Discovery of lipophilic two‐pore channel agonists

Two‐pore channels (TPCs) have been a hot topic in recent literature. Their involvement in various diseases such as viral infections and cancer is of great interest for drug research. Due to their localization in the endolysosomal system and the lack of cell‐permeable activators, complex techniques were required for studying channel functions. Here, we review the first published lipophilic small‐molecule activators of TPCs. In independent high‐throughput screens, several new agonists were discovered, which now allow simple and fast investigation of TPCs in more detail in intact cells and in vivo. Zhang et al. identified tricyclic and phenothiazine antidepressants as TPC1 and TPC2 activators by screening a library of approved drugs. In contrast, Gerndt et al. screened an extensive compound library with mostly new chemotypes and drug structures. The latter resulted in two structurally distinct high‐affinity agonists, which are able to selectively activate TPC2 in either an NAADP‐ or PI(3,5)P2‐like manner. Here, we discuss the advantages and drawbacks of the identified molecules and their structural features. The versatility by which TPCs can be activated indicates many opportunities for future studies.

Despite this, TPCs have emerged in recent years as highly exciting potential novel drug targets for a number of diseases associated with the endolysosomal system. Thus, TPCs have been demonstrated to play a role in various infectious diseases such as Ebola filovirus, Middle East respiratory syndrome coronavirus (MERS-CoV), COVID-19 coronavirus, or HIV-1 retrovirus infections [16][17][18][19][20]. In addition, several bacterial toxins such as diphtheria toxin, anthrax toxin, cholera toxin, or pasteurella multocida toxin [21,22] have been shown to require functional TPCs for trafficking and release of the toxins into the cytosol.
In an attempt to identify more potent/efficacious variants of TPC2-A1-N (1) and TPC2-A1-P (2), several structure modifications were performed and the modified compounds subsequently tested. Surprisingly, none of the 46 modified versions of TPC-A1-N (1) showed significantly increased efficacies or potencies (Table 1). Replacing the p-trifluoromethyl group on the aniline side of the molecule (R 1 ; orange) with other electron-withdrawing groups in para-position did not cause significant changes, even the introduction of electron-releasing groups in para-position was tolerated to some extent (SGA-4 (5), SGA-84 (17)). The apparent enhancement of activity of SGA-85 (18) is explained by the fact that control cells showed increased levels of activation in these experiments as well [15]. For the substitution pattern of the acylated phenyl ring system (R 2 , green), meta-disubstitution patterns are most beneficial. More drastic changes in this aromatic region (replacement by methyl or pyrrole residues), as demonstrated for the approved drugs teriflunomide (26) and prinomide (27) as well as the 4-trifluoromethyl variant of prinomide (SGA-32; 28), led to a complete loss of activity ( Fig. 1). Teriflunomide (26) was introduced for the treatment of multiple sclerosis and prinomide (27) for rheumatoid arthritis [33,34]. TPC-A1-N (1) itself and some of its analogs bearing residues in para-position at the acylated aromatic ring are known anthelmintic agents [35]. In Table 2. Structure variations (selected examples) of TPC2-A1-P (2) and EC 50 values on TPC2. Even slight changes in structure of the original hit lead to decrease or loss of activity. Activity on the ion channel was determined by Fura-2-based single-cell Ca 2+ imaging experiments, as previously described [15]. cy, cyclohexyl. Similarly, modified versions of TPC2-A1-P (2) showed no improvement of efficacy [15]. In a collection of 20 analogs, prepared by systematic variation of the substituents, every change in structure resulted in a decrease or total loss of function ( Table 2). Analysis of structure-activity relationships revealed that the free carboxylic acid is essential for the activating effect, as the ester SGA-140 (32) is no longer active. Possibly, it might serve as a prodrug of TPC-A1-P in living systems, but this has not been investigated yet. Both the trifluoromethoxy and the bromine substituent at the phenyl ring are essential for activating TPC2, as exemplified by the inactive methoxy (SGA-55, 31) and des-bromo SGA-162 (37)  There are only few reports in the literature about the biological activities of TPC2-A1-P-like compounds. TPC2-A1-P (2) itself is mentioned as a precursor in  the synthesis of cannabinoid-1 receptor (CB1R) inverse agonists, whereas the final active compounds contained a carboxamide group instead of the free carboxylic acid function [36]. Phenylpyrrolecarboxamides derived from SGA-50 (30) bind to 5-HT 2A and 5-HT 2C receptors and also the 5-HT transporter, and were thus evaluated as antidepressant compounds [37].
In summary, the two high-throughput screening hits TPC2-A1-N (1) and TPC2-A1-P (2) can be regarded as strong chemical tools with the need of fully analyzing their pharmacological properties.

