Phosphatidylinositol-4,5-bisphosphate is required for KCNQ1/KCNE1 channel function but not anterograde trafficking

The slow delayed-rectifier potassium current (IKs) is crucial for human cardiac action potential repolarization. The formation of IKs requires co-assembly of the KCNQ1 α-subunit and KCNE1 β-subunit, and mutations in either of these subunits can lead to hereditary long QT syndrome types 1 and 5, respectively. It is widely recognised that the KCNQ1/KCNE1 (Q1/E1) channel requires phosphatidylinositol-4,5-bisphosphate (PIP2) binding for function. We previously identified a cluster of basic residues in the proximal C-terminus of KCNQ1 that form a PIP2/phosphoinositide binding site. Upon charge neutralisation of these residues we found that the channel became more retained in the endoplasmic reticulum, which raised the possibility that channel–phosphoinositide interactions could play a role in channel trafficking. To explore this further we used a chemically induced dimerization (CID) system to selectively deplete PIP2 and/or phosphatidylinositol-4-phosphate (PI(4)P) at the plasma membrane (PM) or Golgi, and we subsequently monitored the effects on both channel trafficking and function. The depletion of PIP2 and/or PI(4)P at either the PM or Golgi did not alter channel cell-surface expression levels. However, channel function was extremely sensitive to the depletion of PIP2 at the PM, which is in contrast to the response of other cardiac potassium channels tested (Kir2.1 and Kv11.1). Surprisingly, when using the CID system IKs was dramatically reduced even before dimerization was induced, highlighting limitations regarding the utility of this system when studying processes highly sensitive to PIP2 depletion. In conclusion, we identify that the Q1/E1 channel does not require PIP2 or PI(4)P for anterograde trafficking, but is heavily reliant on PIP2 for channel function once at the PM.


Introduction
Cardiac repolarisation is determined by a number of K + currents, including the slow delayedrectifier potassium current (I Ks ), the rapid delayed-rectifier potassium current (I Kr ) and the inward-rectifier potassium current (I K1 ). The channel underlying I Ks is a heteromultimeric complex comprised of a canonical voltage-gated K + channel α-subunit, KCNQ1 (Q1) (Kv7.1), and a β-subunit, KCNE1 (E1) [1,2]. In a related vein, the Kv11.1 α-subunit forms the human PLOS  ether-a-go-go-related gene (hERG) channel and constitutes the I Kr current in human ventricular myocytes [3]. I K1 is a classic inward rectifier K + current and is likely constituted predominantly by Kir2.1 [4]. I Ks is so named because of its slow kinetics, which allow the current to accumulate at high heart rates. Notably, I Ks is augmented through increased adrenergic drive [5][6][7]. As such, I Ks is thought to be particularly important during exercise, when it plays a major role in action potential and QT interval shortening [8]. This is supported by the fact that most cases of sudden death in patients with long QT syndrome type 1 (LQT1; caused by mutations in Q1) occur during exercise [9].
Numerous ion channels are regulated by phosphatidylinositols [10][11][12]. Phosphatidylinositols are a family of acidic phospholipids, each member differing in the phosphorylation of its myo-inositol ring. These phospholipids are present on the cytosolic face of cellular membranes and are involved in many biological processes, including vesicle formation and trafficking [13]. Phosphatidylinositol (PI) is synthesised in the endoplasmic reticulum (ER) and is subsequently transported via specific transfer proteins or vesicles. PI kinases then phosphorylate PI to generate different PI species in the relevant cellular location. For example, phosphatidylinositol-4,5-bisphosphate (PIP 2 or PI(4,5)P 2 ) is predominantly present in the plasma membrane (PM), whilst phosphatidylinositol-4-phosphate (PI(4)P) is present in the PM but is also enriched at the Golgi [13,14].
