Kir2.1 Interactome Mapping Uncovers PKP4 as a Modulator of the Kir2.1-Regulated Inward Rectifier Potassium Currents

A comprehensive map of the Kir2.1 interactome was generated using the proximity-labeling approach BioID. The map encompasses 218 interactions, the vast majority of which are novel, and explores the variations in the interactome profiles of Kir2.1WT versus Kir2.1Δ314-315, a trafficking deficient ATS1 mutant, thus uncovering molecular mechanisms whose malfunctions may underlie ATS1 disease. PKP4, one of the BioID interactors, is validated as a modulator of Kir2.1-controlled inward rectifier potassium currents.


Kir2.1 Interactome Mapping Uncovers PKP4 as a Modulator of the Kir2.1-Regulated Inward Rectifier Potassium Currents
Sung-Soo Park 1, ‡ , Daniela Ponce-Balbuena 2, ‡ , Rork Kuick 3, † , Guadalupe Guerrero-Serna 2 , Justin Yoon 1 , Dattatreya Mellacheruvu 1 , Kevin P. Conlon 1 , Venkatesha Basrur 1 , Alexey I. Nesvizhskii 1,4 , José Jalife 2,5,6, *, and Jean-François Rual 1, * Kir2.1, a strong inward rectifier potassium channel encoded by the KCNJ2 gene, is a key regulator of the resting membrane potential of the cardiomyocyte and plays an important role in controlling ventricular excitation and action potential duration in the human heart. Mutations in KCNJ2 result in inheritable cardiac diseases in humans, e.g. the type-1 Andersen-Tawil syndrome (ATS1). Understanding the molecular mechanisms that govern the regulation of inward rectifier potassium currents by Kir2.1 in both normal and disease contexts should help uncover novel targets for therapeutic intervention in ATS1 and other Kir2.1-associated channelopathies. The information available to date on protein-protein interactions involving Kir2.1 channels remains limited. Additional efforts are necessary to provide a comprehensive map of the Kir2.1 interactome. Here we describe the generation of a comprehensive map of the Kir2.1 interactome using the proximity-labeling approach BioID. Most of the 218 high-confidence Kir2.1 channel interactions we identified are novel and encompass various molecular mechanisms of Kir2.1 function, ranging from intracellular trafficking to cross-talk with the insulin-like growth factor receptor signaling pathway, as well as lysosomal degradation. Our map also explores the variations in the interactome profiles of Kir2.1 WT versus Kir2.1 D314-315 , a trafficking deficient ATS1 mutant, thus uncovering molecular mechanisms whose malfunctions may underlie ATS1 disease. Finally, using patch-clamp analysis, we validate the functional relevance of PKP4, one of our top BioID interactors, to the modulation of Kir2.1-controlled inward rectifier potassium currents. Our results validate the power of our BioID approach in identifying functionally relevant Kir2.1 interactors and underline the value of our Kir2.1 interactome as a repository for numerous novel biological hypotheses on Kir2.1 and Kir2.1-associated diseases.
The strong inward rectifier potassium channel Kir2.1 is a key regulator of the resting membrane potential of the cardiomyocyte and plays an important role in controlling ventricular excitation and action potential duration in the human heart (1)(2)(3). Both loss and gain of function mutations in KCNJ2, the gene encoding Kir2.1, result in inheritable cardiac ion channel diseases. For example, several KCNJ2 loss of function mutations are associated with the inheritable type-1 Andersen-Tawil syndrome (ATS1), also known as long QT syndrome type 7, which predisposes patients to cardiac arrhythmias and sudden death (4,5). On the other hand, KCNJ2 gain-of function mutations give rise to the type-3 variant of the short QT syndrome, which also results in increased risk of sudden cardiac death (6). Understanding the molecular mechanisms that govern the regulation of inward rectifier potassium currents by Kir2.1 in both normal and disease contexts should help uncover novel targets for therapeutic intervention in ATS1 and other Kir2.1-associated channelopathies.
