Distinct carbohydrate recognition domains of an invertebrate defense molecule recognize Gram-negative and Gram-positive bacteria.

Coelomic fluid of Eisenia foetida earthworms (Oligochaeta, Annelida) contains a 42-kDa defense molecule named CCF for coelomic cytolytic factor. By binding microbial antigens, namely the O-antigen of lipopolysaccharide (LPS), beta-1,3-glucans, or N,N'-diacetylchitobiose present, respectively, on Gram-negative bacteria or yeast cell walls, CCF triggers the prophenoloxidase activating pathway. We report that CCF recognizes lysozyme-predigested Gram-positive bacteria or the peptidoglycan constituent muramyl dipeptide as well as muramic acid. To identify the pattern recognition domains of CCF, deletion mutants were tested for their ability to reconstitute the prophenoloxidase cascade in E. foetida coelomic fluid depleted of endogenous CCF in the presence of LPS, beta-1,3-glucans, N,N'-diacetylchitobiose, and muramic acid. In addition, affinity chromatography of CCF peptides was performed on immobilized beta-1,3-glucans or N,N'-diacetylchitobiose. We found that the broad specificity of CCF for pathogen-associated molecular patterns results from the presence of two distinct pattern recognition domains. One domain, which shows homology with the polysaccharide and glucanase motifs of beta-1,3-glucanases and invertebrate defense molecules located in the central part of the CCF polypeptide chain, interacts with LPS and beta-1,3-glucans. The C-terminal tryptophan-rich domain mediates interactions of CCF with N,N'-diacetylchitobiose and muramic acid. These data provide evidence for the presence of spatially distinct carbohydrate recognition domains within this invertebrate defense molecule.

Invertebrate innate defense strategies are based on pattern recognition receptors that do not discriminate between individ-ual antigens but recognize surface determinants common to potential pathogens (1). Pattern recognition receptors interacting with different saccharide moieties of microbial lipopolysaccharides (LPS), 1 peptidoglycans, and ␤-1,3-glucans were described in numerous invertebrate and vertebrate species. We have characterized a 42-kDa lectin named coelomic cytolytic factor 1 (CCF) from the coelomic fluid of Eisenia foetida earthworms (Oligochaeta, Annelida (2,3)). CCF displays significant amino acid sequence homology with bacterial and animal ␤-1,3glucanases but does not exhibit such enzymatic activity (3)(4)(5). More interestingly, CCF shows homology with the ␣ subunit of the ␤-1,3-glucan sensitive factor G from the horseshoe crab Tachyplesus tridentatus (6) and with the Gram-negative bacteria-binding proteins of various insects (7)(8)(9)(10). More recent reports concern the sequence homology of CCF with LPS and ␤-1,3-glucan recognition proteins from arthropods (11-13, for review see Ref. 14). All these invertebrate homologs have been suggested to play a role in invertebrate innate immunity acting as pattern recognition receptors. Accordingly, CCF binds with cell wall components of Gram-negative bacteria or yeast (the O-antigen of LPS, ␤-1,3-glucans, N,NЈ-diacetylchitobiose), triggering the activation of the prophenoloxidase (pro-PO) cascade (3,15). This pathway results in the formation of cytotoxic and antimicrobial compounds and thus represents an important defense mechanism in a variety of invertebrates (for review see Refs. 16 -18).
Pro-PO cascade is activated within 6 h in E. foetida coelomic fluid by Gram-negative but not Gram-positive bacteria (3). However, E. foetida coelomic fluid incubated for 6 h with Gramnegative or Gram-positive bacteria exhibits a similar antibacterial effect (19). Whether the failure of Gram-positive bacteria to induce the pro-PO cascade resulted from insufficient activation time is not clear.