It is well established that the Q1/E1 channel requires the presence of PIP 2 to function, and it is thought that it acts to stabilise the channel in an open state [15][16][17]. Other KCNQ channels also require PIP 2 for voltage-dependent activation, and PIP 2 depletion is likely the main mechanism by which the muscarinic acetylcholine receptor (M 1 ) induces M-channel (KCNQ2/3) closure [15,18]. In our earlier work we defined a PIP 2 -binding region in the proximal C-terminus of the channel, with the most important residues in this region being Lys358 and Arg360 [19]. Even though PIP 2 bound to the channel with the highest affinity, other phosphoinositide species could also bind [19]. Furthermore, we showed that the homologous residues in KCNQ2/3 are also likely to form a phosphoinositide binding site [20,21]. In a recent structural study of KCNQ1 using cryo-electron microscopy, the voltage sensor domains were uncoupled from the closed pore region in the absence of PIP 2 [22]. Therefore, it is clear that PIP 2 , through direct binding to a KCNQ potassium channel domain, can modulate channel opening [23,24]. What is less clear, however, is whether binding of PIP 2 in the PM and/or PI(4)P in the Golgi plays a role in the anterograde trafficking of KCNQ channels. This is a relevant issue for KCNQ1 as a number of LQT1-causing mutations reside in PIP 2 binding sites [25], and we have shown that the impaired delivery of mutant channels to the PM and their retention in the ER is an important disease mechanism [26,27].
The established methods for assessing the function of phosphoinositides, such as activation of a G q/11 -coupled G-protein coupled receptor (GPCR) or the addition of wortmannin, tend to be indirect and lack specificity. For example, the kinase inhibitory actions of wortmannin are wider ranging than previously thought [28] and the phospholipase C (PLC)-mediated depletion of PIP 2 upon activation of the M 1 receptor (a G q/11 -GPCR) additionally produces diacylglycerol, activates protein kinase C and mobilises Ca 2+ from intracellular stores. The identification of the voltage-sensitive phosphatase (VSP) from Ciona intestinalis (Ci-VSP) [29] has provided a tool that can deplete PIP 2 in a more specific manner. Ci-VSP localises to the PM and can rapidly deplete PIP 2 and phosphatidylinositol-3,4,5-trisphosphate (PI(3,4,5)P 3 ) upon membrane depolarization [30]. VSPs, in conjunction with patch clamp recording, have been used to determine the PIP 2 dependence of a range of ion channels (for example [31] and [32]). However, despite the obvious advantages of VSPs their PM localisation and requirement for membrane depolarisation can be limiting (for example when trying to study cellular processes such as signaling and protein trafficking).
In 2006, a novel chemically induced dimerization (CID) system was developed to enable the selective depletion PIP 2 at the PM without activating PLC or other downstream pathways [33]. This CID system exploits the ability of two protein domains, one from FK506 binding protein (FKBP) and one from mTOR (mechanistic target of rapamycin) FKBP-rapamycin binding (FRB) to heterodimerize upon the addition of rapamycin or a chemical analog iRap (for review see [34]). By fusing the yeast inositol phosphatase (Inp54p) to FKBP (Inp54p-FKBP) and the FRB domain to a PM anchor Lyn 11 (Lyn 11 -FRB), Suh and colleagues were able to rapidly recruit and concentrate Inp54p at the PM upon the addition of iRap [33]. Using this CID system the authors elegantly identified that the depletion of PIP 2 alone, upon recruitment of Inp54p-FKBP to the PM, is sufficient to result in M-channel closure [33].
Similar CID approaches have since been used to examine a variety of biological processes and they provide an exciting tool for studying signaling pathways [35]. In this study, we use a CID system [36] to examine whether PIP 2 and/or PI(4)P play a role in the anterograde trafficking and function of the Q1/E1 channel complex that underlies I Ks .
For the in-cell western assay, cells were washed with PBS + (PBS supplemented with 1 mM MgCl 2 & 0.1 mM CaCl 2 ) then fixed using 3.7% formaldehyde solution (PBS + + formaldehyde) for 20 minutes. After fixing, cells were washed with PBS + , and then permeabilized with 0.1% triton-X100 solution (PBS + + triton-X100). Cells were washed with PBS + and incubated in cell culture media for 30 minutes. Primary antibody (Ab) was added (α-KCNQ1; sc-20816, Santa Cruz) in 200 μl cell culture media/well at a 1:1000 dilution, and the cells incubated on a rocking platform at room temperature for 1 hour. Cells were washed with PBS + before adding secondary Ab (α-Rabbit Dylight 800; 5151S, New England Biolabs) in 500 μl cell culture media/ well at a 1:1000 dilution. During secondary Ab incubation, cells were protected from light and incubated on a rocking platform at room temperature for 1 hour. Cells were then washed 3 times with PBS + , and 500 μl PBS + was added to each well for visualisation.
