The human two-pore channel 1 is modulated by cytosolic and luminal calcium

Two-pore channels (TPC) are intracellular endo-lysosomal proteins with only recently emerging roles in organellar signalling and involvement in severe human diseases. Here, we investigated the functional properties of human TPC1 expressed in TPC-free vacuoles from Arabidopsis thaliana cells. Large (20 pA/pF) TPC1 currents were elicited by cytosolic addition of the phosphoinositide phosphatidylinositol-(3,5)-bisphosphate (PI(3,5)P2) with an apparent binding constant of ~15 nM. The channel is voltage-dependent, activating at positive potentials with single exponential kinetics and currents are Na+ selective, with measurable but low permeability to Ca2+. Cytosolic Ca2+ modulated hTPC1 in dual way: low μM cytosolic Ca2+ increased activity by shifting the open probability towards negative voltages and by accelerating the time course of activation. This mechanism was well-described by an allosteric model. Higher levels of cytosolic Ca2+ induced a voltage-dependent decrease of the currents compatible with Ca2+ binding in the permeation pore. Conversely, an increase in luminal Ca2+ decreased hTPC1 activity. Our data point to a process in which Ca2+ permeation in hTPC1 has a positive feedback on channel activity while Na+ acts as a negative regulator. We speculate that the peculiar Ca2+ and Na+ dependence are key for the physiological roles of the channel in organellar homeostasis and signalling.

The endolysosomal system is composed of a series of internal compartments fundamental for cellular homeostasis and is involved in a variety of different physiological processes from signaling to cell growth up to defence mechanisms, to cite only few 1,2 . The family of two-pore channels (TPC) belongs to the group of endolysosomal membrane proteins and has two members in humans, namely TPC1, which is mainly expressed in endosomes, and TPC2, which mainly localizes to lysosomes 3,4 . From a structural point of view, TPC channels are homo-dimers [5][6][7] . The single monomer is formed by two shaker-type subunits covalently linked. Each shaker-type subunit has six transmembrane domains (S1 to S6) of which S4 contains a series of basic amino acids and is recognized to be the voltage sensor 6 , even though TPC2 is practically voltage-independent [8][9][10][11] . The regions between S5 and S6 (P-loops) form the selectivity filter of the permeation pore that confers cationic selectivity.
The physiological importance of TPC channels is underlined by recent findings of their involvement in different severe pathologies. TPC2 plays a role in neoangiogenesis processes linked to vascularization of solid tumors 12 , in neurodegenerative Parkinson disease 13 and in Ebola virus infections 14 . In the heart, TPC1 is involved in the generation of Ca 2+ waves after a period of ischemia and its deletion in mice is cardio protective 15 .
Despite the physiological relevance, the functional properties of TPC channels are not fully defined. The major problem in the functional characterization of these proteins is represented by their intracellular localization. This renders very difficult the application of electrophysiological techniques because of the sub-micrometric dimension of animal endosomes and lysosomes. Divergent results have been reported regarding the nicotinic acid adenine dinucleotide phosphate (NAADP)/PI(3,5)P 2 sensitivity and the Ca 2+ permeability of the channels. Using intracellular Ca 2+ measurements 10,[16][17][18][19][20][21][22][23] , redirection of the channels from lysosomes to the plasma membrane 24 , reconstitution in lipid bilayers [25][26][27] and planar patch clamp of vacuolin enlarged lysosomes 10,14,28,29 , TPC1 and TPC2 have been reported to be activated by sub-micromolar concentrations of NAADP and to be highly permeable to Ca 2+ , leading to the suggestion that TPC channels represent the long sought for NAADP sensor 3,30 . In contrast, using classical patch clamp of enlarged lysosomes TPC1 and TPC2 were found to be insensitive to NAADP (even though in one study TPC2 could also be activated by NAADP 8
An alternative approach to study mammalian lysosomal channels is to express them in plant protoplasts where they localize to the large lysosomal compartment, the vacuole, which is easy to isolate and amenable to classical patch clamp techniques 9,33-35 . Knocking out the single gene coding for the endogenous AtTPC1 the system is completely TPC-free 9 . Furthermore the vacuole has its cytosolic side facing the external bath solution, which is an ideal experimental situation to investigate cytosolic modulators 35 . Using this system we have previously found that hTPC2 is insensitive to NAADP, activated by PI(3,5)P2, and highly Na + selective 9 , in agreement with Wang et al. 11 . Here, we focused our attention on human TPC1 15,26,27,32,36,37 , which is less characterized than TPC2. We found that the channel is dependent on PI(3,5)P2 but insensitive to NAADP. Currents are strongly voltage-dependent, mainly selective to Na + but also Ca 2+ permeable. Furthermore we found that both cytosolic and luminal Ca 2+ are powerful modulators of hTPC1.

