Inhibition of mammalian carbonic anhydrases I-XIV with grayanotoxin III: solution and in silico studies

Abstract

Grayanotoxin III (GTX3) was investigated for inhibition of all catalytically active mammalian carbonic anhydrase (CA, EC 4.2.1.1) isoforms, i.e. CA I to CA XIV. It showed micromolar inhibition (K_Is of 8.01 and 6.13 µM) for cytosolic isoforms CA I and II, respectively. GTX3 showed a submicromolar inhibition (K_Is in the range of 0.51–2.15 µM) for the remaining cytosolic (CA III, VII and XIII), membrane-associated/transmembrane (CA IV, IX, XII and XIV), mitochondrial (CA VA and CA VB) and secreted (CA VI) isoforms. This inhibition profile is very different from that of the sulfonamide CA inhibitors (CAIs), which possess different clinical applications. A molecular docking study for GTX3 within the active sites of CA I and II assisted to the understanding of molecular mechanism of the ligand. The interesting inhibition profile, coupled with various possibilities of interacting with the enzyme active site make this family of natural compounds attractive leads for designing novel chemotypes acting as CAIs.

• Carbonic anhydrase
• grayanotoxin III
• molecular docking

Introduction

Grayanotoxins (GTXs) are tetracyclic diterpenoid compounds found in many plants belonging to the Ericaceae family and are poisonous to mammals1–4. The most investigated grayanotoxin intoxications involves eating the so-called “mad honey”, i.e. honey contaminated by Rhododendron spp. nectar3. This is one of the common food intoxications encountered in humans and livestock in Turkey and Japan2,3. GTX intoxication symptoms include dizziness, weakness, excessive perspiration, hypersalivation, nausea, vomiting and paresthesias3. The severe intoxication may lead to life-threatening cardiac complications such as complete atrioventricular block, due to a specific increase in resting sodium permeability as well as activation of voltage-sensitive sodium channels3,5. Possible involvement of muscarinic cholinergic neurons in the GTX-III-induced salivation were also suggested5. Recently, honey extracts from the endemic plant in the Black Sea region Rhododendron ponticum, were investigated for their inhibitory effects against the metalloenzyme carbonic anhydrase (CA, EC 4.2.1.1), more precisely the human (h) isoforms hCA I and II, by our groups6. We have hypothesized that these enzyme inhibitory effects may be due to the presence of GTXs (and more specifically, GTX III, abbreviated here as GTX3) in the mad honey samples, since honey from other types of flower than R. ponticum were devoid of such properties6.

The CAs are widespread metalloenzymes found in organisms all over the tree of life, including vertebrates7–10. In such organisms, 16 isozymes have been characterized to date, many of which are involved in critical physiological processes. They catalyze the following reaction7–11: CO₂ + H₂O↔H⁺+ . In humans, CAs are present in a large variety of tissues including the gastrointestinal tract, the reproductive tract, the nervous system, kidneys, lungs, skin and eyes10. The different isozymes are localized in different parts of the cell with CA I and CA II, important isozymes in normal cells, being localized in the cytosol10. Many of the CA isozymes are important therapeutic targets with the potential to be inhibited to treat a range of disorders7–12. CA II plays a role in bicarbonate production in the eye and is, therefore, a target for therapy of eye disease such as glaucoma7. Indeed, CA inhibitors (CAIs) of the sulfonamide type such as dorzolamide DZA or brinzolamide BRZ are topically used antiglaucoma agents7–10, whereas the older drugs, such as acetazolamide AAZ or dichlorophenamide DCP show the same action through systemic administration, which however leads to a wide range of side effects due to inhibition of the enzyme from other organs than the target one, i.e. the eye7 (Scheme 1)11–15. CA IX and CA XII, are transmembrane isoforms with an extracellular active site, which were shown to be overexpressed in many types of tumors.

Scheme 1. 2D molecular structures of GTX3, DZA, BRZ, DCP, and AAZ.

