Agonist-mediated switching of ion selectivity in TPC2 differentially promotes lysosomal function

Ion selectivity is a defining feature of a given ion channel and is considered immutable. Here we show that ion selectivity of the lysosomal ion channel TPC2, which is hotly debated (Calcraft et al., 2009; Guo et al., 2017; Jha et al., 2014; Ruas et al., 2015; Wang et al., 2012), depends on the activating ligand. A high-throughput screen identified two structurally distinct TPC2 agonists. One of these evoked robust Ca2+-signals and non-selective cation currents, the other weaker Ca2+-signals and Na+-selective currents. These properties were mirrored by the Ca2+-mobilizing messenger, NAADP and the phosphoinositide, PI(3,5)P2, respectively. Agonist action was differentially inhibited by mutation of a single TPC2 residue and coupled to opposing changes in lysosomal pH and exocytosis. Our findings resolve conflicting reports on the permeability and gating properties of TPC2 and they establish a new paradigm whereby a single ion channel mediates distinct, functionally-relevant ionic signatures on demand.


Introduction
When ion channels open in response to a given stimulus, they allow the flow of a specific set of ions. The textbook view is that ion selectivity for a given ion channel is a hallmark feature that cannot be changed. But evidence has accrued suggesting that this may not always be the case. P2X receptors have long been proposed to change their ion selectivity in real time upon prolonged activation (Khakh and Lester, 1999) albeit controversially . And in TRPV channels, the selectivity filter is thought to form a second gate thereby coupling channel opening with changes in ion selectivity (Cao et al., 2013). Other channels such as Orai channels (McNally et al., 2012) and the mitochondrial uniporter MCU1 (Kamer et al., 2018) appear to alter their ion selectivity depending on protein partners. Two-pore channels (TPC1-3) are ancient members of the voltage-gated ion channel superfamily (Patel, 2015). TPCs are expressed throughout the endo-lysosomal system where they regulate the trafficking of various cargoes (Grimm et al., 2014;Ruas et al., 2010;Sakurai et al., 2015). Lossand gain-of function of TPCs is implicated in a number of diseases including NAFLD (non-alcoholic fatty liver disease) and Parkinson's (Patel, 2015;Patel and Kilpatrick, 2018). Despite their pathophysiological importance, both the ion selectivity and activating ligand(s) of TPCs are equivocal. Initial studies characterized TPCs as non-selective Ca 2+ -permeable channels activated by NAADP (nicotinic acid adenine dinucleotide phosphate) (Brailoiu et al., 2009;Brailoiu et al., 2010;Calcraft et al., 2009;Grimm et al., 2014;Pitt et al., 2010;Ruas et al., 2015;Schieder et al., 2010). But other studies indicate that TPCs are highly-selective Na + channels that similar to endolysosomal TRP mucolipins (TRPMLs), are activated directly by PI (3,5)P 2 (phosphatidylinositol 3,5bisphosphate), and/or by voltage (Boccaccio et al., 2014;Cang et al., 2014;Guo et al., 2017;She et al., 2018;Wang et al., 2012). Co-regulation of TPCs by these disparate stimuli has been noted in some instances (Jha et al., 2014;Ogunbayo et al., 2018;Pitt et al., 2014;Rybalchenko et al., 2012).
Functional characterization of TPCs is challenging due to their intracellular location. The anionic nature of NAADP and PI (3,5)P 2 prevents sufficient plasma membrane permeability, making them less suitable for experimental use in intact cells. On the other hand, studies of isolated lysosomes may result in loss of crucial accessory factors such as NAADP-binding proteins (Lin-Moshier et al., 2012). To circumvent these issues, we screened for membrane-permeable small molecule activators of TPCs. Here, we report two agonists that surprisingly switch the ion selectivity of TPC2 with different consequences for lysosomal function and integrity.

Identification of novel small molecule agonists of TPC2
Mutation of the N-terminal endo-lysosomal targeting motif in human TPC2 redirects it to the plasma membrane where TPC2 reportedly mediates robust Ca 2+ entry upon activation with NAADP (Brailoiu et al., 2010). We took advantage of this property to screen a cell line stably expressing TPC2 L11A/L12A with a library of 80000 natural and synthetic small molecules using a FLIPR-based Ca 2+ assay ( Figure 1A and B). Two structurally independent hits were identified, termed TPC2-A1-N and TPC2-A1-P ( Figure 1C and D), which reproducibly evoked Ca 2+ signals. The signals evoked by the compounds showed different kinetics whereby the TPC2-A1-N response reached its plateau faster than TPC2-A1-P ( Figure 1E and F). The structures and activities of these hits were confirmed by independent chemical syntheses and subsequent retesting. Full concentration-effect relationships for the plateau response indicated EC 50 values of 7.8 mM and 10.5 mM, for TPC2-A1-N and TPC2-A1-P, respectively ( Figure 1G and H). We thus identified novel small molecules that activate TPC2 and confirm its Ca 2+ -permeability in intact cells.

