Recombinant tandem of pore-domains in a Weakly Inward rectifying K+ channel 2 (TWIK2) forms active lysosomal channels

Recombinant TWIK2 channels produce weak basal background K+ currents. Current amplitudes depend on the animal species the channels have been isolated from and on the heterologous system used for their re-expression. Here we show that this variability is due to a unique cellular trafficking. We identified three different sequence signals responsible for the preferential expression of TWIK2 in the Lamp1-positive lysosomal compartment. Sequential inactivation of tyrosine-based (Y308ASIP) and di-leucine-like (E266LILL and D282EDDQVDIL) trafficking motifs progressively abolishes the targeting of TWIK2 to lysosomes, and promotes its functional relocation at the plasma membrane. In addition, TWIK2 contains two N-glycosylation sites (N79AS and N85AS) on its luminal side, and glycosylation is necessary for expression in lysosomes. As shown by electrophysiology and electron microscopy, TWIK2 produces functional background K+ currents in the endolysosomes, and its expression affects the number and mean size of the lysosomes. These results show that TWIK2 is expressed in lysosomes, further expanding the registry of ion channels expressed in these organelles.

TWIK2 is expressed in several mammalian tissues including secondary lymphoid organs and the vascular system [20][21][22][23] . Its expression in the central nervous system is very weak 3 . TWIK2 knock-out mice exhibit an increased mean arterial and pulmonary blood pressure, depolarized resting membrane potential in freshly dispersed smooth muscle cells and increased vascular contractility 24,25 . In this study, we studied the mechanism of its intracellular distribution in transfected cells.

Expression of TWIK2 protein and current in lysosomes. Recombinant mouse, rat and human
TWIK2s were expressed in Xenopus oocytes. We could record weak but significant K + currents (<1 µA at +50 mV, Fig. 1A). These currents are rapidly inactivating as previously shown for human TWIK2 in transfected mammalian cells (Fig. 1B) [20][21][22][23] . However, they are much smaller than the currents usually produced by other K 2P channels that are expressed at the plasma membrane. Under the same conditions TREK1 (KCNK2, K 2P 2.1) and TASK3 (KCNK9, K 2P 9.1) channels would produce more than 5 µA of current, suggesting that TWIK2 is mainly expressed in an intracellular compartment as shown previously for "silent" TWIK1 19 and THIK2 18 . To study its cellular localization, we expressed TWIK2 into MDCK cells, an epithelial cell line classically used to Figure 1. Comparison of human, rat and mouse TWIK2 channels in heterologous expression systems. TWIK2 currents were recorded in Xenopus oocytes 24 h after cRNA injection. (A) Current amplitude at +50 mV of human, rat and mouse TWIK2 compared to human K 2P channels TREK1 and TASK3. TWIK2 current is small (<1 µA) but significantly different from non-injected cells (non injected n = 56, human TWIK2 n = 23, p < 0.001, rat TWIK2 n = 47, p < 0.001, mouse TWIK2 n = 33, p < 0.001). Each value is the mean ± sem, statistic evaluation was done using Mann-Whitney test. (B) I-V relationships deduced from recordings with voltage pulses ranging from −120 mV to +60 mV in 10 mV steps from a holding potential of −80 mV (non injected n = 56, human n = 23, rat n = 47, mouse TWIK2 n = 33). The "inset" represents the evoked current shape of mouse TWIK2, 20 mV steps. (C) Human and rodent TWIK2 co-localization in transfected MDCK cells. Human TWIK2 was stained with rabbit polyclonal TWIK2 antibody and rat and mouse TWIK2HA with mouse monoclonal HA antibody. Overlap of green and red fluorescence appears in yellow. (C) TASK3 (in green) and TREK1 (in green) in transfected MDCK cells. study trafficking of membrane proteins 8 . Following immunolabeling of TWIK2, these cells display an intracellular staining (Fig. 1C). Human TWIK2 was labeled using a rabbit polyclonal TWIK2 antibody that does not cross-react with rodent channels 22 , whereas rat and mouse TWIK2s were labeled using HA-antibodies directed against a HA-tag fused to their C-terminus (C-ter). Figure 1C shows the co-localization of human TWIK2 co-expressed with rat or mouse TWIK2s, demonstrating that all three isoforms share the same intracellular distribution. A previous study has shown a relocation of rat TWIK2 from an intracellular localization to the plasma membrane after 72 h of expression in CHO cells 21 but we could not confirm this observation in MDCK cells (Suppl. Fig. 1).