Over the last 20 years, analyses of the Kir2.1 channelosome using genetic, pharmacological, and molecular approaches have greatly contributed to our understanding of its function and its role in cardiac diseases (3,7,8). Expression, trafficking, localization, and function of Kir2.1 are regulated by its interactions with multiple proteins. To date, 24 putative Kir2.1 interactors have been identified, out of which 16 are high-confidence Kir2.1 interactors (supplemental Table S1). For example, Leonoudakis et al. showed that the PDZ-binding motif of Kir2.1 interacts with the SAP97, CASK and LIN7C proteins, members of the membrane-associated guanylate kinase (MAGUK) family, which work both as scaffolding proteins for large trol (TM-CTRL), the amino acid sequence IIFRTLFGSLVFAIFLILMIN of the Saccharomyces cerevisiae P25353 protein transmembrane domain, which has no homology to any human proteins, was selected and a humanized codon sequence was synthesized (supplemental Table S2). The protein-encoding Open Reading Frames (ORFs) of Kir2.1 WT (NM_000891.2), Kir2.1 D314-315 , and TM-CTRL were first cloned by Gateway recombination cloning from the cDNA plasmids into the Gateway Donor vector pDONR223 to generate Entry clones, as previously described (21) (primer sequences are shown in supplemental Table S2). The Gateway clone of PKP4 (EL733946) was obtained from the Center for Cancer Systems Biology (CCSB, Harvard Medical School, Boston) human ORFeome collection (21). Protein-encoding ORF/s were then subcloned from the Entry clones into the Gateway Destination vectors pDEST-pcDNA5-FRT-BirA*-FLAG-N-term and pDEST-pcDNA3-HA (18,22). A target sequence of PKP4 for CRISPR/Cas9 (supplemental Table S2) was selected using CHOPCHOP web tool (23) and cloned into pX459 (Addgene, Watertown, MA, 62988), as previously described (24). The following primary and secondary antibodies were used in the Western blot analyses: FLAG (Sigma, Kawasaki, Kanagawa, Japan, A8592) and goat a-rabbit IgG (Cell Signaling, Danvers, MA, 7074).
Cell Facility of the Department of Pathology at the University of Michigan for protein identification. Briefly, upon reduction (10 mM DTT) and alkylation (65 mM 2-chloroacetamide) of the cysteines, proteins were digested overnight with 500 ng of sequencing grade, modified trypsin (Promega, Madison, WI, V5111). Resulting peptides were resolved on a nano-capillary reverse phase column (Acclaim PepMap C18, 2 mM, 50 cm, ThermoFisher Scientific, ES803) using 0.1% formic Kir2.1 Interactome Mapping Using BioID acid/acetonitrile (ACN) gradient at 300 nL/minute (2-25% ACN in 108 min, 25-40% in 20 min, followed by a 90% ACN wash for 10 min and a further 30 min' re-equilibration with 2% ACN) and directly introduced into Orbitrap Fusion Tribrid Mass Spectrometer (Thermo-Fisher Scientific). MS1 scans were acquired at 120k resolution. Datadependent high-energy C-trap dissociation MS/MS spectra were acquired with top speed option (3 s) following each MS1 scan (relative CE ;32%). Proteins were identified by searching the data against Homosapiens database containing both canonical and isoform protein entries (SwissProt v2016-11-30; total number of entries: 42054) using Proteome Discoverer (v2.1, ThermoFisher Scientific, June 2016) with enzyme specificity for trypsin. Search parameters included MS1 mass tolerance of 10 ppm and fragment tolerance of 0.1 Da; two missed cleavages were allowed; carbamidomethylation of cysteine was considered fixed modification and oxidation of methionine, deamidation of asparagine and glutamine were considered as variable modifications. False discovery rate (FDR) was determined using target-decoy strategy and proteins/peptides with a FDR of 1% were retained. A stringent set of criteria were used for the selection of the high-confidence Kir2.1 BioID hits: (1) average SpC (spectral counts) in either the Kir2.1 WT or the Kir2.1 D314-315 BioID experiment . 10; (2) SAINT probability (27,28) . 0.9; and (3) primary fold change FC-A score (28) . 7. The data obtained with the TM-CTRL bait were used as negative controls for the calculations. One of the high-confidence Kir2.1 BioID interactors has been identified with a single peptide, i.e. NACA2. The spectra annotation for NACA2 is provided in the supplemental Fig. S1.  Table S3, columns X-AE). Second, after adding one to each SpC and Log 2 -transformation (columns AF-AK), we calculated the average Log 2 -transformed, normalized SpC for both the Kir2.1 WT and Kir2.1 D314-315 BioID experiments (columns AL and AM). The ratio R was then calculated by performing a power of 2 transformation (column AN). The statistical significances of the difference between the Kir2.1 WT and Kir2.1 D314-315 BioID experiments were computed using a two-sample t test on the Log 2 -transformed data (column AO). Kir2.1 BioID hits with both R . 2 and p , 0.05 were classified as Kir2.1 WT -preferred interactors whereas proteins with both R , 0.5 and p , 0.05, i.e. a 2-fold decrease, were classified as Kir2.1 D314-315 -mutant preferred interactors (column AP). A permutation analysis was performed to estimate the false discovery rate (FDR , 0.007).
