Heteromeric Assembly of Human Ether-à-go-go-related Gene (hERG) 1a/1b Channels Occurs Cotranslationally via N-terminal Interactions*

Alternate transcripts of the human ether-à-go-go-related gene (hERG1) encode two subunits, hERG 1a and 1b, which form potassium channels regulating cardiac repolarization, neuronal firing frequency, and neoplastic cell growth. The 1a and 1b subunits are identical except for their unique, cytoplasmic N termini, and they readily co-assemble in heterologous and native systems. We tested the hypothesis that interactions of nascent N termini promote heteromeric assembly of 1a and 1b subunits. We found that 1a and 1b N-terminal fragments bind in a direct and dose-dependent manner. hERG1 hetero-oligomerization occurs in the endoplasmic reticulum where co-expression of N-terminal fragments with hERG1 subunits disrupted oligomerization and core glycosylation. The disruption of core glycosylation, a cotranslational event, allows us to pinpoint these N-terminal interactions to the earliest steps in biogenesis. Thus, N-terminal interactions mediate hERG 1a/1b assembly, a process whose perturbation may represent a new mechanism for disease.

Potassium channels encoded by the human ether-à-go-gorelated gene (hERG1) 2 are important in controlling cardiac excitability, neuronal firing frequency, tumor cell proliferation, and smooth muscle function (1)(2)(3)(4)(5)(6)(7). Clinically, they are considered the primary target for acquired long QT syndrome (LQTS) (4,5,8) caused when drugs intended for other therapeutic targets block hERG1 channels and trigger catastrophic ventricular arrhythmias and sudden death. Mutations in hERG1 give rise to congenital LQTS (9). Although only a fraction of the more than 200 potential disease-causing mutations in hERG have been analyzed in heterologous expression systems, most reduce surface membrane expression of channels because mutant subunits fail to exit the endoplasmic reticulum (ER) (10 -15). Little is known about why such LQTS-2 mutants fail to mature. In theory, maturation defects could result from misfolding, failed oligomeric assembly, or trafficking defects during translocation of channel complexes to the plasma membrane (16,17). Drugs can rescue a subset of mutants, as if by stabilizing the internal vestibule of the pore and thus compensating for folding or oligomerization defects (18).
We are only beginning to understand how ion channels assemble (19). Studies of certain voltage-gated potassium (Kv) channels and neurotransmitter receptors indicate that assembly of related subunits is specified by N-terminal interactions (20 -22). In homomeric Kv1.3 channels, N termini interact as they emerge from the translocon, well before subunit synthesis is complete (23). In contrast to Kv or ligand-gated channels, where homologous N termini mediate association, hERG subunits present an unusual challenge: they exist as two isoforms that are identical except for structurally divergent N termini. The hERG 1a N terminus comprises ϳ396 residues, whereas the 1b N terminus is a mere 56 residues, the first 36 of which are unique (Fig. 1A). The two subunits assemble in native tissues and in heterologous systems (24) where, in the absence of hERG 1a, 1b subunits fail to form robust homomeric currents (25,26). If the N termini are involved in assembly for hERG channels as for other potassium channels, the underlying mechanisms may differ substantially from those regulating association of homologous domains. Understanding these mechanisms is an important first step to uncovering new determinants of LQTS that occur because of defects in biogenesis.
In this study we tested the hypothesis that cotranslational interactions between the 1a and 1b N termini promote heteromeric subunit association. We show that hERG 1a and 1b N termini interact in cellular and in vitro assays. We utilized a truncated 1b subunit, which is retained in the ER, as a reporter of early biogenic events. Homo-oligomers of this subunit are core-glycosylated, but glycosylation can be disrupted by the introduction of heteromerizing 1a N-terminal fragments. Thus, heteromeric interactions occur before the addition of the glycan group in the ER. Such non-homotypic interactions between structurally dissimilar N termini occur cotranslationally and likely function to facilitate the heteromerization of hERG1 channel subunits.
