A Novel Secreted Endoglycosidase from Enterococcus faecalis with Activity on Human Immunoglobulin G and Ribonuclease B*

The human pathogen Enterococcus faecalis can degrade the N-linked glycans of human RNase B to acquire nutrients, but no gene or protein has been associated with this activity. We identified an 88-kDa secreted protein, endoglycosidase (Endo) E, which is most likely responsible for this activity. EndoE, encoded by ndoE, consists of an α-domain with a family 18 glycosyl hydrolase motif and a β-domain similar to family 20 glycosyl hydrolases. Phylogenetic analysis of EndoE indicates that the α-domain is related to human chitobiases, and the β-domain is related to bacterial and human hexosaminidases. Recombinant expression of full-length EndoE or EndoEα, site-directed mutagenesis of the catalytic residues, mass spectroscopy, and homology modeling shows that EndoEα hydrolyzes the glycan on human RNase B, whereas EndoEβ hydrolyzes the conserved glycan on IgG. Denaturation experiments indicate that the chitinase activity on RNase B is not dependent on the tertiary structure, although it is on IgG. The ndoE gene and secreted EndoE are present in most E. faecalis but not in Enterococcus faecium isolates. Correspondingly, E. faecalis, but not E. faecium, degrades the glycan on RNase B during growth. Thus, we have identified a secreted enzyme from E. faecalis, EndoE, which by two distinct activities hydrolyzes the glycans on RNase B and IgG. Both activities could be important for the molecular pathogenesis and persistence of E. faecalis during human infections.

Enterococcus faecalis is a human commensal bacterium that inhabits the mucosal surfaces primarily in the gastrointestinal tract. Nevertheless, E. faecalis is a major health problem in humans because it can cause a wide variety of diseases by infecting the urinary tract, bloodstream, and endocardium (1). It is one of the leading causes of nosocomial infections and can cause bacteremia with substantial mortality (2). Therapy of infections is severely complicated by the intrinsic low and the high levels of acquired antibiotic resistance. The complete genomic sequence of the vancomycin-resistant E. faecalis isolate, V583, was recently published, demonstrating the importance of genetic mobile elements in development of resistance, and provides an excellent resource for identifying genes with potential roles in virulence and persistence (3).
Little is known about how E. faecalis obtains nutrients and persists in vivo, but it has been shown that it can release high mannose-type N-linked glycan from host glycoproteins that could support growth in vitro (4). This ability is associated with an endo-␤-N-acetylglucosaminidase activity detected in several enterococcal species, including E. faecalis, but not in the closely related Enterococcus faecium (5). No specific gene or protein has been linked to this activity.
Glycosyl hydrolases can be found in all types of organisms, from bacteria to mammals, and are classified by their enzymatic activities and by their primary and structural similarities (6 -8). An excellent up-to-date on-line resource describing the known glycosyl hydrolases and their classification can be found at afmb.cnrs-mrs.fr/CAZY/index.html. Glycosyl hydrolases of family 18 (FGH18) 1 include enzymes that hydrolyze GlcNAc polymers with varying specificity. The FGH18 chitinases degrade chitin, a polymer of GlcNAc, whereas other FGH18s, like the endo-␤-N-acetylglucosaminidases, very specifically hydrolyze the chitin core of various asparagine (N)linked glycans on glycoproteins. These include EndoF 1-3 and EndoH from Flavobacterium meningosepticum that are widely used in glycobiology research to characterize glycans on glycoproteins (9). The catalytic activities of FGH18s are dependent on a glutamic acid residue, serving as a proton donor, and a well defined 8-amino acid motif has been established for this family (10). The family 20 glycosyl hydrolases (FGH20s) consists of hexosaminidases from bacteria to mammals, including human lysosomal hexosaminidases that are important for degradation of glycoconjugates. Hexosaminidases (EC 3.2.1.52) hydrolyze ␤-glycosidically linked N-acetylglucosamine or N-acetylgalactosamine in glycans. FGH20s are also dependent on a Glu residue as proton donor (11,12), but there are no well defined enzymatic motifs. Rather, inclusion into this family is based on overall primary and structural similarities (8).
Endo-␤-N-acetylglucosaminidases have been identified in a number of Gram-positive pathogens. For instance, a secreted Streptococcus pneumoniae enzyme, belonging to FGH20, has activity on human glycoproteins and is suggested to be involved in virulence (13). Furthermore, Streptococcus oralis can hydrolyze and metabolize glycans from mucins and ␣ 1 -acid glycoprotein (14 -16). The important human pathogen Streptococcus pyogenes, which is closely related to E. faecalis, has recently been shown to secrete an endo-␤-N-acetylglucosaminidase (FGH18), EndoS, that exclusively hydrolyzes the complex-type biantennary glycan on the heavy (␥) chain of IgG (17). This activity alters the function of opsonizing IgG through impaired binding to Fc receptors and decreased activation of the classical pathway of complement and thereby increases bacterial survival in human blood (18). Proteolytic and glycolytic modulation of the host immune system seems to be a common theme in the pathogenesis of S. pyogenes as well as other bacterial pathogens (19). Therefore, we asked if E. faecalis might have an enzymatic approach similar to the other Gram-positive pathogens to modulate host immune defense and/or release glycan from glycoproteins for nutritional purposes.
In this study, we utilized the amino acid sequence of the endoglycosidase EndoS from S. pyogenes to probe available bacterial genomic sequences. We identified a novel gene, ndoE, and a corresponding secreted protein, EndoE, in E. faecalis that has glycosyl hydrolase activity on both human RNase B and IgG.

