MUSCLE was used to perform multiple sequence alignment (Edgar, 2004). Software MEGA 7.0 was used for phylogenetic analysis (Kumar et al., 2016). SignalP 4.1 (Petersen et al., 2011) was used to identify the potential signal peptide sequence.

Pseudoalteromonas aurantia DSM6057 was obtained from the DSMZ. The chi23 gene was amplified from the genomic DNA of P. aurantia DSM6057 via PCR using gene-specific primers (Table 1). The amplified fragment was ligated into the vector pET22b to construct the recombinant plasmid pET22b-chi23. Site-directed point mutants were created by the QuikChange mutagenesis method (Xia et al., 2014) with primers containing mutations (Table 1) and with plasmid pET22b-chi23 as the template. All the recombinant plasmids were verified by sequencing.

The wild-type Chi23 protein and all mutants were expressed in Escherichia coli BL21 (DE3) cells and induced by the addition of 1 mM isopropyl-β-D-thiogalactopyranoside at 20°C and 110 rpm for 20 h. Cells were collected and disrupted by pressure in a binding buffer (50 mM Tris-HCl, 100 mM NaCl, pH 8.0) containing 5 mM imidazole. The crude extract was loaded onto Ni-NTA agarose resin (Qiagen, United States), washed with 15 mM imidazole in the binding buffer, and eluted with 350 mM imidazole in the binding buffer. Recombinant His-tagged proteins were further purified by gel filtration chromatography on a Superdex-200 column (GE Healthcare, Sweden) equilibrated with 10 mM Tris-HCl buffer (pH 8.0) containing 100 mM NaCl. The eluted enzyme fractions were checked on SDS-PAGE (12.5%). Zymogram analysis was also performed to confirm the target protein. Chitinase activity of the purified protein was detected on gels by using the fluorescent substrate 4-methylumbelliferyl-β-D-N,N’-diacetylchitobiose [4-MU-(GlcNAc)₂] (Sigma, United States) (Barboza-Corona et al., 1999; Nagpure et al., 2014). Protein concentrations were determined by Pierce BCA Protein Assay Kit (Thermo Scientific, United States).

Colloidal chitin was prepared as previously described (Roberts and Selitrennikoff, 1988) with some modification. Briefly, 10 g of powdered chitin (Carbosynth, China) was dissolved in cold concentrated HCl (200 mL) and stirred at 180 rpm and 10°C for 2 h. 2.5 L of cold deionized water was added to the hydrolysate with continuous stirring for 10 min and the resultant chitin suspension was centrifugated at 9,000 rpm and 4°C for 10 min. The chitin pellets were washed repeatedly with cold deionized water until the pH of the filtrate was about 5.0, and then dissolved in deionized water to the concentration of 50 mg/mL.

Chitinase activity was determined by measuring the production of reducing sugars from colloidal or crystalline chitin. The standard reaction system (220 μL) contained 50 mM sodium acetate buffer (pH 5.0), 0.8–15 μM enzyme and 10 mg/mL colloidal chitin, which was replaced by 60 mg/mL crystalline chitin in a final volume of 1 mL when crystalline chitin was used as substrate. After incubation at 60°C for 15 min (for colloidal chitin substrate) or 20 min (for crystalline chitin substrate), the reaction was terminated by the addition of dinitrosalicylic acid (DNS) into the mixture. The amount of reducing sugars released into the mixture was determined with GlcNAc as the standard using the DNS method (Miller, 1959). One unit of enzyme activity is defined as the amount of enzyme required to release 1 μmol of reducing sugars per min.

