The Bactericidal Activity of the C-type Lectin RegIIIβ against Gram-negative Bacteria involves Binding to Lipid A*

Background: The C-type lectin RegIIIβ has bactericidal activity against certain Gram-negative bacteria. The mechanism had remained unclear. Results: RegIIIβ binds to lipid A through a loop motif, which is also required for the bactericidal activity. Conclusion: Bacterial binding of RegIIIβ to lipid A is critical for its bactericidal activity against Gram-negative bacteria. Significance: This study provides novel insights into the bactericidal mechanism and recognition specificity of RegIIIβ. RegIIIβ is a member of the C-type lectin family called RegIII. It is known to bind peptidoglycan, and its bactericidal activity shapes the interactions with commensal and pathogenic gut bacteria. However, little is known about its carbohydrate recognition specificity and the bactericidal mechanism, particularly against Gram-negative bacteria. Here, we show that RegIIIβ can bind directly to LPS by recognizing the carbohydrate moiety of lipid A via a novel motif that is indispensable for its bactericidal activity. This bactericidal activity of RegIIIβ could be inhibited by preincubation with LPS, lipid A, or gentiobiose. The latter is a disaccharide composed of two units of β-(1→6)-linked d-glucose and resembles the carbohydrate moiety of lipid A. Therefore, this structural element may form a key target site recognized by RegIIIβ. Using point-mutated RegIIIβ proteins, we found that amino acid residues in two structural motifs termed “loop 1” and “loop 2,” are important for peptidoglycan and lipid A binding (Arg-135, Asp-142) and for the bactericidal activity (Glu-134, Asn-136, Asp-142). Thus, the ERN motif and residue Asp-142 in the loop 2 are of critical importance for RegIIIβ function. This provides novel insights into the carbohydrate recognition specificity of RegIIIβ and explains its bactericidal activity against Gram-negative bacteria.

The gastrointestinal tracts of mammals represent highly complex ecosystems colonized with very dense microbial communities called "microbiota." These communities are composed of several hundred different bacterial species (1,2). However, the mechanisms shaping the community structure of the microbiota are still not well understood. Antimicrobial proteins produced by intestinal epithelial cells such as ␣-defensins can affect the composition of gut microbiota and limit opportunistic invasion by symbiotic bacteria (3,4). In addition, antimicrobial proteins also play a pivotal role in protecting from enteropathogenic infections (4,5).
The bactericidal RegIII (regenerating gene family protein III) lectin family plays a crucial role in maintenance of microbiota homeostasis with commensal bacteria and protection of the host from pathogens (6 -11). The matured RegIII lectins have a molecular mass of ϳ16 kDa and are secreted from Paneth cells and epithelial cells of the intestine into gut lumen (6,12). The murine intestine expresses two RegIII family members, RegIII␤ and RegIII␥ (13). Production of RegIII␤ and RegIII␥ is dramatically increased in response to bacterial colonization and pathogenic infection (6,8,12). Similarly, human hepatocarcinoma intestine pancreas/pancreatitis-associated protein (HIP/PAP), 3 a human homolog for RegIII␥, is expressed in the small intestine, and strong HIP/PAP expression has been shown in patients with inflammatory bowel disease (14,15). Hence, RegIII proteins are thought to contribute substantially to the bacteria-host interaction in the gut.
The different RegIII lectins have distinct spectra of antimicrobial activity. RegIII␥ and HIP/PAP recognize peptidoglycan and display bactericidal activity against Gram-positive bacteria, whereas Gram-negative bacteria are not affected. Presumably, this is attributable to the topology of the Gram-negative cell envelope where the LPS-loaded outer membrane shields the access of RegIII␥ and HIP/PAP to the peptidoglycan (6). In contrast, despite sharing a high degree of sequence similarity to HIP/PAP and RegIII␥, RegIII␤ can kill not only Gram-positive bacteria, but also some certain Gram-negative bacterial species. For example, RegIII␤ was found to kill Clostridium butyricum, Lactobacillus reuteri, and different Escherichia coli strains but not Enterococcus faecalis, Lactobacillus murinus, and Salmonella typhimurium (8). The mechanism explaining the activity against some Gram-negative species (but not others) had remained unknown.
As a structural feature, RegIII lectins have two distinct loop regions called loop 1 and loop 2 (16,17). In HIP/PAP, and possibly also in RegIII␥, a canonical EPN tripeptide motif in loop 1 was shown to be essential for carbohydrate recognition and bactericidal activity (18). Although the canonical motif is also * This work was supported by Grant ETH-08-08-3 from the ETH Zurich found in loop 1 of RegIII␤, roles of the motif in RegIII␤-mediated carbohydrate recognition and bactericidal activity have remained unknown. Furthermore, little is known about the carbohydrate recognition specificity of RegIII␤ and how this protein kills bacteria.
