A new group of glycoside hydrolase family 13 α-amylases with an aberrant catalytic triad

α-Amylases are glycoside hydrolase enzymes that act on the α(1→4) glycosidic linkages in glycogen, starch, and related α-glucans, and are ubiquitously present in Nature. Most α-amylases have been classified in glycoside hydrolase family 13 with a typical (β/α)8-barrel containing two aspartic acid and one glutamic acid residue that play an essential role in catalysis. An atypical α-amylase (BmaN1) with only two of the three invariant catalytic residues present was isolated from Bacillus megaterium strain NL3, a bacterial isolate from a sea anemone of Kakaban landlocked marine lake, Derawan Island, Indonesia. In BmaN1 the third residue, the aspartic acid that acts as the transition state stabilizer, was replaced by a histidine. Three-dimensional structure modeling of the BmaN1 amino acid sequence confirmed the aberrant catalytic triad. Glucose and maltose were found as products of the action of the novel α-amylase on soluble starch, demonstrating that it is active in spite of the peculiar catalytic triad. This novel BmaN1 α-amylase is part of a group of α-amylases that all have this atypical catalytic triad, consisting of aspartic acid, glutamic acid and histidine. Phylogenetic analysis showed that this group of α-amylases comprises a new subfamily of the glycoside hydrolase family 13.

Scientific RepoRts | 7:44230 | DOI: 10.1038/srep44230 Twenty years ago, several α -amylases and related enzymes composed of a (β /α ) 7 -barrel (an irregular TIM-barrel domain) were classified into the family GH57; more recently also the family GH119 was established 9,10 . Both of these enzyme families are at present considerably smaller than GH13 and only few members have been characterized in detail 11 . The first determined 3D structure of GH57 was that of the 4-α -glucanotransferase from Thermococcus litoralis (TLGT). X-ray crystallography supported by site-directed mutagenesis of TLGT revealed that it has two catalytic residues, Glu123 and Asp214, as the catalytic nucleophile and the general acid/base, respectively 12 . No 3D structure is currently available for GH119 members. In addition to the structural differences between GH13 and GH57-GH119 family members, there are also distinctive conserved regions between these families 9 . The GH57 and GH119 families share five conserved sequence regions 13 .
Several microbial strains isolated from a unique land-locked marine lake located in Kakaban island, part of the Derawan Islands, East Kalimantan, Indonesia, were screened for the production of α -amylases. The lake originally was the lagoon of an atoll, formed by corals over a period of two million years. As a result of movements in the earth's crust the coral reef was raised above the sea level, trapping 5 km 2 of seawater within a 50 meter high ridge, effectively creating a land-locked marine lake 14 . One of the isolates, Bacillus megaterium NL3, contained an active GH13 α -amylase with only the general acid/base residue (Glu231) at the conserved position. Amino acid sequence alignments and 3D homology modeling showed that the nucleophile may be shifted one position downstream (Asp203) and that the transition state stabilizing residue is not the canonical Asp but a His residue (His294). Phylogenetic analysis clustered this new α -amylase and its homologs, which also possess the incomplete GH13 catalytic machinery, as a separate branch in family GH13, representing a novel subfamily.

Results and Discussion
Screening of Kakaban lake isolates producing extracellular amylases. Eight of twenty bacterial isolates from Kakaban landlocked marine lake tested positive for the hydrolysis of starch by producing a clear halo around their colonies on red-dyed amylopectin agar plates. Isolate NL3 showed the largest clearing zone, indicating a relatively high α -amylase activity and was selected for further study. 16S rDNA sequence analysis showed that strain NL3 was most closely related to Bacillus megaterium. This result was in agreement with biochemical and physiological properties (data not shown) and hence the selected isolate was designated as B. megaterium NL3. The culture medium of strain NL3 showed activity towards soluble starch. The 50-80% ammonium sulphate precipitate of the culture medium gave a single protein band with molecular mass of approximately 55 kDa on SDS-PAGE in combination with activity staining with soluble starch (data not shown).

Molecular identification of the NL3 amylase.
