Chem. Res. Toxicol. 2007, 20, 1269–1276
1269
Structural Evidence of the Similarity of Sb(OH)3 and As(OH)3 with Glycerol: Implications for Their Uptake Alain Porquet†,‡ and Montserrat Filella*,†,§ Schema, 92 rue Principale, L-6990 Rameldange, Luxembourg, Alpine Institute of EnVironmental Dynamics, L’Entropierre, 108 rue du Puy, F-38660 La Terrasse, France, and Department of Inorganic, Analytical and Applied Chemistry, UniVersity of GeneVa, 30 Quai Ernest-Ansermet, CH-1211 GeneVa 4, Switzerland ReceiVed April 7, 2007
Recent experimental results suggest that As(III) and Sb(III) transport in prokaryotes and eukaryotes might be facilitated by aquaglyceroporins. GlpF, the glycerol facilitator in Escherichia coli was the first to be identified as a trivalent metalloid transporter. Quantum calculations have been performed to study the possible existence of common structural properties between As(OH)3 and Sb(OH)3 and glycerol. Because the mechanism of substrate migration is primarily related to the successive formation of hydrogen bonds between the substrate and the hydrophilic part of the channel wall, this study has focused on the structural, thermodynamic, and electrostatic comparison of the main As(III) and Sb(III) compounds present in aqueous solution at physiological pH values, As(OH)3 and Sb(OH)3, with the glycerol conformation closest to the structures of these As- and Sb-containing compounds. This particular glycerol conformation has then been compared to three known experimental conformations of glycerol present in the protein channel. Besides their stoichiometry and electroneutral condition, As(OH)3 and Sb(OH)3 show very strong similarities to both each other and the studied conformation of the glycerol molecule: Namely, they show a similar charge distribution and a slightly smaller volume than glycerol. Their smaller size can be an additional advantage for the transit through the narrowest region of the GlpF channel. However, the metalloid hydroxyl groups lack the flexibility of glycerol, which probably helps this molecule to adapt its conformation to the topology of the GlpF channel. Introduction Arsenic and antimony compounds are widespread in the biosphere. They are introduced in the environment from both natural and anthropogenic sources (1–3). Their biologically and environmentally relevant oxidation states are III and V; the trivalent forms are more toxic than the pentavalent ones. Epidemiological studies provide clear evidence that inorganic arsenic is a human carcinogen affecting the liver, skin, kidney, urinary bladder, and lung (4–6). In India and Bangladesh, millions of individuals dependent on drinking water from arsenic-contaminated wells exhibit symptoms ranging from nodular keratoses on the soles and palms to various forms of cancer (7). Although there is some evidence for the carcinogenicity of certain antimony compounds by inhalation, there are no data to indicate carcinogenicity by the oral route (5, 8). The IARC concluded that there is inadequate evidence for the carcinogenicity of antimony trioxide in humans but sufficient evidence in experimental animals (9). At the same time, arsenic and antimony have a long history of use in medicine (10, 11). Arsenical drugs were first used in the late 19th and early 20th centuries for the treatment of malaria, syphilis, and cholera. Recently, arsenic trioxide has been shown to be an effective treatment for acute promyelocytic leukemia (APL) (12) and, the drug Trisenox has been approved as a chemotherapeutic agent for the treatment of APL. The first choice for treatment of leishmaniasis, a protozoan parasitic * To whom correspondence should be addressed. Tel: +41 22 3796049. Fax: +41 22 3796069. E-mail:
[email protected]. † Schema. ‡ Alpine Institute of Environmental Dynamics. § University of Geneva.
