X-ray crystallographic studies reveal that the incorporation of spacer groups in carbonic anhydrase inhibitors causes alternate binding modes

Human carbonic anhydrases (CAs) are well studied targets for the development of inhibitors for pharmaceutical applications. The crystal structure of human CA II has been determined in complex with two CA inhibitors (CAIs) containing conventional sulfonamide and thiadiazole moieties separated by a —CF₂— or —­CHNH₂— spacer group. The structures presented here reveal that these spacer groups allow novel binding modes for the thiadiazole moiety compared with conventional CAIs.

1. Introduction

Carbonic anhydrases (CAs; EC 4.2.1.1) catalyze the interconversion of carbon dioxide and bicarbonate. The mammalian α-class of CAs (and the CA domains in more complex isoforms) are monomeric, with a molecular weight of approximately 30 kDa. At present, 14 human-expressed isoforms of CA (HCA I–XIV) are known, displaying widely varying tissue and subcellular distributions. They differ with regard to expression patterns, expression levels, kinetic properties and sensitivity to inhibitors (Tufts et al., 2003; Duda & McKenna, 2004).

There is a wealth of information available on various CA inhibitors (CAIs) and their kinetic and structural characterization (for a general review, see Mansoor et al., 2000; Supuran et al., 2004). Briefly, various anions have historically been useful for studying the properties of the Zn active site of CA and it has been shown that most monovalent anions bind and inhibit CA activity to varying extents. These CAIs bind to the Zn²⁺ ion and interfere with the Zn-bound hydroxide coordination by either displacing or replacing the hydroxide, thus disrupting catalysis. A key residue in determining this displacement/replacement is Thr199, which is thought to act as a hydrogen-bond acceptor, stabilizing the Zn-bound hydroxide (Merz, 1990). Small anionic inhibitors (e.g. azide and cyanate) can bind directly to the zinc ion, thus substituting for the Zn-bound hydroxide. Inhibitors that bind in this way maintain hydrogen bonding with Thr199 and participate in the tetrahedral coordination of the Zn²⁺ ion.

The most frequently used pharmaceutical class of CAIs are aromatic and heterocyclic sulfonamides of the form R—SO₂NH₂ or R—SO₂NH(OH). Several crystal complex structures of various sulfonamide inhibitors with various CAs have been determined and show similar interactions. The ionized NH group binds directly to the zinc and simultaneously donates a hydrogen bond to the hydroxyl group of Thr199 and an O atom of the sulfonamide group interacts with the amide backbone of Thr199 and in doing so displaces a second solvent molecule in the active site, termed the `deep water' (Nair et al., 1995). Fig. 1(a) shows the crystal structure of HCA II complexed with the clinically used inhibitor acetazolamide (Nair et al., 1995).

Figure 1 Structure of the active site of HCA II complexed with inhibitors. (a) Acetazolamide (AZM), (b) BB3, (c) TDM; (d) superimposition of all three: BB3, blue; TDM, red; AZM, green. The Zn atom is shown as a black sphere; side chains are as labeled. AZM coordinates were taken from PDB code 1yda (Nair et al., 1995). The red Fo − Fc electron-density map shown in (a) and (b) is contoured at 2σ and was calculated without the inhibitors present. Figures were generated and rendered with BobScript and Raster3D, respectively (Esnouf, 1999; Merritt & Bacon, 1997).

