E3 ubiquitin ligase rififylin has yin and yang effects on rabbit cardiac transient outward potassium currents (Ito) and corresponding channel proteins

Genome-wide association studies have reported a correlation between a SNP of the RING finger E3 ubiquitin protein ligase rififylin (RFFL) and QT interval variability in humans (Newton-Cheh et al., 2009). Previously, we have shown that RFFL downregulates expression and function of the human-like ether-a-go-go-related gene potassium channel and corresponding rapidly activating delayed rectifier potassium current (IKr) in adult rabbit ventricular cardiomyocytes. Here, we report that RFFL also affects the transient outward current (Ito), but in a peculiar way. RFFL overexpression in adult rabbit ventricular cardiomyocytes significantly decreases the contribution of its fast component (Ito,f) from 35% to 21% and increases the contribution of its slow component (Ito,s) from 65% to 79%. Since Ito,f in rabbits is mainly conducted by Kv4.3, we investigated the effect of RFFL on Kv4.3 expressed in HEK293A cells. We found that RFFL overexpression reduced Kv4.3 expression and corresponding Ito,f in a RING domain–dependent manner in the presence or absence of its accessory subunit Kv channel-interacting protein 2. On the other hand, RFFL overexpression in Kv1.4-expressing HEK cells leads to an increase in both Kv1.4 expression level and Ito,s, similarly in a RING domain–dependent manner. Our physiologically detailed rabbit ventricular myocyte computational model shows that these yin and yang effects of RFFL overexpression on Ito,f, and Ito,s affect phase 1 of the action potential waveform and slightly decrease its duration in addition to suppressing IKr. Thus, RFFL modifies cardiac repolarization reserve via ubiquitination of multiple proteins that differently affect various potassium channels and cardiac action potential duration.

Genome-wide association studies have reported a correlation between a SNP of the RING finger E3 ubiquitin protein ligase rififylin (RFFL) and QT interval variability in humans (Newton-Cheh et al., 2009).Previously, we have shown that RFFL downregulates expression and function of the humanlike ether-a-go-go-related gene potassium channel and corresponding rapidly activating delayed rectifier potassium current (I Kr ) in adult rabbit ventricular cardiomyocytes.Here, we report that RFFL also affects the transient outward current (I to ), but in a peculiar way.RFFL overexpression in adult rabbit ventricular cardiomyocytes significantly decreases the contribution of its fast component (I to,f ) from 35% to 21% and increases the contribution of its slow component (I to,s ) from 65% to 79%.Since I to,f in rabbits is mainly conducted by Kv4.3, we investigated the effect of RFFL on Kv4.3 expressed in HEK293A cells.We found that RFFL overexpression reduced Kv4.3 expression and corresponding I to,f in a RING domaindependent manner in the presence or absence of its accessory subunit Kv channel-interacting protein 2. On the other hand, RFFL overexpression in Kv1.4-expressing HEK cells leads to an increase in both Kv1.4 expression level and I to,s , similarly in a RING domain-dependent manner.Our physiologically detailed rabbit ventricular myocyte computational model shows that these yin and yang effects of RFFL overexpression on I to,f, and I to,s affect phase 1 of the action potential waveform and slightly decrease its duration in addition to suppressing I Kr .Thus, RFFL modifies cardiac repolarization reserve via ubiquitination of multiple proteins that differently affect various potassium channels and cardiac action potential duration.
The QT interval is an important diagnostic feature on surface electrocardiograms because it reflects the duration of the ventricular action potential.Genome-wide association studies have provided evidence for the association of the human cardiac QT interval with a genetic variant near the gene encoding the E3 ubiquitin (Ub) ligase RING finger and FYVE-like domain-containing protein (RFFL) on chromosome 17q12 (1)(2)(3).Ubiquitylation (also known as ubiquitination) targets most proteins for degradation by the 26S proteasome.Other functions of ubiquitination include internalization and lysosomal targeting, modulation of protein interactions, alteration of subcellular distribution, regulation of transcription, DNA repair, propagation of transmembrane signaling, and antiviral immune response (4,5).Ubiquitination occurs through the sequential action of three classes of protein: Ub-activating enzymes (E1s), Ub-conjugating enzymes (E2s), and Ubprotein ligases (E3s).
There are multiple targets of the RFFL E3 Ub ligase.RFFL is involved in endocytic trafficking, which is regulated by ubiquitination of cargoes and endocytic machineries.An RFFL dominant-negative mutant induced clustering of endocytic recycling compartments and delayed endocytic cargo recycling (6).RFFL also contributes to tumorigenesis by repressing caspases and tumor suppressor genes (7).Of note, upregulation of RFFL leads to ubiquitination and degradation of the PRR5L subunit of mammalian target of rapamycin complex 2 (8).Importantly, RFFL is responsible for chaperoneindependent ubiquitination of misfolded cystic fibrosis transmembrane conductance regulator (CFTR) (9), since the most common CFTR mutant, ΔF508-CFTR, is removed from the plasma membrane for lysosomal degradation by ubiquitination.Excess of RFFL also inhibits recycling from the endocytic recycling compartment (10), indirectly affects endocytosis, and enhances intracellular protein polyubiquitination (11).