disease that affects 12 million people worldwide, is antimonial drugs (13–17). The complete understanding of the mechanisms of arsenic and antimony toxicity, as well as of their ability to serve as chemotherapeutic agents, requires the identification of the routes of uptake into cells. Because both elements are nonessential ones, it is unlikely that transport systems evolved for their uptake and it is most probable that arsenic and antimony are taken up by transporters for vital elements or biological molecules. A prerequisite to identication of the uptake route is a clear understanding of the speciation in solution of the elements of interest at the relevant pH value. In the case of Sb(III) and As(III), it is now well-established that they are uncharged at neutral pH; the species that are present are As(OH)3 and Sb(OH)3 (18, 19). Although it has long been assumed that antimonite and arsenite enter cells by passive diffusion, it has emerged in these last years that As(III) and Sb(III) transport in prokaryotes and eukaryotes might be facilitated by aquaglyceroporins. GlpF, the glycerol facilitator in Escherichia coli, was the first aquaporin to be identified as a trivalent metalloid transporter, responsible for the uptake of antimonite by Sanders and co-workers (20). These authors showed that E. coli strains with a mutation in the glpF gene coding for an aquaglyceroporin involved in the glycerol channel are more tolerant for Sb(III), suggesting that the GlpF protein was probably involved in antimony uptake. A recent study (21) showing that the disruption of glpF reduces the uptake of 73As(III) and 125Sb(III) clearly demonstrates the role of GlpF in the transport of As(III) inside the cells. Moreover, Fps1p, the yeast homologue of GlpF, has been shown to be the route of uptake of As(III) and Sb(III) in Saccharomyces cereVisiae (22). Inactivation of the uptake
10.1021/tx700110m CCC: $37.00 2007 American Chemical Society Published on Web 08/23/2007
1270 Chem. Res. Toxicol., Vol. 20, No. 9, 2007
system, either by deletion of FPS1 or by increasing the osmolarity of the growth medium, results in enhanced tolerance. Moreover, cells with a constitutively open and unregulated Fps1p channel are hypersensitive to both arsenite and antimonite. The mammalian aquaglyceroporin AQP9 has recently been shown to transport both As(III) and Sb(III) (23, 24). AQP9 expression also modulates the Trisenox sensitivity to leukemic cells (25). Rat AQP9 conducts methylarsonous acid at a higher rate than As(III) (26). Several Leishmania species have an aquaglyceroporin, AQP1, which seems to facilitate Sb(III) uptake (27, 28). A 2.2 Å resolution X-ray structure of GlpF (29) shows homotetrametric architecture, with glycerol and water present inside the channel. Conduction along the channel proceeds randomly by successive formation and breaking of hydrogen bonds between the substrate and the side chains of the amino acids located on the wall of the channel. This process causes small displacements that allow the substrate to cross the channel of the GlpF protein in a given direction. Competition between the various substrates, mainly water molecules but also glycerol and metalloid polyhydroxides, also involves the formation and breaking of hydrogen bonds (30). As pointed out by Rosen and co-workers (20), the structural similarity between the Sb(III) compounds and the substrates usually using this channel needs to be significant for Sb(III) compounds to enter cells by using the GlpF pathway. They suggest that theoretical studies of the structural similarity between the Sb(III) compounds and the substrates are needed. This approach has been followed in this study: Quantum calculations have been performed to study the possible existence of common structural properties between As(III) and Sb(III) compounds and glycerol. Nonspecific interactions between the protein and the substrate have not been studied because the mechanism of substrate migration is primarily related to the successive formation of hydrogen bonds between the substrate and the hydrophilic part of the wall of the protein channel. Instead, the study has focused on the structural, thermodynamic, and electrostatic comparison of As(OH)3 and Sb(OH)3, the main As(III) and Sb(III) compounds present in aqueous solution at physiological pH values, and some conformations of glycerol, the substrate normally transported by GlpF.