TPC2 agonists differentially evoke cytosolic Ca 2+ signals in live cells
To validate the hits further, we performed ratiometric Ca 2+ imaging of single cells transiently transfected with TPC2 L11A/L12A or a 'pore-dead' version (TPC2 L11A/L12A/L265P ). As shown in Figure 1I and J, TPC2-A1-N evoked Ca 2+ signals in cells expressing TPC2 L11A/L12A but not TPC2 L11A/L12A/L265P . These data confirm that TPC2-A1-N evoked Ca 2+ influx through the TPC2 pore. In accord, the responses to TPC2-A1-N were selectively blocked by the recently identified TPC2 blockers tetrandrine (Tet), raloxifene (Ral), and fluphenazine (Flu) (Penny et al., 2019;Sakurai et al., 2015;Figure 1-figure supplement 1) and by removal of extracellular Ca 2+ (Figure 1-figure supplement 2). TPC2-A1-P also induced Ca 2+ signals in cells expressing TPC2 in the presence ( Figure 1K and L) but not absence ( Figure 1-figure supplement 2) of extracellular Ca 2+ . However, the responses were smaller and delayed compared to TPC2-A1-N ( Figure 1I and K), consistent with the results obtained in cells stably expressing TPC2 L11A/L12A ( Figure 1E and F). TPC2-A1-P failed to induce Ca 2+ signals in cells expressing 'pore-dead' TPC2 L11A/L12A/L265P ( Figure 1K and L). Both TPC2-A1-N and TPC2-A1-P also failed to evoke Ca 2+ signals in cells expressing human TRPML1 re-routed to the plasma membrane (TRPML1 DNC ) (Grimm et al., 2010;Yamaguchi et al., 2011;Figure 1M-N and    in response to different concentrations of TPC2-A1-N (G) and TPC2-A1-P (H). Each concentration was tested in duplicates in two to four independent experiments, each. (I) Representative Ca 2+ signals recorded from HeLa cells loaded with the ratiometric Ca 2+ indicator, Fura-2 and stimulated with the indicated concentration of TPC2-A1-N. Cells were transiently transfected with plasma-membrane-targeted human TPC2 (hTPC2 L11A/L12A ) or a pore dead mutant (TPC2 L11A/L12A/L265P ). Colored lines represent the mean response from a population of cells. Traces in grey represent responses of single cells. (J) Statistical analysis of the maximal change in Fura-2 ratio (mean ± SEM) with the number of independent transfections in parentheses. An unpaired t-test was applied. *p<0.05. (K-L) Similar to I and J except that cells were stimulated with TPC2-A1-P. ***p<0.001. Traces in grey represent single cells responding to TPC2-A1-P.
Systematic structure modifications were performed on both screening hits. Surprisingly, none of the modified versions of TPC-A1-N or TPC2-A1-P showed significantly increased efficacies . TPC2-A1-P analogues missing the trifluoromethoxy residue were completely inactive ( Figure 1-figure supplement 6).

