Distinct classes of potassium channels fused to GPCRs as electrical signaling biosensors

Summary Ligand-gated ion channels (LGICs) are natural biosensors generating electrical signals in response to the binding of specific ligands. Creating de novo LGICs for biosensing applications is technically challenging. We have previously designed modified LGICs by linking G protein-coupled receptors (GPCRs) to the Kir6.2 channel. In this article, we extrapolate these design concepts to other channels with different structures and oligomeric states, namely a tetrameric viral Kcv channel and the dimeric mouse TREK-1 channel. After precise engineering of the linker regions, the two ion channels were successfully regulated by a GPCR fused to their N-terminal domain. Two-electrode voltage-clamp recordings showed that Kcv and mTREK-1 fusions were inhibited and activated by GPCR agonists, respectively, and antagonists abolished both effects. Thus, dissimilar ion channels can be allosterically regulated through their N-terminal domains, suggesting that this is a generalizable approach for ion channel engineering.


INTRODUCTION
Ligand-gated ion channels (LGICs) are natural biosensors that transform chemical stimuli into electrical signals. These signals can be recorded by standard electrophysiological techniques or micro-and nano-electronic systems.
LGICs have some limitations that restrain applications in biosensing, such as the low number of recognized ligands and the short-lived signal induced by intrinsic inactivation process. A new class of man-made LGICs has been designed by physical and functional coupling of different G protein-coupled receptors (GPCRs) to an ion channel (Moreau et al., 2008). The GPCR moiety plays the role of the ligand-binding domain while the fused ion channel forms the ion-selective pore. These manmade LGICs, called ion channel-coupled receptors (ICCRs) broaden the properties of LGICs as: (1) they diversify the range of ligands to those recognized by the fused GPCRs, (2) they generate a sustainable signal, (3) they are selective to potassium, and (4) they have a basal activity that preserves a negative membrane resting potential and allows the detection of both activating and inhibiting ligands. The ICCR technology has multiple applications both in applied research as biosensors in interface with nano-electronic systems (Lim et al., 2015) and in basic research for functional studies of ion MOTIVATION For biosensing applications, electrical signals are attractive since they can be recorded by micro-or nano-electronic systems for developing miniaturized detection devices in biomedical or environmental fields. One of the main limiting steps is the design of sensing elements that specifically recognize ligands such as biomarkers or chemical compounds. Ligand-gated ion channels are natural biosensors that specifically recognize ligands and generate an electrical signal. They have naturally evolved to be finely tuned by regulatory domains or proteins resulting in an appropriate electrical signal. The objective of this work is to leverage these natural biosensors by engineering diverse artificial ion channels that have the desired properties for biosensing or basic research applications. The main challenge is to design de novo allosteric regulations between ion channels and physiologically unrelated membrane proteins. Previously, different G protein-coupled receptors were successfully coupled to a specific ion channel, Kir6.2. In this work, we demonstrate that other ion channels (viral and murine) with distinct oligomerization properties can also be fused to GPCRs via their N-terminal domains. These results open avenues for diversifying engineered ligand-gated ion channels. channels Principalli et al., 2017) and GPCRs (Moreau et al., 2015;Niescierowicz et al., 2014).
The ICCRs were initially conceived as versatile tools with simple protein engineering for swapping the fused GPCR with another one. Several ICCRs were successfully designed with the following GPCRs: human M2 (M2) (Moreau et al., 2008), human D2 L (D2) (Moreau et al., 2008), human b2 adrenergic (Caro et al., 2011), bovine rhodopsin (Caro et al., 2012), human OR2AG1 olfactory (Lim et al., 2015), and human oxytocin (OXTR) (Niescierowicz et al., 2014) receptors. ICCRs are created by fusing the GPCR C termini to the N terminus of the Kir6.2 ion channel. Additional engineering steps are required to create a functional coupling between the two proteins in order to regulate the ion channel gating by the conformational changes of the fused GPCR. The length of the linking region must be adjusted (Moreau et al., 2008) by deleting the first 25 residues of Kir6.2  and by adjusting the length of GPCR C termini to that of the M2 or D2 receptor. This empirical approach appeared to be partially applicable to some GPCRs but not to all of them for reasons that are still unknown. We postulate in this work that changing the ion channel part could facilitate the design of ICCRs and broaden the properties of the signal. Kir6.2 was initially selected because it is naturally regulated by large transmembrane proteins, the sulfonylurea receptors (SURs) (Inagaki et al., 1995;Martin et al., 2017;Lee et al., 2017;Ding et al., 2019). While the molecular mechanism of this allosteric regulation is not precisely defined, structural evidence suggests that multiple domain-domain interactions are involved in the propagation of conformational changes from SUR ligandbinding sites to Kir6.2 gates (Martin et al., 2019). Similarly, in ICCRs, both the N-and C-terminal domains of Kir6.2 play a role in the regulation by the fused GPCR (Principalli et al., 2017).
While Kir6.2 naturally evolved to be allosterically regulated by large transmembrane proteins, this is an exception in the ion channel superfamily, and we asked whether dissimilar ion channels can be also regulated through N-terminal fusions. Ion channels devoid of interactions between domains could simplify the engineering of ICCRs and facilitate the development of a standard and rational protocol for their design. To explore this hypothesis, two potassium channels were chosen from different families and with different stoichiometries.