Patch-Clamping Experiments-HEK293 cells were grown in 60mm dishes and, upon reaching ;60-70% confluence, were transi-ently transfected using X-tremeGENE HP DNA transfection reagent (Sigma, 6366244001) following the supplier's directions. The DNA plasmids used in the patch-clamp experiments were: CRISPR/Cas9 construct pX459-gPKP4 and negative control pX459 and protein expression vectors pcDNA3-HA-PKP4 and "empty vector" negative control pcDNA3-HA. A GFP-expression vector was also cotransfected to visualize the transfected cells during the patchclamp experiments. Patch-clamp experiments were performed 48 h after transfection, as previously described (12,13,15). Inward rectifier potassium currents (I Kir2.1 ) were recorded at room temperature (21-23°C) using the whole-cell patch-clamp technique and filtered at half the sampling frequency (12,13,15). Series resistance was compensated manually and 80% compensation was achieved. Under our experimental conditions, no significant voltage errors (,5 mV) because of series resistance were expected with the micropipettes used.
Experimental Design and Statistical Rationale-The total number of samples analyzed and the statistical tests used in the various experiments are indicated throughout the text, e.g. triplicate BioID experiments were performed, the number of transfected cell batches (N) and number of cells (n) used in the patch-clamping experiments are shown, or the statistical significances of the difference between the Kir2.1 WT and Kir2.1 D314-315 BioID experiments were computed using a two-sample t test on the Log 2transformed data.

RESULTS AND DISCUSSION
Kir2.1 Interactome Mapping Using BioID-Using the proximity-labeling technology BioID (16-18), we generated a map of the Kir2.1 interactome (overall procedure is described in Fig. 1). Both Kir2.1 WT and Kir2.1 D314-315 , an ATS-associated mutant protein for which Golgi export is impaired (4,19,20), were used as BirA*-tagged BioID baits. We performed patch-clamping analyses of both the BirA*-tagged Kir2.1 WT and Kir2.1 D314-315 bait proteins (supplemental Fig. S2). As expected, HEK293 cells expressing BirA*-tagged Kir2.1 WT , but not mutant BirA*-tagged Kir2.1 D314-315 cells, exhibited the characteristic ability to strongly rectify, that is to pass K 1 current in the inward direction much more readily than outward. Moreover, BirA*-tagged Kir2.1 WT channels maintain their sensitivity to Ba 21 blocking (supplemental Fig. S2). As a negative control, we used the TM-CTRL construct, a yeast transmembrane domain fused to the BioID BirA* tag. We investigated the subcellular localization of both Kir2.1 WT Table S3). Interestingly, some of the newly identified BioID interactors include products of genes previously associated with physiological or pathophysiological function of the myocardium, e.g. UTRN (29,30) and PKP4 (31, 32), as described in detail below. Curation of the literature and of the functional data reported in PubMed for each of these genes led to the identification of a dozen of known protein complexes or functionally related biological modules in our list of 218 high-confidence Kir2.1 BioID hits, e.g. the cadherin adhesome and the COPII complex.
