al., 2009), and 21 isolates derived from animals, animal feed or obtained from published clade A2 strains studies (Sillanpää et al., 2009; Lebreton et al., 2013). Source information for these isolates is provided in Table S1 . Genotype of strains with deletions is in Table 1 below.

Gene deletion mutants of previously described cell wall anchored (CWA) surface proteins with Ig-like folds (Sillanpää et al., 2008) were given fms designations and were generated in two clinical E. faecium strains (TX0082 and TX2154) using previously published methods (Panesso et al., 2011; Somarajan et al., 2014). Deletions included fms18 (ecbA of E. faecium microbial surface component recognizing adhesive matrix molecules (MSCRAMM) (Sillanpää et al., 2008; Hendrickx et al., 2009); fms11 (predicted pilus-associated) (Sillanpää et al., 2008) and scm (second collagen adhesin of E. faecium) (Sillanpää et al., 2008). Two substitutions (Y176A and F192A) were introduced in acm, a previously described MSCRAMM gene (Nallapareddy et al., 2003; Nallapareddy et al., 2008) using published methods (Liu et al., 2007). These strains were designated as TX6111 (E1162 acm ^Y176A,F192A Δfms18 Δscm), TX6114 (E1162 Δfms15), TX6109 (E1162 Δscm), TX6110 (E1162 acm ^Y176A,F192A Δscm), TX6094 (TX0082 acm ^Y176A,F192A Δfms11 Δscm), and TX 6097 (TX0082 Δfms11 Δscm) and have been listed in strain list in Table 1 .

For the high-throughput, liquid-based pathogenesis assay, 25 synchronized young adult worms were sorted into 384-well plate. E. faecium strains were grown in deep-well plates at 700 rpm for 16 h. S Basal medium was mixed with brain-heart infusion (BHI) medium (10%) and Enterococcus faecalis or E. faecium (final OD600: 0.03) were then added into each well. Plates were incubated at 25°C until assay completion. At various time points, bacterial OD was measured with a Cytation5 (BioTek Instruments) plate reader and then plates were washed three times. Sytox Orange nucleic acid stain was added into each well and incubated for 12 - 16 h to stain dead worms. Plates were then washed three times, worms were transferred into a new 384-well plate, washed three more times, and imaged with Cytation5 automated microscope. Dead worms were quantified with CellProfiler software, similar to previously established pipelines (Conery et al., 2014).

For treatment with catalase, catalase (7.5 mg/mL) was supplemented in S Basal, BHI, Enterococcus mixture that was then distributed into each well of the 384-well plates.

For killing on agar, 50 young adult worms were transferred onto Enterococcus-BHI plates and incubated at 25°C. Worms were scored every day for survival; dead worms were removed from assay plates.

E. faecium strains, TX0016 (clade A) and a commensal isolate, TX1330 (clade B) were tested following our previously published method (Singh et al., 1998). In brief, mice were injected intraperitoneally with appropriate dilutions of bacteria grown in BHI broth, premixed with sterile rat fecal extract (SRFE) and were observed for 5 days for survival. Data (h) were used to compare animal survival/mortality using six mice per group. Comparison of the survival curves was performed using a log-rank test with GraphPad Prism 4 for Windows^®. A p < 0.05 was considered significant. All experiments were approved by the Animal Welfare committee, University of Texas Health Science Center at Houston.

75 worms from liquid-based assay were collected into microcentrifuge tubes. Worms were washed three times to remove residual bacteria. After the final wash excess media was aspirated down to 100 µL. Equal volume of carbide beads was then added into the tubes and vortexed vigorously for 1 min to break worms. Supernatant was serially diluted and plated on BHI plates.

Genomic analysis was performed in the usegalaxy.org platform. FASTQ files generated from NGS were submitted for SNPs analysis to snippy (Galaxy version 4.5.0), a bioinformatic tools for rapid bacterial variant calling. Meanwhile, FASTA files were processed with prokka (Galaxy version 1.14.5) for genome annotation, roary (Galaxy version 3.13.0) for genomic alignment, and RAxML (Galaxy version 1.0.0) to construct a maximum-likelihood phylogenetic tree. Figure 8A was generated with R packages treeio (version 1.14.3) and ggtree (version 2.4.1).

