Crystal structure of quinone‐dependent alcohol dehydrogenase from P seudogluconobacter saccharoketogenes. A versatile dehydrogenase oxidizing alcohols and carbohydrates

Abstract

The quinone‐dependent alcohol dehydrogenase (PQQ‐ADH, E.C. 1.1.5.2) from the Gram‐negative bacterium Pseudogluconobacter saccharoketogenes IFO 14464 oxidizes primary alcohols (e.g. ethanol, butanol), secondary alcohols (monosaccharides), as well as aldehydes, polysaccharides, and cyclodextrins. The recombinant protein, expressed in Pichia pastoris, was crystallized, and three‐dimensional (3D) structures of the native form, with PQQ and a Ca²⁺ ion, and of the enzyme in complex with a Zn²⁺ ion and a bound substrate mimic were determined at 1.72 Å and 1.84 Å resolution, respectively. PQQ‐ADH displays an eight‐bladed β‐propeller fold, characteristic of Type I quinone‐dependent methanol dehydrogenases. However, three of the four ligands of the Ca²⁺ ion differ from those of related dehydrogenases and they come from different parts of the polypeptide chain. These differences result in a more open, easily accessible active site, which explains why PQQ‐ADH can oxidize a broad range of substrates. The bound substrate mimic suggests Asp333 as the catalytic base. Remarkably, no vicinal disulfide bridge is present near the PQQ, which in other PQQ‐dependent alcohol dehydrogenases has been proposed to be necessary for electron transfer. Instead an associated cytochrome c can approach the PQQ for direct electron transfer.

Short abstract

Interactive Figure 1; Interactive Figure 5 | PDB Code(s): 4CVB; 4CVC

Abbreviations

Asd

aldose dehydrogenase

G4

Maltotetraose

GDH

glucose dehydrogenase

MDH

quinoprotein methanol dehydrogenase

PQQ‐ADH

quinone‐dependent alcohol dehydrogenase from Pseudogluconobacter saccharoketogenes IFO 14464

QADH

quinone‐dependent alcohol dehydrogenase

QH‐ADH

quinohemoprotein alcohol dehydrogenase

rmsd

root mean square deviation.

Introduction

Quinoproteins are oxidoreductases containing an amino acid‐derived o‐quinone cofactor, such as pyrroloquinoline quinone (PQQ; also named methoxatin), topaquinone (TPQ), tryptophan tryptophylquinone (TTQ), lysine tyrosylquinone (LTQ), and cysteine tryptophylquinone (CTQ). They have been recognized as the third class of redox enzymes after nicotinamide‐ and flavin‐dependent dehydrogenases,1 oxidizing a wide range of saccharide, alcohol, amine, and aldehyde substrates.2, 3 The best‐studied quinoproteins are PQQ‐containing enzymes. PQQ is a heterocyclic compound with two quinone oxygen atoms, two heterocyclic nitrogen atoms, and three carboxylate oxygen centers. It binds non‐covalently to proteins and can serve to bind metal ions. In the crystallographically characterized enzymes generally a Ca²⁺ ion is observed, coordinated by the PQQ O5, N6, and O7A atoms.4, 5

PQQ‐containing quinoproteins have been classified into two classes on the basis of their primary and tertiary structures.6 One class features a six‐bladed “propeller fold” and includes soluble glucose dehydrogenase (sGDH)5 and some aldose dehydrogenases.7, 8 The other class is characterized by an eight‐bladed propeller fold, and comprises three distinct types (I‐III) of PQQ‐dependent alcohol dehydrogenases.9, 10 Type I enzymes include methanol dehydrogenases (MDHs) and quinone‐dependent alcohol dehydrogenases (QADHs). MDHs are heterotetramers with a α₂β₂ composition; the α‐subunits contain the PQQ cofactor and a Ca²⁺ ion, while the β‐subunit is very small and has no known function.4 They are mostly found in the periplasmic space of methylotrophic bacteria, where they convert methanol into formaldehyde. QADHs are mostly homodimeric enzymes found in Proteobacteria like Pseudomonas, Ralstonia, and Comamonas species. They convert primary and secondary alcohols and aldehydes into the corresponding aldehydes/ketones or acids.11 Type II enzymes are monomeric quinohemoprotein alcohol dehydrogenases (QH‐ADHs), which convert the same substrates as Type I QADHs, but not methanol. They have two distinct functional domains of which one is a cytochrome c domain with a covalently bound heme c group.12, 13, 14 Finally, Type III enzymes are multicomponent membrane‐bound alcohol dehydrogenases present in acetic acid bacteria.10 Membrane‐associated glucose dehydrogenases (mGDH)6, 15 and polyol dehydrogenases16, 17 are also denoted as Type III enzymes.

