Introduction

α-Amylases (EC 3.2.1.1) are endo-acting enzymes catalyzing the hydrolysis of α-1,4-glucosidic linkages in starch, glycogen and related polysaccharides using the retaining mechanism (Janeček et al. 2014). The tertiary structures of α-amylases bear three domains designated as A, B, and C. With being the most conserved one, domain A contains the active site of the enzymes in a typical (β/α)8 TIM-barrel structure (Janeček et al. 1997, 2014). The Asp, Glu and Asp residues, positioned on β-strands of the barrel domain, mostly create the catalytic triad consisting of catalytic nucleophile, proton donor, and transition-state stabilizer functions, respectively (Matsuura et al. 1984; Janeček 2002). There are also seven conserved sequence regions (CSRs), CSR-I to CSR-VII, that are characteristic for the α-amylase family and function in catalytic or substrate-binding activities (Janeček 2002).

According to the actual sequence-based classification system in the Carbohydrate-Active enZYmes database (CAZy; http://www.cazy.org/), (Cantarel et al. 2009), α-amylases represent the largest group among Glycoside Hydrolyses (GHs), (Lombard et al. 2014). α-Amylase specificity is presented mostly in the GH13 family in addition to the GH57, GH119 and eventually in GH126 (Janeček et al. 2014, 2015). The family GH13 also forms the clan GH-H with the families GH70 and GH77 in the CAZy database. GH13 family is divided into totally 42 subfamilies, nevertheless the specific α-amylase activity is present only in the 15 subfamilies of GH13_1, 5, 6, 7, 15, 19, 24, 27, 28, 32, 36, 37, 39, 41, and 42 (Stam et al. 2006; Cantarel et al. 2009; Sarian et al. 2017). Many new subfamilies have also been proposed and awaiting to be assigned into a definite subfamily (Stam et al. 2006).

The starch is abundant in nature; therefore, the starch hydrolyzing amyloytic enzymes are also widespread among the living beings from eukaryotic to the prokaryotic origin, though displaying some differences in their enzyme activities and substrate specificities (Stam et al. 2006). Moreover, all the Bacillaceae family members frequently diverge in nature and human-made environments due to their endo-spore forming capabilities and their resistance to many harsh conditions such as high or low pH and temperature values. The temperature requirements of these bacilli also vary from mesophilic to thermophilic growth (Nazina et al. 2001; Cihan et al. 2014b). Since Cohn had first introduced the mesophilic genus of this family as genus Bacillus in the eighteenth century, its thermophilic members from the genera Geobacillus and Anoxybacillus could only be described after 2000’s (Cohn 1872; Pikuta et al. 2000; Nazina et al. 2001). These explanations clarify why all the commercially available α-amylases are from mesophilic bacilli such as Bacillus licheniformis (MEGAZYME, E-BLAAM), Bacillus amyloliquefaciens (MEGAZYME, E-BAASS), and Bacillus subtilis subsp. subtilis str. 168 (PROZOMIX, PRO-E0403). The α-amylases of the mesophilic species Bacillus aquimaris MKSC 6.2 (BaqA), (Puspasari et al. 2013) and Bacillus megaterium NL3 (BmaN1), (Sarian et al. 2017) are also the well-characterized ones.

After the twentieth century, due to some beneficial features of the thermostable enzymes in the industrial processes, the isolation of novel thermophilic bacilli gained considerable attention. In the two phylogenetic diversity studies dealing with the Bacillaceae family revealed that Anoxybacillus species are more predominant in the extreme hot environments probably owing to their carbohydrate degrading abilities among the other genera members (Derekova et al. 2008; Cihan 2013). Many thermostable α-amylases from the genera Anoxybacillus and Geobacillus were characterized including: Anoxybacillus flavithermus (Bolton et al, 1997; Tawil et al. 2012; Agüloğlu Fincan et al. 2014; Ozdemir et al. 2015, 2016a), Anoxybacillus amylolyticus (Poli et al. 2006) and Anoxybacillus caldiproteolyticus D504 and D621 (Ozdemir et al. 2016b), Anoxybacillus spp. KP1, SK3-4 (ASKA), DT3-1 (ADTA), TSSC-1, IB-A, AH1, and YIM 342 (Chai et al. 2012; Kikani and Singh 2012; Hauli et al. 2013; Matpan and; Güven 2014; Acer et al. 2015; Zhang et al. 2015) in addition to Geobacillus thermoleovorans MTCC 4220 (Gt-amyII), CCB_US3_UF5 (GTA), YN, NP54 (Berekaa et al. 2007; Rao and Satyanarayana 2007; Mok et al. 2013; Mehta and Satyanarayana 2014), G. thermoleovorans subsp. stromboliensis Pizzo (amyA), (Finore et al. 2011), Geobacillus thermodenitrificans HRO10 (Ezeji and Bahl 2006), and Geobacillus sp. IIPTN (Dheeran et al. 2010).

Despite the accumulation of several α-amylase sequence data, isolation and characterization studies, it is worth mentioning that most of these endospore-forming bacilli-originated enzymes have not been validly assigned into any of the current 42 GH13 subfamilies until now, except the presence of α-amylases belonging to Bacillaceae in GH13_5, 19, 28 and 36 subfamilies. Nevertheless, some novel subfamilies containing these α-amylases from Bacillaceae members were proposed but still non-defined. In chronological order, the three undefined bacilli GH13 subfamilies sharing two conserved tryptophan residues (W200-W201, E184aa-A numbering) which are unexceptionally positioned between the loop 3 and β4 strand of the catalytic TIM-barrel structure, were proposed as (i) the mesophilic group including Bacillus aquimaris (BaqA), B. coahuilensis, Bacillus sp. SG-1 and NRRL B-14,911 α-amylases (Puspasari et al. 2013); (ii) the ASKA and ADTA, BaqA, GTA, amyA, and Gt-amyII amylases (Ranjani et al. 2015; Janeček et al. 2015; Sarian et al. 2017); and finally (iii) the non-defined BmaN1 GH13 subfamily which includes the mesophilic enzyme of Bacillus megaterium NL3 having an atypical catalytic triad (Sarian et al. 2017). The thirdly proposed subfamily discriminated from the other two preceding ones by the absence of a complete catalytic machinary. Among these bacilli α-amylases, the X-ray crystal structures of Anoxybacillus sp. SK3-4 α-amylase (TASKA, PDP ID: 5A2B; it is a truncated form of ASKA) and G. thermoleovorans GTA α-amylase (PDB ID: 4E2O) were successfully unveiled (Mok et al. 2013; Chai et al. 2016).

In this study, a total of 15 isolates and reference strains from genus Anoxybacillus were screened for their α-amylase activity and some of them were found to exhibit very high level of thermostable enzyme activities, which possess potential in biotechnological processes and may satisfy the industrial demands. After the PCR amplification and sequence determination of these Anoxybacillus α-amylase genes, their putative protein sequences were subsequently used in BLAST search, phylogenetic analyses and amino acid sequence alignments along with their most closely related sequences belonging to Bacillaceae family members. Phylogenetic analyses clustered these 15 sequences together with the other thermostable Anoxybacillus α-amylases of ASKA, ADTA and GSX-BL and with some putative Anoxybacillus α-amylase sequences, but not with genus Geobacillus or Bacillus related GTA, BaqA, amyA or Gt-amyII enzymes as proposed previously. When all 15 Anoxybacillus sequences and their homolog Bacillaceae family related α-amylase sequences were considered as a whole, the family branched into five separate clusters which exhibit a novel and three reorganized GH13 subfamilies in addition to the undefined “xy” labeled subfamily containing BmaN1 (Sarian et al. 2017). These four representative sequences from the newly proposed or rearranged GH13 subfamilies were subjected to further secondary and tertiary structural comparison analyses via generally used in silico techniques based on their detailed domain and surface structures and maltose binding sites. The comparison of these new Anoxybacillus α-amylase sequences with a wide sequence collection, containing the other endo-spore forming bacilli sequences, depicted directly the evolutionary history of α-amylases from Bacillaceae family as all these bacilli-related clusters share some common features. Moreover, this approach allowed us to group Bacillaceae α-amylases under more accurate groups based on their taxonomicaly related genera, temperature optima, sequence features and their related phylogenetic analyses. In conclusion, we proposed a novel bacilli α-amylase GH13 subfamily in addition to the division of the previously proposed, but still unassigned GH13 subfamily into individual subfamilies, which contained ASKA and ADTA, BaqA, GTA, amyA, and Gt-amyII α-amylases (Ranjani et al. 2015; Janeček et al. 2015; Sarian et al. 2017) originated from different taxonomic groups: the genus Anoxybacillus, Geobacillus, or Bacillus.