Substrate specificity assays of Chi23 were carried out with crystalline chitin, colloidal chitin, glycol chitosan, chitosan and avicel. The optimum temperature (T_opt) for Chi23 and its mutants was measured at temperatures ranging from 0 to 90°C at pH 5.0. For thermostability assay, Chi23 was preincubated at temperatures ranging from 60°C to 80°C for different periods of time, and the residual activity was measured at 60°C. The optimum pH of Chi23 was determined at 60°C in the Britton-Robinson buffer ranging from pH 2.0 to 10.0. For pH stability assay, the enzyme was preincubated in buffers with a pH range of 2.0–13.0 at 0°C for 1 h, and then the residual activity was measured at pH 5.0 and 60°C. The effect of NaCl on Chi23 activity was determined at NaCl concentrations ranging from 0 to 4.0 M. For halotolerance assay, the enzyme was incubated at 0°C for 1 h in buffers containing NaCl ranging from 0 to 4.5 M before the residual activity was measured at 60°C. The effects of selected metal ions on Chi23 activity were examined at pH 5.0 and 60°C in a final concentration of 1 or 10 mM.

Enzyme kinetics assays were carried out in McIlvaine’s buffer (pH 5.0) using 4-MU-(GlcNAc)₂ at concentrations from 0.01 to 0.6 mM. Kinetic parameters were calculated by non-linear regression fit directly to the Michaelis–Menten equation using the Origin8 software. The overall secondary structures of wild-type Chi23 and its mutants were investigated at 25°C using a J-810 circular dichroism (CD) spectropolarimeter (Jasco, Japan). CD spectra were collected from 200 to 250 nm at a scanning rate of 200 nm/min with a path length of 0.1 cm. All proteins used were at a concentration of 9.8 μM in 10 mM Tris-HCl buffer (pH 8.0) containing 100 mM NaCl.

Differential scanning calorimetry (DSC) measurements were carried out over a temperature range of 10 to 110°C at a scanning rate of 1°C/min using a MicroCal VP-DSC microcalorimeter (GE Healthcare, Sweden). The sample solutions contained approximately 29.5 μM enzyme dissolved in Tris-HCl buffer (10 mM Tris-HCl, 100 mM NaCl, pH 8.0), and the Tris-HCl buffer was used as reference solution for accuracy. Prior to loading, all solutions were filtered using 0.22 μm pore size membrane filter and degassed. All data were analyzed using Origin8 software and further processed by fitting to a non-2-state model.

Thin layer chromatography (TLC) analysis was carried out as described by Huang and Chen (2005) with some modification. Reaction mixture (60 μL) containing 3.3 mg/mL of chitooligomers (chitobiose-chitohexaose) and 32.8 nM of Chi23 was incubated at 60°C for 12 h. After boiling the mixtures for 15 min, the reaction products were spotted onto a silica gel plate and developed with n-butyl alcohol-acetic acid-water-ammonia (10:5:5:1, v/v/v/v) for 3 h. To visualize the hydrolysis products, the developed plate was sprayed using a diphenylamine-aniline-phosphate reagent (0.8 g diphenylamine, 40 mL acetone, 0.8 mL aniline and 4 mL of 85% phosphoric acid) and then heated at 105°C for 10 min.

Time course of (GlcNAc)₆ hydrolysis by the wild-type Chi23 was also measured. Hydrolysis of (GlcNAc)₆ by Chi23 was carried out at pH 5.0 and 60°C for 5–120 min and the concentrations of the enzyme and substrate used were 32.8 nM and 6.2 mg/mL, respectively. The resultant hydrolysis products were analyzed by gel filtration chromatography on a Superdex Peptide 10/300 GL column (GE Healthcare, Sweden) at a flow rate of 0.4 mL/min using 0.2 M ammonium hydrogen carbonate as the running buffer. Elution was monitored at 210 nm using a UV detector. LabSolutions software was used to online monitoring and data analysis.