The second chosen ion channel was the mouse Twik-related K + channel 1 (mTREK-1 or mK 2P 2.1) (Fink et al., 1996). It belongs to the family of the two-pore-domain potassium channels (K 2P ) (Gada and Plant, 2019;Enyedi and Czirjá k, 2010), which has the particularity of functioning as dimers while most potassium channels are tetramers. Moreover, in the structure (Lolicato et al., 2017) of mouse TREK-1 (PDB: 6CQ6), the N-terminal domain is a cytoplasmic extension of the transmembrane helix 1 (TM1) pointing toward the cell interior and does not contain the interfacial helix found in Kir channels (Meng et al., 2016;Nishida et al., 2007) and predicted in Kcv. The interfacial helices are amphipathic a helices parallel to the membrane plane and anchored to the inner leaflet of the membrane by hydrophobic residues ( Figure 1A). Consequently, fusion of GPCR's C terminus to mTREK-1 could directly link the receptors to the first transmembrane helix of the channel. Such a configuration has two advantages: (1) an absence of N-terminal cytoplasmic domain and interfacial helix, which should simplify the strategy of design; and (2) an imposed stoichiometry of two GPCRs per pore in the GPCR-mTREK-1 complex.
The challenge in the design of this new type of ICCR is to create de novo allosteric regulations of ion channels with different structures, oligomeric states, and gating.
With regard to the large diversity of ion channels, the ability to fuse different ion channels with GPCRs would open new perspectives to adjust the properties of ICCRs with desired characteristics such as higher signal amplitude, specific regulations (chemical, mechanical, or physical), and different oligomeric states. These man-made LGICs are also novel tools to study the inherent molecular mechanisms of gating through intermolecular allosteric regulation.

M2 receptor regulates the fused Kcv channel
The first attempts to change the ion channel moiety in ICCRs were made with the Kcv channel. This channel is an attractive candidate since it has a predicted short amphipathic N-terminal helix (interfacial helix) that can be directly connected to the C-terminal amphipathic helix VIII of GPCRs ( Figure 1A). The fusion of the receptor and the channel amphipathic helices could simplify the ICCR design by creating a single helix with an optimal length.
With this objective, we deleted the C-terminal nine residues of M2 (M2_Kcv-09-xx) to reach the last residue of the helix VIII, the palmitoylated Cys 457 ( Figure 1B). Then, incremental deletions of the Kcv N-terminal domain were made ( Figure 1C), linking to the receptor the residues from Val 3 to Pro 13 , the latter being at the cytoplasmic end of TM1 ( Figure 1A).
The ICCRs were expressed in Xenopus oocytes and their expression and regulation were assessed by the two-electrode voltage-clamp (TEVC) technique. These results ( Figure 1C) clearly show that two constructs (M2_Kcv-09-03 and M2_Kcv-09-12) significantly generate larger basal currents (1.21 ± 0.10 mA and 0.86 ± 0.21 mA, respectively) compared with non-injected (NI) oocytes (0.024 ± 0.005 mA). Those large currents suggest that the corresponding constructs shift the equilibrium of Kcv toward an open state. The profile of basal currents (generated in absence of ligand) suggests periodicity of about four residues, with bigger amplitudes peaking at deletions À03, À07, and À12 ( Figure 1C). Such a periodicity is in line with our initial hypothesis of creating a single helix by connecting the helix VIII of the GPCR to the putative amphipathic helix of Kcv.
The functional characterization of the ICCRs in presence of the M2 agonist acetylcholine (ACh) at 5 mM ( Figure 1D) reveals that only M2_Kcv-09-03 is clearly functional with an inhibition of 53.4% ± 1.2%. Constructs with larger deletions up to 12 residues did not yield significant responses, which indicates that most of the Kcv N-terminal domain must be conserved to be functionally connected to the M2 helix VIII. Two reasons could explain the lack of response of the other constructs: (1) a loss of receptor activity, or (2) a lack of functional coupling between the fused proteins. To test these hypotheses, all M2_Kcv constructs were co-expressed with the homomeric G protein-activated Kir3.4(S143T) channel (Kir3.4T). Kir3.4T channels are activated by the Gbg subunits of Gi/o proteins, which are endogenously expressed in Xenopus oocytes (Hatcher-Solis et al., 2014), and they thus act as reporters of GPCR activation. The results (Figure 2A) show that the deletions À02 to À06 of the Kcv N terminus did not affect the G protein activation by M2, while larger deletions impaired this function. Consequently, the lack of Kcv regulation is not related to a lack of receptor function for the constructs À02 to À06. By contrast, larger deletions have a clear detrimental effect on the receptor activity, which pre-cludes any functional coupling with the fused channel. The lack of regulation of Kcv in constructs with receptors having the ability to activate G proteins suggests that the activation of G protein pathways is not a mechanism involved in the regulation of Kcv ICCR. To confirm this assumption, controls were made by co-expression of unfused M2 and Kcv and by expression of Kcv alone, and the results (Figures S1A-S1C) demonstrate an absence of regulation of unfused Kcv by M2 and by ACh. Consequently, the regulation of M2_Kcv-09-03 by ACh is mediated by the fusion between the receptor and the channel.
