Antidepressant-induced Ubiquitination and Degradation of the Cardiac Potassium Channel hERG

The most common cause for adverse cardiac events by antidepressants is acquired long QT syndrome (acLQTS), which produces electrocardiographic abnormalities that have been associated with syncope, torsade de pointes arrhythmias, and sudden cardiac death. acLQTS is often caused by direct block of the cardiac potassium current I(Kr)/hERG, which is crucial for terminal repolarization in human heart. Importantly, desipramine belongs to a group of tricyclic antidepressant compounds that can simultaneously block hERG and inhibit its surface expression. Although up to 40% of all hERG blockers exert combined hERG block and trafficking inhibition, few of these compounds have been fully characterized at the cellular level. Here, we have studied in detail how desipramine inhibits hERG surface expression. We find a previously unrecognized combination of two entirely different mechanisms; desipramine increases hERG endocytosis and degradation as a consequence of drug-induced channel ubiquitination and simultaneously inhibits hERG forward trafficking from the endoplasmic reticulum. This unique combination of cellular effects in conjunction with acute channel block may explain why tricyclic antidepressants as a compound class are notorious for their association with arrhythmias and sudden cardiac death. Taken together, we describe the first example of drug-induced channel ubiquitination and degradation. Our data are directly relevant to the cardiac safety of not only tricyclic antidepressants but also other therapeutic compounds that exert multiple effects on hERG, as hERG trafficking and degradation phenotypes may go undetected in most preclinical safety assays designed to screen for acLQTS.


SUMMARY
The most common cause for adverse cardiac events by antidepressants is acquired long QT syndrome (acLQTS) which produces electrocardiographic abnormalities that have been associated with syncope, torsades de pointes arrhythmias and sudden cardiac death. AcLQTS is often caused by direct block of the cardiac potassium current I Kr /hERG, which is crucial for terminal repolarization in human heart. Importantly, desipramine belongs to a group of tricyclic antidepressant compounds that can simultaneously block hERG and inhibit its surface expression. Although up to 40% of all hERG blockers exert combined hERG block and trafficking inhibition, few of these compounds have been fully characterized at the cellular level. Here, we have studied in detail how desipramine inhibits hERG surface expression. We find a previously unrecognized combination of two entirely different mechanisms: desipramine increases hERG endocytosis and degradation as a consequence of druginduced channel ubiquitination, and simultaneously inhibits hERG forward trafficking from the ER. This unique combination of cellular effects, -in conjunction with acute channel block-, may explain why tricyclic antidepressants as a compound class are notorious for their association with arrhythmias and sudden cardiac death. Taken together, we describe the first example of drug-induced channel ubiquitination and degradation. Our data are directly relevant to the cardiac safety of not only tricyclic antidepressants but also other therapeutic compounds that exert multiple effects on hERG, because hERG trafficking and degradation phenotypes may go undetected in most preclinical safety assays designed to screen for acLQTS.
Drug-induced or acquired long QT syndrome (acLQTS) poses a major problem for therapeutic drug use as well as for the development of novel drug compounds, because acLQTS produces electrocardiographic abnormalities that have been associated with syncope, torsades de pointes arrhythmias (TdP) and sudden cardiac death (SCD) (1). In most instances, acLQTS can be traced to a well-understood phenomenon: direct block of the cardiac potassium current I Kr /hERG, which is crucial for terminal repolarization in human heart (2). However, therapeutic compounds such as arsenic trioxide, which is used in the treatment of leukemia (3), or pentamidine, which is used in the treatment of Pneumocystis carinii pneumonia (4), (5) precipitate acLQTS via an unconventional mechanism: drug-induced hERG trafficking inhibition (6). Moreover, up to 40% of all direct hERG-blockers combine conventional hERG block with unconventional hERG trafficking inhibition, i.e. they exert combined hERG activity (7), (8), (9), (10), (11). Few of these compounds have been fully characterized at the cellular level. This is a crucial omission, because hERG trafficking inhibition may be missed in preclinical safety assays that have been designed exclusively for the detection of direct hERG block (12).
Tricyclic and tetracyclic antidepressants (TCA) such as maprotiline, amoxapine, imipramine or desipramine represent a large compound class with combined hERG activity that is notorious for its association with acLQTS (7), (13). TCAs have also been linked to increased risk of SCD, especially following overdosing or with dosing regimens designed to achieve high therapeutic levels (14). While TCAs have been somewhat supplanted by better tolerated serotonin reuptake inhibitors (SRIs), they are still widely prescribed in patients who do not tolerate SRIs, or for certain off-label indications, including panic disorder, migraine headaches, neuropathic pain or eating disorders (15), (16), (17), (18).
The overall goal of the present study is to understand at the cellular and molecular level how hERG surface expression is disrupted by desipramine. We address two major questions: 1) is endocytic internalization and recycling of hERG altered in the presence of desipramine? 2) Is hERG forward trafficking from the ER to the cell surface inhibited by desipramine? Our results provide a more complete mechanistic picture of the multiple effects exerted by desipramine on hERG, which ultimately should help to improve the cardiovascular safety of all TCAs.
Neonatal rat ventricular myocytes (NRVM) were isolated from dissected hearts of 1-to 2-day-old rat pups. All procedures conformed to institutional guidelines for the care and use of animals in research. Briefly, hearts were minced in HBSS and tissue fragments were digested overnight with trypsin at 4 o C. In a second step, trypsinized tissue fragments were treated repeatedly for short periods of time with collagenase at 37 o C followed by trituration. Dissociated cells were filtered, collected via centrifugation and preplated for 2 hrs at 37 o C in DMEM supplemented with 5% fetal bovine serum (FBS) and penicillin/streptomycin (PS) to remove fibroblasts.
Supernatants were collected from 'pre-plating' dishes, and NRVMs were replated in DMEM/5% FBS/PS at a density of 4 x 10 6 cells/60-mm culture dish. NRVM cultures were maintained at 37°C, 5% CO 2 with bromodeoxyuridine added to suppress fibroblast growth.
For electrophysiological studies NRVMs were grown on collagen-coated glass coverslips. Experiments were typically performed 2-4 days after initial plating.