The PQQ‐dependent alcohol dehydrogenase (PQQ‐ADH) from the Gram‐negative Pseudogluconobacter saccharoketogenes IFO 14464 (E.C. 1.1.5.2) is a monomeric enzyme that was initially found to be able to oxidize L‐sorbose to 2‐keto‐L‐gulonic acid via L‐sorbosone.18 However, the enzyme also oxidizes other monosaccharides, primary and secondary alcohols, aldehydes, α‐ and β‐cyclodextrin, oligosaccharides, and polysaccharides, including pectin, carrageenan, guar gum, alginate, and carboxymethyl cellulose, although the larger molecules are oxidized with less efficiency.18, 19 Maltotetraose (G4) and maltoheptaose (G7) can be oxidized at the reducing end C1 atom, yielding the carboxylic acid, as well as at their C6 atom, which yields aldehyde as the major and carboxylic acid as the minor product.19 The enzyme's broad substrate specificity makes it of interest for biocatalytic applications.

In the present study we determined the X‐ray crystal structure of PQQ‐ADH from P. saccharoketogenes IFO 14464, in its native form and in a complex with a polyethylene glycol fragment. These structures reveal an altered, more open active site conformation compared with other quinone‐dependent methanol/alcohol dehydrogenases, which explains its unusually broad substrate specificity. Moreover, the present work supports evidence for Asp333 as the catalytic base, and not the residue at position 220 as proposed previously20 in Type I PQQ‐ADHs.

Results and Discussion

Overall structure

The crystal structure of the quinone‐dependent alcohol dehydrogenase from P. saccharoketogenes IFO 14464 (PQQ‐ADH) was determined at a resolution of 1.72 Å to final R/R _free values of 12.8/14.9%. The final model comprises one protein chain of 562 residues, one calcium ion, one PQQ cofactor, three zinc ions, one chloride ion, one propanoic acid molecule and 614 water molecules. The first 8 residues of the mature protein (residues 37–44) and the last two residues were not visible in the electron density maps. Further details of the refinement are listed in Table 1. Most of the φ/ψ angles fall in the allowed regions of the Ramachandran plot; 96.3% are in the most favored regions, 3.2% in the additionally allowed regions, and 0.5% in the disallowed regions (Val196, Pro432, and Val475). Pro432 is forced into an unusual geometry by the disulfide bond between the adjacent Cys431 and Cys460. Val196, in a β‐turn, and Val475, next to the active site calcium ion ligand Tyr476, are also forced into the disallowed region by steric hindrance.

Table 1.

Data Collection and Refinement Statistics

	Native	G4 soak
Wavelength (Å)	1.54	0.93
Space group	C2	C2
Cell dimensions (Å,°) (a, b, c, β)	127.9, 87.2, 57.1, 90.16	128.2, 87.2, 57.0, 90.29
Crystal mosaicity (°)	0.73	0.90
Resolution range (Å)	30.5–1.72 (1.81–1.72)a	36.0–1.84 (1.93–1.84)a
R merge (%)	8.5 (22.9)a	8.1 (27.1)a
R pim (%)	5.4 (20.9)a	4.1 (12.3)a
Completeness (%)	88.9 (31.3)a	100.0 (99.9)a
〈I/σ(I)〉	11.4 (4.3)a	14.6 (6.2)a
Reflections total (unique)	194,837(59,031)a	269,946 (54,822)a
Multiplicity	3.3	4.9
Wilson B (Å2)	9.5	10.5
Refinement
Total protein atoms	4278	4278
Amino acid sequence numbers	45–606	45–606
Mean B‐factor protein	6.8	7.4
No. water molecules	626	571
Mean B‐factors solvent	16.9	18.4
No. ions	1 Ca2+, 4 Zn2+, 1 Cl−, 1 PPI	7 Zn2+, 1 Cl−, 1PPI
Mean B‐factors ions	14.3, 13.7, 13.0, 15.6	15.1, 12.2, 10.7
Ligands	1 PQQ	1 PQQ, 1 PEG
Mean B‐factors ligands	16.9	12.0, 29.4
R cryst/R free (%)	12.8/14.9	16.1/18.5
RMS deviation bonds/angles	0.010/1.4	0.006/1.1
Ramachandran plot (%) (favored/additionally allowed/disallowed)	96.3/3.2/0.5	97.0/2.5/0.5
MolProbity
Clash‐score/percentile	5.5/93rd	1.3/100th
MolProbity score/percentile	1.55/89th	1.03/100th

^aValues in parentheses are for the highest resolution shell.

