The reference strain used in this study is E. faecalis V19, which corresponds to a plasmid-cured strain derived from the V583 strain of clinical origin (Paulsen et al., 2003). Overnight cultures were achieved in M17 medium supplemented with 0.5% glucose (GM17). Escherichia coli strains TOP10 (ThermoFisher, Waltham, MA, United States), NEB-5α (New England BioLabs, Ipswich, MA, United States), and M15 pRep4 (Qiagen, Hilden, Germany) were used for RNA in vitro production, mutant constructions, and recombinant protein synthesis, respectively. Media were supplemented with chloramphenicol (Cm 10 μg/ml), kanamycin (Kan 50 μg/ml), or ampicillin (Amp 100 μg/ml) when needed (Supplementary Table 1).

Primers used in this study are listed in Supplementary Table 2. All molecular biology techniques were performed following the manufacturer’s recommendations. Q5® High-Fidelity DNA Polymerase (New England BioLabs) and GoTaq DNA Polymerase (Promega, Madison, Wisconsin, USA) were used for the PCR reactions. 5′RACE experiments were performed with the 5′/3′ RACE kit (Sigma-Aldrich, Saint-Louis, Missouri, United States), using SP1 or R (retrotranscription), SP2 (PCR with anchor primers), and SP3 (sequencing) primers and poly-G oligonucleotides. In the case of uncertainty, poly-A tailing was also used. Purifications of PCR products were performed with NucleoSpin Gel and PCR Clean-up kit (Macherey-Nagel, Düren, Germany), and plasmid extractions were achieved using NucleoSpin Plasmid kit (Macherey-Nagel). Digestions were generated using restriction enzymes (Promega, Madison, Wisconsin, United States) and ligations with T4 DNA Ligase (New England BioLabs).

All mutant strains (ΔnagY, ∆5’nagY, ΔhylA) were constructed in E. faecalis V19 (Supplementary Table 1) using E. coli DH-5α as intermediate cloning host. pLT06 vector (Thurlow et al., 2009) was amplified with pLT06_1_bis and pLT06_2 primers, and flanking regions of the region to be deleted were amplified by PCR using oligonucleotides 1 and 2 for the upstream fragment, and oligonucleotides 3 and 4 for downstream fragment (Supplementary Table 2). Primers 1 and 4 have overlapping tails compatible with pLT06_1_bis and pLT06_2 primers, and primers 2 and 3 have overlapping tails compatible with each other. Cloning was performed using the in vivo recombination method (Huang et al., 2017). Deletion was obtained by double crossing over, as previously described (Thurlow et al., 2009), and was checked by PCR using primers 5 and 6 (Supplementary Table 2). Gene deletion and absence of point mutations susceptible to change phenotypes of deleted strains were checked by whole genome sequencing as described in Supplementary Material and Methods, and a summary of the variant detection analysis is listed in Supplementary Table 3.

Cultures of 10 ml were performed at 37°C in carbon depleted medium cdM17 (La Carbona et al., 2007) supplemented with the appropriate sugar to OD₆₀₀ 0.5. Cells were pelleted and lysed using a FastPrep device (MP Biomedicals, Illkirch Graffenstaden, France). RNA extraction and purification were achieved with TRIzol Reagent (ThermoFisher) and chloroform/isoamyl alcohol separation before using Direct-Zol RNA Miniprep kit following the manufacturer’s recommendations (Zymo-Research, Irvine, Californie, United States). RNAs were quantified using Nanodrop™ 2000 (ThermoFisher) and their quality was checked by electrophoresis.

Reverse Transcription for RT-PCR and RT-qPCR assays were performed using QuantiTect Reverse Transcription kit (Qiagen) with L/R oligonucleotides and random primers, respectively (Supplementary Table 2). The GoTaq qPCR Master Mix (Promega) was used for qPCR, as well as the C1000™ Thermal Cycler (Bio-Rad) apparatus, using the following conditions: 95°C for 3 min, and 40 cycles at 95°C for 15 sec, 60°C for 1 min. Normalization was performed using the gyrA reference gene. Standard curves of each gene and qPCR efficiency were obtained using genomic DNA of the E. faecalis V19 strain.