Drugs were added to cell cultures for either 6 hr (short-term) or 16-20 hrs (overnight) prior to Western blot analysis or current recordings.
Stock solutions of desipramine, bafilomycin, astemizole and dynasore were prepared in DMSO. Final DMSO concentrations in drug-containing solutions did not exceed 0.1 %. Amitriptyline stock solutions were prepared in H 2 O. Experiments with dynasore were performed in serum-free DMEM, because dynasore binds to serum proteins and loses its activity (30).

C).
Western blot analysis. A previously described polyclonal anti-hERG antibody, rabbit HERG 519, was used to analyze hERG expression in HEK cells (29). In NRVMs HERG/I Kr expression was analyzed using a polyclonal anti-hERG-GST antibody from Alomone Labs. Briefly, HEK/hERG cells or NRVMs were solubilized for 1h at 4 o C in lysis buffer containing 150mM NaCl, 1mM EDTA, 50mM Tris, pH 7.5, 1% Triton X-100 and protease inhibitors (Complete, Roche Diagnostics, Indianapolis, IN). Protein concentrations were determined by the BCA method (Pierce, Rockford, IL). Proteins were separated on SDS polyacrylamide gels, transferred to polyvinylidene difluoride membranes and developed using the appropriate anti-HERG antibody followed by horseradish peroxidase-conjugated secondary antibody and ECL Plus (GE Healthcare, Piscataway, NJ). For quantitative analysis, signals were captured directly on a Kodak Imager R4000 (Carestream Health).
HERG internalization and recycling assays. Cleavable EZ-Link Sulfo-NHS-SS-Biotin (0.25mg/ml; Pierce) was used to biotinylate cell surface proteins expressed in HEK/hERG cells for 30min at 4 o C in PBS. The biotinylation reaction was quenched with 10mM glycine.
To measure hERG internalization, cells were incubated for designated times ranging from 5 to 60 min in complete DMEM in the absence or presence of desipramine to allow for internalization of biotinylated cell surface proteins. At the end of each internalization period, cells were treated with the membrane impermeable reducing agent MESNA (sodium 2mercaptoethane-sulfonate, 50mM in PBS) for 20min at 4 o C, to strip biotin labels from proteins remaining at the cell surface. Detergent-soluble cell lysates were prepared from MESNA-treated cells and internalized biotinylated proteins were isolated using streptavidin agarose (Thermo). Biotinylated proteins were released from streptavidin agarose by boiling in Laemmli sample buffer, resolved on SDS-PAGE and blotted with anti-hERG antibody. Non-reduced and reduced samples processed prior to internalization were used to determine the total amount of cell surface hERG present as well as any background remaining upon MESNA treatment.
To determine hERG recycling, HEK/hERG surface proteins were biotinylated as described above. Cells were then incubated in complete DMEM at 37 o C for 30min to allow for internalization of biotinylated cell surface proteins. Next, cells were treated with MESNA to remove all biotin labels remaining at the cell surface. Then, cells were washed quickly with pre-warmed complete DMEM to remove MESNA, and incubated with complete DMEM at 37 o C for designated times to allow for recycling of initially internalized biotinylated proteins back to the cell surface. Experiments were performed in the absence and presence of desipramine. At the end of each recycling period, biotinylated proteins were stripped again with MESNA. Cells were washed, lysed, and processed for hERG protein expression as described above. Note, that hERG protein resolved on Western blots has been protected from MESNA. Consequently, the recycled fraction of hERG protein was determined by the difference between initially internalized hERG (t=0) and hERG protein protected from MESNA at the end of recycling periods.

HERG ubiquitination.
In ubiquitination studies stable HEK/hERG WT cells were transiently transfected with either HA-ubiquitin or His 6 -ubiquitin cDNA using Fugene (Roche). Cells were harvested 2 days after transfection.
In experiments with desipramine cells were treated on the second day following transfection for either 1-6 hrs or overnight.
Whole cell lysates were immunoprecipitated with anti-hERG antibody. Duplicate samples of immunoprecipitates were analyzed on Western blots using either antibody to hERG or to the HA epitope fused to ubiquitin. Lysates from HEK/hERG cells transfected with His 6 -ubiquitin were used as negative control.
Briefly, HEK/hERG WT cells were starved for 30 minutes and pulse labeled for 60 minutes in 100-150 μCi/ml [ 35 S]-methionine/cysteine containing medium. Cells were harvested immediately after labeling or following different chase periods in label-free medium. Desipramine was added either for 24hrs prior to labeling or during chase periods. Cells were lysed in a 0.1% NP40 buffer in the presence of protease inhibitor. Immunoprecipitations with anti-hERG antibody (Alomone) were incubated overnight at 4°C and collected with Protein G Dynabeads (Dynal, Lake Success, NY). Immunoprecipitated radiolabeled proteins were eluted from beads via boiling, separated by SDS-PAGE, and analyzed with a STORM PhosphoImager (GE Healthcare). In pulse chase experiments image densities of fullyglycosylayted and core-glycosylated hERG were normalized to the signal of freshly synthesized, core-glycosylated hERG protein isolated immediately after radiolabeling at t=0.
To study hERG-Hsp/c70 interactions, HEK/hERG WT cells were labeled and chased in the absence or presence of desipramine as described above.
Immunoprecipitations reactions were performed either with anti-hERG or anti-Hsp/c70 antibody (Santa Cruz). Eluted samples were separated by SDS-PAGE, and analyzed with a STORM PhosphoImager (29). In some experiments, hERG-Hsp90 interactions were studied following treatment of labeled HEK/hERG cells with the chemical crosslinker dithiobis(succinimidyl propionate) (DSP, Pierce).
Cross-linking was quenched by addition of glycine and resolved by boiling in β-mercaptoethanol /SDS sample buffer to release proteins immunoprecipitated with Hsp90 antibody (Santa Cruz).
To quantify hERG/chaperone interactions at specific time points, image densities of core-glycosylated (cg) and fullyglycosylated (fg) hERG protein bands were determined on autoradiograms after immunoprecipitation with anti-hERG and anti-Hsp/c70 antibodies.
