Structure of a dimeric fungal α-type carbonic anhydrase

Abstract

AoCA binds to AoCA X-ray crystallography (View interaction) Carbonic anhydrase (CA) catalyses the reversible conversion of CO2 into bicarbonate ions and was first purified from bovine blood in 1933 [1] but has since been found in all domains of life [2]. Besides the biological (e.g., as essential enzyme for CO2 fixation in plants [3]) and pharmaceutical importance of CAs (e.g., as drug targets for the treatment of glaucoma [4]), the enzyme has gained attention from an industrial perspective. A promising approach for sustainable CO2 capture from industrial gases (i.e. smoke gas) is the replacement of high-energy requiring absorption chemicals with a biocatalyst. Generally, these CO2 capture techniques operate by bringing an absorber solution into contact with exhaust-CO2 in presence of a soluble or immobilized CA. In previous work, bovine CA [5] and modified human CA II (hCAII) [6] were used as CO2 capture catalysts. CAs are classified in five classes, α, β, γ, δ and ε [2], which are unrelated in sequence and structure, but have all converged to a metal dependent mechanism for catalysis, where the metal is usually Zn. The α-class is predominantly mammalian, monomeric and by far the best mechanistically studied. α-CAs are also the only ones found in mammals. The β-class is found in plants, fungi and prokaryotes while γ-CAs have been so far only identified in archea. δ- [7] and ζ-CAs had until recently only been identified in marine phytoplankton. No α-type fungal CA has been characterized in detail so far to the best of our knowledge. In fungi the intracellular β-class predominates [2] and the biological function of the putatively secreted α-CA enzymes is unknown. Only one structure representative for a fungal CA has been published, the structure of the β-CA from the pathogenic fungus Cryptococcus neoformans [8]. The structure of A. oryzae CA (AoCA) presented here represents the first fungal α-CA. The structure reveals unusual characteristics, which may have implications for the biological function of the α-class in fungi. Cloning, expression and purification: Genomic DNA was isolated from Ao strain CBS 205.89 (Centraalbureau voor Schimmelcultures, Netherlands) according to a modified FastDNA SPIN protocol (MPBiomedical). Oligonucleotides F-Q2TWF5 (5′ACACAACTGGGGATCCACCATGAAGTTCGCCACTACTTTG-3′) and R-Q2TWF5 (5′AGATCTCGAGAAGCTTAACCGAATTGAGTTCAATTTCTG-3’) were used to amplify the AoCA gene from locus AO090010000582 of EMBL: AP007175. InFusion™ PCR (Clontech) was used to clone the PCR product into an Aspergillus expression vector (described in [9]). One PCR error free plasmid was transformed into Ao strain BECh2 using a similar procedure as in [10]. The derived transformants were re-isolated twice under selective conditions on Cove-N minimal media plates. To test expression and secretion of AoCA, transformants were grown in YPM media (1% (w/v) yeast extract, 2% (w/v) peptone and 2% (w/v) maltose) for 3 days at 30 °C and fermentation samples analyzed by SDS-PAGE (NuPAGE, Invitrogen, Carlsbad, USA). A high yielding transformant was grown in YPM media for 3 days at 26 °C and 150 rpm shaking on a New Brunswick Scientific orbital shaker. Culture broth was filtered through a Whatman GF/F 0.7 μm glass fiber filter (Maidstone, UK) and then through a 0.22 μm filter. (NH4)2SO4 was added to this filtrate to 1.6 M final concentration prior to application on a Phenyl-sepharose FF (high sub) column (GE Healthcare) equilibrated in 20 mM HEPES/NaOH, 1.6 M (NH4)2SO4, pH 7.0. AoCA was eluted with a linear gradient between the equilibration buffer and 20 mM HEPES/NaOH, pH 7.0 over three column volumes. The AoCA peak was transferred to 20 mM HEPES/NaOH, pH 7.0 using a G25 sephadex column (from GE Healthcare) and applied to a Q-sepharose FF column (from GE Healthcare) equilibrated in 20 mM HEPES/NaOH, pH 7.0. AoCA was eluted with a linear 0–0.4 M NaCl gradient in the same buffer. Fractions judged as pure by SDS-PAGE were pooled and used for further characterization. Characterization: Intact Mw was analysed with DataAnalysis version 3.3 and measured using a Bruker microTOF focus electrospray MS (Bruker Daltonik GmbH, Bremen, DE) calibrated with ES tuning MIX (Agilent). The protein was N-terminal sequenced using a Procise Protein Sequencer from Applied Biosystems after it was recovered by SDS-PAGE on a PVDF membrane. Analytical gel filtration was carried out on a Superdex 75 (10/300 GL) column using 20 mM HEPES pH 7.0, 0.1 M NaCl as elution buffer at 4 °C. 100 or 300 μg of AoCA prepared as for crystallization were loaded. CA activity was measured essentially by the Wilbur–Anderson method [11] in a slightly modified form as described by Sigma–Aldrich (http://www.sigmaaldrich.com). 