The cellular localization of TWIK2 was further studied by co-expressing rat TWIK2 with different intracellular proteins coupled to GFP. A partial co-localization of TWIK2 with Rab7-GFP 26 and Lamp1-GFP 27 was observed, suggesting expression of TWIK2 in late endosomes and lysosomes ( Fig. 2A). Labeling endogenous Lamp1 using a specific antibody gave the same result confirming the localization of rat, mouse and human TWIK2 in lysosomes (Fig. 2B). Next we analyzed the distribution of a rat TWIK2-GFP fusion protein in living cells. Staining lysosomes with Lysotracker enabled us to observe TWIK2 in the lysosomal membranes (Fig. 2C). Live video microscopy shows the movement of these lysosomes (Suppl. Video). TWIK2 expression did not seem to change mobility of lysosomes but affects their number and mean size (Fig. 2D). Cells expressing TWIK2 have nearly 2.5 times more lysosomes than WT cells (37.00 ± 3.91 lysosomes per section plane vs 13.41 ± 2.15, n = 20, p < 0.01, Student's non-paired t test). Lysosomes of TWIK2-expressing cells are nearly 3 times larger than lysosomes in control cells (0.29 ± 0.02 µm 2 vs 0.12 ± 0.01, n = 20, p < 0.01, Student's non-paired t test) (Fig. 2D).
To record the lysosomal K + permeability of TWIK2, we used a modified conventional whole-endolysosome manual patch-clamp technique. Transfection of TWIK2 generated major K + outward currents (from cytosol to endolysosomal lumen) not observed in control cells ( Fig. 3 and Suppl. Fig. 2). When recorded with symmetrical Na + solutions, outward cation currents were reduced in endolysosomes of HEK293T cells transfected with TWIK2 ( Fig. 3A-G), showing that TWIK2 formed a K + channel on endolysosomal membranes. The current-voltage (I-Ψ) relationship showed an outward rectification under high cytosolic K + condition (high K + gradient) as previously demonstrated 22 . It has also been proposed that TWIK2 currents were not sensitive to extracellular pH (similar to endolysosomal luminal) but slightly reduced by intracellular acidification induced by application of extracellular 2,4-dinitrophenol or carbon dioxide 20 . However, under our endolysosomal recording conditions, cytosolic acidification has no obvious effects on the amplitudes or I-Ψ relationship (Fig. 3H,I). We further confirmed that TWIK2 was not affected by direct intracellular acidification by recording a mutant form of TWIK2 predominantly found at the plasma membrane (TWIK2 Y 308 A IL 289/290 AA, described in the next paragraph). As expected, TWIK2 currents recorded in endolysosomes (Fig. 3A,C) and at the plasma membrane  Signal sequences in the C-ter of TWIK2 control its trafficking/transport to lysosomes. Sorting of transmembrane proteins to intracellular organelles is mediated by trafficking signals embedded in their cytosolic domains, which are recognized by elements of the cell machinery such as adaptor protein (AP) complexes. The C-ter region following the last transmembrane helix (M4) constitutes the longest cytoplasmic sequence of TWIK2. Despite its variability, it contains potential trafficking motifs that are present in all mammalian sequences, and partially conserved in Xenopus tropicalis and Danio rerio (Suppl. Fig. 4). To check the functional relevance of these motifs, we progressively deleted the C-ter of rat TWIK2 (Suppl. Fig. 5A). However, all truncated channels displayed an intracellular localization in transfected MDCK cells (Suppl. Fig. 5B). We next introduced point mutations in the three potential signal sequences. Figure 4A displays