Structural and functional studies of inward rectifier Kir7.1 and its regulation by the Melanocortin-4 receptor Authors: Alys Peisley, Ciria C Hernandez, Naima S Dahir, Laura Koepping, Ashleigh Raczkowski, Min Su, Masoud Ghamari-Langroudi, Xinrui Ji, Luis E Gimenez, Roger D Cone Published: 2026 ABSTRACT AbstractHereditary defects in the function of the inwardly rectifying potassium channel Kir7.1 in the retinal pigment epithelium are associated with the ocular diseases retinitis pigmentosa, Leber congenital amaurosis, and snowflake vitreoretinal degeneration. However, this ion channel is widely expressed, and studies also suggest that Kir7.1 may be directly regulated in a G protein-independent manner by a GPCR, the melanocortin-4 receptor (MC4R), in specific hypothalamic neurons involved in energy homeostasis. To advance our understanding of this unusual direct regulation of an ion channel by a GPCR, it is important to characterize the structure and function of Kir7.1, and the Kir7.1-MC4R complex. Utilizing cryo-electron microscopy, we present structures of human Kir7.1 and demonstrate the structural basis for the blockade of the channel by a small molecule ML418. We show that channel blockade of Kir7.1 depolarizes MC4R neurons in a slice preparation of the paraventricular nucleus of the hypothalamus (PVH), and activates PVH neurons in vivo, inhibiting food intake and inducing weight loss. We further show that Kir7.1 and MC4R are in close association on the plasma membrane and characterize a fusion construct of MC4R and Kir7.1 that provides an experimental tool for study of Kir7.1 regulation by the MC4R.Subject terms: Hypothalamus, Homeostasis, Structural biology, Biophysics Introduction The inward rectifier potassium channel Kir7.1 is expressed in the CNS but is also distinguished by its expression in various epithelial tissues, notably in the retinal pigment epithelium (RPE), crucial for maintaining the blood-retinal barrier1–4. Mutations in Kir7.1 (Kcnj13) are associated with several ocular disorders, including retinitis pigmentosa, Leber congenital amaurosis, and snowflake vitreoretinal degeneration (SVD), with the severity of visual impairment correlating with the extent of channel dysfunction3,5. A missense mutation in Kcnj13 (c.484 C > T, p.Arg162Trp) has been associated with autosomal-dominant SVD3 while a separate heterozygous mutation at the same codon (c.485 G > A, p.Arg162Gln) is associated with milder adult-onset retinitis pigmentosa5. Data from studies in cell culture and hypothalamic slice preparations also suggest that Kir7.1 can be directly regulated by a GPCR, the melanocortin-4 receptor (MC4R) required for energy homeostasis6,7. In one report, the endogenous ligand of the MC4R, α-MSH, was demonstrated to depolarize MC4R PVH neurons in a slice preparation, and pharmacological studies suggested the depolarization was mediated by Kir7.1 and was independent of G protein activation6. Conversely, the endogenous MC4R inverse agonist agouti-related protein (AgRP) rapidly hyperpolarized the neurons6. Subsequent studies, conducted during the determination of the X-ray crystal structure of the MC4R, demonstrated that replacement of the third intracellular loop of the MC4R with a large fusion protein ablated G protein coupling while leaving MC4R ligand-mediated coupling to Kir7.1 fully intact7. While deletion of Kir7.1 from PVH MC4R neurons removed α-MSH mediated depolarization of PVH MC4R neurons in a slice preparation, no early onset metabolic phenotype was seen in the animals, leaving the physiological function of Kir7.1 in MC4R neurons ambiguous8. While evidence for direct ion channel regulation by a GPCR is limited, the regulation of inward rectifier Kir6.2, by the multi-membrane spanning sulfonylurea receptor9 provides a model for how this might occur. Inward-rectifier potassium channels (Kirs) are lipid-gated ion channels that, unlike other K+ channels, preferentially permit inward rather than outward flow of K+ ions. At membrane potentials more positive than the potassium ion reversal potential (i.e. during depolarization), Mg²⁺ and polyamines enter the cytoplasmic side of the channel, physically blocking the pore. This prevents the flow of K⁺ ions, resulting in the cessation of ion conduction. At negative (hyperpolarized) potentials, these blockers dissociate from the pore, allowing inward movement of potassium ions into the cell10. Kir channels exist as dynamic ensembles that sample a range of conformations, categorized into three main functional states: resting, activated, and inactivated. All Kir channels display a tetrameric arrangement of four identical or related subunits that surround a central pore. Each subunit is composed of a transmembrane domain (TMD) tethered to a C-terminal domain that together with the N-terminus constitutes the cytosolic domain (CTD)11,12. A juxta-membrane slide helix connects the N-terminal peptide to the transmembrane domain. The channel pore is constricted at the transmembrane selectivity filter and helix bundle crossing, and the cytosolic G-loop gate, with the latter two serving as functional gates that facilitate K+ permeation12. The gating of all eukaryotic Kir channels is regulated by lipids such as phosphatidylinositol-4,5-bisphosphate (PI(4,5)P2 or PIP2)10,13,14. PIP2 binds at a canonical site at the interface of cytosolic and transmembrane domains governed by strong electrostatic interactions between PIP2 phosphate groups and basic residues in Kir TM1 and tether helix13,14. Kir channels can be additionally regulated by other lipids including cholesterol15,16. Kir7.1, like other eukaryotic Kir channels, requires phosphoinositides for activation and is robustly gated by PI(4,5)P2. Rohács and colleagues placed Kir7.1 in the Group 2 class of Kir channels that prefer PI(4,5)P2 but are also significantly activated by PI(3,4,5)P3, indicating PI(4,5)P2-dependent gating with broader phosphoinositide tolerance than classical Kir2.1 channels, which are highly selective for PI(4,5)P217,18. In addition, Kir7.1 activity is further modulated by endogenous steroids and by MC4R. The sensitivity of Kir channels to intracellular blockade is dictated by a conserved negatively charged residue in the transmembrane domain, known as the rectification controller18. Many small-molecule inhibitors exploit this site, blocking the conduction pathway through interactions with the rectification controller and adjacent pore-lining residues. One such inhibitor, ML418 (iso-Propyl (1-((5-chloro-8-hydroxyquinolin-7-yl) methyl) piperidin-4-yl) carbamate) is a submicromolar (IC50 = 310 nM) inhibitor of Kir7.1 that is equipotent at Kir6.2/SUR1 and displays at least 17-fold selectivity over Kir1.1, Kir2.1, Kir2.2, Kir2.3, Kir3.1/3.2, and Kir4.119. Here we analyze the structures of full-length variants of Kir7.1 in apo form and bound to the channel blocker ML41819. We then utilize ML418 to study the role of the Kir7.1 in regulating MC4R PVH neurons in slice preparation and in regulating feeding behavior, and body weight in vivo. In addition, we develop and characterize MC4R-Kir7.1 fusion proteins as a tool for studying the structural characteristics that govern the interaction between MC4R and Kir7.1. Results Protein expression and characterization of purified WT and R162Q mutant of human Kir7.1 To fully understand the regulation of Kir7.1 by MC4R it will ultimately be critical to determine the structure of the MC4R-Kir7.1 complex. While multiple MC4R structures are available7,20, this long-term goal can be aided by the determination of the structure of