Cell Culture-Human embryonic kidney (HEK)-293 cells were cultured in Dulbecco's modified Eagle's medium at 37°C. A cell line stably expressing 1b⌬CT protein was generated by transfecting HEK-293 cells with Myc-tagged 1b⌬CT DNA and growing in medium containing 1 mg/ml neomycin for selection. Separate cell colonies were selected, and lysate from them was probed with 1b-specific antibody to confirm expression. Cell lines expressing 1b⌬CT were then maintained in 500 g/ml neomycin.
Protein Expression and Purification from HEK-293 Cells-Cells were transfected with appropriate quantity of DNA at 70 -80% confluency using TransIT-LT1 kit (Mirus). Membrane preparations were made, 48 h post-transfection, by solubilizing cells in 150 mM NaCl, 25 mM Tris-HCl, pH 7.4, 20 mM NaEDTA, 10 mM NaEGTA, 5 mM glucose, and 0.5-1% (v/v) Triton X-100 followed by sonication and incubation for 15 min. Lysates were cleared of debris by centrifugation at 10,000 ϫ g for 15 min, and the supernatant quantified using the Bradford assay (DC Protein Assay, Bio-Rad). All steps were carried out at 4°C.
Co-immunoprecipitation-Cell lysates were precleared with 25 l of 25% protein A/G bead slurry (Amersham Biosciences) for 30 min at 4°C. Precleared lysate was incubated with respective antibody for 3 h. Lysates were further incubated with 50 l of 25% protein A/G slurry for 2 h at 4°C. Protein complexes were collected by centrifugation at 2,000 ϫ g. Beads were washed three times in 0.1% Triton X-100 containing solubilization buffer. Immune complexes were eluted in 5ϫ LSB (225 mM Tris-HCl, pH 6.8, 5% SDS, 50% glycerol, 200 mM dithiothreitol) at 65°C for 5 min. Eluted protein complexes, alongside their input lysate, were size-separated by SDS-PAGE and Westernblotted using standard methods.
Co-immunoprecipitation was used to isolate complexes of the 1aNT tm fragment with the truncated 1b⌬CT construct in stably expressing 1b⌬CT cells to observe the effects of association on 1b⌬CT core glycosylation. In contrast, in transient cotransfections the effects could be seen in the lysate without co-immunoprecipitation. We inferred that the inability to observe an effect in stable cell lysates without co-immunoprecipitation is because of the reduced efficiency of co-expression of the two constructs compared with transient transfections in which both constructs are simultaneously introduced and coordinately translated.
Protein Expression and Purification in Escherichia coli-Recombinant proteins were purified as per manufacturer's protocol (Amersham Biosciences and Novagen). Purified 1aNT-His 6 protein was eluted from Ni 2ϩ beads (Qiagen) with 250 mM imidazole and desalted and exchanged in 20 mM HEPES, pH 7.4, 500 mM NaCl by Amicon 10K column (Millipore). Purified proteins were size-separated by SDS-PAGE, stained by Coomassie Blue, and quantified by comparison with bovine serum albumin protein standards.
GST Pull-down Assays-Binding assays were carried out by incubating the indicated amounts of soluble 1aNT-His 6 with 2 M immobilized recombinant fusion proteins, 1bNT or GST. Reactions were incubated with agitation for 2 h at 4°C in a total volume of 150 l of TBS buffer containing 0.5% Triton X-100 (binding buffer). Bound proteins were washed three times with 1 ml of binding buffer, and eluted by boiling in LSB. The entire binding reaction was subjected to SDS-PAGE and Coomassie Blue staining.
Endoglycosidase Analysis-Deglycosylation was performed according to the manufacturer's protocol. 8 g of denatured lysate was treated with either endoglycosidase H (Roche Applied Science) or PNGase F (New England Biolabs) overnight at 37°C. In control reactions, the enzymes were replaced with buffer.