In the present work, we examined whether Gram-positive bacteria trigger the pro-PO cascade in the earthworm coelomic fluid. We report that digestion of the Gram-positive bacteria or peptidoglycan by lysozyme-like activity makes it possible to induce the pro-PO cascade through the recognition of muramic acid by CCF. We also show that the broad specificity of CCF results from the presence of two distinct lectin-like domains within the molecule, a domain located in the central part of CCF implicated in interactions with LPS and ␤-1,3-glucans, and a C-terminal tryptophan-rich domain interacting with N,NЈ-diacetylchitobiose and muramic acid. Finally we observed that the CCF homolog in the E. foetida-related species Lumbricus terrestris displays distinct saccharide specificities and does not recognize N,NЈ-diacetylchitobiose and muramic acid. E. foetida CCF is, to our knowledge, the first invertebrate defense molecule acting as pattern recognition molecule for cell wall components of yeast and Gram-negative and Gram-positive bacteria.
RNA Preparations-Body pieces of adult E. foetida or L. terrestris earthworms kept for 2 days on filter paper soaked with isotonic phosphate-buffered saline in the presence of penicillin/streptomycin (100 units, 100 g/ml) were excised and placed in liquid nitrogen, and total RNAs were prepared using TRIzol (Sigma). All molecular biology procedures were performed according to the manufacturer's recommendations.
Cloning of L. terrestris CCF-An E. foetida CCF (ECCF) homolog was detected in L. terrestris by Northern blot using total ECCF cDNA as a probe; a 1.6-kb mRNA was observed. 2 L. terrestris CCF (LCCF) cDNA was cloned using 5Ј-rapid amplification of cDNA ends (RACE) and 3Ј-RACE (Life Technologies, Inc.) with generated specific primers (GSP, Table I). For 5Ј-RACE, GSP-1 was based on ECCF and glucanase motif homologous sequences (3,14). PCR products were gel-purified and sequenced (Thermo Sequenase radiolabeled terminator cycle sequencing kit, Amersham Pharmacia Biotech). The complete cDNA encoding LCCF was amplified (Titan One tube RT-PCR system, Roche Molecular Biochemicals) using RNA as template and primer pairs (LCCF , Table I) containing a BamHI or a SmaI site at the 5Ј-N-terminal or 3Ј-Cterminal ends, respectively.
E. foetida and L. terrestris CCF Vector Constructions and Bacterial Expressions-The cDNA encoding E. foetida (ECCF, E1-E5) or L. terrestris (LCCF, L2, L4) CCF peptides were amplified (Titan One tube RT-PCR system) using RNA as templates and primer pairs (Table I)   Defense Molecule Recognizes Gram-negative and -positive Bacteria containing a BamHI or a SmaI site at the 5Ј-N-terminal or 3Ј-Cterminal ends, respectively. After digestion, PCR products were ligated in QIAexpress PQE-30 vector containing a N-terminal 6ϫ His-affinity tag (Qiagen). The TOP10FЈ E. coli strain was transformed, grown in the presence of ampicillin (100 g/ml), and induced for 2 h at 37°C in the presence of isopropyl-1-thio-␤-D-galactopyranoside (2 mM). CCF peptides were purified and resuspended in TN buffer as described previously (3). LPS contamination was Ͻ15 pg/mg CCF peptides (QCL LAL test, Bio-Whittaker Europe).
Prophenoloxidase-activating Assay-The level of pro-PO cascade activation was assessed as described previously (3). Briefly, 50 l of the coelomic fluid (without or with 1 mM Pefabloc (serine proteinase inhibitor, Roche Molecular Biochemicals)), 25 l of 0.1 M Tris, pH 8, containing 50 mM Ca 2ϩ and 10 l of L-DOPA (3-(3,4-dihydroxylphenyl)-Lalanine, Fluka, final concentration 1.5 mM) was incubated at room temperature for 6 h in the absence or presence of tested compounds (1 g/ml). The oxidation of L-DOPA was measured at 492 nm and expressed as the difference between the values without and with Pefabloc.