The three‐dimensional structure of PQQ‐ADH shows an eight‐bladed β‐propeller fold (Fig. 1). Each of the eight propeller blades (W1–W8) contains four antiparallel β‐strands (A–D) forming a twisted β‐sheet (also called W‐motif). After W5, three additional strands are inserted (2a, 2b, and 3a), two forming a β‐sheet, and one is part of a three‐stranded sheet formed by the insertion of two strands in W6 after strand B [Fig. 1(B) and Supporting Information Fig. S1]. In addition, a small α‐helix (residues 579–583) is inserted after strand B of W8. All these features are also observed in QADHs and MDHs. Furthermore, an extra strand [W5E in Supporting Information Fig. S1 and Fig. 1(B)] is inserted between W5A and W5B. This strand is a unique feature of PQQ‐ADH. It contains a zinc‐binding site important for crystal contacts [Fig. 2(C)] (see below).

Figure 1.

Three‐dimensional structure of Pseudogluconobacter saccharoketogenes PQQ‐ADH. (A) Ribbon diagram showing the protein as a cartoon in rainbow colors from the N‐terminus in blue to the C‐terminus in red. http://imolecules3d.wiley.com:8080/imolecules3d/review/uEGEGTbxT0msDcChC18eZT435oUEV3Pku1HxNVBcT1co6E2WJygjt1xBbVwTFJpI784/1546. (B) Ribbon diagram showing the protein rotated by 90° compared with (A) with helices colored red and the β‐propeller in blue. The three‐stranded sheet is colored purple and the two‐stranded sheet cyan. Strand W5E (in teal) is a unique feature of PQQ‐ADH. The PQQ cofactor is shown in gray sticks, the calcium ion as a green sphere, the chloride ion as a yellow sphere, the zinc ions as brown spheres and the disulfide bridges as sticks. (C) Stereo image of the active site of PQQ‐ADH. Amino acid residues in the active site are shown as green sticks, the PQQ cofactor as gray sticks and the calcium ion as a green sphere. Hydrogen bonds are indicated with black dashes and metal‐ligand bonds with cyan dashes. An interactive view is available in the electronic version of the article.

Figure 2.

Zinc‐binding sites in PQQ‐ADH. (A–C) Amino acid residues and ligands of native PQQ‐ADH are in shown in green, of symmetry related molecules in dark green and denoted with an asterisk (*). Zinc ions are shown as brown spheres, the chloride ion as a yellow sphere, and water molecules as red spheres. (D–E) Amino acids and ligands of the G4 structure in cyan. Zinc ions 1 to 3 are of the native and zinc ions 5 to 7 are of the G4 structure.

Six of the eight β‐propeller blades contain a conserved tryptophan‐docking motif (Table 2). The invariant residues cause a tryptophan residue to stack between an alanine residue and the peptide bond of a glycine residue located on a neighboring β‐sheet. The tryptophan‐docking motif probably stabilizes the β‐propeller fold.11, 21

Table 2.

Tryptophan‐Docking Motif

Position	Motif	1	2	3	4	5	6	7	8	9	10	11
W1		Ala 127	Ile	Asp	Gly	Lys	Thr	Gly	Ser	Leu	Ile	Trp 137
W2		Ala 178	Leu	Asp	Ala	Lys	Thr	Gly	Lys	Leu	Ala	Trp 188
W3		Gly230	Thr	Asp	Ala	Glu	Ser	Gly	Glu	Glu	Leu	Trp 240
W4		Ala 312	Val	Asp	Pro	Lys	Thr	Gly	Glu	Val	Val	Trp 322
W5		Gln384	Phe	Asp	Ala	Lys	Thr	Gly	Asp	Tyr	Phe	Trp 394
W6		Ala 496	Ile	Asp	Leu	Ala	Thr	Gly	Glu	Thr	Lys	Trp 506
W7		Ala 537	Leu	Asp	Ala	Glu	Ser	Gly	Lys	Glu	Val	Trp 547
W8		Val601	Phe	Ala	Leu604	—	Asp91	Leu	Gln	Leu	Val	Trp 96
Consensus		Ala	—	Asp	Ala	Lys	Thr	Gly	Glu	Leu	Val	Trp

Amino‐acid sequences of the tryptophan docking motif of PQQ‐ADH. The bold residues represent invariant or semi‐invariant residues between motifs. The table has been adapted from Ref. 18.

