ERG1, like other voltage-gated potassium channels, show a 4-fold tetrameric structure containing a voltage sensor (helices S1–S4) and a pore-forming region (S5–S6). ERG1 channels, members of the broader KCNH channel family, have a cytosolic Per-Arnt-Sim (PAS) domain and a cytosolic cyclic nucleic binding homology domain (CNBHD) that interact directly to regulate channel gating (Figure 1; Warmke and Ganetzky, 1994; Morais Cabral et al., 1998; Sanguinetti and Tristani-Firouzi, 2006; Tao et al., 2010; Gustina and Trudeau, 2011; Whicher and MacKinnon, 2016; Harley et al., 2021).

Recent cryo-EM structures of ERG1 provided exciting insight into the basic structure of ERG1 channels. The most notable feature is the positioning of each voltage sensing domain directly against the pore-forming helices (S5 and S6) of their own subunit in ERG1 (Figure 2A; Wang and MacKinnon, 2017; Asai et al., 2021). A very short S4–S5 linker sequence functions as a ligand, binding to the C-terminal end of the inner pore S6 helix below the activation gate, and stabilizes the closed state at rest (Whicher and MacKinnon, 2016; Malak et al., 2017, 2019; Wang and MacKinnon, 2017). This stands in contrast to the broader family of voltage gated potassium channels, which display a domain-swapped voltage sensing domain (Figure 2B).

Two potential accessory subunits, KCNE1 and KCNE2, have been suggested to interact with the pore-forming ERG1 subunit (McDonald et al., 1997; Isbrandt et al., 2002; Abbott et al., 2007). KCNE1 and KCNE2 are single transmembrane helices with extracellular N-termini and cytoplasmic C-termini (Abbott and Goldstein, 2001). Originally reported as modifiers of the KCNQ1 channel (Takumi et al., 1991; Melman et al., 2001, 2004; Panaghie et al., 2006; Chen et al., 2009; Goldman et al., 2009; Strutz-Seebohm et al., 2011), KCNE1 and KCNE2 expression alters ERG1 gating kinetics and protein degradation in native cardiac tissue, as well as heterologous expression systems (Yang et al., 1995; Bianchi et al., 1999; Mazhari et al., 2001; Ackerman et al., 2003; Zhang M. et al., 2012). Both subunits are expressed in the mammalian brain along with KCNE3–KCNE5 (Lundquist et al., 2006; Sjostedt et al., 2020), where they could act to modify ERG1 function. The specific role of these accessory subunits in native ERG1 function is still of debate.

KCNH2 encodes multiple ERG1 distinct transcripts that, when transcribed, combine to form hetero-tetrameric ERG1 channels: ERG1a, ERG1b, ERG1c, and ERG1_USO (Figure 3). All ERG1 subunits share an identical transmembrane core but vary in the length and structure of their N- and C- termini. ERG1a was the first subunit identified and the most studied, however, all five transcripts have been shown to modify ERG1 currents. ERG1a, ERG1b, and ERG1c transcription is initiated from distinct subunit-specific promoter regions making them alternate transcripts of the same gene (Lees-Miller et al., 1997; London et al., 1997; Huffaker et al., 2009). Conducting ERG1 channels of native tissue can comprise at least three subunits: ERG1a, ERG1b, and ERG1c (Lees-Miller et al., 1997; London et al., 1997; Jones et al., 2004, 2014; Huffaker et al., 2009; Calcaterra et al., 2016; Carr et al., 2016; Goversen et al., 2019). 15 exons encode the ERG1a transcript, the first five of which encode the long N-terminal domain that contains the channel’s PAS domain.

In ERG1b, a transcription start site in intron 5 encodes a single 1b-specific exon that replaces exons 1–5 and encodes a shorter, PAS-deficient N-terminal domain (London et al., 1997; Splawski et al., 1998). It was also demonstrated that ERG1a and ERG1b preferentially interact and co-assemble in the ER to form heterotetrameric ion channels at the plasma membrane in cardiac tissue and human stem cell-derived cardiomyocytes (Kagan et al., 2000; Jones et al., 2004; Sale et al., 2008; McNally et al., 2017). Additional work in human heart demonstrated that individual ERG1a and ERG1b mRNA transcripts co-assemble to promote channel translation, thereby enhancing ERG1 surface expression and membrane currents (Eichel et al., 2019).