the complete set of rat TWIK2 mutants. These mutants were either expressed in Xenopus oocytes to compare current amplitudes (Fig. 4B), or transfected and immunolabelled in MDCK cells to study cellular distribution (Fig. 4C). The first potential motif is Step protocols from −100 to +100 mV with a 20 mV step were applied from a holding potential of 0 mV. (B) Currents recorded using a high Na + -containing bath solution (145 mM Na + , pH 7.2). Same step protocol as in a tyrosine-based motif YXXØ at the very end of the C-ter (YASIP in mammals, Y[NS]SIN in toad and zebrafish) 28 . In TWIK2, replacing tyrosine 308 (Y 308 A) or isoleucine 311 (I 311 A) with an alanine promoted a significant increase of cell surface expression in MDCK cells, and a 2 to 3-fold current increase in oocytes (1.19 ± 0.06 µA for TWIK2 Y 308 A and 0.74 ± 0.03 µA for TWIK2 I 311 A versus 0.43 ± 0.02 µA for TWIK2 at 50 mV). The second potential signal sequence is related to a di-leucine-type motif [DE]XXXL[LI] that is also present in TWIK1 and responsible for its constitutive endocytosis 9 . This motif (DEDDQVDIL in rat TWIK2) contains an acidic cluster of amino acids, which may also favor channel sorting to lysosomes. Alanine substitution of isoleucine 289 (TWIK2 I 289 A) and/or leucine 290 (TWIK2 L 290 A), as well as glutamic acids 284 and 285 of the acidic cluster (TWIK2 DD 284/285 AA) increased the expression of TWIK2 at the plasma membrane ( Fig. 4B,C). Mutations of the tyrosine and di-leucine-like motifs show additive effects leading to a 9 to 10-fold current increase (3.8 ± 0.2 µA for TWIK2 Y 308 A IL 289/290 AA, 4.3 ± 0.2 µA for TWIK2 I 311 A IL 289/290 AA versus 0.43 ± 0.02 µA for TWIK2 at 50 mV) and to a strong cell surface relocation (Fig. 4B,C). Finally, the third motif belongs to the DXXLL type (ELILL in mammalian TWIK2s), which is known to promote sorting directly from the Trans-Golgi-Network (TGN) to endosomes without trafficking via the plasma membrane 28 . Alanine replacement of leucines 270 and 271 (TWIK2 LL 270/271 AA) favors cell surface expression (Fig. 4B,C), but with less current expression when compared to the wild type channel (0.31 ± 0.03 µA for TWIK2 LL 270/271 AA versus 0.43 ± 0.02 µA for TWIK2 at 50 mV). The same inhibitory effect is observed when this mutation is combined with mutations of the two other trafficking motifs, even if the mutated channel is fully expressed at the cell surface with no intracellular labeling. These results suggest that the DXXLL motif is an active trafficking signal but that mutations of this motif affect TWIK2 channel activity, probably because of its vicinity with the last membrane-spanning segment M4 that contributes to the internal mouth of the ionic pore.
The results presented so far were obtained using rat TWIK2. By using the same approach, we next confirmed the importance and conservation of tyrosine and di-leucine-like motifs in human TWIK2. Inactivation of these motifs resulted in more expression at the plasma membrane in MDCK cells and higher current amplitudes in oocytes (0.75 ± 0.09 µA for TWIK2 Y 308 A, 1.90 ± 0.14 µA for TWIK2 IL 289/290 AA and 2.59 ± 0,12 µA for TWIK2 Y 308 A IL 289/290 AA versus 0.30 ± 0.02 µA for TWIK2; Fig. 5A,B). We also expressed human wild type and mutated TWIK2 channels in HEK293 cells to carry out immunocytochemistry and electrophysiology in human cells (Fig. 5C,D). TWIK2 Y 308 A IL 289/290 AA produced 18 times more current than TWIK2 (26 ± 2.7 pA/pF, n = 10 at + 60 mV versus 1.4 ± 0.4 pA/pF n = 9; p < 0.001, Student's