elifeeLifeeLife2050-084XeLife Sciences Publications, Ltd5641610.7554/eLife.56416Research ArticleBiochemistry and Chemical BiologyMicrobiology and Infectious DiseasePositively selected modifications in the pore of TbAQP2 allow pentamidine to enter Trypanosoma bruceiAlghamdiAli H1MundayJane C1CampagnaroGustavo Danielhttp://orcid.org/0000-0001-6542-04851GurvicDominik2SvenssonFredrik3OkparaChinyere E4KumarArvind5QuintanaJuan6Martin AbrilMaria Esther1MilićPatrik1WatsonLaura1PaapeDaniel1SettimoLuca1DimitriouAnna1WielinskaJoanna1SmartGraeme1AndersonLaura F1WoodleyChristopher M4KellySiu Pui Ying1IbrahimHasan MS1HulpiaFabianhttp://orcid.org/0000-0002-7470-34847Al-SalabiMohammed I1EzeAnthonius Ahttp://orcid.org/0000-0002-4821-16891SprengerTeresa8TekaIbrahim A1GudinSimon1WeyandSimone8FieldMark69DardonvilleChristophe10TidwellRichard R11CarringtonMarkhttps://orcid.org/0000-0002-6435-72668O'NeillPaul4BoykinDavid W5ZachariaeUlrich2De KoningHarry Phttps://orcid.org/0000-0002-9963-1827Harry.De-Koning@glasgow.ac.uk1Institute of Infection, Immunity and Inflammation, University of GlasgowGlasgowUnited KingdomComputational Biology Centre for Translational and Interdisciplinary Research, University of DundeeDundeeUnited KingdomIOTA Pharmaceuticals Ltd, St Johns Innovation CentreCambridgeUnited KingdomDepartment of Chemistry, University of LiverpoolLiverpoolUnited KingdomChemistry Department, Georgia State UniversityAtlantaUnited StatesSchool of Life Sciences, University of DundeeDundeeUnited KingdomLaboratory for Medicinal Chemistry, University of GhentGhentBelgiumDepartment of Biochemistry, University of CambridgeCambridgeUnited KingdomInstitute of Parasitology, Biology Centre, Czech Academy of SciencesCeske BudejoviceCzech RepublicInstituto de Química Médica - CSICMadridSpainDepartment of Pathology and Lab Medicine, University of North Carolina at Chapel HillChapel HillUnited StatesClaytonChristineReviewing EditorDKFZ-ZMBH AllianceGermanySoldati-FavreDominiqueSenior EditorUniversity of GenevaSwitzerland1108202020209e564162602202006082020© 2020, Alghamdi et al2020Alghamdi et alhttp://creativecommons.org/licenses/by/4.0/This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Mutations in the Trypanosoma brucei aquaporin AQP2 are associated with resistance to pentamidine and melarsoprol. We show that TbAQP2 but not TbAQP3 was positively selected for increased pore size from a common ancestor aquaporin. We demonstrate that TbAQP2’s unique architecture permits pentamidine permeation through its central pore and show how specific mutations in highly conserved motifs affect drug permeation. Introduction of key TbAQP2 amino acids into TbAQP3 renders the latter permeable to pentamidine. Molecular dynamics demonstrates that permeation by dicationic pentamidine is energetically favourable in TbAQP2, driven by the membrane potential, although aquaporins are normally strictly impermeable for ionic species. We also identify the structural determinants that make pentamidine a permeant although most other diamidine drugs are excluded. Our results have wide-ranging implications for optimising antitrypanosomal drugs and averting cross-resistance. Moreover, these new insights in aquaporin permeation may allow the pharmacological exploitation of other members of this ubiquitous gene family.

eLife digest

African sleeping sickness is a potentially deadly illness caused by the parasite Trypanosoma brucei. The disease is treatable, but many of the current treatments are old and are becoming increasingly ineffective. For instance, resistance is growing against pentamidine, a drug used in the early stages in the disease, as well as against melarsoprol, which is deployed when the infection has progressed to the brain. Usually, cases resistant to pentamidine are also resistant to melarsoprol, but it is still unclear why, as the drugs are chemically unrelated.

Studies have shown that changes in a water channel called aquaglyceroporin 2 (TbAQP2) contribute to drug resistance in African sleeping sickness; this suggests that it plays a role in allowing drugs to kill the parasite. This molecular ‘drain pipe’ extends through the surface of T. brucei, and should allow only water and a molecule called glycerol in and out of the cell. In particular, the channel should be too narrow to allow pentamidine or melarsoprol to pass through. One possibility is that, in T. brucei, the TbAQP2 channel is abnormally wide compared to other members of its family. Alternatively, pentamidine and melarsoprol may only bind to TbAQP2, and then ‘hitch a ride’ when the protein is taken into the parasite as part of the natural cycle of surface protein replacement.

Alghamdi et al. aimed to tease out these hypotheses. Computer models of the structure of the protein were paired with engineered changes in the key areas of the channel to show that, in T. brucei, TbAQP2 provides a much broader gateway into the cell than observed for similar proteins. In addition, genetic analysis showed that this version of TbAQP2 has been actively selected for during the evolution process of T. brucei. This suggests that the parasite somehow benefits from this wider aquaglyceroporin variant.

This is a new resistance mechanism, and it is possible that aquaglyceroporins are also larger than expected in other infectious microbes. The work by Alghamdi et al. therefore provides insight into how other germs may become resistant to drugs.

Trypanosoma bruceiaquaporindrug transportpentamidinedrug resistancemelarsoprolResearch organismOtherhttp://dx.doi.org/10.13039/501100000265Medical Research CouncilG0701258De KoningHarry Phttp://dx.doi.org/10.13039/100000002National Institutes of HealthGM111749BoykinDavid Whttp://dx.doi.org/10.13039/501100000265Medical Research CouncilMR/R015791/1De KoningHarry Phttp://dx.doi.org/10.13039/501100000708Scottish Universities Physics AllianceZachariaeUlrichAlbaha UniversityAlghamdiAliScience Without Borders, Brazil206385/2014-5CampagnaroGustavo Danielhttp://dx.doi.org/10.13039/100004440Wellcome204697/Z/16/ZFieldMarkhttp://dx.doi.org/10.13039/501100000265Medical Research CouncilSprengerTeresaCambridge trustSprengerTeresahttp://dx.doi.org/10.13039/100004440Wellcome TrustWeyandSimonehttp://dx.doi.org/10.13039/501100000288Royal SocietyWeyandSimoneThe funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.Author impact statementStudy of TbAQP2 adaptations and substrate interactions shows how this aquaglyceroporin enables cellular entry of large antimicrobial agents in Trypanosoma brucei.Introduction

The Trypanosoma brucei-group species are protozoan parasites that cause severe and fatal infections in humans (sleeping sickness) and animals (nagana, surra, dourine) (Giordani et al., 2016; Büscher et al., 2017). The treatment is dependent on the sub-species of trypanosome, on the host, and on the stage of the disease (Giordani et al., 2016; P. De Koning, 2020). Many anti-protozoal drugs are inherently cytotoxic but derive their selectivity from preferential uptake by the pathogen rather than the host cell (Munday et al., 2015a; P. De Koning, 2020). Conversely, loss of the specific drug transporters is a main cause for drug resistance (Barrett et al., 2011; Baker et al., 2013; Munday et al., 2015a; P. De Koning, 2020). This is the case for almost all clinically used trypanocides, including diamidines such as pentamidine and diminazene (Carter et al., 1995; de Koning, 2001a; de Koning et al., 2004; Bridges et al., 2007), melaminophenyl arsenicals such as melarsoprol and cymelarsan for cerebral stage human and animal trypanosomiasis, respectively (Carter and Fairlamb, 1993; Bridges et al., 2007), and the fluorinated amino acid analogue eflornithine for human cerebral trypanosomiasis (Vincent et al., 2010). The study of transporters is thus important for anti-protozoal drug discovery programmes as well as for the study of drug resistance (Lüscher et al., 2007; Munday et al., 2015a).

In Trypanosoma brucei, the phenomenon of melarsoprol-pentamidine cross-resistance (MPXR) was first described shortly after their introduction (Rollo and Williamson, 1951), and was linked to reduced uptake rather than shared intracellular target(s) (Frommel and Balber, 1987). The first transporter to be implicated in MPXR was the aminopurine transporter TbAT1/P2 (Carter and Fairlamb, 1993; Mäser et al., 1999; Munday et al., 2015b) but two additional transport entities, named High Affinity Pentamidine Transporter (HAPT1) and Low Affinity Pentamidine Transporter (LAPT1), have been described (de Koning, 2001a; de Koning and Jarvis, 2001; Bridges et al., 2007). HAPT1 was identified as Aquaglyceroporin 2 (TbAQP2) via an RNAi library screen, and found to be the main determinant of MPXR (Baker et al., 2012, Baker et al., 2013; Munday et al., 2014). The apparent permissibility for high molecular weight substrates by TbAQP2 was attributed to the highly unusual selectivity filter of TbAQP2, which lacks the canonical aromatic/arginine (ar/R) and full NPA/NPA motifs, resulting in a much wider pore (Baker et al., 2012; Munday et al., 2014; Munday et al., 2015a). Importantly, the introduction of TbAQP2 into Leishmania promastigotes greatly sensitised these cells to pentamidine and melarsen oxide (Munday et al., 2014). Moreover, in several MPXR laboratory strains of T. brucei the AQP2 gene was either deleted or chimeric after cross-over with the adjacent TbAQP3 gene, which, unlike AQP2, contains the full, classical ar/R and NPA/NPA selectivity filter motifs and is unable to transport either pentamidine or melaminophenyl arsenicals (Munday et al., 2014). Similar chimeric genes and deletions were subsequently isolated from sleeping sickness patients unresponsive to melarsoprol treatment (Graf et al., 2013; Pyana Pati et al., 2014) and failed to confer pentamidine sensitivity when expressed in a tbaqp2-tbaqp3 null T. brucei cell line whereas wild-type TbAQP2 did (Munday et al., 2014; Graf et al., 2015).

The model of drug uptake through a uniquely permissive aquaglyceroporin (Munday et al., 2015a) was challenged by a study arguing that instead of traversing the TbAQP2 pore, pentamidine merely binds to an aspartate residue (Asp265) near the extracellular end of the pore, above the selectivity filter, followed by endocytosis (Song et al., 2016). This alternative, ‘porin-receptor’ hypothesis deserves careful consideration given that (i) it is an exceptional assertion that drug-like molecules with molecular weights grossly exceeding those of the natural substrates, can be taken up by an aquaglyceroporin and (ii) the fact that bloodstream form trypanosomes do have, in fact, a remarkably high rate of endocytosis (Field and Carrington, 2009; Zoltner et al., 2016). The question is also important because aquaporins are found in almost all cell types and the mechanism by which they convey therapeutic agents and/or toxins into cells is of high pharmacological and toxicological interest. While TbAQP2 is the first aquaporin described with the potential to transport drug-like molecules, this ability might not be unique, and the mechanism by which the transport occurs should be carefully investigated.

We therefore conducted a mutational analysis was undertaken, swapping TbAQP2 and TbAQP3 selectivity filter residues and altering pore width at its cytoplasmic end. This was complemented with a thorough structure-activity relationship study of the interactions between pentamidine and TbAQP2, using numerous chemical analogues for which inhibition constants were determined and interaction energy calculated. The pentamidine-TbAQP2 interactions were further modelled by running a molecular dynamics simulation on a protein-ligand complex, and in addition, we investigated a potential correlation between the T. brucei endocytosis rate and the rate of pentamidine uptake. Our results provide strong evidence for pentamidine permeating directly through the central pore of TbAQP2. Having identified the essential characteristics that allow the transport of large, flexible molecules through TbAQP2, this should now allow the evaluation of aquaporins in other species for similar adaptations.

ResultsInvestigation of the structural determinants of AQP2 for pentamidine transportPositive selection for pore size

In T. brucei, the AQP2 and AQP3 genes are arranged as a tandem pair on chromosome 10 and have 74% amino acid identity. Whereas TbAQP2 clearly mediates pentamidine uptake, TbAQP3 does not (Baker et al., 2012; Munday et al., 2014), nor do various chimeric AQP2/3 rearrangements that give rise to pentamidine resistance (Munday et al., 2014; Graf et al., 2015). To investigate the origin of the AQP2 gene, a phylogenetic analysis of AQPs in African trypanosomes was performed. The number of aquaporin genes varies: there is a single aquaporin in T. vivax and T. congolense, two in T. suis and three in T. brucei and its derivatives and the most probable tree (Figure 1) is consistent with the evolutionary history of the four species (Hutchinson and Gibson, 2015), and indicates AQP1 as the ancestral AQP present in all trypanosome species. A duplication occurred in the common ancestor of T. suis and T. brucei after divergence from T. congolense and a further duplication, to form AQP2 and AQP3, in the ancestor of T. brucei after divergence from T. suis. Multiple alignment (Figure 1—figure supplement 1) shows that the classical NPA/NPA and ar/R AQP selectivity filter elements are present in all AQPs except T. brucei AQP2. The divergence of T. brucei AQP2 and 3 was investigated by calculating the non-synonymous/synonymous codon ratio (dN/dS) for different AQPs (Figure 1—figure supplement 2). For T. brucei aligned with T. suis AQP1, dN/dS is 0.21 and for AQP3, dN/dS is 0.30 indicating purifying selection. However, comparing T. brucei AQP2 with T. brucei AQP3, dN/dS is 2.0 indicating strong selection pressure for divergence on AQP2 towards an aquaporin with increased pore size. In order to verify any role of amino acids along the TbAQP2 pore in facilitating pentamidine sensitivity and/or uptake, we performed a mutational analysis.

Phylogenetic tree of <italic>Trypanosoma</italic> aquaporins.

The tree is a Neighbour-joining tree produced in Clustal Omega with the lengths of the horizontals proportional to the differences.

Sequence alignment and individual sequences of the <italic>T. congolense</italic>, <italic>T. b. brucei</italic> and <italic>T. suis</italic>.

The T. brucei and T. congolense sequences were obtained from tritrypDB, T. suis sequences (Kelly S, Gibson W and Carrington M. The genome of Trypanosoma suis. In preparation). The alignment was produced with Clustal Omega. The yellow highlighting indicates the N-terminus of the sequences used to determine non-synonymous v synonymous ratios. dN/dS. T brucei AQP1 v T. suis AQP1 0.21. T brucei AQP3 v T. suis AQP3 0.30. T brucei AQP2 v T. brucei AQP3 2.00.

The ratio of non-synonymous v synonymous (dN/dS) codon changes calculated for selected comparisons between <italic>T.</italic>

brucei and T. suis AQPs. The ratios were calculated using a region of high confidence alignments from ~amino acid 60 (highlighted in Figure 1—figure supplement 1) to the C-terminus.

Figure 1 shows that replacing the AQP2 NSA/NPS motif in the selectivity filter of the pore with the consensus NPA/NPA motif of other aquaporins still allows pentamidine to permeate, although at substantially reduced rate, with an EC50 at >50 fold lower than the aqp2/aqp3 null control, while cymelarsan is almost completely blocked (EC50 ~60% of control). Similar observations are made for the L258Y mutant (Figure 1B,C). In contrast, the single mutant L264R did not allow either cymelarsan or pentamidine permeation (Figure 1A, E) and the double mutant I110W/L264R actually enabled a level of cymelarsan (but not pentamidine) permeation (Figure 1F). These findings can be rationalised by the differing structural features, interaction patterns, and charge of the two compounds. The leucine to arginine substitution introduces a further positive charge into the central pore of the aquaporin, which repels the dicationic pentamidine molecule and is therefore likely to decrease its affinity to this variant, whereas the neutral arsenical compound is much less affected by the addition of positive charge into the pore.

Introduction of AQP3 residues into the AQP2 selectivity filter

One highly conserved motif of aquaporins, believed to be essential for permeant selectivity, is NPA/NPA which is present in TbAQP3 but not in TbAQP2, where, uniquely, it is NS131A/NPS263 instead. We therefore constructed a TbAQP2 variant with the classical NPA/NPA motif (TbAQP2S131P/S263A) and expressed it in the aqp2/aqp3 null cell line (Baker et al., 2012; Munday et al., 2014). In this cell line, uptake of 30 nM [3H]-pentamidine was reduced to 4.40 ± 0.71% (n = 4) of the rate in the control line expressing TbAQP2WT (p<0.05, Student’s unpaired t-test), as well as significantly different from the rate measured in parallel in the tbaqp2/tbaqp3 null cells (p<0.01) (Figure 2A). The remaining pentamidine uptake was sufficient to strongly sensitise the TbAQP2S131P/S263A cells to pentamidine in a standard protocol of 48 hr incubation with the drug followed by a further 24 hr in the presence of the resazurin indicator dye (p<0.0001 vs tbaqp2/tbaqp3 null) but the EC50 was still significantly higher than the TbAQP2WT control (p<0.05) (Figure 2B). A similar effect was observed for the melaminophenyl arsenical drug cymelarsan, but there was no change in sensitivity to diminazene or the control drug phenylarsine oxide (PAO), which is believed to diffuse directly across the membrane (Fairlamb et al., 1992; Figure 2B).

The selectivity filter differences between TbAQP2 and TbAQP3 are largely responsible for their differences in pentamidine sensitivity and transport rates.

(A) Transport of 30 nM [3H]-pentamidine by tbaqp2/aqp3 null cells expressing TbAQP2-WT or one of the TbAQP2 mutants as indicated (blue bars). The corresponding brown bars are pentamidine transport in the control tbaqp2/aqp3 null cells assessed in parallel in each experiment. Transport was determined in the presence of 1 mM adenosine to block the TbAT1/P2 transporter. Bars represent the average and SEM of at least three independent experiments, each performed in triplicate. Blue stars: statistical significance comparison, by two-tailed unpaired Student’s tests, between the cells expressing TbAQP2WT and mutants; red stars: statistical comparison between the AQP2-expressing cells and control cells; NS, not significant. (B–F) EC50 values indicated test drugs, expressed as a percentage of the resistant control (tbaqp2/tbaqp3 null), against cell lines either expressing the indicated TbAQP2 mutant or TbAQP2WT (sensitive control). Red stars and green stars: comparison with tbaqp2/aqp3 null or TbAQP2WT-expressing cells, respectively, which were always assessed in parallel in each experiment. (G) Transport of 30 nM [3H]-pentamidine by tbaqp2/aqp3 null cells expressing TbAQP3 or an AQP3 mutant as indicated. (H) EC50 values of the indicated drugs against tbaqp2/aqp3 null cells expressing either TbAQP3 or a mutant thereof, expressed as percentage of tbaqp2/aqp3 null. All data for these graphs are contained in Figure 2—source data 1. All experiments are the average and SEM of at least three independent experiments. *, p<0.05; **, p<0.01; ***, p<0.001, ****, p<0.0001 by unpaired Student’s t-test, two-tailed.

Individual and average EC<sub>50</sub> values and transport rates for <xref ref-type="fig" rid="fig2">Figure 2A-H</xref>.