Activators of TPC1 and TPC2-voltage-dependent gating
Zhang et al. [14] have likewise used a calcium imagingbased high-throughput screening approach to identify activators of TPCs. This was despite their claim in 2012 that TPCs were not calcium-permeable but rather sodium-permeable channels [6]. In contrast to the 80 000 compound-strong Roche Xplore libraries used by Gerndt et al., Zhang et al. screened the Sigma LOPAC library which contains 1280 compounds. In this screening, 23 compounds induced calcium increases in TPC2-expressing cells but not in control cells expressing TRPML1 albeit in the plasma membrane. These hits from diverse chemical classes were further submitted to electrophysiological characterization using whole-cell recordings in TPC2 LL/AA -expressing HEK293 cells similar to those used by Gerndt et al. In this test, only five out of the 23 compounds showed significant currents, all of them belonging to the chemically closely related classes of dibenzazepinetype tricyclic antidepressants (TCAs) and phenothiazine-based antidepressants. Subsequently, other TCAs were also tested. In summary, the TCAs clomipramine (38), desipramine (39), imipramine (40), amitriptyline (41), and nortriptyline (42) (named LyNa-VA1.1 to LyNa-VA1.5 by the authors), as well as the phenothiazines chlorpromazine (43) and triflupromazine (44) (named LyNa-VA2.1 and LyNa-VA2.2), were found to activate TPC2. The EC 50 values were between 43 and 112 µM and thus approximately two orders of magnitude less potent than TPC2-A1-N/ P ( Fig. 2A,B). None of the compounds were found to activate TRPML1 but TCA-induced currents exhibited strong inward rectification, which is characteristic of TRPML channels (TPCs normally show no rectification upon activation). In a separate screen, Zhang et al. identified another compound, riluzole (45, Fig. 2D) which also activates TPC2 but showing linear currents typical for TPC2 as reported before [6,7,10,29]. Currents elicited with TCAs were strongly voltage-dependent while riluzole (45) activation was voltage-independent, suggesting that the voltage dependence of TPC2 can be unmasked by extrinsic agonists rather than being a fixed intrinsic property of the channel. What the origin of the proposed agonistmediated voltage dependence in the otherwise voltageindependent TPC2 is remains unclear. In contrast to Gerndt et al. [15] who identified two agonists (TPC2-A1-N (1) and TPC2-A1-P (2)) altering cation permeability in an agonist-dependent manner, the activators identified by Zhang [44]. (A) Using the experimentally resolved apo-state hTPC2 structure (grey, accession 6nq1) [44], we docked TPC2-A1-P (2, blue), clomipramine (38, pink), and chlorpromazine (43, yellow) to the channel. PI(3,5)P 2 (red) was added to its cryo-EM-resolved site. Residues forming polar bonds with the ligand are highlighted in red letters. PyMOL v2.3.4 was used to assemble the structure. AutoDockTools (ADT) version 1.5.6 Sep_17_14 was utilized to prepare the protein and ligand. The channel pore was excluded from docking analyses by drawing two grid boxes, each demarcating one half of the protein, preserving peripheral pockets. AutoDock Vina 1.1.2 was used to carry out the docking simulation (exhaustiveness = 200). Binding sites were visualized in PyMOL v2.3.4. Following identification of agonist-binding sites, 'sticky' sites were excluded from further analysis: TPC2 agonist classes were previously discovered [15], showing TPC2 to be gated by distinct mechanisms (PI(3,5)P 2 -like, NAADP-like gating). TCAs furthermore display another distinct, voltage-dependent gating mechanism. Since various modes of activation suggest distinct binding sites, sites of single-class binding were kept, and promiscuous binding sites excluded. Tricyclic antidepressants do not directly activate TPC2, rather rendering the channel voltage-gated. Subsequently, sites of individual agonist binding were removed, and binding sites where various activators bound maintained. Agonists were docked de novo within these binding pockets (30 9 30 9 30 A search space), rendering residues within 6 A of the docked agonist flexible. The following free energies were obtained by flexible docking (in kcal per mol): TPC2-A1-P (2, À7.4), clomipramine (38, À8.2), and chlorpromazine (43, À7.0). (B) Human and mouse TPCs were aligned using NCBI Protein BLAST to compare TPC2-A1-P-interacting residues. Residues found by docking to form polar bonds with TPC2-A1-P (bold) are fully conserved between human and mouse TPC2, but differ in TPC1. Red shade indicates positively charged residues, yellow polar residues, grey hydrophobic residues, and green glycine (no side chain). Dots indicate PI(3,5)P 2 -interacting residues (red) and charge transfer center arginines (black), previously described for HsTPC2 (above, [44]) and MsTPC1 [45].