To clarify the effects of RFFL on the QT interval and specifically on the repolarization reserve in larger animals, we initially studied how RFFL affects the major repolarizing current in large mammals, specifically the rapidly activating delayed rectifier potassium current (I Kr ) in adult rabbit ventricular cardiomyocytes (ARbCMs).We have found that RFFL polyubiquitinates the corresponding human-like ether-a-gogo-related gene (hERG) potassium channel in the endoplasmic reticulum (ER), leading to proteasomal degradation of hERG and to an almost complete disappearance of I Kr.This activity of RFFL depended on an intact RING domain, which is necessary for Ub ligase activity.Loss of hERG reduces repolarization reserve and, based on computational studies, increases the QT interval (12).Surprisingly, RFFL has a negative correlation coefficient with mouse cardiac hypertrophy caused by a 3week isoproterenol treatment (13).Of note, it has been found that a congenic strain of Dahl salt-sensitive rat with a genomic segment encoding RFFL from the normotensive Lewis rat had significantly higher expression levels of both RFFL mRNA and protein, correlating with a significantly shorter QT interval than the original Dahl salt-sensitive rats (14).This indicates that the increase in RFFL likely leads to a shortening of the action potential duration (APD) in rat ventricular myocytes, where the major repolarizing current is not I Kr , but the transient outward potassium current (I to ).Therefore, to further clarify the effects of RFFL on the QT interval in the rat and larger animals, we studied RFFL effects on I to in rabbit cardiomyocytes and on corresponding channel proteins expressed in human embryonic kidney (HEK) cells.We have found that RFFL overexpression decreases the fast component (I to,f ) and increases the slow component (I to,s ) of I to in rabbit myocytes.This result correlates with RFFL-induced changes in HEK cells expressing Kv4.3 and Kv1.4 alpha subunits of the corresponding channel proteins.In combination with our previously published findings (12), this article explains the opposite effects of RFFL on QT interval in rodents and larger mammals.

RFFL overexpression reduces I to,f and increases I to,s in ARbCMs
To investigate the effect of RFFL on I to , we transduced ARbCMs with adenovirus encoding GFP or RFFL, and 48 h later, we measured I to in the whole-cell configuration.The holding potential was −70 mV and I to was evoked by a voltage step to +50 mV. Figure 1, A and B show a representative experiment with overlapping I to traces evoked by a paired pulse protocol with varying interpulse intervals (Δt, see details in the Experimental procedures) in GFP-expressing ARbCMs.Figure 1C shows an experiment similar to Figure 1A, but with ARbCMs-expressing RFFL.The top inserts in Figure 1, A and C show two pulses with Δt = 5s, as an example of 1 out of 15 Cardiac RFFL regulates I to currents and matching proteins different pulse protocols with various interpulse durations.The first (noninactivated) I to peak amplitudes were similar in these two groups: 8.7 ± 1.0 pA/pF in GFP-expressing ARbCMs (n = 12) and 8.3 ± 1.6 pA/pF in RFFL-expressing cells (n = 10).
Traces of I to recovery from inactivation after the first pulse are shown in Figure 1, A and C. Yet, the datasets analyzed with a mixed effects model test did not show any significant difference between GFP-and RFFL-expressing cells, due to the large variation of I to amplitudes between individual cells.
The fast inactivation time constants of I to for GFP-and RFFL-expressing cells were 22 ± 1.5 and 19 ± 0.8 ms, correspondingly (p = 0.26), while slow inactivation constants were 180 ± 16 and 218 ± 58 ms, correspondingly (p = 0.32).Thus, we found no significant difference neither in the fast I to inactivation kinetics nor in the slow I to inactivation kinetics between GFP-and RFFL-expressing myocytes.
Using normalized amplitude data in which the second I to peak amplitude was divided by the corresponding first peak amplitude in each paired pulse recording (Fig. 1D), we found a significant difference between the I to recovery curves of GFPand RFFL-expressing cells by utilizing a mixed effects model test (p < 0.05).In addition, at both 400 ms and 600 ms interpulse intervals the I to values in RFFL-and GFP-expressing cells are significantly different (p < 0.05, Student's t test).Thus, RFFL suppresses I to recovery at 0.4 and 0.6 s interpulse intervals (red asterisks in Fig. 1D).
To clarify the origin of the difference between the I to recovery curves, we fit these two curves with a sum of two exponential recovery functions, described by the following equation: I to = I to,f (1 -exp(-Δt/τ f )) + I to,s (1 -exp(-Δt/τ s )).The best fits of the curves gave us recovery time constants for I to,f components (τ f ) equal to 50 ± 10 ms and 18 ± 8 ms, but they were not statistically different (p > 0.05) due to large variation of measured I to,s because of relatively large partially uncompensated capacitive transients in these short recovery time intervals (<100 ms).At the same time, I to,s recovery components (τ s ) were equal to 2.5 ± 0.3 s and 2.8 ± 0.2 s for GFP-and RFFL-expressing cells, respectively (Fig. 1D).Thus, we did not uncover any significant effects of RFFL on the time constants of I to recovery.