For the on-cell western assay, cells were either incubated in humidified cell culture incubator at 37˚C throughout the assay [37˚C live] or were fixed after a 1-hour incubation with the primary Ab at 4˚C [4˚C fixed] as indicated in the figure legends. 37˚C live assay: Primary Ab (α-VSV; V5507, Sigma) was added in 200 μl cell culture media at a 1:500 dilution for 1 hour at 37˚C. Cells were then washed three times for 10 mins in cell culture media. Secondary Ab was then added (α-mouse DyLight800, 5257S, New England Biolabs) in 500 μl cell culture media at a 1:1000 dilution for 1 hour. Cells were washed twice with PBS + , and 500 μl PBS + was added to each well for visualisation. 4˚C fixed assay: Primary Ab (α-VSV) was added in 200 μl ice-cold cell culture media at a 1:500 dilution for 1 hour at 4˚C. Cells were then washed once in icecold cell culture media and twice in ice-cold PBS + . Cells were then fixed using 3.7% formaldehyde for 20 minutes. After fixing, cells were washed with PBS + , and secondary Ab was added (α-mouse DyLight800) in 500 μl cell culture media at a 1:1000 dilution for 1 hour at RT. Cells were washed twice with PBS + , and 500 μl PBS + was added to each well for visualisation.
Plates were visualised using a LI-COR Odyssey Plate Reader. To correct for any background signal not related to specific channel staining, the mean intensity of the wells that were untransfected was subtracted from the intensity of the wells expressing the channel. To normalise the infrared signals between assays, the mean intensity values from the various transfection conditions were normalised to the mean intensity values from the cells transfected with VSV-E1-Q1 + pcDNA3.1.
Rapamycin (5 μM) (sc-3504, Santa Cruz) was added, either 1 hour or 24 hours before the start of the assay. Wortmannin (10 μM) was added 1 hour prior to the start of the assay, and brefeldin A (5 μM) was added 24 hours before starting the assay.

Electrophysiology
Whole-cell patch clamp was used to record ionic currents at room temperature using an Axopatch 200B amplifier (Axon Instruments). Transiently transfected cells (500 ng per construct) were identified using epifluorescence and were patch clamped 48-72 hours after transfection. Extracellular bath solution for all experiments with HEK-I Ks , HEK-I K1 and CHO-K1 cells contained (mM) 150 NaCl, 5 KCl, 10 HEPES, 2 MgCl 2 and 1 CaCl 2 (pH 7.4 with NaOH). Extracellular bath solution for all experiments with HEK-I Kr cells contained (mM) 150 NaCl, 4 KCl, 10 HEPES, 1 MgCl 2 , 1.8 CaCl 2 and 5 glucose (pH 7.4 with NaOH). The intracellular pipette solution contained (mM) 150 KCl, 10 HEPES, 5 EGTA, 2 MgCl 2 , 1 CaCl 2 and 5 (Na) 2 ATP (pH 7.2 with KOH). Pipette resistance, when pipettes were filled with intracellular solution, was 2-3 MO. Pipette tips were coated with SigmaCote (SL2, Sigma) after being filled with intracellular solution, to reduce pipette capacitance. After seal formation and successful rupture of the membrane patch, cells were dialysed for 2 minutes before recording. In all experiments, R series was compensated for by 70% using the amplifier circuitry. The voltage protocols used are indicated in each figure. The liquid junction potentials for both combinations of solutions were relatively small (+4.3 and +4.4 mV respectively) and therefore post-recording adjustments of membrane potential were not performed. In experiments involving the perfusion of rapamycin, extracellular solution containing rapamycin (5 μM) was perfused for 2 minutes, after which perfusion was then returned to rapamycin-free extracellular solution.
Data were analysed using Clampfit (Molecular Devices, CA, U.S.A) and GraphPad Prism (GraphPad Software, Inc., San Diego, CA, U.S.A). Current density (CD) was calculated by normalising the peak current value at the end of each activating pulse to the cell capacitance (nA/ pF). Peak tail current density (PTCD) was calculated by normalising the peak tail current value at the start of the repolarising pulse to cell capacitance (pA/pF).

