Aquaglyceroporin 2 controls susceptibility to melarsoprol and pentamidine in African trypanosomes

T. brucei AQP2 Has an Unusual Selectivity Filter.

The T. brucei locus recently linked to MPXR encodes the closely related proteins, AQP2 and AQP3 (5). These proteins are also related to the human aquaglyceroporins, AQPs 3, 7, 9, and 10 (22, 24). The structural determinants for gating these channels are the focus of intense interest, and because the selectivity of channels is central to their function, we examined the conserved residues that constitute the AQP2 and AQP3 “selectivity filter” (Fig. 1A). AQPs have six membrane-spanning α-helices and two half helices that fold into the center of the channel, with the termini located on the cytoplasmic face of the membrane. The selectivity filter is typically defined by two prominent constrictions in the channel, one formed by the highly conserved “NPA” motifs within the half helices, and the other, usually narrower, formed by an “aromatic arginine” (ar/R) motif (27–29). Schematic representations of AQP2 and AQP3 indicate the location of the selectivity filter residues relative to the predicted membrane-spanning domains (Fig. 1A). Whereas AQP3 encodes a conventional selectivity filter, as does AQP1, AQP2 lacks the NPA and ar/R motifs. Homology models (Fig. 1B) indicate how the unusual selectivity filter of AQP2 likely impacts the lining of the channel. An analysis of more than 1,000 MIPs from 340 organisms in the MIPModDB database (30) revealed T. brucei AQP2 as the only MIP with NSA/NPS or IVLL motifs; 78% contain NPA/NPA and 89% contain the arginine of the ar/R motif. Thus, AQP2 has a unique selectivity filter.

aqp2 Null Cells Are Melarsoprol and Pentamidine Resistant.

Cells lacking both AQP2 and AQP3 were resistant to melarsoprol and pentamidine (5). Because the analysis above indicated major differences between AQP2 and AQP3 that may be of functional significance, we wanted to determine what contribution each protein makes to the drug susceptibility phenotype. We used a gene knockout approach to test our hypothesis that AQP2 is specifically required for drug susceptibility. An aqp2 null strain was generated by sequential transfection with constructs that replace each AQP2 allele with selectable marker genes, deleting the entire AQP2 protein coding sequence but leaving the AQP3 gene intact (Fig. 1C). AQP2 knockout was confirmed by Southern blotting (Fig. 1C). Neither aqp2 null cells nor aqp2/aqp3 null cells displayed any detectable growth defect relative to wild-type cells. Drug susceptibility analysis revealed that aqp2 null cells were resistant to both melarsoprol and pentamidine (Fig. 1D); the 50% effective growth-inhibitory concentration (EC₅₀) was increased almost 2-fold relative to wild-type cells for melarsoprol and by more than 15-fold for pentamidine. Thus, aqp2 null cells are indistinguishable from aqp2/aqp3 null cells (5) in terms of EC₅₀ for either of these drugs.

Differentiation of African trypanosomes from the bloodstream stage to the insect “procyclic” stage, during the life cycle, involves major changes in surface architecture and energy metabolism (31). To probe the function of AQP2 during and/or following this developmental transition, we differentiated aqp2 and aqp2/aqp3 null strains; triggered in vitro by reduced temperature and by the addition of citric acid cycle intermediates to the culture medium. We saw no evidence for a defect in the process of differentiation in either aqp2 null or aqp2/aqp3 null cells. Dose–response curves revealed that the impact of AQP2 on drug sensitivity was similar in both stages. In aqp2 null insect-stage cells, the EC₅₀ for melarsoprol was increased by approximately 3-fold compared with wild-type cells, whereas the EC₅₀ for pentamidine was increased by more than 30-fold (Fig. 1E) and we obtained similar results for aqp2/aqp3 null insect-stage cells. We conclude that AQP2 is constitutively expressed and renders both major life-cycle stages drug sensitive. However, neither AQP2 nor AQP3 is required for viability of the bloodstream or procyclic stage or for differentiation in vitro.