The 3D structure of PQQ‐ADH is very similar to the propeller fold part of Type I MDHs22, 23, 24, 25, 26 and QADHs,11 and Type II QH‐ADHs,12, 13, 14 with rmsd values of less than 1.7 Å and sequence identities around 27 to 30%. Yet, the highest sequence and structural similarity is observed to L‐sorbose dehydrogenase (SDH).27 Unlike Type I MDHs, SDH and QADHs, PQQ‐ADH is monomeric in solution. Phylogenetic analysis of PQQ‐containing methanol/alcohol dehydrogenases shows that PQQ‐ADH belongs to an outgroup with only distant relationship to the other enzymes.28

Active site of PQQ‐ADH

The active site is located near the upper face of the β‐propeller in a hydrophobic cradle on the surface of the enzyme. Its solvent‐exposed position enables it to accommodate a large spectrum of variously sized substrates. PQQ is bound via hydrogen‐bonding interactions to the side chains of Glu106, Arg158, Ser203, Thr218, Gln220, Asp333, Glu335, Lys378, Asp439, Trp440 and Tyr515, and the carbonyl oxygen atom of Phe434 [Fig. 1(C)]. Additional binding interactions are provided via hydrophobic stacking interactions with Leu153 on one side of its molecular plane (see next paragraph) and with the indole ring of the invariant Trp270 on the other side. The reactive orthoquinone group is hydrogen‐bonded to Asp439 and Lys378 with its O4 atom, and to the Lys378 side chain with its O5 atom. The O5 atom is also a ligand of the active site Ca²⁺ ion. The interactions of O5 have been suggested to polarize the C5–O5 bond,29 priming the PQQ C5 atom to receive the hydride ion from the substrate, which is essential for the enzyme's catalytic activity (see below). As in other PQQ‐dependent enzymes, the calcium ion is bound to the O5, N6, and O7A atoms of the PQQ cofactor. Additional coordination is provided by the side chains of Asp333, Gln220, Glu335, and Tyr476. The latter three ligands are unique to PQQ‐ADH. To create a more open active site several active site loops have different conformations, and the calcium ion is coordinated by different residues. For instance, Glu335 replaces Asn255 as a Ca²⁺ ligand, and Gln220 replaces a Glu at a position equivalent to 221 in other quinoproteins. The Ca²⁺‐coordinating Tyr476 side chain has no equivalent in the other PQQ‐dependent alcohol dehydrogenases (Supporting Information Fig. S1). The corresponding Tyr in SDH is located in a β hairpin not involved in Ca²⁺ binding.27 Figure 3 shows a side‐by‐side comparison of the active sites entrances of four PQQ‐dependent alcohol dehydrogenases, rationalizing the different substrate specificities.

Figure 3.

Side‐by‐side comparison of the active sites entrances of four PQQ‐dependent alcohol dehydrogenases. The (transparent) accessible surfaces of the active sites of the four dehydrogenases are depicted. The probe radius used here is 0.95 Å (typically 1.4 Å is used) to visualize the entrances to the active sites. The PQQ cofactor is shown in CPK sticks and the calcium ion as a green sphere (A) PQQ‐ADH shows an open and accessible active site explaining its broad substrate range. (B) QH‐ADH (PDB code: 1KB0) has a slightly narrower entrance accounting for its relatively broad substrate specificity for sterols, PEGs and aldehydes.12 (C) MDH (PDB code: 4AAH) has a rather small cavity suitable for methanol and ethanol.25 (D) QEDH (PDB code: 1FLG) has a small cavity accommodating primary and secondary alcohols.11

The solvent exposed and accessible active site suggests that a cytochrome c_L, the usual electron acceptor in MDHs,21 or a cytochrome c551 in the case of SDH,27 could easily enter the active site to receive the electrons produced during the reaction. Superimposing the structure of PQQ‐ADH on QH‐ADH,12 which has a covalently attached cytochrome c domain, shows indeed that there is enough space in PQQ‐ADH for a cytochrome c to bind in a productive orientation (Fig. 4 and Supporting Information Fig. S3).

Figure 4.

Structural comparison of the active sites of PQQ‐ADH and C. testosteroni QH‐ADH. Active site amino acid residues of PQQ‐ADH are shown as green sticks. The PQQ cofactor is shown in gray sticks and the calcium ion as a green sphere. The amino acids residues, PQQ and calcium ion of QH‐ADH are shown in magenta.

Although PQQ‐ADH has its broad substrate specificity in common with the PQQ‐containing soluble aldose dehydrogenase from Escherichia coli (Asd),7 their sequence identity is only 7% and Asd belongs to the six‐bladed β‐propeller class of glucose and aldose dehydrogenases. The active site environment is also quite different between the two enzymes. PQQ‐ADH lacks the His residue that serves as general base in Asd and sGDH, as well as the Arg involved in substrate binding.7, 30

Disulfide bridges

PQQ‐ADH contains three disulfide bridges, Cys219‐Cys226, Cys336‐Cys377, and Cys431‐Cys460. The Cys219‐Cys226 and Cys336‐Cys377 disulfides stabilize the positions of Gln220 and Glu335, respectively, which are both Ca²⁺ ligands. The Cys431‐Cys460 disulfide is located between β‐strands 3a and 3b and is also present in MDHs but not in QADHs.