The mutant L258Y, which has the AQP3 Tyr-250 half of the highly conserved aromatic/arginine (ar/R) motif, responsible for pore restriction and proton exclusion (Wu et al., 2009), introduced into the TbAQP2 pore, yielded a drug transport phenotype similar to TbAQP2S131P/S263A. The [3H]-pentamidine transport rate was reduced to 6.6 ± 1.4% of TbAQP2WT (p<0.05) but remained above the rate in the tbaqp2/tbaqp3 null cells (p<0.01) (Figure 2A). Pentamidine and cymelarsan EC50 values were also significantly different from both the TbAQP2WT and the tbaqp2/tbaqp3 null controls (Figure 2C).

The ar/R motif is part of the larger selectivity filter, usually WGYR, present in both TbAQP1 and TbAQP3 but uniquely consisting of I110VL258L264 in TbAQP2 (Baker et al., 2013), all non-polar, open chained residues. Cell lines expressing mutations AQP2I110W and AQP2L264R, either alone or in combination, displayed pentamidine transport rates, and pentamidine and cymelarsan EC50 values that were not significantly different from the tbaqp2/tbaqp3 null controls but highly significantly different from the TbAQP2WT drug-sensitive controls, showing that their capacity for pentamidine and cymelarsan uptake had been reduced to zero (Figure 2A,D–F).

We conclude that the unique TbAQP2 replacement of the NPA/NPA motif and all of the WGYR selectivity filter mutations are necessary for the observed pentamidine and melaminophenyl arsenical sensitivity observed in cells expressing wild-type TbAQP2.

Introduction of TbAQP2 selectivity filter residues into the AQP3 pore enables pentamidine transport

An interesting question was whether the introduction of (some of) the critical TbAQP2 residues in TbAQP3 might give the latter the capacity to take up pentamidine. We therefore constructed TbAQP3W102I/R256L and TbAQP3W102I/R256L/Y250L and tested whether tbaqp2/tbaqp3 null cells transfected with these mutant aquaporins were able to take up 25 nM [3H]-pentamidine in the presence of 1 mM adenosine (which blocks uptake via TbAT1/P2). Pentamidine uptake in the tested cell lines was very low compared to the same cells expressing TbAQP2WT (Figure 2G). However, by measuring [3H]-pentamidine uptake over 30 min it was possible to reliably and reproducibly measure radiolabel accumulation in each cell line. This showed that while uptake in TbAQP3W102I/R256L only trended slightly upwards (p>0.05), the mutant AQP3 with all three AQP2 WGYR residues (W102I, R256L and Y250L) accumulated significantly more [3H]-pentamidine than the tbaqp2/tbaqp3 null cells (p<0.01) or the null cells expressing TbAQP3WT (p=0.011). This is further corroborated by comparing the pentamidine sensitivity profile of these cell lines: only TbAQP3W102I/R256L/Y250L conveyed significant sensitisation to tbaqp2/tbaqp3 null cells (p<0.0001; Figure 2H). Thus, TbAQP3 is converted into a pentamidine transporter by the insertion of the AQP2 WGYR residues, although this does not convey as high a rate of pentamidine uptake as TbAQP2. Similar experiments did not show any sensitisation to cymelarsan upon expression of TbAQP3 or mutants TbAQP3W102I/R256L and TbAQP3W102I/R256L/Y250L, although the latter mutant actually showed a 50% higher EC50 concentration (p<0.001); EC50 values were 114 ± 5%, 105 ± 8% and 159 ± 8% of the tbaqp2/tbaqp3 control, respectively (n ≥ 7).

Mutations of amino acids modelled to potentially bind pentamidine or melarsoprol dramatically reduce pentamidine transport

Our previous attempts at modelling the binding of pentamidine and melarsoprol into the pore of TbAQP2 tentatively identified several residues that could be involved in this process (Munday et al., 2015a), from which we selected two residues, Ile190 and Trp192, at the extracellular end of the channel (position shown in Figure 7), to swap with the corresponding residues of TbAQP3, creating TbAQP2I190T and TbAQP2W192G. Both residues were predicted to interact with the substrate(s) via main-chain carbonyl oxygen atoms, but the side chains could nonetheless affect the interactions.

TbAQP2I190T displayed dramatically reduced [3H]-pentamidine uptake, at 2.7 ± 0.7% (p<0.01, n = 4) of the TbAQP2WT control, although significantly higher than the rate of the tbaqp2/tbaqp3 null negative control (p<0.05) (Figure 3A). The reduced rate was the result of a reduced Vmax of the high affinity [3H]-pentamidine uptake, rather than a change in Km; the LAPT1 Vmax and Km were unchanged in cells expressing TbAQP2I190T or TbAQP2WT (Figure 3—figure supplement 1). TbAQP2I190T still conferred some increased pentamidine sensitivity in the standard resazurin test (p<0.0001), although highly significantly less sensitizing than TbAQP2WT (p<0.001); an intermediate sensitivity was also observed for cymelarsan (Figure 3B). Substitution W192G also produced intermediate sensitivity to both drugs (Figure 3C) but the double substitution TbAQP2I190T/W192G displayed no significant pentamidine uptake above tbaqp2/tbaqp3 null (Figure 3A) and did not sensitise to pentamidine or cymelarsan (Figure 3D).

Mutational analysis of TbAQP2 residues I190 and W192.

(A) Transport of 30 nM [3H]-pentamidine by tbaqp2/tbaqp3 null cells or TbAQP2 variants expressed therein. Transport was expressed as a percentage of the rate of the AQP2WT control, performed in parallel. Blue stars are comparison with TbAQP2WT, red stars, comparison with the tbaqp2/tbaqp3 null control. NS, not significant. (B) EC50 values for the indicated drugs against tbaqp2/tbaqp3 null cells, and against TbAQP2WT and TbAQP2I190T expressed therein; values were expressed as % of the tbaqp2/tbaqp3 null (resistant) control. Red stars, comparison with the resistant control; green stars, comparison with the internal sensitive control (TbAQP2WT). The assays for all three strains and all three drugs were done simultaneously on at least three different occasions. (C) As B but for TbAQP2W192G. (D) As B but for TbAQP2I190T/W192G. All data for these graphs are contained in Figure 3—source data 1. *, p<0.05; **, p<0.01; ***, p<0.001, ****, p<0.0001 by unpaired Student’s t-test.

Individual and average EC<sub>50</sub> values and transport rates for <xref ref-type="fig" rid="fig3">Figure 3A-D</xref> and <xref ref-type="fig" rid="fig3s1">Figure 3—figure supplement 1</xref>.

Pentamidine transport analysis for TbAQP2<sup>I190T</sup> and TbAQP2WT.

(A) Transport of 30 nM [3H]-pentamidine by tbaqp2/tbaqp3 null cells expressing TbAQP2I190T, in the presence of unlabelled pentamidine at the indicated concentrations. Incubation time was 15 min, required to ensure sufficient radiolabel for accurate quantification, and uptake was linear and through zero over this period. The inhibition data were plotted to a double sigmoidal curve (Prism 7.0) with the bottom value fixed at 0. The high affinity component displayed an average an IC50 of 30.9 ± 12.2 nM (n = 3) and the lower affinity segment could be converted to a Michaelis-Menten plot for determination of Km and Vmax (inset), yielding an average Km of 59.9 ± 9.1 µM (n = 3), consistent with the Low Affinity Pentamidine Transporter (LAPT1; Bridges et al., 2007). The plot shown is one representative experiment in triplicate of three independent experiments. (B) Like (A) but with tbaqp2/tbaqp3 null cells expressing TbAQP2WT. Incubation time was 20 s (linear phase). The high affinity phase had statistically identical EC50 (41 ± 17 nM; p>0.05) as TbAQP2I190T. The inset shows a zoom-in on the low-affinity part of the curve (LAPT1 contribution), with the dotted line representing a theoretical sigmoid plot for one inhibitor, with the upper limit fixed at the value obtained for 10 μM pentamidine. The low affinity component was also statistically identical in the two strains (TbAQP2-WT Km = 82.7 ± 17.5 µM (n = 3; p>0.05)). Note that the amount of [3H]-pentamidine taken up by the low affinity component is highly similar for the mutant (A) and control (B) cell lines, at approximately 0.0005 pmol(107 cells)−1 s−1. Both frames show one representative experiment of three repeats, each performed in triplicate. Error bars are SEM, when not shown, fall within the symbol. All data for these graphs are contained in Figure 3—source data 1.

The effect of large amino acids at the cytoplasmic end of the pore

To test whether restrictions at the cytoplasmic end of TbAQP2 would impact on pentamidine transport, we selected three leucine residues and exchanged each with tryptophan, creating L84W, L118W and L218W (positions indicated in Figure 7). Expressing each of the L-to-W mutants in tbaqp2/tbaqp3 null cells revealed significantly reduced pentamidine sensitivity compared to the same cells expressing TbAQP2WT (Figure 4A), while also exhibiting dramatically reduced rates of [3H]-pentamidine transport (Figure 4B). This effect was additive, with TbAQP2L84W/L118W not significantly sensitising for pentamidine and displaying no detectable increase in [3H]-pentamidine transport relative to tbaqp2/tbaqp3 null cells (Figure 4A,B). None of these L-to-W mutants sensitised the cells to cymelarsan, diminazene or PAO, except that the L218W mutant sensitised slightly to diminazene (~2 fold, p<0.05; compared to 20-fold for pentamidine, p<0.0001) (Figure 4—figure supplement 1). When the same leucine residues were replaced with methionine instead of tryptophan, variants L84M and L218M were not or barely different from TbAQP2WT with respect to pentamidine sensitisation (Figure 4A) or transport (but highly significantly different from their respective tryptophan variants). For position 118 the Met replacement had similar effects as the Trp variant had, albeit with a significantly higher rate of pentamidine transport (1.88 ± 0.20 (n = 6) versus 9.38 ± 0.63% (n = 3) of TbAQP2WT, p<0.001; Figure 4A,B). The L84M and L218M mutants also sensitised to cymelarsan (p<0.01) and, surprisingly, the L218 M mutants also sensitised somewhat to diminazene (~2 fold, p<0.05) (Figure 4—figure supplement 1).

Analysis of TbAQP2 variants with a leucine-to-tryptophan or leucine-to-methionine substitution near the cytoplasmic end of the pore.

(A) Pentamidine EC50 values (nM) for mutant and WT TbAQP2 expressed in tbaqp2/tbaqp3 cells (aqp2-3 null). The mutants are either a Trp (dark blue bars) or Met (orange bars) substitution at the indicated positions. The resistant control (aqp2-3 null) and sensitive control (AQP2WT) for the separate datasets (Trp or Met) are indicated as hatched bars in the same colours. (B) As (A) but showing transport of 30 nM [3H]-pentamidine by the same cell lines, expressed as percentage of the transport rate in the TbAQP2 control cells. (C) Pentamidine EC50 values for the same mutants as in (A) but expressed in the tbaqp1-2-3 null cells, performed in parallel with the determination of EC50 values for SHAM, shown in (D). As all cell lines were done simultaneously, the resistant and sensitive strain control values are identical for the Trp and Met mutants in this series. All bars represent the average and SEM of at least three independent replicates. *, p<0.05; **, p<0.01; ***, p<0.001, ****, p<0.0001 by unpaired Student’s t-test; ns, not significant; nd, not determined. All data for these graphs are contained in Figure 4—source data 1.

Individual and average EC<sub>50</sub> values and transport rates for <xref ref-type="fig" rid="fig4">Figure 4A-E</xref> and <xref ref-type="fig" rid="fig4s1">Figure 4—figure supplement 1</xref>.

EC<sub>50</sub> values for Cymelarsan, diminazene aceturate and phenylarsine oxide (PAO) against the <italic>tbaqp2-tbaqp3</italic> null cell line.

AQP2-WT and various mutant versions thereof (indicated) were expressed in this cell line. EC50 values were determined using the alamar blue (resazurin) assay. Bars represent the average and SEM for at least three determinations. nd, not done. All data for these graphs are contained in Figure 4—source data 1.

Correlation of the EC<sub>50</sub> value with the rate of pentamidine transport for all 19 cell lines expressing a wild-type or mutant TbAQP2 in the aqp2/3 null <italic>T. b. brucei line</italic>.

All EC50 values are expressed as percentage of the resistant control, aqp2/3 null transfected with an empty vector (no TbAQP2). 2T1 is the parental cell line of the aqp2/3 null. All values are the average of at least three independent determinations; the sensitive and resistant control cell lines were included in each independent experiment and the percentages taken are from the internal control rather than from the grand average over all experiments. All data for these graphs are contained in Figure 4—source data 1.

These results strongly argue that the introduction of large amino acids at the cytosolic end significantly blocks the transport of pentamidine, whereas the change to Leu→Met mutants were more permissive for pentamidine, but not cymelarsan. In order to check whether these variants were still functional aquaglyceroporins, we used the observation of Jeacock et al., 2017 that T. brucei cells lacking all three AQPs are sensitised to the Trypanosome Alternative Oxidase inhibitor SHAM, as a result of cellular glycerol accumulation. By this measure, all of the position 84, 118 and 218 Trp and Met mutants were able to transport glycerol, as each displayed SHAM EC50 values significantly different from the tbaqp1-2-3 null cells (Figure 4C,D); several variants displayed an intermediate SHAM EC50, being also significantly different from TbAQP2WT, indicating some attenuation of glycerol efflux capacity for those mutants. Indeed, uptake of [3H]-glycerol closely mirrored the SHAM observations (Figure 4E).

Overall correlation between [<sup>3</sup>H]-pentamidine transport rate and pentamidine EC<sub>50</sub>

The results presented in Figures 2–4 consistently show that even TbAQP2 mutants that display a large reduction in [3H]-pentamidine uptake rate results can show intermediate pentamidine sensitivity phenotypes (EC50s), due to the nature of the standard drug sensitivity test employed, which involves a 48 hr incubation with the drug prior to a further 24 hr incubation with resazurin: even a much-reduced transport rate will be sufficient to accumulate significant amounts of intracellular pentamidine over 3 days. A plot of [3H]-pentamidine transport rates versus pentamidine EC50, using the data for all 19 TbAQP2 and TbAQP3 mutants for which the transport rates were determined, shows that relatively small changes in EC50 occur, even with up to approximately 95% reduction in transport rates; at >95% reduction large EC50 increases become apparent (Figure 4—figure supplement 2).

Partially blocking endocytosis does not alter the rate of pentamidine transport and pentamidine does not trigger AQP2 endocytosis in bloodstream form <italic>T. brucei</italic>

The knockdown of the CRK12 kinase in T. brucei causes a highly reproducible defect in endocytosis that affects an estimated one third of cells 12 hr after RNAi induction and is ultimately lethal (Monnerat et al., 2013). We utilized this system to investigate whether a link between endocytosis and pentamidine transport exists. At 12 hr of CRK12 RNAi induction with tetracycline, CRK12 mRNA levels were reduced by 42% (p<0.001) relative to uninduced controls as determined by qRT-PCR (Figure 5A). Samples from the culture taken at this time point showed an increased abundance of cells with swelling characteristic of endocytosis defects, although this was hard to quantify as a minority of cells were affected, and to various degrees, as the 12 hr time point was deliberately taken at as early a point as possible, so as not to affect cell viability (Figure 5—figure supplement 1) or cause excessive cellular pathology. We thus performed parallel uptake experiments with [3H]-pentamidine and [3H]-suramin, with suramin acting as positive control as it is known to enter T. brucei bloodstream forms through endocytosis after binding to surface protein ISG75 (Zoltner et al., 2016; Zoltner et al., 2020). After 12 hr of CRK12 RNAi induction, pentamidine uptake was not significantly less than in the T. brucei 2T1 parental cells, whereas uptake of [3H]-suramin was (p=0.019, n = 5; Figure 5B,C).

Disabling endocytosis does not reduce uptake of pentamidine.

(A) qRT-PCR of CRK12, normalised to housekeeping gene GPI-8 (n = 3). (B) Transport of 0.025 µM [3H]-Pentamidine measured in control (non-induced) and CRK12 cell after exactly 12 hr of tetracycline induction; incubation time with label was 30 s. Bar is average and SEM of 5 independent determinations, each performed in triplicate. NS, not significant by unpaired Student’s t-test. (C) As frame B but uptake of 0.25 µM [3H]-suramin over 15 min; average and SEM of 5 independent determinations, each in quadruplicate. **, p=0.0027 by Student’s unpaired, two-tailed t-test. All data for these graphs are contained in Figure 5—source data 1.

mRNA abundance for <xref ref-type="fig" rid="fig5">Figure 5A</xref> plus transport rates for <xref ref-type="fig" rid="fig5">Figure 5B, C</xref> and <xref ref-type="fig" rid="fig5s1">Figure 5—figure supplement 1</xref>.

Growth Curve of CRK12 RNAi cells in full HMI-9 medium at 37°C/5% CO<sub>2</sub>, in the presence or absence of 1 μg/ml tetracycline (tet).

Cell counts were performed with a haemocytometer and the average of duplicate determinations is shown. All data for these graphs are contained in Figure 5—source data 1.

As this approach is necessarily limited to a partial inhibition of endocytosis in BSF T. brucei, it was also investigated whether pentamidine induces the internalisation and turnover of TbAQP2, as could be expected if the protein acts to internalise substrate by receptor-mediated endocytosis. Cells pre-treated with the protein synthesis inhibitor cycloheximide (100 µg/ml) were incubated in the presence or absence of 25 nM pentamidine and the abundance of 3 × HA tagged TbAQP2 was followed over a period of 6 hr by western blot (Figure 6—figure supplement 1). Quantification of the bands showed identical turnover rates with or without pentamidine present in the medium (Figure 6).

Quantification of western blots of <sup>3×HA</sup>TbAQP2.

Cells were induced for expression of 3×HATbAQP2, pretreated with cyclohexidine and subsequently with 25 nM pentamidine (5× EC50). Western blots (Figure 6—figure supplement 1) were performed using anti-HA antiserum in order to quantify the relative amount of TbAQP2 in the cells. The two datasets were not significantly different by Kolmogorov-Smirnov test (p=0.77) and data points at each time point were also not significantly different by Student’s t-test (p>0.05). All data for these graphs are contained in Figure 6—source data 1.

Densitometry readings with Western blotting of TbAQP2 turnover, normalised to 0 h.

Western blots for a <sup>3xHA</sup>AQP2 turnover assay in untreated <italic>T. brucei</italic> 2T1 cells, or in the presence of pentamidine 25 nM.

Two independent experiments are shown. HA, haemagglutinin; tub, β-tubulin; CHX, cycloheximide; α-HA, anti-HA serum. All data for these graphs are contained in Figure 6—source data 1.