Image densities corresponding to cg-and fg-hERG found in immunoprecipitations were normalized to image densities of cg-hERG isolated immedialtely after labeling to assess timedependent changes of cg-hERG and fg-hERG synthesis in the presence or absence of desipramine as well as changes in hERG-Hsp/c70 interactions as a function of drug exposure.
Immunocytochemistry. HEK-hERG cells were grown overnight on poly-lysine coated glass coverslips under control conditions or in the presence of 30 µM desipramine. Following incubation, cells were washed with PBS and fixed in ice cold 4 % formaldehyde/PBS for 30 min. After fixation, cells were washed, permeabilized with 0.1 % Triton X-100 and blocked in 5 % goat serum/PBS for 30 -60 min at room temperature.
For double labeling, permeabilized cells were incubated overnight at 4ºC with rabbit anti-hERG GST antibody (1:100; Alomone Labs, Jerusalem, Israel) and mouse anti-KDEL antibody (1:100; Stressgen Biotechnology, Collegeville, PA, USA). The tetrapeptide KDEL, located at the carboxyterminal sequences of luminal ER proteins, is a common motif expressed in the ER that is well-suited as a compartment marker. Primary antibodies were washed off using PBS, cells were re-blocked in 5 % goat serum (30 min) and incubated for 2 hours at room temperature with secondary anti-rabbit FITC (1:100; Jackson Labs, Bar Harbor, ME, USA) and anti-mouse Rhodamine RedX antibody (1:100, Jackson Labs). Coverslips were mounted with Vectashield and examined using a Leica TCS SP2 laser scanning confocal microscope (Leica).
Detection of hERG internalization by immunocytochemistry. Stably transfected HEK/ hERG WT HA ex cells were grown on glass coverslips coated with ECL attachment matrix (Upstate/ Millipore). To label cell surface hERG-HA ex , cells were treated for 30 min at room temperature with a 1:100 dilution of rat monoclonal anti-HA antibody (Roche, highaffinity clone) in DMEM. Following removal of unbound antibody, cells were incubated at 37 o C/ 5% CO 2 in complete DMEM medium for various amounts of time in the absence or presence of desipramine to allow for channel internalization. Following incubation at 37 o C cells were washed, fixed with 4% paraformaldehyde for 15min, permeabilized, blocked with 5% donkey serum for 30 min and stained with Alexa Fluor 488-conjugated donkey anti-rat secondary antibody for 1 hr in blocking medium. Subsequently, stained coverslips were mounted with Vectashield and inspected using a Leica TCS-SP2 microscope in the fluorescent and brightfield/DIC mode to study changes in the subcellular localization of hERG as a function of incubation time.
To simultaneously detect cell surface and internalized HERG HA ex , cells were labeled with rat anti-HA antibody as described above and incubated for various time periods at 37 o C/ 5% CO 2 to allow for channel internalization. Cells were then washed, fixed with paraformaldehyde, blocked and labeled with anti-rat Alexa Fluor 488-conjugated secondary antibody prior to permeabilization to stain the cell surface pool of hERG-HA ex . Subsequently, cells were permeabilized for 5min at RT with 0.1% Tween-20. After permeabilization, cells were re-blocked with donkey serum in PBS-0.1% Tween-20 for 30min RT and stained with donkey anti-rat Red X conjugated secondary antibody to label the pool of internalized hERG channels.
To co-stain cells for HERG HA ex and the early endosomal marker EEA1, cells were first labeled with rat anti-HA antibody as described and incubated for various times at 37 o C/ 5% CO 2 in complete DMEM medium (± desipramine).
Next, cells were fixed, permeabilized with 0.1% Tween-20/PBS, blocked with 5% donkey serum and labeled with mouse anti-EEA1 antibody (BD Biosciences, 1:100) for 1hr at RT. Following washes with PBS, cells were stained with a combination of anti-rat Alexa Fluor 488-and anti-mouse Red-X conjugated secondary antibodies to co-localize hERG and EEA1 proteins.
To co-localize hERG with LAMP1-EGFP which is used as lysosomal marker (32), LAMP1-EGFP was transfected transiently into HEK/HERG HA ex cells using Fugene (Roche). Twenty four hours later, surface hERG was labeled using rat anti-HA antibody, and cells were incubated for various times in complete DMEM at 37 o C/ 5% CO 2 (± desipramine). After fixation and permeabilization cells were stained with anti-rat Red X conjugated secondary antibody for 1hr at RT to visualize hERG in addition to LAMP1-EGFP.
Data analysis. Data are expressed as mean ± S.E. of n experiments or cells studied.
Differences between means were tested using either a two-tailed Student's t-test or single factor analysis of variance followed by a twotailed Dunnett's test to determine whether multiple treatment groups were significantly different from control. P-values <0.05 were considered statistically significant.

RESULTS
Desipramine reduces hERG surface expression. We have recently reported that desipramine reduces cell surface expression of hERG within 1-2 hours (11), which is not easily reconciled with a half-life of 11 hr for hERG channels at the cell surface under control conditions (29).
To characterize desipramine effects on hERG surface expression in more detail, we exposed HEK/hERG cells for extended time periods of 1-16 hours to 30 μM desipramine, a concentration selected to produce robust shortterm effects thereby facilitating biochemical analysis. In a first series of experiments, we studied time-dependent expression changes by Western blot (Fig. 1A). For comparison, we used HEK/hERG cells treated with 0[K + ] ex , because fast internalization and degradation of hERG channels via endocytotic pathways has been demonstrated under low K + conditions (33). We found that desipramine incubation reduced the fully-glycosylated 155-kDa cell surface form of hERG (fg-hERG) within 6 hrs by 70-80% (Figs. 1A and 1B). However, expression of the core-glycosylated, 135-kDa ER resident form of hERG (cg-hERG) was not altered upon short-term incubation, yet increased significantly on long-term exposure to desipramine (16 hrs), suggesting that two independent pathways may regulate surface expression. The fast decrease in fg-hERG expression seen on Western blots was also mirrored in electrophysiological current recordings performed after desipramine had been washed out for at least 10 min. In these experiments HERG current density was reduced from 62.9±9.1 (n=17) to 27.7±4.2 pA/pF (n=12) within 6 hrs of exposure to 30 μM desipramine (Fig. 1C, D). The observed current changes reflected predominantly altered surface expression with very little residual block present, because wash-out of 30 μM desipramine was best described by a time constant of 53.3 s (n=4). Together these experiments describe a fast, drug-induced change in hERG surface expression that mimics data acquired under low K + conditions (see Fig. 1B).