EXPERIMENTAL PROCEDURES
Bacteria, Plasmids, and Culture Conditions-Bacteria and plasmids used in this study are described in Table I. Enterococcal strains were cultured in Todd-Hewitt broth (Difco) at 37°C without aeration. Escherichia coli strains were cultured in Luria-Bertani broth (Sigma) at 37°C with aeration. 1.5% (w/v) agar (Difco) was used to solidify media when needed, and 50 g/ml carbenicillin (Sigma) was added to E. coli cultures harboring pGEX-5X-3 constructs. For analysis of RNase B hydrolysis during growth, enterococci were grown at 37°C for 24 h in 0.5ϫ chemically defined medium originally developed for S. pyogenes (20) with 0.5ϫ PBS and 0.5 mg/ml human RNase B (Sigma). Culture supernatants were analyzed on 15% SDS-PAGE.
SDS-PAGE and Lectin Blot Analysis-For analysis of glycosidase activity, 10 g of human IgG or RNase B (Sigma) was incubated with 1 g of purified EndoE, EndoE␣, and EndoS or 1 unit of N-glycosidase F (PNGaseF) (Sigma) in a total volume of 20 l of PBS for 16 h at 37°C. IgG samples were separated on 10% SDS-PAGE and RNase B samples on 15% SDS-PAGE followed by staining with Coomassie Blue. For lectin blot analysis, proteins were blotted to Immobilon-P PVDF membrane (Millipore) according to Towbin et al. (22). Membranes were blocked with Tris-buffered saline with 0.1% Tween 20 (Sigma) and incubated with 0.5 g/ml biotinylated GNL lectin (Vector Laboratories, Burlingame, CA). After washing in TBST, membranes were incubated with 1 g/ml peroxidase-labeled streptavidin (Vector Laboratories). After washing, membranes were developed using SuperSignal West Pico Chemiluminescent Substrate (Pierce) and exposed on Blue Bio Film (Denville Scientific, Metuchen, NJ). For estimation of the rate of hydrolysis for EndoE, 1 nM IgG-Fc or RNase B was incubated with dilutions (125 ng to 8 g) of EndoE for 20 h at 37°C. Subsequently, samples were analyzed by 15% SDS-PAGE and spot densitometry using an Alpha Imager 2200 Documentation and Analysis System (Alpha Innotech Corp., San Leandro, CA).
Denaturation of IgG and RNase B-For analysis of EndoE on partially or fully denatured substrate proteins, 10 g of IgG in 10 l of PBS was incubated at 37, 50, 60, 70, 80, and 90°C for 30 min followed by equilibration at 37°C. Subsequently, 1 g of EndoE in 10 l of PBS was added, and samples were incubated at 37°C for 2 h. Samples were analyzed on 10% SDS-PAGE. RNase B (10 g) was denatured by incubation in 20 l of PBS for 2 min at 90°C in the presence of 100 mM ␤-mercaptoethanol and 0.5% SDS as described by Chu (23). After cooling to 37°C, 1 g of EndoE was added, and samples were incubated at 37°C for 2 h, followed by analysis on 15% SDS-PAGE.
PCR Analysis of Enterococcal Strains-To investigate the presence of the ndoE gene in enterococcal isolates, single colonies from enterococci grown on Todd-Hewitt agar were boiled in 1ϫ PCR buffer followed by centrifugation at 10,000 ϫ g for 5 min. Supernatants (1 l) were used as template in a standard PCR using AmpliTaq (Applied Biosystems, Foster City, CA) with an annealing temperature of 48°C and oligonucleotide primers 5Ј-ATG-AAC-GGA-GTG-CAG-AAA-GGA-3Ј and 5Ј-TT-T-AAC-CAC-CAT-ATA-CTC-ACG-ATA-3Ј. These primers generated a 2505-bp product covering the whole ndoE gene from HER1044. Samples were separated on 1% TAE-agarose gel and stained with ethidium bromide.
Generation of Antibodies and Western Blots-For generation of polyclonal mouse antiserum, 200 g of EndoE was incubated with 50 l of 2% Alu-Gel-S suspension (Serva, Heidelberg, Germany) for 20 min at room temperature, followed by centrifugation at 10,000 ϫ g for 10 min and resuspension of pellet in 200 l of sterile PBS. At day 1, 100 l of the suspension were injected subcutaneously in BALB/C mice (Charles River Laboratories, Wilmington, MA). At day 14 mice were boosted with an additional 100 l of suspension. At day 24, mice were sacrificed and bled followed by serum preparation using standard protocols. Reactivity against purified EndoE was tested by Western blots (data not shown). For Western blot analysis of EndoE secretion from enterococcal isolates, cultures were centrifuged at 10,000 ϫ g for 10 min followed by precipitation of proteins in supernatants by using a final concentration of 5% trichloroacetic acid and centrifugation at 10,000 ϫ g for 20 min at 4°C. Samples were separated on 10% SDS-PAGE and electroblotted to PVDF membranes. Membranes were blocked in PBS with 0.1% Tween 20 (PBST) and 5% (w/v) skim milk (PBSTS), followed by incubation with 1:5,000 of EndoE antiserum in PBSTS, washing in PBST, incubation with 1:2,000 of goat anti-rabbit IgG conjugated to horseradish-peroxidase (Bio-Rad), and development using chemiluminescence as described above.
Sequence Alignments, Phylogenetic Analysis, and Homology Modeling-All protein alignments and phylogenetic analyses were performed using the MacVector TM 7.1.1 sequence analysis software suite (Accelrys Inc., San Diego, CA). Protein sequences were aligned using the ClustalW algorithm. Alignments were used for reconstruction of phylogenetic trees by using the neighbor-joining method, ignoring gaps and with uncorrected p values. The trees were validated using bootstrap analysis (1,000 repetitions), and consensus trees were generated.