However, in control, GFP-expressing cells, the I to,f maximal amplitude (black thin line level at 30 s in Fig. 1D) contributed 35 ± 3% to the total I to (black thick line), while I to,s (black dashed line) provided the remaining 65 ± 3%.Whereas, in RFFL-expressing cells, the I to,f amplitude (red thin line) contributed only 21 ± 2%, which was significantly less than in GFP-expressing cells (p < 0.01).Correspondingly, the slow component in RFFL-expressing cells (red dashed line) provided 79 ± 2% to the total current (thick red line), which was significantly larger than in GFP-expressing cells (p < 0.01).Thus, these results imply that RFFL overexpression in ARbCMs has opposite effects on the expression of Kv4.3encodedI to,f , and Kv1.4-encodedI to,s .

RFFL downregulates I to,f and Kv4.3 expression in a RING domain-dependent manner
We used HEK293A cells to test whether we could use this model to recapitulate the effect of RFFL on I to seen in ARbCMs.For the electrophysiological experiments, HEK cells were stably transfected with an expression plasmid encoding Kv4.3, the main pore-forming subunit for I to,f in the hearts of rabbits, and transiently cotransfected with plasmids encoding enhanced GFP, RFFL with GFP (RFFL) or RING domain deleted RFFL with GFP (RFFL-ΔRING) with or without the Kv4.3 accessory subunit Kv channel-interacting protein 2 (KChIP2) with DsRed (see Experimental procedures for details).
We found that RFFL inhibited I to,f in a RING domaindependent manner in the absence of KChIP2 (p < 0.01) and in the presence of KChIP2 (p < 0.05, Fig. 2).Similar to previous studies in HEK cells (15,16), we noticed a robust, that is, 7-fold increase in I to,f caused by coexpressed KChIP2 alone (p < 0.001, solid and dashed green lines in Fig. 2B).In the presence of KChIP2, RFFL coexpression caused a 2-fold decrease in I to,f relative to GFP-only coexpressing cells (p < 0.05).This decrease is similar to the 1.7-fold decrease of the fast-recovering I to component in ARbCMs caused by RFFL transduction (Fig. 1D).
To clarify whether the effects of RFFL on I to,f are due to changes in Kv4.3 and/or KChIP2 trafficking or because of modifications of its intrinsic properties, we analyzed the inactivation kinetics of I to,f measured in HEK cells expressing Kv4.3 and KChIP2.Kv4.3I to,f has fast and slow inactivating components.These components in GFP-expressing cells had 16.9 ± 0.7 and 77 ± 15 ms inactivation time constants.While in RFFL-expressing cells these time constants were 15.0 ± 1.3 and 87 ± 23 ms.The ratio of the amplitudes of the fast to slow inactivating components of I to,f (i.e., amplitude of I to,f,fast-inactivating / amplitude of I to,f,slow-inactivating ) in GFP-and RFFL-expressing cells were 14.0 ± 2.2 and 8.9 ± 3.4, respectively.Thus, there were no significant differences in Kv4.3 produced I to,f kinetics between GFP-and RFFL-expressing cells.This implies that RFFL does not modify the kinetics of inactivation of Kv4.3 produced I to,f .
Next, we wanted to assess whether Kv4.3 levels on the membrane were indeed impacted by RFFL ubiquitination activity.To this end, we transiently expressed Kv4.3, RFFL, its RING deletion (RFFL-ΔRING), which is devoid of ubiquitination activity, or a control plasmid in HEK cells for 48 h.Surface biotinylation assays were conducted to look at the respective surface and total levels of Kv4.3.Cell surface and/or input levels of monomeric Kv4.3, Flag-tagged RFFL, transferrin receptor, and GAPDH are shown in Figure 3A (lanes 1-3).As anticipated, we noticed a robust downregulation of cell membrane (−52%) and total Kv4.3 (−40%) in the presence of RFFL (Fig. 3, B and D).This effect was dependent on the functional Ub ligase domain, as coexpression of RFFL-ΔRING not only prevented the RFFL-dependent downregulation of Kv4.3 levels but rather had the opposite effect, resulting in an approximately 2-fold increase in total and membrane levels of Kv4.3 as compared to control.
As KChIP2 is an important accessory subunit for Kv4.both cell membrane (−34%) and total Kv4.3 levels (−19%), albeit this effect was significantly mitigated compared to the negative effect of RFFL on Kv4.3 in the absence of KChIP2 (Fig. 3, B and D).Coexpressed KChIP2 also abolished the positive effect of the RING deletion on total and surface Kv4.3 levels.We also noticed that KChIP2 levels, in the presence of Kv4.3, were approximately 5-fold increased by RFFL and that this significant effect relied on the intact RING domain of RFFL (Fig. 3C).The biochemical data somewhat supports earlier findings that KChIP2 protects Kv4.3 from degradation in the ER and promotes its forward trafficking (15) and implies a role for overexpressed RFFL in an ER-dependent degradation of Kv4.3.