Analysis of desipramine effects in neonatal rat ventricular myocytes. In addition to our experiments in a heterologous expression system, we have also studied desipramine effects on native I Kr /rERG channels expressed in cultured neonatal rat ventricular myocytes (NRVMs). To assess desipramine effects on rERG protein NRVMs were incubated with desipramine for either 6hr or overnight (17hr).
On Western blots desipramine reduced the fully glycosylated cell surface form of rERG both on short-(6hr) as well as on long-term (17h) drug exposure ( Fig. 2A). In addition, we monitored whether the reduction of cell surface rERG was accompanied by a corresponding decrease in I Kr current amplitudes. NRVM I Kr currents were isolated in symmetrical Cs + -solutions and elicited using 350ms depolarizing test pulses from a holding potential of -80 mV (Fig. 2B). Maximal I Kr tail current amplitudes measured on return to -80 mV were reduced from -14pA/pF under control conditions to -7pA/pF (n=6) following a 6hr incubation with 30 μM desipramine resembling our findings in HEK/hERG cells (Fig. 2C).
Drug-induced hERG internalization. To explore whether desipramine may alter hERG surface expression along endocytic pathways, we used a stable cell line expressing hERG WT with an HA-epitope tag inserted in the extracellular S1-S2 domain (hERG WT HA ex , (29)) to directly visualize channel internalization. In these cells hERG-HA ex was pre-labeled at the cell surface with anti-HA antibody. Subsequently, cells were incubated at 37 o C for various time periods, to allow for hERG internalization prior to fixation, permeabilization and staining with Alexa fluor 488-conjugated secondary antibody (Fig 3). Alternatively, cell surface channels were detected with Alexa Fluor 488-conjugated secondary antibody in non-permeabilized cells, while internalized channels were detected with RedX-conjugated secondary by guest on March 24, 2020 http://www.jbc.org/ Downloaded from antibody following permeabilization (supplemental Fig. 1). When cells were analyzed immediately after pre-labeling of surface channels (t=0), we detected uniform cell surface staining which was largely preserved following incubation at 37 o C under control conditions. This indicated that the majority of hERG channels remained at the cell surface. In marked contrast, incubation with desipramine initiated a fast, timedependent internalization process that resulted in removal of most channels from the cell surface and accumulation in enlarged, intracellular vesicles ( Fig. 3; supplemental Fig. 1).
Specificity of desipramine effects. Because desipramine is a cationic amphiphilic drug known to modify a wide range of physicochemical membrane properties, as well as pH of acidic intracellular vesicles, it is possible that desipramine reduces surface expression of ion channels in a global, nonspecific manner. To the contrary, we have shown previously that cardiac action potentials were prolonged in the presence of desipramine due to a targeted suppression of hERG/I Kr currents (11). To address specificity more directly, we studied heterologously expressed hKv1.5, another major cardiac potassium channel known to undergo fast endocytic recycling processes (34), (35); (36). We found that hKv1.5 currents were neither affected on short-(3-6 hrs) nor on long-term (overnight) exposure to desipramine ( Fig. 4A-C). Lack of effect was not explained by desipramineinduced inhibition of hKv1.5 internalization, because immunocytochemical analysis showed no difference in hKv1.5 internalization in the absence or presence of desipramine (supplemental Fig. 2).
To explore a channel that is more closely related to hERG with a high degree of sequence homology, we tested bEAG and found that current densities were not significantly altered upon short-term as well as overnight exposure to 30 μM desipramine. We recorded 581 ± 82 pA/PF (at +40 mV; n=13) under control conditions, 478 ± 99 pA/pF (n=15) following 3 hr exposure to 30 μM desipramine, 628 ± 107 pA/pF (n=15) following 6 hr drug exposure and 483 ± 52 pA/pF (n=11) following overnight drug exposure (supplemental Fig. 3).
One explanation for the desipramine-induced decrease in hERG surface expression may be increased channel internalization.
To disrupt internalization processes at the cell surface we inhibited the GTPase dynamin, which is essential for pinching off transport vesicles from the plasma membrane along most endocytotic pathways, using dynasore. Dynasore is a specific cell-permeable inhibitor of dynamindependent internalization pathways e.g. in neurons or HL-1 cardiomyocytes. It is most often used at a concentration of 80 μM which is thought to block about 90% of all endocytotic events (30). While short-term incubation with 80 μM dynasore did not significantly alter hERG surface expression under control conditions (Figs 5C and 5D), dynasore partially rescued hERG surface expression and function in the presence of desipramine (Fig 5A-D).
Because our experiments with dynasore indicated that desipramine increased hERG internalization, we used a biotin protection assay to directly measure channel internalization during desipramine exposure (37).
In these experiments a cleavable, membrane-impermeable biotin label was coupled to cell surface hERG prior to incubation of HEK/hERG cells at 37 o C. Biotin labels that remained at the cell surface at the end of incubation periods were stripped with MESNA, a membrane impermeable reducing agent. Subsequently, HEK/hERG cells were lysed, and internalized biotinlabeled hERG was isolated on streptavidinbased affinity columns for Western analysis (Fig. 6A). We determined that under control conditions 21.2% of initially labeled cell surface hERG was internalized within 60 min ( Fig 6B). Surprisingly, in the presence of desipramine, only 9.7 % of cell surface hERG was internalized at 60 min (Fig. 6B). We also measured channel recycling using a variant of the above described biotin protection assay (see Methods), and found that 80% of internalized channels were returned to the cell surface within 2-3 min under control or desipramine-treated conditions (Fig. 6C).
Desipramine-induced ubiquitination and degradation of hERG.