RNA synthesis was achieved with the pTOPO plasmid and in vitro synthesis. DNA regions of interest were amplified by PCR using primers containing tails overlapping with pTOPO: topo85_FP1 and topo85_RP1 for 5’nagY, topo5’3023_FP1/topo5’3023_RP1 for 5’hylA, and topo65_FP1 and topo65_RP1 for SRC65 (Supplementary Table 2). Plasmid was amplified with topo_FP2 and topo_RP2 primers. PCR products were inserted into the plasmid with NEBuilder HiFi DNA Assembly Cloning Kit (New England BioLabs), and products used to transform E. coli TOP10. RNA production was performed on the resulting plasmid, linearized with the SpeI restriction endonuclease, using MAXIScript™ T7 in vitro Transcription Kit (Invitrogen, Carlsbad, California, United States). Unincorporated nucleotides were eliminated by ammonium acetate/ethanol precipitation, as recommended on the kit protocol, and RNA were quantified with Nanodrop™ 2000 (ThermoFisher).

The nagY gene was amplified with primers ef1515_pQE70_SphI and ef1515_pQE70_BglII carrying restriction sites (Supplementary Table 2). pQE70 plasmid and PCR product were digested with SphI and BglII enzymes, ligated together, and used to transform E. coli M15 pRep4. The bacteria obtained were grown at 37°C with agitation in Terrific Broth medium supplemented with kanamycin and ampicillin until OD₆₀₀ 0.5. Transcription induction was triggered with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) during 4 h at 37°C under agitation. NagY protein was purified with Protino Ni-NTA Agarose kit according to the manufacturer’s instructions (Macherey-Nagel), and desalted with PD10 columns (GE Healthcare, Chicago, Illinois, United States). Protein concentration was determined with the BCA test (ThermoFisher Pierce), and its purity checked on 12.5% SDS-PAGE and by mass spectrometry.

Recombinant NagY protein was labeled using the His-Tag Labeling Kit RED-tris-NTA 2^nd generation Monolith (Nanotemper Technologies, München, Germany) following the manufacturer’s recommendations, and diluted in ES-Buffer (10 mM Tris pH8.0, 40 mM NaCl, 10 mM KCl, 1 mM MgCl₂, 0.05% Tween-80) at a final concentration of 25 nM. Before the assay, RNAs were heated for 5 min at 70°C and slowly cooled down at room temperature to allow a proper formation of secondary structures. A series of 1:1 dilution of RNAs were prepared in order to obtain a RNAs concentration ranged from 0.015 nM to 492 nM (16 points). Then each tube prepared with this 16 RNAs different concentration is mixed with the labeled protein (1:1), and filled into capillaries and introduced into the Monolith NT.115^Pico instrument (NanoTemper Technologies). Data of at least three independently pipetted measurements were analyzed (MO.Affinity Analysis software version 1.5.41, NanoTemper Technologies). The data were fitted using the law of mass action from GraphPad Prism version 5, and MicroScale thermophoresis (MST) figures were generated using MO.Affinity Analysis.

Overnight cultures in GM17 of E. faecalis were adjusted to OD₆₀₀ of 0.2 in fresh M17 supplemented with 2% glucose. One hundred microliters of the bacterial suspension were inoculated into CytoOne polystyrene microwells plate coated with 1 μg/ml of hyaluronic acid (Sigma-Aldrich), chondroitin sulfate (Carl Roth, Karlsruhe, Germany), or heparin sodium (ThermoFisher). After 24 or 48 h of incubation at 37°C, the plates were washed with 0.9% NaCl to remove unbound bacteria. Each well was then stained with 0.1% (wt/vol) crystal violet for 15 min at room temperature. Wells were then rinsed two times with 0.9% NaCl. Adherent cells were dissolved in 30% acetic acid, and the OD₅₅₀ was measured using a microplate reader Nano Tecan (Life Science). At least three experiments were performed for each condition.

