A growth-factor-activated lysosomal K+ channel regulates Parkinson's pathology

Abstract

Lysosomes have fundamental physiological roles and have previously been implicated in Parkinson's disease^1-5. However, how extracellular growth factors communicate with intracellular organelles to control lysosomal function is not well understood. Here we report a lysosomal K⁺ channel complex that is activated by growth factors and gated by protein kinase B (AKT) that we term lysoK_GF. LysoK_GF consists of a pore-forming protein TMEM175 and AKT: TMEM175 is opened by conformational changes in, but not the catalytic activity of, AKT. The minor allele at rs34311866, a common variant in TMEM175, is associated with an increased risk of developing Parkinson's disease and reduces channel currents. Reduction in lysoK_GF function predisposes neurons to stress-induced damage and accelerates the accumulation of pathological α-synuclein. By contrast, the minor allele at rs3488217-another common variant of TMEM175, which is associated with a decreased risk of developing Parkinson's disease-produces a gain-of-function in lysoK_GF during cell starvation, and enables neuronal resistance to damage. Deficiency in TMEM175 leads to a loss of dopaminergic neurons and impairment in motor function in mice, and a TMEM175 loss-of-function variant is nominally associated with accelerated rates of cognitive and motor decline in humans with Parkinson's disease. Together, our studies uncover a pathway by which extracellular growth factors regulate intracellular organelle function, and establish a targetable mechanism by which common variants of TMEM175 confer risk for Parkinson's disease.

Extended Data Fig. 1 |. Lysosomal currents activated by growth factors.

Lysosomal currents were recorded using a ramp protocol (−100 mV to 100 mV ramp in 1 s with holding voltage of 0 mV) as illustrated in Fig. 1. a, Sizes of K⁺ currents (I_K) recorded at varying voltages (Ψ, −100 mV to 100 mV) from midbrain neurons with (stv) or without (fed) overnight starvation in DMEM containing no B27 nutrient supplement. Averaged I_K sizes (at 100 mV, recorded with 150 mM K⁺-containing bath) are in the right bar graph. Data are mean ± s.e.m. Numbers of recordings are in parentheses. *P < 0.0001, unpaired two-tailed t-test. b, Na⁺ currents recorded from hippocampal neurons with and without starvation demonstrating that B27 does not activate Na⁺ currents. The averaged current sizes (at −100 mV) are in the right bar graph (P = 0.226, unpaired two-tailed t-test). Solutions used in the recordings were the same as those used to record the I_K in Fig. 1a, b, except that K⁺ in the bath was replaced with Na⁺ and 1 μM PI(3,5)P₂ was added in the bath (lysosomal Na⁺ channel requires PI(3,5)P₂ for maximum activation). c, I_K recorded from TMEM175-transfected HEK293T cells starved in HBSS followed by refeeding with NGF (100 ng ml⁻¹ for 3 h) or BDNF (10 ng ml⁻¹ for 3 h), demonstrating that BDNF and NGF do not activate TMEM175 in HEK293T cells. Activation by insulin (right) was used as a positive control for receptor activation (see Fig. 1f for insulin). Averaged I_K sizes (recorded with K⁺-containing bath, at 100 mV) are in the right bar graph. d, I_K recorded from B27-replete midbrain and cortical neurons cultured from wild-type, heterozygous or TMEM175-knockout mice. Averaged I_K sizes (at 100 mV, recorded with 150 mM K⁺-containing bath) are in the right bar graph. Data are mean ± s.e.m. Numbers of recordings are in parentheses. *P < 0.001, unpaired two-tailed t-test. P values (compared to wild type) are as follows: midbrain, P = 0.0007 for heterozygous, P < 0.0001 for knockout; cortex, P < 0.0001 for knockout. e, I_K recorded from wild-type and TMEM175-knockout midbrain neurons before (fed) or after (stv) starvation (overnight in DMEM). An AKT activator SC79 (10 μM) was applied to the recording bath during some of the recordings (blue and red traces). Bar graphs in the right show averaged I_K sizes (at 100 mV). Data are mean ± s.e.m. Numbers of recordings are in parentheses. Arrows are used to indicate curves that overlap and are not easily distinguished.

Extended Data Fig. 2 |. AKT is required for the reconstitution of human TMEM175 as a functional ion channel.

