Carbonic anhydrases (CAs; EC 4.2.1.1) are zinc-containing metalloenzymes that catalyze reversible hydration of CO₂ to bicarbonate and a free proton (CO₂ + H₂O ↔ HCO₃⁻ + H⁺).They are present in all kingdoms of life and play key roles in diverse biological processes such as CO₂ transport, respiration, and photosynthesis. To date, seven classes of CA genes have been identified (α, β, γ, δ, ζ, η, and θ) (Del Prete et al., 2014a; 2014b; Iverson et al., 2000; Kikutani et al., 2016; Meldrum and Roughton, 1933; Mitsuhashi et al., 2000; Xu et al., 2008). Members of the α-CA class are found in vertebrates, algae, green plants, fungi, and some eubacteria. They are among the fastest enzymes, with k_cat values up to ~10⁶ per second, which is close to the diffusion limit (Hasinoff, 1984; Pocker and Janjić, 1987). The β-CA class are predominantly present in bacteria, yeast, and plant chloroplasts (Neish, 1939; Smith et al., 1999), while γ-CAs are found in eubacteria and archaea (Alber and Ferry, 1994). Both δ- and ζ-CAs have been identified in the marine diatom Thalassiosira weissfogii (Lane et al., 2005; Roberts et al., 1997), and η- and θ-CAs were recently discovered in the pathogenic protozoan Plasmodium falciparum and a diatom Phaeodactylum tricornutum, respectively (Del Prete et al., 2014a; Kikutani et al., 2016). In most cases, CAs are active as zinc enzymes, but γ- and ζ-CAs can use iron (II) (Macauley et al., 2009; Tripp et al., 2004) and cadmium (II) (Xu et al., 2008), respectively.

Among the different classes, crystal structures of α-, β-, γ-, and ζ-CA families have only been characterized to date. α- and β-CAs share common catalytic features such as CO₂ hydration activity and a zinc ion as a metal cofactor, however, significant structural differences are apparent. First, most α-CAs form monomers or dimers, whereas β-CAs occur in various oligomeric states, including dimers, tetramers, hexamers, and octamers (Kimber and Pai, 2000). Second, in α-CA structures, three histidine residues and a water molecule strongly coordinate the zinc ion in the active site (Eriksson et al., 1988; Liljas et al., 1972). Crystal structures of human α-CA II (hCAII) in complex with the potent inhibitor acetazolamide revealed that the inhibitor displaces the zinc-bound water, thereby inactivating the enzyme (Nair et al., 1995). Many structures of α-CAs have enabled to propose the two-step catalytic mechanism where the first step is a deprotonation of zinc-bound water to form a hydroxide ion that actively participates in the catalytic reaction through nucleophilic attack of the CO₂ molecule. The second step is the regeneration of the zinc-bound hydroxide ion to reset the enzyme for a next catalytic cycle (Silverman and Lindskog, 1988). Proton transfer from a water molecule to generate a catalytic hydroxide ion is a two-part process. The first is the proton transfer through the six ordered waters to the solvent-exposed histidine residue (H64 in hCAII). The second is the transfer of the proton between H64 to the solvent (Eriksson et al., 1988; Fisher et al., 2007; Liljas et al., 1972). By contrast, crystal structures of β-CAs revealed two distinct zinc binding environments, denoted as active (R-state) or inactive (T-state) forms (Rowlett, 2010). In the R-state conformation, one histidine, two cysteines, and a water molecule coordinate a zinc ion in the active site. In the T-state, however, a water molecule is substituted with an aspartic acid residue, indicating that this state is catalytically inactive. β-CAs of Mycobacterium tuberculosis reportedly switch between these two states in a pH-dependent manner (Covarrubias et al., 2006; Suarez Covarrubias et al., 2005). Some bacterial β-CAs also alternate between active and inactive states by binding of bicarbonate in an allosteric manner (Cronk et al., 2006), or by redox-dependent disulfide bond formation that triggers release of the zinc ion (Nienaber et al., 2015). Crystal structures of γ-CAs have all been determined from the hyperthermophilic archeon Methanosarcina thermophila and Pyrococcus horikoshii (Iverson et al., 2000; Jeyakanthan et al., 2008). γ-CAs are active as homotrimers. The catalytic metal ion binds at the trimer interface and is coordinated by three histidine residues; Two histidines are from the same monomer and the third from the other (Ferry, 2010). Crystal structures of ζ-CAs are limited to cadmium-containing enzyme, CDCA1, in Thalassiosira weissflogii (Alterio et al., 2015; Xu et al., 2008). The active site of CDCA1 resembles those of the β-CAs; the metal ion is tetrahedrally coordinated by three conserved residues, one histidine, two cysteines, and a water molecule.