Materials and Methods Modeling. Quantum calculations were carried out with the program Gaussian 03 Revision B.01 (31) running on a PC equipped with an Intel processor x86-based PC 3215 MHz. Geometry optimization, single-point energy, and frequency calculations were performed with a functional with three terms including the functionals of local exchange, exchange of Becke, and exchange of Hartree–Fock, combined to the functional of local correlation (VWN) and corrected according to the gradient of Lee, Yang, and Parr, usually known under the acronym B3LYP (32, 33). As antimony is a relatively heavy element (Z ) 51), the speed of the core electrons becomes comparable to the speed of light. Thus, the hypothesis of infinite speed cannot be applied. Corrections to nonrelativistic quantum mechanics are necessary to avoid errors in the expansion of the orbitals. The use of a pseudopotential is an approximation that allows taking into account most of the relativistic effects. Because only valence electrons participate in chemical bonds, it is possible to carry out calculations only for the valence orbitals and to consider a “nonlocal effective potential” for the core electrons. The effective core potential of the Stuttgart–Dresden group (SDD) (34–36) has been used in this study. This “effective core potential” was adjusted such that the “valence energy spectra” of the antimony atom were suitably reproduced. The electronic configuration of antimony is (Kr)4d105s25p3. Only the five valence
Porquet and Filella
Figure 1. (a) Crystallographic structure of the GlpF protein cocrystallized with three molecules of glycerol (28). (b) Detail of one glycerol molecule.
Figure 2. (a) Atoms and dihedron angles used to define the conformations of the glycerol molecule. (b) Definition of surface S. See text for definitions.
electrons (5s25p3) were represented explicitly and associated with a molecular orbital, while the core electrons [(Kr)4d10] were described by a SDD pseudopotential function. SDD pseudopotential functions exist for practically all of the elements of the periodic table. To keep some uniformity in the calculations, a SDD pseudopotential function was also used for As, even if As shows very little relativistic effects. Therefore, as in the case of Sb, As inner electrons [(Ar)3d10] were described by a pseudopotential function and only the five valence electrons (4s24p3) were present in the calculations and allocated to a molecular orbital defined from the atomic orbitals of a database. All electrons were represented explicitly for the O (1s12s22p4), C (1s12s2 p2), and H (1s1) atoms. The database of atomic orbitals used is that of Dunning and Huzinaga of the type double-ζ valence (37). This is the basic set to be used with SDD pseudopotential functions. The polarizable continuum model (PCM) method was used for calculations with a continuum solvent model (38, 39). The PCM method places the solute in a cavity formed by the union of spheres centered on each atom. Because the electrostatic potential of the solute generates an “apparent surface charge” on the surface of the cavity, this method includes an accurate treatment of the electrostatic interaction with the surrounding medium. The solvent is characterized by its dielectric constant, surface tension, size, density, etc. Procedures are included not only for the computation of the electrostatic interaction of the solute with the apparent surface charges but also for the cavitation energy and for dispersion and repulsion contributions to the solvation free energy. All of the initial structures of the substrates were created by computer graphics because the molecules involved are either small organic molecules (i.e., glycerol) or inorganic and mononuclear molecules (i.e., arsenite and antimonite). The views of the interior of the channel (Figure 1a,b) were calculated from the crystallographic structure of the GlpF protein (available on the Brookhaven database, ID 1FX8) by molecular computer graphics by using the VMD program (40). In Figure 1b, a blue color is used to represent the amino acids having either an aliphatic side or aromatic hydrophobic chains (Ala, Leu, Val, Ile, Pro, Phe, Met, and Trp). The remaining amino acids are shown in red. They present a higher hydrophilic character.