A third transcription start site within intron 2 initiates a third ERG1 subunit, ERG1c (KCNH2-3.1). ERG1c lacks the first two ERG1a exons and contains an extended exon 3. Thus, the ERG1c subunit lacks the first 102 amino acids of ERG1a and displays electrophysiological properties similar to ERG1b. ERG1c is associated with cognitive dysfunction (Huffaker et al., 2009; Calcaterra et al., 2016; Carr et al., 2016). ERG1a and ERG1b are expressed in both heart and brain, whereas ERG1c appears to be largely limited to the mammalian brain, particularly in the hippocampus (Warmke and Ganetzky, 1994; Saganich et al., 2001; Guasti et al., 2005; Huffaker et al., 2009; Uhlen et al., 2015; Carr et al., 2016; Ren et al., 2020).

The KCNH2 splice variant, ERG_USO, is generated by alternative splicing leading to an additional 88-amino acids extending from exon 9 that then ends with a 3′ UTR in intron 9, thereby omitting the final six KCNH2 exons. ERG_USO channels can generate from any of the three N-terminal variants (ERG1a_USO, ERG1b_USO, ERG1c_USO) and form non-conducting channels that may play a role in regulating ERG1 current density (Kupershmidt et al., 1998; Guasti et al., 2008; Gong et al., 2010). Future studies may identify additional ERG1 splice variants, as the role of KCNH2 in tissues outside the brain and heart is investigated. Finally, despite improved understanding of ERG1 subunit heteromerization, the native ERG1 subunit stoichiometry in vivo has yet to be determined.

ERG1 channels, like other voltage-gated channels, transition between closed, open, and inactivated states in response to changes in the membrane electric field. It is the concerted movement of the four voltage sensing domains that drives the conformational changes that open (activate) or close (deactivate) the pore (Hodgkin and Huxley, 1952; Armstrong and Bezanilla, 1973; Bezanilla et al., 1982; Piper et al., 2005; Goodchild and Fedida, 2014; Dou et al., 2017). ERG1 activation and deactivation are unusually slow, occurring over hundreds of milliseconds to whole seconds. In contrast, ERG inactivation, which is believed to be a C-type mechanism at the selectivity filter, is remarkably fast (≤10 ms). Thus, activated ERG channels inactivate almost immediately (Kiehn et al., 1996, 1999; Rasmusson et al., 1998). It is this disparate time course of ERG activation/deactivation vs. inactivation that gives rise to the unique current profile of the ERG channel family (Figure 4). During depolarization ERG channels quickly transition into their inactivated state, thereby suppressing current at depolarized potentials. Upon repolarization, ERG channels quickly recover from the inactivated state, but because of the slow deactivation time course, the channels remain in a conductive state.

While the role of the voltage sensing domain is somewhat clear, how voltage sensing domain movements translate to pore opening/closing is still a topic of discussion. ERG1 cryo-EM structures, along with homology models with KcsA, Shaker, Eag1, K_vAP, and K_v1.2 have provided excellent insight into ERG1 channel structure and function. Most K_v channels have a long helical S4–S5 linker that allows the VSD from one subunit to interact with the pore domain from the adjacent subunit – a domain-swapped configuration (Long et al., 2005, 2007). The S4–S5 linker of ERG1 (as well as other KCNH, cyclic nucleotide-gated, and K_Ca channels) is short, tethering the VSD to the pore domain of the same subunit – a non-domain swapped configuration. The positioning of the ERG voltage sensor helps to explain the interesting finding that splitting ERG1 channels at the S4–S5 linker, which resides between the voltage-sensing and pore-forming domains, still produces functional voltage-dependent channels (Lörinczi et al., 2015; de la Peña et al., 2018a,b). These data suggest that the voltage-sensing domains of ERG1 and other non-domain swapped channels couple with the pore through direct interactions between the transmembrane helices, rather than through the S4–S5 linker as seen in Shaker-like potassium channels (Lu et al., 2002; Whicher and MacKinnon, 2016; Hite and MacKinnon, 2017; Wang and MacKinnon, 2017).