Here, we show that the bactericidal effect of RegIII␤ against Salmonella typhimurium, an enteropathogenic Gram-negative bacterium that causes self-limiting diarrhea in humans and diverse animal models, including mice (19), is dependent on the bacterial growth phase. RegIII␤ kills S. typhimurium in logarithmic but not in stationary growth phase. The bactericidal activity correlates with the ability to bind to the bacterial surface. We finally show that RegIII␤ binds to LPS by recognizing the carbohydrate moiety of lipid A via a novel motif in the loop 2 region. Moreover, we found that Asp-142 is also involved in lipid A binding and the bactericidal activity of RegIII␤. Our results provide new insights into the carbohydrate recognition specificity of the RegIII lectin family and explain why RegIII␤ is bactericidal for Gram-negative bacteria.

EXPERIMENTAL PROCEDURES
Bacterial Strains-S. typhimurium SL1344 was used in this study (20). Bacteria were cultured in LB broth under mild aeration. Specifically, the bacteria were grown in 1-cm diameter glass test tubes in 3 ml of LB broth on a rotating wheel (160 rpm) at 37°C overnight, diluted 1:200 into a 100-ml Erlenmeyer flask (20 ml of LB broth), incubated at 37°C in an orbital shaker (160 rpm) and grown to the indicated OD.
In Vitro Killing Assay-The in vitro killing assay was performed as described previously (8). In brief, bacteria grown at various growth phases were washed and resuspended in binding buffer (25 mM MES, pH 6.0, 25 mM NaCl) at a density of 1-3 ϫ 10 6 cfu/ml. The diluted bacterial suspension was exposed to 10 M RegIII␤ or 0.7 M polymyxin B at 37°C for 30 min. Bacteria were then plated on selective LB media. The recovered cfu were normalized for the original cfu of the inoculum, thus yielding the bacterial "survival" (in %). For the competitive in vitro killing assay, we used bacteria from the midlogarithmic growth phase. Preincubation steps were performed for 10 min at 37°C, as indicated.
Bacterial Binding Assay-Fifty microliter aliquots (1.0 ϫ 10 7 cfu) of S. typhimurium grown up to the logarithmic (or the stationary) growth phase were pelleted by centrifugation, washed with binding buffer (25 mM MES, pH 6.0, 25 mM NaCl), suspended in binding buffer, and incubated with 10 g of RegIII␤ (in 50 l of total volume) for 15 min at 37°C. The reaction mixtures were centrifuged at 6000 ϫ g for 5 min, and the resulting supernatant was isolated, mixed with SDS-PAGE sample buffer, and boiled for 5 min. The pellet was washed once with binding buffer and boiled for 5 min in SDS-PAGE sample buffer. The samples were subjected to SDS-PAGE and analyzed by Coomassie Brilliant Blue (CBB) staining.
Gel Filtration Chromatography-Gel filtration chromatography was carried out using a Superdex 75 (10/300) column, which was equilibrated in running buffer (25 mM MES, pH 6.0, 125 mM NaCl). One hundred and fifty micrograms of RegIII␤ and/or 3 mg of LPS were mixed, incubated for 30 min at 37°C in binding buffer, and applied to the column at flow rate of 0.5 ml/min. The elution profile was analyzed by monitoring the absorbance at 280 nm and by SDS-PAGE and Western blot analyses of the fractions using an affinity-purified anti-RegIII␤ (8) and an anti-LPS (Salmonella O antiserum Factor 5, Group B; Difco, Detroit, MI) antibodies.
Lipid A Binding Assay Using Polymyxin B-agarose-To generate "lipid A-coated" beads, 100 g of lipid A (Sigma) were preincubated with polymyxin B-agarose (Sigma). Twenty-five g of RegIII␤ was added to the resulting resin in a total volume of 0.5 ml of binding buffer. After rotation for 2 h at 4°C, the resin was washed five times with 0.5 ml of wash buffer (25 mM MES, pH 6.0, 50 mM NaCl). Bound protein was eluted by boiling the resin in SDS-PAGE sample buffer (50 mM Tris-HCl, pH 6.8, 2% (w/v) SDS, 0.1% (w/v) bromphenol blue, 10% (v/v) glycerol, 100 mM 2-mercaptoethanol) and then resolved by SDS-PAGE through 15% acrylamide gels and analyzed by CBB.
Peptidoglycan Binding Assay-Ten g of RegIII␤ or its pointmutated variants were added to 50 g of insoluble peptidogly-Bacterial Recognition and Bactericidal Activity of RegIII␤ OCTOBER 5, 2012 • VOLUME 287 • NUMBER 41 can from B. subtilis (Sigma) in a total volume of 200 l of binding buffer (25 mM MES, pH 6.0, 25 mM NaCl) and incubated for 2 h at 4°C on a rotational mixer (12 rpm). After the incubation, centrifugation at 6000 ϫ g for 5 min was carried out to sediment insoluble peptidoglycan. The supernatant was concentrated via filter centrifugation to reduce the volume. The pellet (insoluble peptidoglycan plus bound material) was washed twice in binding buffer, and the bound protein was eluted by boiling the peptidoglycan pellet in 2ϫ SDS-PAGE sample buffer (100 mM Tris-HCl, pH 6.8, 4% (w/v) SDS, 0.2% (w/v) bromphenol blue, 20% (v/v) glycerol, 200 mM 2-mercaptoethanol). The samples were subjected to SDS-PAGE through 15% SDS-acrylamide gels and analyzed by CBB staining. For inhibition experiments, the RegIII␤ (10 g) was preincubated for 30 min at 37°C with 100 g of soluble peptidoglycan (InvivoGen, San Diego, CA). After the preincubation, 50 g of insoluble peptidoglycan was added, and binding was analyzed as described above.
