Proton-gated anion transport governs macropinosome shrinkage

Intracellular organelles change their size during trafficking and maturation. This requires the transport of ions and water across their membranes. Macropinocytosis, a ubiquitous form of endocytosis of particular importance for immune and cancer cells, generates large vacuoles that can be followed optically. Shrinkage of macrophage macropinosomes depends on TPC-mediated Na+ efflux and Cl− exit through unknown channels. Relieving osmotic pressure facilitates vesicle budding, positioning osmotic shrinkage upstream of vesicular sorting and trafficking. Here we identify the missing macrophage Cl− channel as the proton-activated Cl− channel ASOR/TMEM206. ASOR activation requires Na+-mediated depolarization and luminal acidification by redundant transporters including H+-ATPases and CLC 2Cl−/H+ exchangers. As corroborated by mathematical modelling, feedback loops requiring the steep voltage and pH dependencies of ASOR and CLCs render vacuole resolution resilient towards transporter copy numbers. TMEM206 disruption increased albumin-dependent survival of cancer cells. Our work suggests a function for the voltage and pH dependence of ASOR and CLCs, provides a comprehensive model for ion-transport-dependent vacuole maturation and reveals biological roles of ASOR.

Our model describes vesicles as spheres with an initial volume V 0 of (1) V 0 = 4/3 * π* r 0 3 with r 0 being the initial radius, and a surface area A 0 of (2) A 0 = 4 * π * r 0 2 = (36 * π) 1/3 * V 0 2/3 Assuming infinite water permeability, vesicle volume changes over time according to the number N osm of luminal osmotically active particles: resulting in changes of the surface area The vesicle has an electric capacitance C(t) given by: with c spec being the specific capacity per membrane area.
We consider five ionic species: Cl -, H + , Na + , K + , and X, in which X (which may be any other ion, including buffers) is initially calculated from the given concentrations of the other species to yield the desired initial membrane voltage U (defined as potential difference to cytoplasm) according to: with Q lum being the sum of the electrical charges in the lumen of the vesicle. The initial value of X is calculated as: (7) X lum (t=0) = U(t=0) * C(t=0)/F -(Na + lum (t=0) + K + lum (t=0) +H + tot lum (t=0) -Cllum (t=0)), Since vesicle shrinkage increases [H + tot ] lum , the lumen acidifies also without transmembrane H + flux if the buffer capacity remains constant. However, the bulk of [H + tot ] lum is bound to buffers, the concentration of which also increases during shrinkage. Therefore, we put: (10) β * (t) = β * (t=0) * (V 0 /V(t)), which eliminates purely shrinkage-driven acidification.
pH is calculated as usual: The amount of free H + (H + free ) is negligible compared to that of other ions and thus must not be considered for the osmotic shrinkage of the vesicle. The number of osmotic particles in the lumen (N osm lum ), which determines the size of the vesicle according to eq. (3), is: Note that the transport of H + into the vesicle in stoichiometric exchange for a different luminal ion may lead to vesicle shrinkage because the osmotic effect of H + is abolished by binding to buffer.
We model vesicles as having various combinations of Clchannels (ASOR), Na + channels (TPCs), 2Cl -/H + -exchangers (CLCs such as ClC-5) and a proton pump (H + -ATPase). As previously 1 , we assume for simplicity that transport rates are proportional to the driving force provided by the respective ion concentration differences and transmembrane voltage (and additionally ATP hydrolysis in the case of the proton pump). To model ASOR and CLCs we additionally multiply by equations that semi-quantitatively describe their steep voltageand pH-dependencies.
We write for the respective ion fluxes (positive when directed into the lumen): Flux through Na + channels (TPCs): (13) J Na = (-U + RT/F * ln([Na + ] cyt /[Na + ] lum )) * g TPC * A Flux through voltage-and pH-dependent Clchannels (ASOR): ] cyt )) * g H leak * A g TPC , g ASOR , g CLC , g ATP and g H leak are scaling factors (dimension mol/(V * sec * m 2 )), U ATP the electrochemical potential reached by ATP hydrolysis (270 mV) 2 , and R,T, and F the gas constant, absolute temperature, and Faraday's constant respectively. A is the surface area of the vesicle which changes over time with shrinkage as given by (3) and (4).
For simplicity, macropinosomal Na + channels, likely embodied by both TPC1 and TPC2 3 , were modeled as being voltage-and pH-independent. Whereas TPC1 currents are strongly voltage-dependent 4,5 , this is not the case for TPC2, but rectification and other properties of either channel strongly depend on the endogenous or artificial agonist used for their activation 4,6,7 . Whereas luminal alkalinization increases TPC1-mediated Na + currents 5 , it decreases TPC2-mediated Ca 2+ currents 7 . Given these uncertainties, we opted for modeling Na + channels as unregulated 'leak' currents. Likewise, we modeled the H + -conductance as an unregulated 'leak'.
Luminal amounts of ions were calculated by numerical integration (using Python and stiff Euler integration) of the differential equations: At each integration step, V(t), A(t), U(t) and pH(t) were calculated according to the above equations and ion concentrations were obtained by dividing ion amounts by the actual volume V(t). This was also done for the non-transported species such as X, increasing luminal concentrations with shrinkage.