The N-terminal PAS domain and the C-terminal CNBHD tightly regulate the kinetics and voltage dependence of ERG channel activation and inactivation. In ERG1a, the PAS domain “docks” with the C-terminal CNBHD, allowing the N-terminal PAS cap to interact directly with the S4–S5 linker. The CNBHD is homologous to the cyclic nucleotide domains of nucleotide gated channels (Guy et al., 1991; Warmke et al., 1991; Zagotta et al., 2003), but the CNBHD of ERG1 is not bound by cyclic nucleotides (Robertson et al., 1996; Brelidze et al., 2009). Instead, amino acids immediately distal to the ERG1 CNBHD (860–862) form an “intrinsic ligand” that binds to the CNBHD at the predicted cyclic nucleotide binding pocket (Brelidze et al., 2012, 2013; Codding and Trudeau, 2019). Disrupting the ERG1 intrinsic ligand disrupts the PAS/CNBHD interaction (Codding and Trudeau, 2019). Disrupting the interaction between the PAS and CNBHD significantly accelerates activation, deactivation, and inactivation recovery (Gustina and Trudeau, 2009, 2012; Gianulis and Trudeau, 2011; Codding and Trudeau, 2019). ERG1b and ERG1c lack fully functional N-terminal domains. As a result, the activation and deactivation kinetics of ERG1a are slow in comparison (Sanguinetti et al., 1995; Trudeau et al., 1995; Wang et al., 1997a). ERG1b and ERG1c homomers have dramatically faster activation, deactivation, and inactivation recovery kinetics (Saenen et al., 2006; Larsen et al., 2008; Larsen and Olesen, 2010; Trudeau et al., 2011; Heide et al., 2012); making the PAS-deficient subunits major promoters of inactivation recovery (Perissinotti et al., 2018).

In addition to ERG1, the brain expresses two additional ERG orthologs: ERG2 and ERG3, encoded by KCNH6 and KCNH7, respectively (Warmke and Ganetzky, 1994; Shi et al., 1997; Wymore et al., 1997; Papa et al., 2003; Polvani et al., 2003; Guasti et al., 2005). All three orthologs display region and neuronal cell type-specific distributions (Warmke and Ganetzky, 1994; Chiesa et al., 1997; Wymore et al., 1997; Saganich et al., 2001; Papa et al., 2003; Figure 5).

Inherited and acquired alterations in ion channel expression and function lead to alterations in electrical function, which ultimately provide substrates for both arrhythmias and seizures. While the phenotype of any given channelopathy may be primarily neuronal or cardiac, as many of these channels are expressed in both the brain and heart, they often include electrical disturbances in both the heart (arrhythmias) and brain (seizures) (Goldman et al., 2009; Johnson et al., 2009; Glasscock et al., 2010; Auerbach et al., 2013, 2016; Kalume et al., 2013; Anderson J. H. et al., 2014). As discussed above, KCNH2/K_v11.1 is expressed in both the heart and brain. Despite KCNH2/K_v11.1 being predominantly studied in the heart, due to its critical role in repolarization, it was initially cloned from a human hippocampal cDNA library (Warmke and Ganetzky, 1994). Furthermore, mutations in the drosophila Kcnh2 homolog are associated with ether-induced seizure-like activity (Ether-à-go-go dance), and heat-induced seizures (Kaplan and Trout, 1969; Titus et al., 1997; Hill et al., 2019).

Long QT syndrome (LQT) is an arrhythmogenic cardiac disorder most notably associated with prolongation of the QT interval measured by a surface ECG (Moss, 2003). At least 17 distinct genes can trigger LQT [for review of LQT genetics see Skinner et al. (2019)]. LQT type 2 (LQT2) is caused by loss-of-function variants in the KCNH2 gene that result in reduced cardiac I_Kr, cardiomyocyte action potential prolongation, and QT_c prolongation on the surface ECG (Sanguinetti et al., 1995; Trudeau et al., 1995). LQT2 patients and animal models of LQT2 exhibit cardiac hyperexcitability, such as early after depolarizations in myocytes, and ectopic activity (e.g., premature ventricular complexes) (Brunner et al., 2008). LQT2 patients are prone to ventricular tachy-arrhythmias, such as torsades-de-pointes, and ultimately sudden cardiac death (Crotti et al., 2008).