Surprisingly, PQQ‐ADH does not contain a vicinal SS‐bond between two adjacent cysteine residues in a Type VIII β‐turn31, 32 [Figs. 1(C) and 4 and Supporting Information Fig. S1]. Such a vicinal disulfide is strictly conserved among quinohemo/quino alcohol dehydrogenases with the exception of lupanine hydroxylase, a Type II QH‐ADH,33 mGDH,17 and SDH.27 It has been shown that the disulfide bond is essential for electron transfer from PQQ to the recipient cytochrome c.34 An alternative suggestion is that the disulfide bond provides conformational rigidity of the protein during substrate binding.12 Instead of the vicinal disulfide bond, Leu153 interacts with the PQQ in PQQ‐ADH [Figs. 1(C) and 4 and Supporting Information Fig. S1], and the amino acid residues surrounding Leu153 adopt different conformations, creating a highly open and exposed active site. As evidenced by the crystal structure of PQQ‐ADH, a vicinal disulfide bond is apparently not strictly required for enzyme activity of PQQ‐containing alcohol dehydrogenases.

Substrate specificity of PQQ‐ADH

PQQ‐ADH (residues 37–606) shows broad substrate specificity, and oxidizes various primary and secondary alcohols, as well as various monosaccharides with good activity (Table 3). In solution, these monosaccharides exist predominantly in the pyranose form,35 and hence they are likely oxidized at their reducing end to the δ‐lactone, which hydrolyzes in water to form the linear 1‐acid end product. However, the monosaccharides are, to a small degree, also present in the open‐chain aldehyde form.35 Therefore, we cannot fully exclude a direct oxidation of the aldehyde by PQQ‐ADH to the 1‐acid, in particular because the enzyme can oxidize linear monosaccharide alcohols like D‐sorbitol (D‐glucitol), xylitol, and D‐mannitol, albeit with somewhat lower activity (5–20%; Table 3). The enzyme's activity on l‐sorbose is less than 1% of the activity on glucose and may be related to l‐sorbose being a 2‐keto‐sugar.

Table 3.

Substrate Specificity of PQQ‐ADH

Substrate	Relative activity (%)
Sugars and sugar alcohols
D‐Glucose	100
D‐Galactose	85
D‐Mannose	77
D‐Xylose	130
Methyl α‐D‐glucopyranoside	5
Xylitol	20
L‐Sorbose	0.6
D‐Sorbitol	9
D‐Mannitol	6
Alcohols
Methanol	3
Ethanol	16
Ethyleneglycol	109
n‐Propanol	120
Iso‐propanol	121
rac‐1,2‐Propanediol	128
Propan‐1,2,3‐triol (glycerol)	25
n‐Butanol	241
rac‐2‐Methyl‐2,4‐pentanediol (MPD)	164
Oligosaccharides
Maltotriose (G3)	31
Maltotetraose (G4)	24
D‐Glucose + 20 mM Zn(Ac)2 *	7
G4 + 20 mM Zn(Ac)2 [Link]	0.02

The enzyme assays were performed as described in Materials and Methods. The enzyme activity toward D‐glucose was assigned as 100 (%).

^a96 h Incubation time (with Zn(Ac)₂).

^b3 h Incubation time.

As is evident from Table 3, the enzyme also has significant activity on tri‐ and tetrasaccharides (20–30% of the activity on glucose). The activity on rac−2‐methyl‐2,4‐pentanediol demonstrates the enzyme's activity on secondary alcohols. The broad substrate specificity can be explained by the open active site, which is accessible for substrates of widely different sizes (Fig. 3).

Zinc ions

Four zinc ions, originating from the crystallization solution, were found at the surface of PQQ‐ADH [Fig. 2(A–D)]. Two of them (Zn1 and Zn3) stabilize the interaction between symmetry related molecules in the crystal, rationalizing why PQQ‐ADH crystals could not be grown in the absence of zinc ions. The other two zinc ions have no interactions with symmetry‐related protein molecules. Three additional zinc ions were found in a 9‐month‐old crystal used to investigate the binding of maltotetraose (G4) [Fig. 2(E)]. An anomalous difference Fourier map showed, in total, seven strong peaks well above the noise level. The three extra Zn²⁺ ions probably result from the high concentration of ions and the prolonged storage time of the crystal. Unexpectedly, one of the strongest peaks was found at the position of the Ca²⁺ ion in the active site (Fig. 5). Refining this ion as a Zn²⁺ ion instead of Ca²⁺ resulted in a tetrahedral coordination arrangement (as preferred by Zn²⁺), with two metal‐ligands (Tyr476 and Asp333 side chains) moved out of the coordination sphere (Fig. 5 and Supporting Information Fig. S2). The remaining ligands refined to typical Zn²⁺‐binding distances of 2.0 to 2.3 Å.36 Together, these data suggest that the calcium ion has been replaced by a zinc ion in PQQ‐ADH‐G4. Activity experiments with addition of 20 mM Zn²⁺ to the reaction medium showed decreased activity toward D‐glucose and G4 (Table 3). However, it is unclear if a 96‐h incubation is sufficient for the full exchange of the Ca²⁺ for Zn²⁺. It remains an intriguing possibility that the enzyme's two to three times enhanced activity in the presence of 0.02 to 0.1 mM Fe²⁺ or Fe³⁺ (Ref. 19) is related to the binding of an iron ion at the position of the calcium ion in the active site.