a–f, Whole-cell currents were recorded from Drosophila S2 cells without (a–c, h) or with (d–f) endogenous Akt (‘dAKT’) knocked down using dsRNA treatment. a, d, Nontransfected S2 cells. b, e, h, S2 cells transfected with human TMEM175 alone. c, f, S2 cells cotransfected with human TMEM175 and human AKT1. A western blot showing the reduction of Drosophila Akt protein by dsRNA treatment is shown in the inset in d. For gel source data, see Supplementary Fig. 1. g, Summary of current amplitudes (at −100 mV, recorded in K⁺-containing bath with or without SC79). Recordings were done using a ramp protocol (−100 mV to 100 mV in 1 s, V_h = 0 mV). Bath solution contained 150 mM K⁺, 150 mM K⁺ with SC79 (an activator of human AKT) or 150 mM NMDG with SC79, as indicated. In a, b, d, e, arrows are used to indicate curves that overlap and are not easily distinguished. Black, recorded from bath containing K⁺; blue, recorded from bath containing K⁺ and SC79; red, recorded from bath containing NMDG and SC79. Data are mean ± s.e.m. Numbers of recordings are in parentheses. In g, *P ≤ 0.05 (compared with those from cells cotransfected with human TMEM175 and human AKT). P values (unpaired two-tailed t-tests) are as follows: without dsRNA treatment, P < 0.0001 for TMEM175 vs nontransfected, TMEM175+AKT vs nontransfected and TMEM175+AKT vs TMEM175; with dsRNA treatment, P = 0.1253 for TMEM175 vs nontransfected, P < 0.0001 for TMEM175+AKT vs TMEM175 and TMEM175+AKT vs nontransfected. h, Whole-cell currents recorded from S2 cells transfected with human TMEM175 alone, with (middle) or without (left) recombinant human AKT1 protein (1 μg ml⁻¹) included in the pipette solution. Right, summary of current amplitudes (at −100 mV; b and g show an additional control without AKT protein in pipette. We further tested the necessary and sufficient role for AKT in I_TMEM175 by reconstituting TMEM175 in the Drosophila S2 cell line, which does not have endogenous TMEM175. In HEK293T cells, cotransfection of AKT1 did not further increase I_TMEM175 (Extended Data Fig. 6a), consistent with the idea that the three mammalian AKTs are already abundantly expressed in the cells. S2 cells do not have TMEM175 (which is absent in insects), and have only one Akt gene, making it easier to knock down. In addition, the Drosophila Akt (dAKT) protein is not well-conserved with the mammalian AKTs (61% identity to human AKT1) and is less likely to support human TMEM175 function. For technical reasons, we were unable to enlarge S2 cell lysosomes suitable for patch-clamp recording. We therefore recorded whole-cell currents as an assay for functional channels. Transfecting human TMEM175 alone into S2 cells generated little current (about −20 pA at −100 mV) above the level of endogenous K⁺ currents (a, b, g). In addition, knocking down Drosophila Akt with dsRNA completely eliminated the small I_TMEM175 (d, e, g). Cotransfecting human TMEM175 with a human AKT1 (hAKT1), however, led to robust I_TMEM175, either with or without the endogenous Drosophila Akt knocked down (c, f, g). The currents were potentiated by SC79 in human AKT cotransfected cells but not in those transfected with TMEM175 alone, supporting the notion that SC79 acts on mammalian AKTs to open the channel. We finally tested whether purified recombinant AKT protein is sufficient to activate TMEM175. In S2 cells transfected with human TMEM175 alone (which had minimal current), application of SC79 into the bath increased the currents to −417.0 ± 53.9 pA (h) (n = 3, −100 mV) when human AKT1 protein was introduced into cytosol via pipette solution dialysis. Together, our data suggest that AKT is obligatory for a functional mammalian TMEM175 channel.

Extended Data Fig. 3 |. mTOR is not required for the activation of TMEM175 by AKT.

a–c, Lysosomal I_K currents were recorded from TMEM175-tranfected HEK293T cells with mTOR (a component of both mTORC1 and mTORC2 complexes) (a), Raptor (an mTORC1 component) (b) or Rictor (an mTORC2 component) (c) knocked down with short hairpin (sh)RNA lentivirus. d, Lysosomal I_K recorded from TMEM175-tranfected HEK293T cells treated with rapamycin (2 μM, 1 h). Figure 1f provides comparisons with those with no knockdown or rapamycin treatment. *P ≤ 0.05. P values (unpaired two-tailed t-tests; K vs K+SC79 groups) are: P < 0.0001 (mTor shRNA), P = 0.0013 (Raptor shRNA), P < 0.0001 (Rictor shRNA). e–h, Control for mTOR knockdown. Lysosomal Na⁺ currents were recorded from HEK293T cells transfected with TPC2 as positive controls for mTOR knockdown. TPC2 is a lysosomal Na⁺ channel that is inhibited by ATP. The ATP sensitivity requires mTORC1 but not mTORC2; knocking down mTOR or Raptor—but not Rictor—reduces the ATP sensitivity. In f, arrows are used to indicate traces that overlap and are not easily distinguished. Black, recorded before 1 mM ATP was added to the bath; red, recorded after 1 mM ATP was added to the bath. Data are mean ± s.e.m. Numbers of recordings are in parentheses.

Extended Data Fig. 4 |. Characterization of TMEM175–AKT interaction and its role in TMEM175 activation.

a, Association between heterologously expressed TMEM175 and AKT. HEK293T cells were transfected with GFP-tagged wild-type or mutant TMEM175 and HA-tagged human AKT1 (HA–AKT), as indicated. Lysates were immunoprecipitated with anti-GFP and blotted with anti-GFP or with anti-AKT. Whole-cell lysate was also blotted for input control. b–d, Experiments demonstrating that S241 is not essential for TMEM175 activation. b, Sequence alignment of the region around S241 of TMEM175. GenBank access numbers are in parentheses. c, Lysosomal I_K recorded from HEK293T cells transfected with S241A/T338A or S241A substitutions. Averaged I_K sizes (at 100 mV) are shown in the bar graphs. Data are mean ± s.e.m. Numbers of lysosomes recorded are in parentheses. d, Association between wild-type or S241A-mutant TMEM175 and AKT. HA-tagged AKT was transfected alone (lane 1) in HEK293T cells or together with GFP-tagged wild-type or S241A-mutant human TMEM175 (lanes 2 and 3). Whole-cell lysates were blotted with anti-GFP (bottom) or anti-HA (middle), or immunoprecipitated with anti-GFP followed by immunoblotting with anti-HA (top). In c, arrows are used to indicate curves that overlap and are not easily distinguished. Black, recorded from bath containing K⁺; blue, recorded from bath containing K⁺ and SC79. e, Mapping the AKT-interacting region to the second repeat of TMEM175. A diagram illustrating the GFP-tagged TMEM175 fragments used in the experiments is shown. Lysates from HEK293T cells transfected with HA-tagged AKT1 (HA–AKT) alone (lane 1) or together with GFP-tagged TMEM175 fragments (lanes 2–6) were blotted with anti-GFP (bottom) or anti-HA (middle) or immunoprecipitated with anti-GFP followed by blotting with anti-HA (top). f, Phosphorylation status of T308 and S473 of TMEM175-associated AKT. HEK293T cells were transfected with GFP-tagged TMEM175 (with nontransfected cells as control). Endogenous AKT associated with TMEM175 was pulled down with immunoprecipitation using anti-GFP and was probed with antibodies against total AKT (pan-AKT), AKT phosphorylated at T308 (pAKT(T308)) or at S473 (pAKT(S473)). Total cell lysates (input) before immunoprecipitation were also loaded for comparison. g, Association between TMEM175(M393T) and TMEM175(Q65P) and AKT. Lysates from HEK293T cells transfected with GFP-tagged wild-type or mutant TMEM175 (with nontransfected cells as control) were blotted with anti-GFP, anti-AKT or immunoprecipitated with anti-GFP followed by blotting with anti-GFP or anti-AKT. For gel source data, see Supplementary Fig. 1.