Because desipramine-induced reduction in hERG surface expression appeared to be due to neither enhanced internalization nor reduced recycling, we asked whether desipramine might cause hERG ubiquitination and degradation upon short-term drug exposure. To test with high fidelity for time-dependent changes in channel ubiquitination, we expressed hERG WT together with HA-or His 6 -tagged ubiquitin. Ubiquitin-transfected HEK/hERG cells were either cultured under control conditions or exposed to 30 μM desipramine for 1, 3 or 6 hrs. Following immunoprecipitation with anti-hERG antibody, immunoprecipitates were blotted with either anti-hERG or anti-HA antibody to detect multi-ubiquitinated hERG proteins. On Western blots we detected a fast, timedependent increase in hERG ubiquitination which mirrored the reduction of fullyglycosylated cell surface hERG seen in the presence of desipramine (Fig. 7A). In fact, the dark smear extending from 150 kDa to the top of the SDS PAGE (Fig. 7A, right panel, black bar), reflecting channel ubiquitination, increased linearly with longer drug exposure times and tripled its density within 6 hours (Fig. 7B).
In marked contrast, Kv1.5 ubiquitination was not significantly affected by desipramine (Fig. 7C), in agreement with our observation that Kv1.5 surface expression was not altered. Importantly, Kv1.5 was strongly ubiquitinated in the presence of 100 nM velcade/bortezomib, a potent proteasomal inhibitor (supplemental Fig. 4).
Because desipramine increased hERG ubiquitination, we next asked whether the channel was subsequently degraded along proteasomal or lysosomal pathways. Hence, HEK/hERG cells were incubated for 6hrs with 30 μM desipramine and either 100 nM velcade/bortezomib, a proteasomal inhibitor, or 10 nM bafilomycin, a lysosomal inhibitor. As a negative control we employed the pharmacological chaperone astemizole, which has been used to restore conformational trafficking defects of hERG, yet is not known to interfere with cellular degradation pathways (38). We found that expression of fullyglycosylated hERG was partially restored in the presence of desipramine by the lysosomal inhibitor bafilomycin as judged from Western blots, while neither velcade/bortezomib nor astemizole were effective (Fig. 8A). Interestingly, rescue of fully-glycosylated hERG expression by bafilomycin was not accompanied by an increase in functional channels at the cell surface, as determined by electrophysiological experiments (Fig.8B-D). This implies that rescued channels are retained in an intracellular compartment and are not restored to the cell surface membrane following co-incubation with bafilomycin. To identify intracellular compartments where hERG channels may traverse in the presence of desipramine, we studied co-localization of endocytosed hERG with either EEA1, a marker for early endosomes, or Lamp1, a marker for lysosomes. We detected colocalization of hERG with EAA1 antigen in the absence and presence of desipramine (supplemental Fig. 5). In marked contrast, colocalization with EGFP-tagged Lamp1 was increased in the presence of desipramine as expected from experiments with bafilomycin ( Fig. 9). Taken together, our data suggest that short-term-exposure to desipramine induces rapid ubiquitination of recycling hERG channels and movement into lysosomes for degradation.
Desipramine inhibits forward trafficking. While alterations in endocytosis may explain the reduction of fullyglycosylated cell surface hERG seen on shortterm desipramine exposure, this mechanism does not explain why the core-glycosylated ER resident form of hERG was increased with long-term drug exposure (see Fig. 1B). Since it has been shown that accumulation of coreglycosylated hERG is indicative of impaired forward trafficking, we performed pulse chase experiments to directly monitor hERG maturation in the presence and absence of desipramine. In these experiments complex glycosylation of hERG represents a welldefined marker for ER export. Accordingly, we analyzed hERG maturation in HEK/hERG cells cultured under control conditions or incubated overnight with 30 μM desipramine. We found that hERG maturation from the initially synthesized 135-kDa ER resident form to the fully-glycosylated 155-kD form was largely blocked by long-term incubation with desipramine, while synthesis and turnover of ER resident cg-hERG was not affected (Fig. 10A, B). To narrow the time course of this effect, newly synthesized hERG was labeled with 35 S, followed by chase with unlabeled media in the presence or absence of desipramine for up to 3 hours (Fig. 10C). These experiments revealed that desipramine inhibited forward trafficking within 1 hour. To test for ER retention as a consequence of longterm drug exposure, we co-labeled HEK/hERG cells with anti-hERG and anti-KDEL (a well-established ER marker) antibodies, under control conditions, and following desipramine exposure for 6hr or overnight (Fig. 11). Under control conditions hERG was stained at the cell surface, in a vesicular intracellular compartment, and in the ER. Following 6hr desipramine exposure cell surface staining was no longer detectable, while hERG was still present both in the vesicular and ER compartments. Finally, following overnight drug exposure hERG staining was restricted to the ER, where it accumulated in bright foci, suggesting that hERG channels aggregated and failed to be exported from the ER.
Does desipramine alter hERGchaperone interactions? Possible explanations for desipramine induced ER retention may be (1) inhibition of chaperone association in the hERG export pathway as described for geldanamycin (29) or (2) druginduced channel misfolding leading to prolonged hERG-chaperone interactions and ER retention as described for trafficking deficient LQT2 missense mutations (29). Consequently, we studied the interaction of hERG WT with the major cytosolic chaperones Hsp/c70 (Fig 12A-C) immediately after synthesis, or following a chase period of 6hr in HEK/hERG cells cultured under control conditions or following overnight incubation with desipramine. Importantly, there was no difference in the formation or stability of hERG/Hsp/c70 complexes in desipramine vs. control conditions (Figs. 12A and 12C). In addition, we assessed the stability of hERG/Hsp90 complexes and found that those complexes were also not disrupted by desipramine (supplemental Fig. 6). Thus, desipramine-induced inhibition of hERG forward trafficking is unlikely to be mediated via impaired channel association with the cytosolic chaperones Hsp70/90. In contrast, we were able to show that hERG ubiquitination levels were increased following overnight incubation with desipramine ( Fig  12D). Therefore, it is possible that channel ubiquitination as a consequence of aggregation may also account for the reduction of ER export seen upon long-term desipramine exposure.