Pulse Chase-HEK-293 cells were grown to 70% confluency and transfected with 10 g of 1b⌬CT DNA with either 10 g of 1aNT tm or 10 g of pcDNA3.1 using Ca 3 (PO) 4 (27). Two days post-transfection, cells were starved for 15 min, pulsed with 250 Ci of 35 S Promix (Amersham Biosciences) per dish for 5 min. Cells were chased with unlabeled medium for the indicated times. Cells were suspended in lysis buffer and incubated for 30 min at 4°C. Radiolabeled lysates were clarified by centrifugation, quantified, and immunoprecipitated with either 1b-specific (for pcDNA3.1-transfected cells) or 1a-specific (1aNT tm -transfected cells) antibodies for 3 h, and immune complexes were isolated with protein A beads for 12-16 h at 4°C. Proteins were eluted with LSB, heated at 65°C for 5 min, and size-separated by SDS-PAGE. Gels were fixed in a 25% isopropyl alcohol and 10% acetic acid solution, dried at 80°C for 2 h, and exposed to a phosphorscreen. Pulse chase for 1b⌬CT stable cells was set up similarly with indicated pulse and chase times.
Densitometry and Statistical Analysis-Coomassie Bluestained gels and x-ray films of Western blots were quantified by measuring optical density using LabWorks Image and Acquisition Analysis Software (Upland, CA). Data were analyzed using PRISM 2.0 software (GraphPad, San Diego).
For in vitro binding assays ( Fig. 3), data from three separate experiments were normalized, plotted, and fitted with variable slope sigmoidal dose-response Equation 1, where X is the logarithm of concentration and Y is the response.
For assessing the degree of 1b⌬CT homo-oligomerization in the absence or presence of the disrupting 1aNT tm fragment (Fig. 5), we used the formula in Equation 2.

Level of homo-oligomerization
This formula allowed us to take in account the variability in 1b⌬CT-GFP and 1b⌬CT-Myc protein expression in different sets of lysates. Data were analyzed by running column statistics for one-sample Student's t test.
For disruption experiments ( Fig.  6), optical density (OD) values for core glycosylated/total 1b⌬CT from lysates of cells co-transfected with 1b⌬CT and each of the disruption fragments were individually compared with the fractional OD value from 1b⌬CTϩpcDNA3.1 co-transfected lysate. Data from three experiments were analyzed using a Student's unpaired t test and a two-tailed p value and are presented as mean Ϯ S.E.; n indicates the number of experiments.

RESULTS
hERG1 Subunits Assemble in the ER-Core glycosylation of proteins has been shown to be a cotranslational event in the ER lumen (29 -31). Subsequently, the glycan groups are further modified in the medial Golgi, rendering fully mature, complexly glycosylated species (32). hERG 1a subunits undergo N-linked core glycosylation in the ER at a single site between the S5 transmembrane domain and the P-region (28). We showed previously that hERG 1b subunits expressed in HEK-293 cells mature from an ϳ85-kDa core-glycosylated species in the ER to a ϳ90-kDa species in the medial Golgi (24), similar to the maturation process for hERG 1a subunits (33). Here we show both the mature and immature glycoforms of hERG 1a and 1b can be co-immunoprecipitated from cells transiently transfected with the corresponding cDNA (Fig. 1B, lane  2). The co-immunoprecipitation of immature 1a with 1b subunits indicates they assemble prior to Golgi glycosylation, likely in the ER. hERG 1a and 1b N Termini Interact in Mammalian Cells-To identify the soluble domains required for channel association, we first tested whether heteromeric interactions were preserved in the absence of the C terminus, which might mediate heteromerization. We found that 1a⌬CT and 1b⌬CT proteins co-expressed in HEK-293 cells could be co-immunoprecipitated using 1a-specific antibodies (Fig. 2B, lanes 1 and 2). This association requires interaction of the truncated proteins in vivo, because 1a⌬CT and 1b⌬CT did not associate in mixed lysates independently expressing the two constructs (Fig. 2B,  lanes 3 and 4). We conclude that the C terminus, which includes a tetramerization coiled-coil (TCC) domain (34), is not required for hERG 1a/1b heteromeric association.