non-paired t test). As expected more channels were detected by immunocytochemistry at the plasma membrane (Fig. 5D). Together these results identify three trafficking motifs in the C-ter of TWIK2 that are responsible for its distribution to lysosomes. Are these signals sufficient to drive the relocation of a membrane protein from the plasma membrane to lysosomes? To address this question, we used a version of the TASK3 channel 29 containing a HA-tag on its extracellular side (TASK3HA, Fig. 6A). We designed a chimeric channel by exchanging the cytoplasmic C-ter of TASK3HA with the corresponding C-ter of rat TWIK2 in TASK3HA (chimera noted TASK3HA-TW2Ct). TASK3HA and TASK3HA-TW2Ct were co-transfected with TWIK2 into MDCK cells, and their localization was analyzed using anti-HA and anti-TWIK2 antibodies. TASK3HA is located at the plasma membrane, whereas TASK3HA-TW2Ct co-localizes with TWIK2 in lysosomes (Fig. 6A). Mutating the tyrosine and di-leucine-like motifs in TASK3HA-TW2Ct promotes a re-expression of the chimeric channel at the plasma membrane (Fig. 6B) further confirming the role of these motifs, and demonstrating that the C-ter of TWIK2 is necessary and sufficient to promote membrane protein targeting to lysosomes. To visualize channel proteins without antibody staining, they were fused to GFP. When the cells are not permeabilized only TASK3HA-GFP, but not TASK3HA-TW2Ct-GFP, can be detected using HA antibody (Fig. 6C). This result suggests that either TWIK2 channels do not transit via the plasma membrane, or that the amounts present at the cell surface at any given time are not enough to enable detection. TWIK1 endocytosis is a dynamin-dependent mechanism 9 . Co-transfection of TWIK2 with a dominant-negative dynamin I (dynamin K 44 A) did not change channel distribution, suggesting a dynamin-independent trafficking mechanism (Fig. 6D).
TWIK1 contains a functional N-glycosylation site on its extracellular side 5 . In TWIK2, asparagines 79 and 85 form potential N-glycosylation sites (N 79 AS and N 85 AS) that fit the consensus motif NXS/T (Fig. 7). To test glycosylation, protein lysates prepared from cells expressing rat TWIK2 were incubated with Endo-glycosidases EndoH and/ or PNGaseF, and analyzed by Western blot (Fig. 7B left panel). Three different bands were detected in the untreated lysates, but only the lower band was detected in the lysates treated with glycosidases. This shows that TWIK2 is glycosylated, and that the two upper bands correspond to glycosylated proteins. To test the role of the potential glycosylation sites N79 and N85, these asparagines were substituted by a leucine to create single (TWIK2 N 79 L and TWIK2 N 85 L) and double (TWIK2 N 79 L,N 85 L) mutants. The apparent molecular weights (MWs) of mutated TWIK2s are identical to the MWs of the two different glycosylated forms (Fig. 7B, right panel). Incubation of each single mutant with Endo-glycosidases leads to the deglycosylation of the other asparagine, with no gel shift for the double mutant. These results show that both N79 and N85 are glycosylated in TWIK2 (Fig. 7C). To study the effect of the mutations on TWIK2 cellular distribution, the mutated channels were transfected into MDCK cells and immunolabeled (Fig. 7D). Single mutants TWIK2 N 79 L and TWIK2 N 85 L display the same intracellular pattern as TWIK2, whereas double mutant TWIK2 N 79 L,N 85 L shows a strong ER-like staining, suggesting that TWIK2 needs to be glycosylated on at least one asparagine to be stably expressed in lysosomes.