Results
hTPC1 channels is functionally expressed in Arabidopsis thaliana vacuoles. In mouse, two TPC1 isoforms have been found, the second one shorter than the first with an N-terminal deletion of 69 aminoacids 38 .
Here we investigated a human isoform of TPC1 similar to the short mouse one.
Human TPC1 (hTPC1) was transiently expressed in mesophyll protoplasts from the Arabidopsis thaliana tpc1-2 mutant (AtTPC1 null background) 39 . Released EGFP-positive vacuoles (Fig. 1a) were viewed and selected under a fluorescence microscope to achieve patch-clamp recordings in the whole-vacuole configuration. Small (mostly background) currents were recorded in control conditions, whereas bath (equivalent to the cytosolic side) application of PI(3,5)P 2 reversibly increased the current in response to a depolarizing step of + 40 mV from a holding potential of − 70 mV (Fig. 1b,c). The current gradually decayed to the basal level upon PI(3,5)P 2 washout (Fig. 1b). PI(3,5)P 2 -evoked currents were observed in hTPC1-EGFP expressing vacuoles but not in untransformed control vacuoles (Fig. 1d), strongly suggesting that PI(3,5)P 2 -responses are due to activation of hTPC1 channels. As NAADP has been proposed to be a TPC-channel agonist 3 , we tested this intracellular calcium mobilizer on EGFP-positive vacuoles. Bath application of 100 nM NAADP on PI(3,5)P 2 responding vacuoles did not elicit an increase of membrane current (Fig. 1d); these data point to a direct interaction of PI(3,5)P 2 but not of NAADP with hTPC1.
The PI(3,5)P 2 -activated current increased with increasing PI(3,5)P 2 concentration with a similar apparent binding constant at + 40 mV and − 40 mV, i.e. 15 and 18 nM (Fig. 1e), suggesting voltage-independent high affinity of PI(3,5)P 2 -interaction with hTPC1 channels. PI(3,5)P 2 -mediated current is voltage-dependent and largely carried by Na + ions. To investigate the voltage-dependence of hTPC1-mediated current, EGPF-positive vacuoles were stimulated by increasing voltage pulses during bath perfusion of 90 nM PI(3,5)P 2 (Fig. 2a). The response was larger at positive than at negative potentials indicating activation of a voltage dependent outward rectifying conductance. When cytosolic Na + was diminished to 10 mM, outward currents were significantly reduced (Fig. 2b). Figure 2c summarizes the different current-voltage relationships of hTPC1 mediated currents when varying the cytosolic Na + concentration; a large shift in the reversal voltage (V rev ) from positive to negative values was recorded upon a cytosolic sodium concentration change from 10 to 200 mM. The full agreement of the experimental V rev with the theoretical Nernst voltage for Na + shown in Fig. 2e strongly points to hTPC1 as a Na + permeable channel. In Fig. 2c the voltage dependent inhibition apparent at positive voltages in 100 mM sodium (the effect was less pronounced in 200 mM Na + ) was due to the presence of 2 mM of cytosolic Mg 2+ : removing cytosolic Mg 2+ eliminated the voltage dependent current inhibition (Fig. 2d). A similar effect has been reported for TPC2 8 . By using a Woodhull approach 40 , we estimated that Mg 2+ binds to a site located along the permeation pore at an electrical distance (δ ) from the cytosol of 0.36 ± 0.3 with an affinity constant at 0 mV (K Mg ) of 20 ± 3 mM (see lines in Fig. 2d).

Potassium and calcium selectivity of PI(3,5)P 2 -evoked current.