To determine whether an intact hydrophobic core is required to mediate association, we made more extensive C-terminal truncations leaving only the N terminus and the first two TM domains (1aNT tm and 1bNT tm , Fig. 2C). Including the first two TMs allows the short N terminus of the 1b subunit to fully emerge from the translocon and attain proper tertiary structure (19). We co-expressed 1aNT tm and 1bNT tm in HEK-293 cells, and immunoprecipitated the cell lysate with 1a-spe- cific antibodies. 1bNT tm protein co-precipitated with 1aNT tm demonstrating the N termini and first two TMs were sufficient to support association between the 1a and 1b fragments (Fig.  2D, lanes 1 and 2). As in the previous experiment using 1a⌬CT and 1b⌬CT, 1aNT tm and 1bNT tm did not bind when lysates individually expressing each construct were mixed (Fig. 2D,  lanes 3 and 4).
Heteromeric N Termini of hERG1 Bind in a Direct Dose-dependent Manner-Attempts to demonstrate association of soluble N termini in the absence of TM were not successful because the short 1b N terminus failed to express in HEK-293 cells. Instead, we turned to in vitro binding assays to determine whether 1aNT and 1bNT polypeptides purified from E. coli could bind directly without the TM domains and the 30-aa region of homology in their N termini. We incubated 2 M bead-bound GST-1bNT with increasing amounts of soluble 1aNT from 0.3 to 10 M in separate reactions, eluted bound complex from the beads, and resolved the eluate using SDS-PAGE. Binding of 1aNT to GST-1bNT was sufficiently robust to be visualized by Coomassie Blue staining (Fig. 3A, lanes  1-6). The low efficiency of 1aNT binding is attributable to the large degree of degradation of 1bNT (visible on gel, under the indicated GST-1bNT polypeptide) and to competing homotypic association of 1aNT fragments (supplemental Fig. S1). Immobilized GST used as a negative control did not show binding to the 1a N terminus at any concentration used (Fig. 3A, lanes  7-12). A dose-response curve plotting the OD of soluble 1aNT bound to GST-1bNT as a function of 1aNT concentration shows binding occurs with an EC 50 of 0.96 Ϯ 0.2 M and a Hill coefficient of 1.52 (Fig.  3B). These results indicate that the 1a and 1b N termini bind each other directly and thus potentially mediate association of heteromeric subunits during assembly.
Truncated hERG 1b Serves as a Reporter for Early ER Events-To determine whether hERG 1a and 1b N termini interactions play a role in biogenesis, we next sought an approach to perturb their interaction and measure the effect on early events in assembly. We discovered that the C-truncated construct 1b⌬CT exists not in core and maturely glycosylated forms, as is the case for full-length 1b (24), but rather in core and unglycosylated forms (Fig. 4). We identified the core glycoform based on its sensitivity to both PNGase F and Endoglycosidase H (Endo H), which collapsed the 38-kDa 1b⌬CT to the same size as the lower, unglycosylated band migrating at ϳ35 kDa (Fig. 4A). Fig.  4B shows mature, full-length hERG 1b is insensitive to Endo H, thus ensuring the specificity of the enzyme treatment. The presence of the core and unglycosylated species presented an opportunity to identify, via perturbation, early events leading up to core glycosylation (see below).