In the past decade, CO₂ capture and storage (CCS) technology utilizing hyperthermostable microbial α-CAs has attracted great attention for addressing global warming and climate change. Because CCS processes inevitably require harsh conditions such as high temperature (> 87°C) and strong alkaline (pH > 9) conditions, enzymes must be thermo-and alkali-stable. For this reason, much effort has been spent on finding highly thermostable CAs and/or engineering enzymes to increase thermal and pH stability. Persephonella marina EX-H1, which belongs to a new genus within the phylum Aquificales, is a marine chemolithoautotrophic thermophile isolated from a deep-sea hydrothermal vent at 9° N and 104° W in the Pacific Ocean (Götz et al., 2002). The α-CA enzyme from this organism lacking a signal peptide was expressed at levels five-fold higher than the wild-type protein (Kanth et al., 2014). In addition, it displayed a broad pH tolerance (pH 4-10) and high thermostability up to 100°C, making it a good candidate for catalyzing a fast, economically affordable, and environmentally friendly CCS process.

Here, we present the first crystal structure of an α-CA enzyme from P. marina EX-H1 (pmCA) in the absence and presence of acetazolamide. Various characteristics of typical hyperthermostable α-CA structures are well-conserved in pmCA, which has a relatively compact dimer conformation, an intramolecular disulfide bond, and a zinc ion coordinated by three histidine residues. Unexpectedly, our structure reveals a large surface patch of negatively charged residues that provides novel calcium binding sites at the crystallographic interface and appears to increase thermostability at alkaline pH range via electrostatic interactions with solvent molecules. This structural feature is unique to pmCA and has not been observed in any other CAs thus far.

Cloning, expression, and purification

Preparation of pmCA protein was performed as previously described (Kanth et al., 2014), with minor modifications. Briefly, the synthetic gene encoding α-CA from P. marina EX-H1 without a signal peptide (residues 20-243; Gene Universal, USA) was cloned between the NdeI and XhoI restriction enzyme sites in the pET21a expression vector, incorporating an N-terminal 6× His tag followed by a thrombin protease cleavage site. The protein was produced in Escherichia coli BL21(DE3) cells cultured in Lysogeny Broth medium at 37°C. When the absorbance (optical density [OD]) at 600 nm (OD₆₀₀) reached 0.6 to 0.8, protein expression was induced by addition of 1 mM isopropyl-β-d-thiogalactopyranoside for 16 h at 20°C. After harvesting by centrifugation at 7,700g for 10 min, cells were frozen in liquid nitrogen and stored at −80°C.

For purification, thawed cells in lysis buffer (20 mM MES pH 5.5, 200 mM NaCl, 100 mM DNaseI, and 0.1 mM phenylmethylsulfonyl fluoride) were disrupted using a microfluidizer. Cell debris was removed by centrifugation at 30,000g for 1 h and the supernatant was loaded onto nickel affinity resin (Incospharm, Korea) equilibrated with lysis buffer containing 20 mM imidazole. The protein was eluted using a gradient of increasing imidazole concentration. After thrombin cleavage, the protein was further purified by HiTrap SP (GE Healthcare, USA) cation exchange chromatography and Superdex 200 Increase (GE Healthcare) gel filtration chromatography. The protein was concentrated to 20 mg/ml for crystallization. All purification steps were performed either on ice or at 4°C.