Statistical Analysis-Statistical analysis was performed using the Mann Whitney U test or unpaired Student's t test. p Ͻ 0.05 was considered to be statistically significant.

RESULTS
Growth Phase-dependent Bactericidal Effect of RegIII␤-We have shown that the C-type lectin RegIII␤ has bactericidal effects against certain Gram-positive and Gram-negative bacteria in vitro (8). Moreover, our findings have suggested that RegIII␤ can promote S. typhimurium gut infection by eliminating the competing gut microbiota.
To analyze the bactericidal mechanism of RegIII␤ in more detail, we have analyzed killing of S. typhimurium cells taken from distinct bacterial growth phases in vitro. Similar to our previous data (8), wild-type S. typhimurium (SL1344) grown up to the stationary phase (A 600 Ն 1.0) was resistant to RegIII␤ ( Fig. 1). In contrast, S. typhimurium cells from the early-(A 600 ϭ 0.25), the mid-(A 600 ϭ 0.5), or the late-logarithmic growth phase (A 600 ϭ 0.75) were susceptible to RegIII␤ (Fig. 1). These results establish that the RegIII␤-mediated bactericidal effect against S. typhimurium is dependent on the bacterial growth phase. Thus, S. typhimurium from the early-or the mid-logarithmic growth phase can be employed to study how RegIII␤ acts on Gram-negative bacteria.
The Ability of RegIII␤ to Bind Gram-negative Bacteria Correlates with Its Bactericidal Activity-Binding to bacterial surface structures (i.e. peptidoglycan) is essential for the antibacterial activity of RegIII lectins such as HIP/PAP (18). This correlation has not yet been established for RegIII␤. Therefore, we compared the binding ability of RegIII␤ to S. typhimurium grown to mid-logarithmic or to stationary growth phase. RegIII␤ was bound more efficiently to S. typhimurium from the mid-logarithmic growth phase than to bacteria from the stationary phase ( Fig. 2, A and B). In contrast, BSA did not bind to S. typhimurium from either growth phase. The results indicate that the binding capacity of RegIII␤ to S. typhimurium correlates with its bactericidal activity.

Insoluble Peptidoglycan and Mannan Can Block the Bactericidal Effect of RegIII␤-Earlier work has demonstrated that
RegIII␤ can bind to peptidoglycan (18). This result also may imply that preincubation with peptidoglycan should inhibit the bactericidal effect of RegIII␤. To test this hypothesis, RegIII␤ was preincubated with insoluble peptidoglycan (10 min, 37°C; "Experimental Procedures") before adding it to mid-logarithmic phase S. typhimurium. Preincubation with insoluble peptidoglycan from B. subtilis (supplemental Fig. 1A) completely abolished the RegIII␤-mediated bactericidal effect (Fig. 3). A similar inhibitory effect was observed upon preincubation with mannan (supplemental Fig. 1B), another known substrate for RegIII␤ binding, whereas dextran (no binding to RegIII␤) (supplemental Fig. 1C) had no the inhibitory effect (Fig. 3). Thus, preincubation with known substrates for RegIII␤ can inhibit its bactericidal activity against Gram-negative Salmonella spp.,  Binding of RegIII␤ to S. typhimurium from the middle logarithmic or from the stationary growth phase. A, bacteria were grown up to the middle logarithmic growth phase or the stationary growth phase (1.0 ϫ 10 7 cfu) and incubated with RegIII␤ (10 g) at 37°C for 15 min (T, total fraction). The RegIII␤ bound or unbound to bacteria was separated by centrifugation (P, pellet; S, supernatant) and then analyzed by SDS-PAGE and CBB staining. B, quantified protein levels (%) of the pellet fractions were determined by defining the relative intensity of the total fraction as 100%. The error bars represent the S.D. of the mean from three independent experiments. **, p Ͻ 0.01; ns, not significant (unpaired Student's t test).
presumably by inhibiting RegIII␤-binding to the bacterial surface.
The Effect of LPS on RegIII␤-mediated Bactericidal Activity-In Gram-negative bacteria such as S. typhimurium, the peptidoglycan is normally not accessible to proteins of Ͼ600 Da, as it is shielded by the outer membrane. The outer leaflet of the outer membrane is composed mostly of LPS, a large, complex glycolipid. We thus hypothesized that RegIII␤ may initially interact with LPS and that this may represent a critical step in its bactericidal effect on Gram-negative bacteria. To test this hypothesis, we investigated whether LPS preincubation may inhibit RegIII␤-mediated killing of S. typhimurium. Indeed, preincubation of RegIII␤ with LPS (10 min, 37°C; "Experimental Procedures") completely abolished the bactericidal activity of RegIII␤ (Fig. 4A).