1a N-terminal Fragments Disrupt hERG1b⌬CT Oligomerization-We exploited ER glycosylation of 1b⌬CT to identify the role of heteromeric N-terminal interactions in early assembly of hERG channels. We hypothesized that core glycosylation might report oligomerization of 1b⌬CT, and that disruption of these events by 1aNT fragments would provide evidence of heteromeric N-terminal interactions. We first tested FIGURE 2. hERG 1a and 1b N termini are sufficient for interaction. A, schematic of C-terminal-truncated 1a and 1b subunits. B, Western blot probed with Myc antibody. Lane 1, cell lysate co-expressing Myc-tagged 1a⌬CT and 1b⌬CT. Lane 2, same lysate immunoprecipitated with 1a-specific antibody. Lanes 3 and 4, IPs were ineffectual from mixed lysates independently expressing 1a⌬CT and 1b⌬CT. Lanes 5 and 6, input and IPs from untransfected controls show nonspecific and IgG bands, respectively. Input represents 5% of the total protein used for IP in each case. The entire IP eluate was loaded for each condition. C, schematic of more extensively truncated 1a and 1b subunits, including only the N termini and the first two TMs. D, Western blots probed with 1a antibody (upper panel) and 1b antibody (lower panel) showing association of 1aNT tm and 1bNT tm . Lane 1, input, and lane 2, IP using 1a antibody showing association of the two subunits. Lanes 3 and 4, mixed lysates independently expressing the two constructs show 1bNT tm failed to co-immunoprecipitate with 1a. The inputs represent ϳ10% of total protein lysate used for IP. whether 1b⌬CT subunits oligomerize by co-expressing 1b⌬CT-Myc with 1b⌬CT-GFP in HEK-293 cells. Using Myc antibodies, we found the larger 1b⌬CT-GFP fusion protein coimmunoprecipitated with the 1b⌬CT-Myc (Fig. 5A, lanes 1 and  2). The two species do not interact in mixed lysates from cells independently expressing the two constructs (Fig. 5A, lanes 4  and 5). Reciprocal immunoprecipitates using GFP antibodies gave similar results (data not shown).
1a N-terminal Fragments Cause Accumulation of Unglycosylated 1b⌬CT-The perturbation of 1b⌬CT oligomerization could occur by one of two mechanisms: the 1aNT tm fragment prevents 1b N-terminal interactions during biogenesis, or it causes dissociation of extant 1b oligomers. Preventing early assembly is expected to disrupt core glycosylation, whereas any disruption of preformed 1b⌬CT oligomers is not. By measuring core glycosylation, we can pinpoint the timing of the 1aNT tm and 1b⌬CT interaction.
We cotransfected 1b⌬CT into HEK-293 cells with one of the following constructs: empty vector (control), 1aNT tm , or Ndel (lacking N-terminal residues 1-354; cf. Fig. 1A). We assayed the effects of the disrupting fragments in 1b⌬CT lysates by measuring the core glycoform as a fraction of the total (core-plus unglycosylated) signal. As compared with the control, cotransfection of 1b⌬CT with 1aNT tm resulted in significant reduction in coreglycosylated 1b⌬CT, suggesting that the heteromeric N termini interacted prior to the glycosylation step (p Ͻ 0.005, n ϭ 3; Fig.  6, A and B). Ndel also reduced core glycosylation but to a lesser extent (p Ͻ 0.05, n ϭ 3), and we could not rule out whether this reduction was because of its competition for the glycosylation machinery. In contrast, the greater enrichment of unglycosylated 1b⌬CT by 1aNT tm could not have resulted from a direct exhaustion of the glycosylation machinery, because 1aNT tm lacks the hERG1 N-linked glycosylation site. The 1aNT tm fragment associated primarily with the unglycosylated 1b⌬CT in co-immunoprecipitation experiments, indicating the increase in unglycosylated 1b⌬CT was because of its direct physical association with the disrupting fragments (Fig. 6C). Together, the results above indicate that N-terminal interactions disrupt oligomerization and core glycosylation, reflecting early biogenic events.