The protonmotive force drives AQP2-mediated pentamidine uptake in bloodstream forms of <italic>T. brucei</italic>

It has been reported that knock-down of the HA1–three plasma membrane proton pumps of T. brucei (which are essential for maintaining the plasma membrane potential), confers pentamidine resistance (Alsford et al., 2012; Baker et al., 2013). Interestingly, this locus only conferred resistance to (dicationic) pentamidine, not to the (neutral) melaminophenyl arsenicals, unlike knockdown of the TbAQP2/TbAQP3 locus (Alsford et al., 2012). We have previously reported that the HAPT-mediated pentamidine uptake in T. brucei procyclics correlates strongly with the proton-motive force (PMF) (de Koning, 2001a). However, it is not clear whether this dependency indicates that pentamidine uptake is mediated by a proton symporter, as known for many T. brucei nutrient transporters (de Koning and Jarvis, 1997a; de Koning and Jarvis, 1997b; de Koning and Jarvis, 1998; de Koning et al., 1998), or reflects the energetics of uptake of cationic pentamidine being driven by the strong inside-negative membrane potential Vm. The absence of an effect of HA1–three knockdown on sensitivity to the neutral melaminophenyl arsenicals strongly argues against a mechanism of proton symport for HAPT1/AQP2 but a (partial) dependency of HAPT1/AQP2-mediated uptake of dicationic pentamidine on PMF or Vm would be expected if the substrate traverses the channel, as opposed to binding a single Asp residue on the extracellular side of the protein, as suggested in the endocytosis model (Song et al., 2016). Here we show that the same ionophores that inhibit HAPT1-mediated pentamidine transport in procyclic cells, and inhibit hypoxanthine uptake in both bloodstream form (BSF) (de Koning and Jarvis, 1997b) and procyclic (de Koning and Jarvis, 1997a) T. brucei, also dose-dependently inhibit [3H]-pentamidine uptake in BSF (Figure 7A). This confirms that pentamidine needs the membrane potential for rapid uptake, as predicted by the dependence on the HA1–three proton pumps. Using [3H]-suramin as an endocytosed substrate (Zoltner et al., 2016), we found that 20 µM CCCP also inhibits endocytosis in T. brucei, by 32.6% (p=0.029; pre-incubation 3 min, plus suramin accumulation over 10 min) (Figure 7B). While that means that the ionophore experiments do not perfectly discriminate between endocytosis and trans-channel transport for di-cationic pentamidine, they do for neutral melaminophenyl arsenicals: the non-dependence of these neutral TbAQP2 substrates on the proton gradient (Alsford et al., 2012) indicates that, unlike suramin, they are not endocytosed.

High affinity pentamidine uptake in <italic>T. b. brucei</italic> is sensitive to ionophores.

(A) Uptake of 25 nM [3H]-pentamidine in s427WT bloodstream forms was measured in the presence of 1 mM adenosine to block the P2 transporter, and in the further presence of various ionophores at the indicated concentrations in µM. Incubation with radiolabel was 5 min after a 3 min pre-incubation with ionophore. Accumulation of radiolabel was expressed as a percentage of the control, being a parallel incubation in the absence of any ionophore. Bars represent the average of 3–5 independent determinations (each performed in quadruplicate) and SEM. (B) Uptake of 0.25 µM [3H]-suramin by T. b. brucei s427WT cells over 10 min. Cells were incubated in parallel, with or without the presence of 20 µM CCCP (plus 3 min pre-incubation). Saturation of the suramin-receptor interaction was demonstrated by including 100 µM unlabelled suramin (blue bars). Bars represent average and SEM or three independent experiments, each performed in quadruplicate. (C) Correlation plot of pentamidine transport rate versus protonmotive force (PMF), r2 = 0.93, p<0.05 by F-test. Concentrations in µM are indicated in the frame. CCCP is shown in red and not included in the regression analysis. Each data point is the average of 4 or more independent repeats performed in quadruplicate. The values for PMF were taken from de Koning and Jarvis, 1997b. (D) Uptake of 0.25 µM [3H]-glycerol by aqp1/aqp2/aqp3 null cells expressing TbAQP2-WT. Dose response with CCCP and pentamidine (PMD), using an incubation time of 1 min. The graph shown was performed in triplicate and representative of three independent repeats. (E) As D but using 0.025 µM [3H]-pentamidine and 30 s incubations. Representative graph in triplicate from three independent repeats. *, p<0.05; **, p<0.01; ***, p<0.001 by Student’s unpaired t-test. All data for these graphs are contained in Figure 7—source data 1.

Individual and average transport rates and PMF values for <xref ref-type="fig" rid="fig7">Figure 7A-E</xref>.

Although there is a good correlation between the proton-motive force and TbAQP2-mediated pentamidine transport (Figure 7C), the effect of CCCP was stronger than expected, and stronger than previously observed for [3H]-hypoxanthine uptake in T. brucei bloodstream forms (de Koning and Jarvis, 1997b) and we thus investigated whether CCCP might have a direct effect on TbAQP2. Indeed, CCCP inhibited uptake of (neutral) [3H]-glycerol in tbaqp1-2-3 null cells expressing TbAQP2-WT, with an IC50 of 20.7 ± 2.6 µM (n = 3) and inhibited [3H]-pentamidine uptake in the same cells with a similar IC50 (Figure 7D,E), showing CCCP to inhibit TbAQP2 directly, irrespective of effects on the membrane potential. Figure 7D also shows that pentamidine, used as a control, inhibits [3H]-glycerol uptake with an EC50 value (Mean of 27.5 nM, n = 2) similar to the EC50 of pentamidine inhibiting uptake of [3H]-pentamidine.

Molecular dynamics modelling of pentamidine interactions with TbAQP2

To further investigate pentamidine binding and permeation in TbAQP2, we used the coordinates of the TbAQP2-pentamidine complex that was modelled in our previous study (Munday et al., 2015a). The stability of the protein model was first confirmed by unbiased atomistic molecular dynamics simulations (Figure 8—figure supplement 1). We then conducted force-probe simulations, in which a moving spring potential was used to enforce unbinding of pentamidine from its docked binding position and subsequently reconstructed the free-energy profile of pentamidine association-dissociation along the pore axis by employing Jarzynski’s equality (Park et al., 2003).

Figure 8A shows that the docked position of pentamidine correctly identified its minimum free-energy binding site inside the TbAQP2 pore. Pentamidine adopts an extended state inside the TbAQP2 pore, adapting its molecular shape to the narrow permeation channel; pentamidine binding poses display inter-amidine lengths in the range 16.5–17 Å. Importantly, our steered simulations reveal that pentamidine can exit the channel in either direction, and that unbinding on the route towards the cytoplasm occurs on a free-energy surface roughly symmetric to that towards the extracellular side. Apart from overcoming the strongly attractive binding interaction in the centre, there are no major further free-energy barriers in either direction. The computed free-energy profile of pentamidine binding to the TbAQP2 structural model slightly overestimates its experimentally recorded binding affinity. However, the pentamidine conformation binding the narrow pore may not be the lowest-energy internal conformation of the small molecule, a factor that may be underrepresented in the profile as simulations were started from the protein-bound state. A further source of uncertainty stems from the protein model, which is expected to be somewhat less accurate than a crystal structure.

Pentamidine binding in TbAQP2 and free-energy profile of permeation (Left).

(A) Docked conformation of pentamidine (blue) bound to the TbAQP2 (wheat). The protein and the ligand were modelled as described.4 The protein pore is shown in grey mesh, and the mutated positions described in the text are in magenta. (Right) Free-energy profile G(L) (solid blue line) along the pore axis of TbAQP2 (L). The membrane voltage of T. b. brucei gives rise to a voltage drop across the membrane (gray dotted line), which alters the free-energy profile (dashed blue line includes Vm effect) and reduces the free-energy of pentamidine exit into the intracellular bulk by ~22 kJ/mol as compared to the extracellular side (black arrow). (B) Close-up views comparing the bound positions of pentamidine (left) and melarsoprol (right) and showing the mutated sites and major interactions with the AQP2 pore lining.

Backbone RMSD of the protein inserted into a lipid bilayer showing convergence to ~3 Å in a simulation of 100 ns length.

Due to the dicationic character of pentamidine, the free-energy profile of the molecule within TbAQP2 strongly depends on the membrane voltage. The voltage drop of −125 mV across the cytoplasmic membrane of T. b. brucei (de Koning and Jarvis, 1997b), with a negative potential inside the cell, results in an overall inward attraction of ~22 kJ/mol (Figure 8A, arrow), that is exit from TbAQP2 into the cytoplasm is substantially more favourable for pentamidine than towards the extracellular side. Taken together, the free-energy profile under membrane voltage explains the strong coupling between pentamidine uptake and Vm observed in the experiments. The high affinity of the binding interaction leads to slow off-rates and a relatively low Vmax (0.0044 ± 0.0004 pmol(107 cells)−1 s−1) (de Koning, 2001a).

We further investigated the bound positions of pentamidine and melarsoprol in AQP2 by docking, in order to rationalise the differential behaviour of pentamidine and melaminophenyl arsenicals observed in the studied AQP2 mutants. The binding modes in the central pore of AQP2 obtained by docking are shown in Figure 8B. They reveal that both drugs are likely to bind to the same general region within the central pore in spite of their different sizes and charge states.

Conversely, the shorter arsenical agent is more affected by introducing the NPA/NPA motif since its terminal polar function intensely interacts with the NSA/NPS motif, whereas the major interactions of pentamidine are seen outside this motif. In the case of the L258Y mutant, the difference can be attributed to the extended flexible linker region in the centre of pentamidine, which is likely to enable it to bend around the added bulky side chain in the mutant, while cymelarsan lacks this level of flexibility in its centre.

Finally, the single and double mutations I190T and W192G have broadly similar effects on pentamidine and cymelarsan permeation across AQP2 (Figure 2) as both positions at the entrance to the pore exhibit similar interaction patterns to both pentamidine and the arsenical agent.

SAR of the pentamidine-AQP2 interaction

In order to study substrate binding and selectivity by the T. b. brucei High Affinity Pentamidine Transporter (HAPT1/TbAQP2), competition assays were performed with a series of pentamidine analogues and other potential inhibitors, in the presence of 1 mM unlabelled adenosine to block diamidine uptake by the TbAT1 aminopurine transporter (de Koning, 2001a; Bridges et al., 2007). High specific activity [3H]-pentamidine was used at 30 nM, below the Km value (de Koning, 2001a). Uptake was linear for at least 3 min (de Koning, 2001a) and we utilized 60 s incubations for the determination of inhibition constants (Ki). At 30 nM [3H]-pentamidine there is virtually no uptake through LAPT1 (Bridges et al., 2007) (Km value ~1000 fold higher than HAPT1) (de Koning, 2001a). The full dataset of 71 compounds is presented in Supplementary file 1, featuring Kis spanning five log units.

The linker length and composition is a strong determinant for high affinity binding of pentamidine

We determined the Ki values for analogues with a 2–8 methylene unit linker (Figure 9A, Table 1). Pentamidine analogues featuring 5–7 units displayed submicromolar binding affinities (5 > 6 > 7), while fewer (3-4) or more (8) only conveyed low micromolar binding affinity, equivalent to a decrease in Gibbs free energy of binding (ΔG0) from 10.2 to 13.0 kJ/mol (Table 1). Energy minimalization using Gaussian16 yielded an elongated conformation for pentamidine, with an inter-amidine length of 17.8 Å (Figure 7B). Replacement of the ether oxygens with S or NH, analogues RT-48 and RT-50, respectively (Table 1) resulted in δ(ΔG0) of 10.0 and 12.9 kJ/mol, respectively, indicating that the ether oxygens potentially act as H-bond acceptors: the NH group serves only as an H-bond donor, as its lone pair is conjugated with the aromatic system, and the sulphur mimics an aromatic NH (Beno et al., 2015). The sulfone analogue (RT-49), which introduces a dihedral angle of 180° between the benzamidine and the linker (Brameld et al., 2008), displayed no binding affinity. We propose that a near-planar conformation of the Phe-O-CH2 segment is required for efficient engagement of the binding site. This is supported by examining the binding affinities found for the analogous benzofuramidine series (e.g. RT-14, Figure 9B), which has a conformationally predefined ether-methylene bond orientation. Replacement of the middle methylene unit of pentamidine with an isosteric oxygen (ethylene glycol derivative DB1699, Table 1) results in a less flexible linker and a remarkable drop in binding affinity (δ(ΔG0)=15.3 kJ/mol).

Correlation between linker chain length and affinity to HAPT1.

(A) A series of pentamidine analogues with different methylene linker length was tested for inhibition of TbAQP2/HAPT1-mediated 25 nM [3H]-pentamidine transport (i.e. in the presence of adenosine to block the TbAT1/P2 transporter). The Ki values are listed in Table 1. All Ki values are shown as average and SEM of 3 or more independent experiments, each performed in triplicate. (B) The distance between the amidine carbon atoms in the lowest-energy conformation was calculated using density functional theory as implemented in Spartan’ 16 v2.0.7. Geometry optimisations were performed with the wB97XD functional and the 6–31G* basis set at the ground state in gas phase. Structures and distances shown represent the dication state that is overwhelmingly prevalent in aqueous solution at neutral pH. The numbered red data points correspond to the propamidine - octamidine series in frame A. (C) Repulsion between free electron pairs (double dots), indicated by curved blue lines for RT-52 in the cis-conformation, causing it to exist overwhelmingly in the anti-conformation. (D) Overview of SAR observations on the binding preferences of TbAQP2 for pentamidine and its analogues. All data for these graphs are contained in Figure 9—source data 1.

K<sub>i</sub> values for <xref ref-type="fig" rid="fig9">Figure 9A and B</xref>.

Correlation between the Resistance Factor (RF; EC<sub>50</sub>(aqp2/3 null)/EC<sub>50</sub>(TbAQP2-WT)) and the K<sub>i</sub> value for inhibition of the High Affinity Pentamidine Transporter (HAPT1) encoded by TbAQP2.

The pentamidine value (bold) is the Km determined with radiolabeled pentamidine. The table lists the data points shown in the plot. The line was made by linear regression (Prism 6.0); correlation coefficient r2 is 0.037. F-test: slope is not significantly different from zero (p=0.33). All values are in µM. All data for these graphs are contained in Figure 9—source data 1.

Pentamidine analogues with an aliphatic linker.
CompoundR1R2R3R4XYZKi (µM)δ(ΔG0) PMD (kJ/mol)
EthamidineAmAmHHCHCH-O-(CH2)2-O->100>19.7
PropamidineAmAmHHCHCH-O-(CH2)3-O-6.63 ± 1.4013.0
ButamidineAmAmHHCHCH-O-(CH2)4-O-3.87 ± 1.3811.7
Pentamidine (PMD)AmAmHHCHCH-O-(CH2)5-O-0.036 ± 0.0006--
HexamidineAmAmHHCHCH-O-(CH2)6-O-0.058 ± 0.0111.3
HeptamidineAmAmHHCHCH-O-(CH2)7-O-0.123 ± 0.0103.1
OctamidineAmAmHHCHCH-O-(CH2)8-O-2.16 ± 0.4810.2
RT-48AmAmHHCHCH-S-(CH2)5-S-2.01 ± 0.8610.0
RT-50AmAmHHCHCH-NH-(CH2)5-NH-6.27 ± 1.3012.9
RT-49AmAmHHCHCH-SO2-(CH2)5-SO2->150>20.7
DB1699AmAmHHCHCH-O-(CH2)2-O-(CH2)2-O-16.6 ± 2.115.3
RT-36AmNH2HHCHCH-O-(CH2)5-O-0.43 ± 0.076.2
CHI/1/72/1AmCH3HHCHCH-O-(CH2)5-O-3.1 ± 0.710.7
CHI/1/69/1AmHHHCHCH-O-(CH2)4-O-2.3 ± 0.510.4
RT-38AmCAHHCHCH-O-(CH2)3-O-NI, 100>19.7
meta-PMDHHAmHCHCH-O-(CH2)5-O-2890 ± 105028.1
RT-32ImImHHCHCH-O-(CH2)5-O-0.40 ± 0.086.0
RT-30MeAmMeAmHHCHCH-O-(CH2)5-O-0.30 ± 0.075.3
StilbamidineAmAmHHCHCH-CH = CH-54.8 ± 3.218.3
FR39G1G1HHCHCH-(CH2)2-41.7 ± 15.217.6
CRMI8G2G2HHCHCH-(CH2)2-52.8 ± 12.718.1
RT-43AmAmHClCHCH-O-(CH2)5-O-0.51 ± 0.156.6
Iodo-PMDAmAmHICHCH-O-(CH2)5-O-2.15 ± 0.048.4
RT-46AmAmH-C(O)NH2CHCH-O-(CH2)5-O->100>19.7
RT-52AmAmHHCHN-O-(CH2)5-O-8.84 ± 0.8813.7
RT-53AmAmHHNCH-O-(CH2)5-O-NI, 250>22

Am, amidine; MeAm, Methyl-amidine; Im, imidazole; CA, carboxylic acid; G1, 2-aminoimidazoline; G2, 1-methoxy-2-aminoimidazoline. PMD, pentamidine, NI, no inhibition at the indicated concentration in µM. Ki is the inhibition constant for [3H]-pentamidine transport by TbAQP2/HAPT1. δ(ΔG0) PMD is the difference in Gibbs Free Energy of interaction of the substrate with TbAQP2 with the same value for pentamidine (PMD). All Ki values are the average and SEM of at least 3–4 independent experiments.

Two amidine groups are required for high affinity binding

Matched-molecular pair analysis of non-symmetric analogues identified that both amidines contribute to high affinity binding (compare pairs pentamidine/RT-36 and pentamidine/CHI/1/72/1; Table 1). Removal of an H-bond donor (as in CHI/1/72/1) leads to a loss in ΔG0 > 10 kJ/mol. The aniline derivative RT-36 can still act as an H-bond donor, albeit with significantly reduced basicity (and thus H-bond acceptor propensity), and accordingly displayed intermediate affinity (δ(ΔG0)=6.2 kJ/mol). Interestingly, the removal of one amidine (compare butamidine and CHI/1/69/1) did not produce a significant effect on the binding affinity (Ki = 3.87 µM and Ki = 2.33 µM, respectively), indicating that the low affinity of butamidine (compare 36 nM for pentamidine) is due to an inability to attain a productive interaction with the second amidine. Capping of the amidine group, resulting in imidazoline analogue RT-32, or methylation (analogue RT-30) reduced binding to HAPT1, probably due to increased steric crowding at the interaction site, impairing H-bonding. Reducing pentamidine to just 4-hydroxybenzamidine removed essentially all affinity (Ki = 2.9 mM; δ(ΔG0)=28.1 kJ/mol), and the replacement of one amidine with a carboxylic group (compare propamidine, RT-38) was highly deleterious for engagement with the binding site. Finally, the orientation of the amidine group is crucial as shown by a meta to para change on the phenyl ring (meta-pentamidine, Table 1). We conclude that for high affinity both amidine groups must be able to interact unimpeded with the transporter, and in the linear (para) conformation.