Sb(OH)3, As(OH)3, and Glycerol Similarity
Chem. Res. Toxicol., Vol. 20, No. 9, 2007 1271
Definitions and Parameters Used in Structural and Electrostatic Analysis. Glycerol conformations are defined by five dihedron angles: δ1 (O1–C1–C2–O2) and δ2 (O2–C2–C3–O3) for the alkyl backbone and χ1 (H1–O1–C1–C2), χ2 (H2–O2–C2–H), and χ3 (H3–O3–C3–C2) for the definition of the orientations of the O1–H1, O2–H2, and O3–H3 bonds (Figure 2a). The dihedral angles δ1 and δ2 present three characteristic states, gauche+ (g+), trans (t), and gauche– (g–), and the dihedral angles χ1, χ2, and χ3 present four, synclinal (s), gauche+ (g+), trans (t), and gauche– (g–). The distances between the oxygen atoms of the OH groups, dOO(dO1O2, dO1O3, and dO2O3) have been used to calculate the difference in average O–O distances, ∆dOO, defined as:
∆dOO )
1 3
{∑ dOO[M(OH)3] - ∑ dOO(glycerol)}
(1)
∆dOO values have been calculated for M ) As and Sb and for the “retracted” conformation of glycerol (see below). ∆dOO has been used to compare the structures of the different compounds. A second parameter that allows an easy comparison of different structures is the surface area of the triangle formed by the three O1, O2, and O3 oxygen atoms, S (Figure 2b). S can be calculated from the relationship of Heron of Alexandria:
S ) √p (p - dO1O2)(p - dO2O3)(p - dO1O3)
(2)
where p is half the perimeter, defined as p ) 0.5(dO1O2 + dO1O3 + dO2O3).
Results The precise knowledge of the conformation of glycerol inside the GlpF channel is a prerequisite in any study on the possible molecular mimicry of As(OH)3 and Sb(OH)3 with the glycerol molecule. Possibly because of its complexity, the conformation of glycerol has been the object of very few studies. In effect, the glycerol molecule combines the high flexibility of its alkyl backbone, composed of a linear sequence of three carbons with low-energy rotation C–C bonds, with a high functionality due to the presence of three hydroxyl groups. As a result, the molecule can present 576 possible conformations (41). Unfortunately, the resolution of the three structures of glycerol cocrystallized in the GlpF protein obtained from X-ray measurements by Fu and co-workers (29) does not allow the unambiguous definition of the conformation of the glycerol molecules inside the GlpF channel. For this reason, optimization calculations of the geometry of the three cocrystallized glycerol structures were performed. This study is the object of a parallel publication (42). It is not possible to directly compare the energy levels of glycerol with the energy levels of compounds such as As(OH)3 or Sb(OH)3, but it is always possible to compare two conformations of glycerol among themselves: The first step of this study was the search for a glycerol conformation as structurally close as possible to the conformations of As(OH)3 and Sb(OH)3. This glycerol conformation was then compared with the glycerol conformations experimentally found inside the GlpF channel. Glycerol Retracted Conformation. The geometry of As(OH)3 and Sb(OH)3 corresponds to a pyramid with a triangular base where the top is occupied by the metalloid element and the base by three oxygen atoms. In this structure, the OH groups are very close. To obtain a structure for glycerol similar to this type of structure, a first optimization of the geometry was performed under the constraint of fixed coordinates of the three oxygen atoms, fixing the positions of the OH groups in the glycerol molecule on the basis of the optimized structure of Sb(OH)3. The resulting structure was then relaxed
Table 1. Values of the Dihedral Angles Calculated for the Different Glycerol Conformationsa conformation
δ1
δ2
χ1
χ2
χ3
retracted 661 662 663
structure optimized in vacuo –50 78 –49 55 –57 79 55 –57 79 52 56 –173
168 79 79 77
–171 50 50 –51
retracted 661 662 663
structure optimized in protein interior (ε ) 5) –58 74 –49 168 60 –63 71 66 57 –71 72 69 64 75 –96 53
–171 59 –57 –168
in protein interior (ε ) 20) 74 –46 172 –65 69 62 –71 70 65 73 –81 45
–175 61 120 –171
structure retracted 661 662 663
optimized –60 61 58 66
structure optimized in a continuum water model (ε ) 78) retracted –59 74 –47 172 –171 661 61 –65 69 61 61 662 58 –71 69 64 –60 663 66 73 –79 44 –172 a Dihedral angles are expressed in degrees. Values for 661, 662, and 663 glycerol conformations are from 42.
