Pentamidine Is Not a Permeant but a Nanomolar Inhibitor of the Trypanosoma brucei Aquaglyceroporin-2

The chemotherapeutic arsenal against human African trypanosomiasis, sleeping sickness, is limited and can cause severe, often fatal, side effects. One of the classic and most widely used drugs is pentamidine, an aromatic diamidine compound introduced in the 1940s. Recently, a genome-wide loss-of-function screen and a subsequently generated trypanosome knockout strain revealed a specific aquaglyceroporin, TbAQP2, to be required for high-affinity uptake of pentamidine. Yet, the underlying mechanism remained unclear. Here, we show that TbAQP2 is not a direct transporter for the di-basic, positively charged pentamidine. Even though one of the two common cation filters of aquaglyceroporins, i.e. the aromatic/arginine selectivity filter, is unconventional in TbAQP2, positively charged compounds are still excluded from passing the channel. We found, instead, that the unique selectivity filter layout renders pentamidine a nanomolar inhibitor of TbAQP2 glycerol permeability. Full, non-covalent inhibition of an aqua(glycero)porin in the nanomolar range has not been achieved before. The remarkable affinity derives from an electrostatic interaction with Asp265 and shielding from water as shown by structure-function evaluation and point mutation of Asp265. Exchange of the preceding Leu264 to arginine abolished pentamidine-binding and parasites expressing this mutant were pentamidine-resistant. Our results indicate that TbAQP2 is a high-affinity receptor for pentamidine. Taken together with localization of TbAQP2 in the flagellar pocket of bloodstream trypanosomes, we propose that pentamidine uptake is by endocytosis.


Introduction
A specific aquaglyceroporin, TbAQP2, is required for high-affinity uptake of pentamidine into human African trypanosomiasis parasites, Trypanosoma brucei [1][2][3]. Additionally, a trypanosome adenosine transporter mediates pentamidine transport, albeit at lower efficiency [2]. Aquaglyceroporins represent a subfamily of the aquaporin water channel proteins and conduct small, uncharged solutes, mainly glycerol and urea [4]. However, at a molecular weight of 340 Da and with two strongly basic, positively charged amidine moieties (pK a 12.1), pentamidine is not a physiological substrate analog and differs from previous examples of drug uptake involving aquaglyceroporins [5,6]. In the treatment of acute promyelocytic leukemia, the human aquaglyceroporin AQP9 [7] serves as an entry site for the drug arsenic trioxide, As 2 O 3 [8], which dissolves into weak arsenous acid, As(OH) 3 (126 Da), resembling the glycerol molecule (92 Da). The analogous antimonous acid, Sb(OH) 3 (173 Da), derived from the antimonial drug pentostam, is the first-line treatment of leishmaniasis and enters Leishmania major parasites via LmAQP1 [9]. The antineoplastic agent hydroxyurea (N-hydroxylated urea, 76 Da) is a permeant of PfAQP, i.e. the single aquaglyceroporin of the malaria parasite Plasmodium falciparum, and of TgAQP from the toxoplasmosis parasite Toxoplasma gondii [10].
Trypanosomes express three aquaglyceroporins, TbAQP1-3, with permeability, among others, for glycerol, urea [11], arsenous and antimonous acid [12]. The physiological roles of TbAQP2-3 are unclear because the phenotype of knockout parasites is inconspicuous [2]. The physicochemical properties of pentamidine seem to be generally incompatible with a direct passage via AQPs, because two conserved filter sites select against positive charge and too large size. The selectivity filter or aromatic/arginine (ar/R) constriction is typically composed of an arginine residue in an aromatic environment (Fig 1A, right) and excludes solutes with diameters larger than 3-4 Å as well as protons [13][14][15]. The central Asn-Pro-Ala (NPA) filter is formed by the NPA-capped ends of two α-helices that emanate a positive electrostatic field and repel cations, such as sodium, potassium, and ammonium, NH 4 + [14,16].