Kcv has well-characterized current-voltage relationships (Gazzarrini et al., 2002). Since our experiments are performed in voltage-clamp conditions, voltage dependence does not affect the regulation of the channels by the fused GPCRs in our experimental conditions. However, the fusion with M2 could have affected the voltage dependence of the fused channels and this could be of interest for basic research on the voltage dependence of Kcv or for applications relying on voltage modulation. Measurements of steady-state currents correlated with Kcv N-ter  (M2), the M2 C terminus truncated of its last nine residues (M2 (À9)), the D2 L C terminus (D2), OXTR C terminus truncated of its last 42 residues (OXTR (À42)) and the Kcv N terminus truncated of its first three residues. The nomenclature of ICCRs are GPCR-channel-x-y with x residues deleted at the GPCR C terminus and y residues deleted at the channel N terminus. Positions of the GPCR helix VIII (blue) and the putative Kcv amphipathic helix (red) are represented above the sequences.
(D) Percentage of current change induced by 5 mM ACh on the indicated ICCRs expressed in Xenopus oocytes. TEVC recordings were performed in external 91 mM K + at Vm = À50 mV. In M2_Kcv, the last nine residues of M2 are deleted (À09) and incremental deletions in Kcv N-terminal domain were performed (À02 to À12). Negative values indicate an inhibition of the fused Kcv. One sample t test; ****p < 0.0001. Number of recordings (N) is indicated beside the bars. Error bars are SEM.
incremental clamped voltages (I/V curves in Figures S1D-S1F) suggest that the current-voltage relationship has been modified by the fusion of Kcv with M2, since the I/V curve is more linear than the curve with Kcv alone. Consequently, the fusion seems to affect the voltage dependence of Kcv, which must be taken into account in voltage-dependent applications of Kcv ICCRs.
The design of only one functional M2_Kcv ICCR out of 11 tested constructs highlights the high precision required in protein engineering for designing a Kcv ICCR regulated by a M2 agonist. This regulation is an inhibition of the channel, which was also observed with the equivalent Kir6.2 construct: M2_Kir6.2-09-25 . Kcv and Kir6.2 ICCRs have similar characteristics. Antagonist sensitivity, previously observed with Kir6.2 ICCRs, is preserved with Kcv as shown in Figures 2B and 2C. Co-application of 1 mM atropine with ACh 5 mM on M2_Kcv-09-03 almost entirely abolished the ACh response. Consequently, the fused Kcv channel is reversibly regulated by the agonist and the antagonist-bound states of M2. Concentration-effect curves ( Figures 2D-2G) with ACh and the synthetic agonist carbachol (CCh), indicate half maximal inhibitory concentrations (IC 50 s) of 3.5 mM and 0.7 mM, respectively, which is in the same order of magnitude as IC 50 s of ACh (1.25 mM) and CCh (1.2 mM) observed with Kir6.2 ICCRs . The Hill coefficients for ACh (0.37) and CCh (0.98) suggest an absence of cooperativity as observed with unfused M2 receptor (Jakubík et al., 1997).
These findings demonstrate that Kir6.2 is not the only ion channel that can be functionally coupled to a GPCR. A minimalist potassium channel, composed only of a short amphipathic helix and two transmembrane helices, can be engineered for GPCR regulation and it improves the ICCR technology with a higher signal amplitude (53.4% ± 1.2% of inhibition) compared with Kir6.2 (27% ± 3%). However, the functional coupling requires a highly precise linker engineering since only one functional construct was obtained over 11 tested deletions.
The Kcv channel is not regulated by D2 and OXTR receptors On the basis of the functional M2_Kcv-09-03 construct, we designed new Kcv ICCRs with other GPCRs. The human dopaminergic D2 receptor was chosen since it has already been successfully coupled to Kir6.2 (Moreau et al., 2008). The C terminus of D2 being nine residues shorter than that of M2 (Figure 1B), the fusion D2_Kcv-00-03 features the same linker size as M2_Kcv-09-03. TEVC recordings of D2_Kcv-00-03 showed neither significant amplitudes of basal currents (0.026 ± 0.005 mA) (Figure S1G) nor any regulation by the newly fused receptor ( Figure S1H). To assess whether the functionality of the D2 receptor moiety persisted within the ICCR, the G protein-activated Kir3.4T channel (Moreau et al., 2015) was co-expressed with D2_Kcv-00-03. The modulation of current upon dopamine application ( Figure S1I) demonstrates that D2 receptor still binds dopamine and triggers the activation of the endogenous Gi/o proteins. Consequently, the lack of ICCR response is not due to a loss of D2 receptor function.