It should be noted that in the Western blot analysis, we measure changes from thousands of cells, while in electrophysiology we measure the currents only from dozens of "healthy looking" cells, and we cannot expect linear correlation between these the data obtained by different methods.However, the expression levels of surface Kv4.3 at all studied conditions (Fig. 3A) changed in the same directions as the corresponding I to,f amplitudes (Fig. 2B).

RFFL upregulates I to,s and Kv1.4 expression in a RING domaindependent manner
To investigate the effects of RFFL on I to,s , HEK cells were stably transfected with an expression plasmid for Kv1.4 and transiently cotransfected with plasmids encoding GFP (control), RFFL, or RFFL-ΔRING mutant.The amplitude of I to,s in HEK cells expressing RFFL was significantly larger than in GFP-expressing cells (p < 0.05, Fig. 4).This RFFL effect required an intact RING domain.The amplitude of I to,s in RFFL-ΔRING-expressing cells was significantly smaller than in the cells expressing RFFL (p < 0.05, Fig. 4).The difference between the I to,s amplitude in RFFL-ΔRING-expressing cells and GFP-expressing cells was not significant (p > 0.05, Fig. 4).
To determine whether increased RFFL changes Kv1.4 trafficking or modifies its channel properties, we analyzed inactivation kinetics of I to,s measured in HEK cells expressing Kv1.4.The fast and slow inactivation time constants in GFP-expressing cells were 19.2 ± 1.2 and 226 ± 10 ms, respectively.Whereas in RFFL-expressing cells, the constants were 18.5 ± 0.8 and 211 ± 11 ms.The ratio of the amplitudes of the fast to slow inactivating components of I to,s in GFP-and RFFL-expressing cells were 5.4 ± 0.3 and 5.6 ± 0.3, correspondingly.Therefore, similar to Kv4.3, RFFL did not modify the Kv1.4-encodedI to,s kinetics of inactivation.This suggests that RFFL primarily affects Kv1.4 trafficking, rather than I to,s kinetics.
To that end, surface biotinylation experiments were conducted to study the effect of RFFL on surface and total levels of coexpressed Kv1.Cardiac RFFL regulates I to currents and matching proteins significantly increased total (+94%) and surface expression (+140%) of both glycosylated forms of Kv1.4 (p < 0.05) (Fig. 5, B and C).The RFFL-dependent increase in the expression of the high-mannose ER-resident form of Kv1.4 would also suggest that RFFL likely exerts its effect on Kv1.4 channels on the membrane of the ER.Deletion of the RING domain of RFFL, however, fully eliminated the upregulation of Kv1.4 at both the surface and total levels (Fig. 5B).Thus, the RING-dependent positive effect of RFFL on Kv1.4 surface expression is in agreement with the corresponding change in I to,s amplitude (Fig. 4).

Computer modeling of RFFL influence on rabbit action potential
Previously, we have shown that overexpressed RFFL downregulates I Kr and hERG protein in adult rabbit cardiomyocytes and HEK293A cells (12).Here, we show that RFFL also downregulates I to,f (Figs.1D and 2B) and Kv4.3 (Fig. 3B) but upregulates I to,s (Figs.1D and 4B) and Kv1.4 (Fig. 5).To determine the effect of RFFL on the cellular excitability and APD at 90% repolarization level (APD90), we conducted electrophysiological experiments with adenovirally transduced ARbCMs.Yet, a large variation of APD90 values in these remodeled cultured cells did not allow us to determine a statistically significant difference in APD90 between the cells expressing GFP and RFFL (data not shown).
Therefore, we utilized our spatially detailed model for rabbit ventricular myocytes (18) and presented here assessments of RFFL effects on the AP at 2.5 Hz stimulation in four different conditions: 1. control (g to,f = 0.054 mS/μF, g to,s = 0.1 mS/μF, and normal g Kr ). 2. RFFL caused full suppression of only I Kr with normal I to (g to,f = 0.054 mS/μF, g to,s = 0.1 mS/μF, and g Kr = 0).3. RFFL modified I to with normal I Kr (g to,f = 0.032 mS/μF, g to,s = 0.122 mS/μF, and normal g Kr ). 4. RFFL modified I to and completely inhibited I Kr (g to,f = 0.032 mS/μF, g to,s = 0.122 mS/μF, and g Kr = 0), (Fig. 6).