50 μl enzyme was diluted in a 20 mM Tris buffer at pH 8.3 and mixed with CO2-saturated water (3:2 v/v). The time, in seconds, for the pH of the mixture to drop to 6.3, was measured using a pH meter and compared to a blank without enzyme. CA esterase activity was determined by using p-NP-acetate as substrate as described in [12]. As a reference for activity measurements, bovine erythrocytes CA (beCA) [1] from Worthington was used. Crystallization: AoCA was dialyzed against 100 mM sodium carbonate buffer pH 8.0 and concentrated using a Microcon YM-10 unit (MWCO 10 kDa) to 103 mg/ml as estimated by A280 assuming an extinction coefficient of 24,535 M-1 cm-1. Screening was performed with the JCSG+, PACT and Index crystallization screens. The crystals used for structure determination resulted from optimization in Linbro plates. Rod-shaped crystals of approximate size 25 μm x 25 μm x > 100 μm grew in hanging drops within few days of setting up. The drop consisted of 3.0 μl protein solution and 1.0 μl of reservoir (2.2 M (NH4)2SO4, 0.1 M Buffer system II pH 6.5 [13]). Data collection: Crystals were mounted in nylon cryoloops in mother liquor containing 33% ethylene glycol. Data were collected at beamlines 911-2 and 911-3 of MAXLAB, Lund, Sweden, at 100 K. All data sets were processed with XDS [14]. A dataset suitable for structure determination was collected and the UCLA anisotropy correction server [15] was used to correct the significant anisotropy observed for the phasing step, but final refinement was performed against the unmodified dataset. A second lower resolution dataset was collected at the Zinc K-edge to obtain anomalous signal. The anomalous correlation from XDS was 17% overall and 32% for data up to 5.5 Å. Although this would not be sufficient for phasing, it could clearly identify the Zn2+ position after molecular replacement (MR) phasing. Data statistics are shown in Table 1 . Structure determination: The structure (Fig. 1 a) was determined by MR using an appropriately truncated structure of N. ghonorroeae CA (NgCA, [16]) as search model, which shares the highest sequence identity with AoCA among available structures. MOLREP [17] found two molecules in the asymmetric unit. Anomalous difference at the Zn K-edge showed clear peaks where active site Zn2+ ions were expected (Fig. 1b). The structure was rebuilt in COOT [18], and refined in REFMAC [19] and PHENIX [20] including TLS refinement and NCS restraints. Final refinement statistics are shown in Table 1. Structure figures were made with PYMOL [21]. Characterization: AoCA was expressed in high yield and purified. The theoretical Mw (Uniprot sequence Q2TWF5) is 29728.5 Da while the apparent Mw on SDS-PAGE was 35 kDa with a broad band suggesting glycosylation. N-glycosylation was confirmed by MS analysis prior and after EndoH treatment indicating a mass difference of 1825 Da, corresponding to 11 glycosyl units. The dominant N-terminal sequence was AAGGLDD, corresponding to a start at position 27 in the Uniprot sequence. A minor sequence, about 20% of the sample, had three additional amino acids removed. The apparent Mw was 78 kDa and 80 kDa for two separate gel filtration runs (Fig. 2 a), most consistent with a dimer as the predominant form in solution, thus confirming the dimer observed crystallographically (see below). CA activity was confirmed by two different assays. The specific activity of AoCA in the Sigma–Aldrich CA assay was 220 U/mg, that is 7% of beCA (3115 U/mg) used as reference. Addition of 2.5 mM imidazole decreased beCA activity by 3% but increased AoCA activity by 21%. The specific activity against pNP-acetate was 20 times lower for the AoCA compared to beCA although the Km values were similar for both (about 12 mM), when using 0.1 M phosphate buffer pH 7.5. Final model and overall quality: The final model includes 236 residues in each of the two chains. Residues 27–34, although confirmed to be present in 80% of the mature protein preparation by N-terminal sequencing, are not visible in the electron density maps, which probably results from a combination of sample proteolysis and disorder in the crystal. The electron density, and thus the resulting model, is less reliable at the chain termini. Final model statistics are in Table 1. 