We next engineered fusions of Kcv with another GPCR: the human oxytocin receptor (OXTR). OXTR also created a functional ICCR with Kir6.2 (Niescierowicz et al., 2014). Compared with the wild-type (WT) OXTR, the fused receptor had two modifications: (1) its third intracellular loop was replaced by the T4 phage lysozyme (T4L) domain (Cherezov et al., 2007) for crystallographic studies, and (2) its last 42 residues were truncated to match the M2 (À09) C terminus length ( Figure 1B). That fused construct (OXTR[T4L]_Kcv-42-03) generated significant basal currents (0.47 ± 0.10 mA) (Figure S1G), reminiscent of the M2_Kcv-09-03 construct. However, applications of oxytocin did not induce any response ( Figure S1J). To determine whether the lack of G protein activation by OXTR [T4L] was involved in the lack of ICCR regulation, a construct with intact third intracellular loop and without T4L domain was designed: OXTR_Kcv-42-25. The results in Figures S1G and S1K show that, while T4L removal decreases the basal activity of OXTR ICCR, it does not induce Kcv regulation, indicating that G protein pathways are not involved in the regulation of the Kcv ICCR.
Consequently, the linker size determined for M2 receptor cannot be directly transferred to two other GPCRs, indicating that the precise design of the functional M2_Kcv specifically suits the sequence and the conformations of the M2 receptor. It is likely that D2, OXTR, and other GPCRs could be functionally coupled to Kcv. However, the identification of the correct linker for each receptor would be fastidious and does not meet the objective of simplifying the design of ICCR. We considered that the most suitable strategy for designing fusions of Kcv with different GPCRs would be based on computational studies. However, to be absolutely accurate, this strategy requires knowledge of the three-dimensional structure of Kcv, which has not been obtained yet. In consequence, we decided to focus our efforts on the creation of fusions with another ion channel, mTREK-1.
The homodimeric mTREK-1 channel is regulated by the M2 receptor As it was initially the case for the Kcv channel, there is no prior evidence that TREK-1 can be allosterically regulated by a receptor through an N-terminal fusion. However, using TREK-1 as the channel moiety would provide desirable properties to ICCRs. First, the higher conductance of TREK-1 (14.8 ± 3.3 mS in K +rich solution [Fink et al., 1996] versus 76.4 ± 1.0 pS for Kir6.2 [Inagaki et al., 1995]) increases the signal amplitude and would allow the detection of weakly expressed ICCRs. Second, as TREK-1 is a dimer and not a tetramer, a new stoichiometry of two GPCRs per pore is imposed. The first seen N-terminal residues of the mTREK-1 dimer are located on opposite sides of the channel. Thus, the fused GPCRs would not have direct contact with each other (Figures S2A and S2B). Third, the absence of an N-terminal amphipathic helix in the structure of TREK channels, the murine channel (mTREK-1) (Lolicato et al., 2017), and the human TREK-2 channel (Dong et al., 2015) may lead to simplified protein engineering. In absence of identified N-terminal amphipathic helix in mTREK-1, the previous optimization of Kir6.2 and Kcv N termini could not be re-exploited and the strategy of fusion between GPCRs and mTREK-1 had to be re-investigated. The M2 receptor was once again used as a GPCR archetype and we tested both a full-length and a C-terminal nine-residue truncated construct. Postulating that close proximity of the receptor to the pore would be the most efficient position to create a regulation of the channel, we linked the receptor C terminus to the proximal region of the first transmembrane helix (TM1) of mTREK-1.
In the mTREK-1 structures (Lolicato et al., 2017;Pope et al., 2020) (PDB: 6CQ6 and 6V36), the residue Ser 37 is located at the cytoplasmic end of TM1 and points away from the pore (Figure S2C). Ser 37 thus seemed a favorable initial position for fusing the M2 receptor without disrupting the channel function. Deletion of D1-66 in TREK-2 (equivalent to D1-41 in mTREK-1) produced functional channel with higher unitary conductance (Simkin et al., 2008). Consequently, the first mTREK-1 ICCR was M2_T-00-37 with zero deletion in the M2 receptor C terminus (full-length receptor) and 37 residues deleted in the mTREK-1 N terminus. Three additional, slightly larger, deletions in the mTREK-1 N terminus were also created, À40, À43, and later À46 ( Figure S3A), with the objective of bringing the receptor even closer to the pore. The periodicity of three residues in the truncations was used to preserve the orientation of the fused residues in the TM1 a helix. The first TREK-1 residues of the deletions À37, À40, À43, and À46 (Ala 38 , Val 41 , Trp 44 , and Val 47 respectively) belong to or surround a conserved region in the TREK-1 family. Unfused N-terminally truncated (À37, À40, and À43) mTREK-1 were initially created as controls. Deletion of the last nine residues of M2 (M2_T-09-43) was also explored in order to directly connect the M2 helix VIII with the mTREK-1 helix TM1.