Wild-type Chi23 for crystallization was diluted to 5 mg/mL in 10 mM Tris-HCl (pH 8.0) containing 100 mM NaCl. Chi23 crystals grew at 18°C in the buffer containing 0.05 M KH₂PO₄ and 20% (w/v) PEG 8000. All the x-ray diffraction data were collected on the BL18U1 Beamline at Shanghai Synchrotron Radiation Facility using Area Detector Systems Corporation Quantum 315r. The initial diffraction data sets were processed by the HKL3000 program (Minor et al., 2006). The crystal structure of Chi23 was solved by molecular replacement using the PPL2 structure (PDB code 2GSJ) as the starting model. Subsequent refinement was performed using Coot (Emsley and Cowtan, 2004) and Phenix (Adams et al., 2002). All structure figures were generated using PyMOL software.

Schrödinger software¹ was used to conduct the Chi23 and (GlcNAc)₅ docking. The crystal structure of Chi23 was first optimized by Protein Preparation Wizard to optimize the structure and minimize the energy to make it more stable. Then Schrödinger LigPrep software (LigPrep, Schrödinger, LLC, New York, NY, United States, 2019) was used to preprocess (GlcNAc)₅ to obtain its low energy three dimensional conformers. Finally, (GlcNAc)₅ was docked into the binding site of minimized Chi23 using the Glide (Friesner et al., 2006) with the standard precision scoring mode. In molecular docking, the candidate with the lowest binding energy was chosen.

Chitin utilization experiment indicated that the marine bacterium P. aurantia DSM6057 (Gauthier and Breittmayer, 1979) could use colloidal chitin as the sole carbon source, suggesting that this strain produces chitinases for chitin degradation. We then sequenced the genome of this strain to find out its chitinases. Based on gene annotation, a gene encoding a putative chitinase was identified from the genome sequence, which was named chi23. The Chi23 protein comprises 280 amino acid residues with a calculated molecular mass of 30.4 kDa, and no signal peptide sequence can be predicted in Chi23 sequence by SignalP 4.1. Among the characterized enzymes, Chi23 is most closely related to the GH18 endochitinase PPL2 from Parkia platycephala seeds (Cavada et al., 2006) with a low sequence identity of 30%. By blasting against the Conserved Domains Database (CDD), Chi23 was found to be a single-domain enzyme, containing only a GH18 catalytic domain. Phylogenetic analysis showed that Chi23 belongs to the subfamily B of the GH18 family (Figure 1). Multiple sequence alignment suggested that the key catalytic residue of Chi23 is Glu117, which is located in the typical DxDxE motif (Matsumoto et al., 1999) of the GH18 chitinases (Figure 2).

Chi23 was over-expressed in E. coli BL21 (DE3) and purified. SDS-PAGE analysis showed that the purified Chi23 displays an apparent molecular mass of approximately 30 kDa, accordant to that predicted from its sequence (30.4 kDa) (Figure 3A). Only one fluorescent band with an apparent molecular mass of 30 kDa was observed in zymogram analysis, indicating that the purified Chi23 is an active chitinase (Figure 3A). Recombinant Chi23 could hydrolyze both crystalline and colloidal chitin, but showed little activity toward glycol chitosan, chitosan or avicel (Table 2). The optimal temperature for Chi23 activity was 60°C (Figure 3B). Chi23 was stable at 60°C, and retained over 40% of its maximal activity after 1 h incubation at 70°C (Figure 3C), suggesting that Chi23 is a thermostable enzyme. Chi23 had the highest activity at pH 5.0 and showed good tolerance over a wide pH range, retaining over 60% activity at pH 2.0–11.0 (Figures 3D,E). Similar to MmChi60 (Stefanidi and Vorgias, 2008), basic solutions with pH values in a range of 8.0–10.0 had small impact on Chi23 stability, but severely decreased or fully abolished the activity of Chi23, suggesting that the protein structure of Chi23 is possibly maintained under basic conditions. Chi23 activity could be stimulated by 3.0 M NaCl by 1.6 folds (Figure 3F), consistent with its marine origin. Moreover, after 1 h incubation in 4.5 M NaCl, Chi23 still retained 84% activity (Figure 3G). Thus, Chi23 is a thermostable, halotolerant, and acidic chitinase. Like other GH18 subfamily B chitinases (Wen et al., 2002; Stefanidi and Vorgias, 2008), Chi23 activity was hardly affected by K⁺, Li⁺, Ba²⁺, Ca²⁺, Mg²⁺ or Ni²⁺ at 1–10 mM concentrations, but severely inhibited by Cu²⁺ at 10 mM concentration (Table 3). Chi23 activity was also strongly inhibited by Co²⁺, Fe²⁺, and Mn²⁺ at 10 mM concentration, which, however, had no effect on the activity of MmChi60 (Stefanidi and Vorgias, 2008).