Kir7.1. Towards this end, the human wild-type (WT) and R162Q Kir7.1 channel were produced in Sf9 insect cells using recombinant baculovirus expressing GB1-tagged Kir7.1 constructs. Attempts were made to isolate the clinically significant R162W variant; however, the protein proved to be only marginally soluble and could not be produced in sufficient quantities for further investigation. Size exclusion chromatography (SEC) profiles of affinity-purified proteins revealed distinct elution patterns (Fig. 1A, B). WT Kir7.1 showed a prominent peak at ~11.3 ml (Fig. 1B, and Fig. S1A), corresponding to the channel tetramer. In contrast, the R162Q mutant exhibited an earlier elution peak at ~10.5 mL (Fig. 1A, and Fig. S1C). The addition of the channel blocker ML418 to the R162Q mutant shifted its elution profile to match that of the apo wild-type channel (Fig. 1A, B, and Fig. S1A, D). ML418 did not affect the elution profile of WT Kir7.1 (Fig. S1A, B). These differences in elution profiles suggest distinct conformational preferences among WT, R162Q mutant, and ML418-blocked channels. To investigate binding of the channel blocker ML418 to Kir7.1 we first obtained cryo-EM reconstructions of apo and ML418-bound WT Kir7.1 which refined to low resolution (~4.9 and ~4.0 Å respectively, Fig. S1E–H). A weak residual density was observed in the central pore cavity in the presence of ML418 indicative of low occupancy. We have performed further classifications of the WT data set (Fig. S2). It is evident that the transmembrane domain is dynamic at the individual subunit level. Although we observe fluctuations between open and closed conformers within the transmembrane domain, the cytosolic domain remains disengaged for all observed cryo-EM volumes, also observed for cryo-EM of GIRK2 homomers21. We found that the introduction of the retinitis pigmentosa-associated R162Q mutation into Kir7.1 induced conformational stabilization, enabling us to obtain reconstructions of 3.5 Å resolution with local resolution approaching 2.8 Å (Fig. S3). This stabilization assisted with improvement of resolution, with reconstructions of the WT Kir7.1 only achieving 4.6 Å (Fig. S2). A strong density in the central cavity was observed for reconstructions of the R162Q mutant in the presence of ML418, likely due to the widened helix bundle crossing allowing for greater access. A similar approach has been utilized for the Kir4.1, employing PIP2 to open the channel in order to obtain structures with a bound channel blocker22. Therefore, to understand the binding mode of ML418, we focused on the R162Q mutation, which has been identified in retinitis pigmentosa patients with less severe disease outcomes5. Structures of apo and ML418 bound R162Q mutant of human Kir7.1. A Overlay of Superdex 200 gel filtration profiles of apo and ML418 bound R162Q mutant Kir7.1. B Overlay of Superdex 200 gel filtration profile of apo wild-type (WT) Kir7.1 compared to ML418 bound to the R162Q mutant. C Cartoon representation of cryo-EM structure obtained of apo R162Q mutant (gray, residues 14-325 (excluding residues 31-36), gray) and R162Q mutant bound with ML418 (blue, residues 14-325 (excluding residues 30-39, 161-165) blue). Downward shift of the cytosolic domain (CTD) relative to the transmembrane domain (TMD) is demarcated. D-F Superposition of apo (gray) and ML418 (blue) bound R162Q mutant structures: side view of TMD and CTD. D extracellular view of TMD (E) and cytoplasmic view of CTD (F). Rotation of the cytoplasmic domains in the presence of ML418 is indicated. G-H Backbone trace of selectivity filter residues showing distances between backbone carbonyls in apo R162Q with potassium ions displayed as purple spheres (G). I ML418 bound R162Q with an unoccupied selectivity filter (as in H). Pore radius calculated between van der Waals surfaces by the program HOLE and plotted as a function of the distance from the central point along the pore axis for R162Q mutant (gray), or for the extended ML418 bound R162Q Kir7.1 (blue). Identities of residues at the narrowest and widest radii along the pore axis are indicated along with associated structural elements. J, K Pore volume shown within structure of apo-R162Q colored as gray cartoon (J) and for ML418 bound R162Q Kir7.1 colored in blue ribbons (K) The central pore region is shown as a dotted surface where red is where the pore radius is too narrow for a water molecule, green is where there is room for a single water, and blue is where the radius is double the minimum for a single water molecule. Structure and conformation of the apo- and ML418-bound R162Q mutant Kir7.1 channel We determined structures for the apo and ML418-bound R162Q mutant Kir7.1, at 3.5 and 3.9 Å resolutions, respectively (Fig. 1C, Fig. S3, S4, and Table S1). The overall structure of the channel is a tetrameric arrangement, with each monomer conforming to the classical fold of the inwardly rectifying potassium channels. A transmembrane domain consisting of two transmembrane helices, M1 and M2, that are linked from the outer M1 helix through the extracellular turret loop regions before descending to the shorter pore-lining helix that then extends a loop region through the selectivity filter 119TIGYGYT124 back to the extracellular side to connect to the inner M2 helix. This arrangement extends both N and C termini to the cytoplasm, with both termini collectively forming the cytosolic domain. The C-terminal domain is composed of beta-sheets arranged in an Ig-like domain fold. The N-terminal peptide region forms a hook-like structure that folds onto the Ig-like domain and extends to the slide helix at the interface of the transmembrane and cytosolic domains. A single disulfide bond is present in each subunit at the extracellular turret region from Cys99 near the pore helix to Cys131 at the apex of the M2 helix. No density was present for the very C-terminal regions of the protein from residues 326-360 presumably due to intrinsic disorder. In addition, regions of the N-terminal region connecting the cytoplasmic domain to the slide helix possessed insufficient density to be accurately modeled. In the ML418 bound structure residues in the C-linker between transmembrane and cytosolic domains were also missing density. The R162Q mutant of Kir7.1, in its apo form, adopts a docked conformation (Fig. 1C) like that of the PIP2-bound structures observed in previous cryo-EM and X-ray crystallography studies of Kir channels13,14,21,23. This conformation is consistent with an open, conducting state of the channel. When mutated to glutamine, residue 162 (Fig. S5A) coordinates with the Met49 (a residue unique to Kir7.1 at this position) backbone in the slide helix near the base of the outer M1 transmembrane helix within the same subunit (Fig. S5A–C). This glutamine residue in the R162Q mutant maintains the helicity of the interdomain tether, thus docking the cytosolic domain to the transmembrane region (Fig. 1C, and Fig. S5A) through restricting the interdomain distance. Three strong oblate lipid densities are observed, two at the inter-subunit interface in the transmembrane domain, bridging the TM1 and TM2 helices at the external face near the selectivity filter in one subunit to the slide helix and base of the TM1 helix in another subunit (Fig. S6A). A third density is observed near the extracellular surface between an inter-subunit TM1 and TM2 interface near the pore helix. We speculate that these densities may be assigned to cholesteryl hemisuccinate (CHS) which is present in the detergent solubilization and purification buffers. CHS, a derivative