1. Introduction
The Carbohydrate-Active enZYme database (CAZy) (
www.cazy.org) is a repository of catalytic modules and domains of enzymes that degrade, modify, or create glycosidic bonds. In conjunction to the CAZy enzymes, auxiliary activity (AA) enzymes provide a helping hand during lignocellulose degradation [
1]. Currently, 17 families of auxiliary activity cover redox enzymes that act in conjunction with CAZymes, namely nine families of ligninolytic enzymes and eight families of lytic polysaccharide monooxygenases. The ligninolytic AA family 3 (AA3) exclusively includes members of the glucose–methanol–choline (GMC) superfamily of FAD-dependent oxidoreductases [
2] and is structured into four subfamilies. Subfamily AA3_1 contains the flavodehydrogenase domain of cellobiose dehydrogenase; subfamily AA3_2 includes aryl alcohol oxidase, glucose oxidase, glucose dehydrogenase and pyranose dehydrogenase; subfamily AA3_3 consists of alcohol (methanol) oxidases and finally, subfamily AA3_4 comprises pyranose oxidases [
3]. Pyranose oxidase (POx, glucose 2-oxidase; EC 1.1.3.10, pyranose:oxygen 2-oxidoreductase) is an FAD-dependent oxidoreductase [
4] and has been studied extensively in the last 20 years mainly from fungal sources, both biochemically and structurally [
5,
6], because of potential bioelectrochemical and biocatalytic applications.
The reaction mechanism of POx involves two half-reactions: the reductive and the oxidative half-reaction. In the reductive half-reaction, two electrons are transferred as a hydride equivalent from the electron donor to the FAD, forming the reduced form of FADH
2. In the oxidative half-reaction, FADH
2 is reoxidized to its initial form while the electron acceptor is reduced. Preferred electron donor substrates for fungal POxs are various aldopyranoses (e.g.,
d-glucose,
d-galactose,
d-xylose and glucono-δ-lactone), while preferred electron acceptors include molecular oxygen, quinones (e.g., p-benzoquinone) and metal ions (e.g., the ferrocenium ion Fc
+). Oxidation of aldopyranoses typically occurs at the C2 position, yielding a 2-ketopyranose, but can also take place at the C3 position, yielding a 3-ketopyranose [
7]. A typical fungal POx monomer is comprised of an FAD-binding domain (a Rossmann domain of class α/β), and a substrate-binding domain (a six-stranded central β-sheet and associated α-helices). Fungal POxs form tetramers, where the access of the substrate to the active site is modulated by a large cavity and two access channels that allow the substrate molecule to enter the active site. Furthermore, its reactivity is fine-tuned by a mobile active-site loop [
8]. The FAD moiety in fungal POx is covalently linked to a histidine residue, resulting in one FAD molecule per POx monomer.
Pyranose oxidases have been characterized mainly from wood-degrading fungi, where they are secreted extracellularly and associated with membrane-bound vesicles or other types of structures of fungal hyphae [
9]. To date, POxs of fungal origin were biochemically characterized from
Aspergillus oryzae, Aspergillus nidulans,
Irpex lacteus,
Lyophyllum shimeji,
Peniophora gigantea,
Peniophora sp.,
Phanerochaete chrysosporium,
Phlebiopsis gigantea,
Trametes multicolor (
ochracea) and
Tricholoma matsutake [
3,
6]. Already, in 2009, it was hypothesized that
pox genes first evolved in bacteria and were then introduced into fungi via horizontal gene transfer [
10]. A newly constructed phylogenetic tree (
Figure 1) confirms the close relationship between putative bacterial and fungal POx sequences, despite their sharing sequence similarities of only 26 to 39%. It also shows that putative bacterial POxs are mainly found in the phyla of Actinobacteria and Proteobacteria.