In some experiments, M. lysodeikticus and E. coli suspensions or M. lysodeikticus peptidoglycan were preincubated with hen egg lysozyme or trypsin (1 g/mg bacteria or 1 g/mg peptidoglycan, up to 5 h), or with coelomic fluid proteins (1 mg/mg bacteria or 1 mg/mg peptidoglycan, up to 12 h) before testing the activation of the pro-PO cascade.
Data Analysis-Data (mean of triplicate Ϯ S.D.) are representative of three independent experiments performed. The validity of the results was assessed by Student's t test.

RESULTS
Pattern Recognition Specificity of E. foetida CCF-The coelomic fluid of E. foetida earthworms displays phenoloxidase (PO) activity within 6 h upon recognition of yeast or Gramnegative bacteria (3,15). With a similar experimental procedure, Gram-positive bacteria did not trigger the coelomic fluid PO activity. However, the coelomic fluid of E. foetida was described to display lysozyme-like activity, with 50 g of coelomic fluid proteins exhibiting an activity comparable with 1 g of hen egg lysozyme in M. lysodeikticus bacteriolytic assay (20, 21). Therefore, we tested whether PO activity could be triggered in E. foetida coelomic fluid by M. lysodeikticus pretreated with coelomic fluid or hen egg lysozyme. As shown in Fig. 1A, whereas intact M. lysodeikticus did not trigger detectable PO activity within 6 h of incubation in coelomic fluid, bacteria pretreated for 12 h did so with considerable efficiency. In addition, bacteria pretreated for 5 h with lysozyme, but not with trypsin (up to 12 h), activated the pro-PO cascade in E. foetida coelomic fluid. Similar results were observed with B. subtilis. 2 In contrast, Gram-negative E. coli suspension triggered similar levels of pro-PO activity whether or not predigested with lysozyme or coelomic fluid proteins (Fig. 1A). On the other hand, peptidoglycan purified from the M. lysodeikticus cell wall did not induce PO activity unless preincubated with hen egg lysozyme for 5 h or with coelomic fluid for 12 h (Fig. 1B). Because lysozyme hydrolyzes the ␤-1,4-glycosidic bond between N-acetylglucosamine and N-acetylmuramic acid of peptidoglycan, we tested the ability of synthetic peptidoglycan analogs (muramyl dipeptide, disaccharide dipeptide, muramic acid, pentapeptide, and N-acetylglucosamine) to trigger the pro-PO activation in E. foetida coelomic fluid. Muramyl residue-containing analogs and muramic acid efficiently activated the pro-PO cascade, whereas pentapeptide and N-acetylglucosamine did not (Fig. 2). The PO-inducing activity of Grampositive bacteria, or peptidoglycan digested with lysozyme or coelomic fluid proteins, 2 and of muramyl dipeptide, disaccharide dipeptide, and muramic acid (Fig. 2) was abolished in CCF-depleted coelomic fluid and reconstituted by adding recombinant CCF.
Together, these data indicate that E. foetida CCF displays a broad pattern recognition specificity for yeast as well as Gramnegative and Gram-positive bacteria.
Comparison of the Primary Structure of E. foetida and L. terrestris CCF Molecules-L. terrestris CCF-like molecule was cloned, and the amino acid sequence was compared with the sequence of E. foetida CCF (Fig. 4). The two molecules displayed 91% of homology, with identity in the putative polysaccharide-binding motif of bacterial glucanase (7,22) and in the putative catalytic site of bacterial endo-1,3(4)-␤-glucosidase (23). Both E. foetida and L. terrestris CCF molecules contained three conserved cysteine residues and a remarkably high number of tryptophan residues (7%), the majority of which are located in the C-terminal part of the molecule.
Peptide constructs covering the two putative functional regions of CCF, as well as the tryptophan-rich domain, were expressed in bacteria to identify the carbohydrate recognition

FIG. 4. Amino acid sequence alignment of E. foetida (GenBank TM accession number AF030028) and L. terrestris CCF (GenBank TM accession number AF395805). Amino acids are numbered from the initial methionine.
