Characterization of a phospholipid‐regulated β‐galactosidase from Akkermansia muciniphila involved in mucin degradation

Abstract The gut microbe Akkermansia muciniphila is important for the human health as the occurrence of the organism is inversely correlated with different metabolic disorders. The metabolism of the organism includes the degradation of intestinal mucins. Thus, the gut health‐promoting properties are not immediately obvious and mechanisms of bacteria‐host interactions are mostly unclear. In this study, we characterized a novel extracellular β‐galactosidase (Amuc_1686) with a preference for linkages from the type Galβ1–3GalNAc. Additionally, Amuc_1686 possesses a discoidin‐like domain, which enables the interaction with anionic phospholipids. We detected a strong inhibition by phosphatidylserine, phosphatidylglycerol, phosphatidic acid, and lysophosphatidic acid while phosphatidylcholine and phosphatidylethanolamine had no influence. Amuc_1686 is the first example of a prokaryotic hydrolase that is strongly inhibited by certain phospholipids. These inhibiting phospholipids have important signal functions in immune response and cell clearance processes. Hence, Amuc_1686 might be regulated based on the health status of the large intestine and could therefore contribute to the mutualistic relationship between the microbe and the host on a molecular level. In this sense, Amuc_1686 could act as an altruistic enzyme that does not attack the mucin layer of apoptotic epithelial cells to ensure tissue regeneration, for example, in areas with inflammatory damages.

while inflammation control, gut peptide secretion and mucus layer thickness were improved. In addition, the gut barrier function was increased by strengthening the enterocyte monolayer integrity (Everard et al., 2013;Reunanen et al., 2015). Moreover, gastrointestinal disturbance of individuals with autism could be linked with a low relative abundance of A. muciniphila (Wang et al., 2011).
Interestingly, most of the described effects required viable cells while heat-killed cells did not improve the metabolic profiles (Everard et al., 2013). Only a few studies suggest a potentially negative effect of A. muciniphila at some circumstances, such as an exacerbation of gut inflammation in Salmonella typhimurium-infected gnotobiotic mice or the potentially harmful effect in conditions of a disrupted epithelial barrier (Ganesh, Klopfleisch, Loh, & Blaut, 2013;Kang et al., 2013).
The mucus layer that covers epithelial cells offers many ecological advantages to gut bacteria as carbon and nitrogen source, particularly in the colon where free carbon sources are limited (Derrien, Collado, Ben-Amor, Salminen, & Vos, 2008;Salyers, West, Vercellotti, & Wilkins, 1977). 16S RNA gene analysis of this complex microbial ecosystem and enrichment cultures of mucin-degrading bacteria from human feces indicated that A. muciniphila is a common and prevalent member of mucin-utilizing organisms in the human intestinal tract (Collado, Derrien, Isolauri, Vos, & Salminen, 2007;Derrien et al., 2008Derrien et al., , 2004. In spite of these obvious benefits that contribute to a decrease of different symptoms associated with metabolic diseases in the mammalian gut, mechanisms of bacteria-host interactions and exact mucin-degrading processes of A. muciniphila remained mostly unclear (Reunanen et al., 2015). It is known that only a fraction of colonic microbes is able to degrade mucin, providing nutrients also for other members of the colonic microbiota (Derrien et al., 2004;Hoskins & Boulding, 1981). Due to the high complexity and diversity of intestinal mucin glycan structures, cooperative action of different mucin-degrading organisms and different enzymes as sulfatases, proteases, and especially different glycoside hydrolases (GH) are necessary for an efficient degradation (Crost et al., 2016;Derrien et al., 2004;Willis, Cummings, Neale, & Gibson, 1996). Examples for these important GH family members are fucosidases (GH29 and GH95), exo-and endoβ-N-acetylglucosaminidases (GH84 and GH85), neuraminidases/sialidases (GH33), endoβ 1-4-galactosidases (GH98), and β -galactosidases (GH2, GH20 and GH42) (Crost et al., 2016).
In this study, a new type of β-galactosidase from A. muciniphila is described which belongs to GH family GH35. We were able to show a strong inhibitory effect of anionic phospholipids on this β-galactosidase, which might indicate a regulatory function of the C-terminal domain and is, to our knowledge, the first characterization of a prokaryotic hydrolase that is strongly inhibited by certain phospholipids. During the degradation of intestinal mucins, this regulatory module could be responsible for the discrimination between apoptotic and non-apoptotic epithelial cells. In this case, the C-terminal domain might increase the specificity of the extracellular hydrolase for epithelial mucins of viable cells.