Choice of parameters
For the calculations shown in Supplementary Note Figs. 1 -4, the following parameters were used: c spec = 0.01 Farad/m 2 , the typical capacitance of biological membranes 8 .
To estimate the order of magnitude for ion flux through ASOR per unit, we considered published values for ASOR plasma membrane currents in different cell lines 10,11,14 , and took a value of I/C = 100 pA/pF at + 100 mV, i.e. I/(U*C) = 10 3 S/F as estimate for maximally activated ASOR conductance per membrane capacitance. Since the specific membrane capacitance is c spec = 0.01 F/m 2 , this translates to 10 S/m 2 . Converting this value from electrical conductance to substance flow, we divide by Faraday's constant F = 96 485 to obtain g ASOR ~ 1* 10 -4 moles/(sec * m 2 * V). This rough estimate rests on the assumptions that channel density per surface area does not differ between the plasma membrane and macropinosomes and that channel properties are not changed by differences in membrane composition. The corresponding values g TPC , g CLC , g ATPase and g H leak were arbitrarily adjusted to yield macroscopic fluxes of the same order of magnitude as J ASOR , while allowing ASOR to yield ≈ 5-fold more maximal currents than CLCs to account for the generally higher conductance of channels vs. transporters. Given the strong voltage-and pH dependencies of ASOR and ClC-5, this approach required smaller values for g TPC and g ATPase .

Model calculations
To obtain basic insights into effects of transport processes, we first examined oversimplified models in which none of the transporters displays regulation by pH or voltage, with transmembrane ion fluxes being proportional to the driving force given by the Nernst potential (Supplementary Note Fig. 1). We used the same scaling factors g as for subsequent calculations in which we introduced voltage-and pH-dependencies for ASOR and CLCs, and set the respective factors f describing these dependencies as being constantly 0.5. The oversimplified models (Supplementary Note Fig. 1) reveal the fundamental effects of transporters and channels that lack explicit voltage-and pH dependencies and to correlate the predicted changes in pH, voltage, and volume with ion fluxes through specific transporters. Fig. 2 depicts the voltage-and pH-dependencies of ASOR and CLC calculated from (19) and (20)  ASOR has a moderate effect on pH, U, and volume V, the mutant markedly accelerates resolution (as observed in the experiments shown in Fig. 7c), makes the lumen more alkaline and shift the luminal potential to more positive values. Both parameters feed back negatively on ASOR currents, rendering resolution remarkably resilient towards ASOR expression levels.