TLC analysis showed that (GlcNAc)₂ and (GlcNAc)₃ were the predominant products when crystalline chitin, colloidal chitin and chitooligomers (chitotriose-chitohexaose) were hydrolyzed by Chi23 (Figures 4A,B). Chi23 could not hydrolyze (GlcNAc)₂, and the products from (GlcNAc)₆ degradation were not all (GlcNAc)₂ (Figure 4B), suggesting that Chi23 functions as an endochitinase rather than an exochitinase. To reveal the action mode of Chi23, time course of (GlcNAc)₆ hydrolysis by Chi23 was performed. Degradation products from (GlcNAc)₆ by Chi23 were (GlcNAc)₂, (GlcNAc)₃ and (GlcNAc)₄ at first, and the resulting (GlcNAc)₄ was subsequently hydrolyzed into (GlcNAc)₂ (Figure 4C), which further indicated that Chi23 is an endochitinase.

To ascertain the structural basis of Chi23 for chitin degradation, the crystal structure of Chi23 was solved at 1.80-Å resolution by molecular replacement using the PPL2 structure (30% sequence identity with Chi23, PDB code 2GSJ) (Cavada et al., 2006) as the starting model. The crystallographic and refinement statistics are summarized in Table 4. The crystal of Chi23 belongs to the P12₁ space group with four molecules loosely packed per asymmetric unit. Gel filtration analysis showed that Chi23 presents as monomers in solution (Figure 5A). Consistent with sequence analysis, structural analysis also showed that Chi23 is a single-domain protein containing only a CaD, which is distinct from other endochitinases of the GH18 subfamily B that are all modular proteins (Payne et al., 2012; Malecki et al., 2013). Like the catalytic domains of the other chitinases from the GH18 subfamily B, the overall structure of Chi23 adopts the classical (β/α)₈ TIM-barrel fold (Figure 5B), most closely resembling the structures of PPL2 (Cavada et al., 2006) and ChiNCTU2 (PDB code 3N11) (Hsieh et al., 2010), with the root mean square deviations of 2.89 Å (202 monomer Cα atoms) to PPL2 and 3.48 Å (217 monomer Cα atoms) to ChiNCTU2. Chi23 comprises 14 α-helices and 10 β-sheets (Figure 5B). The catalytic ¹¹³DxDxE¹¹⁷ motif is located in the loop between β6 and α5, and a hydrogen bond is formed between the catalytic residues Asp115 and Glu117.