Using degenerate α -amylase specific primers and inverse PCR, a DNA fragment of 2.3 kb was obtained from genomic DNA of strain NL3. Analysis of the nucleotide sequence of this fragment showed that an open reading frame of 1515 bp with clear α -amylase sequence similarity was present. This gene was designated as bmaN1. A putative ribosomal binding site (RBS) corresponding to the AGGAGG sequence located 12 nucleotides upstream of the start codon was predicted. A probable catabolite responsive element (CRE) was found together with possible − 10 (TATAAT) and − 35 (TTAACA) regions. The CRE sequence showed only one mismatch in the last position when compared to the consensus sequence (TGT/ AAANCGNTNA/TCA) 15 . The BmaN1 polypeptide deduced is 505 amino acid residues in length with a clear putative signal peptide sequence of 23 residues preceding the mature enzyme, as predicted by SignalP 4.0 Server 16 . The molecular weight and pI of BmaN1 were predicted using ExPASy server (http://web.expasy.org/protparam) as 56934 Da and 9.05, respectively. The full-length DNA sequence of the putative α -amylase gene of B. megaterium NL3, bmaN1, has been deposited in the GenBank database 17 under the accession no. AGT45938.
In silico analysis of BmaN1 and its homologues. The BLAST search using the BmaN1 amino acid sequence as a query resulted in retrieving of more than 30 highly similar sequences of putative α -amylases ( Fig. 1) some of which have already been classified in the family GH13 18 . Although all of them possess variations in the three residues forming the family GH13 catalytic machinery, it is possible to divide them into two groups: (i) the first, larger group (Nos 1-27 in Fig. 1) with Lys202 and His294 in the positions of the catalytic nucleophile and transition state stabilizer, respectively (instead of normally occurring aspartates); and (ii) a second, smaller group (Nos 28-34 in Fig. 1) exhibiting substitutions in positions of the entire catalytic triad, but rather without an obvious regularity (Fig. 2). While the sequences of the members of the former group are almost identical to BmaN1, those of the latter one are slightly different (Fig. 2). Interestingly, there is a strictly conserved aspartic acid residue succeeding the "strange" lysine, which corresponds with the position of the catalytic nucleophile (Fig. 2). The sequences of both these groups, proposed here to constitute a novel GH13 subfamily xy around the α -amylase from B. megaterium BmaN1, are all highly similar to those of α -amylases around the α -amylase from B. aquimaris BaqA (Nos 35-39 in Fig. 1) suggested recently to define also a new and independent GH13 subfamily xx 19 . The main difference between the α -amylases around the BmaN1 and those around BaqA is that the BaqA α -amylase and the members of its subfamily possess the complete catalytic machinery (Fig. 2) characteristic for the α -amylase family GH13 7 . The other feature of interest is the presence of a tryptophan pair in both BmaN1 and BaqA (Fig. 2) between the CSR-V (loop 3) and CSR-II (strand β 4), located in the helix α 3 of the catalytic (β /α ) 8 -barrel 19 .
In addition to the incomplete catalytic machinery mentioned above, the most striking differences of BmaN1 and its close homologues discriminating them from other well-established GH13 subfamilies with the α -amylase specificity (Fig. 2) are the presence of a glutamic acid instead of aspartate at the beginning of the CSR-I (strand β 3) and in addition the position of the histidine in the middle of the CSR-V (loop3), a position usually occupied by aspartic acid 20 . With regard to alignment of the representative α -amylases studied here (Fig. 1), its substantial part covering almost the entire (β /α ) 8 -barrel including domain B clearly document a very close homology of both eventual GH13 subfamilies, i.e. BmaN1 and BaqA. All these sequences (Nos 1-39 in Fig. 1) go well together exhibiting their own pattern of the alignment in comparison to remaining α -amylases that represent well-established GH13 subfamilies (Nos 40-63 in Fig. 1). Of note is also the fact that the small group of putative α -amylases with irregular substitutions in catalytic positions (Nos 28-34 in Fig. 1) exhibits obviously a higher similarity to the BaqA subfamily than to that around BmaN1, especially in domain B (preceding the CSR-V in loop3) as well as in the segment preceding the CSR-IV at strand β 7 (Fig. 2). The cyclomaltodextrinase from Flavobacterium sp. No. 92 21 was added to the comparison as an interesting example since it was recently found to possess the pair of adjacent tryptophan residues 19 , typical for both BmaN1 and BaqA (Fig. 2). This is of interest because the cyclomaltodextrinase belongs to GH13 members intermediate between subfamilies of oligo-1,6-glucosidases and neopullulanases 22 that are closely related to α -amylases from the subfamily GH13_36 that, however, do not possess the tryptophan pair (Fig. 2).