In bloodstream-form cells, the resistance phenotype was specific for pentamidine and melaminophenyl arsenicals, with little or no effect on sensitivity to the veterinary diamidine, diminazene aceturate (Berenil), or the phenanthridine trypanocides, isometamidium and ethidium (Fig. 1F). In addition, wild-type and aqp null cells were equally sensitive to the lipophilic arsenical, phenylarsine oxide, which is known to diffuse across membranes (18). Thus, the deletion of AQP2 renders T. brucei MPXR but resistance does not extend to all diamidines or arsenicals.

AQP2 Restores Drug Sensitivity to aqp2/aqp3 Null Cells.

The results above suggest that AQP2 plays an important role in rendering T. brucei sensitive to melarsoprol and pentamidine. To determine whether AQP3 is required for this effect, we reintroduced an inducible copy of AQP2 into aqp2/aqp3 null cells. In the absence of AQP2 induction, these strains displayed a similar drug sensitivity phenotype to aqp2 or aqp2/aqp3 null cells (compare Fig. 1F with Fig. 2 A and B). In striking contrast, induction of AQP2 expression restored sensitivity to both melarsoprol (Fig. 2A) and pentamidine (Fig. 2B). Importantly, drug sensitivity is restored in the absence of AQP3, indicating that AQP3 expression is not required for melarsoprol and pentamidine sensitivity. A similar experiment, but involving the reintroduction of a copy of AQP3, failed to show any impact on MPXR (Fig. 2 A and B), confirming that AQP3 expression does not confer sensitivity to melarsoprol or pentamidine. AQP2 also restored drug sensitivity in an aqp2 null strain with an intact AQP3 gene. We conclude that, whereas AQP2 is required, AQP3 is neither sufficient nor required for drug sensitivity.

We also used the approach described above to reintroduce AQPs with green fluorescent protein (GFP) fused to the N termini. This approach allowed us to confirm inducible expression of ^GFPAQP2 and ^GFPAQP3 by Western blotting (Fig. 2C) and also to show that both ^GFPAQP2 and ^GFPAQP3 partition predominantly into the membrane fraction, as expected (Fig. 2D). ^GFPAQP2 fully restored drug sensitivity in both aqp2 and aqp2/aqp3 null strains, whereas ^GFPAQP3 failed to restore drug sensitivity in either strain (Fig. 2 E and F), confirming the results above and indicating that AQP2 remains functional when fused to GFP.

AQP2 Is Disrupted in a Pentamidine-Resistant Strain.

We next asked whether exposure to pentamidine could select for mutations that disrupt AQP2 function, thereby producing drug-resistant strains. To address this question, the AQP2 gene was sequenced in the well-characterized KO-B48 strain, which lacks HAPT1 activity, displays reduced pentamidine uptake, and is MPXR (18). Strikingly, an AQP2/AQP3 chimeric gene was found in place of AQP2 (Fig. 2G); a 272-nucleotide segment was replaced by an AQP3-derived sequence that alters four of the predicted residues (underlined) of the selectivity filter described above, from NSA/NPS/IVLL in AQP2 to NSA/NPA/IGYR in the chimera (Fig. S1).

AQP2 Is Restricted to the Flagellar Pocket Specifically in Bloodstream-Form Cells.

AQP3 is known to be dispersed over the plasma membrane (26) but AQP2 localization has not been determined. We used the strains expressing GFP-tagged AQPs described above to determine AQP subcellular localization and observed GFP fluorescence directly to eliminate potential artifacts associated with immunofluorescence detection of membrane proteins following detergent-based permeabilization. In bloodstream-form cells, the two proteins displayed strikingly different localizations; ^GFPAQP2 localized to a focal region adjacent to the kinetoplast (mitochondrial DNA) (Fig. 3A, Upper), whereas ^GFPAQP3 localized to this region and also to the plasma membrane (Fig. 3A, Lower), as demonstrated previously for native AQP3 (26). These results were confirmed by confocal microscopy and were indistinguishable in an aqp2/aqp3 or aqp2 background. Because AQPs typically form tetramers, these results suggest little or no physical interaction between AQP2 and AQP3 and the formation of primarily or exclusively AQP2 and AQP3 homotetramers.