CAIs are commonly prescribed to treat a major symptom of glaucoma: increased intraocular pressure. The inhibition of CA suppresses the secretion of sodium, bicarbonate and aqueous humor, thus lowering intraocular pressure (Maren, 1987). Clinically important drugs for the treatment of glaucoma include Diamox (acetazolamide) and Trusopt (dorzolamide) (Breinin & Gortz, 1954; Lippa et al., 1992). Recently, aberrant expression of HCA IX has been shown to be associated with certain types of cancer, such as non-small-cell lung cancer and renal-cell carcinoma. It is thought that the presence of CA in solid tumors may help cancerous cells to adapt to hypoxic conditions and could increase the invasiveness of these tumors (Wykoff et al., 2000; Giatromanolaki et al., 2001). Currently prescribed CAIs such as acetazolamide are not very soluble and it is believed that a drug with better water solubility would have improved bioavailability for the topical treatment of glaucoma and possibly cancer (Kim et al., 2005). In the search for CAIs with improved water solubility and subsequent bioavailability, the Blackburn laboratory has recently synthesized a series of heterocyclic sulfonamides that have better water solubility than some of the conventional glaucoma treatments yet still bind HCA II with nanomolar affinity (Boyle et al., 2005). These compounds have an additional spacer group, —CF₂— or —CHNH₂—, between the classically employed sulfonamide moiety and the 1,3,4-thiadiazole ring (Figs. 1b and 1c, respectively).

We have determined the crystal structures of two of these compounds, namely 2-amino-1,3,4-thiadiazolyl-5-difluoromethanesulfonamide (BB3) and 2-dimethylamino-5-sulfonamido(aminomethyl)-1,3,4-thiadiazole (TDM), in complex with HCA II to 1.6 and 1.15 Å resolution, respectively. The structures show that the spacer allows the CAIs more conformational freedom and as a consequence alters the binding locations of the 1,3,4-thiadiazole ring within the active site. These alternate binding modes are compared with other CAIs. The use of such spacers may offer a method of making CAIs that (i) are more soluble, (ii) have higher affinity and (ii) provide new binding modes of heterocyclic sulfonamides within the active-site cavity whose ring could be modified to target specific amino-acid differences between the various CA isoforms.

2. Experimental procedures

2.1. Expression and purification

Human CA II was expressed and affinity-purified as described elsewhere (Khalifah et al., 1977). The protein purity of fractions was determined by silver-stained SDS–PAGE and fractions were pooled accordingly. Prior to crystallization, the protein was dialyzed against 50 mM Tris–HCl pH 7.8 and subsequently concentrated to ∼15 mg ml⁻¹ with Amicon Ultra centrifugation filtration devices with a molecular-weight cutoff of 10 kDa. Final concentrations were determined by measuring the optical density at 280 nm and using a molar absorbance of 5.5 × 10⁴ M⁻¹ cm⁻¹.

2.2. Crystallization and diffraction data collection

Crystals of HCA II were grown using the hanging-drop method at room temperature (McPherson, 1982). The crystallization drops were prepared by mixing 5 µl protein (∼15 mg ml⁻¹ in 50 mM Tris–HCl pH 7.8) with 5 µl precipitant solution (50 mM Tris–HCl pH 7.8, 2.6 M ammonium sulfate) and were equilibrated against 1000 µl precipitant solution. Useful crystals appeared within 5 d. The inhibitors BB3 and TDM were solubilized in water to a final concentration of 20 µM. Pre-grown crystals of HCA II were then soaked in the presence of BB3 and TDM (at a final concentration of ∼1 µM) for ∼24 h prior to data collection. The crystals were quick-dipped in cryoprotectant (30% glycerol in precipitant solution) before flash-freezing them at 100 K for X-ray data collection.

Diffraction data for the BB3-soaked HCA II crystals were collected at Brookhaven National Laboratory beamline X29 on a Quantum 315 CCD detector using a wavelength of 1.100 Å. The crystal-to-detector distance was set at 100 mm and oscillation images were collected in 1° steps for a total of 297 images.

Diffraction data for the TDM-soaked HCA II crystals were collected at Cornell High Energy Synchrotron Source beamline A1 on a Quantum 210 CCD detector using a wavelength of 0.978 Å. The crystal-to-detector distance was set at 100 mm and oscillation images were collected in 1° steps for a total of 410 images. Data-collection statistics are given in Table 1.

Table 1
Data-collection, refinement and final model statistics

Values in parentheses are for the highest resolution shell.