Materials and methods

Bacterial isolates and reference strains

The bacterial isolates of Anoxybacillus spp. A321, A3210, D222b, E184aa, E184ab and E208a as well as the reference strains of Anoxybacillus salavatliensis DSM 22626T, Anoxybacillus gonensis NCIMB 13933T, Anoxybacillus ayderensis NCIMB 13972T, Anoxybacillus kestanbolensis NCIMB 13971T, Anoxybacillus kamchatkensis DSM 14988T, Anoxybacillus flavithermus DSM 2641T, Anoxybacillus amylolyticus DSM 15939T, Anoxybacillus thermarum DSM 17141T and Anoxybacillus kamchatkensis subsp. asaccharedens DSM 18475T were used for the α-amylase assays, gene amplifications and phylogenetic analyses. The isolation procedures, characterization studies and 16S rRNA gene sequencing analysis of these bacilli were determined as previously described by Cihan et al. (2014a.)

Qualitative and quantitative α-amylase assays

The isolates and reference strains were screened for their amylolytic activity on Medium-I agar plates containing 1% soluble starch (Suzuki et al. 1976) upon incubation for 24 h at 55 °C. Then, the plates were treated with 0.2% I2 dissolved in 2% KI solution, and the halo zones were measured around the colonies in order to determine the amylolytic enzyme producing ones. Geobacillus stearothermophilus DSM 22T and Bacillus amyloliquefaciens DSM 7T were used as the α-amylase producing references for comparison. Before determining the quantitative α-amylase activity for enzyme production, the bacilli were incubated in a modified Santos and Martins (2003) broth (1.0% tryptone, 0.5% yeast extract, 1.0% soluble starch), (Santos and Martins 2003) by shaking at 150 rpm during 72 h. The incubation temperature values (55–65 °C) and pH of the media (pH 7.0–8.0) were adjusted according to the bacteria. Extracellular α-amylase activity was carried out using 3,5-dinitrosalicylic acid (DNS) method with some slight modifications (Miller 1959). The standard reaction mixture contained 0.5 ml of each 0.2 M sodium phosphate buffer, 2% soluble starch and appropriately diluted enzyme solution. The effects of temperature and pH on α-amylase activities were determined as previously described (Ozdemir et al. 2016b). The reactions were carried out at the required optimal pH and temperature values of bacteria for 10 min and stopped by boiling 5 min after addition of 1 ml DNS. Finally, the absorbance was measured spectrophotometrically at 540 nm. One unit of α-amylase activity was defined as the amount of enzyme that catalyzed the liberation of reducing sugar equivalent to 1 µmol of maltose per min under the assay conditions. The millimolar extinction coefficient was calculated using maltose as the standard. Total amount of protein was determined by the Lowry method (1976) using Bovine Serum Albumin (Lowry and Tinsley 1976). The enzyme assays were performed at least three times.

Amplification of α-amylase genes

Cultures grown on Medium-I plates for 18 h at 55 °C were used for genomic DNA extraction (Fermentas K0512, Genomic DNA purification kit). The α-amylase genes were amplified by PCR using the protocol of Chai and colleagues with some modifications (Chai et al. 2012). The PCR conditions were optimized according to the primers’ annealing temperature (Tm) as 55 or 58 °C and by adjusting the MgCl2 concentration as 2.0 or 2.5 mM which varied for the bacilli used. The PCR products were purified with GeneJET PCR Purification kit (Fermentas K0702) and sequenced by using an ABI 3100 gene sequencer with a Bigdye cycle sequencing kit (Macrogen, Europe).

Sequence collections

The entire protein sequences, deduced from the Anoxybacillus nucleotide sequences, were taken into the query sequence analysis in the BLASTN and BLASTP programs (Altschul et al. 1997). In collecting the sequences, the criteria, proposed by Stam et al. (2006) were used for the identification of distinct subfamilies sufficiently. The caught sequences from the blast search contained the most similar sequences sharing high similarities, appearing at the top on the blast report, displaying a slow and progressive increase in E-values as well as belonging to a closely related taxonomic group from which the sequences were obtained (Stam et al. 2006). These sequences were also checked for the presence of (i) the (β/α)8-barrel architecture (Janeček 2002; Hostinová et al. 2010), (ii) the signal sequences, (iii) all the seven CSRs of the GH13 α-amylase family (Janeček 2002; Janeček et al. 2015), (iv) the catalytic triad, (v) the calcium ion binding sites from 1 to 4 (Mok et al. 2013; Chai et al. 2016), (vi) the substrate binding subsites (Chai et al. 2016), (vii) a pair of tryptophan (W200,W201, E184aa-A) between the CSR-V (loop 3) and CSR-II (β4 strand), (Puspasari et al. 2013), (viii) the consecutively repeated aromatic motifs of phenylalanine and tyrosine residues at the end of the C-terminal segment (Mok et al. 2013; Janeček et al. 2015), (ix) the signature residues with the invariable consecutive lysine-arginine (KR) at the terminus of domain C, and (x) the 4 residues involving in the formation of putative S1 and S2 transmembrane regions. Therefore, in addition to the 15 newly introduced amino acid sequences in this study, additional 78 sequences were retrieved directly from the Universal Protein Knowledgebase (UniProt Consortium 2017), the GenBank (Benson et al. 2014), and the Protein Data Bank (PDB), (Berman et al. 2000) or from the annotated whole genome projects found in the NCBI-genome databases (https://www.ncbi.nlm.nih.gov/genome/). The sequence sets also covered all the well-defined GH13 subfamilies in the CAZy database, having α-amylase specificity including subfamilies GH13_1, 5, 6, 7, 15, 19, 24, 27, 28, 32, 36, 37, 39, 41, 42 (Stam et al. 2006; Cantarel et al. 2009; Lombard et al. 2014), the unassigned cyclomaltodextrinase from Flavobacterium sp. No 92 (GH13_??; (Fritzsche et al. 2003)), and the formerly suggested bacilli related GH13 subfamilies the first one akin to BaqA α-amylase by Puspasari and colleagues (2013), the second one including BaqA, ADTA, ASKA, Gt-amyII and GTA α-amylases by Chai et al. (2012), Chai et al. (2016), Janeček et al. (2015), Sarian and colleagues (2017), and lastly the GH13_xy around BmaN1 by Sarian et al. (2017). So, totally 93 identified enzymes and hypothetical proteins were studied with in silico techniques. The list of all the polypeptide sequences used for this study, their accession numbers from UniProt (UniParc), (UniProt Consortium 2017) and GenBank (Benson et al. 2014) databases, their amino acid lengths, the results of the specific α-amylase activity experiments, and their related classified or newly proposed GH13 subfamilies from a1 to a4 were presented in Supplementary file, Table SI.

