Functional KV10.1 Channels Localize to the Inner Nuclear Membrane

Ectopically expressed human KV10.1 channels are relevant players in tumor biology. However, their function as ion channels at the plasma membrane does not totally explain their crucial role in tumors. Both in native and heterologous systems, it has been observed that a majority of KV10.1 channels remain at intracellular locations. In this study we investigated the localization and possible roles of perinuclear KV10.1. We show that KV10.1 is expressed at the inner nuclear membrane in both human and rat models; it co-purifies with established inner nuclear membrane markers, shows resistance to detergent extraction and restricted mobility, all of them typical features of proteins at the inner nuclear membrane. KV10.1 channels at the inner nuclear membrane are not all transported directly from the ER but rather have been exposed to the extracellular milieu. Patch clamp experiments on nuclei devoid of external nuclear membrane reveal the existence of channel activity compatible with KV10.1. We hypothesize that KV10.1 channels at the nuclear envelope might participate in the homeostasis of nuclear K+, or indirectly interact with heterochromatin, both factors known to affect gene expression.

Introduction K V 10.1 (Ether-à-go-go-1, KCNH1) is a voltage-gated potassium channel expressed almost exclusively in the mammalian central nervous system, where its physiological roles have not been clearly identified. Intriguingly, over 70% human tumors of various origins show K V 10.1 expression [1,2,3,4,5,6,7] the intensity of which correlates negatively with survival in some sarcomas [8] and leukemias [9]. It has also been shown that K V 10.1 not only plays a role in oncogenic transformation and malignancy, but also that the proliferation of cancer cells is dependent on its expression [10]. It nevertheless remains unclear if K + permeation is required for the carcinogenic properties of K V 10.1. On one hand, a K V 10.1 open channel blocker [11] and a monoclonal antibody against K V 10.1 [12] both of which inhibit ion flow, are able to reduce xenograft tumor growth in immunodeficient mice. On the other hand, the Drosophila ortholog eag is also known to induce an increase in cell proliferation independently of ion influx [13], and a nonconducting K V 10.1 point mutant still promotes tumor progression, although to a reduced extent [11]. Also, not every K + channel shows a similar behavior. For example, K V 10.5, the closest relative to K V 10.1, possesses very similar electrophysiological properties, but is not carcinogenic [11].
Interactions of the channels with signaling cascades within the cell could explain these effects without ion permeation. This is in line with the observation that the vast majority of K V 10.1 protein remains in intracellular pools, including the perinuclear region, in either heterologous systems [14], neurons [15] or tumor cells [16].
It has also been reported that K V 10.1 plays a role in oxygen homeostasis and angiogenesis independently of K + flow [11], which would provide an advantage to tumor cells but does not explain the observations reported in vitro.
During the eukaryotic interphase, nucleoplasm and cytoplasm constitute two distinct compartments separated by the nuclear envelope, and trafficking between compartments is allowed by the nuclear pore complex (NPC). Small (,40 KDa) soluble proteins can cross the NPC freely, while larger molecules are active and selectively transported [17]. The nuclear envelope is a membranous structure composed of two layers, the inner and the outer nuclear membranes (INM and ONM, respectively), and the space between is termed the perinuclear space. The ONM is an extension of the ER but shows a distinct protein composition, different from that of the rest of the reticulum [18]. Consequently, the perinuclear space is continuous with the lumen of the ER. The protein composition of the INM is radically different from that of the ONM and the ER. Only a handful of transmembrane proteins are known to reach the INM [19].
As a particular type of complex transmembrane protein, ion channels have also been found in the nuclear envelope [20]. Most reports refer to channels in the ONM, while, to our knowledge, only few channels have been reported to date at INM. Marchenko et al. [21] describe the inositol 1,4,5-trisphosphate (InsP 3 ) receptor, Longin et al. [22] a zinc and a calcium channel, Rousseau et al. [23] report two chloride conductances and a non-selective cation channel, and Fedorenko et al., describe a large conductance nonselective cation channel in both the ONM and the INM [24]; among all of them, only the InsP 3 receptor has an identified molecular identity.
The permeability of the NPC to small ions is still under debate. If it would not be permeable to ions, the ionic composition should be different between the nucleoplasm and cytoplasm, and there could be an electrical potential across the nuclear envelope. However, there are few reports on the ion composition of the nucleoplasm [25,26] and, in most cases, the intranuclear potential is only slightly negative compared to the cytoplasm [20] but is still dependent on the cytoplasmic K + concentration [27]. On the other hand, there are compelling electrophysiological studies where it has been proposed that the NPC can close in a voltagedependent manner in the absence of cargo, and that ion flow is prevented when the pore is occluded by cargo [28]. This makes it possible for channels at the INM to dampen or amplify cytoplasmic Ca 2+ transients into the nucleoplasm [29]. From an experimental point of view, the existence of a diffusion barrier also allows measurement of currents from other smaller channels on either the outer or the INM. It should also be noted that the activity of channels at the INM could participate in the regulation of ion exchange between the nucleoplasm and the perinuclear space, which is continuous with the ER lumen but not the cytosol. In any case, it cannot be ruled out that such ion channels may have functions independent of ion permeation.