CO₂ hydration activity assay

The CO₂ hydration activity of pmCA was measured as previously described using an electrometric method (Carter et al., 1969). Briefly, 4 ml of freshly prepared CO₂-saturated water was injected rapidly into 6 ml of reaction buffer (20 mM TRIS-HCl pH 8.5) in the presence or absence of purified pmCA at varying final concentrations. The CO₂ hydration activity of pmCA was measured by monitoring the time taken for the pH to change from 8.5 to 6.5. All preparations and reactions were performed on ice. The activity of pmCA was calculated in Wilbur–Anderson Units (WAU) using the following formula: WAU = (t₀ − t_c)/t_c, where t₀ and t_c refer to the time taken in the absence and presence of the enzyme, respectively (Wilbur and Anderson, 1948). For acetazolamide or anion inhibition, protein samples (6.15 nM) were mixed with varying concentrations of acetazolamide, nitrate (NO₃⁻), nitrogen dioxide (NO₂²⁻) and sulfate (SO₄²⁻) ions for 1 h on ice before measuring the decrease in pH. All measurements were repeated three times using the same preparation.

CaCO₃ precipitation assay

CaCO₃ precipitation assays were performed as previously described (Mirjafari et al., 2007). Briefly, 500 μl of CO₂-saturated water was mixed with the same volume of reaction buffer (1 M TRIS-HCl pH 9.0 or pH 11.0, 20 mM CaCl₂) containing varying final concentrations of purified pmCA. The reaction cuvette was sealed immediately to prevent CO₂ leakage. CaCO₃ precipitation was monitored by measuring changes in absorbance at 600 nm every 10 s to measure turbidity. For acetazolamide inhibition, protein samples (507 nM) were mixed with varying concentrations of acetazolamide for 10 min on ice before measuring the OD₆₀₀. All measurements were repeated three times using the same protein preparation.

Crystallization, X-ray diffraction data collection, and structure determination

Crystals of pmCA were grown at 22°C by the sitting-drop vapor diffusion method using a reservoir solution containing 100 mM sodium HEPES pH 7.5 (or glycine pH 9.5), 28–32% (w/v) polyethylene glycol (PEG) 400, and 200–400 mM calcium chloride dehydrate. For co-crystallization, 7 mM acetazolamide was added to the concentrated protein solution and incubated on ice for 30 min. Crystals appeared after 3 days and grew to full size within 1 week. Crystals were flash-frozen in liquid nitrogen after placing in reservoir solution containing ethylene glycol as a cryoprotectant. X-ray diffraction data were collected at beamlines 5C and 11C of the Pohang Accelerator Laboratory (PAL). Diffraction data were processed using the HKL2000 package (HKL Research, USA). The pmCA structure was determined by molecular replacement (Phaser, CCP4) using the α-CA structure from the Logatchev hydrothermal field metagenome (LOGACA, PDB code 6EKI) as a search model (Fredslund et al., 2018). Refinement of the structure was performed using REFMAC5 and PHENIX, with non-crystallographic symmetry and secondary structure restraints (Collaborative Computational Project, Number 4, 1994). The atomic model was constructed and manipulated manually after each refinement step using the program COOT (Emsley and Cowtan, 2004). X-ray crystallographic data and refinement statistics are summarized in Supplementary Table S1. All figures in the manuscript were prepared using the program PyMOL (www.pymol.org).

CO₂ hydration and CaCO₃ precipitation activities of pmCA

Direct CO₂ hydration assays were performed to confirm whether pmCA accelerates hydration of CO₂ and production of hydrogen ions. A plot of CO₂ hydration activity against time showed that increasing the pmCA concentration accelerated the drop in pH (Fig. 1A). Based on the Wilbur–Anderson equation (Wilbur and Anderson, 1948), the purified protein exhibited a specific activity of ~2,500 WAU/mg, which corresponds to ~50% of the previous measurements (Kanth et al., 2014). It is likely that the different environmental conditions can alter the rate of reaction caused by the pmCA enzyme.

We further explored the biological activity of pmCA by monitoring conversion from hydrated CO₂ to solid CaCO₃ in the presence of calcium ions. The biomineralization of carbon dioxide using CA enzymes is one of the most efficient ways to permanently store carbon. As expected for a CA enzyme, the rate of CaCO₃ precipitation was much faster with increasing pmCA concentration (Fig. 1B). As reported previously, no precipitate was observed below pH 9.0 (data not shown) because bicarbonate is converted to carbonate at pH > 8.3 (Han et al., 2006). Independent assays also showed that both CO₂ hydration and CaCO₃ precipitation activities were significantly inhibited in the presence of acetazolamide (Figs. 1C and 1D).