of cholesterol that retains the basic steroid nucleus structure but with the hydroxyl group at the 3-beta position replaced by a hemisuccinate group, has been observed in cryo-EM structures of GIRK221. While the addition of this group increases the water solubility of the molecule, the longer hemisuccinate side-group is also available to participate in additional interactions. The lipid densities map to a region that is consistent with that identified from molecular docking studies24,25 yet are at positions distinct from that observed for GIRK2, channels that are uniquely upregulated by cholesterol21. The primary lipid density extends between Leu144 and Phe71 at an inter-subunit TM1-TM2 interface to Arg42 in the slide helix of an adjacent subunit. The second lipid density stacks alongside the first density and extends from Leu74 in TM1 to Arg52 and Arg54 at the base of TM1 from an adjacent subunit. The third lipid density near the extracellular surface extends from Thr107 in the pore helix and Ala139 in the adjacent subunit TM2 to Ser61 in TM1. Superposition of apo-R162Q with PIP2 bound Kir2.2 (3SP1) reveals that the lipid densities near the intracellular interface are closely positioned near the acyl chains of C8-PIP2 (Fig. S6D, E) and potentially overlap binding sites for the extended acyl chains. The third lipid density overlaps the terminal region of the extended acyl chain of long-chain PIP2 bound to Kir6.2 (Fig. S6C). It is possible that the presence of three molecules of CHS assists in stabilizing the docked open conformation of the mutant. When ML418 is present, the predominant class of R162Q Kir7.1 particles adopts an extended conformation in which the cytosolic domain is displaced from the transmembrane region by ~6 Å (Fig. 1C–D) and rotated by ~39° around the central axis of the channel (Fig. 1F). More subtle conformational rearrangements are observed within the transmembrane region (Fig. 1E). In addition to this major extended class, heterogeneous refinement revealed a smaller population (~26% of particles) in an intermediate state between the docked and extended architectures that nevertheless displays clear density for ML418 within the central cavity (Fig. S4C), but this less populated class did not refine to model-building resolution, indicating residual conformational heterogeneity within the class. These ML418-bound conformations resemble the extended conformations observed in cryo-EM structures of other Kir channels21,23,26. This suggests that ML418 binding induces a conformational change that may be associated with channel closure or inactivation (Fig. 1C, I). One of the greatest differences between apo- and ML418-bound R162Q is seen at the selectivity filter which is unoccupied in the ML418 bound structure (Fig. 1G, H). There is a lateral outward splaying of the TM1 and TM2 helices in the apo-R162Q mutant open conformer relative to the ML418-bound form, allowing for altered interdomain contacts and widening of the pore at the helix bundle crossing (Fig. 1I, J, figure S5D). A constriction at the helix bundle crossing is observed in the ML418 bound structure that is absent in the apo R162Q mutant (~4.2 Å vs ~7.2 Å diameter; Fig. 1I–K). In the apo R162Q structure, Val157 is drawn away from the pore cavity, while in the ML418 bound structure, it occupies the pore (Fig. S5H–J). In the ML418 bound structure, a density corresponding to a cholesterol group is observed at a position that appears intermediate between the two cholesterol molecules observed in the apo R162Q mutant structure (Fig. S6B, E, F). Residual lipid density is observed extending from Leu144 and Phe151 in TM2 and towards Trp53 at the base of TM1 in the opposing subunit, oriented to be slightly above the canonical PIP2 site. Interestingly, when the channel is bound by ML418, we observe density consistent with an altered position of Trp45 from the slide helix. Rather than maintaining its typical network of contacts with the adjacent inner TM2 helices, Trp45 is instead oriented to interact with Trp53 at the base of the TM1 helix in a neighboring subunit (Fig. S5E–G, S11). These interactions may preserve the blocked conformation by both stabilizing the outer TM1 helices and sterically occluding the PIP2 binding site. The third lipid density near the extracellular interface remains unchanged (Fig. S6B). Prior molecular dynamics studies have suggested rearrangement of the distribution of bound cholesterol dependent on the conformation of Kir channel27. Functional characterization of the Kir7.1 R162Q mutant Because our functional studies have relied on both automated and manual patch-clamp recordings, we first validated that Kir7.1 behaves similarly across these platforms. The automated patch-clamp (APC) system uses a K⁺-based internal solution containing 110 mM KF and 10 mM EGTA. While these conditions promote high-resistance seal formation, they could potentially influence phosphoinositide-dependent regulation of the channels. To address this, we directly compared the properties of Kir7.1 measured under APC conditions with those previously obtained using manual whole-cell recordings with a KCl-based internal solution (140 mM KCl)7,28. The macroscopic biophysical properties of Kir7.1 were indistinguishable across both platforms. In each recording setup, Kir7.1 produced significant inward currents at hyperpolarized potentials, with a maximal inward current at −180 mV and a reversal potential (Vrev) near −60 mV. The mean current–voltage (I–V) relationships were closely aligned between the KF-based (APC) and KCl-based (manual) internal solutions, indicating that the characteristic inward rectification of Kir7.1 was maintained and not influenced by intracellular fluoride. As described in detail below, the functional coupling between MC4R and Kir7.1 remained consistent across the different recording conditions. The magnitude and direction of responses of Kir7.1 to the MC4R agonist α-MSH and the inverse agonist AgRP, measured using the APC platform, were comparable to those previously reported in manual whole-cell recordings7,28. Collectively, these findings demonstrate that any potential confounding effects due to fluoride-dependent activation of phospholipases do not measurably alter PIP₂-dependent Kir7.1 activity or MC4R coupling under our recording conditions. Cryo-EM analysis of the Kir7.1 R162Q mutant in its apo state revealed that the channel adopts a docked conformation, suggesting that the R162Q mutation intrinsically stabilizes the open conformation of Kir7.1. To functionally assess the impact of this structural change, we expressed either wild-type (WT) or R162Q Kir7.1 channels in HEK293 cells and performed whole-cell patch-clamp recordings. Cells expressing the Kir7.1 R162Q mutant displayed significantly increased peak currents compared to those expressing the WT channel (Fig. S7A). This increase in conductance is consistent with the open conformation observed structurally, indicating a gain-of-function phenotype. Although R162Q resides at the PIP2-binding site, our KF-based whole-cell recordings controls do not implicate differential PIP2 depletion as the cause of its increased conductance relative to WT. However, we cannot exclude enhanced PIP2 sensitivity, which will require direct testing. To further evaluate the functional consequences of the R162Q mutation in Kir7.1, we analyzed current-voltage (I–V) relationships. The I–V curves showed that cells expressing Kir7.1 R162Q exhibited markedly larger whole-cell current densities compared to WT at both hyperpolarizing and depolarizing potentials (Fig. S7B, C). When normalized to cell capacitance (nA/nF), the elevated current densities suggest that the increased current is not due to differences in cell size or surface area but rather reflects a true gain-of-function effect on the channel’s biophysical properties. In addition to the changes in current density, we observed a significant shift in the reversal potential (Vrev) for the R162Q mutant. Under our recording conditions, the apparent Vrev of R162Q lies close to the predicted Nernst potential for K⁺, whereas WT exhibits a less negative Vrev, consistent with lower apparent K+ selectivity. The R162Q mutation produces a ~ 20 mV negative shift in apparent Vrev relative to WT, despite both channels being recorded under identical ionic conditions. Because cryo-EM structures indicate that R162Q Kir7.1 preserves the canonical selectivity filter geometry with well-defined K⁺ binding sites, a true change in the underlying K⁺ equilibrium potential is unlikely. Instead, the very low macroscopic conductance of WT Kir7.1 makes its I–V zero crossing particularly vulnerable to distortion by small non-Kir leak currents and series-resistance errors, whereas the larger R162Q currents are less affected. We therefore interpret the more negative Vrev of R162Q as a more faithful approximation of the true K+ reversal potential, with the WT value representing a leak-biased underestimate rather than evidence that R162Q alters ion selectivity. Consistent with this view, although an early overexpression study suggested that R162W Kir7.1 forms a nonselective cation channel, subsequent work using untagged human R162W established that it behaves as a dominant-negative loss-of-function mutant that abolishes Kir7.1 K⁺ currents without changing ion selectivity3,29,30. In contrast to the apo-R162Q structure, the cryo-EM structure of the R162Q mutant in complex with ML418 revealed a distinct, extended conformation of the cytoplasmic domain, suggesting a closed or inactivated state. To quantify the impact of ML418 on Kir7.1 R162Q activity, we measured the kinetics and extent of ML418-mediated inhibition in HEK293 cells expressing either WT or R162Q Kir7.1 (Fig. S8). In WT channels, ML418 induced a rapid and robust block, with a time constant (τ) of approximately 37 seconds (Fig. S8A), achieving around 60-70% inhibition of whole-cell potassium currents (Fig. S8B–D). In contrast, the R162Q mutant exhibited significantly slower blocking kinetics, with τ ~ 132 seconds (Fig. S8A), and a reduced maximal inhibition of around 30-40% (Fig. S8B–D). These data suggest that the R162Q mutation reduces the functional efficacy of ML418 as a pore blocker, despite the channel’s ability to adopt a closed-like conformation in the presence of the compound. The prolonged blocking kinetics and incomplete inhibition support a scenario in which the mutation shifts the conformational equilibrium toward a more open or semi-open state, weakening the structural transitions required for complete pore occlusion by ML418. Given MC4R’s role in regulating Kir7.1 activity through the neuropeptides involved in energy balance, α-MSH and AgRP, we determined whether the R162Q mutation alters the channel’s sensitivity to endogenous ligands (α-MSH and AgRP) that mediate MC4R-dependent modulation (Fig. S9A–D). We expressed either wild-type (WT) or R162Q Kir7.1 channels in HEK293 cells, co-transfected with the melanocortin-4 receptor (MC4R), and conducted whole-cell patch-clamp recordings. The application of α-MSH, an agonist known to inhibit Kir7.1 through coupling to MC4R, resulted in channel closure in both WT and R162Q-expressing cells. However, the IC50 for α-MSH was slightly right-shifted in the R162Q mutant, indicating that higher concentrations were required to achieve equivalent channel inhibition (Fig. S9B–D). Conversely, the MC4R inverse agonist AgRP, which promotes Kir7.1 opening via direct coupling to MC4R, showed a significantly left-shifted EC50 in the R162Q mutant compared to WT (Fig. S9B–D). This indicates that the mutant channel is more sensitive to AgRP, requiring lower concentrations of AgRP to further enhance channel activity. Together, these structural and functional data provide a cohesive picture of how the R162Q mutation alters Kir7.1 gating and MC4R-dependent modulation. Channel blocker ML418 resides in the central pore region beneath the selectivity filter Pharmacological studies show that the compound ML418 is a blocker of the Kir7.1 channel19. In the presence of ML418, Kir7.1 was observed to assume an extended conformation (Fig. 1C). At reconstruction resolution obtained ~3.9 Å (Fig. S4), we were able to model the protein into the density and make predictions about interactions with the channel blocker and the residues involved in conformational changes. A strong density is observed in the central pore region (Fig. 2A–F), indicating occupancy by ML418. ML418 was initially placed in the central pore region using a top scoring pose generated from AutoDock Vina that was consistent with the EM density and subsequently refined in Phenix. The refined binding pose is located centrally between Thr153 and Glu149 and extends towards the S4 site of the selectivity filter (Fig. 2B). ML418 extends horizontally across the pore with potential hydrophilic interactions Glu149 and Thr153 in diametrically opposed subunits and hydrophobic interactions with the isopropyl and quinoline rings through Leu146 in alternate subunits (Fig. 2B, C). Though the binding pose is consistent with the observed density, there is unaccounted for residual density that can be attributed to symmetry-induced artifacts (Fig. 2D). Strong continuous density from Glu149 in the M2 helix is observed, potentially due to the coordination of the chloride group from the quinoline ring (Fig. 2E, F). This chloride group is located on the same plane as the strontium ion observed in the cryo-EM structure of Kir2.126. Previous mutagenesis studies suggest two pore-lining residues are important for ML418 interaction19. Mutation of Glu149 to glutamine (E149Q) significantly reduced sensitivity to the inhibitor, shifting the IC50 from ~2 μM to ~18 μM and mutation of Ala150 to serine (A150S) also led to a partial loss of inhibition, shifting the IC50 to ~7 μM. Simultaneous mutation of both residues led to an additive loss of inhibitor sensitivity. While Ala150 was not identified as a potential interaction partner in the current study, it does make important contact at gating residues at the helix bundle crossing (discussed below). Location of ML418 binding site. A Cartoon representation of ML418 bound R162Q mutant of Kir7.1 displaying density observed in the central region of pore cavity. B View across the transmembrane region showing the ML418 binding pose and nearby interacting residues beneath the selectivity filter. C Simplified view of ML418 surrounded by residues within interaction distance. D Chemical structure representation of ML418. E ML418 fitted to density (displayed as mesh) observed in cross-section of central pore region. F Density displayed as mesh surrounding ML418 and Glu149 in side view. Conformational changes induced by ML418 binding to R162Q open conformer In the R162Q open conformer, Glu149 at the rectification controller position and Leu146 are in van der Waals contact, maintaining a central cavity network that couples to the selectivity filter and helix-bundle-crossing gates. When ML418 binds within the central cavity and directly engages Glu149 and Leu146, this contact is disrupted and the associated interaction network is reorganized (Fig. S10A–C). Specifically, in