In the present report, we provide evidence showing that K V 10.1 is located to the INM and that currents compatible with K V 10.1 can be detected by patch clamp at the INM. We also show that INM K V 10.1 might interact with heterochromatin. We discuss these findings in light of the role of K V 10.1 in cancer, as the INM has been shown to affect gene expression, genome stability and cell senescence, all of which are relevant to cancer [30].

K V 10.1 localizes to the perinuclear region of native and transfected cells
There is compelling electrophysiological, microscopical and biochemical evidence that K V 10.1 is a plasma membrane channel. However, immunostainings of neurons, transfected cells, tumors and tumor cells, as well as in vivo observation of K V 10.1 labeled with fluorescent proteins (DsRed2, EGFP and mVenus) in different cell types (NIH3T3, CHO, HEK293), revealed perinuclear localization of the protein in 53% of ts20 cells (n = 156) and 11% of CHO cells (n = 203) in the form of a thin line compatible with the nuclear envelope (Fig. 1A, B). Depending on the cell line used, there were roughly 20-40% of cells exhibiting a perinuclear staining pattern possibly indicating an unrevealed role of K V 10.1 in a particular cellular function.
To confirm this observation, we performed co-localization experiments with lamin A/C as a nuclear envelope marker using double staining with antibodies in HeLa cells expressing endogenous K V 10.1. A significant level of co-localization was observed (Fig. 1D). This localization is not dependent on Golgi integrity, since it is not abolished by brefeldin A (50 mM) treatment, nor does it entirely co-localize with the ER, as revealed by co-transfection of DsRed-tagged K V 10.1 and a YFP-ER marker (data not shown). The C-terminus of K V 10.1 exhibits a bipartite nuclear localization signal. We deleted this sequence to test its influence in the perinuclear localization of K V 10.1. For this purpose, we used an EGFP-tagged version of K V 10.1. The deleted mutant still preserves electrophysiological activity when expressed in Xenopus oocytes (data not shown). However, we did not observe abolition of the perinuclear signal (Fig. 1C), and therefore concluded that the identified nuclear localization signal of K V 10.1 is not required for its targeting to the nuclear envelope.
In heterologous systems and tumor cells, although not in the brain, it could be argued that the expression of the channel in the nuclear envelope is an artifact of massive overexpression. If this were the case, one would expect a positive correlation between overall expression levels and INM expression. However, detailed image analysis revealed that the perinuclear fluorescence intensity of K V 10.1 did not correlate with cytoplasmic intensity: 91% of ts20 cells showed higher perinuclear than cytoplasmic fluorescence intensity in the immediately neighboring region (n = 90). In 19% of cells, the perinuclear intensity was at least twice as high than the cytoplasmic one (see examples in Figure 1). Inhibition of de novo protein synthesis by cycloheximide (10 mg/mL for 1-12 hours) did not abolish the perinuclear localization, also arguing against an overexpression artifact (data not shown). This suggests that, even in overexpression systems, the localization of K V 10.1 to the nuclear envelope involves a specific regulatory mechanism.
The proximity of ONM and INM (at the scale of 40 nm [31]) is far beyond the resolution of standard confocal microscopy. We therefore carried out post-embedding immunoelectron microscopy on samples showing endogenous K V 10.1 expression using mAb62 antibodies. Approximately 90% of gold particles labeling K V 10.1 channels were detected on the INM of both the human cancer cell line MCF-7 ( Fig. 2A-C) and rat cerebellum ( Fig. 2D-E) and hippocampus ( Fig. 2F-H). The rest of gold particles were localized either in the ONM or in the adjacent cytoplasm.

The perinuclear localization of K V 10.1 is compatible with the INM
The ONM is continuous with the ER, where integral membrane proteins are synthesized. These proteins could, in principle, diffuse to the ONM and result in a perinuclear distribution pattern. To distinguish between external and INM localizations, digitonin-permeabilization experiments are often employed. Digitonin preferentially permeabilizes cholesterol-rich membranes (e.g. plasma membrane) and consequently the cholesterol-poor ER and nuclear envelope membranes remain largely intact [32]. In the selectively permeabilized cells, antibodies can only access the cytoplasmic side of ER/ONM proteins but cannot reach the INM proteins, which are shielded by the nuclear envelope. Because the ER lumen is topologically equivalent to the extracellular space, the extracellular domains of K V 10.1 should face the ER lumen or the perinuclear space, while the C-terminus is expected to face the cytoplasm [33]. Antibodies could therefore not access the INM K V 10.1, regardless of its orientation, as long as the ONM is intact.
CHO cells transiently transfected with K V 10.1-mVenus were permeabilized either thoroughly with Triton X-100 or selectively with digitonin, and then probed with antibodies recognizing either an extracellular loop (mAb66) or the intracellular C-terminus (mAb33) (Fig. 3). In the Triton X-100 permeabilized cells, signals from the K V 10.1-mVenus channel perfectly overlapped with the immunofluorescent staining ( Fig. 3 Bc,d and Cc,d). In cells permeabilized with digitonin, mAb66 was unable to label the intracellular K V 10.1 (Fig. 3 Ba,b), suggesting that the integrity of ER/nuclear envelope was preserved, and that the extracellular regions of ER K V 10.1 face the lumen. In contrast, mAb33 (Fig. 3 Ca,b) mostly stained the intracellular K V 10.1, but failed to label K V 10.1 located at the perinuclear rim, indicating that perinuclear K V 10.1 localized to a subcellular compartment distinct from the ER/ONM but compatible with the INM. Consistent staining patterns were obtained in CHO cells transfected with non-tagged K V 10.1 (not shown) and are therefore unlikely to be an effect of the mVenus fusion.
