Enzymatic production of defined chitosan oligomers with a specific pattern of acetylation using a combination of chitin oligosaccharide deacetylases

Chitin and chitosan oligomers have diverse biological activities with potentially valuable applications in fields like medicine, cosmetics, or agriculture. These properties may depend not only on the degrees of polymerization and acetylation, but also on a specific pattern of acetylation (PA) that cannot be controlled when the oligomers are produced by chemical hydrolysis. To determine the influence of the PA on the biological activities, defined chitosan oligomers in sufficient amounts are needed. Chitosan oligomers with specific PA can be produced by enzymatic deacetylation of chitin oligomers, but the diversity is limited by the low number of chitin deacetylases available. We have produced specific chitosan oligomers which are deacetylated at the first two units starting from the non-reducing end by the combined use of two different chitin deacetylases, namely NodB from Rhizobium sp. GRH2 that deacetylates the first unit and COD from Vibrio cholerae that deacetylates the second unit starting from the non-reducing end. Both chitin deacetylases accept the product of each other resulting in production of chitosan oligomers with a novel and defined PA. When extended to further chitin deacetylases, this approach has the potential to yield a large range of novel chitosan oligomers with a fully defined architecture.

heterogeneous mixtures of chitosan oligomers. Still, due to the cleavage specificities of the enzymes, the resulting mixture will be better defined than the chitosan oligomer mixtures obtained by chemical or physical depolymerisation 2,12 . A fully controlled method potentially leading to a broad range of fully defined products is chemical synthesis of chitosan oligomers from monomeric building blocks. However, this attempt is time and labour intensive and the yields are rather low 14,15 . Alternatively, chitin oligomers can be efficiently converted into defined chitosan oligomers by the help of specific chitin deacetylases (EC 3.5.1.41) and chitin oligosaccharide deacetylases (EC 3.5.1.-) 9,16 . A number of different chitin deacetylases from different sources have been described [17][18][19][20][21][22] . Among these, two highly interesting candidates for the production of fully defined chitosan oligomers are the highly specific chitin oligosaccharide deacetylases NodB from Rhizobium spp. and COD from Vibrio cholerae. NodB deacetylates exclusively the GlcNAc unit at the non-reducing end, whereas COD deacetylates the second unit from the non-reducing end 18,20,23 . The limitation of this technique lies in the rather low number of recombinant chitin deacetylases available today with a known and fully defined mode of action, leading to a very limited number of defined chitosan oligosaccharides that can be obtained in this way.
To at least partially overcome this limitation, we have expressed cod from V. cholerae as well as nodB from Rhizobium sp. GRH2 in Escherichia coli and purified the recombinant enzymes. Starting with chitin oligomers of defined DP (A n ), we used the single enzymes for the in vitro generation of two different, fully defined mono-deacetylated chitosan oligomers (DA (n-1) and ADA (n-2) ). As a general agreement the first letter of the oligomers mentioned here represents the residue at the non-reducing end, while the last one represents that of the reducing end. As a proof-of-principle study, we then combined the two deacetylases in a single reaction and generated the expected doubly deacetylated oligomers (DDA (n-2) ), showing that this approach opens a way to produce novel chitosan oligomers with hitherto unreported PA.

Results and Discussion
Isolation and analysis of Rhizobium sp. GRH2 nodB. We designed consensus-degenerate hybrid oligonucleotide primers (CODEHOP) 24 based on the sequences of multiple nodB genes from different Rhizobium strains, and used these to identify and subsequently clone the chitin deacetylase gene from Rhizobium sp. GRH2. The analysis of 16S rDNA sequences by BLASTn revealed that Rhizobium sp. GRH2 is most closely related to Rhizobium leguminosarum bv. trifolii (99% identity).
Production of enzymes in E. coli and characterization of purified enzymes. The GRH2 nodB gene was expressed in E. coli BL21 (DE3), whereas the cod gene was expressed in E. coli Rosetta 2 (DE3) [pLysSRARE2] essentially as previously reported 23,25 . Both genes contained a downstream located Strep-tag II encoding sequence for the purification by streptactin affinity chromatography. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and western blot analysis revealed single bands for both enzymes, with apparent molecular masses close to the expected ones of 24.4 and 45.5 kDa for NodB and COD, respectively (Fig. 1).