Confocal microscopy
Live cells transiently transfected with 500 ng of each construct were imaged 48-72 hours after transfection using a Zeiss LSM510 confocal microscope (Mark 4; Carl Zeiss, Oberkochen, Germany) and a Plan-Apochromat 63x oil lens objective. GFP was excited with a multiline argon laser (wavelength 488 nm) and RFP was excited using a helium/neon laser (wavelength 543 nm). Tubby-YFP was imaged using laser settings and filters optimal for excitation of GFP in order to reduce potential cross-talk with RFP. GFP/YFP and RFP/DsRed were excited sequentially to minimise potential crosstalk. Images were median-filtered, converted to RGB format and saved in a JPEG/TIFF format. Colocalisation analysis was performed as described in [45].

Statistical analysis
Data are expressed as mean ± S.E.M. Statistical analysis was performed using GraphPad Prism. A Student's t-test (P <0.05) was used to determine statistical significance for single comparisons, and a one-way ANOVA (P <0.05), with Bonferroni's multiple comparison test, was used to determine significance when multiple groups were compared.

Mutations in a PIP 2 binding domain in KCNQ1 lead to increased retention of the channel complex in the ER
We previously identified charged residues in the proximal C-terminus of KCNQ1 important for phosphoinositide binding [19]. Subsequently, we examined whether the mutation of these residues in KCNQ1-GFP (co-transfected with KCNE1) affected channel trafficking and retention in the ER in transiently transfected CHO-K1 cells. We found that these mutants displayed greater ER retention than the wild-type (WT) channel (Fig 1). This raised the question of whether anionic phospholipid binding was important for the trafficking of KCNQ1 in the secretory pathway. To explore this in a more definitive manner we employed a CID system, described in the materials and methods section, specifically to examine the roles played by PIP 2 and PI(4)P in Q1/E1 channel trafficking and function.
Characterisation of the CID system using a specific PIP 2 reporter Previously we developed a specific fluorescent probe that is able to monitor and report changes in PIP 2 levels at the PM, named Tubby-YFP [41]. Using Tubby-YFP we monitored whether the CID system was working as expected. PJ-DEAD or PJ were transiently expressed with LYN 11 -FRB and Tubby-YFP in HEK293 cells, and imaged using confocal microscopy. When rapamycin (5 μM) was added to cells expressing PJ-DEAD or PJ with LYN 11 -FRB and Tubby-YFP, both PJ-DEAD and PJ translocated from the cytosol/nucleus to the PM in 30-45 seconds (Fig 2A and 2B). In cells expressing PJ, Tubby-YFP underwent a rapid redistribution after the translocation of PJ to the PM (Fig 2A, 2C and 2D). At 45 seconds after rapamycin addition, Tubby-YFP had partially moved from the PM to the cytosol, and at 60 seconds it showed a complete redistribution (Fig 2A, 2C and 2D). In contrast, in cells expressing PJ-DEAD there was no movement of Tubby-YFP after translocation of PJ-DEAD to the PM (Fig 2B, 2E and 2F). The redistribution of Tubby-YFP from the PM to the cytosol when PJ, but not PJ-DEAD, was recruited to the PM indicates that the CID system was working as expected.