Computer modeling shows that at a stimulation of 2.5 Hz, the combined effect of a decrease in I to,f and an increase in I to,s , mimicking RFFL overexpression, reduces APD90 only by 3.0 ms and 3.9 ms with normal and completely inhibited I Kr , respectively (Fig. 6A).The decrease in I to,f (Fig. 6F) led to a smaller AP notch and thicker AP overshoot (Fig. 6A insert), while the increase in I to,s was the major contributor to APD90 shortening both with normal and completely inhibited I Kr (Fig. 6, A, D, and E).However, the major effect of RFFL on APD90 is due to its suppression of I Kr (Fig. 6, A and D), which leads to an increase of 20.6 ms in APD90 when I to,f and I to,s are unaffected (Fig. 6A).With the changes in I to,f and I to,s included (Fig. 6, E and F), the increase in APD90 due to I Kr suppression is slightly reduced to 19.7 ms (dot-dashed versus solid lines in Fig. 6A).Furthermore, the overexpression of RFFL also affects Ca 2+ cycling.The more positive notch when I to,f and I to,s are affected results in a decreased peak of the L-type Ca 2+ current (I Ca ) (Fig. 6C) and, consequently, a reduced peak of intracellular Ca 2+ concentration (Fig. 6B).Additionally, the reduced prolongation of APD90 leads to a slight decrease in intracellular Ca 2+ concentration during the decay phase (Fig. 6B).In other words, the effect of RFFL overexpression induced changes to APD90 via I to,f and I to,s , and the effect of the notch potential on I Ca partially counteracts the impact of I Kr suppression by RFFL overexpression.The APD90 prolongation of 20 ms due to the full I Kr suppression may seem inconsistent with the much more significant prolongations observed in existing experimental LQT2 rabbit models (19).However, it is notable that this LQT2 rabbit model involves an additional 30% inhibition of I Ks , a longer pacing cycle length, and the presence of isoproterenol.As shown in Table S3 produced by our computational rabbit model, an additional 30% inhibition of I Ks increases the APD90 from 217 ms to 266 ms, that is, by 49 ms.Under this condition, RFFL-induced I to changes reduce APD90 by 8 ms (from 266 ms to 258 ms).With the stimulation frequency decreased from 2.5 Hz to 0.25 Hz, which is close to the experimental frequency in Ref. (19), the APD90 increases from 284 ms to 349 ms, that is., by 65 ms, while RFFL-caused I to changes have no noticeable effect on APD90.Furthermore, in the presence of isoproterenol and at the cycle length of 4 s, the full I Kr blockade and 30% suppression of I Ks prolong the APD90 by 94 ms (from 322 ms to 416 ms), and the RFFL-caused I to changes reduce the APD90 by 13 ms (from 416 ms to 403 ms).It is notable that this APD90 prolongation by 94 ms quantitatively agrees with the experimental observations of the LQT2 rabbit model (20,21).In addition to the effect of I Kr suppression, these results also demonstrate that the effect of the RFFL-produced I to modifications on APD90 depends on the background conditions and other channels.

Discussion
Introgression of a congenic strain of Dahl salt-sensitive rats with genomic segments from the normotensive Lewis rat allowed mapping of a quantitative trait locus for hypertension, short QT interval, and cardiac hypertrophy to a 42.5-kb region that contains only the RFFL gene (14).As the transient outward potassium current I to is the major repolarizing current in rat ventricular myocytes (22,23), we set out to look for any functional interaction between RFFL and I to using cultured adult rabbit cardiomyocytes.However, dedifferentiation of cultured rabbit cardiomyocytes (24-26) may confound any findings.To minimize dedifferentiation in this study, we cultured adult rabbit cardiomyocytes no more than 48 h and used only healthy looking, rod-shaped, striated cardiomyocytes for patch-clamp experiments.We also added 0.5 μM cytochalasin D to the medium to preserve the function (e.g., APs and calcium transients) and morphology (rod-shape and T-tubular structure) of cardiomyocytes as published elsewhere (27).In the past, we used 0.5 μM cytochalasin D to prevent dedifferentiation in 3-week-old adult rabbit cardiomyocytes cultured for 48 h (28).Although some functional parameters, such as resting potential and I Na , did not change after a 48-h culture of cardiomyocytes, others did (e.g., APD prolongation, increase in I Ca , and decrease in I K1 or I Kr ).Importantly, I to did not change in these cardiomyocytes cultured for 48 h.Despite the efforts to minimize cardiomyocyte remodeling, we acknowledge the possibility that exogenous RFFL, a RING finger Ub ligase with a plethora of target molecules, may impact the dedifferentiation process in cultured cells, which in turn may explain the reciprocal changes of I to,f and I to,s .Yet, we do believe that there is a direct functional interaction between RFFL and I to in mammalian cardiomyocytes as implied by the aforementioned study of a congenic rat strain (14).Importantly, we were able to reproduce the yin-yang effect of RFFL on I to,f and I to,s in heterologous HEK293A cells.
Our I to recovery time constants in GFP-expressing cultured ARbCMs (Fig. 1D) are similar to previously published parameters obtained in acutely isolated ARbCMs, τ f = 22 ± 1 ms and τ s = 2.78 ± 0.06 s (29).However, the relative contributions of the fast and slow components to the total I to , 16 ± 2% and 83 ± 6%, respectively ( 29), differ from our data.Here, we show that the relative contributions are 35 ± 3% and 65 ± 3%, respectively (Fig. 1D).Therefore, in the experiments with cultured primary cardiac myocytes, we alternated patch-clamp recordings from RFFL-and GFP-expressing cells to equalize the culturing time between the two types of cells before the measurements.