Fully conjugated linking units

Stilbamidine and the short-linker analogues FR39 and CRMI8 (Ríos Martínez et al., 2015) displayed low binding affinity (Table 1). Diminazene also displayed similar low affinity (Ki = 63 µM), and [3H]-diminazene uptake can only just be detected in procyclic T. b. brucei, that is in the absence of the TbAT1/P2 transporter (Teka et al., 2011), potentially indicating a minimal uptake via HAPT1. Stilbamidine and diminazene feature a similar inter-amidine distance, much shorter than pentamidine (12.35 and 12.25 Å, respectively). DB75 (furamidine) likewise displayed low affinity (Table 2) and is only internalised by TbAT1/P2 (Ward et al., 2011). The 2,5-furan linker imposes a fixed, inflexible angle of 136° on the benzamidine moieties and the phenyl rings will adopt a planar orientation with respect to the furan plane. This appears to allow only one benzamidine end to interact with the transporter, as (unlike the flexible linker of aliphatic diamidines, vide supra) the replacement of one amidine group actually increases the binding affinity, presumably by allowing an improved bonding orientation of the remaining amidine. Thus, DB607 (methoxy for amidine) and DB960 (N-methyl benzimidazole for benzamidine) display a somewhat higher affinity than DB75, although the fixed angle was unchanged. Introduction of a pyridine-N in the ortho-position with respect to the amidine functionality (DB994), dramatically reduces the pKa of the amidine moiety (Wang et al., 2010), resulting in a complete loss of binding affinity (Ki = 167 ± 20 µM), while this was not observed for the corresponding meta-pyridine derivative (DB829). The unfavourable furan bond angle is further demonstrated by the distally elongated analogues DB1061 and DB1062 that approximate the inter-amidine distance of pentamidine but showed no improvement in binding affinity (Table 2). Replacement of furan with thiazole (ER1004) or methylpyrrole (DB320), which feature a similar bond angle, also revealed comparable binding affinities. In contrast, a 2,5-substituted thiophene (DB686 and DB1063) or 2,5-substituted selenophene (DB1213) as a bio-isosteric replacement for the furan linker resulted in significantly higher binding affinities when compared to their matched pair analogue (DB1063/DB1061 and DB1213/DB75), which we attribute to a larger benzamidine-benzamidine angle. This is corroborated by the much weaker binding of the 2,4-thiophene derivative DB1077. A terminal amidine cap (imidazoline) reduced affinity as it did for pentamidine (compare DB1061/DB1062 and DB1063/DB1064 (Table 2)). A difuran spacer (DB914) resulted in a high affinity binder (Ki = 0.073 µM) because the two furans orient themselves in a trans conformation, resulting in a near-linear molecule.

Selection of diamidine analogues with aromatic linkers.
CompoundR1R2ArXYKi (µM)δ(ΔG0) PMD (kJ/mol)
DB75AmAmCHCH38.2 ± 10.217.3
DB607AmOCH3CHCH18.1 ± 1.915.5
DB960AmNMBaCHCH16.6 ± 3.515.3
DB994AmAmNCH167 ± 2021.0
DB829AmAmCHN39.9 ± 8.017.4
DB1061EtAmEtAmCHCH32.3 ± 6.016.9
DB10622MeIm2MeImCHCH59.6 ± 11.218.4
ER1004AmAmCHCH68.7 ± 16.018.8
DB320AmAmCHCH71.3 ± 12.118.9
DB686GuaGuaCHCH0.29 ± 0.115.2
DB1063EtAmEtAmCHCH0.40 ± 0.106.0
DB10642MeIm2MeImCHCH3.0 ± 0.8211.0
DB1213AmAmCHCH0.72 ± 0.177.5
DB1077AmAmCHCH13.8 ± 3.114.8
DB914AmAmCHCH0.073 ± 0.0131.8

Am, amidine; Im, imidazole; EtAm, ethylamidine; 2MeIm, 2-methylimidazoline; NMB, N-methyl benzimidazole. aThis compound lacks the second benzene ring and features the terminal NMB moiety instead. Ki is the inhibition constant for [3H]-pentamidine transport by TbAQP2/HAPT1. δ(ΔG0) PMD is the difference in Gibbs Free Energy of interaction of the substrate with TbAQP2 with the same value for pentamidine (PMD). All Ki values are the average and SEM of at least 3–4 independent experiments.

Modifications to the phenyl rings of pentamidine

Substituents in the ortho-position (relative to the alkyloxy substituent) of pentamidine were poorly tolerated, including chloride or iodide (RT-43, iodopentamidine; Table 1); the amide analogue displayed no affinity at all (RT-46). Such substituents will cause an out-of-plane conformation of the alkoxy-group to avoid clashing with the ortho-substituent; high-affinity pentamidine binding appears to require a coplanar arrangement of the first methylene bound to the oxygen. Similarly, the introduction of an ortho-pyridine N (RT-52) led to a δ(ΔG0) of 13.7 kJ/mol. This derivative exhibits a conformational bias towards an anti-orientation of the ether oxygen and pyridine nitrogen (Figure 9C; Chein and Corey, 2010). The regio-isomeric meta-pyridine (RT-53) was completely inactive, reflecting the need for a positively charged amidine, as this analogue has a significantly reduced pKa (Wang et al., 2010) (see furan-spaced analogue DB994, supra).

Non-diamidine trypanocides

The important veterinary trypanocide isometamidium, a hybrid of the phenanthridine ethidium and the diamidine diminazene, inhibited HAPT1-mediated [3H]-pentamidine uptake with a Ki of only 3.5 µM (Supplementary file 1), most probably through an interaction with its benzamidine moiety, as ethidium displayed virtually no affinity (Ki = 97 µM). However, we found no evidence that HAPT1/AQP2 is able to transport the bulky isometamidium molecule. For instance, the 2T1, tbaqp2 null, TbAQP2 expressed in tbaqp2 null, and the tbaqp2/tbaqp3 null strains displayed statistically identical EC50 values for isometamidium (112 ± 12 nM, 103 ± 14 nM, 98 ± 24 nM and 95 ± 12 nM, respectively; p>0.05, Student’s unpaired t-test), and the EC50 values for ethidium were also identical for each of these strains (1.32 ± 0.07 µM, 1.39 ± 0.08 µM, 1.35 ± 0.11 µM and 1.38 ± 0.14 µM, respectively). It is thus likely that isometamidium acts as an extracellular inhibitor rather than a substrate for HAPT/AQP2, as it does for the TbAT1/P2 transporter (de Koning, 2001b). The nitro-heterocyclic trypanocide megazol (Carvalho et al., 2014), curcumin and its trypanocidal analogue AS-HK14 (Alkhaldi et al., 2015) failed to inhibit HAPT1. Two trypanocidal bis-phosphonium compounds, CD38 (Taladriz et al., 2012) and AHI43 (Alkhaldi et al., 2016) did inhibit pentamidine uptake (Ki5–10 µM), whereas two related compounds, CDIV31 and AHI15 (Taladriz et al., 2012), did not. Phloretin, which inhibits human AQP9 and AQP3 (Geng et al., 2017),47 displayed a Ki of 1.76 µM for HAPT1/TbAQP2.

Are all the HAPT1/AQP2 inhibitors transported?

In an uptake-by-endocytosis model some correlation between TbAQP2 binding energy and TbAQP2-mediated uptake rates for each analogue would be expected, although in many instances of receptor-mediated endocytosis factors such as gating and the induction of conformational changes in the carrier might complicate such a correlation. Nevertheless, if we could observe a significant correlation of these parameters it would strengthen the argument for the endocytosis model. We were unable to ascertain the existence of such a correlation directly, for lack of radiolabelled substrates other than pentamidine and diminazene and thus used the Resistance Factor (RF; EC50(aqp2/3 null)/EC50(TbAQP2-WT)) as a proxy: clearly, a compound with a significant RF is internalized by TbAQP2. We observed a poor correlation between HAPT1 binding affinity and the level of resistance in the tbaqp2/tbaqp3 null strain (r2 = 0.039, Figure 9—figure supplement 1; n = 30), with many inhibitors, even those with high affinity, not displaying any significant resistance in the null line. This indicates that many of these compounds inhibit HAPT1/TbAQP2 but are not transported by it. This fails to support a model in which pentamidine binds and is then internalized by endocytosis: the inhibitors do not show resistance in the tbaqp2/tbaqp3 null line, whereas substrates do. The lack of any correlation would be especially problematic for a model of ‘passive’ pentamidine endocytosis, where the drug merely piggy-backs on TbAQP2 as it is internalised in its regular turnover schedule (t1/2 >4 h [Quintana et al., 2020]), without inducing any conformation changes in the receptor protein. The caveat inherent to using the RF instead of rate of transport is that it cannot be excluded that some of the test compounds are AQP2 substrates yet predominantly taken up by transporters other than TbAQP2, and hence show a low RF.

SAR summary

Figure 9D summarises the structure-activity relationship of pentamidine interactions with HAPT1/TbAQP2. No modification in any part of pentamidine improved affinity for TbAQP2, but virtually every modification resulted in a significant loss of binding activity (a similar analysis with melaminophenyl arsenicals was impossible for lack of the required organo-arsenicals). The results clearly demonstrate that at least both amidine groups and one or both ether oxygens are involved in interactions with AQP2, the sum of which adds up to the unusually high binding energy for this substrate-transporter pair (ΔG0 = −42.6 kJ/mol). These results are fully compatible with pentamidine binding in an elongated orientation, and are in complete agreement with the modelling and molecular dynamics, and the mutational analysis presented above, strengthening those conclusions using a completely different approach.

Discussion

There is overwhelming consensus that expression of TbAQP2 is associated with the extraordinary sensitivity of T. brucei to pentamidine and melaminophenyl arsenicals, and that mutations and deletions in this locus cause resistance (Baker et al., 2012, Baker et al., 2013; Graf et al., 2013; Graf et al., 2015; Graf et al., 2016; Pyana Pati et al., 2014; Munday et al., 2014; Munday et al., 2015a; Unciti-Broceta et al., 2015). What has remained unclear, however, is the mechanism underpinning these phenomena – there are currently no documented other examples of aquaporins transporting such large molecules. Yet, considering how ubiquitous aquaporins are to almost all cell types, this question is of wide pharmacological importance: if large cationic and neutral drugs (pentamidine and melarsoprol, respectively) can be taken up via an aquaglyceroporin of T. brucei, what other pharmacological or toxicological roles may these channels be capable of in other cell types? This manuscript clearly shows that changes in the TbAQP2 WGYR and NPA/NPA motifs, which collectively enlarge the pore and remove the cation filter, allow the passage of these drugs into the cell, and thereby underpin the very high sensitivity of the parasite to these drugs.

TbAQP2 has evolved, apparently by positive selection given the high dN/dS ratio, to remove all main constriction points, including the aromatic amino acids and the cationic arginine of the ar/R selectivity filter, and the NPA/NPA motif, resulting in an unprecedentedly enlarged pore size (Baker et al., 2012; Munday et al., 2015a). Whereas the advantage of this to T. b. brucei is yet unknown, the adaptation is stable within the brucei group of trypanosomes, and found in T. b. rhodesiense (Munday et al., 2014; Graf et al., 2016), T. b. gambiense (Graf et al., 2013; Graf et al., 2015; Munday et al., 2014; Pyana Pati et al., 2014), T. equiperdum and T. evansi (Philippe Büscher and Nick Van Reet, unpublished). As such, it is not inappropriate to speculate that the wider pore of TbAQP2 (i) allows the passage of something not transported by TbAQP1 and TbAQP3; (ii) that this confers an a yet unknown advantage to the cell; and (iii) that uptake of pentamidine is a by-product of this adaptation.

It is difficult to reconcile the literature on pentamidine transport/resistance with uptake via endocytosis. For instance, the rate of endocytosis in bloodstream trypanosomes is much higher than in the procyclic lifecycle forms (Langreth and Balber, 1975; Zoltner et al., 2016), yet the rate of HAPT-mediated [3H]-pentamidine uptake in procyclics is ~10 fold higher than in bloodstream forms (de Koning, 2001a; Teka et al., 2011), despite the level of TbAQP2 expression being similar in both cases (Siegel et al., 2010; Jensen et al., 2014). Moreover, in procyclic cells TbAQP2 is spread out over the cell surface (Baker et al., 2012) but endocytosis happens exclusively in the flagellar pocket (Field and Carrington, 2009) (which is 3-fold smaller in procyclic than in bloodstream forms [Demmel et al., 2014]), as the pellicular microtubule networks below the plasma membrane prevent endocytosis (Zoltner et al., 2016). Thus, TbAQP2-mediated pentamidine uptake should be all but impossible in procyclic T. brucei, if dependent on endocytosis. Similarly, the expression of TbAQP2 in Leishmania mexicana promastigotes produced a rate of [3H]-pentamidine uptake more than 10-fold higher than observed in T. brucei BSF (Munday et al., 2014), despite these cells also having a low endocytosis rate (Langreth and Balber, 1975). The Km and inhibitor profile of the TbAQP2-mediated pentamidine transport in these promastigotes was indistinguishable from HAPT in procyclic or bloodstream form T. brucei (de Koning, 2001a).

The experimental Vmax for HAPT-mediated pentamidine uptake in T. brucei BSF and procyclics (de Koning, 2001a) can be expressed as 9.5 × 105 and 8.5 × 106 molecules/cell/h, respectively; given a 1:1 stoichiometry for AQP2:pentamidine the endocytosis model would require the internalisation and recycling of as many units of TbAQP2 and this seems unlikely, especially in procyclic cells, as even in BSF the half-life time for TbAQP2 turnover is >4 hr (Quintana et al., 2020) and procyclic cells have a lower endocytosis rate and cannot easily internalise the aquaporins spread over the cell surface, as discussed above. Given the observed rate of uptake and turnover rate, this would require the presence of ~4 × 106 TbAQP2 units per BSF cell in the flagellar pocket. These observations are all inconsistent with the contention that pentamidine uptake by trypanosomes is principally dependent on endocytosis. Although it is likely that AQP2-bound pentamidine is internalised as part of the natural turnover rate of the protein, this is not likely to contribute very significantly to the overall rate of uptake of this drug.

Furthermore, the Gibbs free energy of −42 kJ/mol for the pentamidine/AQP2 interaction (de Koning, 2001a; Zoltner et al., 2016) is highly unlikely to be the result of the one interaction between one terminal amidine and Asp265 as required in the endocytosis model (Song et al., 2016). For the TbAT1 transporter, a double H-bond interaction of Asp140 with the N1(H)/C(6)NH2 motif of adenosine or with one amidine of pentamidine (Munday et al., 2015b) is estimated to contribute only ~16 kJ/mol to the total ΔG0 of −34.5 kJ/mol for adenosine (−36.7 kJ/mol, pentamidine) (de Koning and Jarvis, 1999). The endocytosis model also does not address the internalisation of melaminophenyl arsenicals, which presumably would equally need access to Asp265, or address why most diamidines including furamidines and diminazene aceturate are at best extremely poor substrates for TbAQP2 (Teka et al., 2011; Ward et al., 2011).

Here we systematically mapped the interactions between the aquaporin and pentamidine (ΔG0 for 71 compounds), yielding a completely consistent SAR with multiple substrate-transporter interactions, summarised in Figure 9D. The evidence strongly supports the notion that pentamidine engages TbAQP2 with both benzamidine groups and most probably with at least one of the linker oxygens, and that its flexibility and small width are both required to optimally interact with the protein. This is completely corroborated by molecular dynamics modelling, which shows minimal energy to be associated with a near-elongated pentamidine centrally in the TbAQP2 pore, without major energy barriers to exiting in either direction, but driven to the cytoplasmic side by the membrane potential. This contrasts with the contention (Song et al., 2016) that pentamidine could not be a permeant for TbAQP2 because it did not transport some small cations and that this proves that the larger pentamidine cannot be a substrate either. There is scant rationale for that assertion: out of many possible examples: there are 5 orders of magnitude difference in affinity for pentamidine and para-hydroxybenzamidine (35 nM vs 2.9 mM; Supplementary file 1); adenine is not a substrate for the T. brucei P1 adenosine transporter (de Koning and Jarvis, 1999), the SLC1A4 and SLC1A5 neutral amino acid transporters transport Ala, Ser, Cys and Thr but not Gly (Kanai et al., 2013), Na+ is not a permeant of K+ channels (Zhorov and Tikhonov, 2013) and some NAT family transporters from bacterial, plant and fungal species display much higher affinity for xanthine and uric acid but not for hypoxanthine (Gournas et al., 2008).

The endocytosis model identifies only two key residues for pentamidine access (Leu264) and binding (Asp265) in TbAQP2 (Song et al., 2016). Yet, multiple clinical isolates and laboratory strains contain chimeric AQP2/3 genes associated with resistance and/or non-cure that have retained those residues and should thus allow binding and internalisation of pentamidine (Graf et al., 2013; Pyana Pati et al., 2014; Unciti-Broceta et al., 2015; Munday et al., 2014). Although we find that introduction of the AQP3 Arg residue in position 264 (TbAQP2L264R) disables pentamidine transport, we would argue that this is because the positively charged arginine, in the middle of the pore, is blocking the traversing of all cations through the pore, as is its common function in aquaporins (Beitz et al., 2006; Wu et al., 2009). Indeed, the W(G)YR filter residues appear to be key determinants for pentamidine transport by AQPs and the introduction of all three TbAQP2 residues into TbAQP3 (AQP3W102I/R256L/Y250L) was required to create an AQP3 that at least mildly sensitised to pentamidine, and facilitated a detectable level of pentamidine uptake. Conversely, any one of the mutations I110W, L258Y or L264R was sufficient to all but abolish pentamidine transport by TbAQP2. Similarly, the conserved NPA/NPA motif, and particularly the Asp residues, present in TbAQP3 but NSA/NPS in TbAQP2, is also associated with blocking the passage of cations (Wree et al., 2011). The unique serine residues in this TbAQP2 motif, halfway down the pore, might be able to make hydrogen bonds with pentamidine. Reinstating the NPA/NPA motif resulted in a TbAQP2 variant with a 93.5% reduced rate of [3H]-pentamidine transport.

Tryptophan residues were introduced towards the cytoplasmic end of the TbAQP2 pore (L84W, L118W, L218W) to test the hypothesis that introducing bulky amino acids in that position would block the passage of pentamidine. Each of these mutants was associated with reduced sensitivity to pentamidine and cymelarsan and a > 90% reduction in [3H]-pentamidine uptake. This effect was size-dependent as the pentamidine transport rate of L84M and L218M was statistically identical to that of control TbAQP2 cells, and L118M also displayed a higher transport rate than L118W (p<0.0001). These mutant AQPs were still functional aquaglyceroporins as their expression in tbaqp1-2-3 null cells made those cells less sensitive to the TAO inhibitor SHAM, and increased glycerol uptake.