Formally, the total RegIII␤ concentration was ϳ3 times lower than that of the LPS in this assay. This might suggest that one LPS molecule might bind more than one RegIII␤ molecule. However, an alternative explanation is equally plausible. The bacterial binding assay (Fig. 2) indicated that, at least in some preparations, only one-third of the RegIII␤ protein does bind to the LPS (and/or other bacterial surface structures). The "inactivity" of the remaining two-thirds of the protein might be explained by incomplete refolding and/or by inhibitory bacterial molecules (LPS, murein, other inhibitors) which were carried over during the purification procedure, thus reducing the active RegIII␤ concentration. Thus, the LPS and the active fraction of the RegIII␤ might be present at approximately equimolar ratios in the experiment shown in Fig. 4A. Therefore, we cannot draw any firm conclusions on the stoichiometry of the RegIII␤-LPS complex from these data.
Polymyxin B was used as a control. This cationic antimicrobial antibiotic is well known to bind LPS (21,22), to disrupt the bacterial membrane, and thus to kill Gram-negative bacteria. Similar to RegIII␤, the bactericidal effect of polymyxin B was inhibited by LPS preincubation (Fig. 4A). Furthermore, we wanted to exclude that micelle formation affected LPS-mediated inhibition. Triton X-100 did not alter the effect of RegIII␤ on Salmonella viability and did not affect the inhibition of RegIII␤ by LPS (supplemental Fig. 2). These results suggest that RegIII␤ may indeed bind to LPS and that this interaction may be critical for the bactericidal effect of RegIII␤.
Lipid A Is Responsible for Counteracting the RegIII␤-mediated Bactericidal Effect-LPS consists of three structural domains, the O-saccharide antigen, the core oligosaccharide, and the lipid A (Fig. 4B). To identify the structural domain required for inhibiting RegIII␤-mediated bacterial killing, we performed a series of inhibition experiments using truncated versions of LPS. Rough LPS Ra lacks the O-specific polysaccharide (OPS, O-antigen), the Re core lacks the OPS and most residues of the core oligosaccharide, and lipid A lacks the OPS and the entire core oligosaccharide (Fig. 4B). Preincubation with LPS Ra and Re inhibited RegIII␤-mediated killing (Fig.  4C). Control experiments were performed with polymyxin B. Similar to RegIII␤, polymyxin B-mediated killing was inhibited by preincubation with LPS Ra or LPS Re (Fig. 4C). These results suggest that neither the OPS nor an intact core oligosaccharide are necessary for inhibiting the bactericidal activity of RegIII␤.
To this point, it had remained unclear whether the two Kdo residues remaining on the LPS Re or the lipid A moiety itself might be involved. Therefore, we performed additional inhibition experiments. Preincubation with lipid A, but not with Kdo, inhibited the RegIII␤-mediated bactericidal activity (Fig. 4D). We also tested the effect of dimethyl sulfoxide (DMSO) as DMSO served as the solvent for lipid A. However, DMSO had no effect on the bactericidal activity (data not shown). These results indicate that lipid A is the LPS moiety inhibiting the RegIII␤-mediated bactericidal activity and imply that RegIII␤ may bind directly to lipid A.
Direct Interaction of RegIII␤ with LPS-To further substantiate the direct interaction of RegIII␤ with LPS, we employed a gel filtration chromatography assay. In this assay, LPS and/or RegIII␤ were preincubated for 30 min in binding buffer and run on a Superdex 75 (10/300) gel filtration column, and the elution profile was analyzed by photometry (A 280 ) and Western blot of the eluted fractions using an anti-RegIII␤ antibody. In the absence of LPS, RegIII␤ eluted at fraction numbers 18 -24, suggesting a molecular mass of ϳ24 kDa (Fig. 5, A and B). This "apparent" molecular mass was slightly higher than the molecular mass of RegIII␤ (16 kDa). Similar observations have been reported for HIP/PAP (23). Overall, these observations indicated that RegIII␤ mainly exists in a monomeric state.
In the absence of RegIII␤, LPS yielded three peaks as observed by photometry. However, only the first peak harbored LPS as indicated by Western blotting (fractions 12-16; Fig. 5, C and D). It remains unclear whether the peaks observed around fractions number 21 and 27 might be attributable to a protein contamination in the LPS purchased from Sigma, a crude LPS extracted via the chloroform method (see also below). Nevertheless, the co-incubation of RegIII␤ with LPS resulted in a shift of the RegIII␤ peak (from fractions 18 -24) into fractions 12-16, as indicated by photometry and by Western blotting  OCTOBER 5, 2012 • VOLUME 287 • NUMBER 41 ( Fig. 5, E and F). These results strongly suggest that RegIII␤ binds directly to LPS.