As some solid tumors grow in cancer patients, hypoxic regions are formed, particularly in the interior of the tumor13. The gene expression profile of a hypoxic cancer cell is different from that of other cancer cells in a normally-oxygenated environment, i.e. in normoxic conditions11–15. Under hypoxia, the distribution of CA isoforms is altered as compared with normoxic cells11–14. CA IX and CA XII are both extracellularly localized on hypoxic tumor cells13–15. These enzymes play various roles in tumorigenesis, by regulating pH inside and outside the tumor cells11, interfering with phosphorylation of various proteins or by playing a role in the cell-cell adhesion11–15. They therefore provide a target for cancer therapy because they are relatively specific to the hypoxic tumor cells and appear to be important in their survival and proliferation11. Indeed, several antibodies targeting CA IX are in Phase III clinical development for the treatment of solid tumors (or for their imaging) whereas some small molecule inhibitors are also in advanced preclinical evaluation16,17.

The classical CA inhibitors (CAIs) are the sulfonamides and their isosteres (sulfamates, sulfamides, etc)7,17,18. However, most of these compounds non-selectively inhibit many of the 16 CA isoforms known to date in mammals7–11. Thus, efforts have been made to find different CAIs, from the sulfonamide, sulfamate and sulfamide ones. Indeed, recently the coumarins19 were discovered as mechanism-based inhibitors which act as prodrugs and bind in a very different mode compared to sulfonamides and their isosteres20, whereas some polyamines (such as spermine)21, as well as a range of natural product phenols22,23 were also investigated and showed interesting such properties and novel mechanisms of inhibition. Following our preliminary communication6 on the effects of mad honey against CA I and II, we report here an inhibition study of all catalytically active mammalian CA isoforms (i.e. CA I–CA XIV) with GTX3. As we were unable to obtain high-quality crystals of the CA–GTX3 adducts (for X-ray crystallography), we also performed a molecular docking study for the GTX3 within the active sites of isoforms CA I and II, in order to understand the underlying inhibition mechanism of the compound against CAs.

Results and discussion

Pure GTX3 has been investigated for the inhibition of all human (h) CA isoforms possessing catalytic activity, i.e. hCA I–hCA XIV. Indeed, three isoforms (hCA VIII, X and XI) are devoid of catalytic activity as they miss one or more of the histidine residues coordinating the catalytically crucial Zn(II) ion from the enzyme active site7–10. We used a stopped-flow, CO₂ hydrase assay for measuring CA activity and inhibition24.

Data of Table 1 show indeed GTX3 compound to be a CAI. For comparison reasons we also present the CA inhibition data of isoforms hCA I–hCA XIV with the sulfonamide CAI in clinical use dichlorophenamide (DCP). The inhibition profile of GTX3 is very particular, as the toxin shows a low micromolar inhibitory activity against the widespread, house-keeping isoforms hCA I and II (K_Is of 8.01 and 6.13 µM, respectively), whereas the remaining isoforms (hCA III-XIV) are inhibited more effectively, with K_Is in the range of 0.51–2.15 µM. The best inhibition was observed against the hCA IV (K_I of 0.51 µM), followed by the cytosolic isozyme hCA XIII (K_I of 0.69 µM). The remaining isoforms (hCA III, VA,  VB, VI, VII, IX, XII and XIV) showed a rather compact behaviour, as the inhibition constant of GTX3 did not changed too much, being in the range of 0.74–2.15 µM. On the contrary, the sulphonamide inhibitor DCP is ineffective against CA III (K_I of 680 µM), has a micromolar activity against hCA I and IV (K_Is of 1.20–15.0 µM), and it is a nanomolar inhibitor of the other isoforms (hCA II, VA, VB and VI-XIV) against which the inhibition constants varies between 21 and 345 nM. It is well known that all sulfonamides investigated to date (as well as their isosteres, such as the sulfamates and sulfamides) coordinate to the Zn(II) ion from three CA active sites by means of the deprotonated sulfonamide nitrogen atom, making in addition a lot of other favourable interactions with amino acid residues from the enzyme active site25–30. Only recently we have discovered that coumarins20 and lacosamide31 bind in a different binding site, at the entrance of the cavity, being around 7 Å away from the metal ion. It should be mentioned that both sulfonamide isoform-selective CAIs as well as coumarin such derivatives, are in preclinical evaluations as antitumor/antimetastatic agents targeting hypoxic cancers11,16,17.