Independence from endocytosis was investigated by employing the tetracycline-inducible CRK12 RNAi cell line previously described to give a highly reproducible and progressive endocytosis defect in T. brucei (Monnerat et al., 2013), with the aim to distinguish between uptake via endocytosis and transporters, as current evidence suggests that in T. brucei all endocytosis, taking place exclusively in the flagellar pocket, is clathrin-dependent and AP-2 independent (Morgan et al., 2002; Allen et al., 2003). This means that the endocytotic mechanisms of TbAQP2 and suramin receptor ISG75, which are both directed to the lysosome after ubiquitylation (Quintana et al., 2020; Zoltner et al., 2015), are likely to be similar enough for a direct comparison. Twelve hours after CRK12 RNAi induction pentamidine transport was not significantly reduced although uptake of [3H]-suramin, which is accumulated by endocytosis through the T. brucei flagellar pocket (Zoltner et al., 2016), was reduced by 33% (p=0.0027), indicating successful timing of the experiment to the early stage of endocytosis slow-down. We also show that pentamidine, at approximately its half-maximal occupancy concentration, did not influence the half-life time of TbAQP2 turnover, as is often the case in receptor-mediated endocytosis. Nor should TbAQP2, which exists as a rigid tetramer of tetramers in T. brucei bloodstream forms (Quintana et al., 2020), be able to undergo the type of conformational change necessary to signal receptor occupancy and internalisation as observed in well-documented examples of ligand-triggered internalisation of transporters (e.g. Gournas et al., 2010; Gournas et al., 2017; Ghaddar et al., 2014; Keener and Babst, 2013). Thus, the combined evidence, taken together, strongly suggests that pentamidine is not taken up by endocytosis, not induce endocytosis of TbAQP2.

Although several ionophores, including CCCP, nigericin and gramicidin strongly inhibited pentamidine uptake, similar to what has been previously reported for transport processes in T. brucei that are linked to the protonmotive force (de Koning and Jarvis, 1997a; de Koning and Jarvis, 1997b; de Koning and Jarvis, 1998; de Koning et al., 1998), this is probably due to the inside-negative membrane potential of −125 mV (de Koning and Jarvis, 1997b) attracting the dicationic pentamidine. This is consistent with the prediction of the molecular dynamics modelling, and the reported role of the HA1–three proton pumps in pentamidine but not melarsoprol resistance (Alsford et al., 2012; Baker et al., 2013). Although CCCP does inhibit pentamidine through direct, competitive inhibition of TbAQP2 as well, this only starts to have a measurable impact above ~5 µM (IC50 of 20.7 µM), whereas its effects after preincubation, i.e. the combination of competitive inhibition and reducing the protonmotive force, shows ~63% inhibition of pentamidine transport at 1 µM and ~90% inhibition at 5 µM, showing that the more important effect of CCCP is via reduction of the PMF. This is consistent with the conclusion from the molecular dynamics analysis that inward pentamidine flux is dependent on the inside-negative membrane potential.

Altogether, we conclude that the primary entry of the sleeping sickness drugs pentamidine and melarsoprol into T. brucei spp. is through the unusually large pore of TbAQP2, rendering the parasite extraordinarily sensitive to the drugs (compare Leishmania mexicana [Munday et al., 2014]). This is the first report providing detailed mechanistic evidence of the uptake of organic drugs (of MW 340 and 398, respectively) by an aquaporin. We show that this porin has evolved through positive selection and identify the adaptations in the constriction motifs that enabled it. We consider that other pore-opening adaptations may have evolved in other organisms, including pathogens, which could initiate the pharmacological exploitation of aquaporins and lead to the design of new drug delivery strategies.

Materials and methods
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Cell line (Trypanosoma brucei)2T1David Horn
Cell line (Trypanosoma brucei)aqp2/aqp3 nullDavid Horn
Cell line (Trypanosoma brucei)aqp1-3 nullDavid Horn
Cell line (Trypanosoma brucei)CRK12 RNAiTansy Hammarton
Recombinant DNA reagentpRPaDavid Hornplasmid for expression in T. brucei
Gene (Trypanosoma brucei)AQP2TriTrypDBTb927.10.14170Sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2S131P/S263AThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2L258YThis paperMutated TbAQP2; sequence inSupplementary file 2
Gene (Trypanosoma brucei)TbAQP2I110WThis paperMutated TbAQP2; sequence inSupplementary file 2
Gene (Trypanosoma brucei)TbAQP2L264RThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2I110W /L264RThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2I190TThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2W192GThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2I190T/W192GThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2L84WThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2L84MThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2L118WThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2L118MThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2L218WThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2L218MThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP2L84W/L118WThis paperMutated TbAQP2; sequence in Supplementary file 2
Gene (Trypanosoma brucei)AQP3TriTrypDBTb927.10.14160Sequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP3W102I/R256LThis paperSequence in Supplementary file 2
Gene (Trypanosoma brucei)TbAQP3W102I/R256L/Y250LThis paperSequence in Supplementary file 2
Chemical compound, drugAll itemised in Supplementary file 1Custom inhibitors and pentamidine analogues
Chemical compound, drugPentamidine isethionateSigma-Aldrich
Chemical compound, drugDiminazene aceturateSigma-Aldrich
Chemical compound, drugSuraminSigma-Aldrich
Chemical compound, drugCymelarsangift from C. Michael Turner
Chemical compound, drug[3H]-pentamidineGE Healthcare Life SciencesRadiochemical
Chemical compound, drug[3H]-SuraminAmerican Radiolabeled ChemicalsRadiochemical
Chemical compound, drug[3H]-glycerolAmerican Radiolabeled ChemicalsRadiochemical
Trypanosome strains and cultures

The drug-sensitive clonal T. b. brucei strain 427 (MiTat 1.2/BS221) (de Koning et al., 2000) was used for all the work on the SAR of pentamidine transport. The tbaqp2/tbaqp3 null cells (Baker et al., 2012) and tbaqp1-2-3 null cells (Jeacock et al., 2017) (both obtained from David Horn, University of Dundee, UK) are derived from the 2T1 strain of T. b. brucei (Alsford and Horn, 2008). The CRK12 RNAi cell line28 was obtained from Dr Tansy Hammarton (University of Glasgow, UK) and is also based on the 2T1 cell line; RNAi expression was induced with 1 µg/ml tetracycline in the medium. All experiments were performed with bloodstream form trypanosomes grown in vitro in HMI-11 medium as described (Wallace et al., 2002) at 37°C in a 5% CO2 atmosphere. Cultures were routinely maintained in 10 ml of this medium, being seeded at 5 × 104 cells/ml and passed to fresh medium at reaching approximately 3 × 106 cells/ml after 48 hr. For transport experiments 150 or 200 ml of culture was seeded at the same density in large flasks and incubated until the culture reached late-log phase.

Materials

A complete list of diamidine analogues and other chemicals used for the SAR study is given as a table with their sources (Supplementary file 1). Ionophores and uncouplers nigericin, gramicidin, carbonyl cyanide m-chlorophenyl hydrazone (CCCP) and valinomycin, as well as the T. brucei proton pump inhibitor N-ethylmaleimide (NEM) were all purchased from Sigma-Aldrich. New compounds synthesised for this study are listed and described in Supplementary file 3.

Transport assays - Transport of [3H]-pentamidine was performed exactly as previously described for various permeants (Wallace et al., 2002; Bridges et al., 2007; Teka et al., 2011) in a defined assay buffer (AB; 33 mM HEPES, 98 mM NaCl, 4.6 mM KCl, 0.55 mM CaCl2, 0.07 mM MgSO4, 5.8 mM NaH2PO4, 0.3 mM MgCl2, 23 mM NaHCO3, 14 mM glucose, pH 7.3). [3H]-pentamidine was custom-made by GE Healthcare Life Sciences (Cardiff, UK) with a specific activity of 88 Ci/mmol. Incubations of bloodstream form trypanosomes with 30 nM of this label (unless otherwise indicated) were performed in AB at room temperature for 60 s (unless otherwise indicated) and terminated by addition of 1 ml ice-cold ‘stop’ solution (1 mM unlabelled pentamidine (Sigma) in AB) and immediate centrifugation through oil (7:1 dibutylphthalate:mineral oil v/v (both from Sigma)). Transport was assessed in the presence of 1 mM adenosine to block uptake through the P2 aminopurine transporter; adenosine does not affect HAPT1-mediated transport (de Koning, 2001a; Bridges et al., 2007). Inhibition assays were performed routinely with 6–10 different concentrations of inhibitor over the relevant range, diluting stepwise by one third each time, in order to obtain a well-defined and accurate sigmoid plot and IC50 value (inhibitor concentration giving 50% inhibition of pentamidine transport; calculated by non-linear regression using Prism 6.0 (GraphPad), using the equation for a sigmoid curve with variable slope). Highest concentration was usually 1 mM unless this was shown to be insufficient for good inhibition, or when limited by solubility. Ki values were obtained from IC50 values usingki=IC50/[1+(L+km)]in which L is the [3H]-pentamidine concentration and Km the Michaelis-Menten constant for pentamidine uptake by HAPT1 (Wallace et al., 2002). The Gibbs Free energy of interaction ΔG0 was calculated fromΔG0=−RTlnKiin which R is the gas constant and T is the absolute temperature (Wallace et al., 2002). Transport of [3H]-glycerol and [3H]-suramin was performed essentially as for [3H]-pentamidine. For [3H]-glycerol (American Radiolabeled Chemicals, 40.0 Ci/mmol), 107 BSF T. brucei were incubated with radiolabel at a final concentration of 0.25 µM, for one minute. When the effect of CCP was studied, CCCP was added 3 min prior to the addition of the radiolabel. [3H]-suramin (American Radiolabeled Chemicals, 20.0 Ci/mmol) was also used at 0.25 µM final concentration, using 15 min incubations in the presence and absence of 100 µM unlabelled suramin (used as saturation) control.

Construction of AQP mutants and transfection

All mutations in the TbAQP2 and TbAQP3 genes were introduced to the relevant backbone WT vector, either pRPaGFP-AQP2 or pRPaGFP-AQP3 (Baker et al., 2012), by site-directed mutagenesis. Use of the pRPa vector for transfection of 2T1-derived T. brucei ensures integration in a prepared locus in the ribosomal rRNA spacer region and a high level of stable expression (Alsford et al., 2005). For mutations S131P, S263A, I110W, L264R, L258Y, I190T and W192G in AQP2, and W102I, R256L and Y250L in AQP3 mutations were inserted using the QuikChange II kit (Agilent, Santa Clara, CA, USA), following the manufacturer's instructions. For mutations L84W, L118W, L218W, L84M, L118M and L218M were introduced using the Q5 Site-Directed Mutagenesis Kit (E0554S), (New England BioLabs) according to manufacturer’s instructions.

The following primer pairs (itemised in Supplementary file 4) were used to insert the named TbAQP2 mutations: for S131P, primers HDK1062 and HDK1063; in combination with mutation S263A, using primers HDK1064 and HDK1065 to produce plasmid pHDK166. For I110W, primers HDK607 and HDK608, to produce pHDK84; for L264R, primers HDK609 and HDK610 to produce pHDK167; the combination I110W/L264R was produced using primers HDK609 and HDK610 on plasmid pHDK84 to give plasmid pHDK78; for L258Y, primers HDK1109 and HDK1110 to produce pHDK168; for I190T, primers HDK1056 and HDK1057 to produce pHDK163; for W192G, primers HDK1058 and HDK1059 to produce pHDK164; the combination I190T/W192G was produced using primers HDK1060 and HDK1061 on plasmid pHDK163 to give plasmid pHDK165; for L84W, primers HDK1276 and HDK1277, producing pHDK210; for L118W, primers HDK1274 and HDK1275, producing pHDK208; for L218W, primers HDK1272 and HDK1273, producing pHDK209; for the combination L84W/L118W, primers HDK1276 and HDK1277 on template pHDK208, producing pHDK227; for L84M, primers HDK1364 and HDK1367, producing pHDK234; for L118M, primers HDK1365 and HDK1367, producing pHDK235; and for L218M, primers HDK1366 and HDK1367, producing pHDK236. To insert the named mutations into TbAQP3, the following primers were used (Supplementary file 5): for W102I, primers HDK511 and HDK512, in combination with mutation R256L, with primers HDK513 and HDK514, to produce plasmid pHDK71; and to add mutation Y250L to this combination, primers HDK795 and HDK796, to produce pHDK121. All plasmids were checked by Sanger Sequencing (Source BioScience, Nottingham, UK) for the presence of the correct mutation(s) and the cassette for integration digested out with AscI (NEB, Hitchin, UK) prior to transfection.

For transfection, 10 µg of digested plasmid and 1–2 × 107 parasites of the desired cell line (either aqp2/aqp3 null or aqp1/aqp2/aqp3 null) were resuspended in transfection buffer and transfected using an Amaxa Nucleofector, with program X-001. After a recovery period (8–16 hr) in HMI-11 at 37°C and 5% CO2, the parasites were cloned out by limiting dilution with the selection antibiotic (2.5 µg/ml hygromycin). In all cases the presence of the construct and its correct integration into the designed rRNA locus was verified by three PCR reactions, one using primers for the amplification of the full-length aquaporin (primers HDK529 and HDK209). The second PCR was performed to amplify the gene with surrounding parts of the expression cassette using (primers HDK1011 and HDK430). The third PCR was to assess whether the expression cassette had linearized and integrated into the T. brucei genome using (primers HDK991 and HDK713).

Drug sensitivity assays

Drug sensitivity assays for T. b. brucei bloodstream forms used the cell viability dye resazurin (Sigma) and were performed exactly as described (Wallace et al., 2002; Bridges et al., 2007) in 96-well plates with doubling dilutions of test compound, starting at 100 µM, over 2 rows of the plate (23 dilutions plus no-drug control). Incubation time with test compound was 48 hr (37°C/5% CO2), followed by an additional 24 hr in the presence of the dye.

Protein turnover and western blotting

T. brucei 2T1 containing 3×HAAQP2 were incubated with 1 µg/ml tetracycline for 24 hr at 37°C/5% CO2 to induce expression. Tetracycline was then washed away by four consecutive washes with fresh supplemented HMI-9, and cells were then incubated with 25 nM of pentamidine (5 × EC50) for 1 hr at 37°C/5% CO2 prior to harvest cells for pulse chase experiments. To determine protein half-life, translation was blocked by addition of cycloheximide (100 µg/ml) and cells were harvested at various times by centrifugation (800 × g for 10 min at 4°C). Cells were washed with ice-cold PBS, then resuspended in 1 × SDS sample buffer (Thermo) and incubated at 70°C for 10 min. Proteins were subjected to electrophoresis using 4–12% precast acrylamide Bis-Tris gels (Thermo) and transferred to polyvinylidene difluoride (PVDF; Sigma-Aldrich) membranes with a iBlot2 system (Thermo) at 23 V for 6 min, exactly as described (Quintana et al., 2020), blocking non-specific binding with 5% (w/v) bovine serum albumin (BSA; Sigma) in Tris-buffered saline (pH 7.4) with 0.2% (v/v) Tween-20 (TBST) and using rat anti-HA IgG1 (Sigma) or anti-mouse β-tubulin (clone KMX-1; Millipore) at 1:5000 or 1:10,000 dilution in TBST, respectively. Membranes were washed five times with TBST and then incubated in TBST/1% BSA with the appropriate horseradish peroxidase (HRP)-coupled secondary antibody (Sigma), at 1: 10,000. Bands were visualised using the ECL Prime Western Blotting System (Sigma) and GE healthcare Amersham Hyperfilm ECL (GE). Densitometry quantification was conducted using ImageJ software (NIH).

Molecular dynamics

Molecular dynamics simulations were performed using the GROMACS software package, version 5.1.1 (Abraham et al., 2015). We used the coordinates from the homology model of TbAQP2 published in Figure 2A in Munday et al., 2015a, which was inserted into POPC/POPE (4:1) membranes, approximately reflecting the membrane composition of T. b. brucei (Smith and Bütikofer, 2010). The membrane models were constructed using the CHARMM-GUI webserver (Jo et al., 2008). Subsequently, extended stability tests of the modelled structure and the bound pentamidine were carried out using unbiased simulations of 100 ns length. The root-mean-square deviation (RMSD) of the protein remained relatively low with a backbone RMSD converging to ~3 Å after 100 ns simulated time (Figure 8—figure supplement 1) bound to the binding site defined previously using molecular docking (Figure 8A; Munday et al., 2015a). For these and all following simulations, we used the CHARMM36 force field (Klauda et al., 2010) pentamidine was parameterised using the CHARMM generalized force field approach (CHGenFF [Vanommeslaeghe et al., 2010]). All simulations employed a time step of 2 fs under an NPT ensemble at p=1 bar and T = 310 K. To obtain non-equilibrium work values for removing pentamidine from the internal AQP2 binding site, we then conducted steered MD simulations with a probe speed of 0.005 nm/ns and a harmonic force constant of 300 kJ/mol nm2, pulling pentamidine in both directions along the pore axis. The free energy profile of pentamidine binding to the AQP2 pore was reconstructed by using the Jarzynski, 1997 equality.

Statistical analysis

All transport experiments were performed in triplicate and all values such as rate of uptake, percent inhibition, Ki, Km, Vmax etc were performed at least three times completely independently. For drug sensitivity tests, all EC50 values were based on serial dilutions over two rows of a 96-well plate (23 doubling dilutions plus no-drug control), which were obtained independently at least three times. EC50 and IC50 values were determined by non-linear regression using the equation for a sigmoid curve with variable slope and are presented as average ± SEM. Statistical significance between any two data points was determined using Student’s t-test (unpaired, two-tailed).

Acknowledgements

This work was supported by the UK Medical Research Council (MRC) [grant G0701258 to HPdK] and by the US National Institutes of Health (NIH) [Grant No. GM111749 to DWB]. DC was supported by an MRC iCASE award [MR/R015791/1]. UZ acknowledges funding from the Scottish Universities Physics Alliance. AHA is funded through a PhD studentship from Albaha University, Saudi Arabia. GDC was funded by a PhD Studentship from Science Without Borders [206385/2014–5, CNPq, Brazil]. TS was funded via a Doctoral Training Programme of the MRC and the Cambridge Trust and SW was funded by a Sir Henry Dale fellowship of the Wellcome Trust and Royal Society. The authors thank Dr Tansy Hammarton for the use of the CRK2 RNAi cell line and Prof David Horn for the use of the aqp1-3 null cell line. This work was supported by a grant from the Wellcome Trust (204697/Z/16/Z to MCF. The authors are grateful to Professor George Diallinas, University of Athens, Greece, for his exceptionally insightful reviewer comments and have adopted several of his arguments in revision.

Additional informationCompeting interests

No competing interests declared

Was an employee of IOTA Pharmaceuticals Ltd at the time

Author contributions

Investigation

Investigation, Methodology

Investigation

Investigation

Investigation, Methodology

Investigation

Investigation

Investigation

Investigation

Investigation

Investigation

Investigation

Validation, Investigation, Methodology

Investigation

Investigation

Investigation

Investigation

Investigation

Investigation

Investigation

Formal analysis, Methodology, Writing - original draft

Investigation

Investigation

Investigation

Investigation

Investigation

Supervision

Conceptualization, Supervision, Funding acquisition, Methodology

Resources

Resources, Supervision

Conceptualization, Investigation, Methodology

Resources, Supervision

Resources, Supervision

Formal analysis, Supervision, Methodology

Conceptualization, Data curation, Formal analysis, Supervision, Funding acquisition, Methodology, Writing - original draft, Writing - review and editing

Additional filesTable of all HAPT/TbAQP2 inhibitors used in this study.Table with all sequences of TbAQP2 and the generated mutants thereof.

Mutated nucleotides and codons are indicated.