Interaction with hERG 1a N Terminus Prevents Maturation of 1b⌬CT-As an alternative explanation for the increase in abundance of unglycosylated 1b⌬CT just described, we considered whether 1a N-terminal fragments might promote degradation of 1b⌬CT rather than prevent core glycosylation. If so, we would expect to see an accumulation of deglycosylated 1b⌬CT species in the cytosol en route to degradation as part of the unfolded protein response (35,36). We used pulse chase labeling to test this possibility. We co-transfected 1b⌬CT with vector (as a control) or with 1aNT tm fragment in HEK-293 cells. We radiolabeled cells for 5 min, and immunoprecipitated  expressed polypeptides from the cell lysate. The 1b⌬CT protein in mock-transfected cells manifests as both the unglycosylated and more abundant core-glycosylated forms (Fig. 7A, left  panel). In the presence of 1aNT tm , the core-glycosylated 1b⌬CT never appeared, even as early as the transition between the 5-min pulse period and the beginning of the chase period (Fig. 7A, right panel). Endo H selectively reduced the upper 1b⌬CT band in control cells, confirming its identity as the core glycoform (Fig. 7B). The rapidity with which association of 1aNT tm caused a loss of the core glycosylated band suggests 1aNT tm prevented core glycosylation of nascent 1b⌬CT, rather than promoting deglycosylation of mature 1b⌬CT en route to degradation. Fractionation studies in a stable 1b⌬CT cell line further support this conclusion by showing that unglycosylated 1b⌬CT is associated with 1aNT tm in the membrane fractions rather than the cytosol (see supplemental data). The perturbation of cotranslational glycosylation confirms that 1a and 1b N termini associate early, prior to the core glycosylation of the S5-P linker as it emerges from the translocon.

DISCUSSION
In the current study, we investigated the biochemical basis for heteromeric assembly of hERG1 channels during biogenesis. We showed that hERG 1a and 1b subunits assemble in the ER. Their N termini interact in vitro in a direct and dosedependent manner. As illustrated in the model in Fig. 8, the 1a N-terminal fragment interacts with the 1b N terminus and prevents oligomerization of a truncated 1b construct (1b⌬CT). Cotranslational glycosylation of 1b⌬CT is concomitantly reduced, suggesting heteromeric N termini interact prior to glycosylation and therefore cotranslationally. Heteromeric interaction and inhibition of glycosylation occurred within 5 min of the onset of protein synthesis as measured with pulse chase. Thus, N-terminal interactions are crucial to early assembly in the biogenesis of hERG1 channels.
Mechanisms controlling homoand hetero-multimeric assembly of membrane proteins have been explored in several other systems. In the distantly related Shaker Kv channels, homo-oligomerization is facilitated by the association of homologous tetramerizing (T1) interaction domains (22,37,38). Nascent N-terminal T1 domains interact cotranslationally, associating as they are produced on the ribosome and well before the pore-forming domains have been synthesized or properly folded (23, 39 -41). The efficiency of this process depends on the presence of the ER membrane, which is thought to concentrate the interacting domains that would otherwise be diluted in the cytosol (20,41). Moreover, the transmembrane anchors facilitate the appropriate folding of the contiguous N terminus (42)(43)(44)(45)(46). The interaction of the T1 domains is also thought to regulate heteromeric assembly: channels from divergent Kv subfamilies (e.g. Kv1.1, Kv2.1) can assemble, albeit inefficiently, only if their T1 domains have been removed (47); if the T1 domains are present, only heteromers from within a subfamily (e.g. Kv1.1, Kv1.3) will be produced (48). In contrast to Kv and ligand-gated channels, where oligomerization occurs via the interactions of homologous N termini (19,21,49,50), our results indicate the assembly of hERG 1a and 1b proceeds by the interaction of structurally dissimilar domains.  , probed with 1b-specific antibodies. B, densitometric readings were obtained for the core-glycosylated and unglycosylated 1b⌬CT proteins bands in each lysate. Error bars represent S.E. from three independent trials. C, Western blot of IP eluates from each lysate (1b⌬CT with vector, 1aNTtm, or Ndel) was probed with Myc antibody to detect coprecipitated 1b⌬CT. A representative example of three independent trials is shown.
Whether hERG 1a channels have an N-terminal oligomerization domain similar to T1 remains controversial. A hERG 1a N-terminal fragment termed NAB associates as tetramers in solution (51), whereas in crystal structures the same protein fragment exists strictly as monomers (52). Although crystallography also revealed a PAS domain predicted to be involved in protein-protein interactions (53), the specific 1a N-terminal domains mediating heteromeric association remain to be determined.
The perturbation of 1b⌬CT oligomerization by the 1aNT tm fragment indicates the 1b N termini interact cotranslationally during homo-oligomerization. This result was not necessarily predicted from analysis of the short, 36-amino acid 1b N terminus, which exhibits an excess of 8 positively charged residues. Our results indicate these charges do not present an electrostatic barrier to 1b N-terminal homo-oligomerization.