| Culture conditions and standard molecular techniques
Escherichia coli DH5α and E. coli BL21 (DE3) were grown in lysogeny broth (Miller, 1972). For plasmid maintenance µg ml -1 ampicillin was added. All standard molecular techniques used in this study were done according to Sambrook, Fritsch, and Maniatis (1989).

| Construction of an amuc_1686 expression system
The β-galactosidase from A. muciniphila encoded by amuc_1686 was amplified via PCR without its native signal sequence and with or without its C-terminal discoidin domain. Each PCR product contained the endonuclease restriction sites SnaBI and XhoI at the 5′and 3′ end respectively. The fragments were ligated into the corresponding restriction sites of pASK-IBA5 resulting in the expressions vectors pASK-IBA5_noSP-amuc_1686 and pASK-IBA5_noSP-amuc_1686_ short (Table 1).

| Overexpression and purification of Amuc_1686 and Amuc_1686_short
Escherichia coli DH5α cells were used for vector cloning and vector amplification. For protein production overnight cultures of E. coli BL21 DE3 (5 ml) harboring plasmids of interest were used to inoculate 1 L LB medium and were incubated at 37°C in shaker flasks at 180 rpm. When the cultures reached an optical density at 600 nm of approximately 0.4, protein production was induced by addition of 0.2 µg ml -1 anhydrotetracycline. Cells were harvested at an optical density between 1.0 and 1.5 by centrifugation at 9,000 g.
Lysis and purification was done as previously described (Kosciow, Domin, Schweiger, & Deppenmeier, 2016). Polyacrylamide gel electrophoresis was performed according to Laemmli (1970) and visualization of the protein bands via silver stain was done as described by Blum, Beier, and Gross, 1987. Native conformation of Amuc_1686 was analyzed by applying the protein to a gel filtration chromatography using a HiLoad 16/60 Superdex 75 pg column (GE Healthcare) connected to an ÄKTApurifier system (GE Healthcare).
Equilibration was done with 50 mM Tris-HCl buffer pH 7, containing 150 mM NaCl.

| Measurement of enzyme activities
Enzyme activity assays for Amuc_1686 were performed in a combined buffer system containing sodium acetate, Tris-HCl, potassium dihydrogen phosphate and dipotassium phosphate (50 mM each). In case of colorimetric p-nitrophenol substrates, the activity was measured photometrically at 420 nm. The optimum pH of Amuc_1686 was determined using a combined buffer in the range from pH 5 to 10 at 30°C. The temperature optimum was determined at optimum pH and temperatures ranging from 20 to 80°C. Enzyme kinetics were measured by varying substrate concentration (0-40 mM) at optimal pH (7.5) and temperature (65°C).
Inhibition experiments were performed at 37°C and a 50 mM combined buffer system pH 7.5 was used. All potentially inhibiting compounds were dissolved in methanol prior to assay application.
The impact of different ions on the enzyme activity was analyzed in 100 mM Tris-HCl buffer pH 7.5.

| Determination of products from noncolorimetric substrates
For substrate spectrum analysis with non-colorimetric substrates, enzymatic assays were performed as described above and incubated for different time periods at 37°C. Samples were applied to HPLC using an Aminex-HPX87H column (BioRad, 300 × 7.8 mm) with 5 mM H 2 SO 4 as mobile phase at 25°C and a flow rate of 0.6 ml min -1 .
The amounts of products and the grade of substrate degradation was determined, using an UV-Vis detector at 210 nm and a refractive index detector. Products were quantified by comparison to calibration curves.