Only mutated amino acids in L. terrestris CCF (lt) as compared with E. foetida CCF (ef) are indicated. Signal peptide in L. terrestris CCF (underlined) was deduced from the known signal sequence of E. foetida CCF (3). Polysaccharide-binding motif (PsBM) and glucanase motif (GM) are boxed. Tryptophan residues are in boldface. domain(s) (Fig. 5). Peptide 1 consisted of the N-terminal part of the molecule, including the polysaccharide-binding and glucanase motifs. Peptide 2 starting from the polysaccharide-binding and glucanase motifs comprised the tryptophan-rich Cterminal part as well. Peptide 3 represented the N-terminal part of CCF but did not include any of the putative functional regions. Peptide 4 corresponded to the 78 amino acids encompassing the polysaccharide-binding and glucanase motifs. Peptide 5 represented the tryptophan-rich C-terminal part of CCF. The entire molecule and all the above-mentioned peptide constructs were derived from E. foetida CCF (ECCF, E1, E2, E3, E4, E5). In L. terrestris, the entire protein and peptides 2 and 4 (LCCF, L2, L4) were used in further experiments.
Binding of CCF and CCF-derived Peptides to ␤-1,3-Glucans and N,NЈ-Diacetylchitobiose-CCF and derived peptides from both E. foetida and L. terrestris were subjected to curdlan (␤-1,3-glucan) or N,NЈ-diacetylchitobiose binding assay. To exclude nonspecific interactions, protein samples were preincubated either with soluble laminarin or N,NЈ-diacetylchitobiose prior to the saccharide binding assay. As shown in Table II, all tested proteins, except for peptides E3 and E5, bound specifically to ␤-1,3-glucans, showing the importance of polysaccharide-binding and/or glucanase motifs for interaction of CCF with ␤-1,3-glucans. On the other hand, only ECCF, E2, and E5, i.e. peptides comprising the C-terminal region of CCF, recognized N,NЈ-diacetylchitobiose. None of the L. terrestris CCF constructs interacted with N,NЈ-diacetylchitobiose.
Identification of CCF Pattern Recognition Domains Involved in Prophenoloxidase Activation-The binding experiments indicated that distinct domains of CCF recognized different saccharide moieties. This assumption was tested in the pro-PO activation test.
E. foetida coelomic fluid, but not CCF-depleted coelomic fluid, exhibited PO activity upon activation with LPS, laminarin, N,NЈ-diacetylchitobiose, and muramic acid (Table III). In CCF-depleted coelomic fluid, the PO activity was completely reconstituted in the presence of the four triggering agents by adding ECCF or E2. Peptides E1 and E4 containing polysaccharide-binding and glucanase motifs reconstituted the pro-PO cascade only in the presence of LPS and laminarin, whereas E5 covering the C-terminal part of ECCF restored the PO activity triggered by N,NЈ-diacetylchitobiose and muramic acid. The CCF N-terminal part E3 did not restore the PO activity assessed by any of the triggering agents in CCF-depleted coelomic fluid.
As mentioned above (Fig. 3), the pro-PO cascade was activated in the coelomic fluid of L. terrestris by LPS and laminarin but not by N,NЈ-diacetylchitobiose or muramic acid (Table IV). LCCF, L2, and L4 reconstituted the sensitivity to LPS and laminarin in L. terrestris CCF-depleted coelomic fluid. Interestingly, ECCF, E2, and E5 conferred the ability to activate the pro-PO cascade in L. terrestris CCF-depleted coelomic fluid by N,NЈ-diacetylchitobiose and muramic acid.