R162Q, Gln117 forms an H-bonded bridge linking the backbone carbonyl of Gly145 in the pore helix to the side chain of Glu149 in TM2, and Trp68 in TM1. Additionally, Leu146 maintains contact with Leu118 at the base of the pore helix. ML418 binding breaks the Gly145-Gln117-Glu149 H-bonded network and displaces Leu146, thus disrupting coordination between the central cavity and the selectivity filter region (Fig. S10B–C). These disruptions produce observable differences in backbone carbonyl spacing at the base of the selectivity filter (Fig. 1G, H). At the intracellular end of the transmembrane bundle, the effect of ML418 is equally pronounced. In the R162Q open conformer, Ala150 from TM2 in an opposing subunit interacts with Val157, the TM2 pore-lining residue at the helix-bundle crossing, and this inter-subunit contact pulls Val157 outward to widen the inner pore (Fig. S10B, S5). Upon ML418 binding and engagement of Thr153, this contact is lost and inter-subunit TM2 interactions shift to occur between Phe156 and Gly154, allowing the helix-bundle-crossing to constrict (Fig. S10C). At the slide-helix/C-linker base in apo-R162Q, Trp45 from the slide helix contacts Ile160 at the TM2 base (Fig. S11), forming an inter-subunit tether that supports the open conformer. In the ML418-bound structure, Trp45 adopts an alternate conformer and instead contacts Trp53 at the base of TM1 in an adjacent subunit (Fig. S11F–H). This switch breaks the original slide-helix–TM2 interactions and weakens the cytoplasmic tether, contributing to transmembrane bundle realignment and occlusion of the PIP2 site. In GIRK1, three conserved aromatic residues in TM1, Phe87, Tyr91, and Trp95, have been identified from molecular dynamics studies to adopt a configuration, termed the hydrophobic wire, that allows their side chains to stack along the helical axis forming a contiguous aromatic network that stabilizes the open conformation31. We observe similar stacking of the equivalent residues in Kir7.1; Phe60, Phe64 and Trp68 from TM1 and extending to Phe156 in TM2. The coupling between the central cavity and transmembrane helix motion is centered on this TM1 hydrophobic wire, particularly Phe60. In R162Q, Phe60 maintains proximity to Met147 in TM2 (Fig. S10E), stabilizing contacts that couple TM1 to TM2 in an opposing subunit and support the open pore. When ML418 binds, Phe60 rotates away from Met147 and instead contacts Met56 within the same TM1 helix (Fig. S10D–F), disrupting the TM1–TM2 inter-subunit coupling and favoring intrahelical interactions that compress the transmembrane bundle. On the external side of the pore near Phe60, the TM1–TM2 contact network also undergoes substantial reorganization (Fig. S11B, C). In the R162Q open conformer, Phe60 stabilizes a path of hydrophobic interactions between Met147 and Phe156, while Met57 links Phe156 and Met147 toward the TM2 base. In the ML418-bound state, this network is replaced by a different configuration in which Leu143 positions Trp68 and Met147 to interact with hydrophobic residues on TM1, creating a more tightly coupled upper-helix network not present in the open conformer (Fig. S11B, C). On the extracellular face of the pore, additional rearrangements occur. In R162Q, Ile135 (TM2) bridges Ser111 and Phe110 from the adjacent pore helix. Upon ML418 binding, contact with Phe110 is lost, weakening TM2 inter-subunit contact (Fig. S10B, C). Furthermore, Phe60 reorientation positions it near Ile152, the Kir7.1 homolog of the cholesterol-sensitive M182 in GIRK432. The gating mechanism converges at two conserved glycine hinge points: Gly145 at the upper hinge position near the selectivity filter, and Gly154 at the lower hinge position at the helix-bundle-crossing level. These hinges act as flexible pivots through which the network rearrangements that occur on ML418 binding; disruption of the central cavity Gln117–Trp68–Glu149 bridge, rotation of the Phe60 hydrophobic wire node, reconfiguration of external inter-subunit TM1:TM2 and pore-helix:TM2 contacts, and breaking of the slide-helix–TM2 tether, are communicated to modulate the spacing and orientation of the transmembrane helices and control pore permeability. ML418 Inhibits Kir7.1 in vivo to inhibit food intake and reduce body weight We next used the Kir7.1 blocker ML418 to test the in vivo activities of Kir7.1 in the regulation of feeding and body weight. We administered ML418 (at 5 or 7.5 mg/kg) or vehicle intraperitoneally (ip) to wild-type mice. ML418 decreased 24-h food intake and induced 24-h weight loss in a dose-dependent manner (Fig. 3A, B), without impacting water intake (Fig. 3C). ML418 also has a high affinity to SUR1/Kir6.2 channels (KATP)19. To test if the anorectic activity of ML418 was due to KATP channel inhibition, we administered A-4166 (nateglinide) (5 mg/kg), a potent Kir6.2 inhibitor33, to wild-type mice. A-4166 did not affect acute feeding in mice (Fig. 3D). Kir7.1 is widely expressed, and global deletion of Kir7.1 leads to postnatal death due to improper development of the lungs and palate34. Thus, to avoid any potential peripheral toxicity, we also tested whether central administration of ML418 affected feeding and body weight. Lean cannulated mice received ML418 or vehicle. MC4R also shows some ability to couple to Kir4.1 in cell culture6. We found that ML418 delivered centrally significantly reduced food intake (Fig. 3E), whereas a Kir4.1 inhibitor, VU013499235, did not affect food intake (Fig. 3F). Next, we tested the potential of ML418 to decrease food intake in DIO mice with increased adiposity. Single central administration of ML418 induced a sustained reduction in both food intake and weight over 5 days (Fig. 3G, H). Rapid rebound is observed, as is the case for most drug-induced weight loss. Inhibition of Kir7.1 decreases food intake and promotes weight loss independent of adiposity. A Food intake of wild-type C57BL/6 J mice over a 24-h period treated with 13% DMSO in saline (vehicle control plotted as 0 mg/kg) or ML418 (5 mg/kg or 7.5 mg/kg); 0 mg/kg n = 13; ML418 5 mg/kg n = 6; ML418 7.5 mg/kg n = 7, one-way ANOVA, F (2, 23) = 21.87, p = 0.0001. B 24-h change in body weight in wild-type mice treated with 13% DMSO in saline (vehicle control plotted as 0 mg/kg) or ML418 (5 mg/kg or 7.5 mg/kg); 0 mg/kg n = 13; ML418 5 mg/kg n = 6; ML418 7.5 mg/kg n = 7, one-way ANOVA, F (2, 22) = 9.988, p = 0.008. C Water intake of wild-type mice over 24-h period treated with vehicle (13% DMSO in saline) or ML418 (7.5 mg/kg), n = 5, Student’s t-test (2-tailed), t(8) = 1.361, p = 0.2106. D Food intake in wild-type mice treated with A-4166 (5 mg/kg; A-4166 results are mostly obscured by vehicle values) or vehicle (13% DMSO in saline), n = 5, two-way ANOVA, F (4, 40) = 0.06315, p = 0.9924, multiple comparisons corrected using the Tukey–Kramer method. E Acute food intake over 24-h in wild-type mice centrally treated with vehicle (1μL DMSO) or ML418 (500picomoles, 1μL), n = 9, two-way ANOVA, F (6, 98) = 2.841, p = 0.0136. F Acute food intake in wild-type mice centrally treated with vehicle (1μL of DMSO) or VU0134992 (500pmoles/1μL), n = 5, F (5, 38) = 0.07213, p = 0.9960. G Food intake of DIO wildtype mice receiving a single icv injection of vehicle (1μL of DMSO) or ML418 (500pmoles, 1μL), n = 9, two-way ANOVA, F (4, 80) = 0.6236, p = 0.6470. H Change in body weight of DIO mice receiving single icv injection of vehicle or ML418, n = 9, two-way ANOVA, F (4, 80) = 3.933, p = 0.0058. ANOVA tests were corrected for multiple comparisons using the Tukey–Kramer method. Mean ± SD are shown. *= p < 0.05, ** = p < 0.01, *** = p < 0.001,**** = p < 0.0001. ML418 activates PVH MC4R neurons, and Kir7.1 in MC4R neurons is required for its effects on food intake Building upon our observation that inhibition of Kir7.1 by channel blocker ML418 reduces food intake and promotes weight loss, we next sought to determine whether the anorexic effects of ML418 are mediated through Kir7.1 in MC4R neurons. Based on our earlier observations that α-melanocyte-stimulating hormone (α-MSH) depolarizes PVH MC4R neurons through closure of Kir7.1 channel in an electrophysiological slice preparation6, we investigated the action of ML418 on the activity of PVH MC4R neurons in the basal and α-MSH treated state. Using whole-cell patch recordings from current clamped PVH MC4R neurons from hypothalamic slices derived from MC4R-GFP mice, we reproduced the finding that bath applied α-MSH reliably activates PVH MC4R neurons. This activation was associated with depolarization of membrane potential (~9.5 mV) and increases in spontaneous action potential firing frequency (Fig. 4A–C). We next examined whether bath applications of ML418 can mimic the excitatory effects of α-MSH on these neurons. Treatment with ML418 also depolarized the baseline membrane potential (~10.3 mV) and increased the action potential firing frequency in the MC4R neurons (Fig. 4D–F). To further investigate that the effects of ML418 are caused by closure of channels and not synaptically driven, we examined the effects of this drug on the MC4R neurons’ input resistance by comparing their responses to applications of consistent current pulses in vehicle and in ML418. Compared to vehicle, ML418 significantly augments the magnitude of depolarization and consequently the number of action potentials evoked (from 2.7 to 4.3 spikes), suggesting closure of channels by this drug (Fig. 4G, H). Furthermore, pretreatment with ML418 markedly blocked the α-MSH-induced depolarization and action potential firing of PVH MC4R neurons (Fig. 4I–K). ML418 activates and depolarizes PVH MC4R neurons and blocks further depolarization by a-MSH. A–C Representative voltage traces recorded in current clamp mode (A) demonstrating the effects of α-MSH (500 nM) on the membrane potential (B, from −75 + /−2.8 mV to −65.4 + /−2.8 mV, p = 1.1E-07) and action potential firing frequency (C, from 0.3 + /− 0.2 Hz to 1.9 + /− 0.6 Hz, p = 0.0079) of PVH MC4R neurons (two tailed paired t test, n = 10 cells). D–F Representative voltage trace (D) demonstrating the effects of bath application of ML418 10 µM compared to baseline vehicle on the membrane potential (E, from −68.9 + /− 1.5 mV to −58.7 + /− 2.3 mV, p = 0.004) and action potential firing frequency (F, from 0.05 + /− 0.02 Hz to 0.25 + /− 0.07 Hz, p = 0.011) of PVH MC4R neurons (two tailed paired t test, n = 9 cells). G, H Representative voltage (upper panel) and current (lower panel) traces (G) demonstrating the effects of ML418 10 µM treatment on the number of action potential firings (H, from 2.7 to 4.3 in 100 ms, p = 0.0002) in response to consistent current pulses in baseline vehicle and in ML418 in PVH MC4R neurons (two tailed paired t test, n = 8 cells). I–K A representative voltage trace (I) demonstrating the effects of α-MSH (500 nM) on the membrane potential (J, from −68.5 + /− 2.8 mV to −68.3 + /− 3.2 mV, p = 0.8) and action potential firing frequency (K, from 0.2 + /− 0.08 Hz to 0.12 + /− 0.04 Hz, p = 0.73) of PVH MC4R neurons after incubating slices with ML418 (10 µM), (two tailed paired t test n = 9 cells). *= p < 0.05, ** = p < 0.01, *** = p < 0.001,**** = p < 0.0001. Using c-Fos as a marker of neuronal activation, we found a significant increase in c-Fos expression in the PVH upon intraperitoneal ML418 administration (Fig. 5A–C). ~70% of MC4R neurons were c-Fos positive after ML418 treatment, compared to fewer than 20% following vehicle treatment, indicating that inhibiting Kir7.1 activates the majority of MC4R neurons, along with a number of other unidentified PVH neurons. As expected, the injection of setmelanotide (Imcivree)36, an MC4R agonist, into the PVH led to decreased food intake and weight loss (Fig. 5D, E). Direct administration of ML418 into the PVH of wild-type mice (Fig. 5) similarly resulted in reduced food intake and weight loss within 24 h (Fig. 5F, G). To determine if Kir7.1 in MC4R neurons plays a role in inhibition of feeding and weight loss following ip administration of ML418, we used the Cre-LoxP system to delete Kcnj13, the gene encoding Kir7.1, from MC4R neurons. As expected, control Mc4r-Cre mice exhibited a significant reduction in food intake and body weight following ip ML418 injection (Fig. 5H, I). Conversely, mice with selective deletion of Kcnj13 from Mc4r-expressing cells had no significant changes in food intake following ip ML418 injection and exhibited less weight loss relative to ML418 treated control Mc4r-Cre animals (Fig. 5J, K). The remaining weight loss observed in ML418-treated MC4RCre:Kcnjflox/flox mice may well be due to the other sites of Kir7.1 expression, suggested by the c-Fos response to ML418 in many non-MC4R neurons (Fig. 5C). Interestingly, no discernible effect on acute food intake or body weight was observed upon tissue-specific deletion of Kcnj13 from MC4R-expressing cells, suggesting potential compensatory developmental mechanisms. ML418 activates PVH MC4R neurons and inhibits food intake, and deletion of Kcnj13 from Mc4r-expressing cells blunts ML418 action. A Representative image of coronal sections of the paraventricular hypothalamus (PVH) from MC4R-GFP mice, systemically stimulated with vehicle (DMSO) or ML418. B Quantitation of cells expressing c-Fos in the PVH after animals were treated with vehicle or ML418, n = 5, Student’s t-test (2-tailed), t(8) = 13.26, p = 0.0001. C Percent of c-Fos-positive cells are MC4R cells, Student’s t-test (2-tailed), n = 5, t(8) = 6.6109, p = 0.0002. D 24 hr food intake of Intra-PVH cannulated wild-type mice receiving central injection of vehicle (0.5μL/side, Ringer’s solution, n = 7), or setmelanotide (250pmoles/0.5μL/side, n = 9), Two-way ANOVA: F(5,84) = 2.258, p = 0.0559; multiple comparisons corrected using the Tukey–Kramer method. E 24-h body weight change of the same mice (n = 7 (vehicle) or n = 9 (setmelanotide)), Student’s t-test (2-tailed), t(15) = 2.908, p = 0.0108. F Food intake of Intra-PVH cannulated wild-type mice receiving central injection of vehicle (0.5μL/side, Ringer’s solution) or ML418 (250pmoles/0.5μL/side), n = 8 Two-way ANOVA: F (5, 84) = 3.09, p = 0.0131, multiple comparisons corrected using the Tukey–Kramer method. G 24-h body weight change of the same mice, Student’s t-test (2-tailed), t(15) = 3.807, p = 0.0019. H Food intake over 24-h period of Mc4r-Cre mice treated with vehicle (13% DMSO in saline, i.p.) or ML418 (5 mg/kg, ip), n = 9, F (4, 80) = 2.212, p = 0.0751 and I change in 24-h body weight in the same mice, Student’s t-test (2-tailed), t(16) = 3.402, p = 0.0036. J Food intake over 24-h period of Mc4r-Cre::Kcnj13flox/flox mice treated with vehicle (13% DMSO in saline, i.p.) or ML418 (5 mg/kg, ip), n = 9, F (4, 90) = 0.4581, p = 0.7662 and K change in 24-h body weight in the same mice, Student’s t-test (2-tailed), t(18) = 2.286, p = 0.0346. ANOVA were corrected for multiple comparisons using the Tukey–Kramer method. Mean ± SD are shown. *= p < 0.05, ** = p < 0.01, *** = p < 0.001,**** = p < 0.0001. MC4R and Kir7.1 physically interact in living cells Prior electrophysiological6 and in vivo data8 strongly support the role of Kir7.1 in the regulation of feeding behavior and body weight