A structure model of the FGH18 domain of EndoE was constructed using Geno3D automated homology modeling (geno3d-pbil.ibcp.fr) (24). This method involves a PSI-BLAST (25) search, selection of Protein Data Bank entries as templates, and calculation of agreement of secondary structure (26); distances and dihedral angle restraints are calculated from the alignment and applied onto the query sequence followed by analysis using the CNS software (27). A model of amino acids 119 -250 from EndoE was generated by using a chitinase from the fungus Coccidioides immitis (Protein Data Bank 1D2K) (28) as a template. Models were edited, and figures were created by using 3D-Mol (component of Vector NTI, Informax, Frederick, MD).
Mass Spectroscopy-For mass analysis, 500 g of RNase B or human IgG Fc fragments (Sigma) were incubated with 1 g of EndoE, 1 g of EndoS, or 10 units of PNGaseF (Sigma) in PBS for 16 h. Samples were desalted on a HiTrap Desalting column (Amersham Biosciences). IgG-Fc samples were reduced with a final concentration of 20 mM dithiothreitol. Samples were analyzed on reducing and non-reducing SDS-PAGE to verify cleavage and reduction of IgG-Fc. 1 l of samples was spotted onto a sample plate. 1 l of saturated solution of ␣-cyano-4-hydroxycinnamic acid (Fisher) in 50% acetonitrile (Burdick and Jackson, Muskegon, MI) and 0.1% trifluoroacetic acid (Sigma) was deposited on top, and the mixture was allowed to air-dry. MALDI-TOF mass spectra were acquired on a PerSeptive Biosystems Voyager-DE RP equipped with a nitrogen laser (337 nm). Mass spectra were obtained in positive ion mode with an acceleration voltage of 25 kV, grid voltage of 88.0%, and delay time of 300 ns and averaged over 256 laser shots. Spectra were calibrated with RNase B or bovine serum albumin.

Identification of a Putative Endoglycosidase in E. faecalis-
The amino acid sequence of the IgG-hydrolyzing endoglycosidase EndoS (GenBank TM accession number AAK00850) from S. pyogenes strain AP1 (17) was used as a probe to search the bacterial genomes by using the comprehensive microbial resource Blast at www.tigr.org. This resulted in one significant hit (p value of 4.0 ϫ 10 Ϫ7 ) in the recently published genome sequence of E. faecalis V583, a vancomycin-resistant isolate harboring the vanB resistance gene (3,29). The corresponding open reading frame encodes an 835-amino acid protein annotated as a ␤-N-acetylglucosaminidase (GenBank TM accession number NP_813917). Oligonucleotide primers constructed from the corresponding DNA sequence were used to amplify and clone a homolog from the E. faecalis strain HER1044. Sequencing of this gene revealed an open reading frame encoding an 838-amino acid protein that is 99% identical to the predicted ␤-N-acetylglucosaminidase from V583. The sequences of the protein (denoted EndoE for endoglycosidase of Enterococcus), and the corresponding gene ndoE, have been deposited in GenBank TM under accession number AY376354. When the EndoE sequence was analyzed using the SignalP web server for predicting signal sequence cleavage sites at www.cbs.dtu.dk/services/SignalP/ (30), it was predicted that the first 55 amino acids constitute a signal peptide with a signal peptidase cleavage site. After removal of the signal peptide, the mature EndoE protein has a predicted size of 88 kDa.
EndoE Belongs to Both Family 18 and 20 of Glycosyl Hydrolases by Similarity-To further analyze the primary sequence of EndoE, a conserved domain search was performed at the NCBI website at www.ncbi.nlm.nih.gov/blast/. This revealed that the carboxyl-terminal portion of EndoE contains a conserved N-acetyl-␤-hexosaminidase region belonging to FGH20s according to Henrissat (7). The amino-terminal region contains no putative conserved domains, but when EndoE was analyzed for protein patterns present in the Prosite data base at us. expasy.org/cgi-bin/scanprosite, a perfect FGH18 motif (7) was identified. Moreover, when the EndoE sequence was aligned with that of EndoS (the IgG glycan hydrolase from S. pyogenes used to identify EndoE), it was apparent that the similarity was confined to the amino-terminal regions of both proteins including the conserved FGH18 motif (Fig. 1A). The carboxylterminal domain instead is highly similar to another FGH20 endo-␤-N-acetylglucosaminidase from S. pneumoniae (Fig. 1B). This pneumococcal enzyme has activity on human glycoproteins and has a predicted role in virulence (13). Based on these similarities, we proposed a two-domain model for EndoE composed of a 391-amino acid FGH18 ␣-domain and a 392-amino acid FGH20 ␤-domain (Fig. 1C), and we hypothesized that the two domains might provide two distinct glycosyl hydrolase activities on host glycoproteins.
EndoE Is Closely Related to Hexosaminidases and Chitobiases from Mammals-Enzymes belonging FGH20 (N-acetyl-␤hexosaminidases or hexosaminidases for short) are highly conserved, and evidence for a common evolutionary origin has been presented. For example a hexosaminidase from Vibrio vulnificus has been shown to be closely related to a human hexosaminidase (31). The ␤-domain of EndoE was aligned with FGH20 hexosaminidases from the bacteria Vibrio vulnificus, Vibrio harveyi, S. pneumoniae, Lactococcus lactis, and Lactobacillus casei; the fungus Candida albicans; the silkworm Bombyx mori; the amoeba Entamoeba histolytica; and the mammals Homo sapiens (HexA and B) and Felix catus. This alignment was used to construct a phylogenetic tree. This analysis ( Fig. 2A) showed that EndoE␤ is clustered in a group with bacterial enzymes of L. casei, L. lactis, and S. pneumoniae. Most interesting, this group is in turn closely clustered with the enzymes from mammals (human and cat) and more distantly related to enzymes from Vibrio spp. This suggests that the FGH20 ␤-domain of EndoE has a common origin with hexosaminidases from mammals, including humans.