Bioinformatics analysis

As in the case of previously published bacilli related new α-amylase subfamily establishment studies, similar, analogous or more comprehensive bioinformatic tools were used by comparing wider and more meaningful amylolytic enzyme sequence sets in this study (Chai et al. 2012, 2016; Puspasari et al. 2013, Ranjani et al. 2015; Janeček et al. 2015; Sarian et al. 2017). The conserved domain types and probable families of 15 new Anoxybacillus originated deduced protein sequences were searched with Pfam (http://pfam.xfam.org/; Finn et al. 2016) and NCBI-Conserved domain databases (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi; Marchler-Bauer et al. 2017). The molecular weights and the isoelectric points of these sequences were predicted using Geneious R10 program (http://www.geneious.com; Kearse et al. 2012). The secondary structures including the α-helices and β-strands, the sugar binding pockets, and the transmembrane residues were predicted by Phyre2 server (Protein Homology/AnalogY Recognition Engine), (http://www.sbg.bio.ic.ac.uh/phyre/; Kelley et al. 2015). The signal peptides from transmembrane associated regions were defined using SignalP 4.0 server (http://www.cbs.dtu.dk/services/SignalP/; Petersen et al. 2011). The extracted sequences containing domains A and B of the compact (β/α)8 TIM-barrel structure in addition to domain C were subjected to a multiple sequence alignment using the CLUSTAL-OMEGA program (https://www.ebi.ac.uk/Tools/msa/clustalo/; Sievers et al. 2011). The similarities were maximized by a manual fine adjustment in order to align the individual CSRs and the borders of CSRs were determined from previous studies (Janeček 2002; Janeček et al. 2015). For the currently organized and suggested 4 novel α-amylase GH13 subfamilies from a1 to a4, the individual sequence logos were created for their seven CSRs and two adjacent aromatic tryptophan motifs typical for the novel subfamilies, using the WebLogo 3.5.0 server (http://weblogo.berkeley.edu; Crooks et al. 2004). All phylogenetic analysis containing the GH13 family members was performed with using the Geneious R10 server. The evolutionary distance matrix and its related phylogenetic tree were constructed using the UPGMA algorithm (Sokal and Michener 1958) with the bootstrap values selected based on 1000 replications (Felsenstein 1985). Additionally, the phylogenetic analyses on 16S rRNA genes of the thermophilic bacilli members from genera Anoxybacillus and Geobacillus were carried out as described previously by Cihan et al. (2014a).

The three-dimensional (3D) structural models of four representative sequences belonging to each of the four suggested subfamilies were constructed using the SWISS-MODEL (http://swissmodel.expasy.org/; Biasini et al. 2014) program server. Models were compared with the template X-Ray structures of Anoxybacillus sp. SK3-4 α-amylase (TASKA, PDB ID: 5A2B), (Chai et al. 2016) and G. thermoleovorans CCB US3 UF5 α-amylase (GTA, PDB ID: 4E2O), (Mok et al. 2013). The predicted 3D models were further visualized and drawn with PyMOL software (The PyMOL Molecular Graphics System, version 1.7.4 Schrödinger, LLC; http://www.pymol.org) and ICM-BrowserPro version 3.8-5 (MolSoft LLC, La Jolla, CA, USA; http://www.molsoft.com/icm_browser_pro.html) programs. RMSD (Å) values were calculated using PyMOL. All computational analyses were performed using Intel® Core™ i5-4570 CPU@ 3.20 GHz processor and 16 GB RAM on Windows7 Enterprise 64x environment.

Results

Extracellular α-amylase activities

The α-amylase screening assay revealed that 15 bacilli were positive for the halo zones formed around their colonies on plates containing soluble starch, but especially the isolates E184aa, E184ab, E208a, D222b and the reference strains of G. stearothermophilus, A. flavithermus, A. salavatliensis and A. amylolyticus produced relatively higher clear zones in diameter. In the quantitative α-amylase assay, extracellular enzyme activities of the bacilli were found to vary from 0.8 U/g to 341.3 U/g (Fig. 1). The ones producing higher zones also displayed the highest amylolytic enzyme activities. It is also noteworthy that at 65 °C, pH 7.0, the measured enzyme activity values of E184aa, E184ab and D222b isolates were the best among the bacilli and the amount of α-amylases produced by these isolates were 1.5 to twofold higher than the control G. stearothermophilus DSM 22T which was known as biotechnologically important.

Fig. 1
figure 1

Extracellular α-amylase activities of the bacilli when 1% of soluble starch was used as the substrate. The isolates and reference strains as well as their optimal temperature and pH values, used during determination of the extracellular α-amylase activities, are as follows: Anoxybacillus sp. E184aa (65 °C, pH 7.0), Anoxybacillus sp. E184ab (65 °C, pH 7.0), Anoxybacillus sp. D222b (65 °C, pH 7.0), G. stearothermophilus DSM 22T (65 °C, pH 7.0), Anoxybacillus sp. E208a (65 °C, pH 7.0), A. flavithermus DSM 2641T (65 °C, pH 7.0), A. salavatliensis DSM 22626T (65 °C, pH 7.0), (A) amylolyticus DSM 15939T (65 °C, pH 7.0), (B) amyloliquefaciens DSM 7T (37 °C, pH 7.0), A. kestanbolensis NCIMB 13971T (55 °C, pH 8.0), Anoxybacillus sp. A3210 (65 °C, pH 7.0), Anoxybacillus sp. A321 (65 °C, pH 7.0), A. gonensis NCIMB 13933T (55 °C, pH 7.5), A. ayderensis NCIMB 13972T (55 °C, pH 8.0), A. kamchatkensis subsp. asaccharedens DSM 18475T (55 °C, pH 7.5), A. thermarum DSM 17141T (60 °C, pH 7.0), and A. kamchatkensis DSM 14988T (60 °C, pH 8.0)

Full size image

Molecular characterization of the nucleotide sequences

The PCR amplicon sizes of the 15 Anoxybacillus were between 1515 and 1518 nucleotides which were compatible with the length of known α-amylase gene sequences, (1518 bp for only A. flavithermus and A. amylolyticus). The amplicons had the DNA G + C contents varying from 42.1 to 43.8 mol%. All the sequenced α-amylase genes were deposited in GenBank database (Benson et al. 2014) under the accession numbers of KY426431 (E184aa), KY426432 (E208a), KY426433 (E184ab), KY426434 (D222b), KY426435 (A3210), KY426436 (A321), KY426437 (A. salavatliensis DSM 22626T), KY426438 (A. gonensis NCIMB 13933T), KY426439 (A. ayderensis NCIMB 13972T), KY426440 (A. kestanbolensis NCIMB 13971T), KY426441 (A. kamchatkensis DSM 14988T), KY426442 (A. flavithermus DSM 2641T), KY426443 (A. amylolyticus DSM 15939T), KY426444 (A. thermarum DSM 17141T) and KY426445 (A. kamchatkensis subsp. asaccharedens DSM 18475T). The nucleotide sequence query in BLASTN showed that the open reading frames (ORFs) of these novel sequences displayed 91.0–99.0% gene sequence similarities with α-amylases of ASKA, ADTA, and GSX-BL. The CDS (Conserved Domain Search) analysis, based on the annotations of the subfamily domain architectures, also revealed that they all belong to AmyAc_bac_CMD_like_2 (cd11339) Conserved Protein Domain Family, which is identified as the α-amylase catalytic domain in bacterial cyctomaltodextinases and the related proteins. Then the nucleotide sequences were converted into deduced amino acid sequences of polypeptides that composed of 504 to 505 amino acids in order to use in further phylogenetic analysis and in silico techniques.