A topological alignment of BmaN1 and the putative α -amylases of B. megaterium DSM319, Bacillus sp. 278922, B. flexus, B. aryabhattai, B. megaterium WSH-002, and GTA was made (Fig. S1). Almost all β -strands and α -helices of the TIM barrel in domain A and the Greek key motif in domain C are conserved in these α -amylases. A model of the 3D structure of BmaN1 was generated by the PHYRE server 23 and visualized by the MacPymol software 24 (Fig. 3). The BmaN1 protein displayed 40% homology (100% confidence, 85% sequence coverage) with the X-ray crystal structure of Geobacillus thermoleovorans α -amylase (GTA, PDB code: 4E2O) 25 which was used as a template for the modeling. The comparison between the model and the GTA crystal structure revealed that the global topology is almost the same (Fig. 3). The BmaN1 protein model folds into three distinct domains: a central A domain of 366 residues harboring a (β /α ) 8 barrel, with an irregular loop domain of 37 residues (domain B) connecting the third β -sheets strand and the third α -helix of the barrel. The C domain of 79 residues has an eight-stranded anti-parallel β -sandwich-like fold (Fig. 3). The (β /α ) 8 barrel is quite similar to the (β /α ) 8 -barrel found in maltogenic amylase from Pseudomonas saccharophila and Bacillus stearothermophilus 26 , in that there is an additional helix between Aα 6 and Aβ 7, which is a three-turn helix lying nearly parallel with the Aα 6 strand.
Superposition of acarbose-bound GTA with the BmaN1 model demonstrated that of the three catalytic residues found in GH13 α -amylases, only residue Glu231 of BmaN1 superimposes with the corresponding residue in Figure 2. Sequence alignment of CSRs of studied family GH13 enzymes with focus on the novel α-amylase subfamily. The two consecutive tryptophans characteristic for the novel α -amylase subfamily are also shown. Colour code for the selected residues: W, yellow; F, Y -blue; V, L, I -green; D, E -red; R, K -cyan; H -brown; C -magenta; G, P -black. The positions of the three catalytic residues are boxed and signified by asterisks under the alignment. The label of the protein source consists of the name of the organisms and the UniProt (UniParc) accession number. The number at the beginning of the protein source label indicates the number of known (already established) GH13 subfamily. For the newly proposed BmaN1 GH13 subfamily, the label "xy" is used; similarly ("xx") for until now non-defined subfamily around the BaqA. The alignment of all 64 enzymes spanning the sequence segment from the beginning of the strand β 2 (CSR-VI) to the end of the strand β 8 (CSR-VII) is shown in Fig. S2.
Scientific RepoRts | 7:44230 | DOI: 10.1038/srep44230 GTA (Glu246), and presumably is the general acid/base in BmaN1 (Fig. 3). As already concluded from sequence alignments, two of the three catalytic residues are not conserved in BmaN1. Lys202 replaces the catalytic aspartate (Asp217 of GTA); however, Asp203 directly downstream of the lysine is positioned nearby and has its carboxylic acid side chain pointing into the presumed substrate-binding groove (Fig. 3). Furthermore, at the position corresponding to the nucleophile, His294 replaces the transition-state stabilizing aspartate residue (Asp314) found in α -amylases. The absence of any one of the catalytic residues normally causes partial or complete loss of hydrolysis activity 27 . Remarkably, the mutant α -amylase from Xanthomonas campestris truncated from the C-terminal part of domain B and thus lacking any of the three conserved catalytic residues, still exhibited starch-hydrolyzing activity 28 , but that observation has never been supported by other examples.