TPC2 agonists differentially evoke Na + currents in isolated endolysosomes
To determine more directly whether TPC2-A1-N and TPC2-A1-P activate TPC2, endo-lysosomal patch-clamp experiments were performed ( Figure 2). TPC2-A1-N elicited currents using Na + as the major permeant ion, in vacuolin-enlarged endo-lysosomes isolated from TPC2-expressing cells ( Figure 2A). The currents were inhibited by ATP as reported previously (Cang et al., 2013). In contrast, no activation was found in cells expressing TPC1 ( Figure 1-figure supplement 7). Endo-lysosomes isolated from cells expressing a gain-of-function variant of TPC2 (TPC2 M484L )  showed larger currents compared to the wild-type isoform upon application of TPC2-A1-N ( Figure 2B). TPC2-A1-P also evoked currents in endo-lysosomes isolated from cells expressing TPC2 and TPC2 M484L ( Figure 2C and D). As with TPC2-A1-N, the currents were potentiated by the gainof-function variant ( Figure 2D). Surprisingly, the currents evoked by TPC2-A1-P were significantly larger than those evoked by TPC2-A1-N ( Figure 2E-G) in both wild-type and gain-of-function variant. Full concentration-effect relationships for the plateau response in endo-lysosomal patch-clamp experiments indicated EC 50 values of 0.6 mM for both TPC2-A1-N and TPC2-A1-P (Figure 2-figure supplement 1A).
Increased Na + currents in the face of reduced Ca 2+ signals ( Figure 1) upon TPC2-A1-P activation prompted us to examine the effects of Na + removal on intracellular Ca 2+ elevation. For these experiments, we used cells stably expressing TPC2 L11A/L12A and agonist concentrations that evoked similar responses. As shown in Figure 2H and J, the amplitude of Ca 2+ signals evoked by TPC2-A1-N were unaffected by removal of Na + from the extracellular solution, although the rate of rise was increased. In contrast, both the amplitude and rate of rise of the responses to TPC2-A1-P were markedly enhanced ( Figure 2I and J). These data indicate that TPC2 is activated by TPC2-A1-N and TPC2-A1-P but that its permeability to Na + differs upon activation.  Figure supplement 1-source data 1. Effect of different TPC2 channel blockers on TPC2-A1-N and ML-SA1 mediated changes in intracellular calcium. Figure supplement 2. TPC2-A1-N and TPC2-A1-P mediated changes in intracellular calcium in the absence of extracellular calcium. Figure supplement 2-source data 1. TPC2-A1-N and TPC2-A1-P mediated changes in intracellular calcium in the absence of extracellular calcium. Figure supplement 3. Effects of TPC2-A1-N and TPC2-A1-P on TRPML1, 2, and 3. Figure supplement 3-source data 1. Effects of TPC2-A1-N and TPC2-A1-P on TRPML1, 2, and 3. (10) 1.00 NAADP PI (3,5) (3) (4) 50 50 Figure 2. TPC2 agonists alter Ca 2+ /Na + permeability of TPC2. (A-B) Representative TPC2-A1-N-evoked currents from enlarged endo-lysosomes isolated from HEK293 cells transiently expressing human TPC2 (hTPC2) (A) or a gain-of-function variant, (hTPC2 M484L ) (B). Currents were obtained before and after addition of 10 mM TPC2-A1-N in the absence or presence of 1 mM ATP. Recordings were carried out using standard bath and pipette solutions and applying ramp protocols (À100 mV to +100 mV over 500 ms) every 5 s at a holding potential of À60 mV. (C-D) Similar to A and B except that TPC2-A1-P (10 mM) was used. (E-G) Statistical analysis of current densities (mean ± SEM) recorded at À100 mV for endo-lysosomes expressing TPC2 or TPC2 M484L for TPC2-A1-N (E) and TPC2-A1-P (F), and the fold-change in current evoked by TPC2-A1-P versus TPC2-A1-N in the two variants (G). Red bars are WT and maroon bars are mutant. An unpaired t-test was applied. *p<0.05, **p<0.01, and ***p<0.001 (n > 10 endo-lysosomes per condition). (H-I) Ca 2+ signals evoked by TPC2-A1-N (H) or TPC2-A1-P (I) in Fura-2-loaded HEK293 cells stably expressing hTPC2 L11A/L12A . Recordings (mean responses from three to four independent experiments) were obtained in standard or Na + -free extracellular solution. (J) Statistical analysis of the maximal change in Fura-2 ratio (mean ± SEM). An unpaired t-test was applied. *p<0.05. (K-N) Agonist-evoked cation currents from enlarged endolysosomes isolated from HEK293 cells stably expressing human TPC2 under bi-ionic conditions: 160 mM Na + in the cytosol (bath) and 105 mM Ca 2+ in the lumen (pipette). Representative current-voltage curves before and after stimulation with either 10 mM TPC2-A1-N (K), 10 mM TPC2-A1-P (L), 50 nM NAADP (4 recordings out of 8 attempts) (M), or 1 mM PI(3,5)P 2 (N). ATP (1 mM) was added at the end of each experiment. Recordings were carried out using ramp protocols (À100 mV to +100 mV over 500 ms) every 5 s at a holding potential of 0 mV. (O) Expanded views of K-N, showing the reversal potentials (E rev ) of currents evoked by the indicated agonist. (P-Q) Statistical analyses of E rev (P) and the calculated relative cationic permeability ratios (P Ca /P Na ) (Q). Shown are mean values ± SEM. (R-S) Representative TPC2-A1-N-(R) or TPC2-A1-P-(S) evoked currents from endo-lysosomes isolated from HEK293 cells transiently expressing hTPC2 L265P under bi-ionic conditions. The online version of this article includes the following source data and figure supplement(s) for figure 2: Source data 1. TPC2 agonists alter Ca/Na permeability of TPC2.  TPC2 agonists alter Ca 2+ /Na + permeability of TPC2 The above analyses raised the intriguing possibility that the selectivity of TPC2 to Ca 2+ and Na + might not be fixed. To directly test this, we measured cation conductance evoked by the two compounds under bi-ionic conditions (luminal side: 105 mM Ca 2+ , pH 4.6, cytosolic side: 160 mM Na + , pH 7.2) ( Figure 2K and L). For these experiments, we used cells stably expressing TPC2. Based on the reversal potential (E rev ), the P Ca /P Na permeability ratio of TPC2-A1-N-evoked currents (I TPC2-A1-N ) was 0.65 ± 0.13 ( Figure 2O-Q). In marked contrast, E rev values for TPC2-A1-P-evoked currents (I TPC2-A1-P ) were more negative such that the P Ca /P Na permeability ratio was 0.04 ± 0.01 ( Figure 2O-Q) and thus Na + -selective. These ratios are highly reminiscent of those obtained for endogenous NAADP-mediated currents in planar patch-clamp analyses (Ruas et al., 2015) and for recombinant TPC2 activated by PI(3,5)P 2 measured using endo-lysosomal patch-clamp experiments where the channel appeared insensitive to NAADP (Wang et al., 2012). We thus examined the effects of NAADP and PI(3,5)P 2 in parallel. NAADP (50 nM) evoked robust currents with an E rev similar to TPC2-A1-N (À8.5 mV for NAADP versus À12 mV for TPC2-A1-N). The P Ca /P Na for NAADP (0.73 ± 0.14) was thus similar to TPC2-A1-N ( Figure 2M-Q). In contrast, the E rev for PI (3,5)P 2 (À53 mV) was similar to TPC2-A1-P (À61 mV) and the previously reported E rev for PI (3,5)P 2 (À68 mV) (Wang et al., 2012), corresponding to a P Ca /P Na ratio of 0.08 ± 0.01 ( Figure 2M-Q). No significant outward Na + or inward Ca 2+ conductances were observed in endo-lysosomes isolated from HEK293 cells expressing 'pore-dead' TPC2 L265P in the presence of TPC2-A1-N or TPC2-A1-P ( Figure 2R-S and Figure 2-figure supplement 1B-C). Moreover, the agonists induced similar shifts in selectivity using luminal solution containing 114 mM Na + and 30 mM Ca 2+ and a bath solution containing 160 mM Na + (Figure 2-figure supplement 1D-F) confirming that the measured P Ca /P Na permeability ratio is independent of the luminal Ca 2+ concentration. These data indicate that TPC2-A1-N and TPC2-A1-P not only differentially affect current amplitudes but also influence the P Ca /P Na permeability ratio ( Figure 2Q). TPC2-A1-N-activated TPC2 shows a higher relative Ca 2+ permeability, similar to NAADP-activated TPC2. In contrast, TPC2-A1-P-activated TPC2 shows a lower relative Ca 2+ permeability, similar to PI(3,5)P 2 -activated TPC2. Ion permeation through TPC2 is thus ligand-dependent providing an explanation for conflicting evidence that TPC2 is a NAADP-activated Ca 2+ release channel (Brailoiu et al., 2010;Grimm et al., 2014;Pitt et al., 2010;Ruas et al., 2015;Schieder et al., 2010) or a PI(3,5)P 2 gated Na + channel (Boccaccio et al., 2014;Guo et al., 2017;Wang et al., 2012).

TPC2 agonists activate TPC2 through distinct sites
The PI(3,5)P 2 binding site in TPC2 has recently been identified (She et al., 2019). To gain mechanistic insight into agonist action, we mutated K204 which is required for activation of TPC2 by PI (3,5)P2 (Kirsch et al., 2018;She et al., 2019) and examined its effect on responses to TPC2-A1-N and TPC2-A1-P. Intracellular Ca 2+ elevation mediated by TPC2-A1-N was similar in cells expressing plasma membrane targeted TPC2 with and without the K204A mutation ( Figure 3A and C). In contrast, the responses to TPC2-A1-P were reduced approximately two-fold ( Figure 3B and C). Similar selective inhibition was observed in GCaMP assays ( Figure 3G). Thus, whereas responses to TPC2-A1-N were comparable for the wild-type and mutant channel, the responses to TPC2-A1-P were reduced by the mutation. Furthermore, currents evoked by TPC2-A1-N were largely unaffected by the K204A mutation ( Figure 3H and J). In contrast, those to TPC2-A1-P were reduced ( Figure 3I and J). Collectively, these data suggest K204 is required for activation of TPC2 by TPC2-A1-P and PI (3,5)P 2 but not TPC2-A1-N. Molecular determinants for agonist action in TPC2 thus differ.