During our microscopy experiments, we noticed that perinuclear K V 10.1 was unevenly distributed, rather than being a smooth line surrounding the whole nucleus. Similar discrete nuclear envelope microdomains, devoid of NPC and therefore termed 'NPC-free islands', have already been reported for nurim [34], emerin [35] and the lamin B receptor (LBR) [36], which has been shown to recruit heterochromatin. To test if the patched distribution pattern of K V 10.1 corresponds to NPC-free islands, we performed double staining of K V 10.1 and NPC constituents. K V 10.1 frequently resided at the region of low or no NPC intensity, as indicated by a Manders Coefficient close to 0 (0.2760.17, n = 15), although NPC is nearly ubiquitous in the nuclear envelope. Perinuclear K V 10.1 was often found to be close to intense DAPI staining, which typically identifies heterochro-matin ( Fig. 4), also in live cells expressing KV10.1-mVenus and stained with Hoechst 33342 (not shown). K V 10.1 localization to NPC-free islands and proximity to heterochromatin may involve either a direct or an indirect interaction with heterochromatin. INM proteins in direct contact with histones can be detected in the 'nuclear envelope-peripheral heterochromatin fraction' [37]. However, this was not the case for K V 10.1 in the HEK-K V 10.1 cell line (data not shown) and we therefore tested for indirect interactions. An indication of heterochromatin formation in the case of certain INM proteins is the loss of lamin A/C signals in immunostaining due to epitope masking [38], although it can also be due to phosphorylation of lamin itself [39,40]. As mentioned before, K V 10.1 is localized to the same areas as lamin A/C. A closer examination revealed that, within those regions, K V 10.1 was consistently localized to regions with low lamin A/C fluorescence intensity. In 11/21 cells, K V 10.1 signal covered 4765% of the perinuclear region, and there was a masking of local lamin A/C signal but no significant effect on the total intensity (compared to the average value in the same field of view); in 5 cells, K V 10.1 signal covered 7268% of the perinuclear region, and there was a global reduction of the lamin A/C fluorescence intensity-typical for ''epitope masking''; in the rest of the cells, K V 10.1 signal occupied 66610% of the perinuclear region, but there was no obvious effect on lamin A/C signal ( Figure 5). This reduction of fluorescence could be rescued partially by extraction with high concentration of salt or incubation with DNase I, and a combination of both treatments resulted in a nearly complete retrieval of lamin A/C fluorescence. This can be interpreted as that the presence of K V 10.1 correlates with the formation of heterochromatin, either locally or globally, which further masks the epitope of lamin A/C. Perinuclear K V 10.1 is resistant to Triton X-100 extraction and shows restricted lateral mobility Higher resistance to non-ionic detergent extraction also characterizes INM proteins. To test this, CHO cells transiently expressing K V 10.1-mVenus were incubated with extraction buffer containing 3% Triton X-100 before fixation and the fluorescence obtained was examined by confocal microscopy. This extraction removed the fluorescence signals from all the cytoplasm except in the perinuclear region as compared to control cells (Fig. 6A,B). The remaining punctate cytoplasmic structures have been suggested to be aggregates resistant to extraction in similar experiments on other INM proteins [41]. Triton X-100 extraction was also applied to isolated nuclei from HEK-K V 10.1 cells and evaluated by western blotting. Efficiency of the detergent extraction of the nuclear fraction was tracked by reduction of the ER/ONM marker NADPH cytochrome c reductase. A significant amount of K V 10.1 remained in the non-extracted (INM) fraction. A fraction of K V 10.1 was also extracted, possibly representing the ER/ONM (or other contaminating organelles) pool of K V 10.1 in this overexpression system (Fig. 6B, C).
INM proteins are typically subjected to various interactions at the lamina layer and their lateral diffusion is therefore restricted as compared to their ER counterparts. To compare the mobility of K V 10.1 in nuclear envelope and ER, we carried out 1-D FRAP experiments in CHO cells expressing K V 10.1-mVenus. Two strips (2 mm wide) were bleached at both the nuclear envelope and ER, and the recovery of fluorescence intensity was monitored for 5 minutes. The ER region regained the fluorescence almost completely within 3 minutes, while the nuclear envelope region recovered slowly and only partially (Fig. 7).
Again, the significant amount of immobile K V 10.1 and the restricted diffusion of the remaining mobile protein suggest to a tight interaction of K V 10.1 with its INM environment.