Utilization of CAs in CCS process requires the enzyme to be stable in the presence of flue gas contaminants such as NOx and SOx. To determine the effect of these anions on the pmCA activity, the major anion components in flue gas, nitrate (NO₃⁻), nitrogen dioxide (NO₂²⁻) and sulfate (SO₄²⁻) ions, were individually added to the enzyme at various concentrations, and incubated on ice for 1 h before measuring the CO₂ hydration. For comparison, enzyme activity in the absence of any inhibitor was set at 100%. As shown in Figure 1E, NO₂²⁻ and SO₄²⁻ anions had no inhibitory activity over the concentration ranges tested. In contrast, pmCA activity was stable in the presence of NO₃⁻ up to 0.1 μM, which corresponds to a pmCA/nitrate molar ratio of 1:16, although it retained only 30% to 57% activity beyond that. Our results suggest that pmCA are very resistant to NO₂²⁻ and SO₄²⁻, but it is weakly inhibited by NO₃⁻. Together, these functional assays demonstrate that purified pmCA under our experimental conditions has apparent in vitro enzymatic activity and a wide range of tolerance to NO₂²⁻ and SO₄²⁻ anions.

Overall structure of pmCA

To establish the molecular basis for thermal stability and CO₂ hydration of pmCA, we crystallized the purified pmCA protein with and without acetazolamide to determine its atomic structure by X-ray crystallography (Supplementary Table S1). The final models showed that two crystal forms of pmCA were obtained in the absence of acetazolamide; one crystallized in the monoclinic space group C2 (form 1) and the other in the orthorhombic space group P2₁2₁2 (form 2). The acetazolamide-bound form also crystallized in the P2₁2₁2 space group (form 3). Similar to other thermostable α-CAs, the crystal structure revealed that pmCA forms a homodimer, consistent with previously reported gel filtration data (Fig. 2A) (Kanth et al., 2014). All crystal forms contain three pmCA dimers within the asymmetric unit (Supplementary Fig. S1). Within each dimer, one monomer is oriented approximately 180° with respect to the other monomer, thereby resulting in overall dimensions of 41.7 × 41.4 × 78.2 Å. The pmCA structure includes typical features found in other α-CAs; namely (i) a 10-stranded β-sheet in the core, (ii) a catalytic zinc ion coordinated in the central active site, and (iii) an intramolecular disulfide bond formed between C44 and C197 (Fig. 2A). The disulfide bridge provides conformational and thermal stability by increasing structural rigidity, as supported by previous cysteine mutagenesis data showing that addition of an intramolecular disulfide bond markedly enhances the thermostability of hCAII and Neisseria gonorrhoeae NgCA (Boone et al., 2013; Jo et al., 2016; Mårtensson et al., 2002).

Structural comparison of pmCA with other thermostable α-CAs from Sulfurihydrogenibium yellowstonense YO3AOP1 (SspCA) (Di Fiore et al., 2013), Sulphurihydrogenibium azorense (SazCA) (De Simone et al., 2015), Thermovibrio ammonificans (TaCA) (James et al., 2014), Photobacterium profundum (PprCA) (Somalinga et al., 2016), NgCA, and LOGACA revealed a similar overall structure, as expected based on the high sequence identity (Supplementary Figs. S2 and S3) (Heinig and Frishman, 2004). However, superposition of pmCA and hCAII (Eriksson et al., 1988) revealed that, similar to SspCA, NgCA, and PprCA, three large surface loops (loops 1–3) are also deleted in pmCA, resulting in a more compact fold compared with the hCAII structure (Supplementary Fig. S4). In particular, loop 2 containing an α-helix formed by six residues (D130-V135) is located on the rim of the hCAII active site. Therefore, deletion of a corresponding α-helix from pmCA may enlarge the cavity entrance, resulting in a more solvent-exposed active site and easier access for the CO₂ substrate to the active site. Hereafter, a single apostrophe signifies residues from a neighboring monomer within a biological dimer or symmetry-related molecule in the crystal.