TPC2 agonists differentially affect lysosomal pH
Lysosomal function and integrity critically depend on the pH in the lysosomal lumen. Previously, it has been reported that NAADP induced pH changes in the lumen of acidic Ca 2+ stores, leading to an alkalinization (Cosker et al., 2010;Morgan and Galione, 2007) but the role of TPC2 in  regulating lysosomal pH is conflicting (Ambrosio et al., 2016;Cang et al., 2013;Grimm et al., 2014;Lin et al., 2015;Ruas et al., 2015). We therefore assessed the acute effects of TPC2-A1-N and TPC2-A1-P on lysosomal pH using a recently described novel lysosomal pH sensor, pH-Lemon-GPI (Burgstaller et al., 2019;Figure 4A-K and Video 1). Results from high-resolution array confocal laser scanning microscopy revealed that TPC2-A1-N, but not TPC2-A1-P, increased pH of single vesicles in cells expressing wild-type TPC2 ( Figure 4A-B and Video 1). In contrast, no pH changes were observed in cells expressing TPC2 L265P . Time-lapse wide-field fluorescence microscopy showed that activation of endogenous TPC2 with TPC2-A1-N modestly increased the lysosomal pH in untransfected cells in a time-dependent manner ( Figure 4C and I). The pH response was markedly potentiated in cells expressing wild-type TPC2 but not the pore-mutant ( Figure 4D,E and I). The response in the former was approximately half that evoked by direct alkalization with NaN 3 /NH 4 Cl ( Figure 4K). Again, vesicular pH was largely unaffected by TPC2-A1-P ( Figure 4F-H and J). Timelapse imaging additionally revealed that vesicle motility was strongly impaired by TPC2-A1-N but not TPC2-A1-P (Video 2). The effect of TPC2-A1-N was reversible and appeared to temporally correlate with the increase in vesicular pH (Video 1). In endo-lysosomal patch-clamp experiments, we probed the proton permeability of TPC2 and found that in contrast to TPC2-A1-P, TPC2-A1-N rendered the channel proton-permeable, thus providing a direct channel-dependent mechanism leading to alkalinization ( Figure 4L-N). Proton permeability did not substantially change our estimates of relative Ca 2+ and Na + permeability as P Ca /P Na values were similar when proton currents were  Representative TPC2-A1-N or -P-evoked currents from enlarged endo-lysosomes isolated from HEK293 cells transiently expressing human TPC2 (hTPC2) or the PI(3,5)P 2 -insensitive mutant, (hTPC2 K204A ). Currents were obtained before and after addition of 10 mM TPC2-A1-N or -P. Recordings were carried out using standard bath and pipette solutions and applying ramp protocols (À100 mV to +100 mV over 500 ms) every 5 s at a holding potential of À60 mV. (J) Statistical analysis of current densities (mean ± SEM) recorded at À100 mV for endo-lysosomes expressing TPC2 or TPC2 K204A using TPC2-A1-N or TPC2-A1-P to activate, respectively. *p<0.05 and **p<0.01, using one-way ANOVA followed by Tukey's post hoc test. The online version of this article includes the following source data for figure 3:  and upon treatment for 10 min with 10 mM TPC2-A1-P (t = 10 min; right images). Scale bar = 5 mm. (C-E) Normalized single cell responses (grey curves) and the respective average response (colored curves) of pH-Lemon-GPI upon treatment with 10 mM TPC2-A1-N in a region of high vesicle density. Shown are ratio curves of pH-Lemon-GPI expressed in control HeLa cells (n = 16 cells; C), in HeLa cells co-expressing wild-type TPC2-mCherry (n = 12 cells; D), and HeLa cells co-expressing TPC2 L265P -mCherry (n = 11 cells; E). (F-H) Experiments as in C-E for TPC2-A1-P. Shown are ratio curves of pH-Lemon-GPI expressed in control HeLa cells (n = 15 cells; F), in HeLa cells co-expressing wild-type TPC2-mCherry (n = 12 cells; G), and HeLa cells coexpressing TPC2L 265P -mCherry (n = 6 cells, (H). (I-J) Columns represent delta ratio values of pH-Lemon-GPI at min six from curves as shown in panels C-E and F-H, respectively. ***p<0.001 using unpaired student's t-test. (K) Average ratio signals over time of pH-Lemon-GPI in HeLa cells expressing wild-type TPC2-mCherry in response to either 0.5% NaN 3 and 50 mM NH 4 Cl (black average curve ± SEM, n = 12 cells) or 10 mM TPC2-A1-N (blue average curve ± SEM, n = 12 cells). Cells were treated with compounds as indicated. (L-M) Representative agonist-evoked inward H + currents from enlarged endo-lysosomes using hTPC2-expressing HEK293 cells. Both bath and pipette solutions contained 1 mM HCl, 5 mM HEPES, 5 mM MES, 150 mM NMDG, pH as indicated, adjusted with MSA. Ramp protocol (À100 mV to +100 mV in 500 ms, holding voltage = 0 mV) was used. Currents were obtained before and after addition of 10 mM TPC2-A1-N (L) or TPC2-A1-P (M). ATP (1 mM) was added at the end of the experiment to block TPC2 (L). (N) Statistical analysis of current densities (mean ± SEM) recorded at À100 mV (inward H + conductance; H + from luminal to cytosolic site) as shown in L and M. *p<0.05, using one-way ANOVA followed by Bonferroni's post hoc test. The online version of this article includes the following source data for figure 4: Source data 1. Effect of TPC2-A1-N on vesicular pH.