PIP 2 /PI(4)P modulation of I K1 and I Kr channel function
We next examined the effects of CID-system-mediated depletion of PIP 2 and PI(4)P at the PM on the cloned equivalents of cardiac currents I K1 and I Kr in HEK293 cell lines stably expressing hKir2.1 (KCNJ2) and Kv11.1 (KCNH2; hERG1a), respectively. In the recording conditions used, the currents were stable for~10 minutes and did not run down (Fig 3A and 3C). The expression of any combination of the CID constructs prior to the addition of rapamycin did not act to significantly suppress either of the currents (S1 Fig and Fig 3B and 3D). Rapamycin (5 μM) was perfused over HEK-I K1 cells expressing PJ-DEAD and LYN 11 -FRB for 2 minutes after the completion of the first I-V recording. The inward rectifier current was recorded every minute for 8 minutes. Over this period, there was no change in peak inward current density (CD) (P = NS) (Fig 3B). The same experiment was repeated in HEK-I K1 cells expressing PJ-INPP5E and LYN 11 -FRB, and after 8 minutes there was no change in the peak inward CD (P = NS) (Fig 3B). Finally, rapamycin was added to cells expressing PJ and LYN 11 -FRB. After rapamycin (5 μM) perfusion for 2 minutes there was a gradual but complete inhibition of the peak inward I K1 CD over approximately 6 minutes (P <0.05 at +1 min rapamycin, and P <0.005 at +2 minutes rapamycin onwards), with only endogenous HEK293 current remaining from 6 minutes (Fig 3B). This current was significantly reduced compared to control (P <0.001 across all time points analysed) and PJ-DEAD (P <0.001 across all time points analysed). In an analogous fashion we examined the effects on I Kr . Rapamycin (5 μM) was perfused over HEK-I Kr cells expressing PJ-DEAD and LYN 11 -FRB for 2 minutes after completion of the first I-V recording. Over a period of 10 minutes, there was a small reduction in peak tail current density (PTCD), which was not significant (P = NS) and was similar to untransfected cells (Fig 3D). When rapamycin was perfused over HEK-I Kr cells expressing PJ and LYN 11 -FRB for 2 minutes, there was a small inhibition of the PTCD (Fig 3D) beyond that seen in untransfected HEK-I Kr cells and cells expressing PJ-DEAD and LYN 11 -FRB, but this difference was not significant (P = NS). These results highlight that I K1 is significantly inhibited by dual PIP 2 / PI(4)P depletion whilst I Kr is not.  Expression of PJ-DEAD and LYN 11 -FRB had no effect on I Ks CD or PTCD in HEK-I Ks cells compared to control (untransfected) cells (Fig 4A, 4B and 4C). Next, PJ-SAC, PJ-INPP5E or PJ were transiently expressed with LYN 11 -FRB in HEK-I Ks cells, and I Ks was recorded in each condition. Expression of PJ-SAC with LYN 11 -FRB had no effect on I Ks CD or PTCD (P = NS), while expression of PJ-INPP5E with LYN 11 -FRB resulted in a reduction of I Ks CD and PTCD by approximately 50% compared to control (P <0.005 across all voltages analysed for both CD and PTCD) (Fig 4A, 4B and 4C). PJ-INPP5E also significantly reduced I Ks compared to PJ-DEAD (P <0.001 across all voltages analysed for both CD and PTCD) (Fig 4A, 4B and 4C). Expression of PJ with LYN 11 -FRB in HEK-I Ks cells caused a near complete loss of I Ks in the absence of rapamycin application (P <0.001 across all voltages analysed for CD and PTCD when compared to control or PJ-DEAD) (Fig 4A, 4B and 4C).
Due to the near complete loss of I Ks that was observed in HEK-I Ks cells expressing PJ with LYN 11 -FRB, a similar experiment was performed in CHO-K1 cells to rule out cell-type-specific effects. The I Ks channel subunits, KCNQ1 and KCNE1, were expressed in CHO-K1 cells with LYN 11 -FRB and PJ-DEAD, and the resulting current was recorded. While expressing PJ-DEAD, I Ks current was detected. However, the expression of PJ in place of PJ-DEAD caused a dramatic loss of the current as previously seen in HEK-I Ks cells, which was significantly reduced (P <0.0001) across all voltages analysed for both CD and PTCD (S2 Fig). A similar suppression of KCNQ1 (Kv7.1) currents occurred if KCNQ1 was expressed in the absence of KCNE1 in CHO-K1 cells (S3 Fig). In addition, when PJ was expressed alone (without LYN 11 -FRB) in HEK-I Ks cells there was a substantial, albeit less prominent, reduction (~50%) in current (Fig 4D and 4E).

The depletion of PIP 2 and/or PI(4)P at the PM and PI(4)P in the Golgi does not affect the cell surface expression level of KCNQ1
We used in-cell and on-cell western assays to measure the surface expression and trafficking of VSV-E1-Q1. When HEK293 cells transiently expressing VSV-E1-Q1 were incubated for 1 hour with wortmannin (10 μM), which inhibits both phosphatidylinositol 4-kinase (PI4K) and 3-kinase (PI3K) activity at this concentration [46], there was no change in the total VSV-E1-Q1 channel expression. Wortmannin incubation did, however, cause a significant reduction in cell-surface expression of the VSV-E1-Q1 construct (P <0.01) (Fig 5). Incubation of VSV-E1-Q1-expressing HEK293 cells with brefeldin A (5 μM) for 24 hours resulted in some cell death. The total cellular expression of VSV-E1-Q1 was significantly reduced (P <0.001) after 24 hours brefeldin A incubation, but it is unclear whether this was solely due to the reduction in total cell number or whether there was an actual reduction in VSV-E1-Q1 protein expression (Fig 5D). Brefeldin A incubation also dramatically reduced the cell-surface expression of the VSV-E1-Q1 construct (P <0.001) (Fig 5C).