mediated by α-MSH and AgRP through the MC4R. Studies also show that Kir7.1 conductance is not modulated by MSH or AgRP in the absence of a functional MC4 receptor6,7. Evidence also indicates that G protein signaling is not required for the MC4R-mediated regulation of Kir7.16,7. To determine whether Kir7.1 and MC4R physically interact at the plasma membrane, we performed a NanoBiT complementation assay (Fig. S12A). In this assay, HEK293T cells co-expressing Kir7.1–N-terminus-SmBiT (Kir7.1-N-SB) and MC4R-C-terminus-LgBiT (MC4R-C-LB) showed a strong luminescence signal, with a 105.8 ± 3.06-fold change (n = 15, SEM). This suggests that the channel and receptor are close enough to reassemble NanoLuc luciferase activity. The signal was significantly higher than in other conditions where the NanoBiT fusion fragments were attached either to the N- or C-terminus of the channels and/or the receptor. Specifically, the results were: - Kir7.1-N-LgBiT + MC4R-C-SmBiT: 11.02 ± 0.42-fold change (n = 15) - Kir7.1-C-SmBiT + MC4R-C-LgBiT: 1.91 ± 0.06-fold change (n = 15) - Kir7.1-C-LgBiT + MC4R-C-SmBiT: 3.46 ± 0.09-fold change (n = 15) (Fig. S12A). These results indicate that MC4R and Kir7.1 are in close proximity, which could reflect a direct interaction or the formation of a stable complex at the plasma membrane. Notably, there appears to be a preference for interaction between the N-terminus of Kir7.1 and the C-terminus of MC4R, consistent with the known intracellular location of these domains. To directly test whether MC4R and Kir7.1 form a functional signaling complex, we created tethered receptor-channel fusion proteins (Fig. 6). Full-length MC4R was fused to Kir7.1 using various flexible linkers (L1–L3: 20, 30, or 40 amino acids) or with no linker at all (L0) (Fig. 6A). All constructs were expressed in HEK293 cells and tested for canonical Gαs signaling with a cAMP biosensor, as well as for Kir7.1 channel activation through electrophysiology. Fusion constructs with short or no linkers (L0–L2) reduced maximal Gαs signaling (lower EMax) but maintained near-normal ligand sensitivity (α-MSH EC50) (Fig. 6B-D). Notably, the L3 construct (40-aa linker) showed full agonist efficacy and potency relative to wild-type MC4R, indicating that tethering does not block GPCR activation when enough linker length is provided. All tethered constructs could still gate Kir7.1 in a ligand-dependent way, showing strong and concentration-dependent rises in inward K⁺ conductance in response to AgRP (Fig. 6E, Fig. S12B-D). While basal conductance was lower each construct responded to AgRP with nanomolar EC50 values, demonstrating that functional coupling of MC4R to Kir7.1 remains intact in the tethered form. Creation of a linkered MC4r-Kir7.1 protein that retains receptor-mediated channel regulation. A Structure of linkered MC4R-Kir7.1 constructs with no linker or linkers of indicated sequence (L1 through L3). B Intracellular cAMP production concentration-response curves in cells transfected with the GScAMP22f luciferase cAMP genetic sensor and the indicated MC4R or Kir7.1 tethered constructs. The data represent the mean ± S.E.M. from fifteen replicates, with the exception of the MC4R curves based on 12 replicates, distributed in three assay plates. The data were fitted to a four-parameter sigmoid model. C pEC50 values and (D) maximum effect (EMAX) relative to the control MC4R construct transfection obtained from the concentration-response curves shown in (B). The data in (D) represent the mean ± S.E.M of EMAX values determined from each of the three assay plates, Significance (D, E) was determined by one-way ANOVA and the Dunnett post-test. *= p < 0.05, ** = p < 0.01, *** = p < 0.001,**** = p < 0.0001. E Representative current traces of the MC4R-Kir7.1 tethered constructs are shown, evoked from a holding potential of −60 mV to a step pulse of −140 applied for 200 ms. F Concentration response curves showing effects of AgRP on currents from cells expressing the MC4R-Kir7.1 tethered constructs. The data represent the mean ± S.E.M., n = 5 for each data point. Together, these two experimental strategies provide converging evidence that MC4R and Kir7.1 form a physical and functional signaling complex. NanoBiT assays reveal close molecular proximity consistent with stable interaction in living cells, while tethered fusion proteins demonstrate that direct physical coupling does not abolish signaling and can even preserve both G protein- and channel-based outputs when appropriate linker lengths are used. These findings support a model in which MC4R and Kir7.1 exist as a preformed or inducible signaling complex, enabling rapid and ligand-specific modulation of Kir7.1 channel activity by MC4R. Discussion This study provides structural insights into the human Kir7.1 potassium channel and further elucidates the molecular mechanism of its inhibition by the small molecule ML418. A key finding is the identification of two conformational states of the R162Q mutant of Kir7.1 - a docked conformation resembling the open, conducting state, and an extended conformation induced by ML418 binding that likely represents a closed or inactivated state. The R162Q mutation, associated with retinitis pigmentosa, stabilizes the channel in an open conformation and confers a gain-of-function phenotype. This provides a structural explanation for the pathogenic effects of this mutation and highlights the importance of proper Kir7.1 regulation in retinal function. The apparent reversal potential (Vrev) of R162Q is shifted by ~20 mV to more negative values compared with WT Kir7.1. Only mutations that alter the relative permeability of K+ versus competing ions are expected to shift the true K+ reversal potential. Within Kir7.1, structural data indicate a canonical GYG selectivity filter with relaxed outer-filter constraints at Met125–Glu115 that permit Cs⁺ permeation while preserving overall K⁺ selectivity, and our cryo-EM analysis shows that R162Q maintains this filter geometry and K⁺ coordination sites, arguing against a genuine change in EK. Instead, we favor a mechanism in which R162Q modifies inward rectification through long-range coupling between the filter and the cytoplasmic vestibule that governs Mg²⁺/polyamine block, thereby shifting the apparent I–V zero crossing to more negative values without altering the underlying K⁺ gradient. The very low conductance of WT Kir7.1 further makes its Vrev especially vulnerable to distortion by small leak and series-resistance errors, whereas the larger R162Q currents likely provide a closer approximation to the true K⁺ reversal potential. Systematically varying intra- and extracellular ion concentrations and directly probing Mg²⁺/polyamine dependence will be important future steps to dissect these contributions in more detail. In our initial trials, we observed a high degree of conformational heterogeneity in preparations of the wild-type unliganded channel, suggesting that Kir7.1 is intrinsically dynamic and oscillates between intermediate conducting and non-conducting states. Additionally, we noted low occupancy by ML418 in wild-type preparations. Thus, use of the open-conformer-stabilized R162Q mutant better enabled determination of the ML418-bound structure of Kir7.1. Our structural analysis of the open-state R162Q mutant revealed lipid densities likely corresponding to CHS, a cholesterol derivative used during protein purification. The positioning of these CHS molecules was nearly identical to that of the acyl chains of PIP2 observed in other Kir structures, indicating that Kir7.1 can accommodate