Bacterial Recognition and Bactericidal Activity of RegIII␤
To further substantiate this and to exclude effects from unknown contaminants, we repeated this assay using pure LPS prepared as described in Ref. 24. As expected, preincubation with the pure LPS completely inhibited the bactericidal activity of RegIII␤ (data not shown). In the gel filtration chromatography assay, LPS from this preparation yielded virtually no peaks around fractions 21 or 27 (data not shown). However, the pure LPS retained the capacity to shift RegIII␤ into fractions 12-16. This lends further support to the notion that RegIII␤ may bind directly to LPS.
Preincubation with Gentiobiose, a Disaccharide Composed of Two ␤-(136)-Linked Units of nD-Glucose, Inhibits the RegIII␤mediated Bactericidal Effect-Published data and our new findings have established that peptidoglycan and lipid A may represent target structures for RegIII␤. Both target molecules include a disaccharide composed of modified glucosamines connected by ␤-(134)or ␤-(136)-linkages, respectively (supplemental Fig. 3, A and B). Therefore, we hypothesized that RegIII␤ might recognize these disaccharide moieties. If so, preincubation with disaccharides resembling those found in peptidoglycan or lipid A should inhibit RegIII␤-mediated killing.
Chitobiose is a dimer composed of two units of N-acetylglucosamine and resembles the carbohydrate structure of peptidoglycan (supplemental Fig. 3A). Preincubation with chitobiose had no effect on the RegIII␤-mediated bactericidal effect (Fig. 6A). The same held true for chitotriose, a trisaccharide composed of ␤-(134)-linked N-acetylglucosamine (supplemental Fig. 3A). Similarly, neither chitobiose nor chitotriose inhibited the polymyxin B-mediated bacterial killing activity (Fig. 6A). Thus, the carbohydrate moiety of peptidoglycan alone seems insufficient for binding by RegIII␤.
Gentiobiose is a disaccharide composed of ␤-(136)-linked D-glucose residues and resembles the carbohydrate structure of lipid A (supplemental Fig. 3B). Preincubation with gentiobiose blocked the RegIII␤-mediated bactericidal activity (Fig. 6B). In contrast, gentiobiose had no effect on the polymyxin B-mediated killing, indicating that the inhibitory effect is specific for RegIII␤. Collectively, these results suggest that the carbohydrate moiety of lipid A represents a key determinant sufficient for targeting by RegIII␤ and suggested that this interaction is critical for the bactericidal effect of this antimicrobial protein.
The EPN and DPT Motifs of Loop 1 Are Not Required for Binding to Peptidoglycan, Lipid A, or for the Bactericidal Activity of RegIII␤-In the human protein HIP/PAP, the canonical EPN motif (in particular amino acid Glu-114) was shown to be of critical importance for binding to peptidoglycan and for the bactericidal activity against Gram-positive bacteria (18). This EPN motif is also conserved in RegIII␤ and RegIII␥ ( Fig. 7A and supplemental Fig. 4). However, it has remained unclear whether EPN contributes to the biological activity of RegIII␤.
To address this issue, we generated an E114Q point-mutated RegIII␤ protein and analyzed possible functional defects. First, we performed the binding assay to insoluble peptidoglycan (see "Experimental Procedures"). (Note that some peptidoglycanassociated proteolytic activity facilitates N-terminal cleavage of RegIII␤ (6,25); this results in a slightly decreased molecular weight but does not affect target binding.) As shown in Fig. 7B, WT RegIII␤ bound efficiently to the insoluble peptidoglycan as it partitioned (almost exclusively) into the pellet fraction. If preincubated with soluble peptidoglycan, a significant fraction of RegIII␤ was retained in the supernatant fraction (Fig. 7B). This validated the assay and confirmed the binding specificity of RegIII␤ to peptidoglycan. Surprisingly, the E114Q variant of RegIII␤ bound with equivalent efficiency as the WT protein (Fig. 7B).
In a second assay, we analyzed the binding of the E114Q variant to lipid A. For this purpose, we coated polymyxin B-agarose beads with lipid A and performed a pulldown assay (see "Experimental Procedures"). Also in this assay, the E114Q variant of RegIII␤ bound with equivalent efficiency as the WT protein (Fig. 7C).
In a third approach, we tested the bactericidal effect of RegIII␤ E114Q. In contrast to the mutant phenotype reported for HIP/PAP (HIP/PAP-E114Q), RegIII␤ E114Q retained the bactericidal activity of the WT protein (Fig. 7D). These results indicate that the EPN motif is not required for either pepti- doglycan or lipid A recognition and for the killing of Gramnegative bacteria by RegIII␤.