The ␣-domain of EndoE was aligned with FGH18 homologs from the bacteria F. meningosepticum (EndoF 1-3 and EndoH) FIG. 1. Alignment of EndoE with FGH18 and FGH20 homologs and a map of EndoE. EndoE was aligned with EndoS, an FGH18 from S. pyogenes (A), or an FGH20 hydrolase from S. pneumoniae (B). Asterisks indicate amino acid identities, and dots indicate similarities. The conserved FGH18 motif is shown in boldface letters. C, a schematic map of EndoE from E. faecalis HER1044. Ss indicates the predicted signal sequence. Numbers above the sequence refer to amino acids of the EndoE sequence submitted to GenBank TM (accession number AY376354). A putative FGH18 ␣-domain, which contains an FGH18 motif, and a putative FGH20 ␤-domain are indicated. (9,(32)(33)(34), Streptococcus equi (open reading frame with similarity to EndoS (19) from the unfinished genome sequence from www.sanger.ac.uk/Projects/S_equi), S. pyogenes (EndoS) (17), Corynebacterium pseudotuberculosis (CP40, originally described as a serine proteinase but contains a perfect FGH18 motif) (35), and the mammalian homologs from H. sapiens, Bos taurus, and Rattus norvegicus (36). When constructing a phylogenetic tree from this alignment, EndoE␣ was grouped together with the mammalian homologs, whereas all the other bacterial proteins were in a distinct group separated from the mammalian proteins (Fig. 2B). This suggests that EndoE␣ is more closely related to the mammalian FGH18s (di-N-acetylchitobiases or chitobiases for short) than to other bacterial FGH18s (chitinases and ␤-N-acetylglucosaminidases).
The Family 18 ␣-Domain of EndoE Is Responsible for the Glycosidase Activity on Human Ribonuclease B-It has been described previously that E. faecalis produces an endo-␤-Nacetylglucosaminidase activity that hydrolyzes human ribonuclease B (RNase B) (4), although no corresponding gene or protein has been identified. RNase B is a glycoprotein with a single N-glycosylation site predominantly occupied by high mannose-type glycans (37). Therefore, we investigated if recombinantly expressed purified EndoE had activity on RNase B. Both full-length EndoE and the ␣-domain were expressed fused to GST. After affinity purification, and factor Xa removal of the GST fusion partner, homogenous preparations of 88-kDa full-length EndoE and 44-kDa EndoE␣ were obtained (data not shown). These preparations were incubated with purified hu-man RNase B for 16 h followed by analysis on 15% SDS-PAGE. As controls, RNase B was incubated with purified recombinant EndoS from S. pyogenes (17), purified PNGaseF, or buffer alone. PNGaseF from F. meningosepticum is a peptide-N-(Nacetyl-␤-glucosaminyl) asparagine amidase that hydrolyzes at the base of the pentasaccharide core of N-linked glycan leaving no remaining carbohydrates attached to the asparagine (38). This revealed that both EndoE and EndoE␣ have activity on RNase B as indicated by a shift in the apparent size of RNase B from the fully glycosylated 15.9-kDa form to the deglycosylated 14.7-kDa form to the same extent as when incubated with PNGaseF (Fig. 3A). In contrast, the negative control EndoS has no activity on RNase B, because no shift as compared with RNase B incubated with buffer could be observed (Fig. 3A,  EndoS and PBS). This indicates that deletion of the ␤-domain of EndoE does not influence this activity, suggesting that the FGH18 ␣-domain is sufficient for hydrolyzing the N-linked high mannose glycans on RNase B. The single high mannose glycan on RNase B occurs in several different forms with varying numbers of mannose residues (5-9) resulting in five RNase B glycoforms (Man 5 -Man 9 ) (Fig. 3B). To determine in detail where in the RNase B glycan structure EndoE hydrolyzes, RNase B incubated with EndoE was analyzed by mass spectroscopy (MALDI-TOF). Analysis of RNase B incubated with buffer alone clearly shows the mass distribution of the five different glycoforms but also a minor peak at the position of completely unglycosylated RNase B (Fig. 3C, RNaseB). In contrast, EndoE hydrolysis shifts the size of RNase B into a single peak of 13,888 Da (Fig. 3C, RNaseBϩEndoE) with no apparent intermediates. This is 205 Da larger than the completely unglycosylated form of RNase (Fig. 3C, RNaseA), suggesting that one monosaccharide is still attached to the protein backbone. Even though one should be cautious with quantitative conclusions from this type of mass spectroscopy data, the fact that the smaller glycoforms of RNase B seems to be shifted to a larger extent than the larger ones could reflect a preference of EndoE for the smaller glycoforms of RNase B.
In order to estimate the rate of hydrolysis of the glycans for EndoE on RNase B, a fixed amount of RNase B was incubated with dilutions of EndoE and analyzed by SDS-PAGE and spot densitometry of hydrolyzed and unhydrolyzed RNase B (data not shown). This showed that EndoE hydrolyzes RNase B in a concentration-dependent manner and that 1 g of EndoE on average hydrolyzes the glycans from 0.55 nM of RNase B in 20 h.
Taken together, these results suggest that the ␣-domain of EndoE is responsible for the previously observed activity of E. faecalis that deglycosylates RNase B (4) and that the high mannose glycan is hydrolyzed in the chitobiose core leaving one N-acetylglucosamine on the protein backbone.