Subcellular fractionation of K V 10.1 is compatible with INM localization
We then set out to confirm the localization of K V 10.1 to the INM using biochemical tools. We purified nuclei from rat brain as described in the Methods section. After ultracentrifugation through sucrose cushions, nuclei were treated with citraconic anhydride, which is described to specifically remove the ONM [45,46]. The presence of K V 10.1 and the well-established INM proteins lamina-associated polypeptide 2 (LAP2) and LUMA was investigated in the different fractions by western blot. K V 10.1, LAP2 and LUMA were all present predominantly in the nuclear fraction in comparison to the total brain homogenate (Fig. 8A). Removal of the ONM by citraconic anhydride (demonstrated by a.10-fold reduction in cytocrome c reductase activity) resulted in a concurrent enrichment of all three proteins, strongly suggesting that K V 10.1 is present in the INM in physiological conditions. In  transfected HEK293 cells, we also observed that a substantial amount of K V 10.1 was detected in the nuclear membrane fraction and could be further enriched with LAP2 and LUMA by citraconic anhydride treatment (Fig. 8B).
The relative abundance of K V 10.1 in nuclear versus extranuclear fractions was much more evident in brain extract than in transfected cells. This is compatible with 'backed up' localization previously observed in the overexpression of other INM proteins (LBR [42], MAN1 [47], nurim [41], LUMA [48] and emerin [43]); this phenomenon occurs when an excess of INM proteins saturates the INM retention sites, forcing accumulation in the ER (also observed by immunofluorescence studies in the case of K V 10.1; data not shown).
Next, we addressed the questions whether K V 10.1 at the INM comes directly from the ER, or rather has been properly targeted to the plasma membrane and only then transported to the nuclear compartment. We performed enzymatic surface-biotinylation experiments using a tagged K V 10.1 channel that carries on an extracellular loop the acceptor peptide (AP) that is recognized and modified by the biotin protein ligase birA [49]. After 20 minutes or 24 hours of biotinylation reaction, the modified proteins were pulled down and analyzed by western blot. Biotinylated K V 10.1 was preferentially found in the nuclear fractions, and a significant fraction was resistant to Triton X-100 extraction, indicating INM localization (Fig. 9).
Altogether, our results suggest a tight retention of perinuclear K V 10.1, probably in association with detergent-resistant structures such as lamina or chromatin.
Single channel activity compatible with K V 10.1 is detected in the INM The above-described data do not answer the question whether the channel located at the INM is functionally active. We therefore set up to record single channels from both membranes of isolated nuclei either directly approaching the envelope for ONM measurement or after removal of the outer membrane with citraconic anhydride for INM recordings. For these experiments, the permeability due to the NPC needs to be suppressed to unmask smaller currents. In the absence of Ca 2+ , which was our initial condition because of the strong inhibition of K V 10.1 current in the presence of Ca 2+ -calmodulin [50], a large (1 nS) conductance was present in virtually every membrane patch. This current was more evident upon de-or hyper-polarization or mechanical stress. These currents could be inhibited by known nuclear pore current blockers (mAb414, wheat germ agglutinin [51], La 3+ and Zn 2+ [52] and are therefore probably due to the NPC. Unfortunately, in our hands neither mAb414 nor wheat germ agglutinin induced complete inhibition. This made us rely on ionic blockers and to remove EGTA from the recording solution, although EGTA was still used throughout the purification process in an attempt to dissociate calmodulin from K V 10.1. Addition of W-7, a calmodulin antagonist, did not significantly increase the chance of finding an active K V 10.1 channel (data not shown), suggesting that calmodulin had indeed been effectively removed during EGTA washing. For recording, EGTA was thoroughly washed out and 10 mM La 3+ was added to the solution. We chose this concentration because it had no evident effect on plasma membrane K V 10.1, which is also inhibited by extracellular La 3+ with an apparent IC 50 of approximately 200 mM.
In asymmetrical potassium at 0 mV, we detected channel activity with a conductance compatible with K V 10.1. In symmetrical potassium, large conductance activity was more frequently observed in untreated, ONM preparations. We were unable to record reliable single channel traces from them and they will not be discussed here further. On the other hand, of 64 citraconic anhydride-treated nuclei, 27 records still showed NPC current and were thereby excluded from further analysis, 17 nuclei showed no activity, while the remaining 20 showed comparatively small channel openings. These openings were compatible with K V 10.1 in terms of conductance (8.160.4 pS; Fig. 10A, B) and voltage dependence, with highest open probability at 260 mV and lowest at +60 mV in the pipette, which would represent a membrane potential of opposite sign in an inside-out configuration, that is, if the extracellular side of the channel faces the pipette and the intracellular domains are located to the nucleoplasm. We did not detect any comparable activity in 11 NPC-current-free preparations from non-transfected cells. In asymmetrical K + , single channels with voltage dependence similar to the one expected for K V 10.1, most remarkably the severe dependence of the activation time constant on the prepulse potential described for this channel and commonly used to identify it (see e.g. ref [10]) and a chord conductance of 8.360.6 pS was detected (Figure 10 B, C). To characterize the pharmacology of the channel, we used astemizole, a H1 histamine receptor inhibitor often used as blocker of K V 10.1, because it has not been reported to block conductances other than those of the eag family including K V 10.1 at a concentration of 2-5 mM. In asymmetrical potassium, we performed continuous recordings at 0 mV to avoid contamination by channels other than those selective for K + ; under this condition, the activity compatible with K V 10.1 virtually disappeared in 7/7 patches after application of astemizole. In those cases that allowed washout of the drug (3/7), the activity was recovered. Figure 11 A depicts a similar experiment, but performed at +60 mV to increase current amplitude. In symmetrical K + , this treatment reduced the open probability at +60 mV (membrane potential) by 94610% (n = 5), also indicating a blockade of the channel by astemizole. Since the extracellular side of the channel appeared to be facing the pipette, we used mAb56, a monoclonal antibody reported to bind to the extracellular loops of K V 10.1 and inhibit current [12]. To do this, we performed layered loading of the pipette by backfilling the antibody-containing solution (45 mg/mL) over antibody-free solution, to allow recording of channel activity prior to the diffusion of the antibody to the pipette tip. The channel activity decreased in all patches tested (n = 6) within 10-20 minutes of the presence of mAb56, a time course compatible with the described action of mAb56 on K V 10.1 whole-cell current in the plasma membrane (Fig. 11B). Under this experimental setup it is not possible to fully discard a spontaneous rundown of the   current, which eventually would happen in any case, but we did not observe rundown in the time range used in these experiments. Due to the selectivity of mAb56, this result strongly supports the molecular identity of the detected channel as K V 10.1, as well as the proposed orientation in the INM implied by the above experiments, namely that the intracellular C-terminus faces the nucleoplasm.