It is known that the manifestation of I to,f and I to,s is not homogeneous in the heart and it exhibits expression gradients of these two I to currents across the free wall, as well as changes in expression of auxiliary subunits and phosphorylation by CaMKII (30,31).Our study shows increased complexity of I to regulation because of a yin and yang effect of RFFL on the fast and slow I to components.While RFFL has effects on both fast and slow I to components (Fig. 1D), the maximal amplitude of total I to remains practically unchanged.Since the major action of RFFL on the repolarization during the AP is the suppression of I Kr (12), the effect of RFFL reported in this study may be considered as a combination of LQT2 syndrome (lack of I Kr ) and a modulation of I to expression.Incidentally, in electrophysiological terms, this situation is similar to the effects of I to in the LQT1 rabbit heart, where the I Ks current is suppressed, while I to kinetics variation leads to polymorphic ventricular tachycardia under isoproterenol (32).When we introduced changes in I Kr and I to corresponding to RFFL overexpression using our spatially detailed model for rabbit ventricular myocytes (18), the APD90 was prolonged by 20.6 ms (i.e., 10%) due to the I Kr suppression alone.On top of that, the effects of I to led to a 3.9 ms (i.e., 2%) decrease in APD90 (Fig. 6).This result demonstrates that the effects of RFFL on APD90 via the changes in I Kr and I to have opposite trends.The dominant contribution of I to to the small APD90 shortening is the increase from 65% to 79% of I to,s relative to total I to in ARbCMs (Figs. 1D, and 6, A and E), which correlates with an increase in Kv1.4 expression in HEK cells (Figs. 4 and 5).In contrast, the RFFL evoked reduction of I to,f in ARbCMs (Figs. 1D and 6F) corroborated the decrease in Kv4.3 expression in HEK cells (Fig. 3) and resulted in a small reduction of the AP notch and a thicker AP overshoot in the computer model (Fig. 6A).Thus, RFFL overexpression leads to the reduction of I Kr and increase in I to,s that have opposing effects on APD90, while the reduction of I to,f slightly broadens the AP peak.The effect of RFFL on the AP will likely be more complicated under isoproterenol (32), which requires additional studies.
The nature of the opposing effects of RFFL on the fast and slow components of I to and on expression of Kv4.3 and Kv1.4 is yet to be determined.The complexity of ubiquitination outcomes is the result of different types of ubiquitinations, such as monoubiquitination, multimonoubiquitination, homotypic ubiquitination, heterotypic ubiquitination, modified ubiquitination, etc., which lead to different consequences (33).In addition, there is the Ub-proteasome pathwaymediated Kv4.3 degradation, which apparently does not involve RFFL (34).There are also other enzymes that affect trafficking of the channel proteins underlying I to , which potentially might also be affected by RFFL (23,35).In contrast, another ring finger protein E3 ligase, RNF207, increases expression of hERG (36), while on the other hand it exacerbates pathological cardiac hypertrophy via posttranslational modification of TAB1 (37).
A further complication of the RFFL effect on I to is that native Kv4.x channels need to assemble with KChIPs to fully reproduce the I to current (38,39).KChIP2 proteins are expressed in human hearts, and KChIP2 overexpression in Kv4.3-transfected HEK293A cells increased I to and slowed the I to decay rate (40), which increases the repolarization reserve.KChIP2 also affects the trafficking of Kv4.2 by masking an Nterminal Kv4.2 hydrophobic domain (41).Since we have shown that RFFL overexpression increased the total expression of KChIP2 (Fig. 3C), and since KChIP2 overexpression increases expression of Kv4.3 (Fig. 3C) and I to in the control (Fig. 2B), then RFFL has two opposing effects on I to,f .On the one hand, RFFL overexpression increases I to,f by increasing KChIP2 (Fig. 3C) and consequently Kv4.3 (Fig. 3A, compare two "R" lanes) expressions, but on the other hand, RFFL reduces total and surface Kv4.3 expression (Fig. 3A, compare "C" and "R" lanes).Given that I to,f in rabbits was suppressed by RFFL overexpression, we conclude that the inhibiting effect of RFFL overexpression on Kv4.3 and I to,f directly dominates over the indirect upregulation of Kv4.3 due to upregulation of KChIP2 in rabbit myocytes.
To our knowledge, this is the first report directly linking ubiquitination to I to currents encoded by Kv1.4 and Kv4.3.Surface biotinylation experiments and patch clamp studies (Figs.2-5) suggest that RFFL expression regulates surface expression of both channels in a reciprocal manner.Importantly, these effects of RFFL are dependent on a functional RING domain.We used Rapid UBIquitination sequence detection, a sequence-based predictor for ubiquitinated lysines (42), to look for potential ubiquitination sites in both channels.