In this study, we asked whether pentamidine is a direct permeant of TbAQP2 and analyzed its interaction with the protein. We found that TbAQP2 is impermeable to even small cations, such as NH 4 + or formamidine, and consistent with this, failed to increase the susceptibility of yeast cells to pentamidine. Strikingly, pentamidine bound avidly to the selectivity filter and fully inhibited TbAQP2 with an IC 50 of 130 nM. This affinity is unprecedented in the on-going search for AQP inhibitors [17] and is derived from a unique interaction between one of the amidine moieties of pentamidine and Asp265 in conjunction with the displacement of water The models are based on the E. coli GlpF structure (PDB# 1FX8). (B) Western blot using membrane protein fractions isolated from yeast expressing TbAQP2 and TbAQP3 that were N-terminally tagged with a hemagglutinin epitope. (C) Phenotypic growth assay of 31019bΔmep1-3 yeast lacking endogenous ammonium transporters on media with NH 4 + as the sole nitrogen source. Growth at pH 6.5 indicates mainly background transmembrane diffusion of uncharged NH 3 ; growth at pH 5.5 is due to transport of charged NH 4 + . LeAMT1;1 is a tomato ammonium transporter, cells without transporter expression (-) served as negative controls. With proline as a nitrogen source growth is even for all cells. (D) Complementary phenotypic growth assay using yeast-toxic methylammonium instead of ammonium yielding opposite effects. (E) Expression of TbAQP2 does not increase the pentamidine-susceptibility of yeast grown on media with glycerol (YPG) as a non-fermentable carbon source. (F) Biophysical light scattering assays with TbAQP2-expressing yeast protoplasts in 300 mM hypertonic solute gradients. The rapid first phase of the traces indicates water efflux, followed by a slower solute influx phase. Red traces indicate permeability for physiological substrates glycerol (left) and urea (right). Traces of control cells without TbAQP2 are black. Structures and traces of 3-amino-propane-1,2-diol (left) and formamidine (right) are labeled dark blue (pH 7.2) or light blue (pH 9.2). The insets show deduced permeability coefficients (P sol ) with S.E.M. from three independent experiments. doi:10.1371/journal.ppat.1005436.g001 molecules from the channel interior. A single TbAQP2 point mutation, Leu264Arg, reversed pentamidine binding and rendered trypanosomes resistant. The localization of TbAQP2 in the flagellar pocket [2], characterized by high rates of membrane turnover, leads us to propose that pentamidine uptake occurs by endocytosis with TbAQP2 acting not as a transporter but a high-affinity pentamidine receptor. Such a mechanism could be exploited as a novel route to shuttle small molecule drugs into trypanosomes in a similar fashion as discussed for macromolecular transferrin protein-conjugated compounds in connection with the trypanosomal transferrin receptor [18,19].

TbAQP2 excludes organic cations
To approach the question whether TbAQP2 is directly permeable for pentamidine, we set up a yeast expression system and initially analyzed permeability for small neutral and positively charged compounds. TbAQP3 carrying a conserved ar/R constriction (Fig 1A) served as a control. TbAQP2 and TbAQP3 were well expressed ( Fig 1B) and functional as shown below.
First, we carried out phenotypic growth assays based on a yeast strain that lacks all three endogenous ammonium transporters (Δmep1-3) [13,20]. This strain does not grow on acidic media with ammonium as the sole nitrogen source (Fig 1C, pH 5.5, second lane). The growth phenotype is pH-dependent due to the chemical equilibrium of NH 3 /NH 4 + (pK a of 9.2) and growth is permitted towards more alkaline pH by increasing transmembrane diffusion of neutral NH 3 . Permeability of transport proteins for charged NH 4 + is indicated by growth at acidic pH, as seen with the tomato ammonium transporter 1;1 (LeAMT1;1) as a positive control ( Fig  1C, top lane). In this assay, neither expression of TbAQP2 nor TbAQP3 enhanced the NH 4 + uptake over that of cells without an ammonium transporter ( Fig 1C). A complementary outcome of the transport assay can be obtained by using yeast-cytotoxic methylammonium, H 3 C-NH 3 + [13]. Here, yeast growth will cease when the neutral methylamine form, H 3 C-NH 