Upon 48 h of infection with E. faecium E007 in the liquid-based assay, worms were washed three times to remove residual bacteria, incubated for one hour with or without 20 µg/mL gentamicin, washed three times, and stained with 40 µM acridine orange for 4 hours. Worms were then washed three more times to remove unbound dye. Worms were imaged using a Zeiss ApoTome.2 Imager.M2 fluorescent microscope (Carl Zeiss, Germany) with a 20x objective magnification.

RNAi-expressing E. coli HT115 were cultured and seeded onto NGM plates supplemented with 25 μg/mL carbenicillin and 1 mM IPTG. 8,000 synchronized L1 larvae were transferred onto RNAi plates and grown at 25°C for 64 hours prior to exposure to pathogens.

RNAi experiments in this study were conducted by using RNAi-competent HT115 obtained from the Ahringer RNAi library (Kamath et al., 2003) and were sequenced prior to use.

With the exception of the mouse infection model, all experiments were performed in at least three independent biological replicates. The number of worms used per biological replicates is listed in the corresponding legends.

Student’s t-test analysis was performed to calculate the p-values when comparing two groups in an experimental setting. p-values were indicated in graphs as follows: NS not significant, *p < 0.05, **p < 0.01, and ***p < 0.001.

Figure 2 was generated with R package ggplot2 (version 3.3.3). Correlation coefficient and p-value between C. elegans survival and E. faecium OD600 were calculated and indicated on the figure.

Figure 5 was generated with R package ggplot2 (version 3.3.3). K-means clustering was performed with the ‘kmeans’ function in R. p-values between Clade A1 and Clade B were indicated on graphs.

To study E. faecium virulence determinants, we developed the first high-throughput infection model for E. faecium that uses C. elegans as a host. In brief (see Figure 1A ), synchronized, young adult C. elegans were sorted into 384-well plates. E. faecium or E. faecalis grown on BHI agar plates were then diluted into 10% BHI medium and added to worms at predetermined concentrations. Worms were infected at 25°C until assay completion and then were washed several times before staining with a cell-impermeant stain to quantify death. The entire assay can be performed within ~ 1 week, and ~90 different strains can be tested in parallel on a single 384-well plate. Unlike in previously published agar-based systems where only E. faecalis killed the worms (Garsin et al., 2001), our liquid assay showed killing within a week by both E. faecalis and E. faecium ( Figures 1B, C ). Importantly, we saw no significant correlation between bacterial growth (as determined by OD) and host death either when we used a pilot set of ~30 isolates or with the full set of ~120 isolates ( Figure 2 ). This suggests that host death is a specific phenomenon related to an interaction between the host and the pathogen, and does not merely result from bacterial growth.

As noted above, E. faecalis-mediated killing of C. elegans in the previously described assay depended on hydrogen peroxide production (Moy et al., 2004). Although our assay does not involve anaerobic growth (which was required for hydrogen peroxide production), we tested whether the addition of catalase to our assay would influence rates of host death. In general, we saw no decrease in killing after the addition of catalase, consistent with catalase-independent pathogenesis ( Figure S1 ).

The PMK-1/p38 MAPK pathway is arguably the most important innate immune pathway for responding to bacterial infection in C. elegans and has been implicated in resistance in most agar-based pathogenesis assays (Kim et al., 2002; Bolz et al., 2010; Shivers et al., 2010; Pukkila-Worley et al., 2012). In addition, pmk-1 mutation has been shown to increase sensitivity to E. faecium (Yuen and Ausubel, 2018). In contrast, the PMK-1/MAPK pathway is detrimental for survival in a liquid-based C. elegans-P. aeruginosa pathogenesis assay (Kirienko et al., 2015; Tjahjono and Kirienko, 2017). We used RNAi to knock down pmk-1 and then tested young adult worms in the liquid E. faecium assay for increased host death. As with P. aeruginosa, loss of PMK-1 function did not increase E. faecium-mediated killing in the liquid-based assay ( Figures 3A, B ). In contrast, disrupting PMK-1 function in the agar-based assay reduced survival ( Figure 3C , S2A ). These results were confirmed using a known loss-of-function mutation, pmk-1(km25) ( Figure S2B ).

Recently, we showed that the Mediator subunit MDT-15/MED15 has a role in host resistance against E. faecalis (Hummell et al., 2021). To probe whether this function is retained in this liquid-based assay, we also used RNAi to knock down mdt-15 expression. Unlike pmk-1(RNAi), mdt-15(RNAi) significantly increased host mortality in both liquid- and agar-based assays ( Figure 3 , S2A ), suggesting that MDT-15 serves an innate immune function in these assays as well.