Discussion
Our results show for the first time that recombinant TWIK2 is located in lysosomes, and that this distribution depends on its cytoplasmic C-ter region. When fused to TASK3, a channel that distributes at the plasma membrane, this domain is sufficient to drive the targeting of the chimeric channel to lysosomes. The TWIK2 C-ter region contains three active trafficking motifs 28 . The first sorting signal is located at the very end of its C-ter, and is a tyrosine-based signal that conforms to the YXXØ consensus motif. This type of motif is essential for rapid internalization from the plasma membrane, but is also implicated in the targeting of proteins to lysosomes, for example in the targeting of Lamp1 30-32 . Lysosomal-sorting YXXØ motifs usually have a glycine residue preceding the critical tyrosine residue and acidic residues at the X positions, but this is not the case for TWIK2. The position of the motif within the cytosolic domain of the membrane protein is another important characteristic determining its endocytic or lysosomal-targeting role. Purely endocytic YXXØ signals are not situated at the C-ter of the proteins, but 10-40 residues next to a transmembrane domain. In contrast, the YXXØ motif in the TWIK2 sequence is located at the extremity of its C-ter with only two residues downstream, as observed in other proteins targeted to late endosomal-lysosomal compartments, including Lamp1 32 . The second signal sequence of TWIK2 is related to the [DE]XXXL[LI]-type. This di-leucine-based motif mediates rapid internalization and targeting to endosomal-lysosomal compartments, suggesting that they are active both at the plasma membrane and in intracellular locations 33,34 . Such a motif is present in TWIK1 and is responsible for its rapid and constitutive endocytosis 9 . Similar to TWIK1, the TWIK2 sequence bears a cluster of acidic amino acids preceding this signal that strengthens its sorting to lysosomes. Both YXXØ-and [DE]XXXL[LI]-type motifs bind to adaptor proteins (AP-1, −2 and −3), and membrane proteins containing these motifs are transported to endosomal-lysosomal compartments [35][36][37] . Interestingly, in TWIK2 these two motifs have synergistic effects suggesting that they do not bind at the same site in AP complexes. The third signal sequence of TWIK2 belongs to the recently discovered DXXLL-type signal 28 . This other type of dileucine-based motif is responsible for cycling between TGN and endosomes [38][39][40] . In contrast to YXXØ-and [DE]XXXL[LI]-type motifs, DXXLL sequences do not bind to APs, but to Golgi-localized, γ-ear-containing, ADP-ribosylation factor-binding proteins (GGAs), which function like adaptor proteins 41 . For TWIK2, our results show that the three trafficking motifs have additional effects, and lead to a specific targeting to lysosomes. As they are vesicles specialized in protein degradation, lysosomal proteins need to be protected against enzymatic proteolysis. One known mechanism is glycosylation, as demonstrated for Lamp1 and P2X4 channels. Glycosylation protects these lysosomal channels from degradation 42,43 . TWIK2 is N-glycosylated in the M1-P1 loop facing the lysosomal lumen. When both glycosylated asparagines are mutated, glycosylation of TWIK2 is prevented. The resulting unglycosylated channel was not detected in lysosomes but in ER, suggesting that the unglycosylated channel is not addressed to lysosomes, or is rapidly degraded in these organelles in the absence of the protection afforded by glycosylation.
Although still under investigation, diverse physiological roles have been suggested for silent K 2P channels. Ion channel selectivity of TWIK1 is dynamically regulated by acidic pH suggesting that its recycling between the plasma membrane and internal acidic stores may determine its inhibitory or excitatory influence on cell    17,44 . TWIK1 may also transport K + and Na + in endosomes and thus influences the biology of this recycling compartment. THIK2 is retained in the endoplasmic reticulum (ER), where its expression may affect calcium storage and release throughout its capacity to transport K + counter-ions 8,16,18 . Here we show that TWIK2 is mainly present in lysosomal membranes of transfected cells. Its expression affects the mean size of the lysosomes suggesting an active role in these organelles. K + plays a role in the electrical potential of endosomes and lysosomes, thereby regulating pH stability and organelle function 45 . Beside K + , there are other cations, Na + and Ca 2+ , as well as the anion Clthat help to counterbalance the H + pumping to generate and maintain the endo/lysosomal acidic pH. A number of channels involved in mediating the counter-ion permeability are known: CFTR for Clconductance, TPC for Na + and P2X4 and TRPM2/TRPMLs for Ca 2+ 45-49 . There was also evidence for a "leak-like" K + current in lysosomal membranes 50 . A recent report describes TMEM175, a newly discovered K + channel that produces this K + current in lysosomes 51 . Unlike any of the other K + channels cloned to date from viruses, bacteria, fungi, plants and animals, TMEM175 has no GYG sequence signature-pore domain. This very unusual structure as well as the absence of reports describing classical K + channels in lysosomes suggested that TMEM175 is a highly specialized organelle channel tailored to the lysosomal function. We show here that TWIK2 is another, voltage-independent K + channel expressed in lysosomal membranes. Both TMEM175 and TWIK2 have a wide tissue distribution suggesting that the two proteins may act together in regulating lysosomal function. However, expression levels of TWIK2 are more pronounced in spleen and PBL suggesting a specialized physiological role in these cells. In phagocyting immune cells like macrophages and antigen-presenting immune cells like dendritic cells, the endo-lysosomal compartment has specialized functions that are different from its general role in degradation and recycling of proteins and amino acids. These specialized vesicles take part in cell activation processes triggered by Ca 2+ -release into the cytosol 52 . They have a crucial function in antigen processing and presentation, as well as cell migration and exocytosis [53][54][55] . During starvation, lysosomes additionally play a role in the autophagocytosis 56 . TWIK2 expression in cells bearing specialized endo-lysosomal compartments may point to a particular role in these vesicles, in addition to the "basic" channel equipment.