	BB3	TDM
Data-collection statistics
Space group	P21	P21
Unit-cell parameters (Å, °)	a = 42.3, b = 41.4, c = 72.1, β = 104.1	a = 42.5, b = 41.6, c = 72.7, β = 103.9
Resolution (Å)	20.0–1.60 (1.66–1.60)	20.0–1.15 (1.19–1.15)
Total No. of unique reflections	32031 (3154)	87379 (7896)
Completeness (%)	99.9 (100.0)	97.9 (89.3)
Redundancy	4.4 (4.3)	7.0 (4.1)
Overall I/σ(I)	17.1 (4.2)	26.4 (4.7)
Rsym†	0.066 (0.209)	0.099 (0.323)
Refinement statistics
Rcryst‡	0.167	0.136
Rfree§	0.212	0.166
R.m.s.d. for bond lengths (Å)	0.009	0.010
R.m.s.d. for angles (°)	1.490	1.315
Average B factors (Å^2)
Main chain	16.7	14.1
Side chains	23.2	17.5
Solvent	31.8	32.9
Inhibitor	21.5	16.5
Ramachandran statistics (%)
Most favored regions	87.9	88.8
Additionally allowed regions	12.1	11.2
No. of protein atoms	2039	2039
No. of solvent atoms	246	310
No. of inhibitor atoms	13	14

†R_sym = .
‡R_cryst = .
§R_free is calculated the same way as R_cryst for data omitted from refinement (5% of reflections for all data sets).

2.3. Structure determination and model refinement

X-ray diffraction data were indexed and scaled with DENZO and SCALEPACK from the HKL program package (Otwinowski & Minor, 1997). The structures were phased using molecular-replacement methods with the software package Crystallography and NMR System (CNS; Rossmann, 1990; Brünger et al., 1998). The structure of HCA II (PDB code 1tbt ) was used as the search model (Fisher et al., 2005). After one cycle of rigid-body refinement, annealing by heating to 3000 K with gradual cooling, geometry-restrained position refinement and temperature-factor refinement, F_o − F_c Fourier electron-density OMIT maps were generated (Figs. 1b and 1c). These maps clearly showed the position of the zinc and inhibitors, which were subsequently built into their respective electron density. The PRODRG server (https://davapc1.bioch.dundee.ac.uk/programs/prodrg/ ) was used to build coordinate and topology files for the inhibitors (Schüttelkopf & van Aalten, 2004). After initial refinement in CNS, R_work and R_free were 0.193 and 0.224 for the BB3 complex and 0.225 and 0.236 for the TDM complex, respectively.

Refinement of both models continued using SHELXL97 in the conjugate-gradient least-squares (CGLS) mode with SHELXL default restraints used for the protein geometry (Sheldrick & Schneider, 1997; Engh & Huber, 1991). After each round of ten cycles of CGLS refinement, Fourier electron-density maps (F_o − F_c and 2F_o − F_c) were calculated and minor manual model manipulations were performed using the molecular-graphics program Coot (Emsley & Cowtan, 2004). After several rounds of refinement, automated water divination was performed with SHELX/WAT using F_o − F_c electron-density OMIT maps with a 4σ cutoff (Sheldrick & Schneider, 1997). For the TDM complex all non-H atoms were refined anisotropically, whereas only the Zn and S atoms of the BB3 structure were refined anisotropically, using the SHELXL97 default anisotropic displacement parameters (Sheldrick & Schneider, 1997). Manual model building included the modeling of alternate conformations for Ile22 and Lys39 for the TDM-complex structure and the addition/deletion of solvent molecules in both structures.

Refinement of the structures continued until convergence of R_cryst and R_free was reached. The final R_cryst and R_free were 0.167 and 0.212 for the BB3 complex (1.6 Å resolution) and 0.136 and 0.166 for the TDM complex (1.15 Å resolution), respectively. The model geometries were analyzed by PROCHECK and inhibitor–enzyme inter­actions were determined with LIGPLOT (Laskowski et al., 1993; Wallace et al., 1995). All data-refinement and final model statistics are given in Table 1.