To assess the sensitivity of our BioID proteomic screen, we measured the proportion of the previously described Kir2.1 interactors that are also present in our list of 218 highconfidence Kir2.1 BioID interactors. A review of several protein interaction repositories [BioGRID (33), IntAct (34), HPRD (35), HuRI (36,37)] identified 24 putative Kir2.1 protein interactions. After literature curation, we selected 16 high-confi-dence Kir2.1 interactions that had been observed in at least two experiments and/or validated for their functional relevance (supplemental Table S1). We note, though, that some of these interactions have been studied using the Mus musculus or Rattus norvegicus proteins, but not the human proteins (columns I-J in supplemental Table S1). Out of all the 24 previously reported putative Kir2.1 interactors, 10 (;42%) were identified in our BioID screen. Similarly, out of the 16 high-confidence Kir2.1 interactions, 7 (;44%) were identified in our BioID screen, i.e. SAP97, CASK, CAV1, LIN7C, AKAP5, Kir2.6 and Kir2.3 (supplemental Table S1 and S3). This analysis suggests a high sensitivity of .40% for our BioID assay. By comparison, the sensitivity of high-throughput protein interaction mapping assays used on a proteomescale ranges from ;20 to ;35% (38). The ATS1-associated mutation Kir2.1 D314-315 blocks Kir2.1 Golgi export (4,19,20). Hence, Kir2.1 WT -preferred interactors may encompass both: i) protein interactors that are directly involved in the normal trafficking of Kir2.1, and ii) protein interactors that, as an indirect consequence of the Golgitrapping, do not interact with Kir2.1 D314-315 simply because it is not present at the plasma membrane. Interestingly, out of the 75 Kir2.1 WT -preferred interactors, 13 of them have been associated (mutation and/or GWAS hit) with one or several heart-related traits or diseases, e.g. atrial fibrillation (AF) or systolic/diastolic blood pressure (supplemental Table S3, column AR). Most ATS mutants, including Kir2.1 D314-315 , exert a dominant-negative effect on Kir2.1 WT channels, as oppose to an haploinsufficiency effect (20). Accordingly, protein interactions involving Kir2.1 D314-315 -preferred interactors may also encompass molecular mechanisms that are not properly regulated in the presence of Kir2.1 D314-315 . For example, the Kir2.1 D314-315 mutation may result in the gain of an interaction (or increased binding affinity) between Kir2.1 D314-315 and a protein present in the Golgi that perturbs normal trafficking and results in the trapping of Kir2.1 D314-315 . On one hand, the increased ability for Kir2.1 D314-315 to interact with some of these Kir2.1 D314-315 -mutant preferred interactors may be an indirect consequence of the D314-315 mutation and these Kir2.1 D314-315 -mutant preferred interactors are thus likely to have low relevance to the disease. On the other hand, the increased ability for Kir2.1 D314-315 to interact with some of these Kir2.1 D314-315 -mutant preferred interactors may be a direct consequence of the D314-315 mutation and some of these Kir2.1 D314-315 -mutant preferred interactors may thus be possible direct contributing factors to the disease. Interestingly, out of the 66 Kir2.1 D314-315 -preferred interactors, 10 of them have been associated (mutation and/or GWAS hit) with one or several heart-related traits or diseases, e.g. atrial fibrillation (AF) or systolic/diastolic blood pressure (supplemental Table S3, column AR). Below, we discuss in details the potential implications associated with some of these Kir2.1 WT -preferred interactors, e.g. UTRN, and Kir2.1 D314-315preferred interactors, e.g. NAC and IGF1R.
The analysis of the variations in the interactome profiles of Kir2.1 WT and Kir2.1 D314-315 allows us to identify protein complexes or groups of functionally related proteins that, for the most part, tend to interact primarily with either Kir2.1 WT or Kir2.1 D314-315 (Fig. 3). For example, most of the proteins in the MAGUK complex interact preferentially with Kir2.1 WT (colored in green in Fig. 3). In contrast, Kir2.1 interactors in the Nascent polypeptide-associated complex (NAC) bind preferentially to Kir2.1 D314-315 (colored in red in Fig. 3). Interestingly, the "Channels-Transporters" group encompasses proteins that interact preferentially with either cytoplasmic membranebound Kir2.1 WT , Golgi-trapped Kir2.1 D314-315 or that are Kir2.1 WT/D314-315 -neutral interactors (colored in gray in Fig. 3). This observation indicates that Kir2.1 not only interacts with other channels-transporters at the cytoplasmic membrane but that the co-trafficking of these proteins in the endoplasmic reticulum (ER) and Golgi apparatus is an important aspect of their function. Thus, the previously described dynamic reciprocity for the regulation of Na v 1.5 sodium and Kir2.1 potassium channel expression, which controls cardiac excitability and arrhythmia (12), likely extends to other channelosome protein complexes beyond Na v 1.5.