Is increased channel ubiquitination a class effect of TCAs? To gain first insight we studied another TCA family member, amitriptyline, which in contrast to desipramine is a tertiary amine. While it has been reported that acute block of hERG by amitriptyline is half-maximal with concentrations of 1-3 μM (hERGAPDbase), it has not been explored whether prolonged drug exposure may also affect hERG surface expression. Therefore, we incubated HEK/hERG cells overnight with increasing concentrations of amitriptyline and found decreasing amounts of fullyglucosylated mature hERG on Western blots (Fig. 13A). In parallel, hERG tail currents were reduced with an IC 50 value of 21.2 ± 0.4 μM (n=7-13; Fig 13B). In addition, we analyzed short-term effects of 30 μM amitriptyline (Fig. 13C). In these experiments we recorded 66.3 ± 7.5 pA/pF (n=10) under control conditions, 52.0 ± 8.6 pA/pF (n=9) following 3 hr exposure to 30 μM amitriptyline, and 33.3 ± 4.8 pA/pF following 6hr drug exposure (n=10; Fig. 13D), which was comparable to current suppression seen on short-term incubation with desipramine (Fig. 13E). Having established amitriptyline effects on hERG surface expression, we tested whether reductions in surface expression were coupled to increased channel ubiquitination as shown for desipramine. To this end, we cultured HEK/hERG cells transfected with HA-ubiquitin under control conditions or in the presence of 30 μM amitriptyline for 3 and  6 hrs, or overnight. Following immunoprecipitations with anti-hERG antibody, we detected on Western blots an increase in hERG ubiquitination following shot-term as well as overnight incubation with amitriptyline (Fig. 13F). Likewise, hERG ubiquitination was increased on incubation with tetracyclic amoxapine, whose multiple effects on hERG we have characterized previously (10) (supplemental Fig. 7).

DISCUSSION
We have expanded on our previous analysis of the tricyclic antidepressant and hERG blocker desipramine (11), and describe here in detail how desipramine inhibits hERG surface expression. We find a previously unrecognized combination of two independent mechanisms: desipramine 1) increases hERG endocytosis and degradation as a consequence of drug-induced channel ubiquitination, and 2) rapidly inhibits hERG forward trafficking from the ER.
Based on Western analysis, electrophysiological current recordings, and immunocytochemical analysis we have shown that hERG channels are rapidly removed from the cell surface upon desipramine exposure. Furthermore, HERG internalization was attenuated by dynasore, an inhibitor of dynamin which is crucial for the endocytosis of many surface proteins e.g. via clathrincoated vesicles (39). In addition, dynamin has also been implicated in several clathrinindependent endocytosis pathways including caveolae-associated mechanisms which is of interest because hERG endocytosis induced under conditions of low [K + ] ex, or in the presence of probucol, appears to be linked to caveolin turnover (39), (40), (41). While we do not address the precise endocytic mechanism(s) in our study, we note that inhibition of dynamin did not affect hERG surface expression under control conditions. One explanation may be that a dynaminindependent mechanism is at play in the absence of desipramine.
Importantly, the pronounced stability of fully-glycosylated cell surface hERG does not imply that channels remain immobile at the cell surface. Instead, they undergo rapid, constitutive endocytic recycling as has been described for several other cardiac potassium channels (35), (42). Under control conditions about 20% of cell surface hERG was internalized within one hour, with 80% of all internalized channels re-appearing at the cell surface within minutes. Thus, we estimate that only about 4% of recycling hERG channels will be permanently lost from the cell surface via cellular degradation pathways over a time period of 1 hr. Interestingly, in the presence of desipramine hERG channels did not accumulate in the cell interior, as judged from our internalization assay, and channel recycling was unaffected.
This strongly suggests that desipramine did not impair endocytic channel recycling in a non-specific manner, e.g. via inhibition of calmodulin which is thought to be involved in basal endocytic recycling processes (43), (44), (45), or via accumulation of cationic amphiphilic desipramine in acidic vesicular compartments, where it may disturb organelle pH (lysosomotropic drug action, (46)). Intact cellular recycling pathways were also demonstrated by our observation that Kv1.5 as well as bEAG cell surface expression were not affected by desipramine exposure.
Since basal endocytic recycling is intact in the presence of desipramine the question remained as to how hERG channels were removed from the cell surface upon drug exposure? We found that fully-glycosylated hERG channels were rapidly and specifically multi-ubiquitinated in the presence of desipramine. Consequently, hERG channels were shifted towards higher molecular weight forms and were no longer detectable either on conventional Western blots or in internalization assays. It remains unclear whether channels are ubiquitinated at the cell surface or in an intracellular compartment. However, our experiments with dynasore suggest that ubiquitination most likely takes place in an intracellular compartment, because fully-glycosylated hERG accumulated in a functional form at the cell surface on exposure to a combination of dynasore and desipramine. It is also not clear as to how hERG ubiquitination is initiated in the presence of desipramine? One possibility may be that desipramine is disturbing cholesterol-rich membranes (47). This could be deleterious to hERG channels which are thought to localize to cholesterol-rich lipid rafts (48). A similar hypothesis has been proposed to explain how probucol, a drug known to deplete cellular cholesterol levels, may induce hERG endocytosis (41). On the other hand, hERG endocytosis cannot be induced by exposure to methyl-β-cyclodextrin, a chelator often used to extract cholesterol from cell membranes (49), (E. Ficker, unpublished data). Alternatively, desipramine may bind directly to hERG thereby altering its conformation. However, should such a binding site exist, it must be distinct from the universal drug binding site of hERG in the conduction pathway, because mutation of residues within this site does not attenuate desipramine effects on hERG surface expression (11). This is in marked contrast to what has been reported for Kv1.5 where druginduced endocytosis was initiated via binding of the blocker quinidine to the conduction pathway of Kv1.5 (50).