Extended Data Fig. 5 |. AKT activates TMEM175 in a catalytic activity-independent manner.

a, Schematic of the canonical AKT activation, with inhibitors used to target each step indicated, modified from previous publications^,. A domain structure of AKT is in the bottom. AKT has a PH domain (PH), a kinase domain (KD) and a C-terminal tail regulatory domain (RD). b, Phosphorylation status of AKT and its two canonical targets GSK3β and PRAS40 following treatments similar to those used in the patch-clamp recording of TMEM175. Lysates from human TMEM175-transfected cells were probed with antibodies as indicated. Treatment conditions were: starvation (0 h, 4 h in HBSS) without (lanes 1 and 2) or with insulin (lane 3, 100 nM, 4 h) treatment afterwards, wortmannin (20 μM, 2 h) with and without SC79 (10 μM, 30 min) treatment afterwards (lanes 5 and 6), BKM120 (10 μM, 3 h; lane 8), GSK2334470 (20 μM, 3 h; lane 10), ARQ-092 (20 μM, 3 h; lane 12), MK2206 (20 μM, 3 h; lane 14), inhibitor VIII (10 μM, 3 h; lane 16), afuresertib (20 μM, 3 h; lane 18), AZD-5363 (20 μM, 3 h; lane 20), GSK690693 (10 μM, 4 h; lane 22). For gel source data, see Supplementary Fig. 1. c–i, Lysosomal I_K was recorded from TMEM175-transfected HEK293T cells pretreated with wortmannin (c), BKM120 (d), GSK2334470 (e), GSK690693 (f), AZD-5363 (g) or afluresertib (h) as in b before lysosomal dissection for recording. Averaged I_K sizes (at 100 mV) are shown in i. Data are mean ± s.e.m. Numbers of recordings are in parentheses. Our finding that SC79 fully activates TMEM175 in starved lysosomes was surprising but is also consistent with the idea that channel activation is independent of the kinase activity of AKT. Upon receptor activation by growth factors, PI3 kinases (PI3Ks) increase the concentration of PIP₃ from PtdIns(4,5) P₂ (PIP₂). SC79 and the endogenous AKT activator PIP₃ both bind to the PH domain of AKTs to induce large conformational changes that lead to the release of the PH domain from the catalytic domain and to an open state of the kinase, which renders AKT accessible to phosphorylation at a key residue in the kinase domain (T308 in human AKT1) by PKD1 and another residue (S473) in the C-terminal regulatory domain by kinases such as mTORC2. Phosphorylation of T308 and S473 is required for the efficient activation of AKT kinase activity to phosphorylate downstream targets such as GSK3β and PRAS40 (a). In nutrient-replete cells, TMEM175-associated AKT is at least partially phosphorylated at the two sites (Extended Data Fig. 4f). We used a panel of AKT inhibitors and mutants to probe the involvement of each of the steps in AKT function in TMEM175 activation. In starved cells and wortmannin-treated ones, AKT kinase activity is minimal because of the lack of phosphorylation of T308 on AKT by PKD1. Indeed, the activation of AKT kinase activity by SC79 on AKT isolated from starved cells requires the addition of PKD1 and ATP. Similarly, SC79 does not induce substantial AKT kinase activity (assayed with target phosphorylation) in wortmannin-treated cells (b). However, SC79 maximally activated TMEM175 in lysosomes isolated from starved cells (Fig. 2d, h) and in those from wortmannin-treated cells (c, i), without PKD1 or ATP in the bath, supporting that the SC79-induced channel activation does not require the kinase activity of AKT. Reduction of PIP₃ by starvation and by inhibition of PI3K with wortmannin or BKM120 (a specific PI3K inhibitor)—treatments expected to keep AKT in a PH-in a closed state—both reduced the phosphorylation of AKT at T308 and S473 and that of the downstream AKT targets GSK3 (S9) and PRAS40 (T246) (b). The treatments also eliminated I_TMEM175, which was restored by SC79 application (b–d, i). These data suggest that AKT activates channel after it switches from the PH-in closed state to the PH-out open state. To further test whether—after AKT switches to a PH-out open state—phosphorylation of T308 in AKT by PDK1 is required for I_TMEM175, we inhibited PDK1 with GSK2334470. Although the inhibitor abolished the phosphorylation of AKT and its downstream targets, it had little effect on I_TMEM175 (b, e, i). Catalytic and allosteric AKT inhibitors both block phosphorylation of downstream targets by AKT, but via different mechanisms (a). Unlike allosteric inhibitors, which probably keep AKT in a closed phosphorylation-resistant state and prevent the activation of TMEM175 (b, Fig. 2f), catalytic inhibitors such as those that compete with ATP binding in the activation loop are downstream of the phosphorylation of AKT T308 and S473 and inhibit AKT kinase activities in an open state. The ATP competitive inhibitors GSK690693, AZD-5363 and GSK2110183 (afuresertib) all inhibited AKT target phosphorylation (b), but—in contrast to the allosteric inhibitors (Fig. 2f)—had little effect on I_TMEM175 (f–i). Together, the inhibitors targeting different steps of AKT activation suggest that conformational changes that lead to AKT in an open state, but not the downstream steps of the catalytic activities of AKT, are required for I_TMEM175.