Enrichment Analyses-To further extend our analysis of the protein complexes and the groups of genes that interact with Kir2.1, we performed a gene set enrichment analysis using DAVID (39) (supplemental Table S4). We observed that  (4,19,20), our data show that ATS1 pathogenesis associated with the Kir2.1 D314-315 mutation underlies malfunctions in the mechanisms governing Kir2.1 trafficking from the Golgi to the cytoplasmic membrane.
Kir2.1 D314-315 -preferred interactors are also linked to ESCRT-0 (GO:0033565, FE = 289, p = 3 3 10 25 ) and lysosome (hsa04142, FE = 7, p = 0.02), suggesting that this ATS1 mutation leads to increased targeting of Kir2.1 to lysosomal degradation. These observations agree with the previous report by Kolb et al. that Kir2.1 degradation is primarily lysosomal dependent and requires ESCRT (40). However, using a pharmacological approach to inhibit either proteasomal or vacuolar proteases, Kolb et al. also observed that Kir2.1 D314-315 mutation targets the channel for proteasomal degradation rather than vacuolar-dependent degradation (40). Though we indeed note the presence of PSMD1, the 26S proteasome nonATPase regulatory subunit 1, among the Kir2.1 D314-315 -preferred interactors, our enrichment analysis does not provide support to the "proteasome hypothesis" but instead, strongly suggests that vacuolar-dependent degradation remains an important route responsible for the increased Kir2.  Table S5). Similarly, out of the top 34 gene sets enriched for Kir2.1 D314-315 -preferred interactors (p , 1 3 10 23 ), only two were also enriched in the Kir2.1 WT -preferred interactors (supplemental Table S5). This observation further underscores the fact that the Kir2.1 WT and Kir2.1 D314-315 proteins obey to dramatically different fates in the cell.

Kir2.1 BioID Interactome: A Repository for Novel Biological
Hypotheses-The Kir2.1 BioID interactome data set represents a repository for numerous, novel biological hypotheses for genes and molecular mechanisms implicated in Kir2.1associated cardiomyopathies. For example, out of the 218 high-confidence Kir2.1 interactors identified in our BioID screen, 37 of them have been associated (mutation and/or GWAS hit) with one or several heart-related traits or diseases, e.g. atrial fibrillation (AF) or systolic/diastolic blood pressure (supplemental Table S3 Table S3). Utrophin deficiency worsens cardiac contractile dysfunction present in dystrophin-deficient mice (29). Interestingly, a race-stratified genome-wide gene-environment interaction association study recently identified polymorphic variations in the UTRN gene locus as potential risk factors in peripheral arterial disease (30). UTRN is part of the Dystrophinassociated proteins (DAP) complex (41) and several other DAP proteins interact with and modulate both sodium and potassium channelosome protein complexes (8,11,42). This novel Kir2.1-UTRN interaction further underlines the role of DAP in regulating these complexes and malfunction of these molecular mechanisms may contribute to pathophysiological conditions of the myocardium.
Nascent-Polypeptide-Associated Complex (NAC)-Our BioID screen identified four NAC-associated proteins as Kir2.1 D314-315preferred interactors, i.e. NACA, NACA2, BTF3 and BTF3L4. For example, the normalized SpC counts for NACA are as follows: Kir2.1 WT : 24.4 and Kir2.1 D314-315 : 270.7 (supplemental Table S3). The heterodimeric NAC protein complex binds to newly synthesized polypeptide chains that lack a signal peptide motif as they emerge from the ribosome, blocking interaction with the signal recognition particle (SRP) and preventing mistranslocation to the endoplasmic reticulum (43). Potential roles for NAC complex thus range from protein folding chaperone to negative regulator of translocation into the endoplasmic reticulum (43). We note, though, that NACA has also been described as a transcription factor (44), e.g. in collaboration with the heart-specific HDACdependent repressor SMYD1 (45). The roles of NACA and NAC in the cell remain to be comprehensively characterized.