91.7% of residues were in the favoured regions and 99.4% in the allowed regions of the Ramachandran as determined by Molprobity [22]. The three Ramachandran outliers are F37 (both chains) and K269 of chain B, in disordered regions at chain ends, hence the difficulties in modelling. Despite their bad geometry, strong difference density appearing when these residues are removed supports their inclusion in the model, in what must be considered approximate positions. Consistent with MS results and bioinformatics predictions, one glycosylation site was observed per monomer. Only one N-acetyl glucosamine unit could be modeled at each site. Overall structure: the structure of the monomer (Fig. 1a), like for other α CAs, is dominated by a central twisted β sheet consisting of eight mostly anti-parallel strands (one exception), with strands β4, β1 and β11 extending the sheet in two separated sections. Six α helices in the periphery complete the secondary structure elements. Strands β5 and β6 bear the residues liganding the Zn cofactor. The active site is at the bottom of a deep cavity and roughly in the center of the molecule. The two monomers in the asymmetric unit are related by a 2-fold NCS axis and are very similar, with a Cα rmsd of 0.28 Å. 1812 Å2 of interface area are buried per monomer at the monomer/monomer interface suggesting a dimer in solution. Dimerization was confirmed by gel filtration (see above). A large part of the dimer interface is formed by N- and C-terminal regions (several residues between 37–65 and 252–266). The C-termini interact with each other and with the N-terminus of the other chain. In addition to this, residues 127 (loop), 134–136 (loop), 160–163 (end of strand β7 and a loop), 237–248 (helix α5 and a loop) complete the dimerization interface. The interface area is rather flat compared to the rest of the protein (Fig. 2c) and comprises both hydrophobic and hydrophilic residues. Because the two monomers interact intimately with each other, water is almost completely excluded, except for a small crevice across the interface, where water-mediated contacts are found. The active sites point away from the dimer interface, and are each formed by residues belonging to a single polypeptide chain, thus they are structurally independent. Oligomerization is a very unusual feature among α-class CAs, which are reported to be monomers with two notable exceptions, the membrane-associated and cancer related human CAs IX (hCAIX) and XII (hCAXII), where dimerization has been reported both in crystals and in solution [23, 24]. In the case of hCAIX, dimerization is stabilized by an intermolecular disulphide bridge as well as non-covalent interactions. Overall comparison to other CAs: The Cα rmsd between AoCA and NgCA (A chains) was 1.65 Å with 206 residues aligned. The AoCA monomer also overlaps well with the extensively studied hCAII, at least for the core secondary structure elements, thus we choose this as one of the mammalian structures for comparison. The Cα rmsd between AoCA A chain and hCAII (A chain, PDBid 3KS3, [25]) was 1.79 Å with 204 residues aligned, and most differences are to be found in inserted or deleted loops. Fig. 3 a shows an overall comparison between the three structures. Additionally we compared AoCA to hCAIX (A chain, PDBid 3IAI [24]) and hCAXII (A chain, PDBid 1JCZ, [23]), the two membrane-associated α-class CA reported as dimers. The Cα rmsd between AoCA A chain and hCAIX was 1.65 Å for 201 residues aligned. The Cα rmsd between AoCA A chain and hCAXII was 1.64 Å for 203 residues aligned. Fig. 3b shows a MUSTANG-MR [26] structure-based sequence alignment, with 11% of residues being identical in the five sequences. When pairwise structure-based sequence alignments were made AoCA showed 26.3% identity with NgCA, 21.9% with hCAII, 19.9% with hCAIX and 18.6% with hCAXII. In AoCA a disulfide bridge is formed between C58 and C219. An equivalent disulphide bridge is formed in many, but not all, α-CAs. In particular, it is conserved in NgCA, used here as MR model, and also in the membrane associated mammalian CAs, including hCAIX and hCAXII (Fig. 3b), where it has been proposed to maintain the orientation of the loop containing the important T199 (hCAII numbering, see below for role). However it must be pointed out that in hCAII, where the disulphide bridge-forming Cys are substituted by the hydrophobic residues Ala and Leu, the T199 conformation is essentially identical as in the disulphide containing structures (3, 4a). In fungal CAs the involved Cys are very highly conserved (Fig. 3c) and as in the mammalian counterparts, might contribute to stability. As seen in the MUSTANG alignment (Fig. 3b), the dimerization interfaces of AoCA and hCAXII only partially overlap, with almost no conservation of the involved residues. The AoCA and hCAIX dimer interfaces do not overlap at all. This is also very clear from Fig. 2b, where the dimer arrangements of the three CAs (AoCA, hCAIX and hCAXII) are compared. The interfaces were analyzed using the PDBsum server [27]. The AoCA dimer interface is by far the most extensive among the three proteins (about 1800 Å2 buried per monomer, with 14 putative H-bonds and 157 non-bonded contacts). hCAXII buries about two thirds of this area in dimerization (about 1100 Å2 buried per monomer, with 15 putative H-bonds and 141 non-bonded