LQT2-associated KCNH2 variants cause either reduced ERG1 channel synthesis (Class 1), disruption in the intracellular transport/trafficking of the ERG1 protein to the cell membrane (Class 2), abnormal ERG1 channel gating (Class 3), or altered ERG1 permeability/selectivity (Class 4) (Delisle et al., 2004; Anderson C. L. et al., 2014; Figure 1). Most LQT2-linked variants reduce the number of channels on the membrane via Class 1/2 mechanisms; others are nonsense mutations, and the majority predict haploinsufficiency through nonsense-mediated RNA decay (Class 1 mechanism) (Splawski et al., 2000; Anderson et al., 2006; Gong et al., 2007; Stump et al., 2013; Anderson C. L. et al., 2014; Mehta et al., 2014, 2018). In a study analyzing 226 LQT2-associated KCNH2 variants, 62% are nonsense mutations, 24% are in-frame insertion/deletion, 7% are splice site mutants, and 3% are in-frame ins/del in the KCNH2 gene (Kapplinger et al., 2009). 32% of the variants resided in the transmembrane and pore-pore domains, 29% in the NH₂ terminal (8% at the PAS/PAC domains), and 31% at the carboxy-terminus (8% at the CNBHD) (Kapplinger et al., 2009). Interestingly, mutations within the pore coincided with an increased risk of cardiac arrhythmia (Moss et al., 2002; Shimizu et al., 2009). A later report studying 167 KCNH2 missense variants demonstrated that 76% of KCNH2 mutations at the ERG1 pore totally abolished surface trafficking of co-expressed wildtype channels – a strictly dominant negative phenotype (Anderson C. L. et al., 2014). This strict dominant negative effect was not observed for mutations in other regions of the channel (Anderson C. L. et al., 2014).

There are multiple case studies of LQT2 patients with a history of epilepsy and seizure events (Omichi et al., 2010; Zamorano-Leon et al., 2012; Partemi et al., 2013, 2015; Li et al., 2016; Miyazaki et al., 2016; Zarroli and Querfurth, 2018; Figure 1 and Table 1). Using two LQT patient datasets, Johnson et al. (2009) and Auerbach et al. (2016), reported a higher prevalence of seizures (i.e., history of seizures/epilepsy or taking anti-anti-seizure medications, ASMs) in LQT2, compared to LQT1, LQT3, and LQT genotype-negative un-/related participants. KCNH2 variants in the pore domain conferred the highest prevalence and risk of seizures (Auerbach et al., 2016). Among LQT patients with EEG evaluation, epileptiform activity, convulsive seizures, and an epilepsy diagnosis were documented at a rate 5-fold higher in LQT2 vs. all other types of LQT (Anderson J. H. et al., 2014). Additionally, LQT2 patients with interictal and ictal epileptiform activity in the temporal lobe has been reported (Anderson J. H. et al., 2014).

Epilepsy in LQT2 patients also appears to be independent of cardiac electrical dysfunction. Despite similar QT_c durations between LQT1 and LQT3 patients (Schwartz et al., 2001; Moss et al., 2007; Shimizu et al., 2009), LQT2 patients have the highest prevalence of seizures [12% LQT1, n = 432; 18% LQT2, n = 420; 9% LQT3, n = 113; 5% LQT^(–), n = 936] (Auerbach et al., 2016). Time-dependent adjusted risk assessments indicated that LQT2 patients are at the highest risk of seizures vs. all other groups (Auerbach et al., 2016). Further, there are multiple reports of epilepsy patients (including SUDEP) carrying LQT2-associated variants with minimal diagnosed cardiac electrical dysfunction (Keller et al., 2009; Anderson et al., 2012; Bagnall et al., 2016). Beta-adrenergic blockade reduces the time-dependent risk of arrhythmias (p = 0.004), but not seizures (p = 0.324) (Auerbach et al., 2016). Similar differences are seen when comparing only patients with both a history of seizure/epilepsy and prescribed ASMs (Auerbach et al., 2016). Collectively, these data suggest that cardiac electrical dysfunction is not a prerequisite for epileptogenesis in the context of KCNH2 variants.

It is also important to recognize that there is an apparent bidirectional relationship between arrhythmias and seizures in LQT2. For example, a KCNH2 variant provides a substrate for arrhythmias, which may result in cerebral hypoperfusion triggered seizures. Similarly, seizure-mediated autonomic dysfunction and central cardiorespiratory depression may provide triggers for arrhythmias (Nashef et al., 2007). This is particularly relevant as abrupt increases in sympathetic function (startle and stress) are triggers for arrhythmias in LQT2 (Schwartz and Priori, 2004). In a LQT2 patient, ventricular ectopy and the initiation of a near-lethal cardiac arrhythmia (torsades-de-pointes) was captured during a seizure (Omichi et al., 2010). Paroxysmal EEG slow waves were consistent with a potential underlying epileptic phenotype, and the coexistence of dual neuro-cardiac pathologies in LQT2 (Omichi et al., 2010). Furthermore, in a mouse model of LQT1 (Kcnq1 mutation), there is a high prevalence and concordance of EEG abnormalities, seizures, autonomic instability, ECG abnormalities, and arrhythmias (Goldman et al., 2009).