The chitin deacetylase activity of NodB was tested against GlcNAc 1-6 . The hydrolysis products were characterized by ultra high performance liquid chromatography -evaporative light scattering detection -electrospray ionization -mass spectrometry (UHPLC-ELSD-ESI-MS), revealing that NodB was not active towards GlcNAc 1 , but converted GlcNAc 2-6 completely into mono-deacetylated chitosan oligomers (Table 1). To our knowledge, this is the first report of obtaining enzymatically active, recombinant NodB without the need of refolding it from insoluble inclusion bodies 20 . COD from V. cholerae is active on short chitin oligosaccharides 18,25 , and we tested COD with GlcNAc 1-6 under the same conditions and analysed the hydrolysis products in the same way as we did for NodB. As expected, COD converted GlcNAc 2-6 completely into mono-deacetylated chitosan oligomers, but not GlcNAc 1 ( Table 1). The optimal pH for NodB was 9 and the optimal hydrolysis temperature was 37uC ( Supplementary Fig. S1 online) as tested with GlcNAc 5 . According to literature, the optimal pH of COD is 8 and the optimal hydrolysis temperature is 45uC 18 . For COD no unspecific hydrolysis products were detected after prelonged incubation, as opposed to NodB, where negligible amounts of double deacetylated chitosan oligomers were detected after prolonged incubation at high enzyme concentrations.
Enzymatic sequencing 23 in combination with UHPLC-ELSD-ESI-MS analysis showed that recombinant NodB deacetylated the first unit from the non-reducing end generating DA 1-5 chitosan-oligomers ( Figure 2), which is in agreement with previous reports on recombinant NodB, refolded from the pellet fraction 20 . Likewise, the same sequencing methodology showed that COD deacetylated the second unit from the non-reducing end generating ADA 1-3 chitosan-oligomers, in agreement with previous reports 18 .
Production and analysis of defined chitosan oligomers. To test whether the chitin oligosaccharide deacetylases NodB and COD are also active on partially deacetylated chitosan oligomers -and whether the enzymes can be used in combination to broaden the spectrum of defined chitosan oligomers with a specific PA -we deacetylated GlcNAc 5 with NodB, removed NodB, and used the mono-deacetylated chitosan pentamer as a substrate for COD. The same was done in reverse order, and in a combined reaction of NodB and COD together. In all cases, GlcNAc 5 was completely converted into doubly deacetylated chitosan pentamers ( Figure 3). The PA of the doubly deacetylated chitosan pentamer was determined using enzymatic/mass spectrometric sequencing which revealed that the first two units from the non-reducing end were deacetylated in the combined as well as in both sequential reactions of NodB and COD ( Figure 4). In addition of GlcNAc 5 , also GlcNAc 2-4 and GlcNAc 6 were converted into the respective doubly deacetylated chitosan oligomers in a combined reaction of NodB and COD (Table 2).  In addition to the in vitro experiments, binding of partially deacetylated chitosan dimers and trimers to COD, was examined in silico. The recently solved three dimensional structures of COD in complex with GlcNAc 2 , and GlcNAc 3 25 were used as reference structures (PDB accession codes: 4NZ1 and 4OUI respectively). The dimeric and trimeric substrates where manually converted to all   Figure 5 shows the crystal structure of COD in complex with the disaccharide substrate AA, and the modelled structure of the complex with the NodB product DA, which properly binds to form a competent complex. The closest amino acid residues to the 2-NAc or 2-NH 2 in the glucosyl unit on the non-reducing end are Arg304 and Asn119.
Removal of the acetyl group in the DA substrate does not alter significantly the interactions map where only one hydrogen-bond with Arg304 is lost. The binding affinity of all acetylated, partially deacetylated, and fully deacetylated compounds to COD was calculated by means of the VINA scoring function 26 (Table 3). Not surprisingly, the fully acetylated compounds are the ones with strongest binding affinity towards COD (29.4 kcal mol 21 for AA, 210.6 kcal mol 21 for AAA). On the other hand, the deacetylated compounds produced by COD (AD and ADA) loose affinity towards the enzyme (increase in energy of roughly 1 kcal mol 21 ). Interestingly, the deacetylated compounds produced by NodB (DA and DAA) are able to bind COD with a stronger affinity than the COD originating products, but the interaction energy to COD is still reduced by 0.8 to 0.4 kcal mol 21 relative to the natural fully acetylated substrates.