Phylogeny of BmaN1 and other α-amylases. The evolutionary relatedness of the α -amylase from B. megaterium BmaN1, representing all its homologues with lysine and histidine in positions of the catalytic nucleophile and transition state stabilizer, respectively ( Fig. 2; Nos 1-27 in Fig. 1), to members of the recently proposed GH13 subfamily around the BaqA (Nos 35-39 in Fig. 1) as well as to representatives of remaining well-established GH13 subfamilies with α -amylase specificity (Nos 40-63 in Fig. 1), is shown in the evolutionary tree (Fig. 4). It is clear that both subfamilies BmaN1 and BaqA are most closely related to each other among all family GH13 α -amylases. Furthermore, a small group of putative α -amylases with irregular substitutions in catalytic positions (Nos 28-34 in Fig. 1) may be considered as an intermediary connection between both BmaN1 and BaqA subfamilies since, despite the lack of complete family GH13 catalytic machinery (similar to BmaN1), they cluster together with representatives of the BaqA subfamily (Fig. 4). One of the most convincing sequence-structural features characteristic for all these α -amylases is the presence of the pair of adjacent tryptophan residues in helix α 3 of the catalytic (β /α ) 8 -barrel (Fig. S2) 19 . Interestingly, the Flavobacterium sp. No. 92 cyclomaltodextrinase (with the tryptophan pair) is positioned in the evolutionary tree between the subfamilies of BmaN1 and BaqA and all other remaining GH13 families with the α -amylase specificity (Fig. 4).
BmaN1 encodes an active exo-acting α-amylase. The gene encoding BmaN1 was cloned in vector pMM1525 and this recombinant plasmid was transformed to B. megaterium MS941. A transformant with clear α -amylase activity, as detected on starch plates by iodine staining, was selected and grown in liquid medium. The culture medium was saturated with 50-80% concentrations of ammonium sulphate to purify the BmaN1 α -amylase enzyme. The molecular weight of the partially purified BmaN1 was estimated to be 55 kDa as judged structure, (C) Active-site region in a superposition of BmaN1 with GTA including the acarbose bound in its subsites − 2 to + 2 (white carbon atoms). Active-center residues of BmaN1 (orange) and GTA (grey) are given as stick models and labeled in orange (BmaN1) and black (GTA).
from activity staining after protein renaturing on SDS-PAGE gels (Fig. S3). In contrast, no band was observed in the culture supernatant of B. megaterium MS941 carrying pMM1525 without any insert (Fig. S3). Amylolytic activity of BmaN1 was measured spectrophotometrically by incubating it with soluble starch and measuring the increase in the amount of reducing sugars released over a period of 40 min (Fig. 5). A clear increase in reducing ends was observed, resulting in an activity of 8.4 U/mg of protein. BmaN1 was found to be most active at 55 °C and pH 6.0. The main end products formed from soluble starch were glucose and maltose, indicating an exo-acting mode of action. Minor amounts of longer chain maltooligosaccharides were also found (Fig. 6). This  mode of action is very similar to that of the amylase from Bacillus sp. IMD 435 that releases glucose and maltose as the major products on hydrolysis of both soluble starch and raw corn starch 29 .
The results presented above indicate that the substitution of an aspartate residue by a histidine, a positively charged amino acid, still gives (some) amylase activity. The reaction mechanism of BmaN1 may be essentially different from the general catalytic reaction mechanism of α -amylases. Further experiments are needed to demonstrate whether the His residue indeed is one of the catalytic residues of α -amylases.

Methods
Materials. All chemicals used were reagent grade and were obtained from either Fermentas (Maryland, USA) or Difco Laboratories (New Jersey, USA).
Screening of α-amylase producing bacteria. Bacterial isolates were inoculated on MB agar plates supplemented with 1.0% (v/v) red-dyed amylopectin 30 and then incubated at 30 °C for 24 h. The appearance of a clear zone against a red background was indicative for the production of α -amylase activity. The positive isolates were then subjected to a second screening round using MB agar plates containing 1.0% (w/v) potato or wheat starch. A clearing zone around the bacterial colony indicated that the starch was hydrolyzed and thus that the isolate produced extracellular amylase activity.