Detection of GAGs-degrading activity were performed as previously described (Kawai et al., 2018), with the following modifications. Overnight cultured cells were centrifuged at 4,500 rpm for 10 min, washed with 1 ml of 0.9% NaCl or GM17, and resuspended in the saline or GM17. The volume (X μl) of the saline or GM17 was calculated by the following formula: X = 200 × OD₆₀₀. Cell suspension was then spotted at the center of the GAG minimal plate [0.2% GAG (hyaluronic acid, chondroitin sulfate, or heparin sodium), 0.1% yeast extract, 0.1% Na₂HPO₄, 0.1% KH₂PO₄, 0.01% MgSO₄·7H₂O, and 1.5% agar] with BSA at 1% and cultured at 37°C for 7 days. After cell growth, 1 ml of 2 M acetic acid was added to form a complex with the remaining GAG and BSA as white precipitates.

Infection assays were performed on Galleria mellonella larva as previously described (Benachour et al., 2012). Bacteria were inoculated with a dose of 2 × 10⁶ CFU to 15 caterpillars per strain for each experiment. At least three experiments were performed for each condition. G. mellonella survival was followed from 16 h post-infection and during 24 h.

The ef1515 gene in E. faecalis V583 is annotated as encoding a BglG/SacY family antiterminator, and we renamed it nagY in reference to sacY, and its following gene nagE (Mao et al., 2009). The EF1515 (EF_RS07320 in the new nomenclature) protein presents 36% identity with SacY from B. subtilis, and 31% amino acid identity with BglG from E. coli (blastP alignment). This gene is followed by the ef1516 (EF_RS07325) gene (Figure 2), recently renamed nagE on V583 genome, encoding a NAG specific EIICBA PTS transporter (Keffeler et al., 2021b). Thus, the NagY and NagE proteins belong to multigene families. In order to unambiguously identify the orthologous genes encoding these proteins in the Enterococcaceae, we performed a phylogenomic analysis of these families.

The PEPPAN pipeline (Zhou et al., 2020) has been used to analyze the pan-genomes of the 81 Enterococcaceae reference genomes. E. faecalis V583 nagY and nagE belong to two OG clusters composed of 47 and 54 members. Both genes appear to be present in the majority of Enterococcaceae studied (Figure 3). They are absent in the genera Enterococcus_J, _H, _E, and _G and weakly represented in the genus Enterococcus_D and in Tetragenococcus. Note that the outgroup position of Enterococcus_J and Enterococcus_H suggests that these genes were absent in the last common ancestor (LCA) of Enterococcaceae. Two nagE paralogs are present in E. thailandicus DSM 21767 and a tblastn search, with the candidate genes as query, identifies two nagE-like sequences in two genomes (E. sp. 9E7_DIV0242 and E. sp. AS17jrsBPGB_10) that were not annotated by Prokka (Figure 3). Chromosomal neighborhood analysis of the nagY genes reveals, in all cases, the presence of a downstream nagE gene. However, in eight genomes, a nagE gene is present without its nagY partner. Five genomes belong to a subtree composed of E. sp. 10A9_DIV0425 and four E. mundtti genomes (Figure 3) suggesting that the nagY gene may have been lost in their LCA. One of the two E. thailandicus DSM 21767 paralogs does not have a nagY gene in its chromosomal neighborhood. Note that nagE genes encode proteins with the EIICBA architecture as PtsG of B. subtilis (and B. cereus) while BglF from E. coli has EIIBCA domain order.

In addition to gene losses, GeneRax predicted that horizontal gene transfers (HGT) may have occurred, with a higher frequency for nagE (17 HGTs for 54 genes) than for nagY (7 HGTs for 47 genes; Supplementary Figures 2A,B). Some HGTs may have occurred between genomes belonging to different genera, as revealed by the splitting of these genera on the protein trees. It should be noted that no conservation of gene neighborhoods was found in the genomes studied. Only a majority of strains of the genus Enterococcus present a conserved genetic context, with the notable exception of the E. faecalis V583 strain (Supplementary Figure 2).

