Instability of aquaglyceroporin (AQP) 2 contributes to drug resistance in Trypanosoma brucei

Defining mode of action is vital for both developing new drugs and predicting potential resistance mechanisms. Sensitivity of African trypanosomes to pentamidine and melarsoprol is predominantly mediated by aquaglyceroporin 2 (TbAQP2), a channel associated with water/glycerol transport. TbAQP2 is expressed at the flagellar pocket membrane and chimerisation with TbAQP3 renders parasites resistant to both drugs. Two models for how TbAQP2 mediates pentamidine sensitivity have emerged; that TbAQP2 mediates pentamidine translocation across the plasma membrane or via binding to TbAQP2, with subsequent endocytosis and presumably transport across the endosomal/lysosomal membrane, but as trafficking and regulation of TbAQPs is uncharacterised this remains unresolved. We demonstrate that TbAQP2 is organised as a high order complex, is ubiquitylated and is transported to the lysosome. Unexpectedly, mutation of potential ubiquitin conjugation sites, i.e. cytoplasmic-oriented lysine residues, reduced folding and tetramerization efficiency and triggered ER retention. Moreover, TbAQP2/TbAQP3 chimerisation, as observed in pentamidine-resistant parasites, also leads to impaired oligomerisation, mislocalisation and increased turnover. These data suggest that TbAQP2 stability is highly sensitive to mutation and that instability contributes towards the emergence of drug resistance.


Detection was carried out by incubating membranes with ECL Prime Western Blotting 185
System (Sigma) and GE healthcare Amersham Hyperfilm ECL (GE). Densitometry 186 quantification was conducted using ImageJ software (NIH). For quantification using 187 the Li-COR system (Li-Cor Bioscience, Lincoln NE), the following antibodies were 188 diluted in Odyssey blocking buffer (Li-COR): goat anti-rabbit IgG: IR Dye680RD and 189 goat anti-mouse or anti-rat IgG: IRDye800CW (Li-COR). All washes were with PBS 190 supplemented with 0.5% Tween20. Quantitative Fluorescence signals were quantified 191 on an Odyssey CLx Imager and processed using Li-COR software (Li-COR). 192

9
To determine whether ubiquitylation is involved in trafficking and turnover of 247 polytopic surface proteins in trypanosomes we addressed whether TbAQP2 is 248 ubiquitylated in vivo. We generated T. brucei cell lines expressing TbAQP2 tagged at 249 the N-( 3×HA AQP2) or C-terminus (AQP2 3×HA ) (Fig 1A and B) using an aqp-null cell line 250 [19] as chassis to prevent heterologous interaction with endogenous AQPs. The 251 hemagglutinin (HA) tag was selected as it lacks lysine residues (as opposed to more 252 bulky tags such as GFP) and therefore is incapable of becoming ubiquitylated and 253 interfering with data interpretation. Both constructs co-localized with ISG75 at the 254 posterior end of the cell, consistent with the location of native AQP2 at the flagellar 255 pocket (Fig 2A). 3×HA AQP2 is predominantly detected as two forms by immunoblotting 256 after SDS-PAGE: a ~38 kDa form, consistent with the monomeric form, and a >120 257 kDa form, likely a homotetramer (Fig 2B, lower panel), as previously reported for 258 other AQPs [16,[49][50][51]. In sharp contrast, AQP2 3×HA was found as two main species 259 of ~35 kDa and ~38 kDa, with no tetrameric form detected (Fig 2B, lower panel). 260 However, under native conditions, both constructs are organized as high molecular 261 weight complexes of ~480 kDa, consistent with a 4x4 conformation under native 262 conditions (Fig 2B, upper panel). 263 Thus, whereas 3×HA AQP2 is readily detectable as a stable tetramer, even under 264 harsh conditions, AQP2 3×HA is comparatively less stable in its tetrameric form (Fig 2B), 265 likely indicating interference by the C-terminal HA tag to oligomerization and/or 266 tetramer stability. To determine the glycerol transport capacity of these proteins, we 267 inhibited the activity of the trypanosome alternative oxidase (TAO) with 268 salicylhydroxamic acid (SHAM) [21]. Inhibition of TAO leads to increased intracellular 269 glycerol, building up to toxic levels that can only be prevented by export via a glycerol 270 transporter such as AQP. Therefore, the absence of functional AQPs renders cells 271 highly susceptible to SHAM. Consistent with stability of the AQP2 3xHA oligomeric form, 272 expression of the 3×HA AQP2 construct in the aqp-null background restored sensitivity 273 to pentamidine and glycerol transport comparable to wild type cells, whereas 274 AQP2 3×HA only partly rescued these phenotypes (Fig 2C and S1 Fig). Both 3×HA AQP2 275 and AQP2 3×HA have long half-lives (>4h) (Fig 2D) indicating that impaired transport 276 activity of AQP2 3xHA is unlikely due to altered turnover or structure. Therefore, although 277 introduction of HA epitopes to either terminus does not alter localization, only the N- We observed high molecular weight adducts when cells expressing 3×HA AQP2 289 were treated with either ammonium chloride (lysosomal activity inhibitor) or MG132 290 (proteasome inhibitor) (Fig 3A), likely representing ubiquitylated intermediates en 291 route to degradation. Subsequent western blotting identified a predominant band of 292 ~55 kDa reactive to anti-ubiquitin antibody upon immunoprecipitation with anti-HA 293 magnetic beads, consistent with the addition of ubiquitin to TbAQP2 (~38 kDa for 294 unmodified protein (Fig 3B). To corroborate these results, we performed an affinity 295 isolation using a commercial ubiquitin binding domain (UBD) resin followed by western 296 blotting with anti-HA antibody. This revealed unmodified monomer together with high 297 molecular weight adducts, likely representing TbAQP2 with various numbers of 298 ubiquitin conjugates; the latter clearly represents a small fraction of total AQP2 299 expressed in these cells (Fig 3C). Interestingly, we noted a band of around ~40 kDa, 300 likely corresponding to monoubiquitylated TbAQP2 (Fig 3C). Collectively, these 301 results indicate that TbAQP2 is modified by ubiquitin in the bloodstream form of T. 302