We demonstrated above that AQP2 confers melarsoprol and pentamidine sensitivity to both bloodstream- and insect-stage cells. To determine whether AQP2 localization differs in these two life-cycle stages, the ^GFPAQP2 and ^GFPAQP3 strains were differentiated to the insect stage. Both proteins localized to the plasma membrane in insect-stage cells (Fig. 3B), indicating that the mechanism blocking access of AQP2 to the plasma membrane is bloodstream-stage specific and developmentally regulated. The focal structure identified by the AQP2 signal in bloodstream-form cells is adjacent to the kinetoplast and may represent the flagellar pocket, an invagination of the pellicular membrane, which is inaccessible to host innate immune effectors and the exclusive site for endocytosis, exocytosis, and specific receptors (32). To test this hypothesis, we stained invariant surface glycoprotein 65 (ISG65), known to identify the flagellar pocket (33), and visualized both proteins by microscopy. ^GFPAQP2 displayed almost complete colocalization with the flagellar pocket marker (Fig. 3C). We conclude that all three T. brucei AQPs display distinct subcellular localization in bloodstream-form T. brucei, with AQP2 displaying specific sequestration in the flagellar pocket.

Strains.

Bloodstream-form T. brucei, Lister 427, MiTat 1.2, clone 221a, and derivatives were maintained as previously described (45); 2T1 (46), KO-B48 (18), and aqp2/aqp3 null strains (5) were described previously. Strains were transfected using a Nucleofector apparatus (Lonza) in conjunction with cytomix or T-cell nucleofection solution. Transformants were selected with blasticidin (10 μg/mL), G418 (2 μg/mL), or hygromycin (2.5 μg/mL). Southern blotting was carried out according to standard protocols. Differentiation to the insect stage was triggered by transferring the cells to glucose-free DTM medium (47) supplemented with citrate and cis-aconitate at 27 °C. AQP and ^GFPAQP expression was induced by exposing cells to 1 μg/mL tetracycline for 24 h. EC₅₀ assays were carried out using alamarBlue as described (48, 49). The insect-stage cells were analyzed ≥6 d after differentiation was initiated. The AQP2/AQP3 chimera was PCR amplified from KO-B48 genomic DNA using a high-fidelity polymerase; two independent products were sequenced using standard procedures.

Plasmid Construction.

The AQP2 locus was disrupted in 2T1 cells by replacement of a 2,535-bp fragment with NPT and BLA selectable markers (the T. brucei genome is diploid). Briefly, AQP2-flanking sequences were inserted on both sides of the selectable marker cassettes and, before transfection, the plasmids were linearized by cleaving at the distal ends of these AQP2 targeting regions. The pRPa^GFPx construct (46) was modified to express native or GFP-tagged AQPs under the control of a Tet-regulated RRNA promoter. Both AQP2 constructs were checked for the presence of the expected SacII cleavage site. Primer sequences are available upon request.

Protein Analysis.

Subcellular fractionation by hypotonic lysis was carried out as described (5). All protein samples were stored in the presence of a protease inhibitor mixture (Roche) and were not boiled. Whole cell lysates and hypotonic lysis fractions were separated by SDS/PAGE and electroblotted according to standard protocols. For immunoblotting, we used an enhanced chemiluminescent kit (GE Healthcare), according to the manufacturer’s instructions. GFP was detected on Western blots and by fluorescence microscopy using polyclonal rabbit α-GFP (Europa; 1:2,000). Immunofluorescence was carried out according to standard protocols. Briefly, cells were permeabilized and ISG65 was detected using rabbit α-ISG65 (1:1,000) primary antibody (50) and a rhodamine-conjugated α-rabbit secondary antibody (1:100). Cells were settled on slides and mounted in Vectashield (Vector Laboratories) containing the DNA counterstain, 4,6-diamidino-2-phenylindole (DAPI). Images were captured using a Nikon Eclipse E600 epifluorescence microscope in conjunction with a Coolsnap FX (Photometrics) charge-coupled device (CCD) camera and processed in Metamorph 5.0 (Photometrics).