Bacterial identification. The isolate showing the largest clearing zone on starch-agar plate was identified by 16SrDNA sequencing. Chromosomal DNA was isolated using Wizard Genomic DNA Purification (Promega). The 16S rDNA gene was amplified by PCR using universal primers UniB1 and BactF1 (Supplementary Table 1). The resulted 1.4 kb fragment was sequenced using the dideoxy-chain termination method (Macrogen, South Korea). The bacterial isolate was identified by aligning the 16s rDNA sequences with other known bacteria using NCBI BLASTn (http://www.ncbi.nlm.nih.gov). 16S rDNA gene sequence was submitted to GenBank. Cloning of the α-amylase-encoding gene and plasmid construction. Two degenerate primers (Table S1) were designed based on the amino acid sequences of the well-conserved regions (region VI-VII) of α -amylases from several Bacilli. The first α -amylase gene fragment was amplified by polymerase chain reactions (PCR), using chromosomal DNA from B. megaterium NL3 as a template and the two degenerate primers. The PCR products were inserted into pGEM-T vector (Promega) and transformed into E. coli TOP10. Plasmid DNA of the transformed E. coli TOP10 was isolated and the nucleotide sequence of the inserted DNA was determined using the dideoxy-chain termination method (Macrogen). The resulting nucleotide sequence was used to design a set of primers, NL3_SP8-invF1 and NL3_SP8-invR1 (Table S1), to amplify parts of α -amylase gene beyond the conserved region. The chromosomal DNA was partially digested with EcoRV and then self-ligated using T4 DNA ligase (Fermentas). Inverse PCR (iPCR) was performed with Dream Taq polymerase (Fermentas) and the primers listed in Supplementary Table 1 using the self-ligated DNA fragment as a template. Analysis of sequence data and sequence similarity searches was performed using the BLAST program of the National Center for Biotechnology Information (NCBI). Primers pMM-NL3-F and pET/MM-NL3-R (Table S1) were used to amplify the complete open reading frame of the α -amylase gene which was designated as bmaN1.
Transformation of B. megaterium. The recombinant plasmid containing the α -amylase gene, pMM1525-bmaN1, was transformed into the expression host, B. megaterium MS941. The transformation procedure was essentially conducted as described by Puyet et al. with some modifications 31 . A 0.5 ml protoplast suspension was added to a tube containing 5.0 μ g DNA and 1.5 ml PEG-P (40% (w/v) PEG6000 in 1x SMMP) for each transformation and incubated for 2 min at room temperature. SMMP medium contains 3.5% (w/v) AB3 (Antibiotic medium no. 3, Difco), 1 M sucrose, 40 mM disodium maleic acid and 40 mM MgCl 2 (pH adjusted to 6.5 before autoclaving for 12 min) and prepared freshly before use. To the mixture, 5.0 ml SMMP was added and mixed by rolling the tube carefully. Cells were harvested by centrifugation at 2,700 × g for 10 min at room temperature and the supernatant was poured off immediately. The pellets were resuspended with 0.5 ml SMMP and incubated at 37 °C for 90 min with gentle shaking or rolling of tubes (max. 100 rpm). Then, 50 to 200 μ l of cells were added into top agar and mixed gently by rolling the tube. The mixture was poured on a pre-warmed plate of LB containing 12 μ g/ml of tetracycline and incubated at 37 °C overnight.
Expression and partial purification of recombinant α-amylase. α -Amylase was produced by growing the B. megaterium harboring recombinant plasmids in 20 ml of LB medium supplemented with 12 μ g/ml tetracycline at 37 °C with shaking. The overnight culture was transferred into fresh media and incubated until the 546-nm absorbance reached 0.8-1.0. Subsequently, expression was induced by adding 1% (v/v) xylose, and the culture was incubated at 18 °C with constant shaking at 150 rpm for 24 h. Cells were removed by centrifugation (6000 × g, 10 min) and the resulted supernatant was subjected to ammonium sulphate precipitation at a saturation value up to 80%. The precipitate was dissolved and dialyzed against 50 mM maleate buffer pH 6.0 at 4 °C. This partially purified α -amylase was used for further studies.
Gel electrophoresis and activity staining. SDS-PAGE was carried out as described by Laemmli 32 and gels were then stained with Coomassie Brilliant Blue (Bio-Rad). For α -amylase activity test, the protein samples were separated by SDS-PAGE containing 1% soluble starch. After electrophoresis, SDS was removed by washing the gel with water followed by 10 min incubation at room temperature. This was repeated twice. The gels were then immersed in the enzyme reaction buffer (50 mM maleate buffer pH 6.0) for 4.0 h at 55 °C and then stained with KI/I 2 solution for 10 min and followed by rinsing with water. The α -amylase activity was detected as a clear zone against a purple background.