2 , diffuses across the cell membrane ( Fig 1D, pH 6.5) or when the protonated, charged form is transported (Fig 1D, pH 5.5, top lane). This assay is particularly sensitive due to exposure of the cells to the toxic compound for several days [13]. Again, TbAQP2 and TbAQP3 appeared impermeable for the positively charged compound form and survived the treatment ( Fig 1D). Pentamidine itself is cytotoxic to yeast when the cells are grown on a non-fermentable carbon source, such as glycerol [21]. Again, expression of TbAQP2 did not increase the susceptibility in comparison to non-expressing and TbAQP3 expressing cells ( Fig 1E). Next, we employed a biophysical light-scattering assay for TbAQP2 permeability [22]. We expressed TbAQP2 in a yeast strain that lacks the endogenous aquaglyceroporin (Δfps1), prepared protoplasts, and challenged them by a hypertonic, test solute-containing buffer. In an initial rapid phase, the protoplasts will release water following the outward osmotic gradient and shrink as indicated by an increase in light scattering. In the second slower phase, volume will be regained if the test solute can pass TbAQP2 and enter the protoplasts. Pentamidine is not testable in this setup because its low solubility prohibits the establishment of a sufficiently concentrated gradient. We used instead two pairs of test solutes, glycerol/3-aminopropane-1,2-diol and urea/formamidine, of which glycerol and urea are established permeants of TbAQP2 [11], whereas the respective partners are bioisosteric nitrogen-derivatives, which are positively charged at neutral pH ( Fig 1F). 3-Aminopropane-1,2-diol has a pK a of 9.2; hence, at a buffer pH of 9.2, which yeast will tolerate for the assay time, 50% of the compound will be deprotonated and neutral. Formamidine is too strong a base (pK a 13) to be amenable to deprotonation in aqueous solution. In the assay, TbAQP2 showed permeability for glycerol and urea ( Fig 1F, red), whereas the positively charged compounds did not pass ( Fig 1F, dark blue) as seen by traces equal to control protoplasts without TbAQP2 (Fig 1F, black). The lack of permeability for the charged compounds is not due to molecule size, because formamidine is even smaller than urea. Indeed, 3-aminopropane-1,2-diol, when partially neutralized by a pH shift to 9.2, passed TbAQP2 at two thirds the rate of glycerol ( Fig 1F, light blue). We conclude that organic cations fail to traverse the TbAQP2 channel.

Pentamidine is a nanomolar inhibitor of TbAQP2
Formally, formamidine represents one of the amidine moieties of pentamidine. We reasoned that if formamidine (44 Da) is excluded from passing TbAQP2, then the much larger pentamidine would be excluded as well. The lack of an increase in susceptibility of TbAQP2 expressing yeast to pentamidine further argues against pentamidine transport via TbAQP2. Consequently, we conducted another set of light scattering assays to determine how pentamidine interacts with TbAQP2.
We switched from hypertonic assay conditions to an isotonic glycerol replacement of 300 mM saccharose from the buffer to omit osmotic water permeability effects and to obtain monophasic solute-uptake traces, which allow for a more accurate quantification especially at low rates [23]. The red traces in Fig 2A show unhindered glycerol influx via TbAQP2 and TbAQP3. Addition of up to 500 μM pentamidine 10 min prior to the assay left the glycerol permeability of TbAQP3 unaffected, whereas TbAQP2 was fully blocked already at 50 μM ( Fig  2A). The inhibition of TbAQP2 by pentamidine was dose-dependent and substantial even in the nanomolar concentration range. We translated the relative glycerol permeability rates into a dose-response curve yielding an IC 50 of 130 nM (Fig 2B). Pentamidine was effective immediately after addition to the protoplasts indicating that the binding site is readily accessible from the extracellular space.
Pentamidine binds to Asp265 of TbAQP2 and displaces water from the channel Since high-affinity inhibition of other AQPs is only achieved by covalent cysteine-modification by metalorganic compounds, e.g. mercurials or gold-containing auphen (IC 50 0.8 μM) [17,24], we set out to investigate the binding mode of pentamidine by determining structure-function relationships and by point mutation of TbAQP2.