To determine the extent to which the orthologous gene pair nagY-nagE is present in E. faecalis strains, we searched the 1949 genomes for orthologs of both gene products. The NagY and NagE proteins were identified in 99.84 and 99.59% of the genomes, respectively, and both are present in 99.38%. Missing genes are due to incomplete genome assemblies.

We experimentally characterized the nagY-nagE operon (represented in Figure 2) by confirming the co-transcription of these two genes by RT-PCR (Supplementary Figure 3). The induction of the nagY-nagE expression in presence of NAG was then revealed by RT-qPCR targeting nagE (Figure 4). This gene is overexpressed in WT strain in presence of NAG as sole carbon source, with a fold-change (FC) of 16.6-fold (±2.15) compared to glucose conditions (p < 0.001), and this induction is almost completely abolished in the ΔnagY strain (FC = 3.1 ± 1.55, p < 0.001).

To investigate the role of NagY in its own operon regulation, we firstly identified the transcription start site by 5′RACE-PCR assay. Thus, consistent with previous studies (Innocenti et al., 2015; Muller et al., 2015; Michaux et al., 2020), the starting base of the RNA was confirmed to be located 172 pb before the predicted initiation codon (Figure 2; Supplementary Figure 4A), showing the existence of a long 5′UTR (named 5′nagY). The search of RAT sequence based on the consensus defined in preceding work in E. coli and B. subtilis (Aymerich and Steinmetz, 1992; Tortosa et al., 2001; Gordon et al., 2015) allowed the identification in the 5′nagY of an imperfect inverted repeat with a low sequence conservation with the consensus sequence (Figure 2). A search performed with MEME (Bailey and Elkan, 1994) in the upstream regions of nagY orthologs reveals the presence of two conserved motifs. The first motif covers the region overlapping the putative RAT sequence identified above in E. faecalis V583. The second motif, located downstream of the first motif, is less well conserved but is characterized by a terminal T-rich region (Supplementary Figure 5). These motifs are conserved with their relative positions upstream of the 46 nagY sequences of Enterococcaceae, with the exception of E. saccharolyticus (Figure 3). Alignment of the first motif with rLocARNA (Will et al., 2012) reveals a conserved stem loop with a two-nucleotide bulge (Supplementary Figure 5A). The stem bases have undergone a large number of compensatory mutations to maintain this secondary structure. The region bounded by the two conserved MEME motifs was extracted from the 5’nagY for the different genomes. The rLocARNA predicts, in all sequences, the presence of a large stem loop of variable length that ends in a T/U-rich region, a structure typical of an independent rho terminator (Supplementary Figure 5B). The foot of this structure overlaps the end of the RAT region (common consensus sequence: GCRUGGA). These two structures are therefore mutually exclusive (Supplementary Figure 5C). This competition between the two structures is typical of what has been observed for BglG/SacY antiterminators. A covariance model was constructed with the alignment obtained with rLocARNA and was used to identify the RAT-terminator motif with a high specificity in genomic sequences (Figure 3).

To confirm that the nagY-nagE transcription-antitermination mechanism is similar to the SacY model in B. subtilis, the direct interaction between NagY and 5’nagY was studied by MST. The purified protein was incubated in presence of the in vitro produced 5’nagY RNA. As shown in Figure 5, we observed a dose–response binding of NagY on 5’nagY, with a Kd of 4.18 nM (±0.42 nM). The SRC65 sRNA (Shioya et al., 2011; Salze et al., 2020b) was used as negative control and did not show any interaction with NagY. These results indicate that NagY has a high affinity for its 5′UTR and regulates both NagE and its own expression by its binding on RNA 5’nagY. Consequently, binding of NagY to RATs can lead to the opening of the hairpin base and therefore make the transcription terminator ineffective.