Collectively, these data suggest that CCF consists of two distinct carbohydrate recognition domains. The first (amino acids 149 -227), surrounding the polysaccharide and glucanase motifs, mediates interaction of CCF with ␤-1,3-glucans and LPS. The second domain located in the C-terminal region CCF recognizes ␤-1,4-N-acetylglucosamine-linked saccharides and muramic acid in E. foetida CCF only. The reconstitution of PO activity in CCF-depleted L. terrestris coelomic fluid by E. foetida CCF peptides suggests that all components triggering the pro-PO cascade except the pattern recognition specificity of CCF are common to both earthworm species. DISCUSSION CCF was previously described to recognize cell wall components of yeast and Gram-negative bacteria that trigger the activation of pro-PO cascade in the coelomic fluid of E. foetida earthworm (3). Under similar experimental conditions, we did not detect activation of pro-PO with intact Gram-positive bacteria. Furthermore, it was reported that CCF agglutinated smooth Gram-negative bacteria, but not rough Gram-negative bacteria or Gram-positive bacteria (3). Here we show that preincubation of Gram-positive bacteria with lysozyme triggers the pro-PO cascade in E. foetida coelomic fluid. Lysozyme treatment does not affect the pro-PO activation by Gram-negative bacteria. The coelomic fluid of E. foetida earthworm exhibits a lysozyme-like activity mediated by a 13-kDa protein displaying considerable homology to other invertebrate lysozymes (20, 21). Accordingly, preincubation of Gram-positive bacteria with E. foetida coelomic fluid for 12 h resulted in pro-PO activation comparable with 5-h lysozyme pretreatment.
Trypsin pretreatment has no effect on the inability of Grampositive bacteria to induce pro-PO cascade suggesting that the ␤-1,4-N-acetylmuramidase activity of lysozyme rather than proteolysis allows the recognition of Gram-positive bacteria cell wall components by earthworm defense molecule(s). Similarly, peptidoglycan, but not the synthetic peptidoglycan analogs muramyl dipeptide, disaccharide dipeptide, or muramic acid, required lysozyme pretreatment to trigger L-DOPA oxidation in E. foetida coelomic fluid. Moreover, synthetic pentapeptide did not exhibit significant PO-inducing activity. Hence, it can be suggested that recognition of peptidoglycan saccharide moi-

TABLE II
Binding of CCF peptides to insoluble ␤-1,3-glucans or insolubilized N,NЈ-diacetychitobiose E. foetida or L. terrestris CCF and CCF peptides, used as such or preincubated with soluble laminarin or N,NЈ-diacetylchitobiose, were incubated with curdlan (insoluble ␤-1,3-glucan) or with agarose-insolubilized N,NЈ-diacetylchitobiose. Bound material was eluted, subjected to SDSpolyacrylamide gel electrophoresis, and Coomassie Blue-stained. The intensity of proteins eluted without or after preincubation with laminarin or N,NЈ-diacetylchitobiose were compared to determine the percent binding inhibition.   Defense Molecule Recognizes Gram-negative and -positive Bacteria eties rather than recognition of amino acid determinants by earthworm defense molecule(s) is required to elicit pro-PO activation. The PO-inducing activity of muramic acid, muramyl dipeptide, and disaccharide dipeptide was abolished in CCFdepleted coelomic fluid and was recovered by recombinant CCF. Hence, CCF plays a central role in the activation of the pro-PO cascade in E. foetida coelomic fluid, recognizing, besides yeast and Gram-negative bacteria cell wall components (3,15), Gram-positive bacteria cell wall components, at least muramic acid.