Replacing cytoplasmic Na + by K + in the presence of luminal Na + resulted in a total disappearance of PI(3,5)P 2 activated outward currents (Fig. 3a), while inward tail currents, reflecting Na + flowing from lumen to cytosol, were still present, heavily suggesting a low hTPC1 permeability for K + . In this experimental condition the reversal voltage was larger than 90 mV as shown in the inset of Fig. 3a and in Fig. 3b pointing to a permeability ratio between K + and Na + lower than 2.8%. To measure the Ca 2+ permeability, we substituted 100 mM Na + with 50 mM Ca 2+ in the pipette solution. In this bi-ionic condition large outward currents, reflecting Na + flowing from the cytosol to the lumen, were present (Fig. 3c, bottom). To focus on Ca 2+ permeability, tail currents were recorded by stepping from + 70 to − 100 mV (10 mV decrement), after a 500 ms activating pule to + 50 mV (Fig. 3d, top). Exploring the responses on an expanded scale around the V rev , small inward tails were detectable. Ion permeability ratio between Ca 2+ and Na + (P Ca /P Na ), measured by reversal potentials (Fig. 3d, bottom) was between 5 and 10% (Fig. 3e). It is worth noting that the reversal voltage was somewhat dependent on the applied protocol: the longer was the pre-pulse to + 50 mV the more positive was V rev (i.e. the higher was the apparent Ca 2+ permeability). We attributed this effect to the entrance of Na + into the lumen and mathematically corrected it (see Supplemental Fig. 1). These data indicate that Ca 2+ can permeate hTPC1 although with a significantly lower permeability than Na + .

Cytosolic calcium ions modulate PI(3,5)P 2 -activated currents.
We wondered if cytosolic Ca 2+ could affect the functionality of hTPC1. We therefore performed experiments adding Ca 2+ in the cytosolic bath solution from nominally 0 up to 200 μ M. Very interestingly, as shown in Fig. 4a, currents increased upon an increase of Ca 2+ from 0 to 20 μ M. Normalized conductance plotted against applied voltage demonstrated that raising the cytosolic Ca 2+ shifted the open probability of the channel toward negative voltages (Fig. 4b). The continuous lines in Fig. 4b were obtained by fitting G norm with a Boltzmann equation: the half activation voltage versus cytosolic Ca 2+ concentration varied by about 50 mV and saturated at cytosolic Ca 2+ near 20 μ M (Fig. 4c). On the contrary, the slope was not significantly affected by Ca 2+ (Fig. 4d). By fitting the time course of current activation and deactivation with a single exponential function (see Supplemental Fig. 2a) we found that the effect of [Ca 2+ ] cyt on the relaxation time constants was pronounced at positive voltages (Fig. 4e), i.e. the channel activated more rapidly at higher [Ca 2+ ] cyt .
The [Ca 2+ ] cyt dependence of the open probability and of the time constants could be described by the mathematical model shown in Fig. 4f: where C 0 and C 1 represent the closed states of the channel respectively without and with a Ca 2+ ion bound, O 0 and O 1 are the open, conductive states without and with Ca 2+ . Transitions between Ca 2+ bound and Ca 2+ free states are supposed to be fast, with apparent voltage-independent dissociation constants K C and K O ; α i and β i are the voltage-dependent rate constants for Ca 2+ free (i = 0) and Ca 2+ bound (i = 1) channels, respectively. Details of this four-state model are presented in the Supplemental Appendix. Continuous lines in Fig. 4c,d,e and in Supplemental Fig. 2b, obtained by a global fitting procedure, are in good agreement with the experimental data.
Besides the shift of the open probability towards negative voltages, cytosolic Ca 2+ increase also induced a voltage-dependent inhibition of the hTPC1 current, as shown in Fig. 4g. The biophysical interpretation of these data suggests that there is at least one site along the permeation pathway that favours Ca 2+ more than Na + binding. We again used a Woodhull approach 40 , and estimated that this Ca 2+ binding site was at an electrical distance of 0.39 from the cytosolic side and had an affinity for cytosolic calcium of 440 μ M at 0 mV (continuous lines in Fig. 4h; see also Supplemental Fig. 2c).