Although Chi23 has similar overall structure to the catalytic domains of the other chitinases from the GH18 subfamily B, some differences in their structures are also observed. All known bacterial chitinases from the GH18 subfamily B contain a small β-hairpin subdomain composed of two antiparallel β-sheets (β7 and β8 in ChiNCTU2) and an extra α-helix (α5 in ChiNCTU2) to extend their substrate-binding clefts (Hsieh et al., 2010; Payne et al., 2012; Malecki et al., 2013), which is absent from all eukaryotic chitinases of this subfamily (Terwisscha van Scheltinga et al., 1994; Cavada et al., 2006; Hurtado-Guerrero and van Aalten, 2007; Schuttelkopf et al., 2010; Masuda et al., 2015). For Chi23, it only has such an α-helix, and lacks the β-sheets for the formation of a β-hairpin subdomain (Figures 5C,D). Therefore, different from other bacterial chitinases from the GH18 subfamily B, Chi23 does not have a β-hairpin subdomain. All chitinases of the subfamily B, except for PPL2, harbor several long flexible loops in different regions. For example, three long loops between β4-α4, α7-α8, and α10-β10 are present in ChiNCTU2 to elongate its binding cleft (Hsieh et al., 2010). However, similar to PPL2 (Cavada et al., 2006), Chi23 only has extremely short loops in these corresponding regions (Figures 5C,D). Moreover, surface residues Phe30-Ser44 in Chi23 form two antiparallel β-sheets, which are conserved in nearly all eukaryotic chitinases of the GH18 subfamily B, but not in the bacterial chitinases of this subfamily where only one or no corresponding β-sheet is formed. Taken together, the structure of Chi23 is distinct from those of the other modular endochitinases from the GH18 subfamily B.

To investigate the substrate-binding mode of Chi23, we tried to obtain the structure of Chi23 binding a substrate. We cocrystallized the wild-type Chi23 and its inactive mutants D115A and E117A with different chitooligomers. Unfortunately, no electron density was observed for the substrate in the structure of Chi23 or its mutants. Instead, we modeled the structure of the wild-type Chi23 in complex with (GlcNAc)₅ by molecular docking. Several candidates were obtained and we chose the lowest energy candidate for the following analysis. In the modeled structure, (GlcNAc)₅ is bound in the negatively charged substrate cleft (Figure 6A).

Five sugar subsites from −3 to +2 are revealed in the substrate-binding cleft of the structure of Chi23 docked with (GlcNAc)₅, and the scissile glycosidic bond between the −1 and the +1 subsites is located close to the side chain of the predicted catalytic Glu117 (Figure 6B). The GlcNAc residue bound at the −1 subsite is in an unfavorable boat conformation, and stabilized by the side chains of Tyr189, Asn190 and Asp229 through hydrogen bonds and by Trp260 through hydrophobic stacking. The boat conformation of the sugar at the −1 subsite is critical for the initiation of the enzymatic cleavage (Brameld and Goddard, 1998). Except for Asp229, other residues involved in the binding of the -1 sugar are highly conserved in the GH18 subfamily B (Figure 2). All the acetamido groups of (GlcNAc)₅ are stabilized by hydrogen bonds. Residues Thr80, Trp260 and Gln261, Tyr189, Gln157 and Gly79 are hydrogen bonded to the C2-acetamido group of GlcNAc unit at subsites from −3 to +2, respectively, mainly through their side chains. The O3 hydroxyl groups of the −2, +1 and +2 sugar residues also form hydrogen bonds with Asn9, Glu117 and Gly79, respectively. There is also a hydrogen bond between the O6 hydroxyl group of the −3 sugar and the main-chain oxygen atom of Asn9.