Extended Data Fig. 6 |. AKT determines the sensitivity of TMEM175 to starvation and activation by phosphatidylinositol lipids.

a–e, HEK293T cells (a, c–e) (lysosomal recording) or S2 cells (b) (whole-cell recording) were transfected with human TMEM175 with or without cotransfection of wild-type or mutant human AKT1. a, b, Experiments demonstrating that AKT with mutations in the kinase domain (K179M alone, T308A/S473A double mutant) are able to support I_TMEM175 as well as wild-type AKT. Extended Data Figure 2 provides a comparison with those from S2 cells transfected with TMEM175 without human AKT1 cotransfection, in which very little currents were generated. A summary of current amplitudes (at −100 mV) is given in the bar graphs. The right panel in a shows the phosphorylation status of transfected AKT and AKT targets GSK3β and PRAS40. The human AKT1 used for transfection has an HA tag and a linker sequence, and can be distinguished from the endogenous AKT (indicated by green arrows) because of its slightly higher molecular weight. For gel source data, see Supplementary Fig. 1. In c–e, stv indicates cells starved in HBSS medium for 2 h before lysosomal dissection for recording. c, PIP₃ (1 μM) was applied to the recording bath during some of the recordings (blue and red traces). d, e, I_TMEM175 recorded from lysosomes with or without cotransfection of the E17K (d) or the D323A/D325A (DA) (e) AKT1 mutants, with or without starvation. PIP₂ in d was applied at 1 μM. Data are mean ± s.e.m. Numbers of recordings are in parentheses. Arrows are used to indicate curves that overlap and are not easily distinguished. f, Positions of the AKT1 mutations used in this figure. We used AKT mutants to further test the nonessential role of kinase activity in I_TMEM175. AKT1 with mutation at K179 (a key residue in the ATP- and Mg²⁺-binding pocket; K179M), or at T308 and S473 (residues key to AKT kinase activation; T308A/S473A) no longer phosphorylates its targets^,. I_TMEM175 from cells co-expressing the mutants, however, was similar to that co-expressing wild-type AKT (a). More importantly, when co-expressed with human TMEM175 in S2 cells, the kinase-inactive mutants fully supported I_TMEM175 in a manner indistinguishable from that of the wild type (b). We then tested whether AKT conformational changes are sufficient to activate the channel. Under physiological conditions, PIP₃ generated on the plasma membrane opens AKT by binding to the PH domain. In lysosomes from starved cells, PIP₃ bath application activated I_TMEM175 (c). The E17K mutation in the PH domain—commonly found in cancers—switches the lipid sensitivity of AKT from PIP₃ to PIP₂, a lipid that is abundant even during starvation. Notably, I_TMEM175 from cells cotransfected with AKT(E17K) was resistant to starvation, and, unlike that without cotransfection, was activated by PIP₂ (d). Thus, AKT also determines the lipid sensitivity of TMEM175. The double mutation D323A/D325A of AKT1 disrupts the locking of the PH domain to the kinase domain and keeps AKT in a constitutively open state^,. I_TMEM175 from cells co-expressing this mutant was resistant to starvation, maximally activated and not further activated by SC79 (e).

Extended Data Fig. 7 |. Association of AKT to TMEM175 is required to maintain I_TMEM175 after the channel is activated.