In addition to the activity attributed to K V 10.1, it is noteworthy that we also detected other conductances in the range of 20-40 pS, which were unaffected by astemizole or mAb56.

Discussion
In contrast to traditional ideas, the nuclear envelope is increasingly viewed as a permeability barrier to ions. Although the NPC is generally believed to be responsible for the most of the permeability of the nuclear envelope, there is no correlation between NPC density and the resistivity of the nuclear envelope [20]. In addition, the intranuclear voltage is dependent on cytoplasmic potassium concentration and independent of membrane potential [27]. Therefore, the NPC should be impermeable to small ions in a physiological environment. The ONM, on the other hand, seems to be quite leaky to small molecules, including ions [53]. In this scenario, the structure responsible for the ionic and electrical differences between the nucleoplasm and cytoplasm should be the INM, and ion channels in this membrane would therefore become relevant.
We report here, for the first time, the presence of a functional voltage-gated (the InsP 3 receptor is a ligand-gated Ca 2+ channel) ion channel with a defined molecular identity in the INM. This statement is based on both optical and electron microscope data as well as biochemical evidence, both in native and heterologous systems. It not only fulfills all the criteria of a transmembrane protein localized to INM but it is also functional as an ion channel conducting potassium. The evidence in favor of an INM localization include i) inaccessibility to channel-specific antibodies after digitonin permeabilization, ii) resistance to Triton X-100 extraction, iii) limited lateral diffusion, iv) co-segregation with established INM markers and v) electron microscopy. We have used citraconic anhydride to remove the ONM. Other reports have used citric acid or Triton X-100 for the same purpose, but both have the disadvantage that they can also extract INM proteins and make electrophysiological recordings more difficult [54,55]. Additionally, our electrophysiological experiments indicate the protein orientation; the extracellular loops of the channel face the pipette, which favors localization at the INM. The identity of the channel as K V 10.1 is supported not only by the fact that it is not found in wild-type nuclei, but also by its voltage dependenceparticularly dependence on the prepulse potential-single channel conductance and pharmacology, most importantly by the inhibition measured in the presence of mAb56, which is the most selective blocker available.
We have devoted substantial efforts to make sure that this distribution is not an overexpression artifact. First, INM localization was also observed in native expression systems with immunoelectron microscopy and subcellular fractionation. Sec-ond, even in heterologous systems, the expression level in the perinuclear region is not correlated with the level in the cytoplasm, and protein synthesis inhibition does not alter the perinuclear localization, arguing against aberrant relocation of excess K V 10.1. Third, a significant fraction of the AP tagged protein at the INM has been exposed to the extracellular medium C. Traces recorded at +60 mV from a holding potential of 260 (blue) or 2100 mV (red). The latency time before the first opening is increased when the holding potential is more negative, as clearly seen in the ensemble currents depicted in the inset. Scale bars: 2 pA, 500 ms. doi:10.1371/journal.pone.0019257.g010 and does not move directly from the ER to the nuclear envelope, arguing against saturation of the transport machinery. We cannot exclude localization also in the ONM, especially in overexpression systems, but the specificity of this localization is more difficult to establish.
Two major types of NLS have been described in the soluble nucleoplasmic proteins SV40 large T antigen and nucleoplasmin. Similar NLS have been predicted in the nucleoplasmic domains of many INM proteins [56] and most of them reside in regions essential for INM localization, as determined by deletion studies (for emerin [43], MAN1 [44] and LEM2 [57], but not for LAP2-b [58]). LBR shows a NLS in its nucleoplasmic domain, but can also be targeted by its transmembrane domain [59]. On the other hand, there is no predicted NLS for nurim, and LUMA has a predicted NLS in its luminal domain [56], and both nurim and LUMA depend on their transmembrane domains for targeting to the INM [34,48]. SUN-2 bears two NLS at the nucleoplasmic domain, but can be targeted by its luminal domain, which possesses no known NLS [60]. In a similar way, while a conserved bipartite NLS can be identified in the nucleoplasmic domain of K V 10.1, its deletion however does not significantly alter the INM localization. Since the size of the nucleoplasmic domain of INM proteins able to pass through the NPC during interphase is small (,75 kDa) [56] and the fact that K V 10.1 is only functional as a tetramer, it is also possible that it is targeted to the INM during reassembly of nuclear envelope in mitosis. However, the relatively fast localization of biotinylated K V 10.1 to the INM (Fig. 9) would not be explained by such a mechanism.