The query results and the phylogenetic analysis

The BLASTP search query using 15 Anoxybacillus amino acid sequences revealed 48 similar putative or biochemically characterized α-amylases belonging to endospore-forming from Bacillaceae family. Sequence collection of GH13 α-amylases used in this study was given in the Supplementary file, Table SI. Interestingly, all of these closely related homologue α-amylases have not been classified properly into a definite GH13 subfamily yet. Therefore, these new Anoxybacillus α-amylase sequences were aligned not only with their 48 related unclassified homologues, but also along with other 30 currently classified prokaryotic and eukaryotic α-amylases under well-defined GH13 subfamilies. Surprisingly, the phylogenetic analysis of those 93 α-amylase amino acid sequences showed that all the 15 Anoxybacillus α-amylase sequences from this study are clustered with other Anoxybacillus α-amylases such as ASKA, ADTA, and GSX-BL, but not with BaqA or GTA. The 63 bacilli α-amylases, including the novel sequences were all originated from endospore-forming bacilli by creating 5 distinct phylogenetic clades, as can be seen in the evolutionary tree in Fig. 2. In the similarity matrices, deduced α-amylase protein sequences of the 15 Anoybacilli displayed 95.1–100% sequence homology to each other. While the protein sequences of A321, E184aa, E184ab, E208a, A3210 and A. gonensis α-amylases completely resembled to each other, the most distinct protein sequence was from A. kestanbolensis (95.1%-97.6%). Based on the protein sequence similarities and the evolutionary relations, four differently proposed subfamilies, which were not belonging to any of the currently classified GH13 subfamilies, were inferred from the phylogenetic analyses. The bootstrap values of a1 to a4 clades were higher than 70% which verified their brach positions in the phylogenetic tree. The sequence identity rates among the members were given in Supplementary Table SII. The firstly proposed subfamily a1 was composed of 27 α-amylases from genus Anoxybacillus including our 15 sequences, Anoxybacillus sp. DT3-1, SK3-4, and GSX-BL enzymes and other 9 putative Anoxybacillus α-amylases annotated in whole genome sequences. The a1 clade showed protein sequence similarities between 72.4% and 100% among the group, having the lowest similarity to BCO1 α-amylase. The a1 clade was mostly related with a second a2 α-amylase cluster containing other four hypothetical Anoxybacillus α-amylases deduced from annotated genome sequences of Anoxybacillus sp. B2M1, B7M1, P3H1B, and UARK-01 as well as with a third a3 cluster composed of 19 endospore-forming, thermophilic bacilli genera from Geobacillus containing GTA, Parageobacillus in addition to genus Anoxybacillus. The members of a2 and a3 clade displayed sequence similarities ranging from 89.2 to 100% and from 74.9 to 100% within their own groups, respectively. All the α-amylases from mesophilic genus Bacillus formed a fourth a4 cluster with 9 members including BaqA which showed 58.3–99.4% similarity values with each other. Finally, the unclassified but proposed four representative aberrant α-amylases from genus Bacillus clustered in a different clade (GH13_xy by Sarian et al. 2017) which diverged from other mesophilic bacilli from a4 clade around the BaqA α-amylase.

Fig. 2
figure 2

The evolutionary tree of the GH13 α-amylase family obtained from 93 protein sequences. The newly proposed a1 (red), a2 (blue), a3 (green) and a4 (orange) members as well as the GH13_xy members (purple) all belonged to Bacillaceae family and none of these amylases have been assigned to any GH13 subfamilies till now. After GH13_ subfamily indication, mostly the GenBank accession numbers, but if not available, the UniProt (UniParc) numbers are given at the end of the species epithet. “*” symbol near the accession numbers display α-amylase gene sequences obtained from this study

Full size image

Surprisingly, two of the α-amylases from A. tepidamans DSM 16325T and A. geothermalis ATCC BAA 2555T, sharing 75.1% similarity, were clustered together with the members of the genus Geobacillus from a3 instead of their original Anoxybacillus a1 and a2 clades. A. tepidamans α-amylase (ATA) was most closely related to P. thermantarcticus M1 (80.8%), G. thermoleovorans MTCC 4220 (Gt-amyII), (76.7%) and G. thermoleovorans CCB US3 UF5 (GTA), (76.2%) α-amylases, whereas A. geothermalis enzyme was mostly related to to Geobacillus sp. LC300 (98.6%) and G. thermoleovorans CCB US3 UF5 (GTA), (94.0%) α-amylases were clustered within the a3 clade. Interestingly, the results of 16S rRNA gene sequence analysis on all the described Anoxybacillus species revealed that members of this genus, harbouring 23 species and 2 subspecies, divide into two phylogenetically diverge branches that share high 16S rRNA gene homologies only within their related groups (Euzéby and Parte 1997; Gul-Guven et al. 2008; Deep et al. 2013). A. tepidamans DSM 16325T (FN428691) and A. geothermalis ATCC BAA2555T (KJ722458) species were clustered together with the second group members of A. vitaminiphilus JCM 16594T (FJ474084), A. calidus DSM 25220T (FJ430012), A. contaminans DSM 15866T (AJ551330), A. voinovskiensis DSM 17075T (AB110008), A. amylolyticus DSM 15939T (AJ618979), A. rupiensis DSM 17127T (AJ879076) and A. caldiproteolyticus DSM 15730T (FN428698). 16S rRNA gene sequences of this second group shared lower sequence homology (< 97.0%) with all the other type species from genus Anoybacillus, whereas there was a strict requirement for DNA hybridization analysis among each other for their species identifications. This second branch arised based on the analysis 16S rRNA genes, which encloses A. tepidamans and A. geothermalis species (98.2% homology to each other), was also seemed to be more closely related to genus Geobacillus than the first 16S rRNA gene group of Anoxybacillus group. In a recent study of Bezuidt et al. (2016) dealing with the comparative analysis on conserved core and flexible genes of 61 Geobacillus, Anoxybacillus and Bacillus genome sequences, A tepidamans PS2 (BioProject Accession: PRJNA214279) clustered within the genus Geobacillus, whereas other 12 Anoxybacillus species branched on their own clustures. They also suggested that A. tepidamans PS2 should be regarded as a species in Geobacillus based on their shared genes and Average Nucleotide Identitity values (Bezuidt et al. 2016).

Beside these findings, the alignment of Anoxybacillus α-amylases with their other counterparts clearly displayed the presence and varieties of the seven CSRs from regions I to VII (Janeček 2002), the catalytic triad, the common (β/α)8-barrel fold containing domains A, B and C as well as a pair of tryptophans in the α3 helix between CSR-V and CSR-II (Mok et al. 2013; Puspasari et al. 2013) within these phylogenetic groups (Fig. 3). The aligned 93 sequences classified under 21 groups and, it was noteworthy that five of the clades including a1, a2, a3 and a4 (totally 59 sequences) and xy (representative 4 sequences) were all composed of endospore formers from Bacillaceae family and shared similarities in terms of these general features above in congruence with the results of phylogenetic analyses.

Fig. 3
figure 3

The α-amylase sequence alignments of seven CSRs (I-VII) belonging to the newly proposed a1, a2, a3 and a4 novel GH13 subfamilies with GH13_xy BmaN1 group, the unassigned cyclomaltodextrinase from Flavobacterium sp. No 92 (GH13_??) and the other well-defined GH13 subfamilies displaying α-amylase specificity including subfamilies GH13_1, 5, 6, 7, 15, 19, 24, 27, 28, 32, 36, 37, 39, 41, and 42. The representative members of a1, a2, a3 and a4 groups are highlighted with gray. +, characteristic alanine (A); Δ, the catalytic triad (D/E/D); !, invariable arginine (R). 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

Full size image

Since the 15 Anoxybacillus putative proteins obtained from this study were all clustered within the phylogenetic group a1, and the highest amylolytic enzyme activity was measured in Anoxybacillus sp. E184aa, E184aa-A (E184aa α-amylase) was chosen as the representative member of a1 clade within the other 26 sequences. Among thermostable α-amylases from Anoxybacillus sp. B2M1, B7M1, P3H1B and UARK-01, only the amylolytic enzyme activities of B2M1 and B7M1 were known as positive (Filippidou et al. 2016), therefore, B2M1-A (B2M1 α-amylase) was selected as the representative sequence of a2 clade for further structural investigations. A. tepidamans DSM 16325T α-amylase (ATA α-amylase) was chosen as the representative member of the a3 clade which contained 19 sequences belonging to three different genera from thermophilic bacilli. Finally, Bacillus aquimaris strain MKSC 6.2 α-amylase (BaqA α-amylase) was selected for being the representative member of the nine α-amylases from a4 clade, all the members of which were from the mesophilic genus Bacillus. A. tepidamans DSM 16325T (Coorevits et al. 2012), and Bacillus aquimaris MKSC 6.2 (Puspasari et al. 2013) were also known as amylolytic strains from previous studies.