| Characterization of Amuc_1686
In this study, we identified and characterized the hypothetical GH Amuc_1686 from the anaerobic, mucin-degrading gut microbe A. muciniphila ATCC BAA-835. The corresponding gene amuc_1686 encodes a protein with a molecular mass of 86 kDa. BLAST analysis of Amuc_1686 using the SwissProt database indicated that 19 of the 20 closest hits were exclusively eukaryotic β-galactosidases from mainly mammalian species such as Homo sapiens (Q8IW92), Mus musculus (Q3UPY5), and Rattus norvegicus (Q5XIL5). Only one protein with significant homology (identity of 44%, query coverage of 77%) was found in the bacterial kingdom, the β-galactosidase Bga from Xanthomonas manihotis (P48982) (Taron, Benner, Hornstra, & Guthrie, 1995). A bioinformatic analysis of the amino acid sequence of Amuc_1686 with SignalP 4.1 (Nielsen, Engelbrecht, Brunak, & Heijne, 1997;Petersen, Brunak, Heijne, & Nielsen, 2011) and Phobius (Käll, Krogh, & Sonnhammer, 2004) revealed a potential signal sequence and a predicted signal peptidase cleavage site between position 17 and 18 at the N-terminus of the protein.
For heterologous protein production in E. coli BL21, the gene amuc_1686 without its native predicted signal sequence was was purified by streptavidin affinity chromatography to apparent homogeneity. A total of 2.5 mg of protein was obtained from 1 L E. coli culture. The enzyme showed a single band at 87 kDa when analyzed by polyacrylamide gel electrophoresis and silver stain which was in agreement with the predicted size of the recombinant tagged protein without signal peptide (Figure 1a,b).
Gel filtration chromatography revealed a monomeric native structure for Amuc_1686. Purified Amuc_1686 showed the highest catalytic efficiency at 65-70°C and at pH 7.5. Substrate spectrum analysis was performed using various artificial chromogenic nitrophenyl-linked sugars and natural di-and trisaccharides (Table 2) at 65°C and 37°C, respectively. The highest activity for chromogenic substrates was observed for para-nitrophenyl-β-d-galactopyranoside, with a V max of 370 ± 43 U mg -1 and a K m value of 0.5 ± 0.3 mM, resulting in a K cat of 530.3 ± 87.1 s −1 and a K cat /K m of 1060.6 ± 174.2 s −1 mM −1 (Figure 1a). The enzymatic activity with ortho-nitrophenyl-β-d-galactopyranoside as substrate was approximately 10 times lower and with para-nitrophenyl-N-acetyl-β-d-galactosaminide only trace activity could be detected. The enzyme was inactive with all other nitrophenyl-linked substrates ( inhibited the reaction completely (not shown).

| Discovery of a discoidin domain in Amuc_1686
In comparison to its closest homolog, the β-galactosidase Bga from X. manihotis (Taron et al., 1995), Amuc_1686 possesses an additional C-terminal domain ( Figure 3). Further analysis with the bioinformatic tools Pfam (Finn et al., 2016) and Prosite (Hulo et al., 2008;Sigrist et al., 2013) revealed that this part of the enzyme represents a truncated discoidin domain, also known as coagulation factor type C F5/8 domain or FA58C domain (Baumgartner, Hofmann, Chiquet-Ehrismann, & Bucher, 1998;Foster, Fulcher, Houghten, & Zimmerman, 1990;Ortel et al., 1994). Discoidin domains are found in blood coagulation factors 5 and 8 as a twice-repeated C-terminal domain of approximately 150 amino acids. These domains promote binding to specific phospholipids on cell surfaces of, for example, platelets or endothelial cells. It is known that in the case of coagulation factor 8 this domain is crucial for enzyme activity and phosphatidylserine-binding (Foster et al., 1990;Kane & Davie, 1988

| The influence of different phospholipids on the enzymatic activity of Amuc_1686 and construction of a mutant enzyme of Amuc_1686 without C terminal discoidin domain
It is known that discoidin domains, for example, found in the carboxylterminus of blood coagulation factors 5 and 8, can promote binding to cell surface phospholipids such as phosphatidylserine and are responsible for enzyme activity (Foster et al., 1990;Kane & Davie, 1988  F I G U R E 2 Enzymatic activity of Amuc_1686 with galacto-N-biose as Substrate. Combined puffer (150 µl) pH 7.5 containing 35 mM galacto-N-biose was incubated with 2.5 µg enzyme for 6 hr at 37°C. Substrate and products were analyzed via HPLC UV detection after an incubation time of 0 hr (a) and 6 hr (b). After 6 hr, more than 98% of galacto-N-biose (a) was hydrolyzed into GalNAc (b). Peak at 15.75 min = internal acetate standard. The experiment was conducted in triplicate using different protein preparations. One representative experiment is shown To answer the question whether the C terminus containing the discoidin domain is responsible for the phospholipid induced inhibition of Amuc_1686, a shortened variant of Amuc_1686 was constructed (Amuc_1686_short), lacking the last 107 amino acids. For inhibition experiments, 1 µg of Amuc_1686_short was applied to a standard assay that was identically performed as with Amuc_1686.
When directly compared with Amuc_1686, the shortened variant exhibited a slightly decreased enzymatic activity but showed no significant inhibitory effect after addition of PS, PG and PA in the same concentration as applied to the assay containing the full-length enzyme (Figure 4).
In order to analyze whether the hydrophilic head groups of the phospholipids have an influence on enzyme activity, l-serine, glycerol, choline and ethanolamine were added to the standard assay up to a final concentration of 3 mmol L -1 . Under these conditions, no inhibition was observed. To identify the type of inhibition, the K m value for para-nitrophenyl-β-d-galactopyranoside in the presence of 0.25 µmol L -1 PG was determined. As the measured values were in the same range as without inhibitor addition, we concluded a non-competitive type of inhibitory effect of the phospholipids on Amuc_1686.