Recently, a second motif (DPQ) was discovered in the loop 1 region of RegIV, another human Reg family lectin (26). RegIII␤ displays a similar sequence (DPT) in loop 1 (Fig. 7A and supplemental Fig. 4), suggesting that this motif might play a role in the interaction with bacterial target structures or the antimicrobial activity. To analyze the role of the DPT sequence, we generated RegIII␤ D108N. However, RegIII␤ D108N did show wild type activity in all three functional assays (Fig. 7, B-D). These results indicate that neither the EPN nor the DPT motif were required for substrate recognition or the bactericidal activity of RegIII␤. (Fig. 7A and supplemental Fig. 4). The similarity to the EPN motif (i.e. loop 1 of HIP/PAP and RegIII␤) prompted us to explore whether the ERN might contribute to substrate recognition or the bactericidal activity of RegIII␤. For this purpose, we generated three variants of RegIII␤, i.e. E134Q, R135T, and N136A. In binding assays to peptidoglycan or lipid A, RegIII␤ R135T was attenuated, whereas RegIII␤ E134Q and RegIII␤ N136A displayed wild-type activity (Fig. 8, A and B). These results indicate that Arg-135 contributes significantly to target binding by RegIII␤. displayed homologs has an EPN and a DPT motif, indicated by magenta or green boxes, respectively. Loop 2 of RegIII␤ and HIP/PAP has a putative ERN motif, whereas RegIII␥ has a putative motif ETN, indicated by a blue box. Amino acid residues examined by targeted mutagenesis in this study are indicated in red. B, peptidoglycan binding assays of the loop 1 mutants of RegIII␤. Ten g of wild-type RegIII␤ (WT) or point mutated RegIII␤ E114Q or D108N proteins were added to 50 g of insoluble peptidoglycan (iPGN) from B. subtilis. After incubation, the mixture was pelleted by centrifugation. The resulting pellet (P) and supernatant (S) fractions were analyzed by SDS-PAGE and CBB staining. For binding competition, 100 g of soluble peptidoglycan (sPGN) from E. coli was preincubated with RegIII␤. Percent of band intensity of the pellet fraction was calculated using NIH ImageJ software. C, lipid A binding assays of the loop 1 mutants of RegIII␤. Purified wild-type RegIII␤ (wt), or point mutated RegIII␤ E114Q or D108N proteins were incubated with lipid A-immobilized polymyxin B-agarose. After washing the agarose, the bound RegIII␤ proteins were eluted by boiling in SDS-PAGE sample buffer and then subjected to SDS-PAGE and analyzed by CBB staining. The band intensity was calculated using ImageJ software by defining the relative intensity of wild-type RegIII␤ bound to the lipid A-immobilized agarose as "1". D, in vitro killing assay of the loop 1 mutants of RegIII␤. S. typhimurium wild-type strain (SL1344) from the middle logarithmic growth phase (A 600 ϭ 0.5) was incubated with purified recombinant RegIII␤ WT or E114Q or D108N (10 M) for 30 min at 37°C, and bacterial survival was quantified by dilution plating. Assays were done at least in five independent experiments. The bar shows the median. **, p Ͻ 0.01; ns, not significant (Mann-Whitney U test). OCTOBER 5, 2012 • VOLUME 287 • NUMBER 41

JOURNAL OF BIOLOGICAL CHEMISTRY 34851
RegIII␤ Mutants E134Q and N136A Have a Reduced Bactericidal Activity-To clarify the role of the ERN motif in the bactericidal activity of RegIII␤, we performed bacterial killing assays with RegIII␤ E134Q, R135T, or N136A. The R135T mutant had bactericidal activity similar to wild-type RegIII␤ protein (Fig. 8C). In contrast, the bactericidal activity of RegIII␤ E134Q and N136A was significantly impaired (Fig. 8C). Thus, Glu-134 and Asn-136 seem crucial for the bactericidal effect of RegIII␤. Together with the data presented above, these results indicate that ERN represents a novel motif and implicates loop 2 in the bactericidal activity of RegIII␤.
Asp-142 of Loop 2 Is Required for Lipid A Binding and Bactericidal Activity-In several C-type lectins, substrate binding is facilitated via asparagine or glutamine residues and/or the COOH group of aspartic or glutamic acid (16,27,28). Loop 2 of RegIII␤ also features such a residue ( Fig. 7A and supplemental  Fig. 4). To address involvement of this Asp-142 amino acid residue, we generated RegIII␤ D142A. In binding assays, RegIII␤ D142A was impaired in binding to lipid A, but not to peptidoglycan (Fig. 8, A and B). Furthermore, RegIII␤ D142A displayed a reduced bactericidal activity (Fig. 8C). These results indicate that the residue Asp-142 of loop 2 contributes to substrate recognition and bactericidal activity of RegIII␤.