In other words, the contribution to the binding affinity of the Nacetyl substituent of the sugar ring at the non-reducing end is only 0.8 kcal mol 21 for chitobiose in COD, in agreement with the lost of one hydrogen bond with Arg304 in the deacetylated substrate ( Figure 5). This contribution is lower for chitotriose (just 0.5 kcal mol 21 ). It is expected that this contribution will be even lower for longer oligosaccharides. Thus, although COD is more active on AA and k cat decreases with increasing oligosaccharide length 18,25 , compounds such as chitotetraose or chitopentaose continue to be substrates of COD even if they are deacetylated at the non-reducing end because the loss of affinity due to the acetyl substituent of the sugar ring at the non-reducing end will presumably be negligible.
The fact that NodB and COD accept the products of each other offers new possibilities for the biotechnological production of defined chitosan oligomers. Assuming that also other chitin deacetylases can accept partially deacetylated products as substrate, a rather large number of different fully defined chitosan oligomers with different PA could be produced by combining a rather limited number of different chitin deacetylases.
In order to produce sufficient amounts of specifically mono as well as doubly deacetylated chitosan oligomers for further downstream bio-testing, we biotechnologically produced chitin pentamer NodB was removed and the obtained mono-deacetylated chitosan pentamer (A 4 D 1 ) was further deacetylated with COD (b). The same was done vice versa: GlcNAc 5 (A 5 ) was deacetylated with COD (c) in the first step, and the enzyme was then replaced by NodB (d). Furthermore, NodB and COD were combined in a single reaction leading to a double-deacetylated chitosan pentamer in one reaction (e). All reactions were carried out in ammonium hydrogen carbonate buffer (pH 8) at 37uC for 2 h. in E. coli, of which we then deacetylated 100 mg in a combined reaction using NodB and COD at mg-scale. The chitosan pentamers thus obtained were purified using size-exclusion chromatography (SEC), separating them from chitosan tetramers, which originated from a by-product of the recombinant production of chitin pentamers in E. coli. The purified products were analysed using UHPLC- To determine the PA, it was first hydrolysed with the GlcNase GlmA TK , which removes exclusively GlcN units from the non-reducing end. The reaction resulted in GlcN 1 (D 1 ) and GlcNAc 3 (A 3 ) (b). In the next step, GlmA TK was replaced by the GlcNAcase BsNagZ, which exclusively removes GlcNAc units from the non-reducing end, resulting in GlcN and GlcNAc monomers (c).The enzymatic sequencing of the simultaneous hydrolysis product of NodB and COD revealed that the first two units starting from the non-reducing end were deacetylated, giving a specific chitosan oligomer with the novel PA DDAAA. ELSD-ESI-MS and proved to be highly pure ( Figure 6). We obtained 27 mg of highly pure doubly deacetylated chitosan pentamer (DDAAA) and 4 mg of equally pure doubly deacetylated chitosan tetramer (DDAA). As this method can be scaled up, it is feasible to produce fully defined chitosan oligomers in sufficient amounts for trial applications in different fields such as cosmetics or bio-medicine. Evaluating the scientific and commercial potential of fully defined chitosan oligomers is not possible at the moment due to the lack of welldefined chitosan oligomers on the market. When the chitin oligomers used as a substrate are of biotechnological origin, as in this study, our method has the added advantage that the chitosan oligomers produced, can be guaranteed to be free of allergenic contaminants.

Methods
Bacterial strains, vectors and culture conditions. E. coli strains TOP10 (Invitrogen, Darmstadt, Germany) and DH5a 27 were used for general cloning, whereas Rosetta 2 (DE3) [pLysSRARE2] and BL21 (DE3) (Novagen, Darmstadt, Germany) were used for protein expression. The vector pCRII-TOPO (Invitrogen) was used for cloning of PCR fragments, whereas pET-22b(1) (Novagen, Darmstadt, Germany) including a StrepII tag sequence upstream of the multiple cloning site (MCS) 23 was used for cloning and expression. E. coli was grown on LB agar, in LB medium, or in autoinduction medium 28 . Media were supplemented with the appropriate antibiotics (34 mg ml 21 chloramphenicol and/or 100 mg ml 21 ampicillin).