In addition, the assay was able to report genetically encoded perturbations that might affect channel trafficking. We have previously shown that the KCNQ1 mutant, KCNQ1-E261D, is retained in the ER and that this behaviour is dominant, i.e. co-expression with WT leads to retention of both mutant and WT [26]. Overexpression of KCNQ1-E261D led to reduction in after rapamycin (5 μM) perfusion (orange bar). PJ-DEAD (n = 10) or PJ (n = 10). Relative current (at 0 mV) from untransfected HEK-I Kr cells (control; n = 9) included for comparison. Data are presented as mean ± S.E.M. * indicates significant difference (P <0.05) from control value.   cell-surface but not total expression (Fig 5). Furthermore, expression of SAR1-H79G, which disrupts ER export, also decreased cell surface but not total expression (Fig 5).
We next tested whether acute or longer term depletion of PIP 2 /PI(4)P at the PM can affect channel trafficking. To do this we transiently expressed VSV-E1-Q1, LYN 11 -FRB and either PJ-DEAD or PJ in HEK293 cells and recruited the lipid depleting constructs to the PM by adding rapamycin for 1 or 24 hours prior to the start of the assay. In the absence of rapamycin the expression of PJ-DEAD or PJ did not have an effect (P = NS) on cell-surface or total expression of the channel (Fig 6). In addition, the application of rapamycin for either 1 or 24 hours, and therefore induction of PIP 2 /PI(4)P depletion at the PM, had no effect (P = NS) on cell-surface or total expression of the channel (Fig 6).
PI(4)P is concentrated in the Golgi and is thought to be important for transport of specific cargo from the Golgi to the PM [37]. To investigate whether Golgi-localised PI(4)P plays a role in Q1/E1 trafficking we transiently expressed VSV-E1-Q1, Tgn38-FRB, and either PJ-DEAD or PJ-SAC in HEK293 cells. In the absence of rapamycin, and therefore Golgi recruitment, the expression of PJ-DEAD or PJ-SAC did not have an effect (P = NS) on cell-surface or total expression of the channel (Fig 7). Similarly, the addition of rapamycin, and therefore depletion of Golgi-localised PI(4)P, for either 1 or 24 hours failed to effect (P = NS) the total or cell-surface expression level of the channel (Fig 7).

Differential sensitivity of cardiac ion channels to PIP 2
In this study, Q1/E1 channel function was sensitive to the depletion of PIP 2 . Even in the absence of rapamycin (the dimerizing agent) the co-expression of PJ and LYN 11 -FRB dramatically reduced I Ks . We reason below that this occurs because of a high sensitivity of the channel to changes in PM PIP 2 levels, which are reduced by the basal constitutive phosphatase activity of the CID system. The I Ks channel β-subunit, KCNE1, confers altered sensitivity to PIP 2 [47]. As such, we additionally investigated the regulation of the homomeric KCNQ1 channel. The loss of this current when expressing PJ and LYN 11 -FRB, in the absence of rapamycin, also indicates that the homomeric KCNQ1 channel has high sensitivity to PM PIP 2 depletion, which supports the findings of Li and colleagues [47].
For comparison, we investigated the PIP 2 regulation of Kir2.1 and Kv11.1 channels (the pore-forming subunits of the I K1 and I Kr currents, respectively) as these are also present in ventricular myocytes and contribute to ventricular repolarisation. In contrast to I Ks , the co-expression of PJ with LYN 11 -FRB, in the absence of rapamycin, had no effect on either of the currents. In HEK-I K1 cells, recruitment of PJ to the PM caused a gradual and complete inhibition of I K1 , and in HEK-I Kr cells PJ recruitment caused a small (but not significant) reduction in I Kr over the course of the experiment. Thus, Kv11.1/I Kr is relatively insensitive to PIP 2 /PI(4) P depletion, whilst Kir2.1/I K1 is sensitive. Our findings agree with those reported by Kruse and colleagues, who showed that Kv11.1 is insensitive to joint depletion of PIP 2 and PI(4)P [32]. Prior studies have reported that Kv11.1 is regulated by PIP 2 , although the effects on channel function are modest [48,49]. Recently it has been shown that the co-expression of KCNE2 [32] or the G q/11 -coupled Muscarinic-1 receptor [50] with Kv11.1 confers increased sensitivity to PIP 2 . This suggests that the presence of accessory subunits (such as the KCNE family) or G q/ 11 -coupled GPCRs, which are expressed in ventricular myocytes, may enhance the PIP 2 sensitivity of the I Kr channel complex.