We synthesized asymmetric variants of pentamidine by replacing one of the positively charged amidine moieties by nitrile, and by shortening the molecule via 4-pentoxy-, 4-propoxy-to 4-methoxy-benzamidine ( Fig 2B). The nitrile compound was as effective as pentamidine in blocking TbAQP2 glycerol permeability (IC 50 180 nM) showing that one positive site is sufficient for full inhibition, most likely by electrostatical interaction with an acidic residue of TbAQP2. Removal of one benzamidine group (4-pentoxy-benzamidine) led to a shift in IC 50 to 87 μM (Fig 2B). Further shortening of the molecule to 4-propoxy-benzamidine resulted in an IC 50 of 2.8 mM and a maximal inhibition of 60%. The methoxy compound was incapable of inhibiting TbAQP2. Since electrostatic interactions exhibit highest affinity when efficiently shielded from water, our structure-function evaluation suggests that the amidine interaction site is located 7-8 Å, i.e. the length of the pentoxy chain, away from the solvent into the channel. This distance coincides with the position of the ar/R selectivity filter region.
TbAQP2 lacks the conserved arginine in the ar/R filter possibly providing access of the acidic Asp265 to the channel surface ( Fig 1A). If the Asp265 sidechain carboxylate (pK a 3.7 in water) interacts with pentamidine, the affinity should be titratable by pH. Hence, we determined inhibition of TbAQP2 by pentamidine at pH 3.5 and 2.5 and observed 3 times (450 nM) and 8 times (1.1 μM) reduced affinity, respectively (Fig 2C, left). To test whether the pH titration is specific to Asp265, we changed the residue by point mutation to alanine, which equally led to a shift in IC 50 to 4.5 μM (Fig 2C, right). Further, this mutant was pH-independent with regard to pentamidine binding (IC 50 5.8 μM at pH 2.5; Fig 2C, right) confirming the identity of the interaction site. Together, the structure-function evaluation and the mutation data indicate that the exceptional affinity of pentamidine to TbAQP2 is derived from an electrostatic interaction of one amidine moiety with Asp265 and shielding from the solvent by displacement of water molecules from the channel by the pentoxy chain.
Mutation of Leu264 to arginine prevents pentamidine from binding to TbAQP2 and renders parasites resistant A selectivity filter arginine followed by an aspartate at the next sequence position is a signature motif of aquaglyceroporins [25]. At about 4.5 Å distance in the protein, both charged residues are positioned for an electrostatic interaction, see for instance TbAQP3 Arg256/Asp257 in Fig  1A [ 26]. Besides TbAQP2, there are other examples of ar/R selectivity filters in which the arginine is replaced by an aliphatic residue, e.g. the T. gondii TgAQP or the AQP subfamily of plant tonoplast instrinsic proteins TIP2;1 [10]. However, in all such cases, replacement of the arginine goes together with an exchange of the neighboring aspartate for a neutral residue [27]. TbAQP2 is, thus, unique in having lost the arginine but having kept Asp265. We concluded that a point mutation of TbAQP2 Leu264 to arginine should re-establish the internal salt bridge with Asp265 and prevent pentamidine from binding.
Expression in yeast of TbAQP2 Leu264Arg yielded a functional water and glycerol channel with about half the permeability of the wild-type protein ( Fig 3A). Importantly, pentamidine failed to inhibit TbAQP2 Leu264Arg (Fig 3A) and this further supported the binding mode suggested above. We next investigated the findings from the heterologous yeast expression system in T. brucei parasites by expressing GFP-tagged TbAQP2 Leu264Arg ( GFP AQP2 L264R ) or wild-type TbAQP2 ( GFP AQP2 WT ) in a T. brucei aqp2-null background (Fig 3B). Despite the mutation, TbAQP2 Leu264Arg localized to the flagellar pocket as reported previously for wildtype TbAQP2 [2] (Fig 3C). While aqp2-null T. brucei expressing wild-type TbAQP2 were highly sensitive to pentamidine (EC 50 = 0.65 nM), T. brucei expressing mutant TbAQP2 remained resistant to pentamidine (EC 50 = 103 nM; Fig 3D). Thus, TbAQP2 Leu264Arg fails to bind pentamidine and fails to sensitize trypanosomes to pentamidine.