Synthesis of all new inhibitors used in this study.List of primers used in the construction of mutations in TbAQP2.Primers used for mutations in TbAQP3.Data availability

All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided that cover all figures and give raw data, averages, statistics etc.

ReferencesAbraham MJMurtola TSchulz RPáll SSmith JCHess BLindahl E2015GROMACS: high performance molecular simulations through multi-level parallelism from laptops to supercomputersSoftwareX1-2192510.1016/j.softx.2015.06.001Alkhaldi AACreek DJIbrahim HKim DHQuashie NBBurgess KEChangtam CBarrett MPSuksamrarn Ade Koning HP2015Potent trypanocidal curcumin analogs bearing a monoenone linker motif act on Trypanosoma brucei by forming an adduct with trypanothioneMolecular Pharmacology8745146410.1124/mol.114.09601625527638Alkhaldi AAMMartinek JPanicucci BDardonville CZíková Ade Koning HP2016Trypanocidal action of bisphosphonium salts through a mitochondrial target in bloodstream form Trypanosoma bruceiInternational Journal for Parasitology: Drugs and Drug Resistance6233410.1016/j.ijpddr.2015.12.002Allen CLGoulding DField MC2003Clathrin-mediated endocytosis is essential in Trypanosoma bruceiThe EMBO Journal224991500210.1093/emboj/cdg48114517238Alsford SKawahara TGlover LHorn D2005Tagging a T. brucei RRNA locus improves stable transfection efficiency and circumvents inducible expression position effectsMolecular and Biochemical Parasitology14414214810.1016/j.molbiopara.2005.08.00916182389Alsford SEckert SBaker NGlover LSanchez-Flores ALeung KFTurner DJField MCBerriman MHorn D2012High-throughput decoding of antitrypanosomal drug efficacy and resistanceNature48223223610.1038/nature1077122278056Alsford SHorn D2008Single-locus targeting constructs for reliable regulated RNAi and transgene expression in Trypanosoma bruceiMolecular and Biochemical Parasitology161767910.1016/j.molbiopara.2008.05.00618588918Baker NGlover LMunday JCAguinaga Andrés DBarrett MPde Koning HPHorn D2012Aquaglyceroporin 2 controls susceptibility to melarsoprol and pentamidine in African trypanosomesPNAS109109961100110.1073/pnas.120288510922711816Baker Nde Koning HPMäser PHorn D2013Drug resistance in african trypanosomiasis: the melarsoprol and pentamidine storyTrends in Parasitology2911011810.1016/j.pt.2012.12.00523375541Barrett MPVincent IMBurchmore RJKazibwe AJMatovu E2011Drug resistance in human african trypanosomiasisFuture Microbiology61037104710.2217/fmb.11.8821958143Beitz EWu BHolm LMSchultz JEZeuthen T2006Point mutations in the aromatic/arginine region in aquaporin 1 allow passage of urea, glycerol, Ammonia, and protonsPNAS10326927410.1073/pnas.050722510316407156Beno BRYeung KSBartberger MDPennington LDMeanwell NA2015A survey of the role of noncovalent sulfur interactions in drug designJournal of Medicinal Chemistry584383443810.1021/jm501853m25734370Brameld KAKuhn BReuter DCStahl M2008Small molecule conformational preferences derived from crystal structure data. A medicinal chemistry focused analysisJournal of Chemical Information and Modeling4812410.1021/ci700249418183967Bridges DJGould MKNerima BMäser PBurchmore RJde Koning HP2007Loss of the high-affinity pentamidine transporter is responsible for high levels of cross-resistance between arsenical and diamidine drugs in african trypanosomesMolecular Pharmacology711098110810.1124/mol.106.03135117234896Büscher PCecchi GJamonneau VPriotto G2017Human african trypanosomiasisThe Lancet3902397240910.1016/S0140-6736(17)31510-628673422Carter NSBerger BJFairlamb AH1995Uptake of diamidine drugs by the P2 nucleoside transporter in melarsen-sensitive and -resistant Trypanosoma brucei bruceiThe Journal of Biological Chemistry270281532815710.1074/jbc.270.47.281537499305Carter NSFairlamb AH1993Arsenical-resistant trypanosomes lack an unusual adenosine transporterNature36117317610.1038/361173a08421523Carvalho ASSalomão KCastro SLConde TRZamith HPCaffarena ERHall BSWilkinson SRBoechat N2014Megazol and its bioisostere 4H-1,2,4-triazole: comparing the trypanocidal, cytotoxic and genotoxic activities and their in vitro and in silico interactions with the Trypanosoma brucei nitroreductase enzymeMemórias Do Instituto Oswaldo Cruz10931532310.1590/0074-027614049724676659Chein RJCorey EJ2010Strong conformational preferences of heteroaromatic ethers and electron pair repulsionOrganic Letters1213213510.1021/ol902536419957955de Koning HPWatson CJJarvis SM1998Characterization of a nucleoside/proton symporter in procyclic Trypanosoma brucei bruceiJournal of Biological Chemistry2739486949410.1074/jbc.273.16.94869545276de Koning HPMacLeod ABarrett MPCover BJarvis SM2000Further evidence for a link between melarsoprol resistance and P2 transporter function in african trypanosomesMolecular and Biochemical Parasitology10618118510.1016/S0166-6851(99)00206-610743623de Koning HP2001aUptake of pentamidine in Trypanosoma brucei brucei is mediated by three distinct transporters: implications for cross-resistance with arsenicalsMolecular Pharmacology5958659210.1124/mol.59.3.58611179454de Koning HP2001bTransporters in african trypanosomes: role in drug action and resistanceInternational Journal for Parasitology3151252210.1016/S0020-7519(01)00167-911334936de Koning HPAnderson LFStewart MBurchmore RJSWallace LJMBarrett MP2004The trypanocide diminazene aceturate is accumulated predominantly through the TbAT1 purine transporter: additional insights on diamidine resistance in african trypanosomesAntimicrobial Agents and Chemotherapy481515151910.1128/AAC.48.5.1515-1519.2004de Koning HPJarvis SM1997aHypoxanthine uptake through a purine-selective nucleobase transporter in Trypanosoma brucei brucei procyclic cells is driven by protonmotive forceEuropean Journal of Biochemistry2471102111010.1111/j.1432-1033.1997.01102.x9288936de Koning HPJarvis SM1997bPurine nucleobase transport in bloodstream forms of Trypanosoma brucei is mediated by two novel transportersMolecular and Biochemical Parasitology8924525810.1016/S0166-6851(97)00129-19364969de Koning HPJarvis SM1998A highly selective, high-affinity transporter for uracil in Trypanosoma brucei Brucei: evidence for proton-dependent transportBiochemistry and Cell Biology7685385810.1139/o98-08610353720de Koning HPJarvis SM1999Adenosine transporters in bloodstream forms of Trypanosoma brucei: substrate recognition motifs and affinity for trypanocidal drugsMolecular Pharmacology561162117010.1124/mol.56.6.116210570043de Koning HPJarvis SM2001Uptake of pentamidine in Trypanosoma brucei is mediated by the P2 adenosine transporter and at least one novel, unrelated transporterActa Tropica8024525010.1016/S0001-706X(01)00177-211700182Demmel LSchmidt KLucast LHavlicek KZankel AKoestler TReithofer Vde Camilli PWarren G2014The endocytic activity of the flagellar pocket in Trypanosoma brucei is regulated by an adjacent phosphatidylinositol phosphate kinaseJournal of Cell Science1272351236410.1242/jcs.14689424639465Fairlamb AHCarter NSCunningham MSmith K1992Characterisation of melarsen-resistant Trypanosoma brucei brucei with respect to cross-resistance to other drugs and trypanothione metabolismMolecular and Biochemical Parasitology5321322210.1016/0166-6851(92)90023-D1501641Field MCCarrington M2009The trypanosome flagellar pocketNature Reviews Microbiology777578610.1038/nrmicro222119806154Frommel TOBalber AE1987Flow cytofluorimetric analysis of drug accumulation by multidrug-resistant Trypanosoma brucei and T. b. rhodesienseMolecular and Biochemical Parasitology2618319110.1016/0166-6851(87)90142-33431564Geng XMcDermott JLundgren JLiu LTsai KJShen JLiu Z2017Role of AQP9 in transport of monomethyselenic acid and seleniteBioMetals3074775510.1007/s10534-017-0042-x28798983Ghaddar KMerhi ASaliba EKrammer EMPrévost MAndré B2014Substrate-induced ubiquitylation and endocytosis of yeast amino acid permeasesMolecular and Cellular Biology344447446310.1128/MCB.00699-1425266656Giordani FMorrison LJRowan TGDE Koning HPBarrett MP2016The animal trypanosomiases and their chemotherapy: a reviewParasitology1431862188910.1017/S003118201600126827719692Gournas CPapageorgiou IDiallinas G2008The nucleobase-ascorbate transporter (NAT) family: genomics, evolution, structure-function relationships and physiological roleMolecular BioSystems440441610.1039/b719777b18414738Gournas CAmillis SVlanti ADiallinas G2010Transport-dependent endocytosis and turnover of a uric acid-xanthine permeaseMolecular Microbiology7524626010.1111/j.1365-2958.2009.06997.x20002879Gournas CSaliba EKrammer EMBarthelemy CPrévost MAndré B2017Transition of yeast Can1 transporter to the inward-facing state unveils an α-arrestin target sequence promoting its ubiquitylation and endocytosisMolecular Biology of the Cell282819283210.1091/mbc.e17-02-010428814503Graf FELudin PWenzler TKaiser MBrun RPyana PPBüscher Pde Koning HPHorn DMäser P2013Aquaporin 2 mutations in Trypanosoma brucei gambiense field isolates correlate with decreased susceptibility to pentamidine and melarsoprolPLOS Neglected Tropical Diseases7e247510.1371/journal.pntd.000247524130910Graf FEBaker NMunday JCde Koning HPHorn DMäser P2015Chimerization at the AQP2-AQP3 locus is the genetic basis of melarsoprol-pentamidine cross-resistance in clinical Trypanosoma brucei gambiense isolatesInternational Journal for Parasitology: Drugs and Drug Resistance5656810.1016/j.ijpddr.2015.04.00226042196Graf FELudin PArquint CSchmidt RSSchaub NKunz-Renggli CMunday JCKrezdorn JBaker NHorn DBalmer OCaccone ADe Koning HPMäser P2016Comparative genomics of drug resistance of the sleeping sickness parasite Trypanosoma brucei rhodesienseCellular and Molecular Life Sciences : CMLS733387340010.1007/s00018-016-2173-6Hutchinson RGibson W2015Rediscovery of Trypanosoma ( Pycnomonas ) suis , a tsetse-transmitted trypanosome closely related to T. bruceiInfection, Genetics and Evolution3638138810.1016/j.meegid.2015.10.018Jarzynski C1997Nonequilibrium equality for free energy differencesPhysical Review Letters782690269310.1103/PhysRevLett.78.2690Jeacock LBaker NWiedemar NMäser PHorn D2017Aquaglyceroporin-null trypanosomes display glycerol transport defects and respiratory-inhibitor sensitivityPLOS Pathogens13e100630710.1371/journal.ppat.100630728358927Jensen BCRamasamy GVasconcelos EJIngolia NTMyler PJParsons M2014Extensive stage-regulation of translation revealed by ribosome profiling of Trypanosoma bruceiBMC Genomics1591110.1186/1471-2164-15-91125331479Jo SKim TIyer VGIm W2008CHARMM-GUI: a web-based graphical user interface for CHARMMJournal of Computational Chemistry291859186510.1002/jcc.2094518351591Kanai YClémençon BSimonin ALeuenberger MLochner MWeisstanner MHediger MA2013The SLC1 high-affinity glutamate and neutral amino acid transporter familyMolecular Aspects of Medicine3410812010.1016/j.mam.2013.01.00123506861Keener JMBabst M2013Quality control and substrate-dependent downregulation of the nutrient transporter Fur4Traffic1441242710.1111/tra.1203923305501Klauda JBVenable RMFreites JAO'Connor JWTobias DJMondragon-Ramirez CVorobyov IMacKerell ADPastor RW2010Update of the CHARMM all-atom additive force field for lipids: validation on six lipid typesThe Journal of Physical Chemistry B1147830784310.1021/jp101759q20496934Langreth SGBalber AE1975Protein uptake and digestion in bloodstream and culture forms of Trypanosoma bruceiThe Journal of Protozoology22405510.1111/j.1550-7408.1975.tb00943.x1117436Lüscher Ade Koning HPMäser P2007Chemotherapeutic strategies against Trypanosoma brucei: drug targets vs. drug targetingCurrent Pharmaceutical Design1355556710.2174/13816120778016280917346174Mäser PSütterlin CKralli AKaminsky R1999Nucleoside transporter from Trypanosoma brucei involved in drug resistanceScience28524224410.1126/science.285.5425.24210398598Monnerat SAlmeida Costa CIForkert ACBenz CHamilton ATetley LBurchmore RNovo CMottram JCHammarton TC2013Identification and functional characterisation of CRK12:cyc9, a novel Cyclin-Dependent kinase (CDK)-Cyclin complex in Trypanosoma bruceiPLOS ONE8e6732710.1371/journal.pone.006732723805309Morgan GWHall BSDenny PWField MCCarrington M2002The endocytic apparatus of the kinetoplastida. part II: machinery and components of the systemTrends in Parasitology1854054610.1016/S1471-4922(02)02392-912482539Munday JCEze AABaker NGlover LClucas CAguinaga Andrés DNatto MJTeka IAMcDonald JLee RSGraf FELudin PBurchmore RJTurner CMTait AMacLeod AMäser PBarrett MPHorn DDe Koning HP2014Trypanosoma brucei aquaglyceroporin 2 is a high-affinity transporter for pentamidine and melaminophenyl arsenic drugs and the main genetic determinant of resistance to these drugsJournal of Antimicrobial Chemotherapy6965166310.1093/jac/dkt44224235095Munday JCSettimo Lde Koning HP2015aTransport proteins determine drug sensitivity and resistance in a protozoan parasite, Trypanosoma bruceiFrontiers in Pharmacology63210.3389/fphar.2015.00032Munday JCTagoe DNEze AAKrezdorn JARojas López KEAlkhaldi AAMcDonald FStill JAlzahrani KJSettimo LDe Koning HP2015bFunctional analysis of drug resistance-associated mutations in the Trypanosoma brucei adenosine transporter 1 (TbAT1) and the proposal of a structural model for the proteinMolecular Microbiology9688790010.1111/mmi.1297925708978P. De Koning H2020The drugs of sleeping sickness: their mechanisms of action and resistance, and a brief historyTropical Medicine and Infectious Disease51410.3390/tropicalmed5010014Park SKhalili-Araghi FTajkhorshid ESchulten K2003Free energy calculation from steered molecular dynamics simulations using Jarzynski’s equalityThe Journal of Chemical Physics1193559356610.1063/1.1590311Pyana Pati PVan Reet NMumba Ngoyi DNgay Lukusa IKarhemere Bin Shamamba SBüscher P2014Melarsoprol sensitivity profile of Trypanosoma brucei gambiense isolates from cured and relapsed sleeping sickness patients from the democratic republic of the CongoPLOS Neglected Tropical Diseases8e321210.1371/journal.pntd.000321225275572Quintana JFBueren-Calabuig JZuccotto Fde Koning HPHorn DField MC2020Instability of aquaglyceroporin (AQP) 2 contributes to drug resistance in Trypanosoma bruceiPLOS Neglected Tropical Diseases14e000845810.1371/journal.pntd.000845832644992Ríos Martínez CHMiller FGaneshamoorthy KGlacial FKaiser Mde Koning HPEze AALagartera LHerraiz TDardonville C2015A new nonpolar N-hydroxy imidazoline lead compound with improved activity in a murine model of late-stage Trypanosoma brucei brucei infection is not cross-resistant with diamidinesAntimicrobial Agents and Chemotherapy5989090410.1128/AAC.03958-1425421467Rollo IMWilliamson J1951Acquired resistance to 'Melarsen', tryparsamide and amidines in pathogenic trypanosomes after treatment with 'Melarsen' aloneNature16714714810.1038/167147a014806401Siegel TNHekstra DRWang XDewell SCross GA2010Genome-wide analysis of mRNA abundance in two life-cycle stages of Trypanosoma brucei and identification of splicing and polyadenylation sitesNucleic Acids Research384946495710.1093/nar/gkq23720385579Smith TKBütikofer P2010Lipid metabolism in Trypanosoma bruceiMolecular and Biochemical Parasitology172667910.1016/j.molbiopara.2010.04.001Song JBaker NRothert MHenke BJeacock LHorn DBeitz E2016Pentamidine is not a permeant but a nanomolar inhibitor of the Trypanosoma brucei Aquaglyceroporin-2PLOS Pathogens12e100543610.1371/journal.ppat.100543626828608Taladriz AHealy AFlores Pérez EJHerrero García VRíos Martínez CAlkhaldi AAEze AAKaiser Mde Koning HPChana ADardonville C2012Synthesis and structure-activity analysis of new phosphonium salts with potent activity against african trypanosomesJournal of Medicinal Chemistry552606262210.1021/jm201425922390399Teka IAKazibwe AJEl-Sabbagh NAl-Salabi MIWard CPEze AAMunday JCMäser PMatovu EBarrett MPde Koning HP2011The diamidine diminazene aceturate is a substrate for the high-affinity pentamidine transporter: implications for the development of high resistance levels in trypanosomesMolecular Pharmacology8011011610.1124/mol.111.07155521436312Unciti-Broceta JDArias JLMaceira JSoriano MOrtiz-González MHernández-Quero JMuñóz-Torres Mde Koning HPMagez SGarcia-Salcedo JA2015Specific cell targeting therapy bypasses drug resistance mechanisms in African trypanosomiasisPLOS Pathogens11e100494210.1371/journal.ppat.100494226110623Vanommeslaeghe KHatcher EAcharya CKundu SZhong SShim JDarian EGuvench OLopes PVorobyov IMackerell AD2010CHARMM general force field: a force field for drug‐like molecules compatible with the CHARMM all‐atom additive biological force fieldJournal of Computational Chemistry3167169010.1002/jcc.2136719575467Vincent IMCreek DWatson DGKamleh MAWoods DJWong PEBurchmore RJBarrett MP2010A molecular mechanism for eflornithine resistance in african trypanosomesPLOS Pathogens6e100120410.1371/journal.ppat.100120421124824Wallace LJCandlish DDe Koning HP2002Different substrate recognition motifs of human and trypanosome nucleobase transporters. selective uptake of purine antimetabolitesThe Journal of Biological Chemistry277261492615610.1074/jbc.M20283520012004061Wang MZZhu XSrivastava ALiu QSweat JMPandharkar TStephens CERiccio EParman TMunde MMandal SMadhubala RTidwell RRWilson WDBoykin DWHall JEKyle DEWerbovetz KA2010Novel arylimidamides for treatment of visceral leishmaniasisAntimicrobial Agents and Chemotherapy542507251610.1128/AAC.00250-1020368397Ward CPWong PEBurchmore RJde Koning HPBarrett MP2011Trypanocidal furamidine analogues: influence of pyridine nitrogens on trypanocidal activity, transport kinetics, and resistance patternsAntimicrobial Agents and Chemotherapy552352236110.1128/AAC.01551-1021402852Wree DWu BZeuthen TBeitz E2011Requirement for asparagine in the aquaporin NPA sequence signature motifs for cation exclusionFEBS Journal27874074810.1111/j.1742-4658.2010.07993.x21205205Wu BSteinbronn CAlsterfjord MZeuthen TBeitz E2009Concerted action of two cation filters in the aquaporin water channelThe EMBO Journal282188219410.1038/emboj.2009.18219574955Zhorov BSTikhonov DB2013Ligand action on sodium, potassium, and calcium channels: role of permeant ionsTrends in Pharmacological Sciences3415416110.1016/j.tips.2013.01.00223375737Zoltner MLeung KFAlsford SHorn DField MC2015Modulation of the surface proteome through multiple ubiquitylation pathways in african trypanosomesPLOS Pathogens11e100523610.1371/journal.ppat.100523626492041Zoltner MHorn Dde Koning HPField MC2016Exploiting the achilles' heel of membrane trafficking in trypanosomesCurrent Opinion in Microbiology349710310.1016/j.mib.2016.08.00527614711Zoltner MCampagnaro GDTaleva GBurrell ACerone MLeung KFAchcar FHorn DVaughan SGadelha CZíková ABarrett MPde Koning HPField MC2020Suramin exposure alters cellular metabolism and mitochondrial energy production in african trypanosomesJournal of Biological Chemistry2958331834710.1074/jbc.RA120.0123553235474210.7554/eLife.56416.sa1Decision letterClaytonChristineReviewing EditorDKFZ-ZMBH AllianceGermanyDiallinasGeorge ReviewerBeitzEric ReviewerChristian-Albrechts-University of KielGermany

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Thank you for submitting your article "Positively selected modifications in the pore of TbAQP2 allow pentamidine to enter Trypanosoma brucei" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Dominique Soldati-Favre as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: George Diallinas (Reviewer #2); Eric Beitz (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, when editors judge that a submitted work as a whole belongs in eLife but that some conclusions require a modest amount of additional new data, as they do with your paper, we are asking that the manuscript be revised to either limit claims to those supported by data in hand, or to explicitly state that the relevant conclusions require additional supporting data.