Once ubiquitinated, ion channels are degraded either via proteasomal or lysosomal pathways. Our data suggested lysosomal degradation, particularly in experiments with the lysosomal inhibitor bafilomycin, which was able to rescue fully-glycosylated hERG on Western blots in the presence of desipramine. While protein ubiquitination is commonly associated with proteasomal degradation, in many instances addition of ubiquitin directs proteins towards lysosomal degradation, particularly with substrates that recycle along endocytic pathways. Examples are the β2-adrenergic receptor, the chemokine receptor CXCR4 or GABA(A) receptors (51,52). Importantly, bafilomycin was not able to restore functional channels to the cell surface. This is similar to what has been described on exposure of hERG expressing cells to low K, where bafilomycin appears to arrest internalized hERG at the level of multivesicular bodies (MVBs; (53)). Nevertheless, there were important differences between low K-and desipramine-induced hERG endocytosis.
For example, direct channel ubiquitination was not unequivocally demonstrated under low K conditions. Moreover, in marked contrast to low K effects desipramine-induced hERG internalization could not be reversed with proteasomal inhibitors (33).
In addition, desipramine was unique in that effects on channel endocytosis were accompanied by rapid inhibition of hERG forward trafficking as shown in pulse chase experiments. We feel strongly that a reduction in fully-glycosylated hERG, as seen in pulse chase experiments, represents failure of ER export, because export pathways that would bypass the Golgi have not been described for hERG and are unlikely to arise in the presence of desipramine as judged from our immunocytochemical analysis.
However, important questions remain with respect to the precise mechanism(s) underlying ER retention which need to be addressed in future experiments. Importantly, desipramine did not affect crucial hERG-chaperone associations. While surprising, this was consistent with observations that trafficking could not be restored with the pharmacological chaperone astemizole which is thought to correct a wide range of conformational trafficking defects in hERG (38). A wide-ranging shut-down of all ER export appears also unlikely, because cardiac action potentials were preserved and neither Kv1.5 nor bEAG surface expression were affected on long-term exposure to desipramine.
Thus, we propose that desipramine may induce aggregation of nearnative hERG channels in the ER. As a direct consequence, aggregated channels may be ubiquitinated and degraded. The proposed mechanism may also underlie fast ubiquitination of cell surface channels and provide a common link for simultaneous changes in channel endocytosis and forward trafficking as observed on drug exposure (54).
Taken together, we have described the first example of drug-induced channel ubiquitination and degradation with direct relevance for the cardiac safety of therapeutic compounds.
Based on our data with amitriptyline and amoxapine, we speculate that the mechanisms described here for desipramine represent a class effect of all triand tetracyclic antidepressants. It seems that a combination of increased endocytosis, inhibition of forward trafficking and acute hERG block could be clinically particularly worrisome, especially in the context of overdosing and intoxications. Collectively, our data clearly elucidate why TCAs are notorious for TdP arrhythmias and SCD. Furthermore, it is conceivable that the mechanisms described here for TCAs may also apply to structurally closely related phenothiazine anti-psychotics such as thioridizine, trifluoperazine or chlorpromazine all of which are known to block hERG and reduce its surface expression at the same time (7). Finally, our data raise the important question whether low-affinity hERG blockers such as TCAs should not by all means be tested for possible effects on hERG surface expression, because acute block and chronic surface expression changes appear to operate in one and the same concentration window. Thus, focusing exclusively on the preclinical assessment of acute hERG blockade may underestimate the 'true' cardiotoxicity of compounds with multiple effects on hERG.

FOOTNOTES
This work was supported in part by National Institute Health Grants T32HL105338 (DN) and 1R01HL096962 (ID). This work was also supported in part by research grants from the Deutsche Forschungsgemeinschaft (FRONTIERS program to DT), the ADUMED foundation (to DT), the German Heart Foundation/German Foundation of Heart Research (to DT).
The abbreviations used are: acLQTS; drug-induced or acquired long QT syndrome; bEAG, bovine ether a-go-go; HEK, human embryonic kidney; hERG, human ether a-go-go-related gene; I Kr , rapidly activating delayed rectifier K current; MESNA, sodium 2-mercaptoethane-sulfonate; NRVM, neonatal rat ventricular myocytes; SCD, sudden cardiac death; TCA, tri-and tetracyclic antidepressants; TdP, torsades de pointes; Effects of 0K + on fg-hERG are included for comparison. Image densities on Western blots were normalized to fg-hERG levels measured at t=0. Note that cg-hERG is increased at t=16h (n=3-4). C, representative hERG current families recorded under control conditions or following a 6 hr exposure to 30 μM desipramine. Currents were elicited using depolarizing voltage steps from -60 to +60 mV. Tail currents were recorded on return to -50 mV. Holding potential was -80 mV. D, time-dependent reduction of hERG tail current densities on exposure to 30 μM desipramine (n=6-17). Data are given as mean ± S.E. Fig. 2. Desipramine suppresses endogenous rERG/I Kr currents in neonatal rat ventricular myocytes (NRVM) on short-term incubation. A, Western blot showing rERG under control conditions, following overnight (17h) exposure to either 10 or 30 μM desipramine, and following a 6h exposure to 30 μM desipramine. B, rERG currents recorded under control conditions or following a 6h exposure to 30 μM desipramine. Currents were elicited in symmetric Cs + solutions using depolarizing test pulses from -60 to + 60 mV. Holding potential was -80 mV. C, quantitative analysis of maximal rERG tail current densities recorded on return to -80mV under control conditions or following a 6 h exposure to 30 μM desipramine (n=6). Data are given as mean ± S.E. Note that current densities were significantly different at p<0.05 level.    