To analyze the roles of the potential residues involved in substrate binding and catalysis in chitin degradation, single-point mutations of the residues to Ala were performed in Chi23, and the activities and kinetic parameters of the mutants were measured and compared to those of the wild-type enzyme (Figures 6C–E and Table 5). Mutations of residues Asp115 and Glu117 in the catalytic ¹¹³DxDxE¹¹⁷ motif led to complete loss of enzymatic activity toward insoluble and soluble chitin, demonstrating their key roles in the catalysis. Except for T80A at the −3 subsite and G79A at the +2 subsite, all mutations on the residues predicted to be involved in substrate binding at subsites from −2 to +1 severely decreased or even abolished the activity of Chi23 against crystalline chitin, colloidal chitin or 4-MU-(GlcNAc)₂, and reduced both the substrate affinity and k_cat of Chi23, indicating that residues Gln157, Tyr189, Asn190, Asp229, Trp260, Gln261 and Asn9 at subsites from −2 to +1 play important roles in Chi23 for the binding and degradation of insoluble and soluble chitin. Mutation T80A had little effect on the enzymatic activity or the k_cat/K_m value of Chi23, suggesting that Thr80 possibly contributes little to the binding of soluble and insoluble chitin. Mutation G79A significantly stimulated 4-MU-(GlcNAc)₂ hydrolysis of Chi23 but had a minor impact on crystalline and colloidal chitin hydrolysis, and similar cases were observed in Vibrio carchariae chitinase A (Suginta et al., 2007) and Bacillus circulans chitinase ChiA1 (Watanabe et al., 2003), suggesting that the binding and degradation of long-chain chitin tends to be influenced by a cluster of the surface-exposed residues in the substrate cleft rather than by a particular residue. In Bacillus circulans chitinase ChiA1, the increase in the hydrolytic activity against chitooligomers by mutations W164A and W285A at subsites +1 and +2, respectively, was perhaps attributed to the weakness of substrate inhibition (Watanabe et al., 2003). However, no inhibitory effect on the Chi23 activity was observed for 4-MU-(GlcNAc)₂ at the tested concentrations ranging from 0.01 to 0.6 mM, implying that other reason may explain its enhanced activity toward 4-MU-(GlcNAc)₂, which needs further study. Mutation G79A influenced only the k_cat of Chi23 in 4-MU-(GlcNAc)₂ degradation but not its K_m, consistent with the main chain of Gly79 involved in the substrate binding. CD spectroscopy analysis showed that the secondary structures of the mutants exhibited little deviation from that of the wild-type Chi23, indicating that the changes in the enzymatic activity and kinetic parameters of the mutants are caused by residue substitution rather than structural changes in Chi23 (Figure 6F).

In the GH18 family, several endochitinases active on crystalline chitin with specific activities ranging from 0.03 to 1.5 U/mg have been reported, including MmChi60, CjChi18C, SmChiC, SpChiC and SsChi18B, which are all modular enzymes containing one or more CBDs in addition to the CaD (Suzuki et al., 1999; Stefanidi and Vorgias, 2008; Purushotham et al., 2012; Monge et al., 2018; Sun et al., 2019). Compared to other GH18 endochitinases, Chi23 lacks a CBD, but has similar efficiency on crystalline chitin degradation (Table 2). To further reveal the mechanism for chitin degradation, especially for crystalline chitin, by Chi23, a comparative analysis of the structures of Chi23 and other GH18 endochitinases was carried out (Figure 7).

Similar to the catalytic domains of other GH18 endochitinases, Chi23 has a shallow groove-like binding cleft (Figure 7A). However, Chi23 displays differences in substrate binding involving residues Asn9, Asp229 and Gln261 compared to other GH18 endochitinases (Figures 7B–E). The side chain of Asp229 in Chi23 is hydrogen bonded to the O6 group of the -1 sugar to keep its boat conformation, which is replaced by a small hydrophobic Ala conserved in all other chitinases of the GH18 subfamily B (Figures 7B–E). However, mutant D229A of Chi23 showed significantly reduced activity to crystalline and colloidal chitin and a much lower activity to 4-MU-(GlcNAc)₂ (Figures 6C–E), demonstrating the importance of Asp229 in Chi23 for the hydrolysis of insoluble and soluble chitin. Gln261 in Chi23 stabilizes the acetamido group of the −2 sugar through its side chain, and mutation analysis indicates that Gln261 is important for both insoluble and soluble chitin degradation (Figures 6B–E). The counterpart of Gln261 of Chi23 is a small hydrophilic Ser that is conserved in most reported chitinases of the GH18 subfamily B (Figures 7B–E). No interaction is found between the corresponding Ser and bound chitooligomers in ChiNCTU2 (Hsieh et al., 2010) or MmChi60 (Malecki et al., 2013). Asn9 participates in the substrate binding of Chi23 through either its side chain or main chain. Asn9 of Chi23 is replaced by a conserved aromatic His in other bacterial chitinases or by a small conserved Gly in eukaryotic chitinases in the GH18 subfamily B. Mutation analysis indicates that Asn9 in Chi23 plays an important role in insoluble and soluble chitin degradation (Figures 6C–E). However, its counterparts in both ChiNCTU2 (His41) (Hsieh et al., 2010) and MmChi60 (His38) (Malecki et al., 2013) are far away from the bound chitooligomers, and only a weak hydrophobic stacking interaction is formed between His38 of MmChi60 and chitopentose, and no interaction forms between His41 of ChiNCTU2 and chitooligomers due to the buried property of this residue. Altogether, these data indicate that the three unique residues Asn9, Asp229 and Gln261 play an important role in Chi23 for chitin degradation, and their cumulative roles in substrate binding and catalysis contribute to the activity of the single-domain Chi23 against both crystalline and soluble chitin.