Discussion
Maintenance of cation homeostasis is crucial for cell functionality and viability [14]. Accordingly, cation channels and transporters are highly regulated and/or selective, whereas non-cation conducting membrane proteins contain exclusion filters. The typical AQP channel layout exhibits two cation filters, which select against both types of cation conductance: a) hydrogenbonded proton-wires are disrupted in the ar/R constriction preventing the Grotthuss mechanism, b) diffusion of non-proton cations is blocked by the electrostatic field generated in the NPA region [13][14][15][16]. Further, the amphipathic channel lining of the AQPs provide coordinating carbonyl oxygens only on one side, and therefore cannot sufficiently compensate for the dehydration penalty derived from the removal of the water shell from the ion [28]. Finally, the AQP protein structure is rigid and does not allow for conformational changes, because even small bends would obstruct the long (20 Å) and narrow (3-4 Å) water and solute channel [26], forbidding transporter-type cation transduction following the alternate access mechanism [29]. As a consequence, naturally occurring AQPs with cation conductance along the water/ glycerol channel have not been identified [28]; there is evidence, though, for cation permeability via the central pore of the human AQP1 tetramer [30].
TbAQP2 lacks an ar/R filter, which could potentially allow for proton leakage. The NPA region, however, is intact and perfectly repels cations, such as NH 4 + and charged formamidine, as shown in this study. Thus, pentamidine containing two formamidine substructures cannot pass TbAQP2. Instead, pentamidine acts as a high-affinity binder and inhibitor of TbAQP2 with one amidine moiety interacting with Asp265 by taking the place of the lacking arginine sidechain. The electrostatic interaction is increased by several orders of magnitude due to the exclusion of water from the channel (see Fig 4A for a proposed binding mode). Inhibition of TbAQP2 water and solute permeability is not the mode of pentamidine action because parasites lacking the aquaglyceroporin thrive well [2]. The obtained high affinity of 130 nM despite a steep glycerol gradient of 300 mM further argues against pentamidine diffusion through TbAQP2, because even high-affinity, alternate access-type transporters typically exhibit weaker, micromolar binding to enable release of the cargo after traversing the membrane [29]. AQP channel-like transporters, such as the malaria parasite's lactate transporter, PfFNT, or the bacterial formate transporter, FocA, show K m values in the high millimolar range, i.e. 87 mM and 96 mM, respectively [31,32]. In fact, nanomolar binding constants are typically associated with high-affinity receptors.
If TbAQP2 acts as a receptor rather than a transporter for pentamidine, how could drug uptake be achieved? TbAQP2 is specifically localized to the flagellar pocket of bloodstream T. brucei parasites, which is characterized by high membrane turnover activity [33]. This activity is indispensable for the parasite and required, among others, for immune evasion and nutrient uptake. From our data we conclude that a likely scenario for TbAQP2-dependent, high-affinity uptake of pentamidine into trypanosomes is via continued, non-selective endocytosis of pentamidine tethered to TbAQP2 as a receptor (Fig 4B). The hypothesis of pentamidine binding to TbAQP2 and subsequent uptake by endocytosis is strikingly supported by observations from the original biochemical characterization of pentamidine uptake [34]. Key findings were: a) a single amidine moiety suffices to initiate high-affinity transport, which is in agreement with the binding mode of pentamidine to TbAQP2 (Fig 4A), b) transport is mainly unidirectional and highly concentrative, and c) the process requires metabolic energy; the latter two points are fully compatible with endocytosis. Moreover, it is intriguing that in the procyclic, insect form of the parasites TbAQP2 is distributed over the entire surface of the plasma membrane [2] and these cells were found to transport slower and accumulate less pentamidine [2,35] further pointing at the importance of processes in the flagellar pocket for efficient pentamidine uptake. Intracellular release of pentamidine from TbAQP2 may be triggered downstream by acidification of the endo-lysosomal system that weakens the interaction with Asp265 (Fig 4B).
It is becoming increasingly clear that highly active processes in the flagellar pocket of African trypanosomes can be exploited for therapeutic purposes [33,36,37]. The classic pentamidine therapy appears to have been doing just that for decades.
Future routes in drug design could foresee fusion scaffolds that contain a minimal pentamidine anchor, e.g. alkoxy-benzamidine, for TbAQP2 binding linked to an anti-trypanosomal cargo to be delivered into the parasite. There is a good chance for achieving sufficient specificity because of the unique TbAQP2 ar/R filter composition. This approach would be similar to the idea of generating drug fusions with the transferrin protein and transport via endocytosis of the transferrin receptor in the flagellar pocket [19]. A recent study targeted pentamidineloaded chitosan nanoparticles via a single domain nanobody to the parasite surface, inducing endocytotic uptake [38]. Addressing TbAQP2, however, would have the advantage that small molecules could be drafted, which, contrary to pentamidine, require only one positive moiety for Asp265 binding. Using small molecules increases the chance that the compounds will cross the blood brain barrier, which is a necessity to treat the severe cerebral stage of trypanosomiasis.