To confirm that these sequences are involved in the nagY-nagE NAG-dependent induction, a deletion of 5’nagY region overlapping both structures was constructed in the WT strain (Figure 2), and the expression of nagE in this mutant was determined by RT-qPCR (Figure 4). No NAG-dependent induction was observed when we compared NAG to glucose culture conditions, but nagE is deregulated whatever culture conditions are, with FC of 33.2 and 34.8 in presence of glucose and NAG, respectively (Supplementary Figure 6). Considering the role of NagE in the NAG transport, the operon could potentially be regulated by catabolic repression. The expression of the operon was also followed in the presence of glucose and NAG, but no difference of expression was observed compared to the condition with NAG only (Supplementary Figure 6). These results establish that the nagY-nagE operon expression is not under the control of the catabolic repression.

As regulator binding on nucleic acid, NagY could potentially regulate other genes expression by recognizing a conserved sequence. To identify potential NagY targets, we searched for RAT/terminator motif with cmsearch software (Infernal package; Nawrocki and Eddy, 2013) in the E. faecalis V583 genome. We obtained two hits, the highest upstream of the nagY gene, and the second upstream of the ef3023 (ef_rs14340) monocositronic gene, named hylA, which shares 88% identity with the nagY RAT-like sequence. This suggests that hylA possesses a 5’UTR (named 5’hylA) on which NagY could potentially bind to regulate the expression of this gene. We confirmed the existence of a 5’hylA of 193 pb long by 5’RACE-PCR (Supplementary Figure 4B), and the interaction between NagY and this RNA was studied by MST (Figure 5). This assay highlighted a binding of the protein on 5’hylA RNA, with a lower affinity than with 5’nagY (7.46 ± 0.63 nM). Moreover, RT-qPCR assays showed that the expression of hylA depends on the presence of the nagY gene (Figure 4). These results are compliant with the hypothesis that NagY also regulates the expression of hylA, suggesting that nagY, nagE, and hylA belong to the same regulon and potentially the same carbohydrate consumption pathway.

HylA was identified as a cell-wall anchored protein, annotated as a polysaccharide lyase 8 [Lyase_8_N (PF08124), Lyase_8 (PF02278), and Lyase_8_C (PF02884.17) domains, Figure 6]: this group of enzymes targets uronic acid-containing polysaccharides such as some GAGs (hyaluronate, chondroitin, or heparin) that are components of the ECM (Sillanpaa et al., 2004, 2009; Lombard et al., 2014). To our knowledge, the function and substrate of E. faecalis HylA are unknown, although it is annotated as hyaluronidase in KEGG database (Kanehisa et al., 2017) and was identified as a MSCRAMM family member mostly extracytoplasmic (Sillanpaa et al., 2004, 2009). In addition to a signal peptide and the lyase regions, HylA possesses other domains: (i) a F5/8 type C domain (discoidin domain) that is a major domain of many blood coagulation factors (F5_F8_type_C PF00754), (ii) a bacterial Ig-like domain (Big_2 PF02368) found in bacterial cell-adhesion molecule, mediating the intimate bacterial host-cell interaction (Kelly et al., 1999), (iii) FIVAR domains (Found In Various Architectures PF07554) mostly associated to binding domains in cell wall associated proteins, and (iv) a LPXTG cell wall anchor motif (Gram_pos_anchor PF00746) presents in virulence factors which are produced by Gram positive pathogens.

We used the two largest conserved domains of the protein to identify homologous sequences in Enterococcaceae (PF08124, Lyase_8_N and PF02278, Lyase_8) and the simultaneous occurrence of these two domains was found in 27 proteins. These proteins also have the PF02884 Lyase_8_C domain. Three of them are significantly longer and have the same domain architecture as E. faecalis HylA, except for the number of FIVAR domains (Supplementary Figure 2C). The formers are referred to as short Lyase_8 and the latter as long Lyase_8. The SignalP program (Almagro Armenteros et al., 2019) predicts a standard secretory signal peptide in all proteins. Two proteins are very partial (155 AA for T. koreensis KCTC 3924 sequence and 131 AA for E. crotali ETRF sequence). Note that both strains of E. faecalis (V583 and OG1RF) have both types of proteins. Prokka does not annotate a hylA gene in the genome of E. faecalis ATCC 19433A, but a search with tblastn reveals the presence of a partial sequence similar to a gene encoding a long Lyase_8 (Figure 3). The long Lyase_8 protein found in the genomes of E. faecalis V583 and OG1RF strains is also found in the genomes of E. hirae ATCC 9790 (Figure 3).