Figure 5.

Stereo image of the binding of a PEG fragment in the substrate‐binding site of PQQ‐ADH. Amino acids are shown as cyan sticks. The PQQ cofactor is shown as gray sticks, the PEG fragment as yellow sticks and the zinc ion as a brown sphere. Metal‐ligand bonds are indicated with black dashes. The final σ _A‐weighted 2F _o‐F _c map electron density map, in gray, is contoured at 1σ. An interactive view is available in the electronic version of the article.

The exchange of the calcium ion by other ions has also been achieved in other quinoproteins. The calcium ion in the Type I QADH from P. aeruginosa could be removed by treatment with a chelating agent and reconstitution by incubation with PQQ and Ca²⁺ or Sr²⁺, but not with Mg²⁺, Mn²⁺, or Cd²⁺.37 Likewise, a functional MDH was obtained after reconstitution with Ba²⁺ or Sr²⁺.38, 39, 40 Moreover, in the crystal structure of Methylacidiphilum fumariolicum MDH a lanthanide ion was found at the calcium position.26 In these other proteins the active site is more compact and the calcium ion is not readily accessible from the solvent. Because of the open, accessible conformation of the active site of PQQ‐ADH the exchange of Ca²⁺ for Zn²⁺ appears to be much easier in the latter enzyme.

Substrate binding to PQQ‐ADH

To determine the substrate‐binding mode we performed soaking studies with several substrates. Only the experiment with G4 revealed positive F _o –F _c electron density near the active site Zn²⁺ ion and PQQ. The electron density was compatible with a partially occupied PEG550 MME molecule from the crystallization solution, but not with a maltotetraose molecule. In the refined structure the hydroxyl group of the PEG molecule is hydrogen bonded to Asp333 (OD2 atom at 2.5 Å), O5 of PQQ (at 2.3 Å), and Zn²⁺ at 2.9 Å. Van der Waals interactions with Phe434, Leu578, Tyr476, Leu153, Leu435, and PQQ complete the binding (Fig. 5). The hydroxyl group has exactly the same position as one of the carboxylate oxygen atoms of the tetrahydrofuran‐2‐carboxylic acid product found in the structure of QH‐ADH,12 after superimposition of the two enzymes. From this we conclude that the PEG fragment is bound at the expected position for a substrate/product molecule.

Mechanism of oxidation

The PQQ ring structure in native PQQ‐ADH is entirely planar, indicating its oxidized orthoquinone23 or reduced hydroquinone PQQH2 form.12 However, in the G4 structure the PQQ‐O4 is out of plane by about 45° (Fig. 5) and is pulled below the PQQ tricyclic plane towards the Nη of Lys378 and the carboxylate atoms of Asp439. This out‐of‐plane position of O4 suggests that the C4 atom is sp 3 hybridized and bears a hydroxyl group. A similar position of O4 has also been observed in the MDHs from M. extorquens 4 and M. fumariolicum.26 Reduced PQQ is usually protonated at both O4 and O541 or singly protonated at O4.42, 43, 44 Likely, in PQQ‐ADH only the first reduction step has occurred with protonation only of the O4 atom, similar to what has been suggested for quinoprotein alcohol dehydrogenase from P. aeruginosa.41 Another possibility is that the PQQ is in the radical semiquinone state, which also has the O4 out of the plane of the cofactor.44

In quinoproteins a hydride transfer mechanism is generally accepted for the oxidation of substrates.5, 44, 45, 46 The mechanism involves proton abstraction from the hydroxyl group of the substrate by a general base catalyst. In M. extorquens MDH and P. aeruginosa QADH the invariant Asp303 (Asp333 in PQQ‐ADH) was proposed as the general base,42, 44, 47 while Glu171 was suggested to have this role in Methylophilus W3A1 MDH.20 The residue in PQQ‐ADH equivalent to Glu171 is Gln220, which is highly unlikely to function as a general base, and its side chain Oε1 atom, coordinating the Ca²⁺ atom, is too far (3.7 Å) from the PEG substrate to abstract a proton from its hydroxyl group (Fig. 5). Since the invariant Asp is also conserved in PQQ‐ADH (Supporting Information Fig. S1), and at hydrogen‐bonding distance to the PEG hydroxyl group, we suggest that this residue (Asp333) is the general base in the reaction mechanism of this class of PQQ‐dependent dehydrogenases. The mutant enzymes D333A, D333N, Q220A, and Q220E were constructed and found to be inactive (data not shown).