Table 1. Inhibition data of mammalian isoforms hCA I-XIV with grayanotoxin III, and the sulfonamide inhibitor dichlorophenamide DCP (as standard) by a stopped-flow CO₂ hydrase assay24.

	KI (µM)*
Compound/Isoform	GTX3	DCP
hCA I	8.01	1.20
hCA II	6.13	0.038
hCA III	0.76	680
hCA IV	0.51	15.0
hCA VA	1.14	0.63
hCA VB	0.74	0.021
hCA VI	0.86	0.079
hCA VII	1.26	0.026
hCA IX	0.98	0.050
hCA XII	2.15	0.050
hCA XIII	0.69	0.023
hCA XIV	1.47	0.34

*Mean from three different determinations. Errors were in the range of ± 10% of the reported values.

As we were unable to obtain good crystals of GTX3 complexed to CAs, for X-ray crystallographic experiments, we performed molecular docking simulations of the molecule within the active sites of hCA I and II, in order to have a detailed insight regarding the inhibition mechanism of these natural compounds, which do not belong to well-known chemotypes inhibiting these enzymes32–35. The Induced-Fit docking (IFD) module which uses the Glide docking program to account for flexibility of the receptor has been used for docking under the Scrodinger molecular modeling package36–42. Coordinate files of the two targets were available in PDB: file 2FW443 for hCA I docking and file20 3F8E for hCA II docking were employed.

As seen from Figures 1–3, the tetracyclic scaffold of grayanotoxin is easily accommodated within the active site of both CA isoforms used for the docking studies, hCA I and II. However, from the beginning we observed a notable difference in the location of the inhibitor molecule, which is buried deeply within the hCA I active site (Figure 1A) but found more towards the entrance of the cavity (Figure 1B) of hCA II. It is interesting to note that the active sites of these two CAs are very much similar except for the presence of a bulky His residue near the Zn(II) ion, His200 in hCA I, which in hCA II is Thr20025,29,31,43. As a consequence, the hCA I active site is tighter at the bottom of the cavity compared to hCA II, whereas towards the middle part and the entrance to the cavity they more or less have the same shape (although there are several amino acid residues which differ between these two cytosolic isoforms)25,29,31,43. The following close contacts between the inhibitor and the following amino acid residues of hCA I are observed (within 3 Å radial cavity of binding pocket amino acid residues from the inhibitor are listed here): His64, His67, His94, Gln92, Phe91, Ala121, Val62, Leu198, His200, Pro201, Pro202 and Trp5. These amino acid residues were reported earlier to be involved in the binding of inhibitors/activators to the CA I active site (Figures 2 and 4)25,29,31,43. On the other hand, GTX3 was found in the coumarin-binding site of hCA II (Figures 3 and 4)20, at around 8 Å away from the zinc ion. Obviously the inhibitor is not coordinated to the metal ion, but interacts by means of hydrogen bonds and van der Waals interactions with the following amino acid residues situated at the entrance of the cavity (within 3 Å radial cavity of binding pocket amino acid residues from the inhibitor are listed here): Asn67, Glu69, Gln92, Phe131, Val121, Ile91, Trp123, Leu141, Val135, Leu198 and Gly132. Three water molecules (HOH278, HOH291 and HOH452) form hydrogen bonds with the hydroxyl moieties found in the GTX3 molecule (Figures 3 and 4). This binding is very much similar to that of the cis-2-hydroxycinnamic acid derivative obtained by the in situ hydrolysis of a coumarin natural product, identified as the first CAI belonging to this new class of inhibitors20. Also for that compound, many of the amino acid residues mentioned above, were observed to be involved in its binding in the enzyme-inhibitor adducts analyzed by means of high-resolution X-ray crystallography20.