a, b, Development of peptide inhibitors that dissociate AKT from TMEM175 in the lysoK_GF complex. a, Sequences of synthesized human TMEM175 peptides around the K336 region that interacts with AKT. Peptide 2 has a K336A mutation. b, To test the ability of each peptide to dissociate AKT from TMEM175, cell lysates from HEK293T cells cotransfected with GFP-tagged human TMEM175 and HA-tagged human AKT1 were incubated with each peptide (10 μM) for 2 h before immunoprecipitation with anti-GFP to bring down TMEM175 and immunoblotting with anti-HA to detect the associated AKT1. Immunoprecipitation without peptide incubation (lane 1) was used as a control. Whole-cell lysate was also blotted with anti-HA and anti-GFP for input control. For gel source data, see Supplementary Fig. 1. The results show that when applied in cell lysates containing the TMEM175–AKT complex, peptides as short as 10 amino acids (peptide 5) in the K336 region were sufficient to dissociate AKT from the preassembled AKT–TMEM175 complex in the immunoprecipitation assay. As a control, K336A mutation (peptide 2) abolished the dissociation ability of the peptide. c, d, Lysosomal I_K were recorded from TMEM175-transfected HEK293T cells (c) and mouse hippocampal neurons (d). After SC79 application to maximally activate lysoK_GF, peptide inhibitor (10 μM) was added to the bath and currents were recorded for about 10 min until they reached steady states. The K336A mutant peptide 2 was used as a negative control. Averaged I_K sizes (at 100 mV) are shown in the bar graphs. In c subpanel ‘peptide 2’, arrows are used to indicate curves that overlap and are not easily distinguished. Blue, from bath containing K⁺ and SC79 before peptide application; red, from bath containing K⁺ and SC79 after peptide application. Data are mean ± s.e.m. Numbers of recordings are in parentheses. *P ≤ 0.05. P values (unpaired two-tailed t-tests) (K+SC79 vs K+SC79 + peptide groups) are as follows. In c, P = 0.8447 for peptide 2, P < 0.0001 for the other treatments; in d, P = 0.0011 for peptide 5. The results show that after the channel was maximally activated by SC79, bath application of AKT-association peptide inhibitors—but not the one with the K336A point mutation—diminished both heterologously expressed I_TMEM175 in HEK293T cells and the native currents in neurons. Given that the bath contained no phosphatase, the action of the peptides is unlikely through a dephosphorylation of the channel target but is rather through the dissociation of AKT from TMEM175—further supporting the idea that it is the conformational change of AKT, and not target phosphorylation, that activates the channel. e, A model for TMEM175 activation by AKT conformational changes. TMEM175 without AKT or associated with inactive AKT in its PH-in closed conformation is closed. The channel opens upon AKT conformational changes triggered by physiological activators such as extracellular growth factors, lipids, or synthetic small molecules such as SC79. The gating mechanism does not require AKT kinase activity. AKT mutations found in cancers that alter the interaction between the PH and kinase domains also alter TMEM175 channel property.

Extended Data Fig. 8 |. Knock-in mice to model common human variants p.M393T and p.Q65P.

a–d, Experiments demonstrating that the TMEM175(I390T) substitution in mice—corresponding to M393T in human TMEM175—leads to reduced lysosomal I_K, accelerated spreading of pathogenic α-syn and increased neuronal damage. a, Sanger sequencing of genomic DNA PCR products from wild-type, heterozygous and homozygous TMEM175(I390T) knock-in mice in the I390-coding region. Residue I390 in mouse TMEM175 is equivalent to M393 in human TMEM175, as shown in the sequence alignment on the right. b, I_K recorded from wild-type (+/+, wild type), heterozygous (I390T/+, het) and homozygous (I390T/I390T, hom) knock-in mouse midbrain neurons. Bar graphs show averaged I_K (at 100 mV). Data are mean ± s.e.m. Numbers of recordings are in parentheses. c, Neuronal damages in wild-type, het and hom midbrain neurons treated with MPP⁺ with concentrations as indicated added to culture medium for 12 h and those followed by incubation with insulin (100 nM in DMEM) for 3 h. Damage index was calculated as the ratio of damaged area to the total neuronal area. d, Cultured wild-type and TMEM175(I390T) (hom) midbrain neurons were seeded with mouse α-syn PFFs and immunostained with anti-pSer129 α-syn (pSyn, red) two weeks later. β-Tubulin (green) was costained to highlight neuronal processes. DAPI (blue) nuclear staining was used to identify number of cells. Left two subpanels, representative images. Right, pSyn signals normalized to number of cells (n = 8 coverslips each). For c, d, see Fig. 5, Extended Data Fig. 10 for more details. P values in c are as follows (unpaired two-tailed t-tests, compared to wild type): 0 μM MPP⁺, P = 0.9318 (het), P = 0.2516 (hom); P < 0.0001 for all the other comparisons. The P value in d (t-test, wild type versus hom) is P < 0.0001. e–h, Characterization of the rs34884217 (p.Q65P) variant. e, Genomic structures of the human and mouse TMEM175 genes around the rs34884217 SNP region. The A-to-C variation in human and the corresponding substitution in the knock-in mouse line are indicated by arrows. f, Agarose gel electrophoresis showing RT–PCR products of the whole open reading frame cDNA (about 1.5 kb) amplified from wild-type and knock-in mouse brains. For gel source data, see Supplementary Fig. 1. g, Sanger sequencing results of the RT–PCR products in f, demonstrating that the A-to-C substitution in mice leads to a Q65P coding change. h, Sequence alignments around Q65 in the I-TM1–TM2 linker. Q65P is located at the luminal side of the channel entrance.

Extended Data Fig. 9 |. TMEM175 contributes to lysosomal pH stability and membrane integrity, regulates lysosomal enzyme activity and controls organelle fusion during autophagy.