Figure 1. B, quantitative analysis of hERG tail current densities measured under control conditions, following exposure to 80 μM dynasore (6h), 30 μM desipramine (6h) or a combination of 30 μM desipramine and 80 μM dynasore (6h) (n=5-8). C, Western blot showing effects of incubation with 80 μM dynasore (6h), 30 μM desipramine (6h) or a combination of 30 μM desipramine and 80 μM dynasore (6h) on hERG protein stably expressed in HEK293 cells. (fg) indicates fully glycosylated, 150-kDa cell surface form of hERG; (cg) indicates core glycosylated, 135-kDa ER-resident form of hERG. D, quantitative analysis of fg-hERG image densities following incubation with 80 μM dynasore (6h), 30 μM desipramine (6h) or a combination of 30 μM desipramine and 80 μM dynasore (6h). Image densities were normalized to fg-hERG levels measured under control conditions (n=3-4). Experiments with dynasore were performed in media that lack both albumin and serum to preserve its activity. Data are presented as mean ± S.E. Asterisks indicate significant differences at p<0.05 level.  (n=3-4). Note that incubation with desipramine appears to reduce hERG internalization. C, time-dependent recycling of internalized hERG back to the cell surface measured under control conditions or in the presence of 30 μM desipramine (n=3). Data are presented as mean ± S.E. Fig. 7. Desipramine increases HERG ubiquitination. A, Western blot analysis of HEK/hERG cells transiently transfected with HA-tagged ubiquitin and treated for 1, 3 or 6h with 30 μM desipramine. Transfection with HIS 6 -ubiquitin was used as negative control. Whole cell lysates, shown in the right part of panel were immunoprecipitated with anti-hERG antibody, resolved by SDS-PAGE and immunoblotted using either anti-hERG antibody (hERG-IP) or an antibody recognizing the HA epitope fused to ubiquitin (HA-Ub). Immunoblotting with anti-HA antibody identifies high molecular weight forms of ubiquitinated hERG that are increased on prolonged exposure to 30 μM desipramine. B, quantitative analysis of hERG ubiquitination as a function of desipramine exposure (n=3). Image densities corresponding to ubiquitinated hERG were measured in a region of HA-Ub Western blots indicated by black bar to the right of panel (A). Images densities were normalized to Ub-hERG levels measured under control conditions. C, Western analysis of HEK cells co-transfected with h-Kv1.5myc and either HA-or HIS 6 -tagged ubiquitin. Shown are immunoprecipitations of hKv1.5 protein with anti-myc antibody under control conditions or following treatment with 30 μM desipramine for 6h. Samples were analyzed using anti-HA antibody to identify putative high molecular weight forms of ubiquitinated hKv1.5. Transfection with HIS 8 -ubiquitin was used as negative control. Asterisks to the right of panel indicate non-specifically stained protein bands. Note that desipramine does not increase HA-Ub staining in high molecular weight region (black bar to the right of panel) where multi-ubiquinated hKv1.5 would be expected. (cg) indicates core glycosylated, 135-kDa ER-resident form of hERG. C, quantitative analysis of fg-hERG image densities following incubation with 30 μM desipramine (6h), a combination of 30 μM desipramine and 10 nM bafilomycin (6h), and with 10 nM bafilomycin alone. Image densities were normalized to fg-hERG levels measured under control conditions (n=3-6). D, quantitative analysis of hERG tail current densities measured under control conditions, on exposure to 30 μM desipramine (6h) and on exposure to a combination of 30 μM desipramine and 10 nM bafilomycin (6h; n=5-7). Note that bafilomycin does not rescue hERG current levels in the presence of desipramine. Data are given as mean ± S.E. Asterisks indicate significant differences at p<0.05 level. Fig. 9. Desipramine increases co-localization of hERG channels with Lamp1. HEK/hERG-HA ex cells transiently transfected with Lamp1-GFP, a lysosomal marker protein, were pre-labeled with anti-HA antibody and incubated for 4h either under control conditions or in the presence of 30 μM desipramine prior to fixation, permeabilization and staining with RedX-conjugated secondary antibody. Shown are representative confocal images. Scale bar: 20 μm. Fig. 10. Desipramine inhibits hERG forward trafficking. A, pulse-chase analysis of hERG maturation in 35 S-labeled HEK/hERG cells under control conditions or following overnight incubation with 30 μM desipramine. Radiolabeled hERG was isolated by immunoprecipitation after chase periods indicated. Arrows indicate position of fully-glycosylated (fg), and coreglyosylated (cg) forms of hERG. B, quantitative analysis of time-dependent changes of fg-and cg-hERG densities measured under control conditions or following overnight incubation with 30 μM desipramine (n=3). C, pulse chase experiment performed with 30 μM desipramine present during chase period but not during synthesis of hERG protein. Note fast suppression of hERG maturation by desipramine. Data are given as mean ± S.E. Fig. 11. Subcellular immunolocalization of hERG protein under control conditions, following 6h exposure to 30 μM desipramine (desip., 6hr), and following overnight exposure to 30 μM desipramine (desip., o/n). HEK/hERG cells were fixed, permeabilized and double labeled with anti-hERG and anti-KDEL antibody which was used as marker of the endoplasmic reticulum (ER). In untreated control cell hERG was localized to the cell surface, an intracellular vesicular fraction and the ER. In cells treated with desipramine for 6h hERG staining was no longer detected at the cell surface. Following overnight treatment with desipramine hERG staining was restricted to the ER where it accumulated in brightly stained foci. Scale bar: 20 μm. Note that Hsp/c70 association is not different between control and desipramine treated cells. D, Western blot analysis of HEK/hERG cells transiently transfected with HA-tagged ubiquitin and treated for 6h or overnight (o/n) with 30 μM desipramine. Transfection with HIS 6 -ubiquitin was used as negative control. Whole cell lysates were immunoprecipitated with anti-hERG antibody, resolved by SDS-PAGE and analyzed using either anti-hERG antibody or an antibody recognizing the HA epitope fused to ubiquitin. Immunoblotting with anti-HA antibody identifies high molecular weight forms of ubiquitinated hERG that are increased following either 6h or overnight exposure to 30 μM desipramine.  (n=7-13). C, representative hERG current families recorded under control conditions or following a 6 hr exposure to 30 μM amitriptyline. Currents were elicited using depolarizing voltage steps from -60 to +60 mV. Tail currents were recorded on return to -50 mV. Holding potential was -80 mV. D, time-dependent reduction of hERG tail current densities recorded on exposure to 30 μM amitriptyline (n=9-10). Note that current density at t=6h is significantly different from control, Dunnett's, p< 0.05. E, normalized timedependent changes in tail current levels recorded in the presence of either 30 μM desipramine or 30 μM amitriptyline. F, Western blot analysis of HEK/hERG cells transiently transfected with