The coelomic fluid of the related earthworm species L. terrestris displays distinct biochemical properties as compared with E. foetida coelomic fluid reflecting most likely the antigenicity of the biotope where they live (soil versus compost). For example, proteolytic and hemolytic activities are considerably lower in L. terrestris than in E. foetida coelomic fluid (24,25). Moreover, in contrast to the E. foetida coelomic fluid (2,15), the coelomic fluid of L. terrestris does not display lytic activity against murine cell lines or African trypanosomes. 2 Furthermore, here we report that the pro-PO cascade in L. terrestris coelomic fluid can be triggered by ␤-1,3-glucans and LPS but not by peptidoglycan treated with lysozyme, muramic acid, or N,NЈ-diacetylchitobiose. These results indicate that the specificity and/or function of pattern recognition molecules in E. foetida and L. terrestris differ. The comparison of the amino acid sequence of CCF from E. foetida and L. terrestris revealed 91% homology with a complete identity in the putative polysaccharide-binding motif and in the catalytic region of bacterial and animal endo-1,3(4)-␤-endoglucosidases (7,22,23). These two regions are highly conserved in invertebrate pattern recognition molecules homologous to CCF that were reported to bind LPS or ␤-1,3-glucans (14). The remarkable number of tryptophan residues located mainly in the C-terminal part of CCF is of particular interest, because tryptophan residues were described to be important for the recognition of N-acetyl saccharides such as N,NЈ-diacetylchitobiose by tachylectin-2, a defense molecule from the horseshoe crab T. tridentatus (26). The crucial role of polysaccharide-binding and glucanase motifs in the recognition of LPS and ␤-1,3-glucans, and of the Cterminal part of CCF for the recognition of ␤-1,4-linked Nacetyl saccharides was confirmed in binding experiments and in the pro-PO activating assay using recombinant CCF-derived molecules from both E. foetida and L. terrestris. Based on these data, we propose the existence of distinct functional carbohydrate recognition domains within the CCF. The domain comprising amino acids 149 -227, present in both E. foetida and L. terrestris, binds LPS and ␤-1,3-glucans from Gram-negative bacteria and yeast, respectively. The C-terminal domain (amino acids 227-384), showing high divergence in E. foetida and L. terrestris CCF, displays specificity for the structurally related ␤-1,4-linked N-acetyl saccharides N,NЈ-diacetylchitobiose, muramyl dipeptide, and muramic acid. Thus, this tryptophan-rich domain allows the recognition of cell wall components of yeast or Gram-positive bacteria by E. foetida but not by L. terrestris CCF. Mutations in the C-terminal part may be responsible for the inability of L. terrestris CCF to bind muramic acid or N,NЈ-diacetylchitobiose. The existence of two distinct carbohydrate recognition domains may extend the protection of E. foetida against a variety of microorganisms or facilitate the interaction of immunocytes with the pathogen and subsequent induction of cellular defense. Indeed, one domain of CCF may interact with the pathogen, while the second domain may interact with a saccharide moiety on a putative immunocyte receptor (27)(28)(29)(30). Alternatively, the N-terminal part of CCF (amino acids 1-148) could allow the binding to a putative receptor mediating its interaction with immunocytes upon pathogen recognition.
Peptidoglycan recognition proteins have been described from insects to vertebrates, and the level of homology points to their common origin (31). Vertebrate peptidoglycan recognition proteins inhibit the growth of Gram-positive bacteria (32). In insects, peptidoglycan recognition proteins are often involved in activation of the pro-PO cascade (33,34). Despite its analogous function, CCF does not show primary structure homology with any described peptidoglycan recognition proteins. In addition, CCF displays no amino acid homology with lysozymes, Toll-like receptors, or vertebrate CD14. Although it recognizes LPS and ␤-1,3-glucans through the polysaccharide-binding and glucanase motifs as other invertebrate LPS-and ␤-1,3-glucanbinding proteins, CCF may recognize peptidoglycan by a different mechanism from the described peptidoglycan recognition proteins. The latter bind peptidoglycan in the absence of lysozyme treatment, whereas CCF requires lysozymelike activity to recognize peptidoglycan. CCF also differs from other peptidoglycan recognition proteins in the specificity for peptidoglycan constituents. The minimal peptidoglycan structure required to induce silkworm antibacterial defenses consists of two N-acetylglucosamine-N-acetylmuramic acid (muramyl-dipeptide) units with peptide side chains (35). Similarly, CD14 binds specifically to (muramyl-dipeptide) n as a minimal structure of peptidoglycan (36,37). CCF recognizes soluble muramyl-dipeptide initiating innate defense reactions against Gram-positive bacteria in E. foetida earthworm coelomic fluid.
In conclusion, we have provided evidence for the presence of two distinct functional domains in an invertebrate defense molecule, allowing the recognition of yeast and Gram-negative and Gram-positive bacteria.