Small molecules that increase terminal glycosylation or activators that increase the ERG1 channel’s open probability have demonstrated the capacity to restore normal channel trafficking and action potential duration (Ficker et al., 2005; Anderson C. L. et al., 2014; Sanguinetti, 2014). These drugs, however, are linked to drug-induced LQT (Spector et al., 1996). Recently, it was found that the use of blockers and activators in combination could increase the functional expression of ERG1 channels (Qile et al., 2019, 2020), raising the possibility that these treatments could be used together to increase channel trafficking to the plasma membrane. Nevertheless, these maneuvers need to be studied further, because there is a high possibility that increasing the mutant ERG1 channels will result in additional problems, such as changes in gating or permeability, which could worsen LQT2 or their impact on neuronal function.

Patients with epilepsy are at a 24-fold increased risk of sudden death compared to the general population (Ficker, 2000). The leading cause of epilepsy-related death is SUDEP (1–6 cases per 1,000 patient/years), and is 2nd among all neurological diseases in years of potential life lost (Thurman et al., 2014). SUDEP is defined as sudden unexpected, witnessed or unwitnessed, non-traumatic, non-drowning death, with or without evidence of a seizure in an individual with epilepsy, excluding status epilepticus, and postmortem examination does not indicate a toxicological or apparent cause of death (Nashef et al., 2012). While the mechanisms for SUDEP remain unknown, cardiac arrhythmias, respiratory disturbances, cerebral hypoperfusion, failed arousal, and autonomic abnormalities are the proposed mechanisms for SUDEP (Devinsky et al., 2016).

In addition to several epilepsy-related gene variants, post-mortem exome sequencing analyses of 61 SUDEP cases identified a high prevalence of genes linked to arrhythmias (21%, n = 13), especially LQT (Bagnall et al., 2016). KCNH2 was among the top 30 genes with the greatest number of rare variants in SUDEP vs. controls (Bagnall et al., 2016). There was a higher prevalence of loss-of-function (3-fold) and rare variant (11-fold) KCNH2 variants in SUDEP cases vs. epilepsy controls (Soh et al., 2021). In a separate study, 13% (6:48) of SUDEP cases had non-synonymous KCNH2 and SCN5A (Na⁺ channel) variants linked to LQT2 and LQT3, respectively (Tu et al., 2011).

Epilepsy is associated with chronic and peri-ictal altered cardiac electrical function, which include QT_c prolongation, T-wave alternans, and conduction defects (Surges et al., 2010a,b,c; Surges and Walker, 2010; Strzelczyk et al., 2011; Auerbach et al., 2013; van der Lende et al., 2015; Verrier et al., 2016, 2020; Shmuely et al., 2020). There is a significantly higher prevalence and risk of arrhythmias in LQT2 patients with vs. without a history of seizures (Auerbach et al., 2016). Thus, a KCNH2 variant linked to LQT2, and epilepsy-mediated electrical remodeling, likely synergistically increase the risk of arrhythmias and SUDEP (Bleakley et al., 2020).

While the above discussion focused on LQT2 variants altering cardiac and neuronal electrical function, many medications have off target ERG1 blocking properties. ERG1 is highly susceptible to open channel block (Vorperian et al., 1996; Mohammad et al., 1997; Drolet et al., 1999; Zhang et al., 1999; Zhou et al., 1999; Sanchez-Chapula et al., 2002; Rajamani et al., 2006; Sanguinetti and Tristani-Firouzi, 2006), and off-target ERG1 block is the primary cause of acquired long QT syndrome (Sanguinetti et al., 1995). Like LQT2 patients, patients with acquired long QT syndrome have an increased incidence of cardiac arrhythmias and sudden cardiac death (Sanguinetti et al., 1995). The FDA mandates that all lead compounds must be screened for off target ERG1 blockade and QTc prolongation, prior to approval (Food and Drug Administration, HHS, 2005). Many ASMs (e.g., phenytoin and lamotrigine) exert off-target I_Kr blocking properties at therapeutic free plasma concentrations (Danielsson et al., 2003, 2005), and would be expected to impact neuronal ERG channels. This is particularly relevant in LQT2 patients, as ASMs, particularly sodium channel blocking ASMs, result in an increase in the recurrent risk of cardiac events (Auerbach et al., 2018). Surprisingly, the impact of off-target ERG1 channel block on neuronal function has not been explored in detail.