Supplementary Note Figure 1. Calculations for simplified vesicle models containing voltage-and pH-independent 2Cl -/H + -exchangers ('CLC'), Clchannels ('ASOR') and Na + channels ('TPC').
Calculations with initial high (159 mM) or low (1 mM) luminal Clconcentrations ([Cl -] lum ). Results of model calculations are shown for luminal pH, membrane potential U (referred to cytoplasm), vesicle volume V, as well as ion fluxes (when applicable) through CLCs, ASOR, and TPC over a time span of 100 (a,b) or 1000 s (c-e). (a) Vesicles containing only a 'CLC' 2Cl -/H + -exchanger quickly reach an equilibrium with inside-positive potential with high luminal Cl -, and a luminal negative potential with low chloride. No significant change of luminal pH because H + is much more efficiently buffered than electrical charge (by membrane capacitance). Virtually no change in vesicle volume. (b) Parallel operation of a 'CLC' and a Clconductance 'ASOR'. Voltage changes with Clgradients are larger than in (a) because Clgradients produce a 3/2-fold larger change in electrochemical potential with a Clchannel than with a 2Cl -/H + -exchanger, as evident from respective Nernst equations. At these voltages, the CLC is initially far from equilibrium, leading to CLC-mediated H + -transport that lead to luminal alkalinization and acidification with high and low luminal [Cl -] lum , respectively, until an equilibrium is reached. The change is pH appears at first counterintuitive, as one would think that an inside-out Clgradient would increase luminal [H + ] by exchanging external H + for luminal Cl .through CLC 2Cl -/H + -exchange -however, and alkalinization is predicted. This effect is explained by the effect of the 'ASOR' Clconductance on U, which changes the transport direction of the CLC. Virtually no change in volume as Clfluxes are not electrically neutralized by cation currents. (c) Parallel operation of a 2Cl -/H + -exchanger and a Na + channel. TPC-mediated Na + efflux renders the lumen more negative, the specific voltage depending on the ratio of CLC over TPC conductance. The lumen-negative potential drives CLC-mediated Clefflux coupled to H + influx, which is of course larger with high luminal Cl -. Note that with high Clthere is strong luminal acidification, contrasting with the moderate alkalinization predicted with CLC + ASOR under the same ionic conditions (b). This difference is largely due to the opposing effects of Na + and Clchannels on U and explains, in principle, the opposite pH changes with low luminal Clwith WT and Tmem206 -/macropinosomes (Fig.  5a, e). (d) Parallel operation of ASOR + TPC + CLC. Luminal acidification is again stronger with higher than with lower [Cl -] lum , mainly because of the changes in U (that depend on the relative conductances of all three transporters). The uncoupled Cltransport pathway provided by the Clchannel leads to more vesicle shrinkage. (e) Parallel operation of Cland Na + channels leads to vesicle shrinkage with high, but not low [Cl -] lum . No change in pH as either channel transports H + . Parameters for calculation: Total time of simulation is 100 s (a-b) or 1000 s (c-e) with 0.001 s step. General parameters as in 'Choice of parameters'. The following values g for the 'strength' of transporters (in mol*s -1 *V -1 *m -2 , see equations (13)-(16)) were used: (a) g CLC = 4*10 -6 . (b) g ASOR = 4*10 -6 , g CLC = 4*10 -6 . (c) g TPC = 1*10 -6 , g CLC = 4*10 -6 . (d) g ASOR = 4*10 -6 , g TPC = 1*10 -6 , g CLC = 4*10 -6 . (e) g ASOR = 4*10 -6 , g TPC = 1*10 -6 . For the other transporters, the respective value of g was set to 0.