In the subsequent step of the reaction mechanism (or concomitantly) a hydride ion is abstracted from the hydroxyl‐bearing carbon atom, and transferred to the electrophilic C5 carbonyl carbon of the PQQ quinone, upon which PQQ tautomerizes to the hydroquinone PQQH₂ 5, 28, 41, 45 and product is released or further converted. The conserved Lys378 (or Arg in some other enzymes) may help in transferring the electron from PQQH₂ to an electron acceptor.6

To further characterize the binding mode of substrates in the active site of PQQ‐ADH, we carried out docking studies with monosaccharides. These studies indicated that the enzyme can easily accommodate these substrates, both in the pyranose and linear conformation, because of its shallow and solvent exposed active site. These studies also suggest that β‐d‐glucose can be oxidized easier than α‐d‐glucose because in β‐d‐glucose the C1 hydrogen atom, which is to be transferred to the PQQ as a hydride ion, has an axial orientation, and faces the C5 atom of the PQQ. Moreover, productive binding of α‐cyclodextrin appeared possible, although the C6 carbon atom is probably not close enough for efficient direct hydride transfer, explaining the low activity with this substrate.19

In conclusion, the structure of PQQ‐ADH presented here is the first structure of a monomeric Type I quinoprotein alcohol dehydrogenase, distinctly different from other Type I enzymes. The presence of a polyethylene glycol fragment bound in the active site suggests Asp333 as the general base in the catalytic mechanism. The active site is open and easily accessible from the solvent and can accommodate linear alcohols, as well as mono‐ and oligosaccharides of different sizes. Monosaccharides can be bound both in the pyranose and linear conformation. Surprisingly, the enzyme has no vicinal disulfide bond between adjacent cysteine residues, which is observed in other quinone‐dependent alcohol dehydrogenases. The absence of this vicinal disulfide may be required to make the active site accessible to substrates of widely different sizes and to create a versatile enzyme.

Materials and Methods

Cloning, expression, and purification

The gene encoding the P. saccharoketogenes PQQ‐ADH, including its own signal sequence, was synthesized as a codon optimized fragment by GeneArt AG (Regensburg, Germany) and cloned into the pDONR/Zeo cassette via the Gateway^® BP recombination reaction (Life Technologies) resulting in the entry vector pENTRY‐ADH. To enable expression of PQQ‐ADH in Pichia pastoris, the gene was cloned from pENTRY‐ADH into pPIC2‐DEST via the Gateway^® LR recombination reaction. pPIC2‐DEST was derived from the commercially available vector pPIC‐DEST by removal of the signal sequence from this vector. The resulting plasmid, pPIC2‐ADH was linearized by SalI digestion, enabling integration of the construct into the HIS4 locus of P. pastoris GS115 upon transformation. This vector contains the P. pastoris AOX1 promoter, allowing strong methanol‐inducible gene expression. For production of PQQ‐ADH, P. pastoris transformed with pPIC2‐ADH was grown in a 2 liter B. Braun Biostat B fermentor, and expression of the enzyme was induced with methanol according to standard P. pastoris fermentation protocols (Life Technologies). After fermentation the major fraction of the expressed PQQ‐ADH was found in the culture supernatant, with levels reaching 100 to 400 mg/L in 172 h after the start of the methanol induction. The N‐terminus of the mature protein AEPSKAGQSA was found to start at position 37 of the coding sequence as determined by Edman degradation and analysis on a Procise^® cLC capillary 491 protein sequencing system (Applied Biosystems, Foster City, CA).

The culture supernatant was extensively dialyzed against 20 mM Tris‐HCl, pH 7.5 (Buffer A). The recombinant PQQ‐ADH was purified by anion exchange chromatography using a 60 mL Q‐Sepharose HP (XK16/20) column (GE Healthcare, Uppsala, Sweden) using a gradient of Buffer A, and Buffer A containing 1.0 M NaCl as Buffer B. Active fractions eluting at 0.35 to 0.40 M NaCl were pooled. Reconstitution with PQQ was carried out at room temperature in 5 mM CaCl₂, 20 mM Tris‐HCl, pH 7.5. The pooled fractions were concentrated and washed with GF buffer (20 mM Tris, pH 8.5, 5 mM CaCl₂, and 150 mM NaCl) by ultrafiltration to 6.7 mg protein mL⁻¹. Further purification was achieved on a HiLoad Superdex75 PG XK26/60 (320 mL) (GE Healthcare) column. The main fraction, containing active PQQ‐ADH with a molecular weight of 68 kDa, was concentrated to 9.5 mg protein mL⁻¹.