contacts). The interface of hCAIX is even less extensive, with only about 800 Å2 buried per monomer and only 2 putative H-bonds and 53 non-bonded contacts. Active site: the active site of AoCA is highly conserved compared to other α CAs. The three His residues that coordinate the catalytic zinc, H123, H125 and H142, are conserved and in essentially identical positions as the equivalent residues in hCAII (H94, H96 and H119). As in hCAII, H123 and H125 use NE2 for coordination to the Zn2+ ion, while H142 uses ND1. The three residues previously observed in hCAII to hold the three histidines in the correct orientations by hydrogen bonding, are conserved in AoCA as Q121, N249 and E140, and are conformationally equivalent. T215 (199 in hCAII) is also conserved in AoCA as is E129 (106 in HCAII), holding T215 in place. T215 normally hydrogen bonds and orients the hydroxide bound to the Zn2+ ion, this hydroxyl being the nucleophile reacting with CO2. In the AoCA structure this Zn2+ ligand is not observed, as binds A conserved hydrophobic where the substrate and product in other CAs is formed by and with hCAII in hence CO2 probably in the same The coordination position of the Zn2+ ions is usually by the nucleophile hydroxide or an equivalent water although a carbonate or from the mother liquor has been observed In our structure from the crystallization mixture this position and the active with the of Zn2+ at a of Å. The in the A (Fig. and molecules are slightly different hydrogen of the different of with the chain of T215 and In each at least one hydrogen is formed by the with each chain. in hCAII) is also considered part of the active site of CAs, where it a of in the position to the from and to the The conformation of is conserved in the AoCA structure as compared to hCAII, the water chain is not observed as is most of the site. The most difference between AoCA and most other mammalian CAs, is that the corresponding to in hCAII, the is a in AoCA (3, which up the same In this the chain from that of hCAII only for chain roughly the same position as in hCAII, from a loop in in AoCA is in a conformation that it from the to the active site is also associated with the dimer interface, although not part of it according to PDBsum being in contact with the of the chain from the other monomer. are normally considered in the hCAII a which function as a nucleophile and CO2 to and a usually as as a active enzyme. The ligand has been to be a hydroxide of is and crystallography have shown at this site This is not clearly in the AoCA structure presented which could from the is bound under the crystallization The in the is to be of a from the water that the in to the hydroxide nucleophile for the is to this step, and two in most structures (Fig. 4 suggesting it the or (e.g., are normally times less but this activity be to a large by the with However and very in AoCA the corresponding to is which would be to function as a In the structure, is in an conformation (Fig. the active site very to No other that could the of is with the of which is also conserved in CAs with the His as Consistent with this the specific activity of AoCA in the and assay was measured to about 7% of the specific activity of beCA which has a His corresponding to hCAII addition of small of imidazole increased the specific activity of AoCA but not of As is associated with the dimer interface, one could dimerization active site since is not an interface and in of other this is The and solution data presented here between AoCA and mammalian membrane associated CAs. In AoCA, hCAIX and hCAXII are all proteins and have a conserved disulphide the glycosylation site and dimerization interfaces are the of the disulphide bridge and both hCAIX and like hCAII, have as is thus at present to these are or they are of as the of dimerization has not been clearly for the membrane-associated CAs. dimerization and disulphide bridge could be In most CAs to the β but α-CAs are also present are in Fig. CAs in fungi [2] are essential for in involved in CO2 and in but the of different are comparison to fungal α-CAs reveals that only few AoCA dimer interface residues are highly with the is not to on the of sequence, dimerization is a predominant feature among fungal α The of a that function as an the of CO2 is the biological substrate for Only in one of the fungal α-CAs in the shown alignment the corresponding of hCAII is a His (Fig. the of the and biological of AoCA and other fungal α-CAs are AoCA is both a very and soluble protein that be and from industrial This AoCA a promising for industrial compared to mammalian CO2 capture is an of a promising and sustainable approach In application AoCA has shown in of and for and for for of at MAXLAB, Lund, Sweden, and is was by the from the for and the under the the