Cardiac RFFL regulates I to currents and matching proteins Not surprisingly, ubiquitination sites were predicted for human Kv1.4 (lysine 268) and Kv4.3 (lysine 71).Both sites are conserved in rabbit.Next, we used UbiBrowser (43), a bioinformatics platform that predicts Ub ligase substrate interactions.Although UbiBrowser did not predict any interactions between RFFL and Kv1.4 or Kv4.3, it identified potential interactions between RFFL and three (co) chaperones (three out of a total of 117 predicted RFFL substrates) involved in protein quality control: HSP5A (BiP), a member of the Hsp70 family involved in the folding of proteins in the ER, BAG3, and BAG6, which are both cochaperones of Hsp70 and are also found localized to the ER (44)(45)(46).The implication of a functional interaction between RFFL and (co) chaperones is particularly interesting given our published results showing that RING finger protein RNF207, another Ub ligase involved in cardiac excitation, regulates hERG trafficking in a heat shock-dependent manner, likely via interaction with cytosolic chaperones such as Hsp70 and Hsp40 on the ER, to assist synthesis, folding, and/or ER export of hERG (36).One can envisage that RFFL ubiquitinates various (co) chaperones in a similar way as it monoubiquitinates effectors of Rab11, a crucial regulator of intracellular membrane trafficking such as recycling of endocytosed membrane proteins (6,47).Indeed, degradation-independent monoubiquitination of Hsp70 and Hsc70 was previously reported for the RING finger Ub ligase parkin (48), though functional consequences were not seen.Monoubiquitination of (co) chaperones may affect their catalytic activities with respect to folding, assembly, forward trafficking, and degradation of client proteins.In the case of Kv1.4,monoubiquitination of (co) chaperones by RFFL may result in increased surface expression of the channel.Alternatively, RFFL may directly ubiquitinate Kv1.4 to increase its stability.However, unlike for hERG (12), we were not able detect any RFFL-dependent ubiquitination of Kv1.4 using immunoprecipitations.It is also plausible that the effect of RFFL may rather be mediated through an indirect effect, for example, by inhibiting a negative regulator of Kv1.4.We previously showed that RFFL-mediated polyubiquitination of hERG channel resulted in its proteasomal degradation, which relied on a functional ER-associated degradation pathway in 293A cells (12).Likely, the same molecular mechanism underlies the degradation of Kv4.3 by RFFL in the absence of KChIP2.As KChIP proteins are known to stabilize Kv4 α subunit channel complexes and promote channel assembly and/or trafficking to the membrane (49-51), we speculate that KChIP2 substantially protects Kv4.3 from RFFL-mediated degradation.The observed positive effect of RFFL harboring a deletion of the RING domain on Kv4.3 surface expression could be due to the formation of inactive complexes that can no longer promote Kv4.3 ubiquitination/degradation.A possible candidate inhibited by the RING deletion of RFFL could well be CHIP, a U-box-dependent Ub ligase, and cochaperone important for the balance between protein degradation and folding (52).Functional interactions between RFFL and CHIP may exist on the ER as both molecules share overlapping functions in the peripheral protein quality control of misfolded CFTR (9,53).For example, CHIP and parkin are known to functionally interact during the degradation of unfolded PaeI receptor in dopaminergic neurons (54).Further research is warranted to delineate the molecular mechanisms underlying the reciprocal regulation of both I to channels by (in) direct ubiquitination through RFFL.Of note, recent experiments showed that brefeldin-A inhibited transport of Kv1.4 from the ER to the Golgi complex (data not shown).
In summary, our data suggest that channel stability (Kv4.3) and forward trafficking to the surface (Kv1.4) are regulated by RFFL-dependent ubiquitination events on the ER.These events result in reciprocal changes in surface expression of both channels underlying I to,f and I to,s .

Preparation of rabbit cardiomyocytes
All animal experiments and procedures were approved by the Rhode Island Hospital Institutional Animal Care and Use Committee (reference numbers: 0188-14 and 5013-17).Septal ARbCMs were isolated from the hearts of 6-to 24-month-old New Zealand White rabbits (both sexes).The filtered cells were maintained in 45 mM KCl, 65 mM potassium glutamate, 3 mM MgSO 4 , 15 mM KH 2 PO 4 , 16 mM taurine, 10 mM Hepes, 0.5 mM EGTA, and 10 mM glucose (pH 7.3) for 1 h.Cells were allowed to sediment for 30 min, and after removal of the supernatant, resuspended in medium 199 (Thermo Fisher Scientific) supplemented with 5% fetal bovine serum (MilliporeSigma), antibiotics, and 0.5 μM cytochalasin D (MilliporeSigma).After plating cells on laminin-coated cover glasses, adenovirus (50 multiplicity of infection) encoding GFP or RFFL was added to the cells.Cells were maintained at 37 C with 5% CO 2 and 48 h later the cells were used for patch clamp and biochemistry.

Rabbit ventricular myocyte model
To study the effect of RFFL in myocytes, we used a physiologically detailed rabbit ventricular myocyte model (18) with 2.5 Hz stimulation frequency.The 100th AP after the beginning of the simulation is shown in Figure 6.
Statistical analysis and curve fitting were performed with GraphPad Prism 8 (https://www.graphpad.com/)and Origin-Pro 2019 (https://www.originlab.com/2019).If there were outliers in a dataset, they were removed for further analysis.Statistical comparison of multipoint curves was performed with a mixed effects model for repeated measures data test.The fitting of the curves with two exponents were performed in OriginPro with the Levenberg Marquardt iteration algorithm.Statistical comparison of two groups was done with Student's t tests (two-tailed).Data were presented as mean ± SEM for electrophysiological recordings and mean ± SD for immunoblots.A difference was considered significant at p < 0.05.