PQQ‐ADH activity was determined as described.18, 48 The enzyme assays were carried out in a reaction mixture (1.0 mL) containing 1% substrate, 160 μM 2,6‐dichlorophenolindophenol (DCIP), 50 μM phenazine methosulfate (PMS), 20 mM Tris‐HCl, pH 7.5, and reconstituted enzyme solution, and incubated at room temperature. Oxidation of the substrate was monitored at 600 nm for 10 min in a UV/Vis spectrophotometer (Ultrospec 4300 pro; GE Healthcare).

Crystallization and X‐ray data collection

Vapor‐diffusion crystallization experiments were conducted using a Mosquito crystallization robot (Molecular Dimensions Ltd, Newmarket, England). In a typical experiment, 0.1 μL screening solution was added to 0.1 μL protein solution in a 96‐well MRC2 plate (Hampton Research, Aliso Viejo), with the reservoir wells containing 60 μL screening solution. Crystals grew from solutions containing polyethylene glycol (PEG) and zinc acetate.

Crystallization conditions were optimized using hanging‐drop set‐ups with 18 to 21% (v/v) PEG550 MME and 20 mM Zn(Ac)₂ in 100 mM PCB buffer (sodium propionate, sodium cacodylate and BIS‐TRIS propane, pH 6.0–7.0)49 as precipitant, and drops containing 2 μL protein solution (9.5 mg mL⁻¹) and 2 μL reservoir solution at 295 K. Crystals were only obtained from conditions with Zn²⁺ concentrations above 20 mM.

Monoclinic crystals grew in space group C2 with cell dimensions a = 128.0 Å, b = 87.2 Å, c = 57.1 Å, and β = 90.2°. With 1 monomer of 65 kDa in the asymmetric unit, the V _M is 2.5 ų/Da50 with a solvent content of 51%. Before data collection, crystals were soaked for 15 s in a cryoprotectant solution, consisting of 25% (v/v) PEG550 MME and 20 mM Zn(Ac)₂ in 100 mM PCB buffer, pH 7.0, directly followed by flash cooling. Data were collected in house at 110 K with a DIP2030H image plate detector (Bruker‐AXS, Delft, The Netherlands) using Cu‐Kα radiation from a Bruker‐AXS FR591 rotating‐anode generator equipped with Osmic mirrors. Data were 95% complete up to 1.81 Å resolution; completeness of higher resolution shells became lower because of detector geometry. All intensity data were processed using iMOSFLM51 and scaled using SCALA from the CCP4 package.52 The PQQ‐ADH crystal was hemihedrally twinned (the monoclinic β angle was close to 90° (90.16°)) with a twin fraction α of 0.2 (meaning partial twinning). The twin operator was h, −k, −l (180° rotation around a). A summary of data collection statistics is given in Table 1.

Structure determination and refinement

The structure of PQQ‐ADH was solved by molecular replacement with Phaser53 using a search model,54 composed of quinohemoprotein alcohol dehydrogenase from Comamonas testosteroni (26% sequence identity, PDB code: 1KB0)12 and quinohemoprotein alcohol dehydrogenase from Pseudomonas putida HK5 ADH‐IIG (25% sequence identity, PDB code: 1YIQ),14 generated by the FFAS server.55 The phases obtained from the model were improved with the ARP/wARP procedure56 combined with manual model improvement with Coot.57 Refinement was done with REFMAC552 with automatic twin refinement and using TLS rigid body refinement.58 Further details of the refinement are listed in Table 1.

For maltotetraose binding studies, a 9‐month‐old crystal was soaked in a solution containing 100 mM maltotetraose (G4) for several minutes and immediately flash cooled in liquid nitrogen. Data were collected from a nontwinned crystal on beam line X11 (EMBL outstation at DESY, Hamburg, Germany). After rigid body refinement with REFMAC5 using the native structure as starting model, σ _A‐weighted 2F _o –F _c and F _o –F _c maps showed some extra, metal ion‐like density in the active site. To identify the ions, we calculated an anomalous difference Fourier map, with data collected at a wavelength of 0.93 Å (2.3 anomalous electrons for zinc, and 0.5 for calcium), using phases obtained from the model without any metal ions.

Analysis of the stereochemical quality of the models was done with MolProbity.59 Figures were prepared with PyMOL (http://www/pymol.org). Atomic coordinates and experimental structure factor amplitudes have been deposited in the Protein Data Bank (PDB) with PDB IDs 4CVB and 4CVC for native and substrate‐bound PQQ‐ADH, respectively.

Supporting information

Supporting Information

Supporting Information Figure 1a.

Supporting Information Figure 1b.

Supporting Information Figure 2.

Supporting Information Figure 3.