Enzyme assay. The amylase activity assay was conducted using the 3,5-dinitrosalicylic acid (DNS) method described by Miller (1955) with a slight modification 33 . Briefly, the assay was performed by adding 25 μ l of enzyme sample into 25 μ l of 1% (w/v) soluble starch (Fermentas, USA) in 50 mM of the appropriate buffer and then incubated at 55 °C for 10 min. To the reaction mixture, 50 μ l of DNS reagent was added. The absorbance at 500 nm was measured and then the amount of reducing sugar-end was calculated using a glucose standard curve. One unit of α -amylase activity was defined as the amount of enzyme that releases 1 μ mol of reducing sugar per min under the assay conditions. The protein concentration was determined using the Lowry method and bovine serum albumin as the standard.

Analysis of sugars.
The starch hydrolysis products were analyzed by high-performance liquid chromatography (HPLC). Aliquots of 100 μ l of enzyme solution were incubated at 55 °C in the presence of 1% (w/v) soluble starch, maleate buffer 50 mM. After specific time intervals, samples were withdrawn and hydrolysis was stopped by incubation at 90 °C for 10 min. After centrifugation at 12000 × g for 10 min at 4 °C, the products were then analyzed by HPLC (Aminex ® HPX-87H system). The separated hydrolysis products were identified by calculating based on peak areas compared to standard glucose, maltose, and purified maltooligosaccharide (Sigma).
Bioinformatics. The sequences eventually forming the new GH13 subfamily xy were collected based on protein BLAST 34 searches against the non-redundant database using the entire amino acid sequence of Bacillus megaterium NL3 α -amylase BmaN1 (UniProt accession No.: T1SIF2) as well as on previous bioinformatics analyses when the BLAST was performed with the Bacillus aquimaris α -amylase BaqA 19,35 . The main criterion applied for the selection was the lack of at least one residue from the catalytic triad characteristic for the α -amylase family GH13. In addition to BmaN1 and its closely related homologues, five experimentally characterized α -amylases from the recently proposed GH13 subfamily around the B. aquimaris α -amylase BaqA 19,35 were added. These α -amylases -BaqA from B. aquimaris 35 , ASKA and ADTA from Anoxybacillus sp. 36,37 , GTA and GTA-II from Geobacillus thermoleovorans 25,38,39 -exhibit a high degree of sequence similarity with BmaN1, but possess the complete GH13 catalytic machinery 19 . The entire set was finally completed by two selected representatives from well-established GH13 subfamilies with the α -amylase specificity, i.e. 1, 5, 6, 7, 15, 19, 24, 27, 28, 32, 36 and 37 19 including also the related but until now unclassified cyclomaltodextrinase from Flavobacterium sp. No. 192 21 so that the final number of studied enzymes and hypothetical proteins was 64 (Fig. 1).
All 64 GH13 sequences were retrieved from GenBank 17 and UniProt 40 sequence databases and the set was aligned using the program Clustal-Omega 41 available at the European Bioinformatics Institute's web-site (http:// www.ebi.ac.uk/). A subtle manual tuning was done in order to maximize similarities, especially with regard to aligning the individual CSRs. The boundaries of the CSRs were defined based on previous bioinformatics studies 19,20 . The evolutionary tree was constructed based on the final alignment of the sequence segment corresponding to a 255-residue long region of BmaN1 α -amylase spanning almost the entire catalytic (β /α ) 8 -barrel domain including the domain B from the beginning of the CSR-VI (strand β 2; starting with Gly79) to the end of the CSR-VII (strand β 8; ending with Ser333). The tree was calculated as a Phylip-tree type using the neighbour-joining clustering and the bootstrapping procedure -the number of bootstrap trials used was 1,000 42 implemented in the Clustal-X package 43 , and then displayed with the program iTOL 44 .
The 3D structure of BmaN1 was predicted by QuickPhyre structure program server (http://www.sbg.bio.ic.ac. uk/phyre) 23 . Structural modeling of the BmaN1 was performed based on the crystal structures of α -amylase of G. thermoleovorans [PDB access code: 4E20]. The generated BmaN1 structures were displayed and drawn by MacPymol.