These constructs were heterologously expressed in Xenopus oocytes and functionally characterized with the TEVC method. Controls with N-terminal deletions of unfused mTREK-1 ( Figures  S3B-S3E) showed a similar profile of I/V curves of the truncated Cell Reports Methods 1, 100119, December 20, 2021 5 Article ll OPEN ACCESS channels to that of the WT channel, with lower current amplitudes for all truncated channels and a lower activation by 10 mM BL1249 for mTREK-1 À43. These results show that incremental N-terminal deletions of mTREK-1 preserve the voltagedependence characteristics and ligand sensitivity of the channel, but with lower current amplitudes correlated with the length of deletions. In ICCRs, the results ( Figures 3A-3E) showed an inverted effect with an increase of the amplitude of ACh response correlated with an increase of the length of deletions from À37 to À43. Thus, a slight activation of M2_T-00-37 by ACh 5 mM (3.9% ± 0.9%) was observed, which increased with larger deletions -00-40 and -00-43 (20.2% ± 3.5% and 49.9% ± 5.1%, respectively) but decreased for the deletion -00-46 (35.2% ± 1.6%). Thus, while the regulation of the gating is decreased in unfused truncated channels when N-terminal deletions increase, shorter distances between the receptor and the channel improve the regulation of the gating in a certain range with an optimum of 43 residues deleted over the four deletions tested. This is consistent with a mechanical transduction of GPCR conformational changes to the channel gate(s), which is improved by closer contact between the two proteins until steric hindrances occur. However, another possibility is a differential effect of the different deletions on the receptor activity. To evaluate this hypothesis, the activation of G proteins by the fused receptor was assessed for each construct. The fusions M2_T-00-37, À40, À43, and À46 were co-expressed with the G protein-activated potassium channel Kir3.4T. The results ( Figures 3F and S4A-S4D) show clear activation of Kir3.4T channels induced by ACh and no significant differences of the amplitudes between the four constructs. Thus, the different deletions did not differentially affect the activation of the fused receptor.
The effect of fusion on the voltage dependence of mTREK-1 was also assessed for potential applications based on voltage modulation. The I/V curves of mTREK-1 and M2_T-00-43 (Figures S4E-S4G) show a similar profile between the fused and unfused channels. The outward rectification is preserved but with a lower current amplitude for the ICCR, which was previously observed with all other fused channels.
As for the Kcv ICCR, atropine was used to test the sensitivity of M2_T-00-43 to antagonists. Co-application of 1 mM atropine to 5 mM ACh inhibited ACh-induced activation (Figures 3G and  3H). The gating of mTREK-1 is thus sensitive to both agonistand antagonist-bound states of the M2 receptor.
Concentration-effect curves ( Figures 3I and 3J) with ACh and CCh confirm the correlation between agonist concentration and signal amplitude. As observed with the Kcv ICCR, the values of EC 50 and the Hill coefficient (nH) for CCh are in the same order of magnitude as previously observed. For ACh, the EC 50 is decreased (0.34 mM versus 1.25 and 3.5 mM for Kir6.2 and Kcv ICCRs, respectively), while the Hill coefficient (0.76 versus 0.37 for Kcv ICCR) increases but remains lower than 1, indicating an absence of cooperativity. This higher apparent affinity for ACh with M2_T ICCR could be related to the application of high sodium buffer (2K + buffer) before measuring ligand response. Friedman et al. (2020) demonstrated that removal of Na + decreases the potency of ACh on the M2 receptor. High sodium buffer has been specifically used as reference before and after ligand applications with mTREK-1 ICCRs for the baseline determination since this channel is only partially blocked by barium. Kir6.2 and Kcv being efficiently blocked by extracellular barium, the barium-sensitive currents were used as reference in recordings of ICCRs composed of these channels.
In a second round of design, ICCRs bearing a deletion of the last nine residues from the M2 receptor moiety were also constructed since such a deletion generated functional ICCRs with both the Kir6.2 and Kcv channels. The design of M2_T-09-40 and -09-43 creates a direct contact between the last residue (Cys 457 ) of the amphipathic helix VIII of M2 and the TM1 of mTREK-1 (at Val 41 and Trp 44 , respectively). TEVC recordings showed no significant regulation of M2_T-09-40 and -09-43 by ACh 5 mM (Figures 4A-4C). The control with the G protein-activated potassium channel Kir3.4T demonstrated that, despite its close proximity with TREK-1, the M2 receptor in M2_T-09-40 and -09-43 was still able to activate Gi/o proteins in presence of 5 mM ACh ( Figures S4H and S4I). Consequently, the absence of ACh-induced regulation of M2_T-09-40 and -09-43 is not due to a loss of receptor function.
The regulation of mTREK-1 by the fused M2 receptor is not carried out by intracellular effectors WT TREK channels can be activated by the Gai protein-protein kinase A pathway in mammalian cells (Cain et al., 2008;Lesage et al., 2000). To rule out the possibility that the functionality of the M2_T ICCR is not directly due to the fusion of the two partners but related to the activation of the Gai protein pathway (or other intracellular effectors) and a subsequent indirect effect on the mTREK-1 moiety, four controls were performed. First, the unfused receptor and channel were co-expressed. Figures 4D  and 4E shows a slight (À8.9% ± 0.5%) but not significant inhibition in presence of ACh 5 mM (p = 0.0625). Therefore, mTREK-1 is not regulated by the M2 receptor when the two proteins are not fused. This result indicates that the activation of mTREK-1 ICCRs does not rely on intracellular effectors in our experimental conditions (Xenopus oocyte, 91 mM extracellular K + , Vm = À50 mV, 0.3 mM niflumic acid). The second control is based on the irreversible inactivation of Gi/o proteins by the pertussis toxin (PTX) (Hatcher-Solis et al., 2014). When mRNAs coding for PTX (S1 subunit) and M2_T-00-43 were co-injected, amplitudes of ICCR activation were similar with and without PTX (49.3% ± 4.4% and 49.9% ± 5.1%, respectively; Figures 4D, 4F, and G), indicating that Gi/o proteins are not involved in the activation. Expression and function of PTX were confirmed by the inhibition of Kir3.4T activation by the M2 receptor ( Figures 4H-4J). These controls were performed on the same day as the recordings with ICCR. Third, key phosphorylation sites of mTREK-1 (S300 and S333) (Murbartiá n et al., 2005) were mutated in alanine and the results (Figures S5A-S5D) show an increase of the basal current amplitude as previously reported for these mutants and a preserved activation of the ICCR by ACh. While the activation of the S300A S333A double mutant by ACh is significant, the percentage of ACh-induced activation is lower than the WT, likely due to the high basal activity of the mutated channel or a partial involvement of the G protein-mediated phosphorylation. To conclusively rule out the potential role of G protein pathways in the response of mTREK-1 ICCR, a fourth control was made by inserting the T4L domain in the third intracellular loop of the  Figures 4D and 4K) demonstrate that the G protein-uncoupled ICCR is still activated by ACh with the same amplitude as WT, confirming the G protein-independent regulation of the ICCR. Altogether, these results demonstrate that the activation of M2_T ICCRs relies on the direct fusion between the two proteins and does not involve indirect intracellular signaling in our experimental conditions.