Many previous studies have increased our understanding of the molecular mechanisms of ERG1 channel block. These studies identified several key residues as critical drug binding determinants: Thr623 and Ser624 at the inner mouth of the pore helix, and Tyr652 and Phe656 of the S6 helix (Lees-Miller et al., 2000; Mitcheson et al., 2000; Sanchez-Chapula et al., 2003; Fernandez et al., 2004; Kamiya et al., 2006; Durdagi et al., 2010). Access to these binding sites occurs following channel activation (open channel block) (Sanchez-Chapula et al., 2002; Sanguinetti and Tristani-Firouzi, 2006). A subset of ERG1 channel blockers display preferential binding to the ERG1 inactivated state (modulated receptor hypothesis) (Yang et al., 2004). KCNH2 mutations that impair or remove ERG1 inactivation reduce the affinity of numerous clinically relevant drugs such as cisapride, quinidine, and dofetilide (Ficker et al., 1998; Lees-Miller et al., 2000; Yang et al., 2004; Perrin et al., 2008). Similarly, modulating external cations to disrupt inactivation also reduces drug binding (Wang et al., 1997b; Numaguchi et al., 2000). Last, ERG1 subunit abundance, which modifies channel gating, was also shown to mediate the efficacy of block (Sale et al., 2008; Abi-Gerges et al., 2011; El Harchi et al., 2018). Thus, the dynamic nature of the drug-binding site of ERG1 channels is dependent on both the channel’s conformational changes and subunit stoichiometry. Additional positions at the extracellular regions and the outer mouth of the selectivity filter, can bind ERG inhibitors but are far less dependent upon channel gating (Korolkova et al., 2002; Milnes et al., 2003; Zhang et al., 2003, 2007; Liu et al., 2012; Yu et al., 2014). It is highly likely that these characteristics of ERG1 block are conserved in ERG2 and ERG3 given their similar gating and near perfect homology at the pore and S6 helices (Figure 7; Shi et al., 1997).

Therapeutics with off-target inhibitory effects on ERG1 have shown enhanced efficacy in treating schizophrenic patients when ERG1c was upregulated, such as Risperidone (Heide et al., 2016) and antidopaminergic drugs (Apud et al., 2012). Targeted ERG inhibition to treat neuronal dysfunction would promote life-threatening cardiac arrhythmia and is not a viable therapeutic strategy. Activators of ERG1 and its orthologs, ERG2 and ERG3, may represent a more viable treatment for neuronal diseases, such as epilepsy.

ERG1 activators, including RPR260243, ICA105574, and PD-118057 (Kang et al., 2005; Zhou et al., 2005; Gerlach et al., 2010), enhance ERG1 current by inhibiting channel inactivation, promoting activation, or delaying deactivation (Sanguinetti, 2014). ERG1 activators that disrupt inactivation can dramatically shorten the action potential duration in cardiomyocytes (Bentzen et al., 2011; Zhang H. et al., 2012; Yu et al., 2016; Perry et al., 2019; Qile et al., 2019), but are associated with an unexpected increased incidence of arrhythmia in vivo (Bentzen et al., 2011). One report suggested that small molecules that disrupt ERG channel inactivation share a non-selective pharmacophore (Schewe et al., 2019). NS-1643, which disrupts ERG inactivation (Casis et al., 2006; Bilet and Bauer, 2012), was shown to reduce epileptogenesis in mice (Xiao et al., 2018). Unfortunately, ERG1 plays a minimal role in murine cardiac repolarization, so this finding cannot predict potential pro-arrhythmic effects in humans. It does, however, highlight the therapeutic potential of ERG activators in the brain. Harley et al. (2016), demonstrated that selective disruption of the ERG1 PAS domain increased ERG currents and shortened the action potential duration in human cardiomyocytes, suggesting that the ERG PAS domain could work as an alternative target to enhance ERG current amplitude. ERG1 activators that slow channel deactivation have shown limited effects on cardiac repolarization (Zhang H. et al., 2012; Shi et al., 2020a), but could be a useful tool to dampen neuronal excitability. It should be acknowledged that the high homology between orthologs suggests that the development of any ERG1-specific modulator would be challenging.