Pyranose oxidases of bacterial origin were studied only lately, and the first biochemically characterized enzymes originate from
Arthrobacter siccitolerans (
AsPOx) and
Kitasatospora aureofaciens (previously known as
Streptomyces aureofaciens) (
KaPOx) [
11,
12]. In the POx phylogenetic tree,
KaPOx is positioned in the clade closest to fungal POx sequences, whereas the sequence of
AsPOx is positioned in a clade completely separated from these (
Figure 1). Recently, the FAD-dependent enzyme C-glycoside 3-oxidase from
Microbacterium sp. 5-2b (CarA) and related species (
Arthrobacter globiformis and
Microbacterium trichothecenolyticum (
MtCarA)) was reported to be closely related to bacterial POxs [
13]. The sequence of
MtCarA is indeed found within a Micrococcales clade of pyranose oxidases, closely positioned to that of
AsPOx (
Figure 1). The crystal structure of
MtCarA was solved at 2.4-Å resolution (PDB accession ID 7DVE). Its structure shared similarity with that of a subunit of a fungal POx from
Peniophora sp. (PDB 1TZL) with an rmsd value of 2.5 Å (454 Cα atoms; 28% amino acid sequence identity) [
13]. In accordance with the fungal POx monomers, the structure of bacterial
MtCarA showed the typical FAD and substrate-binding domains.
Bearing in mind that only two bacterial pyranose oxidases have been characterized so far, we studied a not yet investigated bacterial POx in order to get a better insight into this class of enzymes. We selected the sequence of Streptomyces canus pyranose oxidase (ScPOx), which is positioned in a separate, hitherto unexplored POx clade containing sequences from Actinobacteria and Streptomycetales, as this would expand our knowledge of the sequence space of bacterial pyranose oxidases. Our study provides a detailed biochemical characterization of ScPOx, with the emphasis on substrate specificity, steady-state kinetic parameters and stability. Additionally, a comparison of the ScPOx structural model with other structures was performed.
3. Discussion
Pyranose oxidase was first described in 1968 in the basidiomycete
Spongipellis unicolor (
Polyporus obtusus) and was subsequently studied from several wood-decomposing fungi [
3]. The first report on a bacterial POx was published in 2016, and only two bacterial enzymes have been characterized as pyranose oxidases so far—those from
K. aureofaciens and
A. siccitolerans [
11,
12]. Genome data provided a wealth of information on putative POx sequences of both bacterial and fungal origin, yet this wide sequence space is only poorly explored experimentally. A phylogenetic analysis of these putative sequences showed that bacterial pyranose oxidases form various distinct clades (
Figure 1). One of these clades, containing sequences from Streptomycetales and Pseudonocardiales, clusters closely with the fungal POx sequences, and contains
KaPOx as one of its members.
AsPOx (as well as
MtCarA) belongs to a clade of Micrococcales sequences that is clearly separated from the above-mentioned clades. Our phylogenetic analysis revealed several additional clades of putative bacterial POx sequences that are completely unexplored to date. To gain a better understanding of biochemical properties of bacterial POxs, we selected the sequence of a member of a Streptomycetales clade, POx from
Streptomyces canus (
ScPOx), for a more detailed study to further explore the wide sequence space of bacterial POx.
ScPOx is phylogenetically closer to
AsPOx and
MtCarA than to
KaPOx and the fungal POxs, and this is also reflected in its properties.
ScPOx is a monomeric enzyme with a molecular mass of 52 kDa (
AsPOx, 54 kDa;
MtCarA, 55 kDa) whereas
KaPOx is homodimeric (two subunits of 61 kDa each), and fungal POxs are typically homotetrameric (four subunits of ~65 kDa). The FAD is non-covalently attached in
ScPOx as in
AsPOx and
MtCarA, whereas it is tethered to a His in both
KaPOx and fungal POxs. Interestingly, this FAD-binding His residue is well conserved even in bacterial POx sequences (
Supplementary Materials Figure S3). It has been suggested that the STHW flavinylation motif of fungal POx sequences (GTHW in
KaPOx) is needed for flavinylation and since it is not present in full in the bacterial sequences [
11], the covalent attachment is not found.