Plasmid constructs and western blots
TbAQP2, TbAQP2-L264R, TbAQP-D265A, and TbAQP3 were cloned via BamH I and HinD III into the pRS426Met25 vector [39]. The constructs carry an N-terminal hemagglutinin (HA) epitope-tag for detection by Western blot using a monoclonal mouse anti-HA antibody (1: 5000, Roche) and a horseradish peroxidase secondary anti-mouse antibody (1: 2000, Jackson Immuno Research) for ECL detection (Amersham). For T. brucei expression, the pRPa GFPx construct [1] was modified to express GFP-tagged AQP2 under the control of a Tet-regulated RRNA promoter. Site directed mutagenesis was carried out using the QuikChange Multi Site-Directed Mutagenesis Kit (Agilent technologies). Point mutations were confirmed by sequencing. Primer sequences are available upon request. T. brucei whole cell lysates were stored in the presence of a protease inhibitor cocktail (Roche) and were not boiled. GFP was detected using polyclonal rabbit α-GFP (Europa; 1:2000).
Biophysical protoplast light scattering assays For the assay, protoplast suspensions were diluted to an OD 600 of 2. All measurements were done using a stopped flow apparatus (SFM-300, BioLogic, Claix, France) with a dead time around 10 ms, total flow rate of 14 ml/s, total volume of 202 μl, at 20°C. For hypertonic solute permeability assays, protoplasts were rapidly mixed with the same volume of buffer supplemented with 600 mM solute (glycerol, urea, 3-amino-1,2-diol, or formamidine), generating a 300 mM hypertonic solute gradient. For isotonic glycerol permeability measurements, protoplasts were mixed with the same volume of buffer in which 0.6 M saccharose were replaced by glycerol, generating an isotonic 300 mM glycerol gradient. Pentamidine and derivatives were added 10 min prior to the assay. Protoplast volume changes were monitored by measuring the intensity of 90°light scattering at 546 nm. In each experiment, 6 to 9 single curves were averaged, exponentially (rapid rates) or linearly fitted (slow rates) and the resulting time constants were related to solute permeability. The solute permeability coefficient P sol was calculated using P sol = |dI/dt| Á (V 0 Á C out )/(S 0 Á C diff ) [16], with dI/dt being the slope of the intensity curve, V 0 the initial mean protoplast volume (65.45 μm 3 ), S 0 the initial mean protoplast surface area (78.54 μm 2 ), C out is the total external solute concentration (1.5 M), and C diff is the chemical solute gradient (0.3 M).

Pentamidine derivatives
Pentamidine and 4-methoxy-benzamidine were from Sigma. 4-Pentoxy-and 4-propoxy-benzamidine were generated by Williamson ether synthesis from 4-hydroxy-benzonitrile and 1-bromopentane or 1-bromopropane, respectively, and subsequent conversion of the nitrile to amidine using CH 3 Al(Cl)NH 2 [42]. The asymmetric pentamidine derivative carrying one amidine and one nitrile moiety was generated in a first step by Williamson ether synthesis of 4-hydroxy-benzonitrile and 1,5-dibromopentane to obtain the di-nitrile compound. The second step was done according to Pinner [43] in molecular sieve-dried ethanol using HCl gas as an acidic catalyst and stopped before full conversion of the nitriles to imino esters. The semiconverted product was isolated by chromatography and the single amidine group was obtained by reaction with gaseous ammonia in ethanol. Nitrile-pentamidine hydrochloride: 1

Microscopy
T. brucei immunofluorescence microscopy was carried out according to standard protocols. Briefly, cells were settled on slides and mounted in Vectashield (Vector Laboratories) containing the DNA counterstain, 4,6-diamidino-2-phenylindole (DAPI). Images were captured using a Nikon Eclipse E600 epifluorescence microscope in conjunction with a Coolsnap FX (Photometrics) charge-coupled device (CCD) camera and processed in Metamorph 5.0 (Photometrics).