a, Lysosomal pH from wild-type and TMEM175-knockout cultured mouse hippocampal neurons with or without starvation (overnight in DMEM). Distribution of pH values are in the first two subpanels. Average values are in the bar graph. Numbers of lysosomes analysed are in parentheses. b, Comparison of mature CatD levels using anti-CatD antibody. Lysate from cultured wild-type or TMEM175-knockout hippocampal neurons, with or without starvation, were probed with anti- CatD, or anti-β-actin for loading control. The mature form of CatD (bottom band) is cleaved from pro-CatD (top band) and the cleavage is pH-dependent. Left, representative blot. Right, signal density of mature CatD normalized to that of pro-CatD (bottom) or to that of β-actin (top). n = 5 repeats. *P ≤ 0.05. P values (unpaired two-tailed t-test, wild type vs knockout) are as follows: normalized to β-actin, P = 0.0119 (without starvation), P = 0.0011 (with starvation); normalized to pro form, P = 0.0002 (without starvation), P = 0.0015 (with starvation). For gel source data, see Supplementary Fig. 1. c, Comparison of mature CatD levels using BODIPY FL pepstatin A (a fluorescent cathepsin indicator that binds to CatD in a pH-dependent manner). Cultured wild-type and TMEM175-knockout hippocampal neurons, with or without starvation, were treated with BODIPY FL pepstatin A (1 μM, 1 h) and imaged for pepstatin A and nuclei. Left, representative images of pepstain A (green), DAPI (blue) and merge between the two. Right, pepstain A signals normalized to numbers of cells (n, number of coverslips imaged). d, Autophagosome–lysosome fusion assay with an mCherry- and GFP-tagged LC3 (tandem tag, illustrated) transfected into wild-type and TMEM175-knockout hippocampus neurons with or without starvation. Left, representative images for GFP (indicative of autophagosome puncta before fusion with lysosome; GFP is invisible after fusion with lysosome because of its pH sensitivity), RFP (indicative of total number of autophagosome puncta before and after phagosome–lysosome fusion; RFP is visible under both neutral and acidic pH) and DAPI (nucleus), and merge among the three. Bar graph indicated averaged numbers of GFP and RFP puncta upon treatments as indicated (middle) and numbers of GFP-positive puncta normalized to those of RFP-positive ones (right). Number of cells analysed are in parentheses. e, Experiments demonstrating that deficiency in lysoK_GF leads to accelerated spreading of pathogenic α-syn and lysosomal permeability in neurons. Cultured wild-type and TMEM175-knockout hippocampal neurons were seeded with mouse α-syn PFFs and costained with anti-pSer129 α-syn (pSyn, red) and anti-galectin 3 (green) two weeks later. DAPI (blue) nuclear staining was used to identify the number of cells. Representative images are in the left panels and averaged pSyn and galectin 3 signal density (normalized to number of cells) are in the right bar graph (n = 6 coverslips each). ***P ≤ 0.001. P values (unpaired two-tailed t-tests; knock out versus wild type) are P < 0.0001 for α-syn and galectin 3. In the bar graphs, data are represented as mean ± s.e.m.

Extended Data Fig. 10 |. LysoK_GF protects mouse neurons from stress-induced damage in a gene-dosage-dependent manner, and a loss-of-function variant in the human TMEM175 gene is nominally associated with increased rates of cognitive and motor function decline in patients with Parkinson’s disease.

a–e, Cultured mouse neurons were challenged with MPP⁺ toxin, B27 deprivation (in DMEM) and/or H₂O₂ treatment. Images were taken for damage analysis. Blebs and fragmentation in neuronal processes were counted as damage, as indicated by arrows in representative pictures in a. The damage index was calculated as the ratio of damaged area to the total neuronal area. a, Treatment of wild-type and TMEM175-knockout midbrain neurons with MPP⁺ added to culture medium for 12 h, with concentrations as indicated. Left, representative images of neurons with bleb damages indicated by arrows. Right, damage index. b, c, Hippocampal (b) or midbrain (c) neurons were starved in DMEM for duration as indicated. H₂O₂ (100 μM) treatment for the period indicated was carried out in DMEM after starvation. d, Effects of insulin. In the group of neurons treated with 14-h starvation and insulin, without H₂O₂, neurons were starved for 14 h in DMEM followed by incubation with insulin (100 nM in DMEM) for 3 h. In the groups treated with H₂O₂, neurons were starved in DMEM for 0 or 14 h, followed by a 3-h incubation with or without insulin (100 nM). During the last hour of incubation, H₂O₂ (100 μM) was added to the medium. e, Comparison between wild-type and TMEM175(Q65P)-carrying neurons with treatments similar to those in c. In the bar graphs, data are mean ± s.e.m.; numbers of neurons analysed are in parentheses. *P ≤ 0.05. P values (unpaired two-tailed t-tests) are as follows. In a, wild type versus knockout: P < 0.0001 for all comparisons. In b, wild type versus knockout: P = 0.0192 (0 h, starved), P = 0.0004 (3 h, starved), P < 0.0001 for the other comparisons. In c, wild type versus knockout: P = 0.0017 (0 h + H₂O₂), P < 0.0001 for the other comparisons. In d, compared with wild type: P = 0.264 (HET, 0 h), P < 0.0001 for the other comparisons. In e, compared with wild type: P = 0.8121 (+/Q65P, 0 h), P = 0.545 (Q65P/Q65P, 0 h), P = 0.0935 (+/Q65P, 3 h), P = 0.2012 (Q65P/Q65P, 3 h), P < 0.0001 for the other comparisons. f, Clinical effect of the rs34311866 (p.M393T) variant in the human TMEM175 on longitudinal course of Parkinson’s disease. Longitudinal effects for carriers of no (blue), one (green), or two (red) p.M393T variants are shown for mixed-effects linear regression models without multiple testing correction; 95% confidence intervals are indicated by coloured bands. In the UPenn cohort, p.M393T carriers had faster cognitive decline (two-tailed P = 0.005, as captured by lower DRS-2 scores) and faster motor decline (two-tailed P = 0.032, higher UPDRS-3 scores). In the PPMI cohort, p.M393T carriers trended (two-tailed P = 0.066) towards faster motor decline, although no statistical significance was reached (presumably because of shorter observation period, of 5 years after diagnosis). As TMEM175 variants are associated with neurodegenerative diseases and AKT is known to protect neurons against stress-induced damage^,,, we tested whether lysoK_GF protects neurons. To do this, we used a neuronal culture model in which damage induced by insults such as the neurotoxin MPTP, H₂O₂ and nutrient removal can be quantified well^–. MPTP interferes with mitochondrial metabolism, inducing parkinsonism in humans and in animal models. Consistent with previous findings^,, incubating midbrain neurons with MPP⁺ (the active form of the neurotoxin) led to perturbations and beading of neuronal processes in a dose-dependent manner. When the size of such damaged areas normalized to the total neuronal area (that is, damage index) was quantified, TMEM175-knockout neurons exhibited much more severe damage compared to wild type (a). Both nutrient removal and reactive oxygen species (such as H₂O₂) cause neuronal damage. Under each combination of H₂O₂ treatment after B27 removal, TMEM175-knockout neurons exhibited more severe damage than wild type (b, c). Growth factors such as insulin are known to mitigate neuronal damage but the mechanisms are not well-understood. In wild-type neurons, insulin reduced the damage induced by B27 removal and/or H₂O₂ treatment. Notably, this protection was nearly absent in TMEM175-knockout neurons (d), which suggests an essential role for TMEM175. In heterozygous neurons, the degree of damage was intermediate between that of wild-type and TMEM175-knockout neurons (d). Thus, Tmem175 gene dosage has an incremental effect on neuronal protection. Decreasing I_TMEM175 by about 50% is sufficient to compromise the protection. In support of this finding, neurons from TMEM175(I390T) knock-in mice were also more prone to MPP⁺-induced damage than were neurons from wild-type mice (Extended Data Fig. 8c). We also tested whether p.Q65P provides additional protection. When challenged with B27 deprivation and H₂O₂, neurons from both heterozygous (Q65P/+) and homozygous (Q65P/Q65P) mice had much less damage than those from wild type (e). The difference was even more notable between TMEM175(Q65P)-carrying neurons and TMEM175-knockout neurons. For example, after a 14-h starvation the TMEM175-knockout neurons had 27% damage and wild-type neurons had 15%—but those from Tmem175_Q65P/+ mice had only 6%. Similarly, H₂O₂-induced damages were mitigated in TMEM175(Q65P)-carrying neurons (c, e). These data suggest that p.Q65P is neuroprotective and constitutes a potential mechanism that underlies a reduced susceptibility to developing Parkinson’s disease.