Our expectation is that the authors will eventually carry out the additional experiments and report on how they affect the relevant conclusions either in a preprint on bioRxiv or medRxiv, or if appropriate, as a Research Advance in eLife, either of which would be linked to the original paper.

Summary:

Mutations of Trypanosoma brucei aquaglyceroporin 2 (AQP2) are known to result in resistance to some anti-trpyanosomal drugs. Although it was originally assumed that AQP2 is a transporter for the compounds, it was recently suggested that instead, AQP2 acts as a receptor and the drugs are subsequently taken up by endocytosis. In this paper, the authors provide various lines of evidence that support the original, transporter, hypothesis.

Essential revisions:

The major revisions are to tone down the claims, since definitive proof is not provided. Both reviewers say this; you need to say it too and remove the emphatic adjectives (see reviewer 3). Since both reviewers are aquaglyyceroporin experts and the reviews contain quite a lot of insights they are included in full.

Reviewer #2:

Summary:

The present manuscript attempts to reject the 'receptor endocytosis' mechanism through the following logic. First, they perform a relative phylogenetic analysis of AQPs that shows that TbAQ2 orthologues have evolved from more canonical AQPs in a way that they possess non-canonical selectivity elements. They propose that these particular selectivity elements of TbAQ2 enlarge the pore size and allow access of pentamidine and other larger cations into the substrate trajectory. Given that the TbAQ3 paralogue, unlike TbAQ2, is not involved in pentamidine uptake, they then try to genetically interconvert the selectivity filters of TbAQ2 and TbAQ3, and test whether this is sufficient to confer gain or loss of sensitivity/resistance to pentamidine.

Their evidence supports that the exchanged selectivity elements does play a role in pentamidine uptake and sensitivity. These results are not discussed further in respect to some specificity changes observed in respect to different drugs (see my specific comment later).

Subsequently, they employ a rational mutational analysis of amino acids modelled to potentially bind pentamidine (and melarsoprol) or act as putative molecular barriers at the cytoplasmic end of the pore (i.e. inner filter or gate). They show that mutations in the putative pentamidine binding site dramatically reduce pentamidine transport and some of them also the sensitivity to the drug, and mutations introducing bulkier than the natural residues at the inner exit of pore also show much reduced rates of pentamidine transport and relative reduced pentamidine sensitivity. Noticeably, mutations reducing pentamidine transport had a minor effect on glycerol efflux, which seems to be the physiological function of TbAQ2. At this point they describe a number of subsequent experiments which, they claim, provide evidence against the involvement of endocytosis in pentamidine uptake (through the use of a strain partially blocked in endocytosis in blood stream forms) and supporting the need for a proton motive force to drive TbAQP2-mediated pentamidine uptake.

What follows is an in silico approach for 'docking' pentamidine in its proposed binding site in the modelled structure of TbAQ2, followed by Molecular Dynamics supporting that pentamidine can exit the channel in either direction. Finally, the authors perform an extensive SAR analysis of the pentamidine-AQP2 interactions, using a plethora of designed and chemically made pentamidine analogues or similar drugs, which shows that both amidine groups and one or both ether oxygens in pentamidine are involved in interactions with AQP2. This analysis, they claim, unambiguously demonstrates pentamidine binding occurs in an elongated orientation and is in complete agreement with the modelling and molecular dynamics, and the mutational analysis presented before.

Essential revisions:

- I will start form the end, saying that the basic conclusion of the authors (TbAQ2 is a specific aquaporin transporting pentamidine), is well justified from their phylogenetic and mutational analyses, modelling, MDs and SAR, as well as, the fact the TbAQ2 is a channel. I am quite convinced that pentamidine uptake is via bona fidae channel activity and not by endocytosis.

- On the other hand, their experiments aiming to show that endocytosis is not involved in pentamidine uptake (part 2) is very weak and not convincing. First of all, the strain used is only moderately blocked in endocytosis and the results they get (Figure 4) concerning the uptake of pentamidine versus suramin are statistically unconvincing.

- Most importantly, drawing conclusions on specific endocytosis of TbAQ2 based on the rate of pentamidine versus suramin uptake is conceptually false. Endocytosis is a cargo-dependent process, a notion that seems to escapes to many, so that different protein cargoes use different mechanism of endocytosis (e.g. clathrin-dependent or clathrin-independent, AP2-dependent or AP2-independent, polarized or non-polarized, cargo conformation-depended or not, etc.).

- If the authors wished to formally dismiss the idea or receptor-endocytosis they should have directly accessed microscopically and by westerns, using GFP-tagging, the internalization/turnover of TbAQP2 in the presence or absence of pentamidine and other drugs or analogues.

- Substrate- or transport activity-dependent endocytosis of transporter is a well-known phenomenon in fungi and seemingly also in plants and mammals (Gournas et al., 2010; Adv Exp Med Biol, 2016; Gormal et al., 2018). Multiple evidence suggests that upon binding of their substrates or ligands transporters undergo conformational changes related to the alternation from an outward to an inward conformation, but also related to the opening and closing of outer or inner gates. In some cases, the entire transport cycle is needed to be complete to elicit endocytosis, but in other cases an intermediate conformation (e.g. inward-facing, substrate occluded conformation) is sufficient to promote endocytosis (Gournas et al., 2010; Ghaddar et al., 2014; Gournas et al., 2017). The generally accepted idea is that a specific substrate/ligand-induced transient conformation make the transporter accessible to ubiquitination and thus to endocytosis. What is important to realize is that it is not possible to predict a priori whether substrate binding is sufficient to trigger endocytosis, or whether a full transport cycle in needed, or even cases where substrates do not cause endocytosis. Each transporter might behave differently in respect to endocytosis by substrates or ligands, particularly under distinct physiological, developmental or stress conditions, and endocytosis of a specific transporter should not be related with endocytosis of other cargoes or general endocytic activity, without targeted experiments.

- If the authors have performed the relative endocytosis studies on TbAQ2 they would have known whether pentamidine or other drugs elicit endocytosis. In case pentamidine elicits endocytosis this can still be unrelated to the mechanism of pentamidine uptake, as several transporters are endocytosed by their substrates. If a TbAQ2 endocytosis serves accumulation of pentamidine this would, in principle, be a fast process. In contrast, in cases endocytosis is the result a feedback mechanism for controlling excess substrate accumulation, this usually is slower becoming visible after a significant time period of transport activity. Moreover, in case pentamidine does not elicit internalization of TbAQ2, then they could easily dismiss endocytosis as the mechanism of uptake. I also suggest to the authors that if they wish to do these experiments to use the procyclic forms of T. brucei, if possible, where TbAQ2 is localized all along the PM and it will thus be easier to detect and quantify endocytosis microscopically and relate it to transport activity measurements., In any case, I propose to remove (or transfer to supplementary material) their experiments describe in part 2. In fact, the authors give convincing explanations in their discussion why it is difficult to reconcile the literature on pentamidine transport/resistance with uptake via endocytosis (Discussion section).

- I also propose to the authors to remove the experiments on the role of the proton motive force in AQP2-mediated pentamidine uptake, as CCCP has both pleiotropic and specific effects on TbAQ2, difficult to put under a rigorous conclusion

- An information I could not find in the present manuscript or in Song et al., (2016) is whether pentamidine can inhibit competitively (or non-competitively) TbAQ2-mediated H3-glycerol uptake/efflux. I think this is a very critical experiment, which together with a direct assessment of endocytosis of TbAQ2 and TbAQ3 (as a control) by pentamidine, will be critical in formally showing that TbAQ2 functions as a channel rather that a receptor for the uptake of pentamidine.

In order to be able to have an opinion on the alternative hypothesis of receptor endocytosis-mediated pentamidine uptake, I also read the Song et al., (2016) article. The evidence that TbAQP2 does not transport pentamidine when expressed in yeast is indirect through growth tests and is not convincing (they have not tested directly uptake of radiolabeled pentamidine). In addition, there are no data showing that TbAQ2 is properly sorted to the PM of yeast (the western shown lacks controls). This would need functional tagging of TbAQ2 with GFP. Moreover, expression in yeast might not reconstitute full activity and physiological kinetics. The aforementioned problems are very well known bottlenecks in efforts to express transporters in heterologous systems. Furthermore, a main criterion used by Song et al. is that "if a solute is excluded (e.g. formamidine) from passing, then a much larger (pentamidine) would be excluded as well". This is surely a misconception, well known to people working on the molecular basis of transporter specificity. Finally, Song et al., argues that since in BSF TbAQ2 localizes in the trypanosome flagellar pocket, which is a site of very efficient exocytosis and endocytosis, then this is compatible with the idea of endocytosis-mediated pentamidine uptake. As discussed before general endocytosis cannot be related with endocytosis if a specific transporter without relative experimental evidence. Overall, the idea of Song et al., that pentamidine-induced endocytosis of TbAQP2 acts as vehicle for uptake seems to me purely speculative and rather unlikely.

The phylogenetic and mutational analyses, modelling, MDs and SAR approaches support the most logical scenario, that pentamidine is transported via a channel and not via receptor endocytosis. AQP2 is a channel and I find no rigorous evidence by Song et al., (2016) that it can function as receptor, nor that it is endocytosed in the presence of pentamidine. Furthermore, the criterion of size in respect to specificity in transport via a channel or transporter does not hold true. Specificity cannot be predicted a priori as it depends not only on elements of substrate binding site, but also on major conformational changes and the dynamics of gating, all triggered by substrate binding and release. Several examples of 'surprising' specificities exist in the literature. For example, specific bacterial, fungal or plant NAT transporters can translocate with very high affinity xanthine and uric acid, but not hypoxanthine, which is a very similar solute of smaller in size (Gournas et al., 2008). Also, a fungal NAT transporter specific for xanthine/uric acid can transport with extremely high-affinity the drug allopurinol (a hypoxanthine analogue), but surprisingly, transport of allopurinol does not follow Michelis--Menten kinetics (Diallinas and Biochimie, 2013). Finally, NCS1 or APC transporters with practically identical substrate binding site residues transport distinct substrates via the involvement of distinct selective outer an inner gates (Krypotou et al., 2015 Sioupouli et al., 2017; Gournas, 2016).

The authors also give convincing explanations why is difficult to reconcile the literature on pentamidine transport/resistance with uptake via endocytosis in the Discussion section.

- However, if they wanted to formally reject the scenario of endocytosis they should have directly accessed microscopically and by westerns, using GFP-tagging, the internalization/turnover of AQP2 in tricyclics, in the presence or absence of pentamidine and other drugs or analogues, where the channel is evenly localized in the PM and is more active.

The present work can be published in eLife after a major revision addressing some of the above points.

Reviewer #3:

The presented multi-author study evaluates experimental data and arguments as a basis to decide whether uptake of the anti-trypanosomal agent pentamidine occurs by passing the channel of an aquaporin isoform, TbAQP2, or by endocytosis of a TbAQP2-pentamidine complex. To this end, the authors generated a number of point mutations of TbAQP2 and of a second isoform, TbAQP3; assayed the effect of ionophores; partially blocked endocytosis by a knock-down approach; simulated docking and molecular dynamics; and synthesized numerous pentamidine-related compounds for structure-activity relationships. Quantitative data derive from radiolabeled pentamidine, effective concentrations (EC50) on parasite growth, and binding competition experiments (K_i), or were computed from the simulations. A tremendous amount of work went into the study.

- The major hurdle to ultimately answer the question seems to be the extremely low rate at which pentamidine is transported into the cells. The authors provide a number in Figure 4B, i.e. about 6 fmols per 10^7 cells per second, which translates into about 23 molecules per cell per second.

- Later, in the Discussion section, the authors state a number from another reference of 9.5 x 10^5 per cell per hour, which is about one order of magnitude higher (about 260 molecules per cell per second). If one guesses the number of aquaporin proteins in a cell to be 1000, which is probably too low as erythrocytes were estimated to contain 200,000 copies, this would mean that the uptake of one pentamidine molecule takes about 50 seconds (or even longer if more than 1000 aquaporins are present).

- There are several references in the manuscript to a study with yeast-expressed TbAQP2 showing inhibition of glycerol permeability by pentamidine (Song et al.). This setup would not pick up such slow transport rates, and, consequently, pentamidine would act as an inhibitor. This would also be true for other cations tested in the yeast. Water permeability of aquaporins is in the 1-nanosecond range and glycerol is slower by two orders of magnitude; hence, there is still a gap of 8 orders of magnitude compared to pentamidine uptake rates into trypanosomes. Another consequence of the long residence time of pentamidine at the aquaporin is that both processes, i.e. membrane-potential driven leakage of pentamidine through the aquaporin and internalisation of the complex, may jointly contribute to uptake (taking into account the numbers obtained for suramin endocytosis in Figure 4C).

1) Results section: in several instances, small differences in the relative uptake numbers are observed for the various TbAQP2 mutants, but potential differences in expression levels or available copies at the plasma membrane are not considered. Measured effects are not necessarily directly connected to the mutation.

2) Subsection “1.4 Mutations of amino acids modelled to potentially bind pentamidine or melarsoprol dramatically reduce pentamidine transport”, and Figure 2A: Presence and absence of significance in the assays using TbAQP2 I190T and TbAQP2 I190T/W192G do not derive from differences in their activity (equal numbers) but from the background level that varies between 0.4% and 1%. This variation lets one wonder if the indicated significance in Figure 1G is true (background of 0.5%; at 1% the stated difference would not be significant).

- Sounds like there is an issue here. Maybe modify conclusion.

3) Subsection “5.5. Non-diamidine trypanocides”, and Figure 9—figure supplement 1: Calculating a resistance factor appears reasonable to estimate the proportion of inhibitor in the cytosol. However, it is questionable whether the comparison of resistance factors is meaningful if the EC50 values differ largely between different compounds. This alone may well explain the lack of correlation.

Therefore, please add the absolute values EC50_TbAQP2/3null and EC50_TbAQP2wt to the table in Figure 9—figure supplement 1. The data do not seem to allow one to draw conclusions on the transport-vs-endocytosis question, see e.g. compounds R14 and R52. R14 exhibits almost equal affinity to TbAQP2, binding of R52 is 200-fold weaker, yet R52 similarly depends on the presence of TbAQP2 as pentamidine, whereas R14 is equally effective with or without TbAQP2. This does neither support transport nor endocytosis.

4) General: the authors quite often use strong wording ("unequivocally", "greatly" etc.) to make their case. This should be toned down a bit as all presented evidence on pentamidine transport is indirect. This is not a shortcoming of the study but inherent to the problem. It would be desirable to have direct, unequivocal transport assays for such very slow events in the future.

10.7554/eLife.56416.sa2Author response

Reviewer #2:

Essential revisions:

- I will start form the end, saying that the basic conclusion of the authors (TbAQ2 is a specific aquaporin transporting pentamidine), is well justified from their phylogenetic and mutational analyses, modelling, MDs and SAR, as well as, the fact the TbAQ2 is a channel. I am quite convinced that pentamidine uptake is via bona fidae channel activity and not by endocytosis.

- On the other hand, their experiments aiming to show that endocytosis is not involved in pentamidine uptake (part 2) is very weak and not convincing. First of all, the strain used is only moderately blocked in endocytosis and the results they get (Figure 4) concerning the uptake of pentamidine versus suramin are statistically unconvincing.

We certainly acknowledge that this experiment is no definitive proof and is not as statistically strong as we would have liked. The problem is that a more complete disabling/inhibition of endocytosis in Trypanosoma brucei bloodstream forms is very quickly lethal. The rate of exocytosis remains very high, leading to a rapid enlargement of the flagellar pocket (FP), which is the only area of the cell where endocytosis and exocytosis takes place and can take place. This in turn leads to gross distortions of the cell shape, and many cellular pathologies, which would invalidate any conclusions drawn from the experiment. We thus had no choice but to go for as early a time point after RNAi induction as allows any observation of diminished endocytosis. This is inherent to the organism being studied. Regardless, we believe the experiment is valid as far as it goes: an indication that endocytosis from the FP is inhibited (3H-suramin control, P= 0.019) but pentamidine uptake is not. The level of reduction for suramin uptake is consistent with the published effects of CRK12 RNAi at 12 h in the reference given (Monnerat et al., 2013), and the reason the 12 h point was chosen. Still, we have moved the figure to the Supplemental data and revised any statements in the Discussion section that give too much weight to the statement.

- Most importantly, drawing conclusions on specific endocytosis of TbAQ2 based on the rate of pentamidine versus suramin uptake is conceptually false. Endocytosis is a cargo-dependent process, a notion that seems to escapes to many, so that different protein cargoes use different mechanism of endocytosis (e.g. clathrin-dependent or clathrin-independent, AP2-dependent or AP2-independent, polarized or non-polarized, cargo conformation-depended or not, etc.).