The function of K V 10.1 at this location remains only speculative. It could partly underlie the oncogenic properties of the channel. INM proteins play a role in gene expression regulation [61] either by sequestering transcription factors, repressors or other regulators or by direct interaction with chromatin. Within the chromatin in the vicinity of the nuclear envelope, genes encoded in areas close to the NPC are being actively transcribed, while those located away from the NPC (in heterochromatin) are often silent. We did not observe a direct physical interaction of K V 10.1 with heterochromatin, but the presence of the channel correlated with the absence of NPC and epitope masking of lamin A/C, which is compatible with enrichment in heterochromatin. The physical structure of heterochromatin influences not only transcription efficiency but also splicing [62,63]. Additionally, K + ions increase the stability of G-quadruplex structures [28,64] (inter or intramolecular non Watson-Crick pairs in guanine-rich areas), which can act as transcriptional repressor elements (for example of the myc oncogene [65]). There is also abundant evidence of crosstalk between INM proteins and proteins involved in signaling pathways associated with cell cycle or carcinogenesis [66]. Results from our laboratory unequivocally point to a crucial role of K V 10.1 channels at the plasma membrane, because a functional antibody that blocks the current shows antitumor activity in vivo, and it can exert its action only on exposed channels [12]. The present report indicates a cross talk between the pools of K V 10.1 at the plasma membrane and at the INM. This is conceptually similar to the role of the C-terminus of a voltage-gated calcium channel, which is cleaved and acts as a transcription factor [67]. There are also descriptions of transmembrane proteins that are translocated from the plasma membrane to the INM [38]. Further experiments to clarify this question are certainly warranted.
EGFP-K V 10.1 was produced by cloning the open reading frame of K V 10.1 into pEGFPC2 (Clontech). To do this, an EcoRV site was introduced at the start codon of K V 10.1. K V 10.1-DsRed and K V 10.1-DsRed2 were generated by introducing a SacI site at the termination codon of K V 10.1 and then cloning the full ORF into pDsRedN1 and pDsRed2N1 (Clontech). The nuclear localization signal was deleted by PCR. The pCDNA3.1 K V 10.1-mVenus construct was generated by replacing the stop codon of K V 10.1 by three alanine residues and inserting the coding sequence of mVenus in frame (a mutant Venus-L221K-with less tendency to multimerize). The full coding region was sequenced to ensure that no additional mutation(s) had been introduced.
To mark surface expressed K V 10.1, the acceptor peptide for biotin ligase (AP) was inserted between the S3 and S4 loops of the channel. This peptide is recognized and biotinylated by the enzyme (BirA), which has only access to exposed parts. The sequence: TGSSGSGSGGLNDIFEAQKIEWHEGGAGGAAG-GTG was inserted after residue E317 of K V 10.1 to produce K V 10.1-AP.
All tagged-K V 10.1 exhibit the electrophysiological characteristics of K V 10.1 in heterologous expression systems (data not shown). For transfection experiments, cells were plated on glass coverslips and transfected at 70-90% confluence using Lipofectamine 2000 (Invitrogen) following the manufacturer's instructions. Immunofluorescence staining and photobleaching experiments were carried out 16-30 hours after transfection.

Immunofluorescence
Mouse monoclonal and rabbit polyclonal anti-K V 10.1 antibodies were generated in our laboratory and have been described elsewhere [1]. mAb33 and polyclonal antibodies (2413 and 9391) recognize the C-terminus of K V 10.1, while mAb62, mAb66 and mAb56 bind to the extracellular S5-S6 linker. For double staining of the NPC and K V 10.1, a commercial rabbit polyclonal antibody was used (Alomone, Jerusalem, Israel).
Other antibodies used were goat anti human lamin A/C (Santa Cruz Biotechnology, Santa Cruz, CA), mouse anti-NPC monoclonal antibody (mAb414, Abcam, Cambride, UK), rabbit polyclonal anti-LAP2, which recognizes only the transmembrane isoforms in LAP2 family, and anti-LUMA antibodies (kindly provided by Henning Otto, Freie Universitä t Berlin, Germany).
Lamin A/C staining and unmasking was performed as described in (

Detergent extraction and permeabilization experiments
For Triton X-100 extraction experiments, CHO cells transfected with K V 10.1-mVenus were incubated on ice for 5 minutes with or without 3% Triton X-100 in extraction buffer [41] (10 mM HEPES, 80 mM KCl, 16 mM NaCl, 1.5 mM MgCl 2 , 1 mM DTT, 30% glycerol, and protease inhibitor mixture -Complete, Roche Applied Science, pH 7.9). The cells were then fixed in 4% p-formaldehyde at room temperature for 15 minutes and incubated with 50 mM NH 4 Cl in PBS for 10 minutes and mounted.