Fig. 1 |. A growth-factor-activated lysosomal K⁺ channel, lysoK_GF.

a, Schematic of the recording of neuronal lysosomes. b, Currents (I) recorded at varying voltages (Ψ) from mouse hippocampal neurons with (starved) or without (fed) overnight starvation in DMEM containing no B27 nutrient. c, Currents from neurons with starvation followed by refeeding (in DMEM medium) with insulin (100 nM for 4 h), NGF (100 ng ml⁻¹, 3 h) or BDNF (10 ng ml⁻¹, 3 h). d, Reconstituting lysoK_GF by TMEM175 transfection in HEK293T cells. Currents were recorded before, 2 h after starvation or after a 2-h starvation followed by refeeding with amino acids (10× for 10 min), with serum (10%, 4 h), with serum inactivated by boiling for 10 min or with insulin (100 nM, 4 h). e–i, Comparison between lysoK_GF from wild-type and TMEM175-knockout neurons. e, Sequencing of the knock out, showing deletion of parts of exons 3 and 4. f, Total brain proteins (100 μg) from wild type (WT), heterozygous (het) and homozygous TMEM175-knockout (KO) mice were immunoblotted with anti-TMEM175 and—as a loading control—reblotted with anti-actin. For gel source data, see Supplementary Fig. 1. g, Currents from B27-replete hippocampal neurons. h, Currents from B27-starved (overnight in DMEM) knockout neurons refed with insulin (hippocampal neurons), NGF (hippocampal) or BDNF (midbrain). Black (K) and red (N-methyl-D-glucamine (NMDG)) traces are currents recorded with bath solutions containing K⁺ or NMDG, respectively. Averaged I_K sizes (at 100 mV, with K⁺-containing bath) are in bar graphs (b–d, i). Data are mean ± s.e.m. Numbers of recordings are in parentheses. P values shown are from unpaired two-tailed t-tests.

Fig. 2 |. AKT is necessary and sufficient to activate TMEM175.

a–f, Currents from TMEM175-transfected HEK293T cells. In b, c, a dominant-negative AKT (triple-mutant AKT(K179M/T308A/S473A); AKT(KD3)) was cotransfected. In a, cells were pretreated with wortmannin (20 μM, 2 h), an allosteric AKT inhibitor (MK-2206 (20 μM, 3 h) or ARQ-092 (20 μM, 3 h)) or AKT inhibitor VIII (10 μM, 3 h) before lysosomal dissection for recording. In a, b, e, cells were not starved; in c, d, cells were starved in Hank’s balanced salt solution (HBSS) medium for 2 h. g, Currents from TMEM175-transfected cells without starvation pretreated with small interfering (si)RNAs against AKT1, AKT2 and AKT3 or with a control siRNA. Inset, western blot showing the efficient knockdown of AKT. For gel source data, see Supplementary Fig. 1. IB, immunoblot. h, i, Currents from wild-type and TMEM175-knockout mouse hippocampal neurons before or after starvation (overnight in DMEM). The AKT activator SC79 (10 μM) was applied to the bath during some of the recordings (blue and red traces). Bar graphs in f, g, i show averaged current sizes (at 100 mV). Data are mean ± s.e.m. Numbers of recordings are in parentheses. Arrows are used to indicate curves that overlap and are not easily distinguished. Colours denote conditions for recording: from a bath containing K⁺ (black), K⁺ and SC79 (blue) or NMDG and SC79 (red).