The alignment results revealed the importance of CSR-V in the loop 3 which shares the conserved motif of LPDLx and a pair of tryptophans between loop 3 and β4 strand among the enzymes belonging to these Bacillaceae members (Janeček et al. 2015; Ranjani et al. 2015). All the thermostable α-amylases from a1, a2 and a3 clades had an alanine residue (A184, E184aa-A numbering) at the end of LPDLx signature, whereas mesophilic members contained an asparagine residue (N185, BaqA) at this site. The two consecutive tryptophans were also shared among these Bacillaceae members only with an exception in (A) flavithermus subsp. yunnanensis containing W200 and R201. By combining these phylogenetic analyses and alignment results, the specific CSRs logos defining the currently proposed or revised GH13 subfamilies of a1, a2, a3, and a4 were created and presented in Fig. 4. Among these logos, the catalytic triad which consists of aspartic acid (D213∆) residue serving as the catalytic nucleophile in CSR-II, glutamic acid (E242∆) playing the proton donor role in CSR-III, the transition-state stabilizer aspartic acid (D310∆) in CSR-IV, and the invariable arginine (R211) in CSR-II were all conserved except in members from a2 subfamily (E184aa-A numbering). Nevertheless, the logo of a2 was the most different among the common CSRs motifs of α-amylases from other well-known subfamilies as well as a1, a3, and a4 groups. B2M1-A, B7M1-A, P3H1B-A and UARK-01-A members from a2 clade, harboured an abnormal catalytic machinery as in the case of an atypical α-amylase GH13_xy subfamily (Sarian et al. 2017). Solely, the catalytic nucleophile aspartate (D215∆, B2M1-A) was preserved in a2 clade, but the proton donor of E242∆ (E184aa-A) was replaced with a glycine G244∆ (B2M1-A) residue in CSR-III (β5 strand) like in the Bacillus sp. 2 A57 CT2 from GH-13_xy (Sarian et al. 2017). But there was also an aspartate residue (D246, B2M1-A), shifted two positions downstream, which might be the potential catalytic nucleophile instead of usual glutamate. Moreover, in CSR-IV residing on β7 strand, the transition-state stabilizer aspartate replaced with a strange arginine (R312, B2M1-A) residue similar to the Bacillus sp. 1NLA3E α-amylase from aberrant GH13_xy subfamily. However, there was also an aspartate residue (D311∆, B2M1-A), which was shifted one position upstream to this abnormal arginine. In addition, although the invariant arginine (R211, E184aa-A) in CSR-II was fully conserved among all GH13 subfamilies except almost all members of GH13_xy, it was changed to a tyrosine (Y213, B2M1-A) residue as in the case of BmaN1 α-amylase of (B) megaterium NL3 belonging to the atypical subfamily.

Fig. 4
figure 4

Sequence logos made with regard to the CSRs of the four newly proposed GH13 subfamilies from a1-a4 all belonging to Bacillaceae family. The positions of CSRs regions from I-VII, two consecutive tryptophans in addition to the catalytic triad (Δ) are also presented. The logos were created by using 27, 4, 19, and 9 protein sequences for the subfamilies a1, a2, a3 and a4, respectively. The residues are numbered according to their representative α-amylase sequences: E184aa-A (a1), B2M1-A (a2), ATA (a3), and BaqA (a4)

Full size image

In silico analyses on the secondary structures

The deduced polypeptides of 15 Anoxybacillus sequences contained 504 to 505 amino acids starting with a 23 residues long putative signal peptide. The predicted molecular weight and pI values of these enzymes ranged between 58.7 and 59.0 kDa and 5.98–6.26, respectively. From this point, totally 59 sequences from endo-spore forming bacilli which comprised the currently suggested a1, a2, a3, and a4 clades, including the 15 Anoxybacillus α-amylases, were used for further in silico secondary and tertiary structure analyses. The characteristic features of these proposed GH13 subfamilies obtained from in silico techniques were also presented in Supplementary file, Table SII. The highest proline contents, predicted molecular weights and pI values of these enzymes were observed in a3, a4 and a2, respectively. The only exception was the pI value of ATA which increased the maximum limits of a3 from 7.91 to 9.31. All the members of a1, a2, a3 and a4 had also signal peptide sequences aligning between residues 1 and 23. The bacilli amylolytic enzymes were predicted to be transmembrane proteins as they shared two putative components embedded in the hydrophobic membrane: S1 and S2, which crossed the membrane from cytoplasmic to extracellular sides with four responsible amino acid residues (K330, Y346, I482 and F498, E184aa-A numbering) in their helix structure. An example of transmembrane helix image, created for E184aa-A α-amylase, was shown in Supplementary file, Fig. S1.

As deduced from topological alignment of primary and secondary (2D) structures of the representative members from the a1-a4 clades (with GTA from a3), all the enzymes consisted of three domains typical for α-amylases from GH13 family (MacGregor et al. 2001). The catalytic domain A containing the exact (β/α)8-barrel structure, the domain B connecting the β3 strand, and α3 helix, and finally the domain C succeeding domain A, which contain eight antiparallel β-sheets were all shared by enzymes of these a1, a2, a3 and a4 clades. The predicted signal sequences, β-strand and α-helix numbering, the catalytic triad, the CSRs, and additionally the possible pockets for substrate-binding were all displayed on the 2D structures of E184aa-A, B2M1-A, ATA, and BaqA obtained by Phyre2 server (Supplementary file, Fig. S2a-S2d).

When all these data obtained from CSRs sequence alignments, Phyre 2D topological alignments from this study as well as the previous experimental analyses of the most related α-amylases of ASKA (Chai et al. 2016), GTA (Mok et al. 2013) and BaqA (Puspasari et al. 2013) were considered, the entire sequences of 63 α-amylases belonging to Bacillaceae family were aligned eventually in order to compare the entire amino acid residues forming the signal sequences, the catalytic triad, the CSRs, the possible calcium and sugar binding residues, the possible sugar pockets for substrate specificities or sugar recognitions, the transmembrane helix regions, the tyrosine and phenylalanine repeats and the consecutive lysine and arginine residues at the end of C termini. However, only the sequence alignments of the 2 representatives of each proposed subfamilies that are E184aa-A and ASKA (a1), B2M1-A and P3H1B-A (a2), ATA and GTA (a3), BaqA and MKU004-A (a4) α-amylases were presented in Fig. 5 to demonstrate these characteristic properties.

Fig. 5
figure 5

The comparison of primary structures of four newly proposed subfamilies, presenting two representatives from each group. Colour code: (1) Conserved sequence regions (CSR I-VII) are highlighted in gray. (2) Residues for Ca1, Ca2, Ca3 and Ca4 binding sites are highlighted in yellow and marked by red, blue, green and black (*) asterisks, respectively. The conserved amino acids in calcium binding sites are indicated with yellow. (3) Residues involved in maltose binding are indicated by blue, and the related maltose binding pockets are abbreviated as F, W1, H-YW, H–Y, W2, LF-L, R-DTVKH, E-W, L-Y / V-F+ (+; only found in a2 and a4), HDTV, I-Y and ED–NR. (4) The catalytic triad are signified by red triangles (Δ). (5) The invariantly conserved position of the arginine in the CSR-II is highlighted in turquoise and marked by a hashtag (#). (6) Residue A in CSR-V, only found in thermophilic groups is indicated with (!). (7) The two adjacent characteristic tryptophans, positioned between CSR-V and CSR-II, are highlighted in pink. (8) The seven conserved tyrosine and phenylalanine residues at the C terminus are highlighted in turquoise. (9) The invariable KR residues at the C-termini are indicated in dark yellow. (10) The putative transmembrane regions and their related residues are indicated with violet and abbreviated as S1C:S1 region-cytoplasmic, S1E: S1 region-extracellular, S2C: S2 region-cytoplasmic, and S2E: S2 region-extracellular

Full size image

In our phylogenetic investigations, ASKA and GTA thermostable α-amylases grouped under proposed a1 and a3 subfamilies. Their calcium and maltose binding sites as well as the other related residues were previously examined in detail by X-ray crystal structure analyses (Mok et al. 2013; Chai et al. 2016). Therefore, the amino acid residues associated with the 2 calcium binding sites of GTA and four sites of TASKA were screened for the newly suggested four subfamily members. As can be seen in the alignment of the representative sequences in Fig. 5, the residues involved in the formation of Ca1, Ca2 and Ca4 calcium binding sites, detected in the crystal structures of GTA and TASKA, were mostly conserved among the proposed a1, a3 and a4 subfamilies. Nevertheless, there were some amino acid substitutions, which were expected to have relatively minor effects on calcium binding like neutral (E/D) transitions or N/D changes from positive to negative charge the latter of which may increase the binding affinity. But the most critical changes from negative to positive charge were also detected as D/P, E/P, E/A, and E/Q mutations, pocessing high possibility of decreasing the sensitivity and affinity for calcium ion (Tien et al. 2014). The conversion of E173/Q174 (E184aa-A/BaqA) in Ca1 sites of a4 members were observed as mutations that may have an effect on binding of calcium ion. The amino acid substitutions of N46/D48 (E184aa-A/in all a2, a3 and a4 members) in Ca2 sites, E109/D111 (E184aa-A/in all a2, a3 and a4 members) and E110/P112 (E184aa-A/B2M1-A and ATA numbering) residues in Ca3 sites, and additionally E283/A285 (E184aa-A/B2M1-A) and E283/D284 (E184aa-A/BaqA) in Ca4 sites were also noticed.