ScPOx was shown to oxidize monosaccharides that are typical POx substrates—
d-glucose,
d-galactose and
d-xylose—albeit with catalytic efficiencies that are significantly lower, by up to four orders of magnitude, compared to values typically reported for
KaPOx or fungal POxs (e.g., catalytic efficiencies for the substrate pair
d-glucose/oxygen are ~0.2, 10,000 and 73,000 M
−1 s
−1 for
ScPOx,
KaPOx and
TmPOx, respectively). This is mainly because these sugars show Michaelis constants in the molar range for
ScPOx (
Table 1), which is a strong indication that they are in fact not the natural substrates of
ScPOx.
Recently, Kumano et al. [
13] reported that bacterial enzymes closely related to POxs, termed FAD-dependent C-glycoside 3-oxidase (CarA), catabolize both various C-glycosides and O-glycosides, but not
d-glucose [
13]. They showed that these enzymes oxidize the glucose moiety of C-glycosides such as carminic acid, primarily at the C-3, but also at the C-2 hydroxyl group, which is in accordance with the regiospecificity of oxidation of fungal POxs, which preferentially oxidize at C-2, but can further oxidize C-2 oxidized
d-glucose or
d-galactose at C-3 as well [
5,
17]. In fact, the sequence of
MtCarA from
M. trichothecenolyticum clusters with that of
AsPOx (
Figure 1) and shows sequence identity to this sequence and the sequence of S
cPOx (
Supplementary Materials Figure S3), so “FAD-dependent C-glycoside 3-oxidase” and “pyranose oxidase” refer to enzymes of the same sequence space (homologs). Since we presumed that monosaccharides are not the natural substrates of
ScPOx, we also tested various C- and O-glycosides as possible substrates. Based on the determined specific activities, some of these are in fact much better substrates than
d-glucose for
ScPOx, and we obtained the highest specific activity of 7.35 U/mg for the C-glycoside puerarin. Interestingly,
MtCarA from
M. trichothecenolyticum and closely related enzymes from
A. globiformis (
AgCarA) and
Microbacterium sp. 5-2b (CarA) showed no activity with puerarin, the C8-glucoside of daidzein. Kumano et al. [
13] stated that the position of the sugar moiety on the aglycon is important for activity, and that their CarA homologs strongly preferred C-6-glucosylated compounds, which we did not confirm for
ScPOx. Hence, it seems that bacterial POxs can show activity on a range of different glycosides, and that the activity to oxidize glycosides is found in various bacterial POx clades. Adding to that the relatively low sequence identity shared between different POx clades, we can probably expect even more functional variation to occur in the hitherto unexplored clades of bacterial POxs.
C- and O-glycosides are naturally found in plants and are known to be metabolized by glycosyltransferases and glycoside hydrolases [
18,
19]. Intestinal microorganisms deglycosylate C-glycosides by a two-step reaction: an oxidation of the sugar moiety, albeit with NAD(H)-dependent oxidoreductases, and a subsequent enzyme-catalysed C-C bond-cleaving step [
20]. It was recently suggested that CarA and its homologs—and hence probably also bacterial POxs—play a crucial role in the metabolism of C-glycosides in soil bacteria by catalysing the first step of an equivalent two-step reaction, however with oxygen-dependent enzymes [
13]. An ancestor of pyranose oxidase could thus have mainly functioned for the metabolization of plant-derived, sugar-containing compounds such as the C- and O-glycosides. Certain bacterial pyranose oxidases, such as the one from
K. aureofaciens, or fungal pyranose oxidases, which were suggested to have been acquired by horizontal gene transfer from bacteria, could then have evolved and specialized over time to oxidize lignocellulose-derived sugars such as
d-glucose,
d-xylose or
d-galactose.