TPC2 agonists differentially affect lysosomal exocytosis
To further probe the physiological relevance of TPC2 activation by the agonists, we assessed the effect of TPC2-A1-N and TPC2-A1-P on lysosomal exocytosis ( Figure 5). Lysosomal exocytosis is involved in a plethora of physiological and pathophysiological processes including release of lysosomal enzymes and inflammatory mediators, clearance of lysosomal storage material and pathogens, plasma membrane repair and cancer progression (Lopez-Castejon and Brough, 2011;Machado et al., 2015;Miao et al., 2015;Xu and Ren, 2015). The Ca 2+ -dependence of this process has been mostly ascribed to activation of TRPML1 (Xu and Ren, 2015). Here, we used murine macrophages which express high endogenous levels of TPC2 (Cang et al., 2013). To assess lysosomal exocytosis, we quantified translocation of LAMP1 to the cell surface. As shown in Figure 5A-E, TPC2-A1-N was without effect on lysosomal exocytosis. This was not due to lack of efficacy of TPC2-A1-N on the murine channel because Ca 2+ measurements with mouse TPC2 fused to GCaMP6 (MmTPC2 GCaMP ) confirmed that TPC2-A1-N evoked larger responses than TPC2-A1-P (Figure 5figure supplement 1A) similar to the human channel ( Figure 1). In contrast, TPC2-A1-P evoked robust lysosomal exocytosis in a time-and concentration-dependent manner ( Figure 5A-E), providing a corollary to the effects of the agonists on lysosomal pH. To further corroborate that TPC2-A1-P triggers TPC2-dependent vesicle fusion in primary macrophages, electrophysiological measurements of membrane capacitance via the whole-cell patch-clamp technique were performed ( Figure 5F-H). These data showed that TPC2-A1-P caused a modest increase in cell size. In the final set of experiments, we examined the effects of TPC2 knockout on agonist responses. As shown in Figure 5-figure supplement 1, both TPC2-A1-N and TPC2-A1-P evoked measurable endogenous TPC2-like currents in macrophages from wild-type animals. The amplitudes of the currents were larger for TPC2-A1-P than TPC2-A1-N, again recapitulating agonist effects on human TPC2 (Figure 2). Importantly, currents were reduced in macrophages derived from TPC2 KO cells ( Figure 5-figure supplement 1) and the effects of TPC2-A1-P on lysosomal exocytosis and cell size were abolished ( Figure 5A-E). TPC2 knockout however failed to affect lysosomal exocytosis in response to ionomycin attesting to specificity ( Figure 5A-E). Collectively, these data further identify agonist-selective effects of TPC2 on lysosomal physiology in an endogenous setting.

Discussion
We demonstrate that the ion selectivity of TPC2 is not fixed but rather agonist-dependent. To the best of our knowledge, TPC2 is a unique example of an ion channel that conducts different ions in response to different activating ligands. The novel lipophilic, membrane permeable isoform-selective small molecule agonists of TPC2 (TPC2-A1-N and TPC2-A1-P) characterized herein mimic the physiological actions of NAADP and PI(3,5)P 2 most likely through independent binding sites. As such, our data reconcile diametrically opposed opinion regarding activation of TPC2 by endogenous cues and the biophysical nature of the ensuing ion flux. Whether TPC1 also switches its ion selectivity remains to be established. A previous bilayer study found that PI(3,5)P 2 did not gate TPC1 but induced modest shifts in relative permeability to different cations when TPC1 was activated by NAADP (Pitt et al., 2014).
Physiologically, we demonstrate inverse effects of the agonists on key lysosomal activities. While TPC2-A1-N increases the pH in the lysosomal lumen in a TPC2-dependent manner, TPC2-A1-P has no significant effect on lysosomal pH. An alkalinizing effect of TPC2-A1-N is in accordance with effects described previously for NAADP, which likewise induces alkalinization (Cosker et al., 2010;Morgan and Galione, 2007) and may be related to agonist-specific effects on proton permeability. The acute nature of these experiments in live cells, made possible by the lipophilicity of TPC2-A1-N, circumvents possible compensatory effects of TPC2 knockout which may have confounded steadystate analyses in previous studies (Cang et al., 2013;Grimm et al., 2014;Ruas et al., 2015). In contrast, TPC2-A1-P but not TPC2-A1-N promoted lysosomal exocytosis. The lack of effect of TPC2-A1-N is surprising given the Ca 2+ -dependence of lysosomal exocytosis but may relate to its inhibitory effect on lysosomal motility. Boosting lysosomal exocytosis is gaining traction as a strategy to combat disease (Lopez-Castejon and Brough, 2011;Machado et al., 2015;Miao et al., 2015;Xu and Ren, 2015). Selective activation of TPC2 in 'PI(3,5)P 2 -mode' with TPC2-A1-P offers novel scope to achieve this.
In sum, TPC2 can mediate very different physiological and possibly pathophysiological effects depending on how it is activated. TPC2 therefore emerges as an integrator of adenine nucleotideand phosphoinositide-based messengers with an intrinsic ability to signal through switchable ionic signatures. High-throughput screening

Materials and methods
The high-throughput screen was performed using a custom-made fluorescence imaging plate reader built into a robotic liquid handling station (Freedom Evo 150, Tecan, Mä nnedorf, Switzerland) as previously described (Urban et al., 2016). HEK293 cells were used that stably expressed RFP fusion proteins of human TPC2 (C-terminally tagged) or human CLN3 (N-terminally tagged) rerouted to the plasma membrane. Targeting was achieved by mutation of the endo-lysosomal targeting motifs (hTPC2 L11A/L12A , hCLN3 L253A/I254A ). Mutants were generated by site-directed mutagenesis using the QuikChange II XL protocol (Agilent), according to the manufacturer's instructions. Stable HEK293 cell lines were generated using 400 mg/mL geneticin (G418, Sigma). If G418-resistant foci were not identified after 3-4 days, the concentration of G418 was increased to 800 mg/mL. After 2-3 weeks cells were picked from G418-resistant foci and colonies were expanded in six well plates. RFP expression was assessed using confocal microscopy when cells were >50% confluent. Colonies with more than 95% RFP positive cells were selected, grown to >90% confluency, split and further expanded. For HTS experiments, cells were cultured at 37˚C with 5% of CO 2 in Dulbecco's modified Eagle medium (Thermo Fisher), supplemented with 10% fetal calf serum (Biochrom, Berlin, Germany), 2 mM L-glutamine, 100 U/mL penicillin, 0.1 mg/mL streptomycin, and 400-800 mg/mL G418.