The secondary structure analysis of the four representative members using Phyre2 server also presented some hints, which might be associated with sugar binding pockets. These hints, which were not used in other studies before, were found to be very useful for further detailed structural analyses. When these hints were combined with the previous findings on maltose binding residues from GTA and TASKA, the conserved regions of CSR-II (β4), III (β5), and IV (β7) were thought to be important both in the catalytic activity and substrate binding. Whereas, the CSR-VI (β2), I (β3), V (loop 3), and VII (β8) regions probably play roles in enzyme specificity and substrate binding of these a1 to a4 subfamilies (Janeček 2002; Mok et al. 2013; Chai et al. 2016). Accordingly, the evaluation of 59 bacilli sequence alignments under 4 clades revealed 12 pockets having possible functions in enzyme activity and substrate specificity: (1) F pocket in β1 strand (in all), (2) W (in a1, a3 and a4) or R (in a2) pockets in CSR-VI, (3) H-YW (in a1, a3 and a4) or Q-KK (in a2) pockets downstream to CSR-VI, (4) H–Y (in a1 and a3) pocket in CSR-I, (5) W (in a1, a2 and a3) or F (in a4) pockets upstream to CSR-V, (6) LF-L in CSR-V (in a1), LN-L (in a2) or LY-L (in a3 and a4) pockets, (7) R-DTVKH (in a1), Y-DDAGH (in a2), R-DAMKH (in a3) or R-DTVRH (in a4) pockets in CSR-II, (8) E-W (in a1 and a3), G-D (in a2) or E-F (in a4) pockets in CSR-III, (9) L-Y (only in a2) and V-F (solely in a4) pockets downstream to CSR-III, (10) HDTV (in a1 and a3), DRTV (in a2) or HDME (in a4) pockets in CSR-IV, 11 I-Y (in all) pocket in CSR-VII and, 12) ED–NR (in a1 and a4), KA–NH (in a2) or ND–NR (in a3) pockets downstream to CSR-VII (Fig. 5 and Supplementary file, Table SII).

In addition to these calcium and maltose binding sites, some specific residues found at the end of the sequences also took attention when 2D structures and the whole sequence alignments were compared. At the C-termini, two novel consecutive tyrosine residues (Y457, Y469, E184aa-A numbering) were detected in addition to the previously described repeated aromatic motifs of tyrosine (Y489, Y497, E184aa-A numbering) and phenylalanine (F481, F492, F495, E184aa-A) by Janeček et al. (2015), which could be evaluated as additional stop signals in all 59 bacilli sequences. Moreover, the residues in helix structures associated with the formation of predicted S1 and S2 regions crossing the cell membrane were nearly conserved among all the sequences (Fig. 5 and Supplementary file, Table SII). Residues that combined S1 region spanning from extracellular environment (S1-E: K330, E184aa-A) to cytoplasm (S1-C: Y346, E184aa-A) were all preserved in a1-a4, whereas the amino acid residues conjoining S2 region from extracellular to cytoplasm varied in S2-E: I482/I485 (in a1/a2 and a3) or L485 (in a4), and S2-C: F498 (in a1), Y501 (in a2 and a3) or L501 (in a4). Finally, all these α-amylase sequences from Bacillaceae family were found to be ended with two conserved, consecutive and positively charged lysine and arginine (K501R502, E184aa-A) residues only with the exceptional RR or KK residues found in 5 of the α-amylases from a3 and a4.

Tertiary structure predictions

The crystal structure analyses of G. thermoleovorans GTA α-amylase (PDB ID: 4E2O) and Anoxybacillus sp. SK3-4 α-amylase (TASKA, PDB ID: 5A2B) from the members of the rearranged subfamilies of a1 and a3 were already investigated in detail by Mok et al. (2013) and Chai et al. (2016), respectively. The 3D models of the representative E184aa-A (a1), B2M1-A (a2), and BaqA (a4) α-amylases were predicted by SWISS-MODEL and visualized by PyMOL and ICM-Browser-Pro using the best template as TASKA among nearly 240 candidates. Only in the case of ATA (a3), the best template model was preferred as GTA α-amylase in terms of higher sequence identity and lower RMSD values to GTA (79.60%, 0.085 Å) rather than TASKA (77.88%, 0.673 Å). The calculated RMSD values, obtained from the structural alignments of E184aa-A, B2M1-A and BaqA models with TASKA template were 0.078 Å, 0.116 Å, and 0.153 Å, in addition to 0.085 Å RMSD value in the case of ATA model with GTA template. According to homology report, the sequence identities, coverage and QMEAN values of E184aa-A (96.92%, 0.90 and − 0.38), B2M1-A (69.45%, 0.89, -0.82), and BaqA (64.10%, 0.89, -1.73) to TASKA as well as ATA (79.60%, 0.88, -0.66) to GTA were all given in parenthesis, respectively. In Supplementary Fig S3a, the folded 3D structure models of E184aa-A, B2M1-A, ATA and BaqA were presented by giving the three domain structures arranged as domain A, B and C. The superimposed structures (green) of these four α-amylases both with TASKA (red) and GTA-II (blue) obviously pointed out the overall similarity of the catalytic (β/α)8-barrel structure exists in GH13 α-amylase family (MacGregor et al. 2001), (Supplementary file Fig. S3b). The surface views of these α-amylases were also displayed in Supplementary Fig. S3c, which depicted a big groove for the active site region associated with maltose binding.

The substrate binding sites

The four representative α-amylase models (green) were overlapped with TASKA (red) and the active site regions as well as the residues possibly associated with maltose binding were illustrated in Fig. 6. In all models, maltose bound to substrate interacting subsites of -1 and − 2 similar to TASKA. The substrate binding pockets, detected by the hints of the previous topological secondary structure analyses, revealed that these residues are directly interacting with sugar ring or the ones responsible for substrate specificity and stabilization. When the structural maltose binding region of E184aa-A α-amylase superimposed with TASKA, model, completely matched to the template as they were both proposed to be the members of a1 subfamily. The catalytic triad of E184aa-A, consisting of Asp-Glu-Asp triad, were seemed to be acting on sugar ring in subsite − 2. Among these triplets, D213∆ serve as a catalytic nucleophile, E242∆ is the proton donor and D310∆ act as transition state stabilizer. Moreover, invariable arginine, positioned in the β4 strand, was located at R211 in E184aa-A α-amylase. These conserved catalytic residues and invariable arginine were all preserved in ATA (D215∆, E244∆, D312∆ and R213) and BaqA (D214∆, E243∆, D311∆ and R212) as a common feature of GH13 α-amylase family. But two of the catalytic residues (E242∆/G244∆ and D310∆/R312∆ transitions) in addition to the invariable arginine (R211/Y213) were replaced in the case of B2M1-A like BmaN1 α-amylase. The various residues which may involve in maltose binding of E184aa-A, B2M1-A, ATA and BaqA α-amylases were listed in Supplementary Table SII.