Full-length EndoE but Not EndoE␣ Has Glycosidase Activity on Human IgG-EndoS, an N-acetylglucosaminidase from S. pyogenes, has a specific activity on human IgG; it hydrolyzes the conserved complex-type biantennary N-linked glycan on the ␥-chain reducing its size by ϳ3 kDa (17). Because EndoE was identified by its similarity to EndoS and is similar to human glycosyl hydrolases, we investigated whether EndoE has activity on human IgG. Purified IgG was incubated with EndoE and EndoE␣ followed by analysis on SDS-PAGE and by lectin blot analysis using Galanthus nivalis (GNL) lectin. GNL specifically binds to mannose residues on the N-linked glycan on IgG (Fig. 4A) and has been used previously to analyze endoglycosidase activity on IgG (17). Separation on SDS-PAGE revealed that EndoE activity shifts the IgG ␥-chain ϳ3 kDa, comparable with what can be seen when incubated with EndoS ( Fig. 4B, Stain, EndoE and EndoS) (17). In contrast, EndoE␣ did not shift the ␥-chain as compared with IgG incubated with buffer alone. The 44-kDa protein in this lane marked with an asterisk represents EndoE␣ and not hydrolyzed IgG (Fig. 4B, Stain, EndoE␣ and PBS). PNGaseF, which completely removes the glycan on the ␥-chain, lead to a size shift of ϳ4 kDa (Fig.  4B, Stain, PNGaseF). It should be noted that the limited resolution of the SDS-PAGE only gives an indication of glycan hydrolysis that needs to be confirmed by other methods. Therefore, to verify that the size shifts correspond to hydrolysis of the glycan on the ␥-chain, samples were analyzed by lectin blot using GNL. This clearly showed that IgG incubated with EndoE did not react with GNL (Fig. 4B, Blot, EndoE and EndoE␣). The reduction in signal was comparable with what could be seen in IgG samples incubated with EndoS or PN-GaseF (Fig. 4B, Blot, EndoS and PNGaseF). In contrast, IgG incubated with EndoE␣ was not affected in its reactivity with GNL compared with IgG incubated with buffer alone (Fig. 4B, Blot, EndoE␣ and PBS).
To get an estimate of the rate of hydrolysis of the IgG glycans for EndoE, a fixed amount of IgG Fc fragments (IgG-Fc) was incubated with dilutions of EndoE and analyzed by SDS-PAGE and spot densitometry (data not shown). We used IgG-Fc instead of intact IgG because of difficulties in discriminating between hydrolyzed and unhydrolyzed full-length IgG on SDS-PAGE. This showed that EndoE hydrolyzes IgG-Fc in a concentration-dependent manner and that 1 g of EndoE on average hydrolyzes the glycans from 1.2 nM of IgG-Fc in 20 h.
To analyze the glycan hydrolysis in more detail, human IgG Fc fragments were incubated with EndoE, EndoS, and PNGaseF and analyzed by mass spectroscopy (MALDI-TOF). The rationale for using IgG-Fc is that we could not obtain satisfactory resolution of the mass spectra within the size range of intact IgG. Prior to MALDI-TOF analysis, samples were reduced to separate the ␥-chains held together by disulfide bonds, desalted, and analyzed by non-reducing SDS-PAGE. This revealed that EndoE, EndoS, and PNGaseF shifted the size of IgG-Fc in a manner comparable with what could be seen with intact IgG, although hydrolysis was not complete even after 20 h of incubation (Fig. 4C). MALDI-TOF spectra of samples showed that both EndoE and EndoS shifts IgG-Fc to single peaks of 25,307 and 25,286 Da, respectively, corresponding to a protein backbone with one N-acetylglucosamine and a core fucose (Fig. 4D, IgG-FcϩEndoE, EndoS). In contrast, PN-GaseF shifts IgG-Fc to a single peak of 24,936 Da representing a protein backbone with no remaining carbohydrates complying with the described activity for PNGaseF (Fig. 4D, IgG-FcϩPNGaseF). Taken together, these data indicate that EndoE hydrolyzes the conserved glycan on IgG in the chitobiose core, leaving one N-acetylglucosamine and a core fucose on the protein backbone, and that the ␤-domain is necessary for this activity.
Homology Modeling of the FGH18 Domain of EndoE-The structures of a number of glycosyl hydrolases have been solved, and despite a wide variety of folds, their overall topology can be divided into three main groups where the active site is within a pocket, a cleft, or a tunnel, respectively (6). In an attempt to predict the topology of EndoE by using homology modeling, it was possible to create a model of amino acids 119 -250 covering the predicted FGH18 active site using a chitinase from the pathogenic fungus C. immitis (28) as a template. A space-filled model of this part of EndoE represents the putative active site forming a tunnel with the catalytic Glu-186 residue at the orifice (Fig. 5A). Furthermore, when analyzing the model in detail, the Glu-186 residue is held in close proximity to Asp-184 (Fig. 5B). In the chitinase from C. immitis, the corresponding Asp residue is important for catalytic activity by electrostatically stabilizing the reaction transition state (39). All known FGH18s hydrolyze the glycosidic bonds by an overall retention with the nucleophilic catalytic base close to the carbohydrate anomeric carbon with an average distance between the catalytic residues (Glu and Asp) of 5.5 Å (40). In the generated model, the calculated distance between Glu-186 and Asp-184 is 4.1Å (Fig. 5B).
Taken together, even though interpretation of homology models should be considered with great caution, it appears that the FGH18 domain of EndoE forms a catalytic tunnel with Glu-186 and Asp-186 at a distance in concordance with a catalytic retention mechanism.
Site-directed Mutagenesis of FGH18 in EndoE-The glutamic acid in the FGH18 motif has been shown to be essential for enzymatic activity (10), and our homology modeling also indicated that Glu-186 is part of the FGH18 active site. Therefore, we investigated if a site-directed mutation in the ␣-domain of Glu-186 (Fig. 1) into glutamine affected the enzymatic activity of EndoE and on IgG and RNase B. Purified EndoE, EndoE␣, or EndoE(E186Q) was incubated with RNase B followed by analysis of size shifts on SDS-PAGE. This revealed that EndoE(E186Q) could not hydrolyze the glycans on RNase B, whereas EndoE and EndoE␣ hydrolyzed the glycans to the same extent (Fig. 6A), confirming that the EndoE FGH18 ␣-domain is responsible for the hydrolysis of RNase B and that Glu-186 is essential for this activity.
Furthermore, to examine the effect of E186Q mutation on IgG hydrolysis, purified IgG was also incubated with EndoE, EndoE␣, or EndoE(E186Q) and analyzed by SDS-PAGE and lectin blots. This revealed that EndoE and EndoE(E186Q) sizeshifted the ␥-chain to the same extent, whereas EndoE␣ did not shift compared with the other samples (Fig. 6B). Moreover, when the samples were analyzed using the mannose-recognizing lectin GNL, the EndoE and EndoE(E186Q) samples had no reactivity, whereas the EndoE␣ sample had strong reactivity with the lectin (Fig. 6B, GNL). This indicates that the FHG18 active site is not responsible for the activity on the IgG glycan, but rather that this activity can be attributed to the FGH20 ␤-domain of EndoE.
Identification of the FGH20 Active Site in Endo␤ by Mutagenesis-Our findings indicated that the FGH20 ␤-domain of EndoE is responsible for the hydrolysis of the IgG glycan. Therefore, we attempted to identify the catalytic FGH20 residue. As discussed previously there is no well defined enzymatic motif for FGH20, but it has been shown by site-directed mutagenesis that a Glu residue is essential for activity of a ␤-Nacetylhexosaminidase from Streptomyces plicatus (12). This is also the case for human ␤-hexosaminidase B (HexB), where an aspartic acid preceding the Glu residue also is important for activity (11). We aligned the 15 amino acids surrounding the Asp and Glu active site residues in hexosaminidases from S. plicatus and HexB with EndoE. This identified the sequence GGDE that is conserved in all three proteins, where Glu represents position 662 of the EndoE sequence (Fig. 7A). Therefore, we hypothesized that Glu-662 is the proton donor essen- tial for the FGH20 activity on IgG. To address this, we mutated Glu-662 in EndoE into Gln and analyzed the activity on RNase B and IgG. IgG or RNase B was incubated with EndoE or EndoE(E622Q) followed by SDS-PAGE analysis and GNL lectin blot of the IgG samples. This revealed that EndoE(E662Q) hydrolyzes the glycans on RNase B to the same extent as full-length EndoE (Fig. 7B), confirming that the EndoE FGH18 ␤-domain is not important for the hydrolysis of RNase B glycans.
In contrast, the E662Q mutation had effects on IgG hydrolysis, because EndoE(E662Q) could not size-shift the ␥-chain as full-length EndoE is determined by SDS-PAGE analysis (Fig.  7C). Moreover, when the samples were analyzed using the mannose-recognizing lectin GNL, IgG incubated with EndoE had little to no reactivity, whereas the EndoE(662Q) sample had strong reactivity with the lectin (Fig. 7C, GNL). As expected, the ␤-domain alone had no activity on RNase B, but surprisingly it had no activity on IgG either (data not shown). Because the activity on IgG is not dependent on the active site (Glu-186) of the ␣-domain, this suggests that the ␤-domain needs the whole or parts of the ␣-domain structure for its activity. Taken together, this indicates that the activity on the IgG glycan can be attributed to Glu-662 of the FGH20 ␤-domain of EndoE, and that even though the catalytic site of the ␣-domain is not necessary, the whole structure is required for activity.
Effect of Denaturation of RNase B and IgG on EndoE Activity-Glycosidases with activity on glycoproteins, such as the commonly used EndoF 1-3 , EndoH, and PNGaseF, require or are enhanced when the substrate glycoprotein is denatured (9). Therefore, we investigated whether denaturation of RNase B or IgG had any effects on EndoE activity. IgG was denatured by increasing temperature followed by 2 h of incubation at 37°C with full-length EndoE and analysis by SDS-PAGE. The SDS-PAGE analysis revealed that the apparent molecular mass of the IgG ␥-chain was shifted, as compared with incubation with buffer alone, when preincubated at 37-70°C, whereas incubation at 80 -90°C abolished the activity of EndoE (Fig. 8A). This indicates that EndoE, as EndoS, does not have activity on denatured IgG, unlike many other endoglycosidases.
We also determined the effect of RNase B denaturation on its hydrolysis by EndoE. Because RNase B is a very thermostable molecule, it was denatured and reduced in the presence of SDS as described previously (23) for analysis of PNGaseF activity on RNase B. Subsequently, denatured RNase B was incubated with purified EndoE, and samples were analyzed by SDS-PAGE for size shifts. This revealed that denatured RNase B was hydrolyzed to the same extent as native RNase B (Fig. 8B,

FIG. 6. Analysis of EndoE(E186Q) activity on RNase B and IgG.