Several observations link NACA to heart biology and diseases. For example, GWAS studies have identified NACA gene variants as potential risk factors in atrial fibrillation (AF) (46,47), or as a regulator of myocardial mass (48). In both Drosophila melanogaster flies and Danio rerio zebrafish, the knock-down of NACA, i.e. fly NACa and zebrafish skNAC, results in severe cardiac muscle defects (48,49). In mouse, Naca deficiency results in embryonic lethality with cardiac developmental defects (50). Park et al. originally linked this phenotype to the role of NACA as a transcription factor and as a major binding partner for SMYD1 in the developing heart (50). The Kir2.1-NACA interaction suggests a more complex role for NACA in the myocardium. Indeed, Kir2.1 deficiency results not only in the loss of inward rectifier potassium currents but also in developmental defects and dysmorphic features, e.g. low-set ears, wide-set eyes, small mandible, clinodactyly and syndactyly (4,5,(51)(52)(53). The molecular mechanism underlying these dysmorphic features is poorly understood. The Kir2.1-NACA interaction provides a biophysical bridge between the respective roles of these two proteins during biological development in general and cardiomyogenesis in particular.  Table S3). GWAS studies have identified IGF1R gene variants as potential risk factors in atrial fibrillation (AF) (46,47), as well as a regulator of myocardial mass Kir2.1 Interactome Mapping Using BioID (48) and global electrical heterogeneity ECG (54). We hypothesize that the Kir2.1-IGF1R interaction is part of a molecular mechanism that underlies the functional associations between IGF1R and cardiac physiology.

Cross-talks between
Could the Kir2.1-IGF1R interaction also help better understand the potential role of Kir2.1 in diabetes? A functional link between Kir2.1 and the insulin-like growth factor receptor signaling pathway has been previously proposed. As early as 1998, Wischmeyer et al. observed that treatment of Xenopus oocytes with insulin was found to robustly suppress I Kir2.1 (55). Electrical activity plays a central role in glucose-stimulated insulin secretion from human pancreatic b-cells. We note that Kir2.1 is expressed in human pancreatic islets (56) and that functional Kir2.1 currents are present in human b-cells (57). Riz et al. predicted that blocking Kir2.1 channels increases the rate of insulin secretion and that hyperactive Kir2.1 channels may lead to reduced insulin secretion (57). Similarly, in a recent mathematical model, Kir2.1 was predicted to be a critical actor in the regulation of oscillations in cellular activity that underlie insulin pulsa-tility in pancreatic islets cells lacking K(ATP) channels (58). On the same note, several publications link abnormal QT interval regulation and diabetes (59)(60)(61). Moreover, patients with long QT syndrome associated to potassium channel deficiency, e.g. KCNQ1 mutants, suffer from over-secretion of insulin, hyperinsulinemia, and symptomatic hypoglycemia (62). Our BioID data suggest that these functional associations between Kir2.1 and the insulin pathway need not be indirect but may also be mediated by biophysical interactions between Kir2.1 channels and insulin pathway proteins.
Desmosome-Several desmosome-associated proteins interact with both Kir2.1 WT and Kir2.1 D314-315 , e.g. JUP, DSG2, DSC2 and PKP4 (supplemental Table S3). Desmosomes act as mechanical cell-cell adhesion junctions maintaining the structural integrity of the tissue. Mutations in the JUP, DSG2 and DSC2 genes have been associated with Arrhythmogenic Right Ventricular Cardiomyopathy (ARVC) (63). Actually, with ;half of ARVC patients carrying a mutation in one of the genes encoding desmosomal proteins expressed in the heart, ARVC can be considered a disease of the desmosome (63). Mutations in the JUP genes have also been associated with Naxos disease, a diffuse nonepidermolytic palmoplantar keratoderma with wooly hair and cardiomyopathy (64). We also note that JUP mRNA is highly differentially expressed in AF (65).   6). PKP4 is a multifunctional armadillo protein coordinating cell adhesion with cytoskeletal organization that also has a role in vesicle transport processes (66). Though less frequent than the ones observed for the JUP, DSG2 and DSC2 genes, mutations in the PKP4 gene have been observed in ARVC patients (31,32). Similarly to many other Kir2.1 interactors (8), PKP4 carries a PDZ binding site at its carboxyl terminus. We hypothesize that the Kir2.1-PKP4 interaction contributes to the regulation of Kir2.1.