Discussion
In this work we examined in depth the functional characteristics of human TPC1 by using a novel heterologous system. First, we succeeded in expressing the channel in isolated vacuoles from mesophyll cells of Arabidopsis lacking the endogenous TPC1. Noteworthy, to obtain a measurable activity of the channel we had to wait at least two days after protoplast transformation. The phosphoinositide PI(3,5)P 2 was a powerful activator of hTPC1 with an apparent binding constant of about 15 nM, more powerful than for hTPC2, which in similar experimental conditions showed a two-fold larger binding constant 9 . The activation of TPC1 by PI(3,5)P 2 found here is in qualitative and quantitative agreement with the results of Cang et al. 32 . In similar agreement, we found TPC1 activity to be unaffected by NAADP. Evidence from other groups that TPC channel mediated Ca 2+ release from acidic organelles is stimulated by NAADP has been rationalized by the hypothesis of an accessory NAADP sensitive protein [41][42][43] . Such an accessory protein is unlikely to be present in the plant systems, which could explain the NAADP insensitivity of TPC1 observed here and that of TPC2 described earlier 9 .
Our work concentrated on the voltage-dependence, ion selectivity and the Ca 2+ regulation of hTPC1. The channel was found to be voltage dependent, with an apparent gating valence of about 1 elementary charge, similar to results reported by Cang et al. 32 . Currents are highly Na + selective with marginal K + permeability. Experiments in bi-ionic conditions showed that Ca 2+ can permeate through the channel even though its permeability is significantly lower than the one for Na + .
In addition the voltage-dependent Ca 2+ block observed at rather high cytosolic Ca 2+ indicates that Ca 2+ ions are able to enter the pore and bind with higher affinity than Na + to a site at an electrical distance of around 0.4, with an apparent affinity constant larger than 200 μ M at 0 mV. This binding site must be separated from the luminal end of the pore by a large energetic barrier to impede significant Ca 2+ flux.
In addition to the intrapore site, we discovered that TPC1 harbours a rather high affinity cytosolic Ca 2+ binding site whose occupation allosterically activates the channel by shifting its voltage-dependent characteristic towards negative voltages, and by accelerating the activation time course. Saturating at 20 μ M of [Ca 2+ ] cyt , this effect is compatible with physiological concentration changes of cytosolic Ca 2+ ; a simple mathematical model is able to quantitatively describe the measured shift together with the kinetic effects.
Finally, we showed functional evidence of a third Ca 2+ binding site on the luminal side of the channel. In this case the binding of Ca 2+ induced a decrease of the activity by shifting the open probability to positive voltages, with a Δ V half of about + 60 mV when [Ca 2+ ] lum was changed from 1 μ M to 1 mM. This might be physiologically relevant since the average luminal concentration of Ca 2+ has been reported to be around 500 μ M and it may vary from 1 mM to less than 1 μ M for example upon experimental manipulation of lysosomal pH 44 . Our mathematical model can be nicely extended to include both the cytosolic as well as the luminal Ca 2+ effect.
Recently the structure of the Arabidopsis plant counterpart of hTPC1, namely AtTPC1, was published 6,7 . Interestingly, like hTPC1, AtTPC1 is activated by cytosolic calcium 39,45 . Calcium binding is mediated by two EF-hands in the cytosolic loop linking the two monomers 46 . Since hTPC1 has no EF-hands, the mechanism of calcium regulation in the plant channel appears to be very different from that of human TPC1. Similarly, luminal calcium is also decreasing the activity of AtTPC1 by binding to an aspartate residue (D454) located in a luminal loop between transmembrane domains S7 and S8 47 . However, this amino acid is not conserved in hTPC1. Thus, the molecular basis of hTPC1 calcium modulation will require further investigation. On the contrary, the selectivity filter between the plant and the human TPC channels is enough conserved 48 . A recent work 48 identified two key residues for Na + -selectivity in the second pore loop and confirmed that human TPC channels are sodium channel with low calcium permeability.
Overall, our data show that calcium permeation mediated by hTPC1 channels provides a positive-feedback mechanism amplified by both  In this context Na + is actually a negative regulator since its release from the luminal to cytosolic side depolarizes the luminal membrane and limits Ca 2+ release. We can speculate that this interplay between Ca 2+ and Na + is necessary to generate a precise shape and dynamic of Ca 2+ release or, in other words, a defined Ca 2+ signature. Our conclusion is therefore that hTPC1 is a Na + and slightly Ca 2+ permeable, outwardly rectifying channel, working as a sensor of luminal and cytosolic Ca 2+ .