Ca 2+ imaging
Single cell Ca 2+ imaging experiments were performed using Fura-2. HeLa cells and HEK293 cells were cultured at 37˚C with 5% of CO 2 in Dulbecco's modified Eagle medium (Gibco), supplemented with 10% fetal bovine serum, 100 U/mL penicillin, and 0.1 mg/mL streptomycin. Cells were plated onto poly-L-lysine (sigma)-coated glass coverslips, grown overnight and transiently transfected for 18-24 hr with plasmids using lipofectamine 2000 (Invitrogen) or TurboFect (Thermo Fisher) according to the manufacturer's instructions. For Ca 2+ influx experiments, cells were transfected with human TPC2 (C-terminally tagged with GFP or RFP) (Brailoiu et al., 2010) or TRPML1 (N-terminallytagged with GFP (Yamaguchi et al., 2011) or C-terminally tagged with YFP). They were targeted to the plasma membrane by mutation/deletion of the endo-lysosomal targeting motifs. For Ca 2+ imaging experiments with TRPML2 or TRPML3 (both C-terminally tagged with YFP Grimm et al., 2010), the latter two sufficiently locating to the plasma membrane when overexpressed. A pore-dead mutant of plasma membrane, GFP-tagged TPC2 (hTPC2 L11A/L12A/L265P ) was generated by site-directed mutagenesis. Transfected cells were loaded for 1 hr at room temperature with Fura-2 AM (2.5 mM) and 0.005% (v/v) pluronic acid (both from Invitrogen) in HEPES-buffered solution 2 (HBS2) comprising 1.25 mM KH 2 PO 4 , 2 mM CaCl 2 , 2 mM MgSO 4 , 3 mM KCl, 156 mM NaCl, 10 mM D-glucose and 10 mM HEPES (adjusted to pH 7.4 with HCl). After loading, cells were washed three times in HBS2, and mounted in an imaging chamber. All recordings were performed in HBS2. Ca 2+ imaging experiments were also performed in HEK293 cells stably expressing C-terminally RFP-tagged hTPC2 L11A/L12A (see above). These cells were plated, loaded with Fura-2 and recorded as with HeLa cells, except for that in indicated experiments, NaCl in HBS2 was replaced with NMDG. For lysosomal Ca 2+ release experiments, HeLa cells were transfected with human TPC2, C-terminally tagged with GCaMP6(s) (hTPC2 GCaMP ). The hTPC2 sequence was amplified by PCR from the pcDNA3.1 plasmid encoding YFP-tagged hTPC2 as described previously , using a forward primer carrying a NheI restriction site followed by a Kozak sequence, and a sequence binding the first 17 base-pairs of hTPC2: TATGCTAGCGCCACCATGGCGGAACCCCAGGC. The reverse primer also contained a NheI restriction site, preceded by a GSG-linker coding sequence, and the final 20 base-pairs of hTPC2, excluding the TPC2 stop codon: ATAGCTAGCACCAGAACCCCTGCACAGC-CACAGGTG. The PCR product was cloned into the GCaMP6(s) (Addgene ID 40753) vector by insertion into a NheI site, yielding plasmids encoding 6xHis tag-HsTPC2-GSG-Xpress tag-GCaMP6 fusion proteins. The ligated plasmid was sequenced by Sanger sequencing using standard CMV forward and pEGFP reverse primers, to confirm the insertion took place as desired. A pore dead mutant (hTPC2 GCaMP/L265P ) was generated by site-directed mutagenesis using the QuikChange II XL protocol (Agilent), according to the manufacturer's instructions, using the forward primer sequence GAGTC TCTGACTTCCCCCCTGGTGCTGCTGAC (reverse primer reverse complement of the former). All recordings were performed in nominally Ca 2+free HBS2. Images were acquired every 3 s at 20X (Fura-2) or 40X magnification using a cooled coupled device camera (TILL photonics) attached to an Olympus IX71 inverted fluorescence microscope fitted with a monochromator light source. Fura-2 was excited at 340 nm/380 nm, and GCaMP6(s) was excited at 470 nm. Emitted fluorescence was captured using 440 nm or 515 nm long-pass filters, respectively. https://elifesciences.org/articles/54712#video2

Endo-lysosomal patch-clamp experiments
Manual whole-endo-lysosomal patch-clamp recordings were performed as described previously . HEK293 cells were plated onto poly-L-lysine (Sigma)-coated glass coverslips, grown over night and transiently transfected for 17-25 hr with plasmids using TurboFect (Thermo Fisher) according to the manufacturer's instructions. Cells expressing wild-type (hTPC2) and a gainof-function variant (hTPC2 M484L ) of human TPC2 tagged at their C-termini with YFP were used . Cells were treated with either vacuolin or YM201636 (1 mM and 800 nM overnight, respectively) to enlarge endo-lysosomes. Currents were recorded using an EPC-10 patch-  (7) (8) (5)    clamp amplifier (HEKA, Lambrecht, Germany) and PatchMaster acquisition software (HEKA). Data were digitized at 40 kHz and filtered at 2.8 kHz. Fast and slow capacitive transients were cancelled by the compensation circuit of the EPC-10 amplifier. Glass pipettes for recording were polished and had a resistance of 4-8 MW. For all experiments, salt-agar bridges were used to connect the reference Ag-AgCl wire to the bath solution to minimize voltage offsets. Liquid junction potential was corrected as described . For the application of agonists, cytoplasmic solution was completely exchanged. Unless otherwise stated, the cytoplasmic solution comprised 140 mM K-MSA, 5 mM KOH, 4 mM NaCl, 0.39 mM CaCl 2 , 1 mM EGTA and 10 mM HEPES (pH was adjusted with KOH to 7.2). Luminal solution comprised 140 mM Na-MSA, 5 mM K-MSA, 2 mM Ca-MSA, 1 mM CaCl 2 , 10 mM HEPES and 10 mM MES (pH was adjusted to 4.6 with MSA). 500 ms voltage ramps from À100 to +100 mV were applied every 5 s, holding potential at 0 mV. The current amplitudes at À100 mV were extracted from individual ramp current recordings. For MmTPC1 measurements a one step protocol was applied (+140 mV over 2 s, holding potential of À70 mV) and the cytoplasmic solution contained 140 mM Na-gluconate, 5 mM NaOH, 4 mM KCl, 2 mM MgCl2, 0.39 mM CaCl2, 1 mM EGTA and mM 10 HEPES (pH 7.2). The luminal solution contained 140 mM Na-MSA, 5 mM K-MSA, 2 mM Ca-MSA, 1 mM CaCl 2 , 10 mM HEPES and 10 mM MES (pH was adjusted to 4.6 with MSA). All statistical analyses were done using Origin8 software.
For analyses under bi-ionic conditions, HEK293 cells stably expressing hTPC2 tagged at its C-terminus with YFP were used. The cytoplasmic solution comprised 160 mM NaCl and 5 mM HEPES (pH was adjusted with NaOH to 7.2) and the luminal solution comprised 105 mM CaCl 2 , 5 mM HEPES and 5 mM MES (pH was adjusted to 4.6 with MSA). The permeability ratio (P Ca /P Na ) was calculated according to Fatt and Ginsborg (1958): for bi-ionic test ( Figure 2Q and  (2006),