brucei. 303
Next, we sought to investigate the mechanisms by which TbAQP2 is degraded. 304 Imaging suggested that TbAQP2 is predominantly located at the flagellar pocket 305 together with ISG75, but a proportion is also in close proximity to early endosomes 306 (positive for Rab5A and Rab5B) but less so for recycling endosomes (Rab11) (Fig 4A)  307 suggesting transit of TbAQP2 through early endosomes. Moreover, TbAQP2 308 displayed strong overlap with p67, a lysosomal marker, suggesting that TbAQP2 is 309 delivered to the lysosome via endocytosis (Fig 4A). Similar observations were made 310 with cells expressing AQP2 3×HA (S2 fig), once more indicating that the C-terminal tag 311 does not impair trafficking but rather hinders oligomerisation. Further, pulse-chase 312 analysis showed that ISG75 has a half-life of ~3.6 h, consistent with earlier studies of ~6 h (Fig 4B), which together with partial juxtaposition with Rab11, suggests 316 possible recycling. To determine whether TbAQP2 is degraded in the lysosome or the 317 proteasome we treated cells with bafilomycin A1 (BafA1; inhibitor of the lysosomal v-318 ATPase) or MG132 (canonical proteasome inhibitor with broad-range inhibitory 319 capacity against serine proteases and calpain-like proteases [54]). In untreated cells, 320 TbAQP2 was reduced by ~50% after 1 h as expected, but in cells treated with BafA1 321 or MG132, less that 20% of the protein was degraded (Fig 4C). It is important to note 322 that MG132 can also impair degradation of proteins delivered to the lysosome as it 323 acts as a broad range inhibitor for lysosome-specific proteases [54]. Overall, these 324 data indicate that TbAQP2 is ubiquitylated and delivered to the lysosome for 325 degradation, albeit with a pool of longer-lived protein that may constitute a recycling 326 population. 327

Intracellular N-terminal lysine residues are essential for oligomerisation and 328 channel function of TbAQP2 329
Predictions of TbAQP2 topology [55] suggested cytosolic localisations for both 330 N-and C-termini, as is known for the mammalian orthologues (S3 Fig, [15,56]). AQP2 331 has five lysine residues that are exposed to the cytosol, at positions 19, 45, 54, 147, 332 and 234 (Fig 1B). To better understand the potential ubiquitylation sites in TbAQP2, 333 we used UbPred (http://www.ubpred.org) [57] to predict lysine residues as candidate 334 ubiquitin acceptors. UbPred suggested that lysine residues in position 19, 45, and 54 335 are potential ubiquitylation sites in TbAQP2, with prediction scores of 0.65, 0.73, and 336 0.88, respectively. All three residues are located within the N-terminal cytoplasmic 337 region of AQP2 (Fig 1B). 338 To dissect the contribution of these residues to TbAQP2 localisation and 339 function, we generated a cell line expressing N-terminally tagged AQP2 in which all 340 three of these lysine residues were simultaneously mutated (AQP2 3K>R ). 341 Unexpectedly, while the wild-type protein located in the posterior end of the cell, 342 AQP2 3K>R was mislocalized (Fig 5A) and failed to restore pentamidine sensitivity and 343 glycerol transport (Fig 5B). Furthermore, whereas AQP2 WT co-localises with ISG75 at 344 the posterior end of the cells, AQP2 3K>R was retained in the endoplasmic reticulum 345 (ER), as suggested by co-localisation with the ER marker TbBiP (Fig 5C). Blue native-346 for TbAQP2 folding and hence anterograde trafficking and that their replacement by 351 arginine triggers entry into an ER-associated degradative (ERAD) pathway [58][59][60]. 352