Fig. 6
figure 6

Overlapping the active sites of the 3D models (green) of Anoxybacillus sp. E184aa (E184aa-A), Anoxybacillus sp. B2M1 (B2M1-A), A. tepidamans DSM 16325T (ATA) and B. aquimaris MKSC 6.2 (BaqA) amylases with TASKA template (red, PDB ID: 5A2B). Model and template residue numbers are coloured in black and red, respectively. All the superimpositions with TASKA-maltose (yellow) were bound to active site region at subsites − 1 and − 2

Full size image

In concordance with the secondary structure analysis of E184aa-A dealing with the detected pocket sites, the amino acid residues of H98, Y100 and W101 (H–YW pocket downstream to CSR-VI), H140 and Y143 (H–Y pocket in CSR-I), F178 (LF–L pocket in CSR-V), D358 and R362 (ED–NR pocket downstream to CSR-VII) were in close relationship with maltose by their side chains in subsite − 2, whereas amino acids of R211 and D213∆ (R-DTVKH pocket in CSR-II), E242∆ (E–W pocket in CSR-III), H309 and D310∆ (HDTV pocket in CSR-IV) served for the catalytic activity directly in the subsite − 1 of the sugar binding groove. Moreover, there were some additional residues which were probably reside in the substrate binding subsite − 1, not being in contact with maltose, as T214 and H217 residues in the R-DTVKH pocket, the aromatic side chains of F36 and W85 in the HDTV pocket, and W166 with H217 in the R-DTVKH pocket. Additionally, W244 which seemed to render the E-W pocket with its bulky side chain, and finally, the residues of I344 and Y346 which appeared to form a loop may have an importance in the substrate specificity.

In substrate binding groove of B2M1-A, despite many residual variations, the overall catalytic system was conserved as Anoxybacillus sp. B2M1 is known to possess amylolytic activity (Filippidou et al. 2016). The amino acid residues of Q100, K102, K103 (Q-KK), H142, S145 (H–S), N180 (LN-L), A360 and H364 (KA–NH) bound to sugar ring at subsite − 2, and Y213, D215∆, D216 (Y-DDAGH), G244∆ (G-D), D311, and R312∆ (DRTV) were all positioned in subsite − 1 of the maltose binding groove in B2M1-A. Other than the similar residues present in E184aa-A, the additional two loops between the residues L265, Y267 (downstream to CSR-III) and residues W168, N169, I174 (upstream to CSR-V) were only peculiar to B2M1-A (Supplementary Table SII). In the case of ATA and BaqA enzymes, most of the amino acid residues interacting directly or indirectly with sugar ring were more conserved similar to E184aa-A with some exceptions. Superimpositions of ATA and BaqA to TASKA revealed the Y143/P145, F178/Y180, T214/A216 amino acid residue modifications in ATA and aromatic Y143/P145, F178/Y179, W244/F245, W166/F167 residue changes in BaqA α-amylase. Besides, the residues of V264 and F266, (pocket downstream to CSR-III) and F167 were also only unique to BaqA.

Discussion

Thermostable α-amylases have been used in several industrial applications as they possess thermal stability to harsh industrial processes including elevated temperatures (Demirjian et al. 2001). Starch degradation, baking, brewing, production of glucose and fructose syrups, fruit juices, alcoholic beverages, papers, pharmaceuticals, α-amylase assay kits, detergents and textiles are the major areas of utilization for amylolytic enzymes in the industry (Klein et al. 1970; Vieille and Zeikus 2001; Van der Maarel et al. 2002; Gupta et al. 2003). Anoxybacillus species are thought to be widespread in thermal habitats rather than other Bacillaceae members (Deep et al. 2013), having heterogeneous intra-species 16S rRNA gene similarity values varying from 93.8 to 99.7% (from this study). Moreover, whole genome sequences of 27 Anoxybacillus were available on Genomes OnLine Database-GOLD v.6 (Mukherjee et al. 2017) and GenBank (Benson et al. 2014) databases now. We screened the starch hydrolysing activities of some newly isolated Anoxybacillus strains in this study, which could be suggested to novel starch hydrolysis applications. Although all the bacilli were found to be amylolytic, Anoxybacillus sp. E184aa, E184ab and D222b isolates stand out from the others by means of their α-amylase production capabilities. Additionally, α-amylase production capacity of A. salavatliensis was also experimentally proved to be similar with that of A. flavithermus (Bolton et al. 1997; Tawil et al. 2012; Agüloğlu et al. 2014; Ozdemir et al. 2015, 2016a) and A. amylolyticus (Poli et al. 2006) species which were already known as amylolytic enzyme producers. Totally, 15 novel Anoxybacillus α-amylase gene sequences were introduced to databases with this study and their preliminary BLASTP queries displayed (≥ 91.0%) gene sequence similarities to only well-known ASKA, ADTA and GSX-BL amylases originated from Anoxybacillus species, which were formerly proposed in a single subfamily within other Bacillaceae enzymes including GTA, Pizzo, Gt-amyII and BaqA from genus Geobacillus and Bacillus (Janeček et al. 2015; Ranjani et al. 2015; Chai et al. 2016; Sarian et al. 2017). Nevertheless, the protein sequence homologies of these Anoxybacillus amylases to GTA (≤ 69.4%) and BaqA (≤ 61.3%) were found to be relatively lower. As the putative protein and genome sequences were accumulated in databases, it is obvious that the accuracy of GH13 sequence-based classification system would increase and their related subfamilies might be classified under more meaningful groups. Thereby, the 15 Anoxybacillus sequences were analysed in detail with blast query and phylogenetic investigation by adding 30 α-amylase sequences from formerly defined GH13 subfamilies and 48 endospore-forming bacilli sequences (hypothetical or experimentally characterized) with the recommendations of Stam et al. (2006). It was interesting that any of these 48 α-amylase sequences from Bacillaceae family still could not be validly assign to any of the defined GH13 subfamilies. The constructed phylogenetic tree in Fig. 2 has completely drawn the picture of the phylogenetic relations between the amylases from Bacillaceae family and the other described GH13 subfamilies. Bacilli amylases were divided into totally 5 distinct branches far from other 15 well-defined GH13 subfamilies. The reorganized a1 and a3 clades, composed from E184aa-A and ATA, and the newly proposed a2 clade including B2M1-A representative were all thermostable members. Whereas the a4 clade which appeared as akin to BaqA and the formerly proposed “xy” non-defined subfamiliy containing BmaN1, formed the other two mesophilic bacilli originated amylases belonging to Bacillaceae family. These bacilli amylases could easily be separated both from the other GH13 subfamilies and from the taxonomic genera within the Bacillaceae family except the a3 clade. Only the clade a3 contained species from different genera: Anoxybacillus, Geobacillus and Parageobacillus, all of which were thermophilic. This clade included two exceptional Anoxybacillus species, A. geothermalis and A. tepidamans. Nevertheless, it must be noted that these two species displayed higher sequence similarities to a3 clade members than to a1 and a2 Anoxybacillus amylases, and they shared high 16S rRNA gene homologies to each other compare to the other genera members (this study, Coorevits et al. 2012; Bezuidt et al. 2016; Filippidou et al. 2016). Moreover, a taxonomic revision of A. tepidamans species was recently proposed and this species was transferred from genus Geobacillus to Anoxybacillus (Schäffer et al. 2004; Coorevits et al. 2012). The differential situation of A. tepidamans α-amylase and its non-conserved calcium binding sites were also previously mentioned by Chai and colleagues (2016). Therefore, these explanations clearly elucidate why A. geothermalis and A. tepidamans α-amylases were positioned in the a3 clade.