To study the contribution of identified substrate-binding residues to the thermostability of Chi23, the apparent melting temperatures (T_m) and optimal temperatures (T_opt) of Chi23 and its mutants were measured and compared. The T_m and T_opt values of mutants N9A, G79A, T80A and Q157A were similar to that of Chi23 (a T_opt of 60°C and a T_m of 77°C), suggesting that these residues may contribute little to the thermostability of Chi23. In contrast, mutation Q261A at the −2 subsite and all mutations including Y189A, N190A, D229A and W260A at the −1 subsite led to 6–14°C reduction in the T_m of Chi23 (Figure 8A), indicating the important roles of these substrate-binding residues at the −1 and −2 subsites in stabilizing the protein structure of Chi23. Among these five mutants, the T_opt values of mutants Y189A and W260A are undetectable due to the complete loss of their activity, mutant N190A had an unchanged T_opt, and mutants D229A (45°C) and Q261A (50°C) had much lower T_opt values than the wild-type Chi23 (60°C) (Table 5). Moreover, both D229A and Q261A became very unstable at 60°C, further supporting that residues Asp229 and Gln261 play an important role in maintaining the structural stability of Chi23. A detailed structural analysis of Chi23 showed that the side chain of Asp229 forms four hydrogen bonds with the solvent-exposed residues Thr226 and Ser264, and that the side chain of Gln261 is hydrogen bonded to residues Ser10 and Asn9 lining on the substrate cleft surface (Figure 8B). The mutation of the hydrophilic Asp229 or Gln261 to small hydrophobic Ala led to the disruption of the hydrogen bond networks involving these two residues and their surrounding residues in Chi23, and therefore the protein structure of Chi23 became less rigid and less stable at high temperatures. The contribution of hydrogen bonds to protein thermostability was also reported in other glycoside hydrolases (Vogt et al., 1997; Han et al., 2018). In addition, the absence of long flexible loops may also benefit the high thermostability of Chi23.

Among the reported GH18 endochitinases, Chi23 is most closely related to the eukaryotic single-domain endochitinase PPL2 (Cavada et al., 2006), with a low sequence identity of 30%, suggesting that Chi23 is a new member of the GH18 endochitinases. In addition to sequence, Chi23 is also different from reported GH18 bacterial endochitinases in structure due to the lack of the CBD and the β-hairpin subdomain (Figure 5). Blasting analysis revealed that more than 140 homologs of Chi23 are found in the NCBI non-redundant (nr) protein database, which are all potential single-domain proteins and distributed in various bacterial species (Figure 9A). Phylogenetic analysis showed that Chi23 and its homologs are clustered as a group separate from the cluster of the reported GH18 endochitinases (Figure 9B). Chi23 and its homologs are more closely related to GH18 eukaryotic endochitinases than to reported GH18 bacterial modular endochitinases (Figure 9B). Thus, based on the differences in sequence and structure between Chi23 (and its homologs) and other GH18 bacterial endochitinases, Chi23 and its homologs are suggested to represent a new group of GH18 endochitinases.