Why glycosylation is disrupted in the presence of heteromerizing N termini is not known. The hERG subunit has a single glycosylation site, and the core glycosylation machinery may require a double glycan moiety not present in the 1b⌬CT-1aNT tm complex. Alternatively, failure to fully oligomerize or global folding disturbances may reduce accessibility to the core glycosylation machinery. The heteromerizing fragment may also prevent topological changes that allow exposure of the glycosylation site to the lumen.
A tangential but important result from this experiment is that truncation of the C terminus results in accumulation of the core glycoform and reveals a requirement for the C terminus in Golgi-mediated hERG 1b glycosylation. Expression of 1a⌬CT gave similar results (supplemental Fig. S3), indicating that the hERG1 C terminus is of general importance in Golgi-mediated glycosylation. We speculate that the absence of the C terminus deprives truncated hERG1 proteins of interactions with forward trafficking partners that ensure their arrival at the Golgi. Alternatively, C-terminal truncation may prevent interaction with proteins normally required to detain hERG1 in the Golgi long enough to ensure appropriate glycosylation (54,55). Golgi-mediated glycosylation of hERG1 channels is not a requirement for function or trafficking but it does enhance channel stability (28). Therefore, indirectly the C terminus may play a critical role in channel stability.
Recent studies have established the 1b subunit as an important component of heteromeric hERG1 channels in cardiac as well as neoplastic tissues (24,56,57). hERG1 channels have been implicated in cell proliferation and invasiveness, where their ectopic or enhanced expression imparts a growth advantage to cancerous cells (6, 58 -61). In several tumor cells, such as neuroblastomas and leukemias, both 1a and 1b subunits are present as heteromeric populations (56). This heteromeric assembly, facilitated by early N-terminal interactions, may be a critical determinant of interaction with other protein partners and subcellular localization of hERG1 channels. For instance, ␤ 1 integrin interacts selectively with hERG 1a and not hERG 1b in neuroblastoma cell lines, where it triggers signaling cascades important in cell proliferation, adhesion, and migration (62).  Model showing preferential heteromeric association with 1a N terminus prevents homo-oligomerization of 1b⌬CT and interferes with cotranslational glycosylation. A, two 1b⌬CT polypeptides (N termini in red) interact as they emerge from the translocon (blue) while they are being translated on the ribosomes (yellow). B, properly assembled and folded 1b⌬CT subunits are flipped to allow for cotranslational glycosylation (inverted blue Y) between the S5 and P-loop in the ER lumen. C, in the presence of the 1aNT tm (green), cotranslational association between the heteromeric N termini prevents the homotypic 1b⌬CT interaction. D, improperly assembled subunits fail to core glycosylate.
The presence of 1b might regulate 1a N-terminal interactions with the ␤ 1 integrin and in turn determine the localization of hERG 1a/␤ 1 integrin complexes in caveolae-containing lipid rafts. hERG1 channels have been intensively studied because of their role as a target for LQTS. Trafficking defects have been identified as a major mechanism underlying congenital LQTS (10,63). Unraveling the assembly process may provide clues to the varied mechanisms by which hERG1 mutations, many of which are unique to the 1a N terminus, can result in LQTS. Mutations could reduce the efficiency of heteromeric hERG1 channel assembly, or mediate dominant-negative interactions between 1a and 1b subunits, resulting in ER retention, enhanced degradation, reduced surface expression and lower current magnitudes, all processes capable of reducing repolarizing current and causing LQTS. Additionally, C-terminal mutations could interfere with complex glycosylation thereby destabilizing channels at the cell surface and potentially reducing hERG1 current.
In summary, our results demonstrate that the N termini interact in a direct, dose-dependent manner and mediate heteromeric association of hERG1 subunits during channel biogenesis. This interaction, which takes place between structurally divergent N-terminal domains of the respective 1a and 1b subunits, is a critical step in the assembly of this unique channel.