Genome analysis of E. faecium strains indicates that they belong to at least two different phylogenetic clades: A (comprised of hospital-acquired strains and strains mostly derived from animal origins), and B (mostly human commensal strains) (Palmer et al., 2012; Galloway-Peña et al., 2012; Lebreton et al., 2013). Although earlier phylogenetic evidence suggests that clade A may be broken into two subgroups (Palmer et al., 2012; Galloway-Peña et al., 2012; Lebreton et al., 2013), a more recent, larger analysis called this conclusion into question (Raven et al., 2016). Pathogenicity of isolates from clade A1 ( Figure 4A ), clade A2 ( Figure 4B ), and clade B ( Figure 4C ) varied, ranging from 45-70% host mortality. Although we tested a larger number of strains from clade A1, all three clades showed similar patterns of variability in virulence. We determined median mortality for each clade and identified strains that were significantly more or less pathogenic for further analysis.

An initial analysis of the pathogenicity of ~30 isolates in our liquid-based model was performed. K-means clustering was then used to group the strains based on virulence, resulting in strong preferential grouping of A1 and A2 clades in high virulence and B strains in low virulence ( Figure 5A ). However, a more robust analysis, using a total of ~100 strains, abolished this difference, with each clade being approximately equally represented in the high- and low-virulence groups ( Figure 5B ).

Colonization is the first step in any infection, including bacterial infections of C. elegans with E. faecalis or E. faecium (Tan et al., 1999; Aballay et al., 2000; Garsin et al., 2001; Kurz et al., 2003; Goh et al., 2017). In contrast, our previously reported liquid-based P. aeruginosa assay, which is based on intoxication rather than infection, showed relatively few bacteria in the intestine, more consistent with transit of undigested bacteria, rather than intestinal colonization (Kirienko et al., 2013). However, in liquid-based C. elegans – Candida albicans model, host colonization was observed (Rosiana et al., 2021). To test whether colonization is related to virulence in C. elegans- E. faecium liquid assay, we selected six strains from Clade A1, including the three with the highest (TX2029, TX2051, and TX2416) and the lowest (S447, TX2022, and TX2025) virulence, and infected C. elegans with them in the liquid assay ( Figure 6A ). After 48 hours of infection, some of the worms were harvested, washed thoroughly to remove extracorporeal bacteria, and then physically disrupted to release intestinal bacteria. Lysates were serially diluted and plated on BHI media to determine bacterial titer via CFU counting, allowing colonization to be directly measured. We found that the high-virulence strains as a group had a better ability to colonize C. elegans than low-virulence strains ( Figure 6B and Figure S3 , p < 0.001, pooled low- and high- virulence groups). To ensure that only bacteria from the worm alimentary track, but not those that adhere to the cuticle, contributed to CFU counts, we performed the same assay except that worms were treated with gentamicin for 1 h after the first three washes ( Figure S4 ). Worms were then incubated with 40 µM acridine orange, a dye commonly used to stain bacteria (Neeraja et al., 2017). Fluorescent imaging confirmed that bacteria were only detected in the worms’ gastrointestinal tract, both for gentamicin-treated and untreated sub-populations, but not on the cuticles ( Figures S4A, B ). These results suggested that the ability to persist in C. elegans gut may be an important contributor to E. faecium virulence in this liquid assay.

In a previously-published liquid-based C. elegans pathosystem, we showed that toxin secretion, particularly of the siderophore pyoverdine, was very important for host killing by P. aeruginosa (Kirienko et al., 2015; Kang et al., 2018). In this case, spent media or purified pyoverdine were capable of killing worms even in the absence of live bacteria ( Figure S5 ). To test the ability of secreted factors to affect host survival during E. faecium infection, the same panel of six E. faecium isolates was grown under assay conditions, but in the absence of worms, for four days. Then bacteria-free filtrates were produced and C. elegans were exposed to them for ~100 h ( Figure 6C ). Virtually no host death was observed, suggesting that live bacteria and host-pathogen interactions are the driving force of pathogenesis in this assay.