The pmCA dimer interface

Analysis of the pmCA structure using the Proteins, Interfaces, Surfaces, and Assemblies (PDBePISA) server (http://www.ebi.ac.uk/pdbe/pisa/) shows that the pmCA homodimer interface covers an area of ~1,024 Ų and is semi-oval in shape (Fig. 2B). The dimer interface is highly hydrophilic, similar to the solvent-exposed surface, indicating that the pmCA dimer is held together mainly by symmetrical hydrogen bonding in the core region, surrounded by additional ionic and hydrophobic interactions at the periphery (Fig. 2B, Supplementary Fig. S5). These strong, specific interactions appear to contribute to the excellent thermal stability of pmCA. Furthermore, a number of water molecules are bound at the interface; these likely participate in additional interactions in the hydrogen bonding network, further stabilizing the dimer interface. Among them, two well-ordered symmetric water molecules are located at the center of the dimer interface that form multiple hydrogen bonds with S189, the backbone oxygen of G190, the backbone nitrogen atoms of S189 and G200, and N49′ (Fig. 2C). Such completely buried water molecules are commonly found in other bacterial α-CAs, and the coordinating residues are highly conserved, except for S189 (Supplementary Fig. S3). TaCA (James et al., 2014) and SspCA (Di Fiore et al., 2013) also have serine residues in the corresponding position, while NgCA (Huang et al., 1998) and Helicobacter pylori HpCA (Modak et al., 2015) have alanine and asparagine, respectively. Interestingly, PprCA also has an asparagine at the equivalent location, which acts as a key residue for chloride ion coordination (Somalinga et al., 2016).

The pmCA calcium binding site

In our structures, calcium ions from the crystallization buffer were bound at the crystallographic interface of pmCA with known geometrical features (Fig. 3A). To the best of our knowledge, this feature has not been observed in previous α-CA structures. As shown in Figure 3B, there is a large patch of negative surface charge formed by a cluster of glutamic acid and aspartic acid residues. This feature facilitates direct or water-mediated binding of calcium ions in close proximity by acting as a ‘molecular glue’ to create crystal contacts by neutralizing the negatively charged and otherwise repulsive surfaces. In our structures, two different types of Ca²⁺ binding sites were observed. In crystal form 1, binding sites of five calcium ions starts near the center of the crystal interface, which is bracketed by a pair of two-fold symmetry-related interfaces (Fig. 4A). Ca1 interacts with four ligands, including D182 and D184 and their two-fold symmetry-related residues in the crystal. Binding site for Ca2 is coordinated by the side chains of D60, E242, E208′, and E209′. Ca3 is coordinated by the side chains of D60, E208′, and E209′, the backbone oxygen of E209′, and a water molecule. Ca2 and Ca3 are only ~4.3 Å apart, and are interlinked through the side chain oxygen atoms of D60, E208′, and E209′. In crystal forms 2 and 3, however, binding sites of six calcium ions are formed by two interfaces with two-fold symmetry (Fig. 4B). Ca1 interacts with six ligands (comprising the side chains of D184, E207, and D182′) and three water molecules in an approximately octahedral geometry. The binding site for Ca2 is coordinated by the side chains of E242, E208′, E209′, and four water molecules. Ca3 is coordinated by the side chains of D60, E208′, and E209′, the backbone oxygen of E209′, and three water molecules. In crystal forms 2 and 3, calcium ions are 5 to 6 Å apart from each other and are interlinked through bridging water molecules. Compared to hCAII, enrichment of negative charges on the solvent-accessible molecular surface of pmCA may increase the thermal and/or alkaline pH stability because acidic residues are easily deprotonated and have a high water-binding capacity under alkaline pH conditions (Supplementary Fig. S6) (Britton et al., 2006; Frolow et al., 1996; Kuntz, 1971). Therefore, negatively charged residues on the surface form the protective hydration shell that can confer high pH stability to pmCA. This structural feature is consistent with previous data showing that ~60% of pmCA activity is retained at pH 10, compared with the maximum activity at pH 7.5 (Kanth et al., 2014). Similarly, it has been reported that alkaline active phosphoserine aminotransferase and xylanase have large number of negatively charged residues on their surfaces (Dubnovitsky et al., 2005; Mamo et al., 2006; Manikandan et al., 2006).