There is quite solid evidence that there is no AP-2 dependent endocytosis in trypanosomes, and that endocytosis is clathrin-dependent (Allen et al., 2003; Morgan et al., 2002, 2003). Profs. Mark Field and Mark Carrington are the leading experts on this and co-authors of the current submission. While we agree with the reviewer that we cannot formally rule out that the receptor for suramin (ISG75) and TbAQP2 are endocytosed via a similar mechanism, there is nothing in the extensive literature on this subject that suggests multiple, independent-acting endocytotic mechanisms being employed by T. brucei bloodstream forms. The suramin receptor ISG75 and TbAQP2 are both ubiquitinated and directed to the lysosomes (Zoltner et al., 2015; Quintana et al., 2020).

- Substrate- or transport activity-dependent endocytosis of transporter is a well-known phenomenon in fungi and seemingly also in plants and mammals (Gournas et al., 2010; Adv Exp Med Biol, 2016; Gormal et al., 2018). […] Each transporter might behave differently in respect to endocytosis by substrates or ligands, particularly under distinct physiological, developmental or stress conditions, and endocytosis of a specific transporter should be related with endocytosis of other cargoes or general endocytic activity, without targeted experiments.

- If the authors have performed the relative endocytosis studies on TbAQ2 they would have known whether pentamidine or other drugs elicit endocytosis. […] In fact, the authors give convincing explanations in their discussion why it is difficult to reconcile the literature on pentamidine transport/resistance with uptake via endocytosis (Discussion section).

We genuinely thank the reviewer for his expert comments and insights. One reason we did not include direct endocytosis of aquaporin experiments in our manuscript was that this was being pursued in parallel, in collaboration, by Prof. Mark Field and his team, and constitutes a separate and challenging project, now accepted for publication (Quintana et al., 2020). Prof. Field is the acknowledged expert on trafficking of membrane proteins in trypanosomes. There are two reasons the suggestion of the reviewer to use procyclic cells for such experiments was not done. First, the relevant system to look at drug resistance mechanisms is the clinical form of T. brucei, the bloodstream form. More important from a cell biology point of view, is that, while the TbAQP2 protein is indeed spread out over the plasma membrane of procyclic T. brucei, it can only be internalised from the flagellar pocket, as stated in the Discussion section: “Moreover, in procyclic cells TbAQP2 is spread out over the cell surface (Baker et al., 2012) but endocytosis happens exclusively in the flagellar pocket (Field and Carrington, 2009) (which is 3-fold smaller in procyclic than in bloodstream forms (Demmel et al., 2014)), as the pellicular microtubule networks below the plasma membrane prevent endocytosis (Zoltner et al., 2016).” This makes it actually a less effective form to study TbAQP2 endocytosis, as only a small proportion of the protein is available for endocytosis. But we make the point that, despite most of procyclic TbAQP2 being unavailable for endocytosis, the rate of HAPT1/AQP2-mediated 3H-pentamidine uptake is ~10-fold higher in the procyclic form: “..the rate of endocytosis in bloodstream trypanosomes is much higher than in the procyclic lifecycle forms (Langreth and Balber, 1975; Zoltner et al., 2016), yet the rate of HAPT-mediated [3H]-pentamidine uptake in procyclics is ~10-fold higher than in bloodstream forms (De Koning, 2001a; Teka et al., 2011), despite the level of TbAQP2 expression being similar in both cases (Siegel et al., 2010; Jensen et al., 2014).”

Importantly, the paper on TbAQP2 endocytosis from the Mark Field laboratory has now been accepted for publication (Quintana et al., 2020). This paper shows that TbAQP2, like ISG75 (the suramin receptor), is ubiquitylated and that this is essential for its stability and for internalisation and trafficking – making it even more likely that suramin and pentamidine are endocytosed using similar mechanisms. The paper also showed ISG75 and TbAQP2 co-localisation by immunocytochemistry. Another noteworthy discovery reported in that paper is that TbAQP2 exists in the trypanosome as a tetramer-of-tetramers, making it highly unlikely that the AQP2 protein is able to undergo substantial conformational changes – nor are these known from aquaporins in general. Finally, the paper shows that TbAQP2 has a half-life time >4 hours, which seems completely incompatible with receptor-mediated endocytosis of pentamidine when the uptake must be measured in a scale of seconds (as opposed to [3H]-suramin, 15 minutes to get a reliable signal).

The reviewer asks us to include an assessment of the effect of pentamidine on the rate of turnover of TbAQP2. With the Quintana et al., paper now in press and available online we can include this experiment, which had already been performed, as new Figure 4. The figure shows the rate of turnover of AQP2, as quantified by Western blot after cycloheximide pre-treatment. The original blots are included in the supplemental materials. At the concentration of pentamidine used, close to the Km and 5× the EC50 value, the drug had no effect at all on the turnover rate of TbAQP2. This information was not included in the Quintana et al., manuscript, and because of this contribution Juan Quintana and Mark Field have been added to the author list.

In conclusion, we believe that, in the revised version, there is very substantial though arguably not conclusive proof arguing against a link between TbAQP2 endocytosis and pentamidine uptake. We have, as requested, qualified the conclusions to reflect that these various observations individually should not be considered as definitive proof. However, we do believe that the totality of evidence and the number of unrelated, independent approaches together, do make a very compelling argument that pentamidine is taken up through the pore rather than via endocytosis.

- I also propose to the authors to remove the experiments on the role of the proton motive force in AQP2-mediated pentamidine uptake, as CCCP has both pleiotropic and specific effects on TbAQ2, difficult to put under a rigorous conclusion

Again, we agree with the reviewer that these results on their own do not constitute complete proof of either model. However, the results with the ionophores nigericin and gramicidin do establish that the plasma membrane potential Vm is essential to drive the pentamidine into the cell, in close agreement with the molecular dynamics modelling, our previous results with [3H]-pentamidine transport obtained in procyclic T. brucei (De Koning, 2001), and the observation that knockdown of plasma membrane proton ATPases leads to reduced sensitivity to cationic pentamidine but not to neutral arsenicals (Alsford et al., 2012; Baker et al., 2013). Even without CCCP there is a clear correlation between [3H]-pentamidine uptake rate and the protonmotive force (PMF) (Figure 5C). As to CCCP we realised and acknowledged the pleiotropic effects of direct inhibition of TbAQP2 and we believe that this is important to note, but the correlation with PMF is not dependent on the inclusion of CCCP in the dataset, neither in this paper nor in the previous paper on pentamidine uptake in procyclic cells (De Koning, 2001). Given the importance of the ionophore results for the molecular dynamics, and for the interpretation of the published observations regarding the HA1-3 proton ATPases, we believe the experiments with ionophores are important for the details (“loose ends”, if you will) of pentamidine permeation into trypanosomes, rather than for the principal evaluation of permeation model, and we would prefer to keep them in the manuscript.

- An information I could not find in the present manuscript or in Song et al., (2016) is whether pentamidine can inhibit competitively (or non-competitively) TbAQ2-mediated H3-glycerol uptake/efflux. I think this is a very critical experiment, which together with a direct assessment of endocytosis of TbAQ2 and TbAQ3 (as a control) by pentamidine, will be critical in formally showing that TbAQ2 functions as a channel rather that a receptor for the uptake of pentamidine.

The information on pentamidine inhibition of [3H]-glycerol uptake is in fact included in the manuscript. Figure 5 frames D and E compare side-by-side the effects of pentamidine (and CCCP) on the uptake of [3H]-glycerol and [3H]-pentamidine, showing, among other things, that pentamidine does inhibit glycerol flux though TbAQP2, with an EC50 value (27.5 nM, average of 2 exp.; the CCCP EC50 was n=3 but only two pentamidine EC50 values were obtained and as such the value was not included in the manuscript) that is similar to that of its inhibition of [3H]-pentamidine uptake. This had not been explicitly drawn attention to in the text, hence it was overlooked; this has now been remedied. But I am not sure how the observation distinguishes between the channel and endocytosis models, which is why we did not use it in the Discussion section of the model. Incidentally, up to 1 mM glycerol does not inhibit [3H]-pentamidine uptake, which we attribute to the vast difference in binding affinity and the fleeting association of glycerol with the channel-lining amino acids.

In order to be able to have an opinion on the alternative hypothesis of receptor endocytosis-mediated pentamidine uptake, I also read the Song et al., (2016) article. […] Overall, the idea of Song et al., that pentamidine-induced endocytosis of TbAQP2 acts as vehicle for uptake seems to me purely speculative and rather unlikely.

We thank the reviewer for his comments. We have made similar points in the Discussion section, particularly concerning the notion of size exclusion.

The phylogenetic and mutational analyses, modelling, MDs and SAR approaches support the most logical scenario, that pentamidine is transported via a channel and not via receptor endocytosis. […] Finally, NCS1 or APC transporters with practically identical substrate binding site residues transport distinct substrates via the involvement of distinct selective outer an inner gates (Krypotou et al., 2015 Sioupouli et al., 2017; Gournas, 2016).

The authors also give convincing explanations why is difficult to reconcile the literature on pentamidine transport/resistance with uptake via endocytosis in the Discussion section.

- However, if they wanted to formally reject the scenario of endocytosis they should have directly accessed microscopically and by westerns, using GFP-tagging, the internalization/turnover of AQP2 in tricyclics, in the presence or absence of pentamidine and other drugs or analogues, where the channel is evenly localized in the PM and is more active.

The present work can be published in eLife after a major revision addressing some of the above points.

We thank the reviewer for his considerable input and constructive, fair criticisms. We have certainly taken the revision very seriously and believe we have been able to accommodate the reviewer’s directions.

Reviewer #3:

- The major hurdle to ultimately answer the question seems to be the extremely low rate at which pentamidine is transported into the cells. The authors provide a number in Figure 4B, i.e. about 6 fmols per 10^7 cells per second, which translates into about 23 molecules per cell per second.

- Later, in the Discussion section, the authors state a number from another reference of 9.5 x 10^5 per cell per hour, which is about one order of magnitude higher (about 260 molecules per cell per second).

The number in Figure 4B was obtained with the CRK12 RNAi line obtained from Dr Tansy Hammarton in our Institute, whereas the published Vmax value was obtained in 2001 using our own clonal line of Tbb Lister s427. More to the point, the Vmax value is of course the maximum rate at saturation concentration of [3H]-pentamidine, whereas the rate of uptake in Figure 4B was obtained with a label concentration of just 25 nM, i.e. a concentration below the Km, as is good practice. Therefore, it is to be expected that this rate is considerably lower than the Vmax.

If one guesses the number of aquaporin proteins in a cell to be 1000, which is probably too low as erythrocytes were estimated to contain 200,000 copies, this would mean that the uptake of one pentamidine molecule takes about 50 seconds (or even longer if more than 1000 aquaporins are present).

I am not sure that I follow this calculation or can discern how it is relevant. Yes, the rate of uptake can be expressed as 265 molecules per cell per s. Thus, if 1000 AQP units are present at the cell surface at any given time it seems to me that each takes up a pentamidine every 3.8 (i.e, 265/1000) seconds, not 50. All this means is that, the small substrates such as water and glycerol, that use TbAQP2 as a channel, have a much faster permeation rate, which is entirely as expected. The difference in substrate affinity is phenomenal, with the Km of pentamidine as low as ~35 nM, compared to an affinity for glycerol that is no-doubt in the millimolar range, so it is expected that the off-rate for pentamidine is very much slower than for glycerol. And yes, that means that pentamidine does occlude the channel for what is a relatively long period, thereby inhibiting glycerol passage as we show, yet it also does penetrate into the cell, at a sufficient rate to have a pharmacological effect.

For the endocytosis route, of course, the rate of uptake would be even much slower. As stated in the Discussion section, “given a 1:1 stoichiometry for AQP2:pentamidine the endocytosis model would require the internalisation and recycling of as many units of TbAQP2 and this seems unlikely”. If we use the t1/2 for TbAQP2 of Quitana et al., (2020) of 4 h, the uptake of 265 molecules of pentamidine per second would require a presence of 3.8×106 copies of TbAQP2 on the cell surface, all in the flagellar pocket of a single cell – an impossibility in my view [9.54×105 molecules per cell per h transported × 4 h]. And again, in procyclic cells the Vmax is ~10-fold higher despite a much lower endocytosis rate. These further considerations have been added to the Discussion section.

- There are several references in the manuscript to a study with yeast-expressed TbAQP2 showing inhibition of glycerol permeability by pentamidine (Song et al.). This setup would not pick up such slow transport rates, and, consequently, pentamidine would act as an inhibitor. This would also be true for other cations tested in the yeast. Water permeability of aquaporins is in the 1-nanosecond range and glycerol is slower by two orders of magnitude; hence, there is still a gap of 8 orders of magnitude compared to pentamidine uptake rates into trypanosomes. Another consequence of the long residence time of pentamidine at the aquaporin is that both processes, i.e. membrane-potential driven leakage of pentamidine through the aquaporin and internalisation of the complex, may jointly contribute to uptake (taking into account the numbers obtained for suramin endocytosis in Figure 4C).

These issues are partly already addressed in the previous point. Again I am not sure where the reviewer is going with his comment on the yeast-expressed TbAQP2, or how it is relevant to the interpretation of our results presented in the current manuscript, using [3H]-pentamidine as direct measurement, in the original cells. We are happy to concede that the total accumulation of pentamidine is a combination of direct permeation and endocytosis, it was never denied, and has been further added explicitly to the Discussion section. However, we remain convinced, based on the evidence and reasoning presented that by far the larger of these two is permeation through the channel: “Although it is likely that AQP2-bound pentamidine is internalised as part of the natural turnover rate of the protein, this is not likely to contribute very significantly to the overall rate of uptake of this drug.”

1) Results section: in several instances, small differences in the relative uptake numbers are observed for the various TbAQP2 mutants, but potential differences in expression levels or available copies at the plasma membrane are not considered. Measured effects are not necessarily directly connected to the mutation.

This is certainly true, and just measuring expression levels by qRT-PCR, or Northern blot would not necessarily give a quantification of actual TbAQP2, nor establish whether that protein is functionally present in the flagellar pocket. Moreover, The constructs were N-terminally GFP-tagged but Western blots and immunofluorescence microscopy are not quantitative enough to account for small changes in sensitivity and could also not be used as controls; Western blots would also not provide information about the cellular localisation of the protein. Incidentally, Baker et al., (2012) already investigated carefully whether N-terminal tagging of TbAQP2 and TbAQP3 affects their functionality with respect to pentamidine and melarsoprol sensitivity and showed it does not.

Thus, such controls could properly only be functional, which we did by showing functional expression quantified by the EC50 of TAO inhibitor SHAM and the uptake rate of [3H]-glycerol. While further limitations of that approach can also be identified, this really is the best way this could have been done. Moreover, for most of the mutants the changes in transport were not “small differences”, and gave a very consistent set of results, including the reciprocity of the swap of ar/R amino acids between TbAQP2 and TbAQP3, which lead to a gain of pentamidine transport function for the altered TbAQP3.

2) Subsection “1.4 Mutations of amino acids modelled to potentially bind pentamidine or melarsoprol dramatically reduce pentamidine transport”, and Figure 2A: Presence and absence of significance in the assays using TbAQP2 I190T and TbAQP2 I190T/W192G do not derive from differences in their activity (equal numbers) but from the background level that varies between 0.4% and 1%. This variation lets one wonder if the indicated significance in Figure 1G is true (background of 0.5%; at 1% the stated difference would not be significant).

- Sounds like there is an issue here. Maybe modify conclusion.

Yes, the level of pentamidine uptake in the TbAQP2I190T and TbAQP2I190T/W192G mutants is very low, approximately 2.5% of the AQP2-WT control in both cases, as depicted in Figure 2A. It is also agreed on that because of the difference in the aqp2/aqp3 null controls performed in parallel with each experiment, and thus separately for the two mutants, happens to result in the P>0.05 threshold difference from the null control being reached for TbAQP2I190T but not for TbAQP2I190T/W192G (P=0.18, n=4, unpaired t-test), all as indicated in the Figure. The issue here, in any case, is not whether pentamidine transport by the mutants was significantly different from the null background, but from the WT AQP2 control, which it was, again as indicated. As to whether the acknowledged inter-assay variation of pentamidine accumulation in the null mutant, when expressed as percent of WT control, between 0.5 and 1.0%, this is precisely why it is essential to perform the same experiment with the null control in parallel, with cultures grown in parallel to the same density and using the same solutions, buffers and radiolabel cocktail on the day, every single time; the same goes for the WT control. As such, having done so, the data presented in Figure 1G can be taken as genuine. In three independent experiments, each in triplicate) the [3H]-pentamidine transport rate of the triple mutant was approximately double the rate of the null control or single mutant, resulting in P<0.01 in an unpaired test whether percentage of control or pmol/107 cells/s-1 was used. It is correct that the greatest care is required when interpreting small differences like this but we believe we have done so to the maximum extent possible.

3) Subsection “5.5. Non-diamidine trypanocides”, and Figure 9—figure supplement 1: Calculating a resistance factor appears reasonable to estimate the proportion of inhibitor in the cytosol. However, it is questionable whether the comparison of resistance factors is meaningful if the EC50 values differ largely between different compounds. This alone may well explain the lack of correlation.

Therefore, please add the absolute values EC50_TbAQP2/3null and EC50_TbAQP2wt to the table in Figure 9—figure supplement 1. The data do not seem to allow one to draw conclusions on the transport-vs-endocytosis question, see e.g. compounds R14 and R52. R14 exhibits almost equal affinity to TbAQP2, binding of R52 is 200-fold weaker, yet R52 similarly depends on the presence of TbAQP2 as pentamidine, whereas R14 is equally effective with or without TbAQP2. This does neither support transport nor endocytosis.

The requested data has been added to as an extended Table to Figure 9—figure supplement 1. The reviewer will see that the majority of the compounds displayed sub-μM activity against T. brucei WT (2T1 strain) although there were some with lower activity. These were included as the HAPT Ki values were available, determined because the structure was of interest for the SAR analysis. However, I am not sure what the reviewer means by “it is questionable whether the comparison of resistance factors is meaningful if the EC50 values differ largely between different compounds. This alone may well explain the lack of correlation.” We believe that for a correlation plot such as Figure 9—figure supplement 1 it is in fact necessary to cover a range of values on both axes: the more diverse the dataset is, the better. and there is a substantial range and diversity in both of the parameters plotted against each other. I am genuinely at a loss to answer the query. However, we have already adapted the text around this figure in response to reviewer 2 (see above) and hope that the reviewer will be satisfied with the new phrasing.

4) General: the authors quite often use strong wording ("unequivocally", "greatly" etc.) to make their case. This should be toned down a bit as all presented evidence on pentamidine transport is indirect. This is not a shortcoming of the study but inherent to the problem. It would be desirable to have direct, unequivocal transport assays for such very slow events in the future.

We have done as requested and removed / toned done the adjectives. Nonetheless, we stand by our assertion that, taken all together, we provide strong evidence through many independent methodologies, for the model that pentamidine does pass through the channel of TbAQP2. As to the desirability of measuring ‘such slow events’, we believe we have done exactly that, using custom-made radiolabelled pentamidine of the highest specific activity that was possible. Our method of measuring pentamidine transport has been very accurate and highly reproducible.