Preparation of genomic DNA from Rhizobium sp. GRH2 and 16S rDNA analysis. Genomic DNA was extracted from Rhizobium sp. GRH2 as described by Rainey et al. 29 . The 16s rDNA was amplified with Phusion Hot Start II High Fidelity proof reading DNA Polymerase (Fisher Scientific, Schwerte, Germany) using three different primer pair combinations (27F/1525R, 27F/926R and 16S-F/16S-Rev). All primer sequences are listed in Supplementary Table S1. Only 27F/926R yielded a product, which was purified, sequenced and analysed using BLASTn.

Identification of unknown nodB gene from GRH2 and plasmid construction.
Consensus-degenerate hybrid oligonucleotide primers (CODEHOP) [24] were designed against known nodB genes from different Rhizobium strains as previously described. All primer sequences used during the study are listed in Supplementary  Table S1. The genes were amplified using Mango-Taq DNA Polymerase (Bioline, Luckenwalde, Germany) and different combinations of CODEHOPs (B-for, D-rev, C-for, E-rev). Obtained PCR products were cloned in pCRII-TOPO for sequencing, followed by sequence analysis using BLASTx and alignment with CloneManager (Scientific & Educational Software, Cary, USA). Inverse PCR was used to isolate the upstream and downstream regions from genomic DNA. Total DNA was each digested with either SacI or HindIII and circularized with rapid T4 DNA ligase.
The circulized DNA was used as template for a PCR with different inverse primer combinations. Amplified PCR products were ligated into pCRII-TOPO vector for sequencing. Promising sequences were aligned with CloneManager to generate a contig. The full-length nodB gene was isolated by standard PCR using a further CODEHOP primer (BC-for) designed for known nodB genes from different Rhizobium strains and specific primers that bind within the partially identified regions of the nodC gene from Rhizobium sp. GRH2, which is located upstream of nodB. In the last step the full length nodB gene was amplified by using specific primers and Phusion Hot Start II High Fidelity proof reading DNA Polymerase (Fisher Scientific) to exclude possible mutations, which may had occurred during the different PCR amplification steps with the non-proof reading Mango-tag polymerase.   by SDS-PAGE in a 12% (w/v) gel 30 . Separated proteins were visualized with zincon/ ethyl violet staining 31 or transferred to a nitrocellulose membrane (GE Healthcare Europe GmbH, Freiburg, Germany) using a semi-dry transfer procedure 32 . Strep-tag II fusion proteins were detected with Strep-Tactin-horseradish peroxidase (HRP) conjugate and chemiluminescent reaction was developed according to the instruction of the manufacturer (IBA GmbH, Göttingen, Germany).
Determination of pH and temperature optimum for NodB. The pH and temperature optima of NodB were determined indirectly, by measuring the amount of acetate released during the enzymatic reaction, using an acetic acid assay kit (Rbioharm AG, Darmstadt, Germany) adapted for microtiter plates. The pH optimum was determined over the pH range 4-12 at 37uC for 2 h, in buffer containing 100 mM NH 4 HCO 3 , 20 mM TEA, 20 mM KH 2 PO 4 and 20 mM Na 2 HPO 4 . Teorell and Stenhagen buffer (100 mM citric acid, 100 mM phosphoric and 100 mM boric acid 33 ) was used for high-pH conditions, overlapping at pH 10. The temperature optimum was determined at 4uC, 22uC, 37uC, 45uC and 55uC at pH 9 using the buffer and conditions described above. The reaction volume was set to 400 ml, comprising 50 mM buffer, 1 mM chitin pentamer (Megazyme, Bray, Ireland) and 0.5 mM purified NodB. The reactions were stopped and the amount of released acetic acid was measured directly. Determination of the PA of the generated chitosan oligomers. The PA of the different chitosan oligomers generated by NodB and COD, or by a combined action of both enzymes was determined by enzymatic sequencing 23 . Samples were analysed by UHPLC-ELSD-ESI-MS analysis instead of HPTLC. The composition of the shake flask and minimal batch fermenter medium was the same as described by Waegeman et al. 36 . Fed-batch medium consisted of 500 g l 21 glucose, 1 g l 21 MgSO 4 37H 2 O and 30 g l 21 NH 4 Cl. All media were supplied with 0.1 g l 21 ampicillin. The recombinant E. coli strain was pre-cultured in two shake flasks of 2 l total volume, filled with approximately 0.2 l shake flask medium and grown at 30uC while constantly shaking at 200 rpm. After 24 hours, 100 ml of