4. Materials and Methods
4.1. Chemicals, Solutions, Buffers
The Escherichia coli expression strain BL21(DE3) was ordered from New England Biolabs (Frankfurt, Germany). The media used for all bacterial cultures was Luria Bertani (LB) broth (10 g/L of peptone from casein, 5 g/L yeast extract, 5 g/L NaCl) with an addition of ampicillin (Roth; Karlsruhe, Germany) to its final concentration of 100 µg/mL. Lactose monohydrate for the induction of gene expression was from Roth. p-Benzoquinone (p-BQ) and ferrocenium hexafluorophosphate ([Fe(C5H5)2]PF6) were purchased from Sigma-Aldrich (St. Louis, MO, USA) and 2,6-dichlorophenolindophenol (DCIP) from Fluka (Buchs, Switzerland). Horseradish peroxidase and 10-acetyl-3,7-dihydroxyphenoxazine (Amplex Red) were produced by Sigma-Aldrich and Chemodex (St. Gallen, Switzerland), respectively. N- and O-glycosides were obtained from abcr (Karlsruhe, Germany), except for carminic acid, which was from Glentham Life Sciences (Corsham, UK). The following buffers were used for protein purification: “buffer A” for affinity chromatography (50 mM Tris-HCl, pH = 7.5, 150 mM NaCl, 5% glycerol, 30 mM imidazole), “buffer B” for affinity chromatography (50 mM Tris-HCl, pH = 7.5; 150 mM NaCl; 5% glycerol and 300 mM imidazole), “storage buffer” for storing purified proteins (50 mM Tris-HCl, pH = 7.5; 150 mM NaCl; 10% glycerol), and a size exclusion chromatography buffer (50 mM Tris-HCl, pH = 7.5; 150 mM NaCl). The “universal buffer” used for determining pH optima and pH profiles was Britton Robinson buffer (50 mM boric acid; 50 mM phosphoric acid and 50 mM acetic acid titrated to a desired pH value). All other solutions were of standard recipe unless otherwise stated.
4.2. Bacterial Pyranose Oxidase Genes and Expression
Ten different coding sequences from various phylogenetic clades of the bacterial pyranose oxidase phylogenetic tree were selected. These sequences, predicted to code for pyranose oxidases, originated from the following organisms:
Frankia alni (Uniprot accession number Q0RGV3)
, Paenibacillus alvei (S9TU08),
Domibacillus aminovorans (
A0A177L2D9),
Klebsiella pneumoniae (A0A1S8Y799),
Rhizobium hainanense (A0A1C3VGT0),
Deinococcus aerius (A0A2I9D0D5),
Microbacterium testaceum (A0A147F038),
Streptomyces canus (A0A117Q443),
Geodermatophilus amargosae (A0A1I6Z5L5)
and Pseudomonas frederiksbergensis (A0A291AJ39). Coding sequences for all ten bacterial genes were ordered and cloned into the expression vector pET-21d+ between the restriction sites
NcoI and
HindIII by the company BioCat (Heidelberg, Germany). Gene codons were optimized for
E. coli. All constructs carried a C-terminal His
6 tag as well as ampicillin resistance. Calcium-competent
E. coli strain BL21(DE3) was transformed with the constructs using the heat-shock transformation method. Cells were grown in 250 mL and 500 mL LB medium with ampicillin inoculated with overnight cultures diluted 1:90. For the expression of the recombinant genes, bacterial cultures were grown at 37 °C with agitation (130 rpm) until OD
600 reached 0.6–1. After that, induction of gene expression was started by adding 10 mM lactose, and the temperature was decreased to 18 °C. Overexpression lasted for approximately 20 h, and cell pellets were collected by centrifugation (20 min, 5000 rpm, 8 °C, centrifuge Beckman Coulter Avanti J-26 XP, rotor JA-10 (Brea, CA, USA)). These cultivations yielded ~8.5 g of wet cell pellet per L of cell culture. Molecular properties of the recombinantly produced proteins were calculated by ProtParam (
https://web.expasy.org/protparam/, accessed on 20 October 2019) [
21].