Supplementary Note Figure 3. Role of individual ion transporters explored in vesicle model considering voltage-and pH dependencies of ASOR and CLC.
A vesicle was modeled to contain 'ASOR' Clchannels and 'CLC' 2Cl -/H + -exchangers, both described to be pH-and voltage-regulated, as well as an unregulated Na + -conductance ('TPC'), an unregulated H + -conductance and an H + -ATPase, using simplified equations in which fluxes are proportional to the electrochemical driving force and parameters as specified in 'model description'. Calculations that explored the impact of luminal Cl -concentrations (159 vs 9 mM as on our experiments) were performed for: (a) the complete model vesicle (see Fig. 7a). (a) Simulation for complete model vesicle (Fig. 7a) with high or low [Cl -] lum (blue line and red lines, respectively). Left three panels show luminal pH, volume V, and voltage U, with smaller right panels displaying Na + -, Cl --and H + -fluxes through respective transporters as function of time. Vesicles are acidified with both high and low luminal Cl -. Vesicles are more acidic under low than under high Clconditions, as observed experimentally (Fig. 5a). This relative acidification is caused by the large difference in luminal potential caused by ASOR-mediated Clcurrents that render the lumen ≈50 mV more negative with low compared to high [Cl -] lum , This difference in U increases the driving force for electrogenic H + uptake with low vs. high [Cl -] lum . Vesicle models lacking various H + -transporters suggest that the acidification by Clremoval occurs in the presence of either an H + -ATPase or an H + -leak (see panels e, f, g, i, j below), but not when CLCs are the only H + transporters (h). Note that net ion flow through CLC 2Cl -/H + -exchangers is most prominent during the first minutes. As expected and found experimentally (Fig. 1c, d), low [Cl -] lum almost completely abolishes resolution that normally proceeds at nearly unchanged rate for a prolonged time.
(b) Effect of deleting the 'ASOR' Clchannel. Luminal voltage U is now negative irrespective of [Cl -] lum . It is largely determined by 'TPCs' and the inside-out Na + -gradient. ASOR KO leads to more acidic pH lum is more acidic than in the 'WT' under high [Cl -] lum ,, which is shown for comparison (taken from(a)), as observed experimentally (Fig. 5a). In contrast to 'WT' (a), and as observed (Fig. 5e), lowering [Cl -] lum leads to relative alkalinization. Both effects can be attributed to CLC 2Cl -/H + -exchange, which accumulates H + in the lumen in exchange for luminal Clin the presence of high [Cl -] lum , but not in its absence. This change in pH lum occurs also in the absence of an H + -ATPase (c) and an H + -conductance ('leak') (d). The experiment of Fig. 5e thus strongly indicates the presences of a CLC 2Cl -/H + -exchanger, likely ClC-5, on MPs. Note that the combination TPC/CLC can support vesicle shrinkage in the absence of ASOR, although at a much reduced rate. Indeed, even overexpression of ClC-5, which prominently localized to MPs, could not compensate for the loss of TMEM206 (Fig. 4j).
(c) Calculation for vesicle without 'ASOR' and H + -ATPase as model for Tmem206 -/-MPs in the presence of bafilomycin (Fig. 5d). Absence of proton pump activity leads to slight alkalinization compared to (b), and ΔpH lum between high and low [Cl -] lum is similar to (b). (e) Vesicle lacking CLC exchanger (blue lines) shows slightly delayed and reduced acidification and resolution compared to 'WT' (black dashed lines, from a). Low [Cl -] lum (red line) leads to relative acidification, primarily owing to changes in luminal potential as in (a). (f) Vesicle lacking H + -ATPase (blue lines) shows reduced acidification and slightly reduced resolution compared to 'WT' (black dashed lines, from a). Low [Cl -] lum (red line) again leads to relative acidification. (g) Vesicle lacking H + -leak (blue lines) shows enhanced acidification and slightly increased resolution compared to 'WT' (black dashed lines, from a). Low [Cl -] lum (red line) again leads to relative acidification.
(h) Vesicle expressing CLC as only H + -transporter (blue lines) show almost unchanged acidification and resolution compared to 'WT' (black dashed lines, from a). Importantly, low [Cl -] lum (red line) does not lead to relative acidification, because with the CLC exchanger the outside-in Clgradient counteracts the increased driving force for electrogenic H + entry provided by the ASOR-and TPC-generated lumen-negative potential.
(i) Vesicle expressing H + -ATPase as only H + -transporter (blue lines) shows delayed, but in the end enhanced acidification and slightly delayed, but in the end normal resolution compared to 'WT' (black dashed lines, from a). Low [Cl -] lum (red line) leads to relative acidification owing to the more negative luminal potential (j) Vesicle expressing an H + -leak as only H + -transporter (blue lines) shows markedly acidification and only moderately reduced resolution compared to 'WT' (black dotted lines, from a). Low [Cl -] lum (red line) leads to relative acidification due to the negative luminal potential. This vesicle models Clcn5 -/-MPs in the presence of bafilomycin (Fig. 5f, g) (k) Vesicle lacking any acidifying transport process (blue lines) shows markedly reduced resolution compared to 'WT' (black dashed lines, from a) or compared to a model vesicle expressing a 'proton leak' (j). Even with the strong pH-dependence of ASOR, as modeled in Supplementary Note Fig. 2, ASOR currents are non-zero. In our experiments with NH 4 Cl (Fig.  5b, c) the lumen most likely achieves more alkaline pH, leading to a virtual shutdown of ASOR and resolution. Fig. 3:

Conclusion from calculations in Supplementary Note
ASOR and TPC are essential for an efficient resolution of the model vesicle. While Cltransport through CLCs can, in principle, partially replace ASOR in this task, it is much less efficient, as experimentally confirmed by ClC-5 overexpression in Tmem206 -/-BMDMs. Given the pH-dependence of acid-sensitive ASOR/TMEM206 channels, resolution depends on luminal acidification which can occur through several mechanisms. Even an H + -conductance, which acts as 'proton leak' with sufficiently acidic lumen (generated e.g. by the H + -ATPase), can acidify the lumen to a degree that yields a resolution rate that is only moderately smaller  Fig. 2), in the presence of TPC, CLC, and V-type ATPase. Note moderate effect of WT ASOR overexpression, and a much stronger effect of the mutant which is accompanied by an alkaline shift in pH lum and a shift to lumen-positive potentials. (b) Plot of pH lum , voltage U lum , volume V and shrinkage rate (dV/dt) at 500 s as function of ASOR expression levels. (c) Plot of f(pH) ASOR , f(U) ASOR (eq. (19) and (20) in model description) and f(pH) ASOR * f(U) ASOR at t=500 s as measure of negative feed-back on ASOR activity. Note the contribution of both pH and U in suppression of ASOR currents with higher ASOR expression levels. Parameters used for calculations: Total time of simulation, 1000s with 0.001s step, general parameters as in 'Choice of parameters'. (a) g ASOR (1x) = 4*10 -5 , g ASOR (5x) = 2*10 -4 , g TPC = 1*10 -6 , g CLC = 5*10 -8 , g ATP = 4*10 -9 (b-c) 1x ASOR expression level g ASOR = 4*10 -5 , changed in 2-fold steps as indicated; g TPC = 1*10 -6 , g CLC = 5*10 -8 , g ATP = 0.