| D ISCUSS I ON
The Gram-negative gut bacterium A. muciniphila is able to utilize the mucus layer that covers colonic epithelial cells in the human large intestine. Mucin is used by this organism as carbon and nitrogen source which has an ecological advantage due to limitation of free carbon sources in this specific gut region (Derrien et al., 2008;Salyers et al., 1977). The degradation of the highly complex mammalian mucin glycan structures involves microbial cooperative action and a set of differently specialized GH (Crost et al., 2016;Derrien et al., 2004;Willis et al., 1996). Among other enzymes, especially β-galactosidases play a major role in the efficient degradation of the oligosaccharide chains of mucins (Crost et al., 2016).
Therefore, the characterization of the β-galactosidase Amuc_1686 was done to get more insight into the mucin-degrading mechanisms of A. muciniphila.
The only hydrolytic activity of Amuc_1686 with a non-chromogenic substrate was detected with Galβ1-3GalNAc (Table 1).
Therefore, we conclude that this type of glycosidic bond (also known as T antigen) is the target structure for the enzyme within the oligosac-  (Baumgartner et al., 1998;Sauer et al., 1997;Villoutreix & Miteva, 2016). During these events the discoidin domain itself is responsible for integrin receptor binding, phospholipid binding or carbohydrate chain binding (Borisenko, Iverson, Ahlberg, Kagan, & Fadeel, 2004;Hidai et al., 1998;Poole, Firtel, Lamar, & Rowekamp, 1981). In this work, we could show a strong inhibitory regulation of Amuc_1686 from A. muciniphila by the anionic phospholipids PA, PG, PS and LPA. It became evident that the C-terminal discoidin domain of the enzyme is responsible for this regulatory effect because the shortened variant of Amuc_1686, missing the discoidin domain, showed no significant inhibition by these lipids.
In eukaryotic cells, enzyme inhibition by phospholipids was previously described (Stace & Ktistakis, 2006). The γ isoform of the human protein phosphatase-1 catalytic subunit (PP1cγ) is a high-affinity target of the bioactive lipid second messenger PA which inhibits the enzyme non-competitively and dose dependently with an IC 50 of 15 nM (Jones & Hannun, 2002). Moreover, PS and PA were shown to inhibit the Ca 2+ -ATPase of the sarcoplasmic reticulum (Dalton et al., 1998) It is known that A. muciniphila positively influences and promotes wound healing and tissue regeneration (Alam et al., 2016).
Due to its metabolism, which includes the degradation of intestinal mucins that act as innate host defense and protective barrier to infections, the gut health-promoting properties seem to be contradictory (Dharmani, Srivastava, Kissoon-Singh, & Chadee, 2009;Linden, Sutton, Karlsson, Korolik, & McGuckin, 2008). However, when apoptosis due to infection or tissue damage of epithelial cells is induced, mucin-like structures play a key role in many events crucial for immune response, cell renewal and tissue regeneration ( Figure 5). One of the first events during programed cell death in epithelial cells is the exposure of PS, which constitutes 5%-10% of total cellular lipid, on the outer leaflet of the membrane (Stace & Ktistakis, 2006;Yamaji-Hasegawa & Tsujimoto, 2006). The normal asymmetrical architecture and the rapid externalization of PS is one of the most important "eat me" signals in the clearance process. In addition, specific mucin-like structures on the epithelial cell surfaces are necessary for a proper cell-cell interaction with dendritic cells and macrophages (Ravichandran, 2010). In view of the mostly mutualistic relationship between A. muciniphila and its host, a regulation of the extracellular mucin-degrading enzyme Amuc_1686 on protein level is conceivable ( Figure 5). An interaction of Amuc_1686 with exposed PS might lead to a down-regulation of the hydrolytic activity in regions with increased numbers of apoptotic cells or wounded tissues, which would result in an unimpeded immune response and therefore an improved tissue regeneration and wound healing.