4.3. Protein Purification
Cell pellets were resuspended in buffer A (1 mL of buffer per 1 g of wet cell pellet), and then disrupted by sonication with ultrasonic homogenizer Bandelin Sonoplus HD 60 (Berlin, Germany) at 120 V and 30% cycle for 5 min. Sonication was repeated 3 times with 5 min breaks on ice in between. The soluble cell extract was separated from cell debris by centrifugation (1 h, 25,000 rpm, 4 °C, centrifuge Beckman Coulter L1F 737, rotor 70Ti). Affinity chromatography was used to purify the His-tagged proteins from this crude cell extract, which was loaded onto a pre-equilibrated 5 mL HisTrap NP Ni Sepharose column (Cytiva; Marlborough, MA, USA) with flow rate of 2 mL/min. After washing the column with buffer A, elution was carried out with a linear increase to 100% buffer B during 15 min with the flow rate of 1 mL/min. When using methods such as SEC-SLS and DSC, size exclusion chromatography (SEC) was used as an additional purification step. Pooled fractions from affinity chromatography were loaded onto a 120-mL Superdex 75 (Cytiva) size exclusion column with the flow rate of 0.5 mL/min. The same flow rate was used to separate the proteins of interest from residual proteins. For both purification methods, fractions that were coloured bright yellow and showed absorption at 280 nm, as well as 389 and 450 nm, were pooled and concentrated/desalted using Amicon Ultra centrifugal filter units (MWCO 30 kDa). The purity of protein samples was checked by SDS-PAGE, and the presence of the His-tag was confirmed by Western blotting. Protein concentrations were determined with both the Bradford assay using commercially available bovine serum albumin (BSA; Thermo Scientific Pierce; Waltham, MA, USA) as standard and measuring absorbance at 280 nm (Agilent 8453 Diode Array spectrophotometer; Santa Clara, CA, USA). Both methods agreed very well and we used the mean values of these two independent measurements. Purified proteins were stored in “storage buffer” at −80 °C.
4.4. Analysis of Molecular Properties, Oligomeric State, and Protein Concentration
Routinely, the analysis of the approximate protein (subunit) mass was done by SDS-PAGE. The electrophoresis system used was from BioRad (Hercules, Clearwater, FL, USA), and appropriate precast gels were used (Mini-PROTEAN TGX Stain-Free Precast Gels, 4–16%). Preparation of protein samples as well as the electrophoresis procedure were done according to the manufacturer’s recommendations. Visualization was carried out in a GelDoc (BioRad). When analysing protein samples by Western blotting using penta-His-Tag monoclonal antibodies (Qiagen; Germantown, MD, USA), a published protocol was used [
22]. To determine the oligomeric state of native proteins, analytical size exclusion chromatography coupled with right-angle light scattering (SEC-SLS) was carried out. SEC-SLS analysis was conducted on an OMNISEC multi-detector GPC/SEC instrument (Malvern Panalytical; Malvern, UK) equipped with refractive index, right-angle light scattering (RALS) and a UV/VIS diode array detector. Protein samples were applied to a Superdex S200 increase 10/300 GL column (Cytiva; Marlborough, MA, USA) maintained at 25 °C, using phosphate-buffered saline as an isocratic mobile phase at a flow rate of 0.5 mL/min. The injection volume varied between 20 and 100 μL. Protein concentrations were measured online by using the refractive index detector. The instrument was calibrated using commercially available BSA (2 mg/mL).
4.5. Spectroscopic Analysis and Identification of the Cofactor
The UV-Vis spectra of purified proteins (protein concentration of 20 mg/mL) were recorded between 250 and 500 nm with an Agilent 8453 Diode Array spectrophotometer at room temperature. Reduction of the protein upon incubation with 4 M d-xylose was performed in the anaerobic environment of a glove box equilibrated with N2 prior to measurements. The final concentration of O2 in the glovebox was <0.2%. Extraction of the cofactor was done as follows: 5% v/v trichloroacetic acid (TCA) was added to 20 mg/mL of protein sample. The solution was then incubated at 600 rpm and 25 °C for 1 h, and subsequently centrifuged at 13,000 rpm for 10 min. The supernatant was neutralized with solid Na2CO3 (Roth) to the final pH value of 7. The spectra of these supernatants were measured as described and additionally analysed by MALDI-TOF mass spectrometry. To this end, samples were spotted with the dihydroxybenzoic acid matrix on the MALDI target plate in various dilutions. MALDI-TOF mass spectrometry was performed with a Bruker Autoflex MALDI-TOF instrument in the positive ion reflection mode. Scans were recorded in the range of 500–2000 m/z. Manual data analysis was done by using the software flex-Analysis v 1.3 (Bruker; Billerica, MA, USA).