The D2 receptor also regulates the fused TREK-1 channel To determine whether the regulation of the fused mTREK-1 is not limited to the M2 receptor, as observed with Kcv, we constructed fusions with the D2 receptor. The C terminus of the D2 receptor being nine residue shorter than that of the M2 receptor, the construct D2_T-00-43 connects the last residue of the D2 helix VIII (Cys 443 ) to the first helix of mTREK-1 (Trp 44 ) ( Figure 5A). According to the results obtained with M2_T-09-43, such a connection should not create a functional ICCR. To mimic the linker length of the functional M2_T-00-43 ICCR, the D2 receptor C terminus was extended with the last nine residues from the M2 receptor, yielding a construct named D2_T+09 M2 -43 ( Figure 5A). TEVC experiments confirmed that only the latter construct with the extended C terminus (+09 M2 -43) was activated by 5 mM dopamine (12.1% ± 1.7%) while the D2_T-00-43 construct showed a weak inhibition (À4.1% ± 0.6%) that is not significantly different from the control with co-expression of the unfused proteins (À2.2% ± 1.3%) ( Figures  5B-5E). The concentration-effect curve on D2_T+09 M2 -43 (Figure 5F) indicates an EC 50 of dopamine (0.72 mM), which is in the same order of magnitude as the EC 50 (0.4 mM) measured on the equivalent Kir6.2 ICCR (D2_K+09 M2 -25) (Moreau et al., 2008). The Hill coefficient is 0.76, which is similar to the Hill coefficient observed for ACh on M2_T-00-43. These results demonstrate the ability to build functional TREK-1-based ICCRs with two different GPCRs using a similar design of the linker.
Fused mTREK-1 preserves its pharmacological regulation and mutants offer additional properties to the signal generated by the ICCRs Adding diversity to the repertoire of ion channels that can be turned in ICCRs would open new possibilities beyond the simplification of the linker design. We checked whether salient pharmacological properties of mTREK-1 were preserved within a fusion protein. To determine whether the fused mTREK-1 was still regulated by pharmacological compounds, the responses of ICCRs to the activator BL1249 (Pope et al., 2018) and to the inhibitor spadin (Mazella et al., 2010) were assessed. M2_T-00-43 was expressed in HEK293 cells by transfection and its functional characterization was performed by whole-cell patch-clamp recordings. The results, depicted in Figures 6A, 6B, S5E, and S5F, show an activation of M2_T-00-43 by 10 mM BL1249 (880% ± 190% increase of the current) and an inhibition of BL1249 activation by 1 mM spadin (56.8% ± 9.1%). These results confirm that the mTREK-1 moiety retains its native pharmacology.
Numerous structure-function studies demonstrated the possibility to tune the properties, or to create specific properties, of ion channels by site-directed mutagenesis (Subramanyam and Colecraft, 2015). In mTREK-1, a new mutation G293E, inspired by the loss-of-function G293R variant inventoried in the SNPs databank COSMIC, yields a hyperactive channel with higher basal activity, while the selectivity filter mutation I252T confers sodium permeability and a hypersensitivity to stretch activation (Decher et al., 2017) (the latter being causal of ventricular tachycardia in humans). Note that our numbering nomenclature is based on the sequence of the isoform mTREK-1a, which is 15 residues shorter than mTREK-1b. Hence, the I252T mutation is equivalent to the I267T mutation studied by Decher et al. (2017).