The complete loss of I Ks even before rapamycin addition indicates that this channel is extremely sensitive to PIP 2 depletion and suggests that the channel complex has a relatively low apparent affinity for PIP 2 . In contrast, Kir2.1/I K1 was only significantly inhibited when PJ  was recruited to the PM and both PIP 2 and PI(4)P were depleted. The inability of PJ-INPP5E recruitment to inhibit Kir2.1/I K1 could have been because the level of PIP 2 depletion achieved was not high enough to inhibit channel function, or alternatively it could be because PI(4)P is equally able to activate the channel in the absence of PIP 2 (especially since PJ-INPP5E converts PIP 2 to PI(4)P). A recent study directly addressed this question and found that the application of PI(4)P, in giant excised patches, does not activate Kir2.1 [51]. Therefore, we feel that the differences in the ability of PJ and PJ-INPP5E to inhibit Kir2.1/I K1 directly reflect how strongly they can deplete PIP 2 . In comparison to I Ks , the inhibition of Kir2.1/I K1 channel activity requires a greater reduction in PIP 2 . This implies that Kir2.1 has a higher relative apparent affinity for PIP 2 . To our knowledge the relative affinities of Q1/E1 and Kir2.1 for PIP 2 have not been directly compared (in the same study). However, Kir2.1 has been postulated to have an approximately seventeen-fold higher affinity than KCNQ2 for PIP 2 [32]. Based on our findings we speculate that the activity of I Ks , but not I K1 or I Kr , is likely to be strongly affected by even small fluctuations in the levels of PIP 2 . This is similar to our previous observations regarding the PIP 2 sensitivity of ATP-sensitive potassium channels [52], and is consistent with the findings by Kruse and colleagues that only a specific subset of ion channels are highly sensitive to PIP 2 depletion [32, 53].

Drawbacks of the CID system
Manipulation of cellular PIP 2 levels using methods such as excised-patch recordings [16], G q/ 11 -coupled receptor activation [54] and exogenous application of PIP 2 analogues [25] give insights into the dependence of the Q1/E1 channel on PIP 2 for function. However, what these experimental methods do not allow is a direct, isolated and specific manipulation of PIP 2 levels within the cell. In theory, the CID system should overcome these issues, and it has indeed been used to investigate a range of ion channels including the Q1/E1 channel and other members of the KCNQ family [32]. However, when we expressed PJ or PJ-INPP5E with LYN 11 -FRB, inhibition of I Ks occurred even before rapamycin was added.
Could a low level of lipid depletion at the PM occur by expressing active lipid-depleting constructs, even in the absence of rapamycin? The phosphatase domains of the lipid-depleting constructs are constitutively active, which could confer a low level of cellular PIP 2 depletion (and/or PI(4)P in the case of PJ and PJ-SAC, respectively). In our imaging experiments it is clear that the lipid-depleting constructs are expressed throughout the cytosol. The cytosolic expression of the phosphatase constructs could allow a low level of lipid depletion by constructs that are coincidentally close to the PM, as has been discussed previously [35]. Even without LYN 11 -FRB, the expression of PJ in HEK-I Ks cells caused a 50% reduction of current.
Interestingly, the fact that LYN 11 -FRB expression confers increased inhibition of I Ks indicates that there may be a low level interaction between these two constructs, possibly between the FRB and FKBP domains in the absence of rapamycin. A 2.2 Angstrom resolution structure of the FRB-rapamycin-FKBP complex showed multiple residues involved in the protein-protein interaction [55]. A transient and weak interaction could somewhat concentrate the lipiddepleting constructs at the PM. Overall, our findings highlight that the constitutive phosphatase activity of the CID system is an inherent problem when trying to study highly sensitive PIP 2 effectors, such as the Q1/E1 channel.