Digitonin permeabilization experiments were performed as described in [59], with modifications. CHO cells transfected with K V 10.1-mVenus were fixed with 4% paraformaldehyde for 10 minutes at room temperature and incubated for 10 minutes in 50 mM NH 4 Cl in PBS. After that, the cells were incubated for 5 minutes at 4uC in 40 mg/mL digitonin (diluted from a 20 mg/mL DMSO stock), or 0.5% Triton X-100 in PBS. After a blocking step with 10% BSA in PBS for 30 minutes, the cells were incubated with the desired primary antibodies (either mAb66 or mAb33, 0.5 mg/mL) for 1 h in TBS and anti-mouse AlexaFluor 546 secondary antibodies (1:1000, 30 minutes).
Fluorescence signals were collected with the equipment indicated in each Figure. Wide field epifluorescence images were obtained in a Zeiss Axioskop 2 microscope equipped with a SPOT 1.3 camera and using the SPOT software. Confocal microscopy was performed in a laser scanning confocal microscope (TCS-SP2; Leica) using an oil immersion objective (HCX PL Apo, 636/ NA = 1.4) or a Zeiss LSM 510 Meta device using a Plan-Neofluar 40x/1.3 Oil DIC objective. ImageJ [68] and Adobe Photoshop were used for offline image processing. No non-linear image modifications were performed. The JACoP plug-in [69] was used to quantify the co-localization of K V 10.1 and NPC, and the ''Straighten Curved Objects'' plugin [70] for the straightened view shown in Figure 4C.

Nuclear isolation, surface labeling and western blot
Whole cell lysates were obtained by homogenization in denaturing lysis buffer (50 mM NaHCO 3 , 15 mM Na 2 CO 3 , 2% SDS). For Triton X-100 extraction experiments, nuclear preparations were obtained using the NucBuster Nuclear Protein Extraction Kit (Novagen, Darmstadt, Germany) according to the manufacturer's manual except for the addition of an extraction step of the nuclear pellet in extraction buffer +0.5% Triton X-100. Cytochrome c assay was performed as described in [71].
For nuclear protein preparation, the nuclei were prepared by homogenizing HEK-K V 10.1 cells in hypotonic solution or minced rat brain in nuclear isolation medium (NIM, 0.25 M sucrose, 25 mM KCl, 5 mM MgCl 2 , 10 mM Tris/HCl, pH 7.4). The crude nuclear extracts were then washed twice in NIM. The nuclei were resuspended in one volume of NIM and then mixed with two volumes of a sucrose density barrier (SDB: 2.3 M sucrose, 25 mM KCl, 5 mM Tris/HCl, pH 7.4). The whole mixtures were laid on top of SDB and centrifuged at 100,000 xg for one hour. The pellets containing nuclei were subjected to citraconic anhydride to remove the ONM as described by [45]. Then the nuclear proteins were prepared by the low-ionic-strength method [71].
To label surface-expressed K V 10.1-AP by enzymatic biotinylation, stably transfected HEK293 cells were incubated for 20 minutes at 37uC in 10 mM biotin and 120 nM BirA (Avidity, USA) in PBS (containing 4 mM MgCl 2 and 1 mM ATP). For 24 hours continuous labeling, after the described incubation, culture medium supplemented with 1 mM ATP, 10 mM biotin, 30 mM di-sodium phosphocreatine and 30 U/mL creatine phosphokinase was added to the reaction.

Electrophysiological Recordings
Nuclei from HEK-K V 10.1 cells were prepared essentially as those for biochemical analysis with the addition of 10 mM EGTA to all solutions. After sedimentation through a 2.3 M sucrose cushion, nuclei were resuspended in NIM for ONM measurements. For INM measurements, nuclei were further treated with citraconic anhydride in modification buffer (200 mM HEPES/ NaOH pH 8.5, 1 mM MgCl 2 , 250 mM sucrose) and washed twice in NIM solution. The nuclei were stored at 4uC in NIM and measured within 36 hours of preparation.
Single channel recordings were performed using bath solution 150 mM KCl, 5 mM MgCl 2 , 10 mM LaCl 3 , 10 mM HEPES/ KOH pH 7.4. The pipette solution for symmetrical K+ recordings was identical to the bath solution except that 200 mM CaCl 2 was added to improve seal formation. For recording in asymmetrical K+, the pipette contained 160 mM NaCl, 2.5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 10 mM LaCl 3 , 8 mM glucose, 10 mM HEPES/NaOH, pH 7.4. The osmolarity of all solutions was 300-330 mmol/kg. Nuclei in suspension were allowed to attach to the plastic bottom of the experimental chamber or to a poly-L-Lysinecoated coverslip for 5 minutes and were then gravity perfused extensively with bath solution to remove trace amounts of EGTA. Ellipsoid nuclei without attached membrane debris were selected visually for measurement. Currents were recorded using an EPC9 patch-clamp amplifier (HEKA Elektronik, Lambrecht, Germany) in the ''nucleus-attached'' configuration. This is equivalent to the standard cell-attached configuration for the INM, but outside-out for the ONM. Pulse generation and data acquisition were controlled with Pulse software (HEKA Elektronik). Data were filtered at 1 kHz and digitized at 5 kHz. Patch pipettes were pulled from WPI.PG10165-4 glass (World precision Instruments) or Hilgenberg thick-wall (0.5 mm) borosilicate with a resistance of 7-12 MV. Offline data analysis was performed using TAC (Bruxton Corp. Seattle, US.). Astemizole (Sigma) was prepared in 10 mM stock solution in DMSO and used at 1:5000 dilution shielded from light. Solution exchanges were completed in 5 minutes. For antibody blockage, the tip of pipette was loaded with normal pipette solution and the pipette was then backfilled with 300 nM mAb56. All electrophysiological experiments were performed at room temperature (20-22uC).