Materials and Methods
Plant material and protoplast transformation with human TPC1 sequence. The hTPC1 coding sequence was PCR amplified from HEK293 (human embryonic kidney) cells cDNA with the Thermo Scientific ® Phusion high fidelity polymerase using the following primers: 5′-CATGAAGCTTCATGGCTGTGAGTTTGGATGACGA-3′ and 5′ -CATGGAATTCCGCGGTAACGGTCTGGGAGCGCTGG-3′ (underlined are HindIII and EcoRI restriction sites used for cloning).
The PCR product was subsequently digested and ligated into the plant expression vector pSAT6-EGFP-N1 in the Multiple Cloning Site between the HindIII and EcoRI sites downstream of the 35SS strong constitutive promoter in frame with an enhanced GFP. The construct was verified by sequencing and identified as the human TPC1 which is the two pore calcium channel protein 1 isoform X1 coded by XM_011538492.1:246.2696 (mRNA transcript variant X4) that encodes for one of the shorter TPC1 isoforms (816 amino acids).
The standard pipette (luminal side) solution contained (in mM): 100 NaCl, 2 MgCl 2 , 10 MES, pH 5.5 (with NaOH). The standard bath (cytoplasmic side) solution contained (in mM): 100 NaCl, 10 Hepes, pH 7.5 (NaOH). In the selectivity experiments of Fig. 2, MgCl 2 2 mM was added in the cytosolic solution. NaCl 10 mM was obtained by substituting 90 mM of NaCl with equimolar concentration of KCl. In the K + -based bath solution of Fig. 3a,b, 100 mM NaCl was replaced with equimolar KCl. To investigate Ca 2+ -permeability, the luminal concentration of NaCl (100 mM) was substituted by 50 mM CaCl 2 . In this condition we observed that the presence of VRS as bath solution increased the quality of the seals. We measured a liquid junction voltage between the two solutions of 9.9 ± 0.6 mV and corrected offline (V = V applied − 9.9 mV). For the sake of clarity in Fig. 3d and Supplemental Fig. 1b,d, the voltage was approximated to the nearest integer. When calcium was not added in the ionic solutions, Ca 2+ concentration was determined by Plasma Emission Spectrometry (ICP-OES), instrument model Vista PRO Varian (Springvale, Australia), with the following main operating conditions: RF Power: 1100 W; Plasma gas flow rate: 15.0 L min −1 ; Sample uptake rate: 0.8 mL min −1 . Free calcium concentration in the presence of EGTA was calculated by using a dedicated program 52 . The osmolarity of the luminal and cytoplasmic solutions was adjusted to 550 mOsm and 600 mOsm, respectively, by the addition of D-sorbitol. Dithiothreitol (DTT; 2 mM) was added to the bath solution prior to the measurements 53 . DTT was prepared as 1 M stock solution the day of the experiment and stored in ice. PI(3,5)P 2 was purchased as dioctanyl ester (diC8) from AG Scientific or Echelon Biosciences Inc (USA). Other chemicals were purchased from Sigma-Aldrich (Italy, Germany). PI(3,5)P 2 and NAADP were prepared as 0.9 mM and 1 mM stock solutions respectively and stored at − 20°.
A total number of more than 95 vacuoles expressing hTPC1 (responding to PI(3,5)P2) were investigated in this work.

Data analysis.
Positive currents correspond to cations flowing from the cytoplasmic side of the vacuole to the lumen or anions moving in the opposite direction. Unless otherwise indicated, data are reported as mean ± sem. For the voltage dependent inhibition mediated by cytosolic calcium, we used the Woodhull model 40 . This approach assumes that the voltage dependence of the effect is due to the existence of a binding site for the ion along the electric field across the membrane. In terms of the model, the exponential voltage-dependence allows to determine the "electrical" distance of the binding site from the membrane surface. Data analysis and figure preparation were done with IgorPro software (Wavemetrics, Lake Oswego, OR, USA).