Both the G293E and I252T mutants of the M2_T-00-40 ICCR were expressed in Xenopus oocytes and characterized by TEVC. The results, depicted in Figure 6C, show a higher amplitude of the basal current generated by the mutant G293E (4.42 ± 0.58 mA) compared with WT and with the I252T mutant (0.77 ± 0.07 mA and 0.66 ± 0.11 mA, respectively). Thus, the hyperactive G293E mutant considerably increases (>5-fold) the ICCR signal amplitude, a very desirable property for weakly expressed ICCRs. Application of ACh leads to an activation of the hyperactive ICCR (32.5% ± 3.6%) in a proportion not significantly different from its WT counterpart (20.2% ± 3.5%) ( Figures  6D-6F), suggesting that only the basal state of TREK-1 seems affected by the mutation and not the allosteric regulation by the M2 receptor. The mutation I252T affects ion selectivity and increases the stretch sensitivity. In the absence of extracellular Na + and in symmetrical K + concentrations, the basal current of the mutated I252T ICCR was similar to that of the WT ( Figure 6C). The activation induced by ACh (84.7% ± 11.2%) was increased more than 4-fold compared with that of the WT counterpart (20.2% ± 3.5%) ( Figures 6D, 6E, and 6G). The hypersensitivity of the mutant to stretch stimuli could explain the higher activation of the ICCR by M2. Indeed, the conformation changes of M2 are also mechanically transduced to the mTREK-1 gate(s). While the underlying mechanism remains to be elucidated, such a drastic increase of the ligand-induced responses is of great interest for ICCRs. Altogether, the characteristics of two mTREK-1 mutants are preserved within ICCRs and can be exploited to adjust both the basal and ligand-induced levels of the signal.

DISCUSSION
The activity of ion channels (gating) and GPCRs (intracellular signaling) is generally regulated by "long-distance" conformational changes coupling the signal reception sites (ligand-binding sites or physical sensors) and the signal generating sites (gates or intracellular effector-activating sites) (Changeux and Christopoulos, 2016). Those structural rearrangements are governed by highly precise molecular mechanisms optimized along the long evolution of living organisms. Inspired by these natural sensors, numerous studies developed artificial ligand-, light-, or voltage-gated ion channels (Arrigoni et al., 2013;Cosentino et al., 2015;Alberio et al., 2018;Ohndorf and MacKinnon, 2005;Schö nrock et al., 2019;Difrancesco et al., 2015;McCoy et al., 2014;Wang and Zocchi, 2011;Coyote-Maestas et al., 2019;Gorman et al., 2019) in order to extend the diversity of recognized ligands or to help understand gating mechanisms or to forge optogenetic tools instrumental in neuroscience. A common challenge in these studies is to artificially create an allosteric transition between two exogenous domains: one for signal recognition and one for signal generation. In the case of ICCR, the concept was initially based on an ion channel (Kir6.2) that naturally evolved toward an allosteric modulation by a large regulatory protein, the sulfonylurea receptor (SUR) (Inagaki et al., 1995). Functional and structural studies highlighted the role of the channel N-terminal domain in SUR-mediated regulation, suggesting that this domain is one of the elements required for the propagation of sulfonylurea-induced conformational changes from SUR toward the channel gate(s). Fusion of GPCRs to this Kir6.2 N-terminal domain indeed created artificial allosteric regulations of Kir6.2 channel, which is related to "discrete, reversible, conformational change in the [receptor] structure" as defined as "allosteric transition" (Changeux and Christopoulos, 2016).
In the case of the virus-encoded Kcv channel, the situation is reversed. Kcv has not evolved toward allosteric regulation, but, on the contrary, it has preserved the most minimalist sequence enabling potassium selectivity and permeation, in order to create a potassium leak that facilitates viral infection. The channel adopts a constitutive open state, which has been observed in several heterologous expression systems (Gazzarrini et al., 2003). Neither the "gating hinge," formed by a glycine involved in transmembrane helix bending required for channel opening, nor bundle-crossing gates in the cytoplasmic side of transmembrane helices are observed in Kcv (Gazzarrini et al., 2003). Despite its minimalist architecture, the artificial regulation of Kcv could be achieved through fusion of exogenous domains. Light- (Cosentino et al., 2015;Alberio et al., 2018), calcium- (Gazzarrini et al., 2003), and voltage-sensitive (Arrigoni et al., 2013) Kcv channels have been engineered by N-terminal fusion of sensor domains. Fusion with the M2 receptor demonstrates that the conformational changes of a transmembrane receptor can also regulate Kcv function via its N-terminal domain. However, the size of the linker differs from those of all previous artificially regulated Kcv channels. Fusion of the D2 L dopaminergic receptor to Kcv, with the same linker length, did not generate a functional ICCR. The molecular mechanisms governing the allosteric regulation thus depend both on the fused regulatory domain and on the linker size, and their full understanding would require structural studies revealing the transmission mechanisms from sensors to channels. In terms of engineering, the simplicity of Kcv contrasts with the diversity of allosteric regulations it is amenable to. However, in the absence of a linker transferable from one GPCR to another, Kcv is not the appropriate ion channel for creating large sets of ICCRs with our current knowledge.