Salt Concentrations of Ն75 mM NaCl Alleviate Binding and Antimicrobial Activity of RegIII␤-Our results from the binding assays suggest that polar, hydrogen, or ionic interactions might contribute to the binding of RegIII␤ to peptidoglycan and to LPS. These bonds should be sensitive to the salt concentration of the buffer. To test this hypothesis, we performed the in vitro killing assay using the binding buffers with increasing NaCl concentrations. In the presence of low NaCl concentrations (0, 25, 50 mM), RegIII␤ had bactericidal activity (Fig. 9A). After incubation, the mixture was pelleted by centrifugation. The resulting pellet (P) and supernatant (S) fractions were analyzed by SDS-PAGE and CBB staining. Percent of band intensity of the pellet fraction was calculated using ImageJ software. B, lipid A binding assays of the loop 2 mutants of RegIII␤. Purified wild-type RegIII␤ or point mutated RegIII␤ E134Q, R135T, N136A, or D142A were incubated with lipid A-immobilized polymyxin B-agarose. After washing the agarose, the bound RegIII␤ proteins were eluted by boiling in SDS-PAGE sample buffer and then subjected to SDS-PAGE and analyzed by CBB staining. The band intensity was calculated using ImageJ software. The relative intensity of wild-type RegIII␤ bound to the lipid A-immobilized agarose was defined as "1". C, in vitro killing assay of the Loop 2 mutants of RegIII␤. S. typhimurium wild-type strain (SL1344) from the middle logarithmic growth (A 600 ϭ 0.5) was incubated with purified recombinant RegIII␤ wildtype E134Q, R135T, N136A, or D142A (10 M) for 30 min at 37°C, and bacterial survival was quantified by dilution plating. Assays were performed in at least five independent experiments. The bar shows the median. **, p Ͻ 0.01; ns, not significant (Mann-Whitney U test). In contrast, no bactericidal activity was detected at 75 and 100 mM NaCl.
In a second approach, we compared the binding capacity of RegIII␤ to mid-log phase S. typhimurium in the presence of low (25 mM) or high (100 mM) concentrations of NaCl. In the presence of 100 mM NaCl, RegIII␤ bound with reduced efficiency (Fig. 9B; compare proteins in the pellet fractions "P"). Together, these results indicate that the ionic strength of the medium has a significant impact on the biological activity of RegIII␤.

DISCUSSION
We have previously shown that RegIII␤ can contribute to Salmonella-induced colitis by eliminating competing gut microbiota (8). This was attributed to the susceptibility of many commensals to RegIII␤-mediated killing, whereas S. typhimurium was found to be resistant in vitro (8). However, our results shown here reveal that S. typhimurium is not resistant per se. Instead, resistance depends on the bacterial growth phase: S. typhimurium taken from early-to mid-logarithmic growth phase are killed by RegIII␤ (this work), whereas bacteria from the stationary growth phase are resistant (as shown previously in Ref. 8). The susceptibility of S. typhimurium from the logarithmic growth phase correlated with enhanced binding to RegIII␤, suggesting that killing might be attributable to the mode and/or the efficiency of RegIII␤ binding. Here, we have found that LPS is (besides peptidoglycan) a key target of RegIII␤ in the Gram-negative bacterial cell envelope, and we determined the molecular domains of LPS and of RegIII␤, which are required for binding and for the antimicrobial activity against mid-log phase S. typhimurium.
Historically, RegIII␤ was found to bind peptidoglycan (18). However, in the context of the Gram-negative cell envelope, peptidoglycan should be shielded from antimicrobial peptides present in the surrounding environment. Therefore, we have hypothesized that RegIII␤ may also bind to LPS, as this complex lipid features a large number of carbohydrate residues and makes up the bulk of the outer leaflet of the outer membrane. Indeed, we found that RegIII␤ directly binds to the lipid A moiety of LPS and that this is an important step in the bactericidal activity against S. typhimurium. Most likely, the carbohydrate moiety of lipid A represents the main structural feature recognized by RegIII␤, as gentiobiose (but not chitobiose or chitotriose), which mimics the carbohydrate backbone of lipid A, can protect S. typhimurium from RegIII␤-mediated killing.
Is there a functional link between the binding of RegIII␤ to LPS and to peptidoglycan? Despite all progress, we still cannot answer this important question. One may consider at least three alternative scenarios: (i) "LPS-mediated access to peptidoglycan". RegIII␤ binding to LPS may compromise the outer membrane integrity. This facilitates access of additional RegIII␤ molecules to the peptidoglycan in a second step. (ii) "Parallel access". RegIII␤ might (at least initially) access the periplasm through pre-formed (growth-related) gaps in the outer membrane. Thus, RegIII␤ may target LPS and peptidoglycan at the same time. (iii) "Two-step-mechanism". Alternatively, the very same RegIII␤ molecules, which initially bind to the LPS, might go on to penetrate the outer membrane and access the periplasm where they bind to peptidoglycan. Deci-phering these steps will be an important topic for future research.
Similarly, the mechanism allowing RegIII␤ access to lipid A is not entirely clear. According to our findings of the growth phase-dependent killing effect of RegIII␤, it would be conceivable if lipid A (and also maybe peptidoglycan) was accessible to RegIII␤ at the septum or the growth zone, i.e. features which may expose lipid A to access by RegIII␤ and which exist only in growing bacteria, but not in stationary phase cultures. Furthermore, the OPS production by S. typhimurium has been shown to depend on the bacterial growth phase, i.e. S. typhimurium grown up to the stationary phase has higher expression of OPS than bacteria from the logarithmic growth phase (29). Our previous data also have shown that the OPS-deficient S. typhimurium is more susceptible to RegIII␤ than the wild-type strain (8). It is conceivable that the rigid LPS layer limits the access of RegIII␤ to lipid A, especially in S. typhimurium grown up to the stationary phase. Revealing the mechanism underlying the growth phase-dependent bactericidal activity of RegIII␤ will be an important topic for future work.