by eliminating the constraint imposed on the positions of the oxygen atoms. In the structure thus obtained, named retracted, the relative disposition of the oxygen atoms is nearly unchanged, mainly due to the formation of intramolecular hydrogen bonds. According to the nomenclature adopted in this study, the dihedral angles of this retracted conformation have the following values (in the order of definition): -50, 78, -59, 152, and -161°. The three oxygen atoms are located on the same side of the plane formed by the alkyl backbone. The OH orientation allows the formation of two hydrogen bridges O2–H2···O1 and O1–H1···O3. This conformation corresponds to the conformation “γγ2” according to the nomenclature of Chelli and co-workers (43). It shows a relatively low electronic energy (+0.67 kcal/ mol), as compared with the most stable conformation in the gaseous phase according to calculations performed at the B3lyp/ 6-311+G(3df,2p) level (42). Normally, (g, g+) conformations of the alkyl backbone, such as the retracted conformation studied here, are not favored in the absence of intramolecular hydrogen bonds (44). The analysis of the potential energy surfaces clearly shows that the conformational region (g, g+) corresponds to a situation where the three electronegative OH groups are situated on the same side of the plane (42). There is then a sufficiently high electrostatic repulsion between the groups to make this region the highest one from an energy point of view. However, the formation of one, or better two, intramolecular hydrogen bonds, favored by the proximity the OH groups, considerably lowers the energy level of (g, g+) conformations. The flexibility of the glycerol molecule allows the hydroxyl groups to orient to form as many hydrogen bonds as possible with the vicinal water molecules. The observed macroscopic effect will be a high miscibility in water. In the absence of proton acceptor groups, as in the vacuum, the retracted conformation allows the hydroxyl groups to interact among themselves. At most, three intramolecular hydrogen bonds can be formed. This conformation corresponds to the most stable conformation in the gas phase. Comparison of Glycerol “Active” and Retracted Conformations. The values of the dihedral angles (Table 1) and of various types of energy (Table 2) have been calculated
1272 Chem. Res. Toxicol., Vol. 20, No. 9, 2007
Porquet and Filella
Table 2. Values of Energy and Thermodynamic Variables Calculated for the Different Glycerol Conformationsa conformation
Escf (a.u.)
retracted 661 662 663
–344.75083600 –344.74810035 –344.74810036 –344.75037651
retracted 661 662 663
–344.77742467 –344.77846891 –344.77663424 –344.77337865
retracted 661 662 663
–344.77754382 –344.77859728 –344.77673873 –344.77368662
retracted 661 662 663
–344.77752133 –344.77860319 –344.77674850 –344.77371136
∆(Escf – Escf(retracted)) (kcal/mol)
∆(H° – H(retracted)°) (kcal/mol)
∆(G° – G(retracted)°) (kcal/mol)
Esolv (kcal/mol)
structure optimized in vacuo +1.72 +1.72 +0.29
+1.51 +1.51 +0.13
+0.94 +0.94 –0.27
structure optimized in protein interior (ε ) 5) –0.65 +0.50 +2.54
–1.08 –0.05 +1.19
–1.61 –0.90 +0.43
–19.73 –17.21 –21.34 –19.25
–1.28 –1.27 +0.20
–20.47 –21.79 –27.22 –24.84
–1.25 –1.15 +0.43
–20.34 –23.05 –28.88 –26.28
structure optimized in protein interior (ε ) 20) –0.66 +0.50 +2.42
–1.68 –0.03 +1.66
structure optimized in a continuum water model (ε ) 78)
a
–0.68 +0.48 +2.39
–1.70 –0.06 +1.63
Values for 661, 662, and 663 glycerol conformations are from 42.
Figure 3. Molecular volumes of As(OH)3, glycerol, and Sb(OH)3 (left to right). Molecular volumes are defined as the volume with a density