The rationale behind the use of Kir6.2 and Kcv channels in ICCRs was based on the existence of modulating subunits for Kir6.2 and earlier protein engineering with soluble sensors for Kcv. By contrast, the second approach, which uses the mTREK-1 channel, was counter-intuitive for three reasons. First, the main natural regulatory domain of mTREK-1, responsible for pressure, temperature, pH, and lipid and fatty acid sensing, lies at the C-terminal end of the transmembrane helices (Soussia et al., 2018;Noë l et al., 2011;Pavel et al., 2020), while we have designed N-terminal fusions. Second, mTREK-1 is devoid of an interfacial helix at the N-terminal domain, which is deemed instrumental for the efficient GPCR to channel coupling in Kir6.2 and Kcv. Third, the channel is dimeric, not tetrameric like most other potassium channels, implying a different regulation of the gating, and possibly precluding the ability of the fused GPCRs to contact each other. However, attaching GPCRs to the truncated N terminus of mTREK-1 created positive allosteric regulation of the channel when the first 40, 43, or 46 residues were truncated, the 43 being optimal. Attempts to reduce the linker size and directly connect the GPCR helix VIII to the mTREK-1 TM1 (M2_T-09-XX) abolished the regulation, in contrast with Kir6.2-and Kcv-based ICCRs.
The straightforward transfer of design to constructs with the dopaminergic D2 receptor confirmed that mTREK-1 can be fused and regulated by different GPCRs, provided that their C terminus extends nine residues beyond the palmitoylated cysteine of the helix VIII. These results demonstrate the potential of mTREK-1 to create functional ICCRs with different GPCRs, using a simplified and consensus strategy. This ion channel is consequently a promising candidate for designing a new generation of ICCRs. Moreover, mutations of mTREK-1 expand the properties of the ICCR signal by increasing $6-fold the amplitude of the basal current and $4-fold the amplitude of the ligand-induced response with the mutants G293E and I252T, respectively.
Another particularity of mTREK-1 is its oligomeric state. While Kir6.2 and Kcv ICCRs contain four receptors per pore, mTREK-1 ICCRs impose a new stoichiometry of two GPCRs per pore and demonstrates that two receptors on two opposite sides of the pore are sufficient to open the channel. The Hill coefficients for ACh and CCh were under 1 for both the tetrameric . In all panels, recordings were performed by TEVC as described in Figure 1B. Mann-Whitney test, ****p < 0.0001. Error bars are SEM.
Kcv and dimeric mTREK-1 ICCR, indicating that no cooperativity occurs in the regulation of the fused channels by M2. The imposed oligomeric states of the GPCR could be used as a new approach to evaluate the role of receptor oligomerization in other GPCR-based mechanisms. Moreover, an increasing number of studies demonstrate the physiological role of K 2P channels heterodimerization (Levitz et al., 2016;Blin et al., 2016). In the absence of reliable tools (e.g., specific antibodies, direct assays) for studying specific K 2P channels, ICCRs could be an alternative method.
In perspective, the similarity of structure between different K 2P (TWIK-1, Miller and Long, 2012; TASK-1, Rö dströ m et al., 2020; TASK-2, Li et al., 2020;and TRAAK, Brohawn et al., 2012 suggests that the ICCR technology could be effectively extrapolated to other members of the K 2P family with the objectives of (1) creating pharmacological regulation based on the large A D E F C B Figure 5. D2_TREK-1 is also ligand regulated with an extended linker length (A) Sequences of GPCR C termini that are fused to the N terminus of mTREK-1, which is truncated of its first 43 residues. In the construct D2_T+9 M2 -43, the C terminus of D2 is extended with the last nine residues of M2 to match the same linker length as M2_T-00-43. (B) Percentage of current change induced by 5 mM dopamine on the unfused D2 L and mTREK-1 (D2+mTREK-1, grey bar), the fused D2_T-00-43 (light green bar), and the ICCR with extended D2 C terminus: D2_T+9 M2 -43 (dark green bar). The number of recordings is five, 32, and 17, respectively. One way ANOVA post Dunnett's test; n.s., p = 0.55. ****p < 0.0001. Error bars are SEM. (C) Representative TEVC recordings of D2 co-expressed with unfused mTREK-1. repertoire of GPCR ligands to compensate the paucity of known pharmacological compounds acting on these K 2P channels, (2) broadening the properties of ICCR signals, or (3) studying the gating regulation of specific members of the family.
An additional benefit offered by the diversity of ion channels is the ability to modify the properties of ICCRs for increasing the signal amplitude for adding sensitivities to different stimuli (e.g., ligands, temperature, stretch) or for blocking the signal with specific chemicals or toxins. mTREK-1 has the distinction of being regulated by a large diversity of stimuli (pH, mechanical activation, temperature, volatile anesthetics, poly-unsaturated fatty acids, and phosphorylation), which offers a large choice of tests to control the presence and function of ICCRs in cells or artificial environments such as biosensing platforms.
This study demonstrates that various potassium channels from different families can be allosterically regulated via their N-terminal fusion with a GPCR. This fusion strategy could be extended to a broader range of ion channels with a cytoplasmic N terminus, even if this domain has no known role in gating regulation. In consequence, these results open new horizons for the generation of various man-made LGICs with tunable electrical signal, stimuli, or GPCR stoichiometries.
Limitation of the study The tested receptors belong to the class A (rhodopsin-like) GPCRs. Fusing GPCRs from others classes could generate different results. In particular, the fusion of mTREK-1 to class C GPCRs should disrupt the formation of mandatory dimers. In most cases, the C terminus length of the GPCR must be adjusted, which could affect interactions with cognate proteins, internalization, recycling, and phosphorylation of the fused and engineered receptor.
The surface expression level of the ICCR is dependent on the fused receptor, which could be a limiting factor for measuring significant current amplitude.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: