Identification and Characterization of a β-N-Acetylhexosaminidase with a Biosynthetic Activity from the Marine Bacterium Paraglaciecola hydrolytica S66T

β-N-Acetylhexosaminidases are glycoside hydrolases (GHs) acting on N-acetylated carbohydrates and glycoproteins with the release of N-acetylhexosamines. Members of the family GH20 have been reported to catalyze the transfer of N-acetylglucosamine (GlcNAc) to an acceptor, i.e., the reverse of hydrolysis, thus representing an alternative to chemical oligosaccharide synthesis. Two putative GH20 β-N-acetylhexosaminidases, PhNah20A and PhNah20B, encoded by the marine bacterium Paraglaciecola hydrolytica S66T, are distantly related to previously characterized enzymes. Remarkably, PhNah20A was located by phylogenetic analysis outside clusters of other studied β-N-acetylhexosaminidases, in a unique position between bacterial and eukaryotic enzymes. We successfully produced recombinant PhNah20A showing optimum activity at pH 6.0 and 50 °C, hydrolysis of GlcNAc β-1,4 and β-1,3 linkages in chitobiose (GlcNAc)2 and GlcNAc-1,3-β-Gal-1,4-β-Glc (LNT2), a human milk oligosaccharide core structure. The kinetic parameters of PhNah20A for p-nitrophenyl-GlcNAc and p-nitrophenyl-GalNAc were highly similar: kcat/KM being 341 and 344 mM−1·s−1, respectively. PhNah20A was unstable in dilute solution, but retained full activity in the presence of 0.5% bovine serum albumin (BSA). PhNah20A catalyzed the formation of LNT2, the non-reducing trisaccharide β-Gal-1,4-β-Glc-1,1-β-GlcNAc, and in low amounts the β-1,2- or β-1,3-linked trisaccharide β-Gal-1,4(β-GlcNAc)-1,x-Glc by a transglycosylation of lactose using 2-methyl-(1,2-dideoxy-α-d-glucopyrano)-oxazoline (NAG-oxazoline) as the donor. PhNah20A is the first characterized member of a distinct subgroup within GH20 β-N-acetylhexosaminidases.


Identification of Putative β-NAHAs in P. hydrolytica and Organization of Vicinal Genomic Regions
The marine bacterium P. hydrolytica degrades effectively many different polysaccharides [2] and its genome exhibits potential for the degradation of chitin and chitooligosaccharides. P. hydrolytica was grown in marine mineral medium supplemented with a mixture of chitooligosaccharides (GlcNAc) 1-6 as the sole carbon source, which were hydrolyzed to GlcNAc (Supplementary Information 1, Figure S1A,B). P. hydrolytica, however, did not hydrolyze α-chitin from crab shells used to supplement the marine mineral medium, as neither GlcNAc nor chitooligosaccharides appeared during the incubation ( Figure S1C). β-NAHA activity from P. hydrolytica was detected by a hydrolysis of the chromogenic 5-bromo-4-chloro-3-indolyl N-acetyl-β-d-glucosaminide (X-GlcNAc) on a complex marine agar medium ( Figure S1D). These results indicated that the bacterium produced at least one β-NAHA which was active towards chitooligosaccharides.
Top hits of protein BLAST, showing up to 54% to PhNah20A and up to 49% sequence identity to PhNah20B, were GH20 β-NAHAs or chitobiases from phylogenetically closely related marine and soil bacteria belonging mostly to the same order as P. hydrolytica-Alteromonadales (Table S1). None of these proteins, encoded by genes from Paraglaciecola or related bacteria (Table S1), had been recombinantly produced or characterized.
The closest relatives of PhNah20A are GH20 β-NAHAs from Bowmanella denitrificans and Lacimicrobium alkaliphilum with 53%-54% sequence identity (Table S1). PhNah20A contains two domains, the GH20 catalytic (β/α) 8 -barrel domain (Pfam: PF00728) and the N-terminal GH20b domain (also referred to as GH20 domain 2; Pfam: PF02838) of a predicted zincin-like fold similar to zinc-dependent metalloproteases [49] consisting of four antiparallel β-strands and an α-helix [27,50]. These two domains are typical for GH20 enzymes [50], and importantly they constitute an active and stable minimum functional unit of GH20 enzymes, thus requiring both a catalytic GH20 and a GH20b domain [50]. PhNah20A has no predicted signal peptide sequence and most probably is not secreted, whereas a 28 residues N-terminal signal peptide was predicted for the hypothetical PhNah20B ( Figure 1A). Therefore, during the growth of P. hydrolytica on chitooligosaccharides, PhNah20B probably performs the initial degradation of these substrates. PhNah20B, in addition to the GH20b and GH20 domains, contains a putative carbohydrate binding domain of the CHB_HEX superfamily (Pfam: PF03173) having a predicted β-sandwich structure similar to cellulose binding domains in cellulases [51], and a C-terminal CHB_HEX_C domain (Pfam: PF03174) of unknown function resembling an immunoglobulin-like fold [50,51]. A similar four-domain architecture was seen in the crystal structure of a chitobiase from S. marcescens [51], and has only been reported for bacterial GH20 enzymes [50,51]. Based on its protein sequence identity and domain architecture, PhNah20B resembles a biochemically uncharacterized GH20 chitobiase from Aliiglaciecola lipolytica and β-NAHAs from other phylogenetically close marine bacteria (Table S1). It can be concluded that one of the reasons for low sequence identity, i.e., 23%, between two putative GH20 enzymes of P. hydrolytica, was the different domain architecture of PhNah20A and PhNah20B ( Figure 1A), as PhNah20B has two additional domains besides the GH20 catalytic domain and an N-terminal GH20b domain. The identity between the two proteins remained low when only the predicted GH20b and GH20 domain sequences were compared, as some regions are not aligning between proteins (Supplementary Information 2).    Genomic regions flanking the two annotated P. hydrolytica β-NAHAs, PhNah20A and PhNah20B, were examined for the presence of operons ( Figure 1B), but were found not to be organized similarly to the operon responsible for chitobiose-utilization in Escherichia coli [52]. Surrounding putative genes, however, encoded proteins potentially participating in the modification of acetylated compounds, the transporter function and transcription regulation ( Figure 1B; Table S2). Notably, a predicted operon of six genes that harbors PhNah20A ( Figure 1B) included a putative amino acid deaminase, d-aminoacylase and the RidA (reactive intermediate/imine deaminase A) family protein, possibly associated with the processing of acylated compounds or amino acids [53]. A two-gene operon was predicted to harbor PhNah20B and a putative ATPase ( Figure 1B, Table S2). Thus, GH20 β-NAHAs genes of P. hydrolytica were not situated adjacent to genes encoding proteins directly coupled to β-NAHA activity, but flanking genes may be important for regulation or substrate transport.

Phylogenetic Analysis of PhNah20A and PhNah20B
Sequences of PhNah20A, PhNah20B and 41 characterized GH20 enzymes were aligned (Supplementary Information 2). PhNah20A and PhNah20B shared a low sequence identity with the other GH20 enzymes (up to 34.1% for PhNah20A and 37.9% for PhNah20B) and only a few highly conserved regions were identified among these GH20 members (Supplementary Information 2). The closest homologs of PhNah20A were Hex2 of an uncultured Bacteroidetes (34.1% identity) and ExoI of the marine bacterium V. furnissii (33.1% identity). Remarkably, GH20 sequences from eukaryotes (human and mouse) were 31.3% and 30.9% identical and more similar to PhNah20A than most other included bacterial sequences. The PhNah20B sequence was most similar to chitobiases from S. marcescens (37.9% identity) and V. harveyi (36.4% identity). The evolutionary relationship illustrated by a radial phylogenetic tree ( Figure 2; for bootstrap values see Figure S2) showed that bacterial GH20s segregate into three groups.
PhNah20B clustered with β-NAHAs from water-living bacteria from the phylogenetically close species such as V. harveyi, P. piscicida and A. hydrophila. However, PhNah20A did not cluster with characterized bacterial β-NAHAs but seems to represent a new distinct group of GH20 enzymes situated between predominantly water-living bacteria and the eukaryotes ( Figure 2).
NagA of the slime mold Dictyostelium discoideum which clusters not far from PhNah20A (Figure 2), is a lysosomal enzyme that maintains the size of pseudoplasmodia [54], and shares 28.5% sequence identity with PhNah20A. According to the BLAST analysis, PhNah20A has higher sequence identity to biochemically uncharacterized β-NAHAs from phylogenetically close marine bacteria (Table S1). Additionally, protein sequences with 44-47% identity to PhNah20A were found in compost, hydrothermal vent and marine sediment metagenomes (Table S1) highlighting unexplored resources harbouring a new group of β-NAHAs.
According to the literature, substrate specificities and biochemical features (e.g. pH and temperature optima) are reported for 41 β-NAHAs of GH20 [4,32] mostly from terrestrial organisms. The few enzymes being from marine bacteria comprise ExoI and chitobiase from Vibrio sp. [33,37], Hex99 and Hex86 from P. piscicida [21,35] and Nag20A and NagB from A. hydrophila [34,36]. The limited knowledge on GH20 from marine organisms motivated the present characterisation of β-NAHA from P. hydrolytica S66 T . Figure 2. Schematic phylogenetic tree of PhNah20A, PhNah20B (both marked with red circles) and 41 biochemically characterized GH20 (EC 3.2.1.52) enzymes. Evolutionary analyzes were conducted, and the tree was composed and visualized using MEGA v 7.0.26 [55]. Protein sequences were aligned with Clustal Omega and the BLOSUM62 protein weight matrix was used. Evolutionary relationships were calculated using the Neighbor-Joining method. Evolutionary distances were computed using the Poisson correction method. All positions containing gaps and missing data were eliminated, and there was in total 292 positions in the final dataset. The tree is in scale with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. Bacterial (○), fungal (□), plant (Δ), insect (▲) and mammal (∎) sequences. Amoebae and C. elegans sequences are marked with a filled diamond (♦). Characterized GH20 enzymes from marine organisms are underlined.

Figure 2.
Schematic phylogenetic tree of PhNah20A, PhNah20B (both marked with red circles) and 41 biochemically characterized GH20 (EC 3.2.1.52) enzymes. Evolutionary analyzes were conducted, and the tree was composed and visualized using MEGA v 7.0.26 [55]. Protein sequences were aligned with Clustal Omega and the BLOSUM62 protein weight matrix was used. Evolutionary relationships were calculated using the Neighbor-Joining method. Evolutionary distances were computed using the Poisson correction method. All positions containing gaps and missing data were eliminated, and there was in total 292 positions in the final dataset. The tree is in scale with branch lengths in the same units as those of the evolutionary distances used to infer the phylogenetic tree. Bacterial ( ), fungal ( ), plant (∆), insect ( ) and mammal ( ) sequences. Amoebae and C. elegans sequences are marked with a filled diamond ( ). Characterized GH20 enzymes from marine organisms are underlined.

Cloning and Production of β-NAHA
From the two candidate β-NAHA genes ( Figure 1A), only recombinant PhNah20A was successfully produced in E. coli ( Figure 3). PhNah20B cloned without the N-terminal signal peptide ( Figure S3) was not obtained despite expression attempts in three E. coli strains [BL21(DE3), BL21(DE3)∆lacZ and Rosetta], using different induction methods: isopropyl thio-β-d-galactoside (IPTG)-induction in lysogeny broth (LB) or auto-induction. PhNah20B was not found in the insoluble fraction by analyzing whole cells from IPTG-induced cultures ( Figure S4). The yield of PhNah20A was modest, probably due to a low expression level. Using different strains and induction strategies resulted in the highest β-NAHA activity of 6 µmol p-nitrophenol released per min and per mg protein in the E. coli cell lysate for IPTG-induced BL21(DE3) transformants in LB medium ( Figure 3).

Cloning and Production of β-NAHA
From the two candidate β-NAHA genes ( Figure 1A), only recombinant PhNah20A was successfully produced in E. coli ( Figure 3). PhNah20B cloned without the N-terminal signal peptide ( Figure S3) was not obtained despite expression attempts in three E. coli strains [BL21(DE3), BL21(DE3)ΔlacZ and Rosetta], using different induction methods: isopropyl thio-β-D-galactoside (IPTG)-induction in lysogeny broth (LB) or auto-induction. PhNah20B was not found in the insoluble fraction by analyzing whole cells from IPTG-induced cultures ( Figure S4). The yield of PhNah20A was modest, probably due to a low expression level. Using different strains and induction strategies resulted in the highest β-NAHA activity of 6 µmol p-nitrophenol released per min and per mg protein in the E. coli cell lysate for IPTG-induced BL21(DE3) transformants in LB medium ( Figure 3). Previously, an increased expression of GH20 β-NAHAs from a metagenome [18] was achieved in E. coli strains BL21(DE3), Turner, C41 or C43 grown in an auto-induction medium ZYM-5052 [56], but this medium gave a very low yield of PhNah20A ( Figure 3) and failed to lead to PhNah20B production.
Up to 2 mg of PhNah20A was purified in two chromatographic steps from one liter of E. coli BL21(DE3) culture (see Section 3.4). Expression of truncated PhNah20A and PhNah20B, containing only the catalytic and not the GH20b domain (see Figure S3), did not result in protein production which is in agreement with previous findings that GH20b is essential for enzyme production and activity [50]. Attempts to produce PhNah20B without the CHB_HEX domains ( Figure S3) also gave no detected protein or β-NAHA activity.

Enzyme Stability
The activity of PhNah20A decreased immediately after dilution to the low concentration of 5 µg mL −1 , even when kept on ice ( Figure 4). By contrast, 1 mg mL −1 PhNah20A retained activity at least four months at 4 °C in 50 mM sodium phosphate pH 7.0, 0.3 M NaCl and 0.02% NaN3. The presence Previously, an increased expression of GH20 β-NAHAs from a metagenome [18] was achieved in E. coli strains BL21(DE3), Turner, C41 or C43 grown in an auto-induction medium ZYM-5052 [56], but this medium gave a very low yield of PhNah20A ( Figure 3) and failed to lead to PhNah20B production.
Up to 2 mg of PhNah20A was purified in two chromatographic steps from one liter of E. coli BL21(DE3) culture (see Section 3.4). Expression of truncated PhNah20A and PhNah20B, containing only the catalytic and not the GH20b domain (see Figure S3), did not result in protein production which is in agreement with previous findings that GH20b is essential for enzyme production and activity [50]. Attempts to produce PhNah20B without the CHB_HEX domains ( Figure S3) also gave no detected protein or β-NAHA activity.
Notably, Hex, the commercial S. plicatus β-NAHA, is produced as a fusion with maltose-binding protein to secure stability and the Hex reaction mixture contained 0.3% of BSA to maintain activity [38]. 5A), whereas 0.5 and 2 M NaCl had no effect ( Figure 4). This behavior and the absence of a signal peptide suggest PhNah20A is an intracellular enzyme. Without a stabilizing agent, 5 µg·mL −1 PhNah20A was completely inactivated within 5 min at 50 °C, while 50% and 3% activity were retained after 20 min and 4 h, respectively, in 0.5% BSA ( Figure S5), and activity was fully retained after 4 d at 37 °C. β-NAHAs from E. coli [57], Prunus serotina [58], Bos taurus [59], Hordeum vulgare [60] and Streptomyces plicatus [61] were similarly found to lose activity by dilution. BSA has been identified as an activating compound to some β-NAHAs, e.g., from Mus musculus [41] and human plasma [42]. Notably, Hex, the commercial S. plicatus β-NAHA, is produced as a fusion with maltose-binding protein to secure stability and the Hex reaction mixture contained 0.3% of BSA to maintain activity [38]. PhNah20A on ice. The retained activity was measured at 37 °C in McIlvaine buffer, pH 6.0, using 2 mM pNPGlcNAc as the substrate. PhNah20A, BSA, NaCl and Triton X-100 were further 10 times diluted in the activity assay done at 0.5 µg mL −1 PhNah20A, 0.05% or 0.5% BSA, 0.05% or 0.5% Triton X-100, 0.05 or 0.2 M NaCl.
Kinetic parameters for PhNah20A hydrolyzing pNPGlcNAc and pNPGalNAc (Table 1) were very similar, kcat being slightly higher on pNPGalNAc. This identified PhNah20A as an Nacetylhexosaminidase rather than either an N-acetylglucosaminidase or an N-acetylgalactosaminidase. Most β-NAHAs, especially bacterial GH20 enzymes, prefer pNPGlcNAc (Table 1) and are referred to as N-acetylglucosaminidases. For instance, S. marcescens β-NAHA showed only 28.1% activity on pNPGalNAc compared to pNPGlcNAc [66]. Similarly, HexA from the ameba E. histolytica had 38% activity on pNPGalNAc compared to pNPGlcNAc [67]. Nag20A from A. hydrophila had very similar KM for pNPGlcNAc and pNPGalNAc, but Vmax for pNPGalNAc was only
Kinetic parameters for PhNah20A hydrolyzing pNPGlcNAc and pNPGalNAc (Table 1) were very similar, k cat being slightly higher on pNPGalNAc.

Transglycosylation by PhNah20A
There are a few reports on LNT2 formation by GH20 catalyzed transglycosylation with (GlcNAc)2 or pNPGlcNAc as donors and lactose as the acceptor [15,18,64] (see Figure 6). Hydrolysis of LNT2 by PhNah20A, an HMO core structure [70], warranted the investigation of the transglycosylation with (GlcNAc)2 and the GH20 reaction intermediate NAG-oxazoline [2-methyl-(1,2-dideoxy-α-D-glucopyrano)-oxazoline] [71] as a donor and lactose as an acceptor ( Figure 7A). We here also demonstrated transglycosylation by the commercial GH20 N-acetylglucosaminidase from S. plicatus (SpHex) [38] (see Figure S6), which has not been previously reported. Notably, the protein sequence identity between SpHex and a bacterial transglycosylating enzyme Hex1 isolated from a metagenome [18] was as high as 53.6%. Transglycosylation by β-NAHAs has been rarely investigated, and in one case there is a report on a bacterial GH20 enzyme for which no transglycosylation was detected [72], indicating that not all GH20 enzymes have the ability to transglycosylate. A GH20 chitobiase Hex99 from the Alteromonas sp. strain O-7 (currently classified as P. piscicida) of the order Alteromonadales formed β-GlcNAc-1,6-GlcNAc from (GlcNAc)2 by transglycosylation. It is to date the only marine GH20 enzyme reported to produce GlcNAc-containing oligosaccharides [21]. Notably, P. piscicida belongs to the same bacterial order as P. hydrolytica. Hex99 has a unique substrate specificity, as it hydrolyzed only chitobiose and pNP(GlcNAc)2, but neither other chitooligosaccharides nor pNPGlcNAc. PhNah20A transglycosylated lactose with NAG-oxazoline as the donor ( Figure 7A), resulting in three trisaccharides (Figure 8). 2, purified by gel permeation chromatography (GPC) ( Figure S7) migrated similarly to LNT2 in thin-layer chromatography (TLC), and nuclear magnetic resonance (NMR) spectroscopy confirmed the product structure ( Figure S9). 1 was determined to be a nonreducing trisaccharide, β-Gal-1,4-β-Glc-1,1-β-GlcNAc (Figure 8 and Figure S8; Tables S3 and S4), once reported as a transglycosylation product of a β-NAHA from Aspergillus flavofurcatis CCF 3061 [73]. For full NMR assignment as well as all measurable 3 JH,H coupling constants of 1, see Tables S3 and S4. The 1,1-linkage was supported by heteronuclear multiple-bond correlation spectroscopy (HMBC) and rotating frame nuclear Overhauser effect spectroscopy (ROESY) correlations between the two anomeric positions as well as by lack of a reducing end. Lastly, the β-configuration was determined of the anomeric positions using the 3 JH,H coupling constants between the anomeric proton and the neighboring proton (Table S4). A third trisaccharide (3) was detected, but not fully characterized due to low abundance. Based on chemical shifts of 3 ( Figure S10), however, it seemed unlikely that the galactose moiety in lactose acted as an acceptor, as none of the corresponding chemical shifts were affected. Consequently, most probably the glucose moiety was the acceptor. As O6 was determined to be unsubstituted and glucose was the reducing end residue, therefore either β-Gal-(β-GlcNAc)-1,2-Glc or β-Gal-(β-GlcNAc)-1,3-Glc was produced ( Figure S10  The overall transglycosylation yield for trisaccharides was estimated from the high-performance anion exchange chromatography with pulsed amperometric detector (HPAEC-PAD) chromatogram to 3.8% obtained with 200 mM acceptor and 100 mM donor. Since other trisaccharides were formed, no further optimization of transglycosylation conditions were pursued, even though LNT2 was the major product. Notably, the three trisaccharides were not completely separated by gel permeation chromatography (GPC) ( Figure S7), but thin-layer chromatography (TLC) and HPAEC-PAD analysis ( Figures S7 and S11) showed products consistent with trisaccharides 1 and 3 (Figure 8).
The acceptor specificity of PhNah20A was explored using D-galactose, D-glucose, 2-deoxy-Dglucose or L-fucose as an acceptor and NAG-oxazoline as a donor. These monosaccharides all proved to be transglycosylated ( Figure 7B and Figure S12) with the similar velocity and transglycosylation Transglycosylation by β-NAHAs has been rarely investigated, and in one case there is a report on a bacterial GH20 enzyme for which no transglycosylation was detected [72], indicating that not all GH20 enzymes have the ability to transglycosylate. A GH20 chitobiase Hex99 from the Alteromonas sp. strain O-7 (currently classified as P. piscicida) of the order Alteromonadales formed β-GlcNAc-1,6-GlcNAc from (GlcNAc) 2 by transglycosylation. It is to date the only marine GH20 enzyme reported to produce GlcNAc-containing oligosaccharides [21]. Notably, P. piscicida belongs to the same bacterial order as P. hydrolytica. Hex99 has a unique substrate specificity, as it hydrolyzed only chitobiose and pNP(GlcNAc) 2 , but neither other chitooligosaccharides nor pNPGlcNAc.
PhNah20A transglycosylated lactose with NAG-oxazoline as the donor ( Figure 7A), resulting in three trisaccharides (Figure 8). 2, purified by gel permeation chromatography (GPC) ( Figure S7) migrated similarly to LNT2 in thin-layer chromatography (TLC), and nuclear magnetic resonance (NMR) spectroscopy confirmed the product structure ( Figure S9). 1 was determined to be a non-reducing trisaccharide, β-Gal-1,4-β-Glc-1,1-β-GlcNAc (Figure 8 and Figure S8; Tables S3 and S4), once reported as a transglycosylation product of a β-NAHA from Aspergillus flavofurcatis CCF 3061 [73]. For full NMR assignment as well as all measurable 3 J H,H coupling constants of 1, see Tables S3 and S4. The 1,1-linkage was supported by heteronuclear multiple-bond correlation spectroscopy (HMBC) and rotating frame nuclear Overhauser effect spectroscopy (ROESY) correlations between the two anomeric positions as well as by lack of a reducing end. Lastly, the β-configuration was determined of the anomeric positions using the 3 J H,H coupling constants between the anomeric proton and the neighboring proton (Table S4). A third trisaccharide (3) was detected, but not fully characterized due to low abundance. Based on chemical shifts of 3 ( Figure S10), however, it seemed unlikely that the galactose moiety in lactose acted as an acceptor, as none of the corresponding chemical shifts were affected. Consequently, most probably the glucose moiety was the acceptor. As O6 was determined to be unsubstituted and glucose was the reducing end residue, therefore either β-Gal-(β-GlcNAc)-1,2-Glc or β-Gal-(β-GlcNAc)-1,3-Glc was produced ( Figure S10). products being detected in the most cases already after 0.03 h (2 min) incubation. Therefore, 2 h incubation was sufficient to assess the transglycosylation ability of PhNah20A (Figure 7 and Figure  S12). PhNah20A thus showed unusual promiscuity towards acceptor molecules, but due to the low yields and formation of several products as seen by TLC (Figure 7, Figure S12), purification and NMR analysis were not pursued. Remarkably, however, the ability to transglycosylate a wide range of acceptors has very rarely been reported for GH20 enzymes [18] and perhaps is associated with the marine origin and the unique phylogenetic relation of PhNah20A. S. marcescens Chb (see Section 2.2) is able to transglycosylate several alcohols, albeit sugar alcohols were not effective acceptors [66]. Some bacterial and fungal β-NAHAs can use lactose as their acceptor [15,18,64], and two Hex enzymes from uncultured bacteria were reported to transfer GlcNAc to D-glucose, D-galactose, sucrose and maltose [18].   The overall transglycosylation yield for trisaccharides was estimated from the high-performance anion exchange chromatography with pulsed amperometric detector (HPAEC-PAD) chromatogram to 3.8% obtained with 200 mM acceptor and 100 mM donor. Since other trisaccharides were formed, no further optimization of transglycosylation conditions were pursued, even though LNT2 was the major product. Notably, the three trisaccharides were not completely separated by gel permeation chromatography (GPC) ( Figure S7), but thin-layer chromatography (TLC) and HPAEC-PAD analysis ( Figures S7 and S11) showed products consistent with trisaccharides 1 and 3 (Figure 8).

Bacterial Strains and Media
The acceptor specificity of PhNah20A was explored using d-galactose, d-glucose, 2-deoxy-d-glucose or l-fucose as an acceptor and NAG-oxazoline as a donor. These monosaccharides all proved to be transglycosylated ( Figure 7B and Figure S12) with the similar velocity and transglycosylation products being detected in the most cases already after 0.03 h (2 min) incubation. Therefore, 2 h incubation was sufficient to assess the transglycosylation ability of PhNah20A (Figure 7 and Figure S12). PhNah20A thus showed unusual promiscuity towards acceptor molecules, but due to the low yields and formation of several products as seen by TLC (Figure 7, Figure S12), purification and NMR analysis were not pursued. Remarkably, however, the ability to transglycosylate a wide range of acceptors has very rarely been reported for GH20 enzymes [18] and perhaps is associated with the marine origin and the unique phylogenetic relation of PhNah20A. S. marcescens Chb (see Section 2.2) is able to transglycosylate several alcohols, albeit sugar alcohols were not effective acceptors [66]. Some bacterial and fungal β-NAHAs can use lactose as their acceptor [15,18,64], and two Hex enzymes from uncultured bacteria were reported to transfer GlcNAc to d-glucose, d-galactose, sucrose and maltose [18].  Ireland). A mixture of chitooligosaccharides, (GlcNAc) 1-6 , was from Koyo Chemicals (Osaka, Japan). All other chemicals were purchased from Sigma-Aldrich (Merck, Darmstadt, Germany) and used without further purification. S. plicatus β-NAHA in fusion with maltose-binding protein was purchased from New England Biolabs (Ipswich, MA, USA).

Molecular Cloning and Plasmids
P. hydrolytica genomic DNA was purified using the Gentra Puregene Yeast/Bact kit B (Qiagen, Venlo, The Netherlands) and plasmid DNA was isolated using the GeneJET Plasmid Miniprep kit (Thermo Fisher Scientific, Waltham, MA, USA). DNA content was determined on NanoDrop Lite (Thermo Fisher Scientific, Waltham, MA, USA). Two putative P. hydrolytica β-NAHA-encoding genes were amplified from genomic DNA by Phusion high-fidelity polymerase (Thermo Fisher Scientific, Waltham, MA, USA) using specific primers (Table S5). Genes were cloned as full-length or truncated variants (see Figure S3) into the pURI3TEV vector by PCR cloning [79].
DNA sequencing (Eurofins Genomics, Ebersberg, Germany) verified that cloned sequences matched the sequences in the P. hydrolytica genome. Plasmids were transformed into E. coli DH5α or BL21(DE3) by electroporation.

Activity Assays
PhNah20A activity was routinely determined on 2 mM pNPGlcNAc at 37 • C in two-fold diluted McIlvaine buffer pH 6.0 (63 mM Na 2 HPO 4 ; 18 mM citric acid), containing 0.05% BSA. The reaction (total volume 500 µL) was performed in McIlvaine buffer, pH 6.0 (250 µl), 100 µl milliQ water and 100 µL of substrate was added. The reaction was initiated by adding 50 µL of PhNah20A (prepared immediately before use in McIlvaine buffer, pH 6.0, 0.5% BSA, and kept on ice) to the reaction mixture yielding a final concentration of 0.3-5 µg·mL −1 . The reaction was stopped typically after 2-5 min by 250 µL 1 M Na 2 CO 3 and the product was measured spectrophotometrically at 400 nm (Ultrospec 3100 pro UV/Visible spectrophotometer, GE Healthcare, Uppsala, Sweden) using pNP (ε 400 = 18,000 M −1 ·cm −1 ) as the standard. One U of activity was defined as the amount of enzyme releasing 1 µmol pNP per min from 2 mM pNPGlcNAc. pH activity optimum was determined for PhNah20A in McIlvaine buffers (pH 4.0-8.0) at 37 • C towards 2 mM pNPGlcNAc and the temperature optimum was determined from the initial rates of pNP release at temperatures in the range 10-65 • C at pH 6.0.

Kinetics
PhNah20A (final concentration 0.3-1.2 µg·mL -1 ) was added to initiate the hydrolysis of 0.05-2 mM pNPGlcNAc (six concentrations) and 0.1-2 mM pNPGalNAc (five concentrations) in 500 µL two-fold diluted McIlvaine buffer pH 6.0, 0.05% BSA at 37 • C. The reaction was stopped at suitable time points by the addition of 250 µL 1 M Na 2 CO 3 and quantified spectrophotometrically as above. Initial rates calculated from pNP formation versus time were plotted against substrate concentration and fitted to the Michaelis-Menten equation using OriginPro 2015 (OriginLab, Northampton, MA, USA) to obtain k cat and K M . The k cat /K M values were either calculated or determined from rates of hydrolysis at low substrate concentration.