4.6. Kinetic Analysis
All activity assays were carried out with a PerkinElmer Lambda 35 spectrophotometer (Waltham, MA, USA). Apparent steady-state kinetic parameters for the electron acceptors p-BQ and DCIP were determined by varying the concentration of the electron acceptors while keeping the electron donor concentration constant (4 M
d-xylose for
ScPOx).
d-Xylose was selected as the constant electron donor as it had the lowest Michaelis constant K
m. Wavelengths and extinction coefficient values were used as previously described [
22]. Steady-state kinetic constants for O
2 were determined as reported [
12].
Steady-state kinetic data for different electron donors were measured by varying the concentration of electron donors (50–2000 mM) while keeping the DCIP concentration constant (0.3 mM). When using oxygen (air) as the constant electron acceptor, the peroxidase-coupled assay with Amplex Red was used to measure hydrogen peroxide, with horseradish peroxidase and Amplex Red in concentration of 7.15 U/mL and 0.05 mM, respectively. Absorbance was measured at 560 nm with an extinction coefficient of 55.5 mM
−1 cm
−1 for Amplex Red. All assays were performed at 30 °C in triplicates using 50 mM potassium phosphate buffer, pH = 6.5. The concentration of
ScPOx in the assays was 5.6 µg/mL. The amount of enzyme that oxidizes 1 μmol of substrate per minute was defined as one unit of enzymatic activity. The steady-state kinetic constants were calculated using the software SigmaPlot v 12 (Systat Software; San Jose, CA, USA). v
max and K
m values were determined using nonlinear least-square regression by fitting the observed data to the Michaelis-Menten equation:
The turnover number kcat was calculated using vmax values and the molecular mass of the ScPOx monomer.
4.7. Stability Measurements
All kinetic enzyme stability measurements were performed with an activity assay using 0.3 mM DCIP as electron acceptor as well as 600 mM
d-glucose as an electron donor and were carried out as described in the Kinetic Analysis section.
d-Glucose was preferred here, as
d-xylose solutions of high concentrations (4 M) were very viscous and not well suitable for routine measurements, while 600 mM
d-glucose was convenient for this purpose. pH optimum determinations (dependence of the activity on the pH value) were performed in 50 mM universal buffer from pH 4.5 to 12 with pH increments of 0.5. pH stability measurements were performed by incubating the enzyme at 30 °C for 30 min in universal buffer of a given pH value. After that, residual activity was measured using the standard setup for enzyme activity assays. The temperature of half inactivation, i.e., the temperature where activity drops to 50% of the initial value within a given time (T
5030’), was measured by incubating the enzyme for 30 min at the certain temperature in a thermoblock, starting at 25 °C. After that, the residual activity was measured using standard assay conditions. The measurements were repeated in the range of 25 to 55 °C in incremental steps of 3 °C. Lastly, the half-life of activity (t
1/2) was determined for 40 and 50 °C by incubating an enzyme sample at a set temperature in a thermoblock and varying the time of the incubation. Again, residual activity was measured using standard assay conditions. The data were fitted to the equation:
where k
d is a rate constant of deactivation. The half-life of activity was calculated with the following equation:
The effects of different ions and compounds were investigated by introducing the desired component into the solution and measuring the activity as described in the
Section 4.6.
To determine the thermal transition temperature (melting temperature, Tm), i.e., the temperature at which 50% of protein is in its denatured state, differential scanning calorimetry (DSC) was used (MicroCal PEAQ-DSC Automated, Malvern Panalytical). The enzyme sample (325 μL in 50 mM KPP, pH = 6.5) was heated from 20 to 85 °C, with increments of 1 °C/min. The raw data were fitted with the software MicroCal PEAQ-DSC v 2.2 (Malvern Panalytical; Malvern, UK), using the fitting algorithm “non-two state”.
4.8. Structural Modelling
RoseTTaFold (
https://robetta.bakerlab.org/, accessed on 13 October 2021) [
23] was used to determine a structural model of
ScPOx. The best model, annotated as “Model 1”, was used for further structural analysis. The structural model was explored in the software PyMOL v 2.5.2, educational license (Schrödinger; New York, NY, USA).