Fig. 3 |. AKT activates TMEM175 via catalysis-independent and interaction-dependent mechanisms.

a, Association between native TMEM175 and AKT. Top two panels, total proteins prepared from wild-type and TMEM175-knockout mouse brains (lanes 1–3) or SH-SY5Y human neuroblastoma cells (lanes 4 and 5) were immunoprecipitated (IP) with anti-TMEM175 or with a control antibody (anti-UNC79) and blotted with anti-AKT and anti-TMEM175. Bottom two panels, total protein blotted for input control. b, Lysates from HEK293T cells transfected with GFP-tagged wild-type or mutant TMEM175 were immunoprecipitated with anti-GFP and blotted with anti-GFP or with anti-AKT. Whole-cell lysates (WCL) were also blotted for input control. c, Sequence alignment of the T338 region. The locations of T338 and K336 are indicated on the human TMEM175 structure (Protein Data Bank code (PDB) 6WCA). d, Pull down of AKT with human TMEM175–glutathione-S-transferase (GST) fusion protein. Cell lysates from human AKT1-transfected HEK293T (lanes 1–3) or S2 (lanes 4–6) cells were incubated with glutathione-agarose-bound GST (as a negative control), GST fusion protein of the 19 amino acids of human TMEM175 containing the wild-type TM2–TM3 linker of domain II (GST–IIL2/3; sequence: HHSLFLHVRKATRAMGLLN) or that with the K336A substitution. Top, bound proteins were probed with anti-AKT (top). Bottom, GST fusion proteins were visualized with Ponceau-S staining of the membrane after immunoblotting. e–j, Currents recorded from HEK293T cells transfected with TMEM175 mutants as indicated with (g) or without starvation (wild type shown in Fig. 2d–f). k, Currents recorded with or without Mg-ATP (2 mM) in the bath. l, Averaged I_K sizes (at 100 mV) recorded under conditions as indicated. Data are mean ± s.e.m. Numbers of recordings are in parentheses. In e–k, arrows are used to indicate curves that overlap and are not easily distinguished. For gel source data, see Supplementary Fig. 1.

Fig. 4 |. Common TMEM175 variants associated with susceptibility to Parkinson’s disease bidirectionally regulate function of the lysoK_GF channel.

a, Variants in human TMEM175 with minor allele frequency (MAF) > 5% (from https://gnomad.broadinstitute.org/). The two single-nucleotide polymorphisms (SNPs) are associated with increased (rs34311866, meta-analysis odds ratio 1.26, 95% confidence interval [1.22, 1.31], P = 6.00 × 10⁻⁴¹) or decreased (rs34884217, odds ratio 0.74, 95% confidence interval [0.69, 0.81], P = 1.56 × 10⁻¹²) susceptibility to Parkinson’s disease (PD) (data from http://www.pdgene.org/). b–d, Characterization of p.M393T. b, Sequence alignments of M393 region (indicated on structure PDB 6WCA). c, Protein expression levels of wild type and human TMEM175 mutants. Total protein from nontransfected HEK293T cells (lane 1) and cells transfected with GFP-tagged wild-type or mutant human TMEM175 (lanes 2–5) was probed with anti-GFP or anti-actin (for input control). For gel source data, see Supplementary Fig. 1. d, Currents from HEK293T cells transfected with wild-type and/or mutant TMEM175 with or without starvation in HBSS medium for 1 or 2 h, as indicated. e, f, Characterization of the TMEM175(Q65P) variant in HEK293T cells (e) and knock-in mouse neurons (f). e, Currents from wild-type and TMEM175(Q65P) cDNA-transfected HEK293T cells before (0 h, fed) or after (1 or 2 h) starvation. f, Currents from wild-type (+/+), heterozygous (Q65/+, het) and homozygous (Q65P/Q65P, hom) knock-in mouse midbrain neurons before (0 h, fed) or after (3, 6 or 9 h) starvation in DMEM. In d–f, averaged I_K sizes (at 100 mV) are shown in the bar graphs. Data are mean ± s.e.m. Numbers of recordings are in parentheses. P values shown are for unpaired two-tailed t-tests.

Fig. 5 |. LysoK_GF deficiency leads to accelerated spreading of pathogenic α-syn, loss of dopaminergic neurons and impaired motor function in mice.

a, α-Syn spread assay. Cultured mouse hippocampal neurons were seeded with α-syn preformed fibrils and immunostained two weeks later with anti-pSer129 α-syn (pSyn, red), α-tubulin (green) and DAPI (blue). Bar graph shows pSyn signals normalized to the number of cells (n = 10 coverslips). Scale bar, 50 μm. AU, arbitrary units. b, Loss of dopaminergic neurons in TMEM175-knockout mice. Left, tyrosine hydroxylase (TH) immunostaining at the level of ventral tegmental area (VTA) and substantia nigra pars compacta (SNc) from wild-type (top) and homozygous TMEM175-knockout (bottom) littermates (18–22 months old). n = 5 mice for each genotype. Right, quantification of TH-positive neurons and total NeuN-positive cells. Scale bars, 200 μm. c, Brains from wild-type and heterozygous littermates (about 12 months old; n = 5 mice for each genotype) were whole-tissue immunolabelled with anti-TH or anti-dopamine transporter (DAT) and imaged with light-sheet microscopy. Bar graphs show the numbers of dopaminergic neurons and width of axonal bundles. Scale bars, 1 mm. d, Behavioural tests using wild-type and heterozygous littermates (about 12 months old). n = 12 mice for each test. In a–d, P values given are from two-tailed t-tests, except for rotarod in d (two-way analysis of variance).