5288
The found that some TCAs also activate TPC1 in a voltage-dependent manner, namely clomipramine (38) and desipramine (39), while the phenothiazine chlorpromazine (43) and riluzole (45) inhibit TPC1. Unfortunately, the authors did not perform systematic structure variations of the hit compounds. Thus, within the seven identified structures it is barely possible to analyze structure-activity relationships. Nevertheless, the dibenzazepine carbamazepine and native phenothiazine (Fig. 2C) did not activate TPC2, highlighting the necessity of the aminoalkyl side chain at the central ring of the tricyclic core. At this stage, more detailed structure-activity analyses would be desirable, as only slight changes in these structures are most likely to convert an activator into an inhibitor. The phenothiazines triflupromazine (44) and fluphenazine (46), recently published by Penny et al. [18], are a striking example for this phenomenon: Triflupromazine (44) activates TPC2, while fluphenazine (46) inhibits TPC2 currents evoked by PI(3,5)P 2 with an IC 50 of 8 µM (Fig. 3) [18].
Riluzole (45) blocks TTX-sensitive sodium channels, kainite receptors, and NMDA receptors [37][38][39][40]. At higher concentrations, it also strongly potentiates  (45) has neuroprotective effects and it is currently approved for the treatment of amyotrophic lateral sclerosis (ALS) [41]. Riluzole (45) is the only drug to prolong survival for ALS, and it is associated with a 35% reduction in mortality [42]. TCAs were introduced into clinics in the 1950s and are used to treat, for example, depression, bipolar disorder, panic disorder, chronic pain, and insomnia. TCAs inhibit monoamine (serotonin, norepinephrine, dopamine) reuptake and block cholinergic, histaminic, and alpha-adrenergic transmission. Although TCAs have a wide range of unwanted effects, they served as first-line treatment for depression for 30 years, until the selective serotonin reuptake inhibitors (SSRI) were introduced. Of note, amitriptyline (41), imipramine (40), and clomipramine (38) are also potent CYP450 inhibitors, significantly inhibiting CYP450 2C19 and 1A2 [43].

Modeling of TPC activators
The recently published TPC2 agonists from both groups [14,15] were docked to the apo-state hTPC2 structure [44] to compare their binding sites (Fig. 4). We previously demonstrated that the TPC2 K204A mutation located within the PI(3,5)P 2 binding pocket blocks TPC2-A1-P (2) activity. Accordingly, our docking results show TPC2-A1-P (2) to dock in close proximity to experimentally resolved PI(3,5)P 2 . Based on our docking results, K204 does not directly interact with TPC2-A1-P (2), but is likely still required to transduce TPC2-A1-P-evoked signals to the channel pore. The putative amino acids required for TPC2-A1-P binding are highlighted in Fig. 4. K143, N155, Q197, N198, and S200 highlighted in red appear to form hydrogen bonds with the docked agonist. In TPC1 (based on the alignment performed by She et al. [45]), these residues correspond to K143K (conserved), N155H, Q197V, N198D, and S200G. Thus, only one of the TPC2-A1-P-interacting amino acids is conserved in TPC1, while they are 100% conserved between human and mouse TPC2. This may potentially explain why TPC2-A1-P activates TPC2 but not TPC1. In contrast to TPC2-A1-P, TPC2-A1-N (1) did not dock within the PI(3,5)P 2 pocket. Supporting distinct binding sites, TPC2-A1-N (1) activity is unaffected by the TPC2 K204A mutation [15] and thus likely binds elsewhere on the channel. Similarly to TPC2-A1-N (1), riluzole (45) is also unaffected by the TPC2 K204A mutation [14]. Further investigations into the channel permeability evoked by riluzole (45) would be interesting, as its mechanism of action might mimic NAADP rather than PI(3,5)P 2 . The voltage-dependent TPC2 activators clomipramine (38) and chlorpromazine (43) share one binding site, located between IS5, ISPH, and IIS6. This site is particularly interesting, as the homologous site forms the target of agonists activating the distantly related TRPML channels-ML-SA1 for TRPML3 [46] and ML2-SA1 for TRPML2 [47]. While several of these findings still require further validation, our analysis identifies numerous binding pockets on the TPC2 protein, and highlights the various means by which TPC2 can be activated, each with different outcomes.
Comparing the results of both screenings, there is comprehensive knowledge on the pharmacological profiles (including undesired effects) of the repurposed TCA/phenothiazine-type TPC2 activators due to their long history in therapy, while the new activators identified by Gerndt et al. still need full pharmacokinetic and pharmacological characterization.
The two reviewed publications illustrate that researchers now have the opportunity to choose from an impressive and highly diverse collection of new lipophilic small-molecule activators for either TPC2 only or both TPC1 and TPC2 with the caveat that some of the compounds also block TPC1. With cellpermeable small-molecule activators, an important milestone has been reached as physiology and pathophysiology of TPCs can now be studied in more detail. Most importantly, the novel tools allow studies in intact cells and they may also be applicable for in vivo studies and perhaps even for therapeutic purposes.