Figure 1. Binding of grayanotoxin III (GTX3) to hCA I (A) and hCA II (B). The protein backbone is represented with residue position default colors in Maestro, the catalytic Zn(II) ion as the pale brown sphere and GTX3 is shown as stick model (carbon gray, oxygen red, hydrogen white).

Figure 2. Detailed interactions of grayanotoxin III (GTX3) when bound to the hCA I active site. Docking score at hCA-I: −8.21 kcal/mol (Glide Induced Fit Docking). The protein backbone is represented with residue position default colors in Maestro, the catalytic Zn(II) ion as the pale brown sphere and GTX3 is shown as stick model (carbon gray, oxygen red, hydrogen white). Amino acid residues and water molecules involved in the interaction with grayanotoxin III are also evidenced (CPK colors; hCA I numbering system).

Figure 3. Detailed interactions of grayanotoxin III (GTX3) when bound to the hCA II active site. Docking score at CA-II: −8.29 kcal/mol; (Glide Induced Fit Docking). The protein backbone is represented with residue position default colors in Maestro, the catalytic Zn(II) ion as the pale brown sphere and GTX3 is shown as stick model (carbon gray, oxygen red, hydrogen white). Amino acid residues and water molecules involved in the interaction with grayanotoxin III are also evidenced (CPK colors; hCA I numbering system).

Figure 4. Ligand interaction diagrams for (A) hCA I and (B) hCA II. Residues are labeled and colored according to residue type (green, hydrophobic; cyan, polar; red, negative; purple, positive). Yellow markers on the ligand denote per-atom solvent accessible surface area and marker size shows the amount of exposure. Labeling of residue (font sizes) is used to represent the depth of a residue in the three dimensional view (i.e. “closer” to the viewer are shown in a large font, whereas, residues that are “far away” are shown with small font. While hydrogen bonds to side chains are represented with dashed, hydrogen bonds to backbones are represented as solid lines.

Docking scores are summarized in Table 2. Parallel to the experimental results, Glide XP score is slightly favourable for CA II (−8.29 kcal/mol) compared to CA II (−8.21 kcal/mol). IFD scores also showed same trend with Glide XP scores that CA II (−503.87 kcal/mol) is slightly more favorable than CA I (−494.58 kcal/mol).

Table 2. Details of molecular docking results of GTX3 within the active sites of hCA I and hCA II using Induced Fit Docking (Glide/XP). Energy units are in kcal/mol.

	Glide_ Gscore	Glide_Lig_ efficiency In	Glide_ energy	Glide_ hbond	Glide_ lipo	Glide_ Ecoul	Glide_ Einter	Glide_ Evdw	Glide_Lig_ Efficiency	Glide_Lig_ eff_SA	Prime_ Coulomb	Prime_ Covalent	Prime_ VDW	Prime_ Sol_SA	Prime_ Sol_GB	Prime_ Energy	XP_ Hbond	IFD_ Score
hCA-1	−8.21	−1.93	−39.54	−0.91	−1.50	−13.81	0.91	−25.73	−0.32	−0.94	−8801.40	1346.85	−1325.02	59.50	−1007.33	−9727.40	−2.82	−494.58
hCA-2	−8.29	−1.94	−38.67	−0.51	−1.30	−13.23	0.89	−25.44	−0.32	−0.95	−8812.78	1342.83	−1401.30	96.11	−1132.16	−9911.30	−2.29	−503.87

In conclusion, we report here that grayanotoxin III is micromolar inhibitor (K_Is of 8.01–6.13 µM) of the widespread cytosolic isoforms CA I and II, and a submicromolar inhibitor (K_Is in the range of 0.51–2.15 µM) of the remaining cytosolic (CA III, VII and XIII), membrane-associated/transmembrane (CA IV, IX, XII and XIV), mitochondrial (CA VA and CA VB) and secreted (CA VI) isoforms. This inhibition profile is very different from that of the sulfonamide inhibitors, which have many clinical applications (antiglaucoma, diuretic, antiobesity, anticonvulsant or antitumor agents). The interesting inhibition profile, coupled with various possibilities of interacting with the enzyme active site make this natural compound an attractive lead for designing novel chemotypes acting as CAIs.