broth was transferred into two shake flasks of 5 l total volume filled with 2 l shake flask medium at 30uC while constantly shaking at 200 rpm. After 24 hours the resulting fermentation broth (4 l in total) was inoculated into a 150 l fermenter (Sartorius, Göttingen, Germany). Fermentation conditions were; temperature: 30uC, stirrer speed: minimum 500 rpm and if pO 2 drops below 30% stepwise increased, aeration: 10 slpm, pH: 7, pressure: 500 mbar. Feeding was started when glucose was depleted in the batch phase and the feed rate applied was 4 l h 21 . The total fermentation time was 110 h. After fermentation, the broth was harvested and cells were separated from supernatant by ceramic tangential flow microfiltration (Tami Industries, Nyons, France, 0.45 mm). Both the cell and supernatant fraction contained the fully acetylated pentamer and were further purified. Cells were disrupted by a homogeniser (GEA Process Engineering, Mechelen, Belgium, 10 l h 21 ) and cell debris was removed by tangential flow microfiltration (Kleenpak, Pall, Zaventem, Belgium, 0.45 mm). In both fractions, salts were removed by ion exchange (Amberlite, Dow, Tessenderlo, Belgium) and the solutions were further concentrated by wiped film evaporation (Carl Canzler, Germany, 150 l h 21 , 60uC, 70 mbar) and finally spray-dried (Xedev, Zelzate, Belgium). After purification, 371 g of product was obtained, which consisted for 85% of fully acetalyated pentamer (GlcNAc 5 ). Purified chitin pentamer (100 mg) was incubated with 0.2 mM NodB and 0.6 mM COD in 100 ml 50 mM NH 4 HCO 3 buffer (pH 8) at 37uC overnight. The sample was freeze dried and solved in water before size exclusion chromatography (SEC). After SEC the sample was diluted with water and freeze dried. The obtained product was analysed using UHPLC-ELSD-ESI-MS.

UHPLC-ELSD-ESI
Size-exclusion chromatography (SEC). Deacetylated chitosan oligomers were purified on three HiLoad 26/600 Superdex 30 pg columns (GE Healthcare Europe GmbH, Freiburg, Germany) in a row with an overall dimension of 2.60 3 180 cm. Ammonium acetate buffer (0.15 M, pH 4.5) was used as mobile phase and the flow rate was set to 0.8 ml min 21 . The effluent was monitored with an online refractive index detector (1260 Infinity Refractive Index Detector, Agilent Technologies Deutschland GmbH, Böblingen, Germany) which was coupled to a datalogger 13,37 . Fractions containing oligomers were pooled and analysed by UHPLC-ELSD-ESI-MS.
Calculation of relative binding affinities of partially deacetylated chitobiose and chitotriose to COD structure. Three dimensional structures of COD in complex with chitobiose and chitotriose have been recently obtained 25 and deposited in the Protein Data Bank with accession codes 4NZ1 and 4OUI respectively. Polar hydrogens were added to the receptor protein structure with AutoDockTools 38 . AutoDock4.2 atom typing was used. Gaisteger partial charges were computed for each atom with AutoDockTools. Ligand structures (chitobiose, AA and chitotriose, AAA) were extracted from the corresponding PDB files. For each ligand, a series of partially and fully deacetylated compounds were generated by removing the corresponding N-acetyl group with AutoDockTools while keeping the overall geometry of the molecules. Thus, three dimensional structures AA, AD, DA, DD, AAA, ADA, AAD, ADD, DAD, DDD were obtained. Every ligand was parametrized in the same way as the receptor. Each ligand was docked onto COD structure by means of AutoDock VINA algorithm 26 . A grid-box of 30 3 26 3 30 A 3 centered at the active site was used as the search space for docking. Interaction energies between ligands and COD receptor were calculated with VINA scoring function 26 . Reported values in Table 3 are an estimation of the differences in free energy upon binding for those docking poses in which the ligand binds in a productive orientation at the active site. The uncertainty of these calculations is 0.6 kcal mol 21 . This was estimated as the average range of VINA score values obtained for a series of similar docking poses (structures within 1.5 A of the lowest score one). This work is licensed under a Creative Commons Attribution 4.0 International License. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder in order to reproduce the material. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ www.nature.com/scientificreports SCIENTIFIC REPORTS | 5 : 8716 | DOI: 10.1038/srep08716