Isolation of murine alveolar macrophages
For preparation of primary alveolar macrophages, mice were deeply anesthetized and euthanized by exsanguination. The trachea was exposed and cannulated by inserting a 20-gauge catheter (B. Braun). Cells were harvested by eight consecutive lung lavages with 1 mL of DPBS (Dulbeccos's Phosphate-Buffered Saline) and cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum and 1% antibiotics. For experimentation, alveolar macrophages were directly seeded onto 12 mm glass cover slips and used within 5 days after preparation.

Lysosomal exocytosis experiments
Alveolar macrophages (3 Â 10 4 ), isolated from wild-type and TPC2 KO mice (described in Grimm et al., 2014) were seeded on 8-well plates (Ibidi) and cultured overnight. Cells were washed once with Minimum Essential Media (MEM) supplemented with 10 mM HEPES and then treated with TPC2-A1-N or TPC2-A1-P as indicated. Ionomycin (4 mM for 10 min) was used as positive control. Following treatment, cells were incubated with an anti-LAMP1 antibody (1:200,SantaCruz) in MEM supplemented with 10 mM HEPES and 1% BSA for 20 min on ice. Cells were then fixed with 2.6% PFA (Thermo Fisher) for 20 min and incubated with Alexa Fluor 488 conjugated secondary antibody (Thermo Fisher) for 1 hr in PBS containing 1% BSA. Nuclei were stained with DAPI. Confocal images were acquired using an LSM 880 microscope (Zeiss) with 40X magnification.

Capacitance measurements
Vesicle fusion of wild-type and TPC2-deficient primary alveolar macrophages was analyzed by measuring cell membrane capacitance using the patch-clamp technique as previously described (Zierler et al., 2016). In brief, vesicle fusion was recorded using the automated capacitance cancellation function of the EPC-10 (HEKA, Lambrecht, Germany). Measurements were performed in a tight seal whole-cell configuration at room temperature. Membrane capacitance values directly captured after breaking the seal between the membrane and the glass pipette were used as a reference for the initial cell size. Further recorded capacitance values were normalized to this initial determined capacitance. Extracellular solution contained: 140 mM NaCl, 1 mM CaCl 2 , 2.8 mM KCl, 2 mM MgCl 2 , 10 mM HEPES-NaOH, 11 mM glucose (pH 7.2, 300 mosmol/L). Internal solution contained: 120 mM potassium glutamate, 8 mM NaCl, 1 mM MgCl 2 , 10 mM HEPES-NaOH, 0.1 mM GTP (pH 7.2, 280 mosmol/L). At 180 s TPC2-A1-P (30 mM, diluted in external solution) was applied via an application pipette.

Lysosomal pH measurements
Cell culture and transfection: DMEM (Sigma Aldrich) containing 10% fetal bovine serum, 100 U/mL penicillin, 100 mg/mL streptomycin, and 2.5 mg/mL Fungizone (Thermo Fisher) was used to grow HeLa cells (all obtained from ATCC, Guernsey, UK). Transfection of cells in six-well (Greiner-Bio-One, Kremsmü nster, Austria) was performed using PolyJet transfection reagent (SignaGen Laboratories, Rockville). HeLa cells stably expressing pH-Lemon-GPI (NGFI, Graz, Austria) were generated by selection with 800 mg/mL of G418 (Sigma Aldrich, St. Louis, USA) and FACS sorting for CFP (excitation at 405 nm). Wide-field fluorescence microscopy experiments were performed using an OLYM-PUS IX73 inverted microscope (OLYMPUS, Vienna, Austria) using a 40X objective (UApo N 340, 40X/ 1.35 Oil, ¥/0.17/FN22, OLYMPUS). Illumination was performed using an OMICRON LedHUB High-Power LED Light Engine, equipped with a 455 nm and 505 nm LED light source (OMICRON electronics, Vienna, Austria), and 430 nm and 500 nm excitation filters (AHF Analysentechnik, Tü bingen, Germany), respectively. Images were captured at a binning of 2 using a Retiga R1 CCD camera (TEL-EDYNE QIMAGING, Surrey, Canada) and emissions were separated using an optical beam splitter (DV2, Photometrics, Arizona). Array Confocal microscopy was performed using an array confocal laser scanning microscope (ACLSM) built on a fully automatic inverse microscope (Axio Observer.Z1, Zeiss, Gö ttingen, Germany) using a 100x objective (Plan-Apochromat 100x, 1.4 Oil M27). Excitation was performed using laser light of diode lasers (Visitron Systems): CFP of pH-lemon was excited at 445 nm and 505 nm respectively. Emitted light was acquired with emission filters ET480/40 m for CFP and ET525/50 m for EYFP (Chroma Technologies, VT). A Photometrics CCD camera (CoolSnap HQ2) was used to capture all images at a binning of 1. Device control and image acquisition was performed using VisiView image acquisition and control software (Visitron Systems, Puchheim, Germany) for both devices. Images were processed using MetaMorph analysis software (Molecular Devices, San Jose). Data and statistical analyses were done using GraphPad Prism Software.