The phylogenetic tree shown in Fig. 2 and the comparison of the sequence alignments as well as the secondary and tertiary structure analyses of these 63 bacilli amylases definitely supported the findings below as presented in Fig. 5 and Supplementary file, Table SII. According to these results, five of the Bacillaceae α-amylase family clusters from a1 to xy share some common features including high sequence homologies to each other and displaying slowly increased E-values within the related group as suggested by Stam et al. (2006). At least one or more of their members are known to possess amylolytic activity experimentally, including the thermostable representatives which are E184aa-A (a1), B2M1-A (a2), ATA (a3) and the mesophilic BaqA (a4) and BmaN1 (xy) enzymes (Chai et al. 2012; Coorevits et al. 2012; Mok et al. 2013; Puspasari et al. 2013; Filippidou et al. 2016; Sarian et al. 2017). All the members contained signal peptide sequences (Chai et al. 2012; Mok et al. 2013) and predicted to be cellular components of the membrane that harboured residues in relation with cytoplasmic to extracellular sides. Although the two consecutive tryptophans in the loop 3 of domain A were preserved among Bacillaceae family, the 3D modelling analyses revealed that these aromatic residues were positioned far from the catalytic site (data not shown) and might play an unknown role instead of sugar binding as proposed before (Mok et al. 2013; Puspasari et al. 2013). The Phyre2 server also did not recognize two consecutive tryptophans as sugar binding pockets. When the surface view and overall fold of the 3D structure models were superimposed both with TASKA and GTA, they mostly covered the similar (β/α)8 TIM-barrel structure and three domain organization as arranged in the GH13 α-amylase family (MacGregor et al. 2001), (Supplementary Fig. S3). The 3D-modelling also supported the position of ATA and GTA α-amylases within the same a3 subfamily because of the derived homology report displaying the highest sequence identity and the lowest RMSD values to each other. These bacilli sequences contained well-defined α-amylase family specific conserved regions from CSR-I to CSR-VII with various amino acid residue differences, and shared the characteristic LPDLx motif in their CSR-V regions (Janeček 2002; Ranjani et al. 2015). Besides the F481, Y489, F492, F495, Y497 (E184aa-A numbering) residues, involved in the conserved aromatic motifs at the C-terminus as previously reported (Mok et al. 2013; Janeček et al. 2015), Y457 and Y469 were also additionally found to be preserved among the Bacillaceae amylases. The end of their C-termini also contained consecutive lysine and arginine residues (K501-R502, E184aa-A).

Beside these general features, Bacillaceae family α-amylase clusters from a1 to xy displayed significant differential characteristics from each other (Fig. 5, Supplementary Table SI and SII). The a1 and a2 members from genus Anoxybacillus in addition to the a3 members from gerera Anoxybacillus, Geobacillus and Parageobacillus were all thermophilic groups of Bacillaceae family. The signature sequence of LPDLx motif in CSR-V was LPDLA in these thermophiles. In the case of mesophilic genus Bacillus counterparts of a4 and xy, this residue changed to alanine in their LPDLN motif. The function of alanine residue (A184) of ASKA was already displayed by experimental mutation analysis (Chai et al. 2016) and its presence only in the thermophilic bacilli α-amylases indicates its importance in enzyme thermostability. The enzyme activity and stability of α-amylases from a1 to a3 clusters were also higher in elevated temperatures when compared with a4 and xy members due to their temperature requirements. As there were some residue differences in the seven CSR regions of α-amylases from Bacillaceae families, the created specific logos peculiar to each subfamily in this study would be useful for placing a novel bacilli α-amylase properly to its relevant subfamily in further studies. In 2017, a novel GH13_xy subfamily akin to BmaN1 α-amylase, having catalytic activity despite its atypical catalytic residues, was described by Sarian et al. (2017). The catalytic triad including the traditional Asp, Glu and Asp (D213∆, E242∆, D310∆, E184aa-A) residues as well as the invariable Arg (R211, E184aa-A) are all preserved in a1, a3 and a4 members like other defined GH13 subfamilies, but in the case of a2 (D215∆, G244∆ and R312∆, B2M1-A) and xy (K202∆, E231∆, and H294∆, BmaN1) subfamilies, an irregular catalytic triad was identified in addition to the residue changes in invariable Arg position (Y213/Y200, B2M1-A/BmaN1). Although the amylolytic activities of both B2M1-A and BmaN1 enzymes from a2 to xy clusters were experimentally proved, they differed from the other three Bacillaceae related groups, by the lack of complete catalytic machinery and this invariable Arg position (Filippidou et al. 2016; Sarian et al. 2017). Moreover, the pI values of a2 and xy members (> 9.0) were predicted higher than the others groups.

The previous studies confirmed that calcium ion could affect amylases by enhancing both their catalytic activities and structural stabilities (Declerck et al. 2000; Tao et al. 2008). Calcium had no effect on the enzyme activity of GTA α-amylase, but increased its thermostability (Mok et al. 2013). In the case of ASKA and ADTA, calcium ion increased both their enzyme activity and thermostability (Chai et al. 2012). The two calcium-binding sites in GTA and four sites in TASKA were already clarified with their crystal structure analyses. Therefore, when these residues involved in calcium binding were compared using sequence alignments of the a1, a2, a3 and a4 representatives, the possible Ca1, Ca2 and Ca4 binding sites were found to be mostly conserved among these bacilli amylases. In Ca2 of a2, a3 and a4 members, an Asn residue (N46, E184aa-A) was changed to Asp residue, from neutral to positive charge which may increase the affinity to calcium. In contrast, amino acid substitutions, having the possibility of decreasing the Ca2+ ion binding efficiency, was observed in a2, a3 and a4 members with Ca3 being the least conserved region among these bacilli. In the secondary structure alignment studies with Phyre2 server, hints for 12 substrate-binding pockets in relation with substrate specificities or sugar recognitions were also found out (Supplementary Fig. SIIa-d). The presence and roles of these pockets in maltose binding groove were confirmed by the in silico 3D models of E184aa-A (a1), B2M1-A (a2), ATA (a3), and BaqA (a4) α-amylases, which superimposed with both GTA and TASKA templates. Also, some additional residues for maltose binding, different from the previously reported ones were determined (Mok et al. 2013; Chai et al. 2016). The differences of the residue changes for 12 sugar pockets and 4 calcium-binding sites peculiar to the proposed subfamilies were also listed in Supplementary Table SII.

Novel thermostable α-amylases were introduced in this study that would be biotechnologically important and investigation of these Anoxybacillus α-amylases collectively with a great number of other bacilli originated enzymes provided a better look for the big picture of α-amylases from Bacillaceae family. Despite some other hypothesis, the evolution of α-amylase genes was thought to be occurred via divergent evolution (Jespersen et al. 1993; Janeček 2002). The number of the recognized α-amylase sequences from Bacillaceae family was considerably increased during the last decade and this survey described the basic story of divergent evolution of endospore-forming bacilli originated α-amylases. Thereby, we suggest the necessity that the still non-defined GH13 subfamily members containing the ASKA, ADTA, GTA, Pizzo, Gt-amyII and BaqA α-amylases which were formerly proposed in a single subfamily (Janeček et al. 2015; Ranjani et al. 2015; Chai et al. 2016; Sarian et al. 2017) should be classified into appropriately separated subfamilies and the examined 63 homologous Bacillaceae family related α-amylases could be grouped under more accurate and manageable GH13 subfamilies as they shared some similarities peculiar to the endospore-forming bacilli, but also contained significant differences that required dividing them into 5 individual subfamilies. This proposal was based upon the phylogenetic findings both on α-amylase and 16S rRNA genes, the genera level taxonomic origins as well as the temperature requirements of these amylolytic enzymes, the high sequence homologies of our 15 Anoxybacillus α-amylases to only Anoxybacillus related ASKA, ADTA and GSX-BL enzymes, their distant relatedness to GTA, Gt-amyII and BaqA, the separate branches and high boostrap values of GTA and BaqA in the cladogram, the comparisional sequence alignments and structural analyses including their 7 CSR regions, 12 sugar-binding and 4 calcium-binding sites, the presence or absence of the complete catalytic machinery in addition to the currently unassigned status of these bacilli α-amylases to a proper GH13 subfamiliy. Consequently, the proposed Bacillaceae family related subfamilies were the new a2 group clustered around α-amylase B2M1-A from Anoxybacillus sp. B2M1, the a1, a3 and a4 subfamilies (including the representatives E184aa-A, ATA, and BaqA) all of which were composed from the division of the formerly grouped single subfamily clustered around α-amylase BaqA, and finally the xy subfamily previously designated by Sarian et al. (2017) that clustered around amylolytic emzyme BmaN1 from B. megaterium.