PEG molecules are bound at the pmCA crystallographic interface

Together with the calcium ions described above (Fig. 3A), two fragments of PEG molecules were also visible at the crystallographic interface, and the protein-PEG interactions also appear facilitate crystal packing (Fig. 4C). Specifically, we observed a pair of symmetrical PEG binding sites close to the calcium binding sites (Fig. 3A). The binding mode of PEG molecules are practically identical in all crystal forms. Binding of PEG molecules is stabilized predominantly through hydrophobic interactions with L116′ and Y121′ (Fig. 4C). F142′, located behind Y121′, is not directly involved in PEG coordination, but contributes to orient Y121′ via π-π interactions, enabling Y121′ to interact effectively with PEG. Moreover, five oxygen atoms of PEG engage in hydrogen bonding and ionic interactions with K117 and Y186 of one pmCA monomer, and Y121′ of a neighboring molecule in the crystal.

The pmCA active site

The catalytic active site of pmCA resembles those of previously studied α-CAs. The active site is located in a cone-shaped cavity in which a zinc ion is bound at the bottom (Fig. 5A). The zinc ion is tetrahedrally coordinated by three histidine residues (H107, H109, and H126), and by a replaceable fourth ligand. In crystal form 1, a strong, positive Fo-Fc peak (> 4σ), which is planar and triangular in shape, was observed in each active site. We concluded that this electron density corresponded to bicarbonate, a hydrated CO₂ species (Fig. 5A). In crystal form 2, a small globular peak was observed in the corresponding position in the Fo-Fc map (> 3σ), and was modeled as a water molecule (Fig. 5B). The crystal structure of pmCA complexed with acetazolamide revealed its binding mode in the active site that is nearly identical to that observed in hCAII, NgCA, and SspCA enzyme-inhibitor structures (Di Fiore et al., 2013; Huang et al., 1998; Vidgren et al., 1990) (Fig. 5C). Specifically, (i) the nitrogen atom of the sulfonamide group of acetazolamide displaces the zinc-bound water and the deeply buried water molecule in the active site, (ii) the sulfonamide group forms hydrogen bonds with the backbone nitrogen and side chain hydroxyl group of T193, (iii) two nitrogen atoms of the thiadiazole ring interact with T194 via hydrogen bonds, and (iv) the acetylamide group is bound to Q105.

To evaluate the surface chemistry of the pmCA active site, we analyzed the spatial distribution of hydrophobicity and hydrophilicity on the molecular surface. Analysis of the surface characteristics revealed polarization of surface hydrophobicity; half of the active site wall is predominantly polar and hydrophilic whereas the other half of the cavity contains a single linear hydrophobic patch (Fig. 5D). Several residues on the hydrophilic half of the cavity are catalytically important. For example, residues Y25, H82, E113, and T193, along with a zinc-bound water (or hydroxide), form a hydrogen bonding network and transfer protons from the active site to the bulk solvent (Fig. 5D) (Eriksson et al., 1988; Silverman and Lindskog, 1988; Silverman and Vincent, 1983; Tashian, 1989). By contrast, residues responsible for CO₂ binding, including V128, V138, L136, L192, V201, and W203, constitute the hydrophobic half of the cavity, and mutating these residues results in a small loss of enzyme activity (Merz, 1990; 1991; Xue et al., 1993). Our structure revealed that the key elements for substrate binding and catalysis are well-conserved, suggesting that pmCA performs the same CO₂ hydration mechanism carried out by other α-CAs.

In summary, we determined the first crystal structure of pmCA and provide direct evidence of dimerization, extensive intermolecular interactions, an intramolecular disulfide bond, and a large negatively charged surface patch. These characteristics explain its high pH and thermostability. These findings could be exploited to engineer α-CAs with improved pH and thermostability for use in harsh conditions such as those employed in CCS.