Site-directed mutation of cytoplasmic lysine residues of TbAQP2 leads to 353 protein instability 354
To determine whether the effects observed for AQP2 3K>R could be attributed to 355 a single lysine residue we generated a construct in which all cytoplasmic lysine 356 residues were mutagenized to arginine (AQP2 5K>R ) (Fig 1B). Using this construct as 357 template, we reverted each lysine individually using site-directed mutagenesis, 358 generating cell lines expressing N-terminally tagged mutant TbAQP2 with only one 359 lysine residue reinstated (AQP2 R19K , AQP2 R45K , and AQP2 R54K ). None of these 360 mutants formed oligomers (S4A Fig.) and were retained in the ER, as demonstrated 361 by co-localisation with TbBiP (Fig 6A). Moreover, AQP2 5K>R , AQP2 R19K , AQP2 R45K , 362 and AQP2 R54K turn over faster than AQP2 WT and are stabilised by MG-132 (Fig 6B  363 and Table 1), consistent with the absence of detection by BN-PAGE analysis and the 364 lack of sensitivity to pentamidine and glycerol transport observed in these mutants 365 Residue 147 is located between TMD4 and TMD5, potentially indicating that mutation 374 of this residue leads to a far more unstable protein than the other constructs, and in 375 good agreement with the MD simulations. In TbAQP2 WT , K147 is predicted to interact 376 with Y151 and N70 on TMD1 and maintain the TMD3-TMD1 interface (S4G Fig, right   although in some cases the selectivity pore is mutated, many chimeric AQP2/3 alleles 391 do not have altered amino acids in the selectivity pore, but rather replacement of TMD 392 regions with sequences from TbAQP3 (Fig 1C), [29][30][31][32]. Moreover, the AQP2/3 393 chimeras characterised so far display a subcellular localisation resembling TbAQP3 394 at the plasma membrane, in contrast with an expected flagellar pocket localisation for 395 TbAQP2 (Fig 2A) [27]. However, it is unclear if TbAQP2 chimerisation impacts 396 additional features beyond subcellular localisation. 397 We generated cell lines expressing tetracycline-regulated N-terminal tagged 398 TbAQP1, TbAQP2, TbAQP3 and the chimeric AQP2/3 40AT (40AT) (Fig 1C), isolated 399 from relapse patients from the Democratic Republic of Congo [29]. One of the main 400 AQP2 TMD4 and 40AT are turned over more rapidly (<1 h) than TbAQP1 and TbAQP2 419 (Fig 7D and Table 2), explaining the lack of glycerol transport in cells expressing these 420 constructs (Fig 7C). 421 The localisation of TbAQP3 and the chimeric AQP2-3 proteins is reminiscent of 422 the subcellular localization observed in the lysine-to-arginine TbAQP2 mutants. Based 423 on these observations we hypothesised that these constructs are likely to be retained 424 in the ER, at least to a level comparable to that of the lysine-to-arginine TbAQP2 425 mutants. We observed that whereas TbAQP1 and TbAQP2 show poor co-localisation 426 with TbBiP, the signal of TbAQP3 and 40AT partly co-localised with this ER marker, 427 indicating some degree of retention within this organelle (Fig 8A). Moreover, TbAQP3 428 and 40AT turnover was faster than TbAQP1 and TbAQP2 and was significantly 429 impaired in the presence MG132 but not bafilomycin A1 (Fig 8B), indicating that these whereas TbAQP3 is comparatively short-lived (t1/2 ~1 h) and localises mainly to the 476 cell body surface, suggesting a connection between oligomerisation, stability, 477 subcellular localisation and transport activity [67,69-71]. Furthermore, replacement of 478 TMD4 or 6 in TbAQP2 by TbAQP3 sequences (we were unable to generate TMD5 479 chimeras), as observed in clinically relevant chimeric AQP2-3, led to impaired is highly sensitive to mutation and/or chimerisation, which results in failure to correctly 485 fold and ER-retention. This mechanism most likely accounts for many instances of 486 clinically observed pentamidine and melarsoprol resistance.    The 3xHA-tag is omitted for simplicity. Details of the selectivity pore for each of these 948 proteins (12Å) are also included. For TbAQP2, AQP2 TMD4 and 40AT constructs, the 949 selectivity filter is composed of the "NSA/NPS" motif, whereas TbAQP1 and TbAQP3 950 contain a "NPA/NPA" motif. 951