The elution was done with (A) water; (B) 1 M NaOH; (C) 200 mM NaOH + 800 mM NaOAc isocratically using 7.5% B in A (25 min) followed by 100% C (1 min) and column re-equilibration (9 min) at 7.5% B in A at 1.0 mL·min −1 . Oligosaccharides in water (10 µL) containing 9 µM l-fucose as standard were injected by autosampler kept at 5 • C. LNT2, glucose, galactose, lactose, GlcNAc, (GlcNAc) 2 and chitooligosaccharides were used as standards for calibration. Reaction mixtures (0.5 mL) containing approximately 10 mg oligosaccharides were separated by GPC (Bio-Gel P-2, Bio-Rad Laboratories, Hercules, CA, USA; 16 × 900 mm XK16/100 mounted on an ÄKTAprime plus chromatography system, GE Healthcare, Sweden), eluted by degassed milliQ water at flow rate of 0.1 mL·min −1 at RT and pressure limit set to 0.3 MPa. Reducing sugar in collected fractions (2 mL) were quantified by the Nelson-Somogyi method [83] using glucose and GlcNAc as standards. Fractions containing trisaccharides were dried (SpeedVac, Thermo Fisher Scientific, Waltham, MA, USA) at 50 • C, dissolved in 50 µL milliQ water and subjected to TLC for the preliminary identification of transglycosylation products. For NMR analysis, identical trisaccharide-containing fractions from two GPC runs were pooled, dried (SpeedVac) and dissolved in 0.5 mL D 2 O (Sigma-Aldrich, USA). Each fraction contained a major component and trace amounts of one or two other products.

Nuclear Magnetic Resonance (NMR)
All NMR spectra were recorded on an 800 MHz Bruker Avance III (799.75 MHz for 1 H and 201.10 MHz for 13 C) equipped with a 5 mm TCI cryoprobe. Acetone was used as internal reference (2.22 ppm and 30.89 ppm for 1 H and 13 C, respectively). The following experiments were used for the structure elucidation: 1 H with presaturation, double quantum filtered correlation spectroscopy (DQF-COSY), rotating frame nuclear Overhauser effect spectroscopy (ROESY), heteronuclear single-quantum correlation spectroscopy (HSQC), heteronuclear single-quantum correlation spectroscopy-total correlation spectroscopy (HSQC-TOCSY) and heteronuclear multiple-bond correlation spectroscopy (HMBC) all performed using standard Bruker pulse sequences. LNT2 and lactose were used as reference compounds. Structural elucidation was carried out by first identifying all 1 H and corresponding 13 C chemical shifts using 1 H with presaturation and HSQC. Subsequently, the different signals belonging to each position in each monosaccharide were determined, primarily using DQF-COSY and HSQC-TOCSY, and finally the connections between the monosaccharides were determined using HMBC and ROESY, as well as comparing chemical shifts to reference compounds.

Conclusions
The genome of the marine bacterium P. hydrolytica S66 T encodes two putative GH20 β-N-acetylhexosaminidase (EC 3.2.1.52) having protein sequences that differed remarkably from earlier characterized β-NAHAs (≤30% identity). PhNah20A was positioned on a phylogenetic tree between β-NAHAs of water-associated bacteria, i.e., Vibrio furnissii and Aeromonas hydrophila, and unicellular eukaryotes (amobae). PhNah20A, produced in E. coli, was unstable if diluted, but was stabilized by BSA or Triton X-100. PhNah20A is a genuine β-NAHA with essentially the same catalytic efficiency for pNPGlcNAc and pNPGalNAc, and thus differs from most of the previously studied bacterial β-NAHAs, which prefer pNPGlcNAc as a substrate while some eukaryotic GH20 prefer pNPGalNAc. PhNah20A also hydrolyzed LNT2, a core structure of human milk oligosaccharides, and showed biosynthetic activity (transglycosylation) which is a poorly studied aspect of GH20 β-NAHAs, especially from eukaryotes and water-living prokaryotes. PhNah20A was able to form LTN2 by transglycosylation using NAG-oxazoline as a donor and lactose as an acceptor, LNT2, β-Gal-1,4-β-Glc-1,1-β-GlcNAc and β-Gal-1,4-(β-GlcNAc)-1,2/3-Glc being identified by NMR as main transglycosylation products. Several monosaccharides were also recognized as acceptors by PhNah20A. To date, based on pH and temperature optima, kinetic parameters or stability characteristics alone, no clear distinction can be made between eukaryotic versus prokaryotic or terrestrial versus aquatic GH20 β-NAHAs. However, this may be due to the very limited number of characterized β-NAHAs of salt or fresh water origin. PhNah20A is the first characterized member of a distinct group of GH20 β-NAHAs located phylogenetically between eukaryotic and prokaryotic enzymes.