Positively selected modifications in the pore of TbAQP2 allow pentamidine to enter Trypanosoma brucei

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 K_m 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 V_m 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 V_max 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 K_m, as is good practice. Therefore, it is to be expected that this rate is considerably lower than the V_max.

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 K_m 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 t_1/2 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 V_max 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 TbAQP2^I190T and TbAQP2^I190T/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 TbAQP2^I190T but not for TbAQP2^I190T/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 K_i 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.