Replicates
Throughout the study, we used biological replicates, that is, different numbers of animals or different frozen HEK cell stocks (as indicated by "N"), and technical replicates (as indicated by "n").

Figure 1 .
Figure 1.RFFL overexpression reduces I to,f but increases I to,s in ARbCMs.A, overlapping representative traces of I to in ARbCMs adenovirally expressing GFP obtained with double-pulse protocol from a holding potential of −70 mV to 0.5 s test pulses of +50 mV.The second pulse was applied after a varied recovery interval (Δt) from 10 ms to 15 s.The voltage protocol in the insert is an example of the interpulse time of 5 s with short green arrows corresponding to the beginning of the first and second I to stimulations and the long green arrows corresponding to the ends of the stimulating pulses.The red dashed curve reveals a trend of I to recovery for this experiment.B, I to traces shown in (A) are zoomed in to show I to peaks at shorter recovery intervals.C, overlapping representative traces of I to in ARbCMs adenovirally expressing RFFL obtained with the same protocol as in (A).D, the normalized I to recovery data (mean ± SEM) of GFP-expressing (N = 5, n = 12) and RFFL-expressing (N = 5, n = 10) ARbCMs are significantly different when compared by a mixed effects model for repeated measures data test (p < 0.05).The differences between GFP and RFFL curves are also significant at recovery time points of 400 and 600 ms (* −p < 0.05, shown by red double headed arrows).GFP and RFFL normalized I to recovery curves were fitted with the equation: I to = I to,f (1 -exp(-Δt/τ f )) + I to,s (1 -exp(-Δt/τ s )).The fitting curves are depicted by thick solid lines for GFP-(black) and RFFL-expressing cells (red).The I to,f components of the fittings presented by thin solid lines and I to,s components are presented by dashed lines.The I to,f maximal amplitudes in RFFL-versus GFP-expressing cells are significantly different, as well as I to,s maximal amplitudes (** −p < 0.01).ARbCM, adult rabbit ventricular cardiomyocyte; RFFL, RING finger and FYVE-like domain-containing E3 ubiquitin protein ligase.

BFigure 2 .
Figure 2. RFFL downregulates I to,f in HEK cells stably expressing Kv4.3 in a RING domain-dependent manner independently of KChIP2.A, representative traces of I to,f in Kv4.3-expressing HEK cells transiently expressing GFP (control), RFFL and GFP, or RFFL-ΔRING and GFP.Top inset shows voltage protocol.Bottom inset shows cartoon of the constructs.B, cumulative I-V curves of I to,f in Kv4.3-expressing HEK cells transiently expressing GFP (control, green lines and symbols), RFFL and GFP (red), RFFL-ΔRING and GFP (blue), in the presence of KChIP2 and DsRed (solid lines and symbols) and the absence of KChIP2 but with DsRed (dashed lines and open symbols).The first number in parentheses in the legend corresponds to the number of experiments (N) and the second number to the total number of the cells for each condition (n).A mixed effects model for repeated measures data test indicates statistically significant differences between the I-V curves, where * corresponds to p < 0.05, ** −p < 0.01, *** −p < 0.001.ARbCM, adult rabbit ventricular cardiomyocyte; HEK, human embryonic kidney; KChIP2, Kv channel-interacting protein 2; RFFL, RING finger and FYVE-like domain-containing E3 ubiquitin protein ligase.

Figure 4 .
Figure 4. RFFL upregulates I to,s in HEK cells stably expressing Kv1.4 in a RING domain-dependent manner.A, representative traces of I to,s in Kv1.4expressing HEK cells transiently expressing GFP (control), RFFL and GFP, or RFFL-ΔRING and GFP.B, cumulative I-V curves of I to,s in HEK cells expressing GFP (green), RFFL and GFP (red), or RFFL-ΔRING and GFP (blue).A mixed effects model for repeated measures data test indicates statistically significant differences between the I-V curves, where * corresponds to p < 0.05.HEK, human embryonic kidney; RFFL, RING finger and FYVE-like domain-containing E3 ubiquitin protein ligase.

Figure 6 .
Figure 6.Computer modeling shows that the change of I to,f and I to,s associated with RFFL overexpression affects action potential duration for both normal and fully suppressed I Kr .A, simulated AP traces (400 ms cycle length) in a control cell (thin red line) and a cell under different I Kr and/or I to conditions associated with RFFL overexpression, which suppresses I Kr , downregulates g to,f , and upregulates g to,s .The thick red line represents the AP with I Kr suppressed only.The dash-dotted thin blue line shows the effect of g to,f downregulation and g to,s upregulation only.The dash-dotted thick blue line exhibits the combination effect on AP by suppressing I Kr , decreasing g to,f , and increasing g to,s , that is, the consequence of overexpressing RFFL.During an AP shown in (A), the time dependences of [Ca 2+ ] i , I Ca,L , I Kr , I to,s , and I to,f are shown in (B-F), respectively.AP, action potential; RFFL, RING finger and FYVE-like domaincontaining E3 ubiquitin protein ligase.