As a first step toward validating the role of PKP4 as a regulator of Kir2.1, we investigated the subcellular localization of both proteins by immunofluorescence (IF) staining in freshly isolated adult rat ventricular myocytes. We observed that Kir2.1 and PKP4 co-localize at intercalated disks and at zdisks near the cardiac sarcomeres (Pearson correlation coefficient = 0.85) (Fig. 4).
Finally, to assess the functional relevance of PKP4 to the modulation of I Kir2.1 , we performed patch-clamp in the voltage clamp configuration (12,13,15) in a stable HEK293 cell line expressing Kir2.1 upon genetic perturbation of PKP4 (Fig. 5). Upon PKP4 overexpression, we observed a gain of I Kir2.1 density (Fig. 5A). Accordingly, upon CRISPR/Cas9mediated depletion of PKP4, we observed a loss of I Kir2.1 density (Fig. 5B). Therefore, in addition to interacting with Kir2.1 (Fig. 2) and to being mutated in an ARVC patients (31,32), PKP4 is a positive regulator of I Kir2.1 density. CONCLUSION Evidence accumulated over the last 20 years strongly indicates that potassium channels function depends on a complex and dynamic ensemble of molecular events occurring in various organelles in the cell that help in the assembly and delivery of the right channel subunits to the right place in the cell membrane at the right time. Understanding such molecular events should help uncover novel insights into how channelosome proteins and accessory proteins participate in the control of cardiac excitability and the mechanisms of arrhythmias. Using a BioID proteomic approach, we have generated the most comprehensive Kir2.1 interactome map known to date with 218 high-confidence Kir2.1 BioID hits, the vast majority of which are novel putative Kir2.1 interactors. Moreover, our approach allowed us to identify novel proteins that interact preferentially with either Kir2.1 WT or the ATS1 mutant Kir2.1 D314-315 , thus uncovering novel molecular mechanisms underlying Kir2.1 function/dysfunction in ATS1. Our Kir2.1 interactome map represents a repository for numerous novel biological hypotheses. Using patch-clamping, we verified the functional relevance of one of our hits, PKP4, to the modulation of I Kir2.1 , thus providing a molecular mechanism by which PKP4 may be involved in ARVC. This result validates the power of our BioID interactome approach in identifying functionally relevant Kir2.1 interactors and modulators of I Kir2.1 .

DATA AVAILABILITY
The MS raw data and Proteome Discoverer protein identification have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (26) with the data set identifier PXD011004.
Acknowledgments-We thank Drs. Anne Claude Gingras (The Lunenfeld-Tanenbaum Research Institute, University of Toronto, Canada) and Brian Raught (Princess Margaret Cancer Center, University of Toronto, Canada) for providing us with BioID reagents. We thank Dr. Robin Kunkel for her help with the cell illustration shown in Fig. 3. We thank Dr. M. Vidal and members of the CCSB (Harvard Medical School, Boston) for sharing the PKP4 ORFeome clone.
Funding and additional information-This work was supported by the National Institutes of Health (NIH) through the National Heart, Lung, and Blood Institute (NHLBI) grant R01HL122352 awarded to J.J., as well as the National Institute of General Medical Sciences (NIGMS) grant R01GM094231 and the National Cancer Institute (NCI) grant U24CA210967 awarded to A.I.N. R.K. is supported by the NCI support grant P30CA046592 awarded to the University of Michigan Rogel Cancer Center. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Author contributions-J.F.R. and J.J. conceived and directed the project. S.S.P., J.Y., D.M., K.P.C., V.B. and A.I.N. designed and performed the BioID experiments. R.K. performed the interactome profiling and enrichment analyses. D.P.B. and G.G.S. designed and performed the patch-clamping and the IF analyses. J.F.R. wrote the manuscript, with contributions from other coauthors.