Fluorescence recovery after photobleaching (FRAP)
CHO cells plated on 40 mm coverslips were transfected as described above. Immediately before FRAP experiments, cells were incubated with Hoechst 33342 (50 ng/mL, 5 minutes at room temperature) to stain the nuclei and washed twice with TBS. FRAP was performed using a 40x, HCX PL Fluotar, 1.25 NA oil immersion objective on a Leica TCS SP II confocal microscope installed in a temperature controlled chamber. Parameters were identical in all experiments; cells were kept at 37uC in a Focht Live-Cell Chamber System (Bioptechs Inc, Butler, US.) in extracellular solution (160 mM NaCl, 2.5 mM KCl, 1 mM MgCl 2 , 2 mM CaCl 2 , 8 mM Glucose, 10 mM HEPES/NaOH pH 7.4.). At a spatial sampling frequency of 40 nm, two 2 mm strips either on the perinuclear region or through the cytoplasm were chosen as regions of interest (ROI). The cells were scanned at 1400 Hz with Ar 514 laser at transmission of 7% every 0.47 seconds for 5 times before bleaching as the reference images, in 12-bit, 5126512 pixels format and then ROI were bleached 20 times at 100% transmission. The first 10, 15 and the last 15 postbleached images were taken every 0.47, 5 and 20 seconds correspondingly. The emission light was collected in the range of 527 nm to 608 nm.
All the images were background-subtracted and then all the post-bleach images were corrected for photobleaching on the principle that the fluorescence intensity of whole post-bleached images should stay the same as the intensity outside the ROI in the reference images. Then the time-lapse images were corrected for lateral drift using StackReg [72].
The fluorescence intensities of the ROIs were normalized to a function of time [73].
F (t) is the fluorescence intensity of the ROI at time t. F 0 is the ROI intensity of the first post-bleach image and F pre is the average ROI intensity of the first 5 reference images. F (t) ' is the ROI fractional fluorescence intensity at time t. The diffusion coefficient D was calculated by fitting the data with [42]: where D is the one-dimensional diffusion constant; v is the width of the strip or the square, (2 mm in our case). t 0 was taken as the midpoint of the last bleach. F(') is the value of asymptote when t approaches infinity. All the data fitting was processed by Igor Pro 6.0.2 (WaveMetrics, Inc. Lake Oswego, OR).
Based on the fitting curves, the mobile fractions (M f ) were calculated as: where F 0 is the fractional intensity at t 0 .
Freeze-substitution and postembedding immunogold labeling MCF7 cells (DSMZ, see above) and hippocampal and cerebellar slices from 3 Sprague-Dawley rats (Harlan Research Laboratories) were used for freeze substitution. Cell cultures were immersion fixed with 4% paraformaldehyde and 0.5% glutaraldehyde in 0.1 M phosphate buffer, pH 7.15, for 1.5 hours. Rats for the hippocampal and cerebellar analysis were transcardially perfused and post-fixed with the same fixative. After fixation, a freeze substitution protocol similar to that described previously was used [74,75]. Hippocampal and cerebellar regions were carefully dissected and processed for freeze-substitution and low-temperature embedding. For post-embedding immunocytochemistry, ultrathin sections (80 nm in thickness) on nickel grids were incubated in sodium borohydride and glycine in Tris-buffered saline solution with Triton X-100. After being pre-blocked with serum, the sections were incubated with affinity-purified primary antibody mAb62 (1 mg/mL; 1:200 dilution). Primary antibody was detected with secondary antibodies conjugated to 5 nm gold particles (1:20; Amersham, Arlington Heights, USA). In some ultrathin sections from cerebellum, the gold conjugated secondary antibody was silver intensified using the Aurion silver enhancement kit (Aurion, Wageningen, The Netherlands).The specificity of the antibody has been established previously [1]. Controls included omitting mAb62 and pre-absorption of mAb62 with the corresponding blocking protein (10 mg/mL final concentration). Ultrathin sections were counterstained with uranyl acetate and lead citrate and studied with a transmission electron microscope. Electron micrographs were taken at 30,000x magnification and scanned at a resolution of 3600 dpi using a Linotype-Hell scanner (Heidelberg, Germany). Image processing was performed with Adobe Photoshop using only the brightness and contrast commands to enhance gold particles.

Ethics
None of the experiments reported here required human material, and therefore ethical approval was not necessary. The experimental procedures on rats were approved and supervised by the University of Connecticut Institutional Animal Care and Use Committee (IACUC-A08.037) in accordance with NIH guidelines.