3. Results and discussion

Here, we present the crystal structures of HCA II in complex with two novel CAIs termed BB3 and TDM. These compounds differ from classical CAIs in that they contain a spacer group, —CF₂— or —­CHNH₂—, between the sulfonamide moiety and 1,3,4-thiadiazole ring for BB3 and TDM, respectively (Figs. 1b and 1c).

For both compounds the sulfonamide group maintains the same anchoring function as for other inhibitors in that the ionized NH group displaces the Zn-bound hydroxide and makes a hydrogen bond with Thr199 OG1. The N3 of the sulfonamide group is 2.8 and 2.9 Å from Thr199 OG1 in the BB3 and TDM structures, respectively. However, it was observed that the spacer and 1,3,4-thiadiazole ring of these compounds both adopt conformations in the active site of HCA II that differ from those of other CAIs (Fig. 1).

For BB3, the —CF₂— spacer permits conformational rotational freedom between the sulfonamide and thiadiazole ring. Subsequently, the thiadiazole ring of BB3 folds back and adopts a planar conformation and forms a π-stacking interaction with His94 (Figs. 1b and 2a). This is a very different configuration to that observed for other CAIs, where the heterocyclic ring adopts a more extended conformation that leads out of the active site. Acetazolamide is given here as a contrasting example (Figs. 1a and 1d).

Figure 2 Interactions between HCA II and inhibitors. (a) BB3, (b) TDM. Hydrogen bonds are indicated by dashed red lines (distances given in Å), side chains are as labeled and the Zn atom is shown as a black sphere. Figures were generated and rendered with BobScript and Raster3D, respectively (Esnouf, 1999; Merritt & Bacon, 1997).

There are also several other observed interactions that mediate the binding of BB3 to HCA II. The most interesting is the inferred hydrogen bond (2.5 Å) between F2 of BB3 and Thr200 OG1 (Fig. 2a). In addition to the fluorine hydrogen bond, the N1 and NH₂ of the thiadiazole ring directly hydrogen bond with Gln92 OE1 and Asn67 OD1, respectively. An indirect hydrogen bond also exists between NH₂ of the thiadiazole ring and Asn62 ND2 through a solvent molecule W369 (Fig. 2a).

In order to assess whether the π-stacking interaction with His94 of the thiadiazole ring of BB3 is specific to the —CF₂— spacer group, the structure of compound TDM, which has a —CHNH₂— spacer group, was investigated (Fig. 1c). In this complex the thiadiazole ring did not exhibit the π-stacking interaction with His94, but instead the ring adopted a puckered conformation and a different position relative to both BB3 and acetazolamide (Fig. 1d). N2 and N3 of the thiadiazole ring hydrogen bond to solvent molecules W316 and W599, which are in turn hydrogen bonded to Asn67. Interestingly, unlike the fluorine hydrogen bond observed for compound BB3, the —CHNH₂— spacer group does not appear to participate in any stabilization interactions (Fig. 2b).

The observed differences between the binding modes of BB3 and TDM may also in part arise from the dimethyl moiety of TDM, which is located in the hydrophobic region of the active site that contains amino acids Phe131 and Leu198 (Fig. 2b).

4. Conclusion

Carbonic anhydrases are a well studied target for the development of new inhibitors for pharmaceutical research and applications. CAIs are prescribed for the treatment of various diseases, such as glaucoma, and have potential for cancer therapy. Historically, CAIs have had poor solubility and bioavailability. We have determined the crystal structures of two novel water-soluble CAIs in complex with HCA II. These CAIs contain conventional sulfonamide and thiadiazole moieties with the addition of spacer groups. The spacers confer novel binding modes of thiadiazole within the active site that have not been observed in other sulfonamide-based CAIs. The ability to change the binding location of the heterocyclic ring of sulfonamide CAI may provide a useful tool in the development of CA isoform-specific inhibitors.

Footnotes

‡Present address: Rady School of Management, Pepper Canyon Hall, 3rd Floor, 9500 Gilman Drive, MC 0093, La Jolla, CA 92093-0093, USA.