E. faecium infection is a multifactorial process that includes attachment, colonization, and biofilm formation. To determine which virulence factors are important for killing in liquid, we measured the pathogenicity of available E. faecium strains harboring deletions of targeted virulence factors ( Table 1 ), including genes from the gls (Choudhury et al., 2011), fms (Sillanpää et al., 2008; Sillanpää et al., 2009; Sillanpää et al., 2010), wxl (Galloway-Peña et al., 2015), and ccpA (Somarajan et al., 2014) genes and gene families.

Deletion of genes encoding general stress proteins (Gls) gls33-glsB alone or doubled with gls20-glsB1 did not appear to affect host survival ( Figure S6A ). This also seemed to be the case with mutants harboring deletions of genes in the WxL loci (Galloway-Peña et al., 2015). The WxL loci, are a series of three loci that each encode between 3 and 6 genes, including peptidoglycan-binding proteins (Brinster et al., 2007). Three strains with a single mutation each and one strain with triple mutations of the WxL loci did not significantly alter worms’ survival as compared to the parental strain TX0082 ( Figure S6B ).

On the other hand, the E. faecium surface proteins (Fms) of the MSCRAMM (microbial surface component recognizing adhesive matrix molecules) family were more likely to be involved in pathogenesis in this E. faecium – C. elegans killing assay. fms genes are tightly associated with E. faecium clinical isolates (Sillanpää et al., 2008). We tested multiple mutants with one (scm, acm, or fms15), two (Δfms21-20, acm ^Y176A,F192A Δscm, and Δfms11 Δscm), or three genes mutated (acm ^Y176A,F192A Δfms18 Δscm, ΔebpABCfm, and acm ^Y176A,F192A Δfms11 Δscm). Among the strains tested, the single deletion of fms15, an adhesin gene, significantly reduced E. faecium virulence ( Figure 7A ). Moreover, triple mutation of acm ^Y176A,F192A Δfms11 Δscm and deletion of pilus subunit proteins encoded by the three-gene locus ebpABCfm (fms1-5-9) reduced pathogenicity as well ( Figure 7B ). As the MSCRAMMs play a role in host-pathogen adherence, these results corroborated our finding that colonization and persistence in C. elegans gut is important for virulence.

Interestingly, deletion of ccpA, a regulator of carbon catabolite repression that affects E. faecium growth and is important for virulence (Somarajan et al., 2014), increased pathogenicity ( Figure 7B ). This result was unexpected as ccpA deletion mutant showed reduced biofilm formation, growth, and attenuated infection in endocarditis model (Somarajan et al., 2014).

To further understand E. faecium virulence mechanisms, we obtained whole genome sequencing data for eight strains in the higher virulence and six strains in the lower virulence groups. Strains from Clade A1 were selected for this analysis as they are most commonly associated with human infection. Phylogenetic tree construction showed close relationships among the fourteen strains, except for TX2046 (lower virulence) and TX2051 (higher virulence) ( Figure 8A ).

Genomic analysis showed that TX2046 contained two frameshift mutations: a one nucleotide insertion in sorB (a PTS system sorbose-specific EIIB component) and another in mazE (the gene encoding the MazE Type II antitoxin). TX2051 acquired a frameshift in araQ (L-arabinose transport system permease protein) and a stop codon in wecA (UDP-N-acetylgalactosamine-undecaprenyl-phosphate N-acetylgalactosamine phosphotransferase).

Toxin-antitoxin pairs are commonly found in bacterial pathogens, and mutation of the antitoxin mazE would result in greatly diminished bacterial survival. We performed growth analysis of the OD 600 for the two strains TX2046 and TX2051. We observed differences in growth rates, with TX2051 growing faster than TX2046 ( Figure 8B ). Thus, TX2051 might be more virulent due to faster growth and colonization rate. This result also showed that, at least in selected cases, pathogenicity may depend on bacterial growth. However, these two parameters did not correlate in a larger group ( Figure 8B and Figure S3 ).

Most importantly, our C. elegans – E. faecium pathogenesis model recapitulated virulence in a murine peritonitis model. Two strains that differ in origin, TX0016 [or DO, an endocarditis isolate with one of the first E. faecium genomes sequenced (Qin et al., 2012)] and TX1330 [a community-derived isolate (Qin et al., 2012)] were used to infect C. elegans or mice. Pathogenesis was noticeably reduced in TX0016 strain compared to TX1330 in both C. elegans ( Figure 9A ) and mice ( Figure 9B ).