A, purified RNase B was incubated with EndoE, EndoE␣, or EndoE(E186Q) and separated on 15% SDS-PAGE as indicated. Arrows to the right indicate the position of intact, fully glycosylated RNase B and RNase B without glycans, respectively. B, purified IgG was incubated with EndoE, EndoE␣, or EndoE(E186Q) and separated on 10% SDS-PAGE and stained, or electroblotted onto PVDF membranes and analyzed with GNL lectin. Asterisks indicate the positions of EndoE, EndoE␣, and EndoE(E186Q), respectively. lanes A and B). As a control, only denaturation of RNase B did not lead to a shift in size (Fig. 8B, lane C). This indicates that EndoE is not enhanced in its activity by denaturation of RNase B.
Distribution of the ndoE Gene and Secretion of EndoE among Enterococcal Isolates-In order to investigate the presence of the ndoE gene in other enterococcal isolates, PCR experiments using oligonucleotide primers based on the sequence of ndoE from HER1044 were performed on chromosomal DNA from several E. faecalis and E. faecium isolates. This revealed that a 2,500-bp product could be amplified in 11/11 of the E. faecalis strains, but no product could be generated in 3/3 E. faecium strains (data not shown). This suggests that the ndoE gene is harbored by most E. faecalis isolates, including vancomycinresistant strains, but is not present in E. faecium. This is in concordance with previous studies (5) describing that an Nacetylglucosaminidase activity hydrolyzing the glycans on RNase B could be detected in several enterococcal species, including E. faecalis, but not in E. faecium.
To investigate whether EndoE secretion correlates with the presence of the ndoE gene as determined by PCR, Western blot analysis on culture supernatants from all isolates using polyclonal mouse antiserum against recombinantly expressed fulllength EndoE was performed. This revealed that all E. faecalis strains secrete an ϳ88-kDa protein that reacts with the antiserum (Fig. 9A). The varying signal strengths between strains could represent production levels of EndoE but could also be explained by different reactivity of the antiserum because of sequence variations. In contrast, no signal could be detected in the E. faecium strains. This indicates that E. faecalis secretes EndoE as a single stable product during growth in vitro. Moreover, the temporal production of EndoE was investigated by Western blot analysis of culture supernatants withdrawn during growth of E. faecalis strain HER1044. As shown in Fig. 9B, EndoE secretion starts in mid-exponential growth phase, continues to accumulate during stationary phase, and is present as a stable 88-kDa protein after 24 h of growth. In summary, these data indicate that all tested E. faecalis isolates harboring the ndoE gene also secrete an EndoE protein during growth, and this protein is not degraded after prolonged stationary phase incubation in vitro.
Human RNase B Is Hydrolyzed during Growth of E. faecalis-In order to investigate whether enterococcal isolates secrete functional EndoE during growth, bacteria were grown in a poor medium with added human RNase B. Culture supernatants were subsequently analyzed for hydrolysis of the glycoprotein by SDS-PAGE. This revealed that all E. faecalis strains have the ability to hydrolyze RNase B into the unglycosylated form, but to somewhat varying extents. Strain B722 had apparent lower activity compared with most strains, whereas strain D76 completely shifted RNase B into the unglycosylated form, and these findings were consistent when the experiment was repeated (Fig. 10, HER1044 to EF25). These variations could not be fully explained by EndoE production levels as determined by Western blots (Fig. 9A) and might therefore indicate that there are EndoE variants with different activities on RNase B. In contrast, no detectable shift of RNase B could be observed during growth of the E. faecium strains (Fig. 10, EFSK-2 to EFSK-33). This indicates that E. faecalis but not E. faecium can hydrolyze the human glycoprotein RNase B during growth, and this activity most likely originates from the secreted enzyme EndoE identified in this study.

DISCUSSION
By using the sequence of a known secreted endoglycosidase, EndoS, from S. pyogenes as a virtual probe, we identified a putative homolog encoded in the recently published genome of the vancomycin-resistant E. faecalis V583. Subsequently, we also identified and sequenced a nearly identical open reading frame in the E. faecalis strain HER1044, and we denoted its protein product EndoE for endoglycosidase of E. faecalis. By analyzing the primary amino acid sequence of EndoE, we proposed a two-domain model with an ␣-domain containing a perfect FGH18 motif and a ␤-domain similar to FGH20s (7,8).
Phylogenetic analysis of EndoE␣ indicates that it is more closely related to mammalian FGH18s (chitobiases) than to other bacterial FGH18s (chitinases and N-acetylglucosaminidases) including EndoS from S. pyogenes. The mammalian chitobiases are lysosomal glycosidases involved in the degradation of N-linked glycoproteins (36). Phylogenetic analysis of EndoE␤ suggests that it is closely related to FGH20s (hexosaminidases) from both bacteria and mammals. The hexosaminidases HexA and HexB from humans are lysosomal proteins involved in degradation of glycoproteins, glycolipids, and proteoglycans. These enzymes are crucial for the degradation of gangliosides, essential outer-membrane lipids, and mutations in the corresponding genes leads to Sandhoff or Tay-Sachs disease with accumulation of gangliosides in neuronal lysosomes resulting in severe neurodegeneration (41)(42)(43).
Taken together, EndoE contains two distinct domains, which are related to human lysosomal hexosaminidase and chitobiase, respectively, which are important for turnover of glycoproteins and glycolipids, but with distinct functions. Human lyso- Arrows to the right indicate the position of intact, fully glycosylated RNase B and RNase B without glycans, respectively. C, purified IgG was incubated with EndoE or EndoE(E662Q) and separated on 10% SDS-PAGE and stained, or electroblotted onto PVDF membranes and analyzed with GNL. somal hexosaminidases are primarily active on N-linked glycans of high mannose type with little activity on complextype N-linked glycans (44), whereas human chitobiase has activity at the reducing end of complex-type N-linked glycans after prior removal of the glycan from the asparagine by a ␤-D-N-acetylglucosamine asparagine hydrolase (45). EndoE represents the first example of a bacterial enzyme combining a hexosaminidase domain and chitobiase domain closely related to human enzymes, which could have disease implications when secreted by the human pathogen E. faecalis.