Synthesis of the compounds
All chemicals used were of analytical grade and were obtained from abcr (Karlsruhe, Germany), Fischer Scientific (Schwerte, Germany), Sigma-Aldrich (now Merck, Darmstadt, Germany), TCI (Eschborn, Germany) or Th. Geyer (Renningen, Germany). HPLC grade and dry solvents were purchased from VWR (Darmstadt, Germany) or Sigma-Aldrich, all other solvents were purified by distillation. Hydrophobic phase separation filters (MN 617 WA, 125 mm) were purchased from Macherey Nagel (Dü ren, Germany). All reactions were monitored by thin-layer chromatography (TLC) using pre-coated plastic sheets POLYGRAM SIL G/UV254 from Macherey-Nagel and detected by irradiation with UV light (254 nm or 366 nm). Flash column chromatography (FCC) was performed on Merck silica gel Si 60 (0.015-0.040 mm). NMR spectra ( 1 H, 13 C, DEPT, COSY, HSQC/HMQC, HMBC) were recorded at 23˚C on an Avance III 400 MHz Bruker BioSpin or Avance III 500 MHz Bruker BioSpin instrument, unless otherwise specified. Chemical shifts d are stated in parts per million (ppm) and are calibrated using residual protic solvents as an internal reference for proton (CDCl 3 : d = 7.26 ppm, (CD 3 ) 2 SO: d = 2.50 ppm) and for carbon the central carbon resonance of the solvent (CDCl 3 : d = 77.16 ppm, (CD 3 ) 2 SO: d = 39.52 ppm). Multiplicity is defined as s = singlet, d = doublet, t = triplet, q = quartet, sext = sextet, m = multiplet. NMR spectra were analyzed with NMR software MestReNova, version 12.0. 1-20560 (Mestrelab Research S.L.). High-resolution mass spectra were performed by the LMU Mass Spectrometry Service applying a Thermo Finnigan MAT 95 or Joel MStation Sektorfeld instrument at a core temperature of 250˚C and 70 eV for EI or a Thermo Finnigan LTQ FT Ultra Fourier Transform Ion Cyclotron Resonance device at 250˚C for ESI. IR spectra were recorded on a Perkin Elmer FT-IR Paragon 1000 instrument as neat materials. Absorption bands were reported in wave number (cm À1 ) with ATR PRO450-S. Melting points were determined by the open tube capillary method on a Bü chi melting point B-540 apparatus and are uncorrected. Microwave-assisted reactions were carried out in a Discover (S-Class Plus) SP microwave reactor (CEM GmbH, Kamp-Lintfort, Germany). HPLC purities were determined using an HP Agilent 1100 HPLC with a diode array detector and an Agilent Poroshell column (120 EC-C18; 3.0 Â 100 mm; 2.7 micron) with acetonitrile/water as eluent (60:40 acetonitrile/water + 0.1% formic acid).

Preparation of the TPC2-A1-N series
General procedure A -Amide coupling According to Sjogren et al. (1991) the appropriate aniline (1.0 eq.) and 2-cyanoacetic acid (1.0 eq.) were dissolved in DMF and cooled to 0˚C. DCC (1.0 eq.) was added portion wise. The mixture was warmed up to rt over 1 hr and subsequently diluted with hexanes/EtOAc (1:1). Precipitates were removed by filtration and the filtrate was extracted once with 1 M aq. HCl and thrice with EtOAc. The combined organic layers were washed with sat. aq. NaCl solution, dried over Na 2 SO 4 , filtered and concentrated in vacuo. Recrystallization from EtOH yielded the desired amides.

General procedure B -synthesis of N-aryl cyanoacetamides
According to Sjogren et al. (1991) the appropriate amides received from general procedure A (1.0 eq.) were dissolved in dry THF, the solution was cooled to 0˚C and NaH (dispersion in paraffin, 60%, 2.3 eq.) was added. After stirring for 15 min, the appropriate benzoyl chloride (1.1 eq.) was added. The mixture was stirred at 0˚C for 1 hr and then cautiously treated with 1 M HCl. The precipitate was collected by filtration, washed with ice water and cold EtOH and recrystallized from toluene to give the desired cyanoacetamides.
If the appropriate benzoyl chloride was not commercially available, it was prepared by refluxing the appropriate benzoic acid (1.1 eq.) in SOCl 2 (55 eq.) for 1 hr and concentrating in vacuo. The resulting acid chloride was immediately transferred to the reaction.
Preparation of the TPC2-A1-P series General procedure C -Paal-Knorr pyrrole synthesis Following a general procedure published by Kang et al. (2010) the appropriate b-ketoester (1.1 eq.) was dissolved in dry THF and cooled to 0˚C, before NaH (dispersion in paraffin, 60%, 1.5 eq.) was added portion wise. After the suspension was stirred for 30 min, a solution of appropriate halogenated acetophenone (1.0 eq.) and KI (1.0 eq.) in dry THF was added dropwise. The reaction mixture was allowed to warm up to rt over 2 hr, then poured on water and extracted three times with diethyl ether. The combined organic phases were washed with sat. aq. NaHCO 3 solution, dried over Na 2 SO 4 , filtered and concentrated in vacuo. The residue was dissolved in acetic acid and the appropriate primary amine (2.0 eq.) was added dropwise. The reaction mixture was stirred at 80˚C for 18 hr. The solvent was removed in vacuo, the residue disperged in water and extracted three times with diethyl ether. The collected organic phases were washed with sat. aq. NaHCO 3 solution, dried over Na 2 SO 4 and concentrated in vacuo. Purification was accomplished by FCC and recrystallization from EtOH if not otherwise specified.
General procedure D -alkaline deprotection of the pyrrolecarboxylic esters LiOH (10 eq.) was added to a solution of the appropriate ester (1.0 eq.) in dioxane/H 2 O (5:1) and the reaction mixture was stirred in a closed vessel under microwave irradiation (p max = 8 bar, P max = 200 W, T max = 180˚C) for 1-18 hr. The suspension was diluted with water to thrice original volume and aq. 2 M HCl was added dropwise under vigorous stirring until the mixture was strongly acidic. The formed precipitate was collected by filtration, washed with water and dried. If necessary the acids were recrystallized from EtOH to yield the pure products.