In addition to PS, a strong inhibitory regulation of Amuc_1686 was observed by the anionic phospholipids PA, PG and LPA. PA is known to serve as a lipid second messenger (McPhail et al., 1999;Stace & Ktistakis, 2006;Voelker, 1991) and is a central pro-inflammatory mediator whose generation is initiated by inflammatory cytokines such as tumor necrosis factor-α and interleukin-1β (Greaves & Camp, 1988;Sturm, 2002). LPA as a derivative of PA can be formed extracellularly by the action of inflammatory type II phospholipase A2 or phospholipases from pathogens by hydrolysis of PA from membrane surfaces in intestinal epithelial cells. LPA shows pro-inflammatory properties similar to PA (Fourcade et al., 1995;Moolenaar, 2000). In wounded tissues, PA and LPA can be present nearby apoptotic stressed epithelial cells, occurring on the surface of small shed microvesicles (Fourcade et al., 1995). Thereby PA and LPA could act as sensible inhibitors of Amuc_1686 leading to the protection of the mucin layer in regions of the large intestine with tissue irritation, injury, or infection ( Figure 5).
An inhibitory effect on Amuc_1686 was not detected in the presence of PC and PE. These phospholipids are found in high abundance in membranes of viable eukaryotic cells, constituting about 40%-50% and 20%-50% of total phospholipids, respectively (Chaurio et al., 2009;Yamaji-Hasegawa & Tsujimoto, 2006).
Another aspect is that carbohydrate chains of the cellular surface function as a signal for apoptotic cell removal (Beppu, Eda, Fujimaki, Hishiyama, & Kikugawa, 1996;Yamanaka, Eda, & Beppu, 2005). It is known that a transient accumulation of the mucin-like major sialoglycoprotein CD43 (also known as leukosialin) occurs on membranes of apoptotic cells. The binding of CD43 results in a condensation of the carbohydrate chains on the cellular surface of the apoptotic cell leading to the formation of target structures for the binding of macrophages (Eda, Yamanaka, & Beppu, 2004;Yamanaka et al., 2005). In the mucin-like structures of CD43, linkages from the type Galβ1-3GalNAc are very common (Piller, Deist, Weinberg, Parkman, & Fukuda, 1991). Interaction of exposed PS with the discoidin domain of Amuc_1686 would inhibit enzyme activity ( Figure 5) avoiding a degradation of this silyl polylactosaminyl saccharide ligands. Such a regulation would be highly beneficial for the host, as a degradation of these oligosaccharide chains would lead to a deterioration of macrophage recognition and therefore a worse clearance and tissue regeneration. Moreover, a slowed removal of apoptotic cells would cause further inflammation and autoimmune responses against intracellular antigens released from dying cells (Kobayashi et al., 2007;Savill & Fadok, 2000).
A further example for a mucin-like structure involved in immune response is the glycoprotein lactadherin which recognizes and binds to PS by a discoidin domain. This protein contains mucin-like glycosidic structures, enhances phagocytosis and is crucial for repair of intestinal epithelium (Bu et al., 2007). Additionally, T cell immunoglobulin mucin proteins 1 and 4 (TIM-1, TIM-4) expressed on mammalian macrophages and dendritic cells specifically bind PS on the surface of apoptotic cells and mediate T cell activation and apoptotic cell uptake. Structurally TIM 1 and 4 are glycoproteins as well, exhibiting important mucin-like glycosidic structures (Kobayashi et al., 2007). Hence, the inhibition of Amuc_1686 by PS would lead to a protection of mucin-like structures found in important compounds of the immune system.
In summary, Amuc_1686 can be classified as a phospholipid-regulated enzyme that might not attack apoptotic epithelial cells to ensure tissue regeneration and to avoid a disorder of the clearance process.
This mutualistic behavior would support health of the host but also implicates that substrate for growth of A. muciniphila is limited from damaged areas of the large intestine. However, further experiments have to be done in the future to verify or to refute the proposed hypothesis.

ACK N OWLED G EM ENT
We thank Thomas Franke for helpful discussions and Prof. Paul Schweiger for critical proofreading of the manuscript. Additionally, we want to thank Dr. Fabian Grein for the provision of phospholipids and Jennewein Biotech GmbH (Rheinbreitbach) for the provided 2'fucosyllactose.

CO N FLI C T O F I NTE R E S T
There is no conflict of interest for all authors.

AUTH O R S CO NTR I B UTI O N
Uwe Deppenmeier and Konrad Kosciow conceived and designed the research. Konrad Kosciow performed the experiments and wrote the initial manuscript draft. Uwe Deppenmeier supervised the research and revised the manuscript draft.

E TH I C S S TATEM ENT
None required.

DATA ACCE SS I B I LIT Y
Data available upon request from the authors.