Because it is known that E. faecalis can release high mannose glycans from glycoproteins (4), we hypothesized that En-doE might be responsible for this activity. We could show that recombinantly expressed EndoE hydrolyzes the high mannose glycan on human RNase B leaving one N-acetylglucosamine on the protein backbone. Furthermore, we could show that the ␣-domain was sufficient for this activity. This suggests that we have identified the enzyme responsible for the previously reported E. faecalis activity on high mannose N-linked glycans, and site-directed mutagenesis confirmed that the activity is dependent on the FGH18 active site in the ␣-domain. We could also show that all tested E. faecalis strains, but not E. faecium, hydrolyze the RNase B glycan during growth, confirming the results from earlier studies (5). This was further supported by our finding that the ndoE gene and secreted EndoE are present in all E. faecalis isolates and not in E. faecium isolates. We cannot exclude however that there may be other enzymes secreted during growth of E. faecalis that could contribute to glycan hydrolysis. An ndoE deletion mutant would certainly be of interest, because this might further verify the contribution of EndoE to glycan hydrolysis in vitro and also enable in vivo virulence and persistence studies. However, despite several attempts we have not been able to generate such mutants in the E. faecalis strains we are studying for yet unknown reasons. This might be clarified, however, by our ongoing efforts in finding a successful mutagenesis strategy, which would allow us to investigate the properties of EndoE-negative strains in infection models in the future.
Besides the FGH18 activity on human RNase B, full-length EndoE, but not EndoE␣, has FGH20 activity hydrolyzing the N-linked glycan on IgG. This activity is not dependent on the family 18 active site as indicated by site-directed mutagenesis (E186Q). Instead, we identified Glu-662 in the ␤-domain as being essential for IgG glycan hydrolase activity by using sitedirected mutagenesis (E662Q).
Most interesting, EndoE, as well as EndoS, does not require prior hydrolysis by an ␤-D-N-acetylglucosamine asparagine hydrolase that is required for activity of the related human chitobiases (45), in order to hydrolyze the complex-type glycan on the ␥-chain. Furthermore, EndoE does not have activity on denatured IgG in contrast to many other glycan hydrolases that have enhanced activity or require denaturation of the substrate glycoprotein (9). This suggests that the N-linked glycan on denatured IgG is not sufficient for activity but rather that EndoE is dependent on the tertiary structure of glycosylated IgG. This resembles the activity of the S. pyogenes FGH18 IgG glycan hydrolase EndoS (21), but interestingly the EndoE activity on the IgG glycan is mediated by the FGH20 ␤-domain that is not very similar to EndoS. EndoS does not contain any domain similar to FGH20s, indicating that S. pyogenes and E. faecalis use glycosyl hydrolases belonging to different families to achieve the same goal, hydrolysis of the conserved glycan on IgG.
The conserved IgG glycans attached to asparagine 297 are located in the cavity between the two ␥-chains and are thought to stabilize the molecule (46, 47) (Fig. 3). The glycan is important for several IgG functions including binding to Fc receptors, activation of complement, and clearance of immune complexes from circulation (48 -51). Furthermore, defects in IgG glycosylation have been associated with several autoimmune disorders such as Crohn's disease, rheumatoid arthritis, and systemic lupus erythematosus (52)(53)(54). It was recently shown that hydrolysis of this glycan by the S. pyogenes endoglycosidase EndoS increased the streptococcal survival in blood containing type-specific opsonizing antibodies. The impairment was mainly because of reduced binding of hydrolyzed IgG to Fc receptors and reduced activation of the classical pathway of complement (18). EndoE activity might also have effects on enterococcal survival within the host, especially during septicemia, because antibody titers against this commensal organism seems to be common in the normal population (55). Furthermore, it has been shown that neutrophil-mediated killing FIG. 8. EndoE activity on denatured IgG and RNase B. A, purified IgG was preincubated at temperatures of 37-90°C as indicated prior to incubation with EndoE (ϩ) or buffer (Ϫ) for 2 h at 37°C followed by separation on 10% SDS-PAGE. B, RNase B was denatured by incubation at 90°C under reducing condition in the presence of SDS prior to incubation with EndoE for 2 h at 37°C. Samples were analyzed on 15% SDS-PAGE. ϩ and Ϫ below the gel indicate if RNase B was denatured or not and if EndoE was added or not, respectively. Arrows to the right indicate the positions of intact and deglycosylated RNase B.
of E. faecalis is enhanced by specific antibodies to the organism (56), suggesting that modulation of IgG functionality might be an efficient bacterial strategy to persist in the bloodstream.
It is fascinating that E. faecalis has acquired two enzymatic domains related to human enzymes and incorporated them into a secreted protein for its own purpose, release of N-linked FIG. 9. Analysis of EndoE secretion in enterococcal isolates and analysis of the temporal production of EndoE. A, Western blot analysis using EndoE antiserum on precipitated culture supernatants separated on 10% SDS-PAGE and blotted to PVDF membranes. E. faecalis and E. faecium strains are as indicated above the blot. B, optical density of E. faecalis HER1044 measured during growth in Todd-Hewitt broth. Samples were withdrawn at indicated time points; culture supernatants were precipitated and separated on 10% SDS-PAGE followed by Western blot analysis using EndoE antiserum. glycans from host glycoproteins for nutritional purposes or possibly also immunomodulation. These activities might be important for the virulence and persistence of E. faecalis during human infections.