GeneRax predicted that the hylA sequences of E. hirae and E. faecalis are originated from HGTs, but presumably from genomes that are not sampled in our study (Supplementary Figure 2C). cmsearch of the upstream region of the E. hirae hylA gene reveals the presence of the RAT-containing motif with the T-rich region just downstream, as observed in the 5’nagY sequence. This conservation suggests similar regulation of hylA by NagY in E. hirae and E. faecalis strains. It also has to be noted that RAT sequence in the 5’hylA is conserved only in the presence of the long Lyase_8 (Figure 3).

To better understand the origin of HylA in E. faecalis, we ran a blastP on the NCBI website with the EF3023 sequence. Even though many sequences are from E. hirae, sequences from Lacticaseibacillus genomes and from different species of Listeria, Staphylococcus, and Streptococcus are found. If some of them are partial, others like those of S. agalactiae or L. monocytogenes have a domain organization similar to those of E. faecalis sequences. The high sequence conservation between these distant species suggests recent HGTs.

To determine the extent to which the orthologous gene hylA is present in E. faecalis strains, we searched for occurrences of hylA in the 1949 E. faecalis proteomes. We identified 1,520 occurrences among which 458 sequences are partial with a length of their gene shorter than 3,000 nt (Figure 7). The hylA genes appears to be well distributed in E. faecalis strains, however we can observe its absence in some closely related species forming subtrees on the species tree, suggesting that it has been lost in their LCA. Similarly, we can note that partial sequences are found in genomes closely related on the species tree, which could indicate pseudogenization of these genes for subsets of related genomes. These hylA genes encode HylA protein but with a variable number of FIVAR domain(s) (from 1 to 9, but centered on four copies). The short version of the protein is present in 1,243 E. faecalis proteomes with 26 genes fissions.

To determine HylA functions, the ΔhylA mutant was constructed and characterized. We observed that this mutant was not affected in its growth in the presence of NAG as sole carbon source (Supplementary Material and Methods; Supplementary Figure 7). HylA has putative hyaluronidase domain, and hyaluronic acid is a polymer made up of alternating NAG and glucuronic acid residues linked by glycosidic bonds (Hynes and Walton, 2000), that could make the functional link between hylA and nagY-nagE operon. As shown on Figure 8A, ΔnagY, and ΔhylA cannot degrade this substrate compared to the WT strain, as well as heparin sodium and chondroitin sulfate, confirming that the encoding proteins are involved in the use of these GAGs.

Given the adhesin domains identified in the HylA protein, biofilm formation was assessed using the microtiter plate assay on a coating of hyaluronic acid, chondroitin sulfate, or heparin sodium (Figure 8B; Supplementary Figure 8). While no difference was observed after 24 h, a significant 2.3 and 3.2-fold decrease in crystal violet staining was observed after 48 h for the ΔnagY and ΔhylA mutants on microtiter plates coated with hyaluronic acid, respectively (Figure 8B), and 1.9/1.7 and 1.5/1.6-fold decrease on plates coated with heparin or chondroitin compared to the WT strain (p < 0.0001; Supplementary Figure 8).

To investigate the role of NagY and HylA in virulence in vivo, we monitored G. mellonella larvae survival infected by WT, the two mutant strains and the 5’nagY mutant strain in which NagY is deregulated and overexpressed whatever conditions are (Figure 9; Supplementary Figure 9). The experiment showed that the deletion of ΔnagY does not affect larval mortality (Supplementary Figure 9). However, larvae infected by ΔhylA and ∆5’nagY strains showed a significant increase in survival relative to the parental WT strain (p < 0.001 and p < 0.01 respectively, Figure 9).