RegIII␤ seems to employ different moieties for target binding than other RegIII family proteins. Lectins of the RegIII family (except for RegIII␣) display an EPN motif in loop 1. In the case of HIP/PAP, this motif was shown to be required for carbohydrate/peptidoglycan recognition (18). In contrast, the EPN motif in loop 1 of RegIII␤ is dispensable for peptidoglycan recognition and for its bactericidal activity. Instead, our results indicate that the carbohydrate recognition and the bactericidal activity (as well as lipid A binding) of RegIII␤ rely on a novel motif in loop 2, the ERN tripeptide, and residue Asp-142. This may resemble to some extent the carbohydrate recognition by mannose-binding lectin and asialoglycoprotein receptor. In these C-type lectins, EPN or QPD motifs of the long loop 2 mediate the recognition specificity to GlcNAc and mannose or Gal, respectively (27). The EPN and the QPD motif act as hydrogen bond donors (asparagine and glutamine) or acceptors (glutamic acid and aspartic acid) during carbohydrate recognition (16). In conclusion, the mode of carbohydrate recognition and bactericidal activity of RegIII␤ differs from that of HIP/ PAP and RegIII␥. It is tempting to speculate that these differences may explain the distinct antimicrobial spectra of the different RegIII family members.
Which target structures are recognized by the RegIII lectins? Peptidoglycan-binding was observed for HIP/PAP, RegIII␥, and RegIII␤, but not for RegIII␣ (6,18 and in this study). This was a first indication that substrate specificities do differ between the different RegIII family members. If binding is observed, the carbohydrate backbone of the peptidoglycan seems to represent the most important target structure. This is indicated by affinity assays and the inhibition of bacterial killing using chitin, a ␤-(134)-linked N-acetyl-glucosamine polymer (6,18). However, the individual RegIII proteins seem to differ in their requirement for target oligomerization. The bactericidal activity of HIP/PAP is inhibited by chitobiose (a ␤-(134)linked N-acetyl-glucosamine dimer) (6), whereas RegIII␥-mediated killing effect is blocked by chitotetraose, but not by chitobiose (6). Similarly, we found that the bactericidal activity of RegIII␤ is not inhibited by chitobiose or chitotriose.
Besides the carbohydrate backbone, at least some of the RegIII family proteins also seem to recognize the linker peptides. Weak binding affinities of L-Ala-D-Glu-mDAP and L-Ala-D-Glu-L-Lys to HIP/PAP have been demonstrated by NMR (18). This binding does not seem to involve loop 1 (i.e. Glu-114, Glu-118), suggesting that different moieties of HIP/PAP are responsible for binding to the carbohydrate and the peptide moieties of peptidoglycan (18). Currently, we do not know whether RegIII␤ binding may also involve an additional epitope such as the linker peptides of the peptidoglycan. Such a requirement is known for peptidoglycan recognition proteins, which recognize a muramyl-tripeptide (30 -32). Overall, the current data support the notion that different antimicrobial lectins of the RegIII family do differ in their target preference and in their affinities toward monomeric or oligomeric peptidoglycan units.
In addition to peptidoglycans, RegIII␤ can also bind to LPS. Again, the carbohydrate moiety seems to be the key binding determinant of this target molecule as demonstrated via gentiobiose inhibition assays. In this case, a carbohydrate dimer was sufficient for detectable binding. Future work will have to determine whether/how the carbohydrate binding site of RegIII␤ may allow discriminating between different carbohydrate moieties. The second residue (Arg-135) of the ERN motif affected both, binding to lipid A and to peptidoglycan and thus might constitute part of a "common" recognition mechanism.
The bactericidal activity of RegIII␤ is quite sensitive to NaCl concentrations of Ͼ50 mM. So far, we do not know whether this effect is attributable to the osmolarity or to the shielding of ionic/polar interactions. In the gut lumen, the total osmolarity is thought to range around the equivalent of 300 mM NaCl (33). However, in this environment, NaCl is only one of many components contributing to the total osmolarity. Based on these observations, it is tempting to speculate that salt concentration or osmolarity might represent a mechanism for controlling RegIII␤ activity in the gut. This will be an important topic for future work.
In summary, we propose a novel bacterial target for RegIII␤, identify a novel functional motif involved in target recognition, and propose a mode of action that differs in important aspects from that of the other well characterized members of the RegIII family. Specifically, RegIII␤ was found to bind LPS by recognizing the carbohydrate moiety of lipid A via ERN and Asp-142 of loop 2. Our results provide novel insights into the carbohydrate recognition specificity and bactericidal mechanism of the RegIII lectin family. This will enable detailed structural studies aiming to clarify the bonding patterns explaining carbohydrate recognition by RegIII␤. Such information may empower systematic approaches for developing new antimicrobial agents.