In summary, this study provides the first evidence of a K 2P channel expressed in lysosomes, likely functioning as a lysosomal K + channel in certain types of cells. Further studies using the knock-out mouse model of TWIK2 should help to decipher the subcellular localization of the native TWIK2 channels and their physiological role in lysosomes.

Methods
Antibodies and Reagents. Rabbit anti-TWIK2 polyclonal antibody has already been described 22 . Mouse anti-HA monoclonal antibody (clone HA.7) and rabbit anti-Lamp1 polyclonal antibody were purchased from Sigma-Aldrich (France). Lysotracker was from Life technologies (France). Cell nuclei were stained using Hoechst 33342 (ThermoFisher, USA). Molecular Biology. Mouse TWIK2 was cloned from lung tissue cDNA into pcDNA, pIRES-GFP and pLIN.
Rat and human TWIK2 clones were kindly provided by Dr Amanda Patel (Valbonne, France). The HA-tag was added in frame at the 3′ end of the TWIK2 coding sequences by PCR. Point mutations, deletions and chimeras were obtained by PCR using standard procedures. All constructs were verified by sequencing. Dynamin-GFP and dynaminK 44  All experiments were performed at room temperature (21-22 °C). K 2P currents were observed using RK400 patch clamp amplifier (Bio-Logic Science Instruments), low-pass filtered at 3 kHz and digitized at 10 kHz using a 12-bit analog-to-digital converter Digidata-1322 (Axon Instrument, Sunnyvale, CA, USA). For current visualization and stimulation protocol application, we used commercial software (Clampex 8.2). The patch pipettes were double-step-pulled from haematocrit-capillaries (Hirschmann Laborgeraete, Germany) using a vertical puller (PC-10, Narishige International, London, UK). Filled pipettes had resistances of 2-4 MΩ. Whole-cell patch-clamp configuration was obtained at a holding potential of −80 mV. A voltage step protocol was applied consisting of 900 ms depolarizing steps going from −80 mV to +100 mV in 20 mV increments. Current amplitudes were expressed as current densities (pA/pF) and results are presented as mean ± standard error of the mean (SEM).
Immunocytochemistry. Cells were grown and transfected on glass coverlips. 24 h, 48 h, or 72 h post transfection, cells were fixed 10 min with 4% paraformaldehyde and blocked and permeabilized with PBS containing 0,1% Triton-X and 5% horse serum for 15 min. Cells were incubated 3 h with primary antibodies, then 1 h with fluorescence labeled secondary antibodies at RT and in the dark. Fluorescence images were acquired and analyzed using an Axioplan2 Imaging Microscope (Carl Zeiss) and the Metavue software.
Live cell imaging. Cells were plated on 25 mm glass coverslips and transfected with rat TWIK2-GFP-encoding plasmid. 24 h post transfection, Lysotracker (Molecular Probes) was added in the medium (final concentration 20 nM), and after 10 min cells were observed on a Zeiss LSM7800 confocal microscope.
Lysosome size. MDCK cells were transfected with empty pIRES2eGFP vector or pIRES2eGFP-rTWIK2. 24 h after transfection, cells were fixed in 1.6% glutaraldehyde (Sigma-Aldrich) in 0.1 M phosphate immediately after medium removal, respectively. Samples were rinsed with cacodylate buffer 0.1 M buffer and then post-fixed in osmium tetroxide (1% in cacodylate buffer 0.1 M) for 1 h. After rinsing with distilled water, they were then dehydrated through an increasing ethanol series and embedded in epoxy resin. Ultrathin sections (70 nm) were collected on Formvar-coated copper grids, stained with uranyl acetate and lead citrate and examined with a Jeol JEM 1400 transmission electron microscope equipped with a SIS MORADA camera. For each condition 10 to 15 fields were analyzed. Results are the mean ± SEM.