PIP 2 and PI(4)P do not regulate the cell-surface expression level of Q1/ E1
The observation that the CID system was not optimal for investigation of Q1/E1 channel function does not preclude using it to study the role of PIP 2 and other anionic phospholipids in channel trafficking. To calibrate the in-cell and on-cell western assays a number of positive control experiments were performed. These were known to affect broad trafficking pathways or specifically channel surface expression. The expression of the SAR1 mutant, SAR1-H79G, and the KCNQ1 dominant negative (DN) mutant, KCNQ1-E2691D, both caused a reduction in the cell-surface expression of the VSV-E1-Q1 construct whilst preserving total channel expression. SAR1-H79G has a DN effect over WT SAR1 and prevents vesicle trafficking between the ER and Golgi [42], explaining the reduction in VSV-E1-Q1 cell-surface expression. KCNQ1-E261D is retained in the ER and has a DN effect over the trafficking of the WT channel [26,56]. These cellular manipulations indicate that the in-cell/on-cell western assays can faithfully report changes in channel trafficking to the PM.
To find out if acute or long-term depletion of PIP 2 and/or PI(4)P at the PM affected trafficking of the Q1/E1 channel, cells expressing the lipid-depleting constructs were incubated with or without rapamycin for 1 hour or 24 hours. None of these conditions resulted in any changes in the cell-surface or total expression of the channel, suggesting that the reduction of PIP 2 and/or PI(4)P at the PM does not affect trafficking or expression. PI(4)P is present at the Golgi in higher concentrations than the rest of the cell, and is suggested to be involved in the anterograde trafficking of proteins destined for the PM. Thus, we depleted PI(4)P at the Golgi using PJ-SAC expressed with Tgn38-FRB [37, 57]. We found that there was no change in channel cell-surface or total expression when PJ-SAC was recruited to the Golgi with either 1 or 24 hour rapamycin incubation, despite Golgi recruitment of CID constructs (S4 Fig). In contrast, wortmannin incubation (for 1 hour) reduced the cell-surface expression of VSV-E1-Q1 (by~50%) without altering total channel expression. Given that the depletion of PIP 2 and/or PI(4)P (upon recruitment of PJ to the PM and PJ-Sac to the Golgi) did not alter Q1/E1 trafficking we feel that it is unlikely that the effects of wortmannin are mediated through PI4K inhibition. At both high and low concentrations wortmannin is a potent inhibitor of PI3Ks [58]. PI3Ks control the production of a number of phosphoinositides including phosphatidylinositol-3-phosphate (PI(3)P), phosphatidylinositol-3,4-bisphosphate (PI(3,4)P 2 ) and phosphatidylinositol-3,4,5-trisphosphate (PI(3,4,5)P 3 ) [59]. Wortmannin-mediated disruption of PI3K activity could therefore lead to reduced levels of these phosphoinositides. PI (3)P is highly concentrated on early endosomes and is involved in the recycling of membrane proteins [13], which may explain why the trafficking of the channel is perturbed. Additionally, the reduction of PI(3)P could also lead to reduced levels of phosphatidylinositol-3,5-bisphosphate (PI(3,5)P 2 ). Intriguingly, an enhanced production of PI(3,5)P 2 has been suggested to be central in mediating an increased rate of Q1/E1 channel exocytosis upon activation of the serum and glucocorticoid-inducible kinase 1 [60].
Our findings suggest that the increased degree of channel retention in the ER seen upon charge neutralisation of residues in a PIP 2 binding site (Fig 1) is not due to perturbed binding to PIP 2 and or PI(4)P. It is possible that the increases seen in ER channel retention are solely due to non-specific effects of these mutations on protein stability/folding. However, the inhibitory effects of wortmannin on Q1/E1 trafficking highlight that other phosphoinositides, such as PI(3)P and PI(3,5)P 2 , could play a role and this warrants future investigation.
In conclusion, we identify that the Q1/E1 channel does not require PIP 2 and/or PI(4)P for anterograde trafficking, but is heavily reliant on PIP 2 for function once at the PM. It has been postulated that the relatively low level of PIP 2 in the secretory pathway is important for silencing (or activating) specific ion channels whilst in transit [11,61]. For Q1/E1, our findings support this hypothesis and it is likely that the channel becomes activated by PIP 2 once it reaches the PM.