Among such botanical species which produce toxic nectar, all belonging to the Ericacee family, five are Rododendrum. Rhododendron ponticum subsp. baeticum is one of the most extensively cultivated rhododendrons in western Europe44, and this and other related species became highly invasive over most Western Europe and New Zealand. They are on the other hand endemic in Turkey, in particular, R. ponticum and R. flavum, and populate the forests of the North of Anatolia. Rhododendron ponticum was introduced in 1763 in United Kingdom and thereafter widely diffused as invasive species44. The non controlled diffusion of R. ponticum is therefore a major concern for the public authorities in order to avoid the risk of alimentary contamination for the consumers and for the protection of the relevant economic sector of honey production45. Territory safeguard, biodiversity conservation, ecological function of the pollination, maintenance of the traditional and cultural heritage are only some of the issues connected with apiculture. In particular it represents one of the strategic sectors for the rural development in Turkey45. At the international level Turkey is at the second place for the number of bee colonies and at the third place for the production of honey, giving work to 240 000 people, with economic revenue of about 205 million dollars/year45. The need to control the expansion of R. ponticum, in addiction to be an internal problem, represents an emergency that Turkey has to face towards the international community as 95% of the produced honey is exported. A contamination of the honey would therefore produce, beside a relevant economical damage, a loss of image difficult to recovery at the international level45.

Indeed mad honey may lead to severe intoxications, as shown in the introduction of this paper. Our study thus, may shed light regarding the biological activity of grayanotoxins present in this type of honey, for the inhibition of enzymes not normally associated with these toxins, but it also offers some interesting opportunities for the drug design of non-sulfonamide CAIs.

Experimental section

Chemistry. GTX3 and DCP were from Sigma-Aldrich (Milan, Italy) whereas all CA isoforms were recombinant enzymes obtained in-house as reported earlier20.

CA inhibition. An Applied Photophysics stopped-flow instrument has been used for assaying the CA catalysed CO₂ hydration activity. Phenol red (at a concentration of 0.2 mM) has been used as indicator, working at the absorbance maximum of 557 nm, with 20 mM Hepes (pH 7.5) as buffer, and 20 mM Na₂SO₄ (for maintaining constant the ionic strength), following the initial rates of the CA-catalyzed CO₂ hydration reaction24 for a period of 10–100 s. The CO₂ concentrations ranged from 1.7 to 17 mM for the determination of the kinetic parameters and inhibition constants. For each inhibitor at least six traces of the initial 5–10% of the reaction have been used for determining the initial velocity. The uncatalyzed rates were determined in the same manner and subtracted from the total observed rates. Stock solutions of inhibitor (0.1 mM) were prepared in distilled-deionized water and dilutions up to 0.01 nM were done thereafter with distilled-deionized water. Inhibitor and enzyme solutions were preincubated together for 15 min at room temperature prior to assay, in order to allow for the formation of the E-I complex (GTX3 was incubated also for longer periods, of 1–24 h, but no differences of activity have been detected). The inhibition constants were obtained by non-linear least-squares methods using PRISM 3, as reported earlier16,17, and represent the mean from at least three different determinations. All CA isofoms were recombinant ones obtained in house as reported earlier20,21.

Molecular docking studies. The IFD default protocol was used, consisting of: (i) Constrained minimization of the receptor with an RMSD cutoff of 0.18 Å. (ii) Initial Glide docking (SP) of each ligand using a soft potentials (0.5 van der Waals radii scaling of non-polar atoms of ligands and receptor using partial charge cutoff of 0.15). (iii) Derived docking poses were then refined using the Prime module of Schrodinger36. Residues within the 5.0 Å of ligand poses were minimized in order to form suitable conformations of poses at the active site of the receptor. (iv) Glide XP re-docking of each protein-ligand complex36.

Declaration of interest

This research was financed by an FP7 EU project (Metoxia, to CTS) and the TAGEM-11/AR-GE/15.2011 project (to SK). The authors report no conflicts of interest. The authors alone are responsible for the content and writing of this article.