sterol‑like lipids at annular sites adjacent to the canonical PIP2 binding region. In the ML418-bound structure, the channel adopts an extended, non-conducting conformation and we observe associated rearrangements of nearby lipid densities such that the canonical PIP2 site is partially occluded. These observations suggest that pore block is accompanied by changes in both the TM–CTD interface and local lipid organization, although we cannot yet determine whether these lipid movements are causal for channel closure. Interestingly, the putative CHS binding pockets in Kir7.1 differ from those described for GIRK2, a channel that is uniquely upregulated by cholesterol, raising the possibility that endogenous cholesterol or related sterols modulate Kir7.1 via channel-specific sites. The structural and pharmacological evidence that the R162Q variant adopts the open conformation and exhibits significantly increased sensitivity to AgRP-mediated channel opening suggests that this variant may be a useful tool for further studies of the physiological functions of Kir7.1 in the regulation of feeding behavior. Fast-induced food intake is mediated in part by AgRP action on MC4R neurons, and thus introducing the R162Q variant of Kir7.1 into MC4R neurons could be predicted to make animals more responsive to fast-induced refeeding. The structural analysis of ML418 binding reveals a pore-blocking mechanism of inhibition. The compound binds in the central cavity below the selectivity filter, interacting with pore-lining residues L146, E149 and T153. This binding mode is consistent with previous mutagenesis studies and explains the strong dependence of inhibitor potency on these residues. The orientation of ML418 in the pore indicates that it not only occupies the permeation pathway, reducing ion flux, but also perturbs key pore-lining interactions. In addition to this pore occupancy, ML418 binding disrupts residue network interactions that stabilize the open state and thereby promotes transition to the closed conformer. This closed or inactivated state may resemble that of the Mg2+/polyamine blocked state. Pharmacological studies have suggested that Kir7.1 can be directly gated by interaction with the MC4R in a G-protein-independent manner, with the endogenous MC4R agonist α-MSH resulting in channel closure and the endogenous antagonist AgRP opening the channel6,7. However, in vivo data for the physiological coupling of MC4R to Kir7.1 have been lacking. For example, the deletion of Kir7.1 from MC4R neurons produced only modest late-onset effects on body weight, perhaps due to developmental compensation8. The demonstration here that the channel blocker ML418 binds to Kir7.1, confirming previous pharmacological data, led us to test the in vivo consequences of Kir7.1 blockade on feeding behavior and body weight. ML418 treatment reduced the resting membrane potential of MC4R PVH neurons in a slice preparation and blocked further depolarization of the cells by α-MSH. Peripheral treatment of mice with the compound activated many neurons in the PVH, including around 70% of the MC4R neurons, as indicated by induction of c-Fos expression. Peripheral administration or stereotaxic injection of ML418 into the PVH inhibited food intake at magnitudes comparable to the administration of the MC4R agonist setmelanotide, and also induced weight loss, and these effects were dependent on the expression of Kir7.1 in MC4R neurons. Thus, in contrast to tissue specific deletion of Kir7.1 from MC4R neurons8, pharmacological blockade of Kir7.1 in MC4R neurons does indeed appear to mimic the effects of MC4R agonists on the activation of MC4R neurons and the subsequent inhibition of food intake and weight loss37. The demonstration here that MC4R and Kir7.1 are in close proximity in the plasma membrane, and that fusion proteins containing both MC4R and Kir7.1 retain MC4R-mediated regulation of Kir7.1 conductance provides additional evidence for the direct coupling of MC4R to Kir7.1. However, these data do not exclude the possibility that an as yet unknown MC4R signaling intermediate secondarily regulates the activity of Kir7.1. Data also suggests that the MC4R-Kir7.1 complex may be critical to the obesogenic action of certain atypical antipsychotics. Recently, it was found that acute application of the atypical antipsychotic risperidone to murine hypothalamic slices inhibits these neurons in an MC4R-dependent manner38. When Li and colleagues sought to determine the target of action of these drugs, they discovered that both risperidone and clozapine enhanced AgRP induced opening of Kir7.1 by the MC4R. Further, genetic ablation of Kir7.1 from MC4R neurons, or pharmacological blockade of Kir7.1 by ML418, as shown here, blocks clozapine-induced weight gain in mice (see Li et al.38). Finally, the structural data presented here suggests a hypothesis for the coupling of MC4R to Kir7.1. The Kir7.1 channel displays parallels to the voltage-gated channels in the selectivity filter, and it is possible that they may also share some features of the gating mechanism as well. The voltage-gated channels have, in addition to the usual K+ channel pore architecture, a four-helix transmembrane domain known as the voltage sensing domain that is connected to the pore helices with a linker helix that lies parallel to the membrane in a manner similar to the slide helix in the Kir channels. Vertical movement of the S4 helix of the voltage sensing domain tugs on the linker helix and results in the widening of the pore helices and opening of the channel. The coupling of movement in one transmembrane domain to channel gating has also been observed in complexes of the KATP channel with the sulfonylurea receptor39. Movement of transmembrane helices in SUR1 in response to ligand binding and alternate engagement and disengagement of the N-terminal peptide of KATP by SUR1 correspond to transitions between docked and extended conformations of the KATP channel8. Future studies using the MC4R-Kir7.1 fusion protein described here may, for example, use mutagenesis of the Kir7.1 slide helix to test this model. Methods Ethical statement All animal experiments performed in these studies were approved by the University of Michigan Institutional Animal Care and Use Office (protocol PRO00009710), or the Vanderbilt University Medical Center Institutional Animal Care and Use Committee (slice electrophysiological studies, Fig. 4, protocol M/10/358). Expression and purification Full-length wild-type and mutant human Kir7.1 constructs (amino acids 2-360, minus initiating methionine) were cloned into LIC-pFastBac vector encoding an N-terminal GB1 tag and TEV protease cleavage site and expressed in Sf9 insect cells, with point mutations introduced using the Q5 Site-Directed Mutagenesis Kit (New England Biolabs, Ipswich, MA). For preparation of the ML418-bound R162Q mutant, 30 µM ML418 was included during membrane extraction and maintained in all subsequent purification buffers. Cell membranes were solubilized in 50 mM Hepes, pH 7.5, 150 mM KCl, 2% DDM and 0.2% CHS. Supernatant was incubated overnight at 4˚C with IgG sepharose beads. Beads were collected and washed with 50 mM Hepes, pH 7.5, 150 mM KCl, 0.1% DDM and 0.1% CHS. Resin was exchanged into MNG detergent on the column prior to overnight incubation with TEV protease. Unbound material was collected and passed over a TALON affinity column. Unbound fractions were collected, concentrated, and loaded on a Superdex 200 size exclusion column previously equilibrated in 50 mM Hepes, pH7.5, 150 mM KCl, 0.005% MNG and 0.0005% CHS. Peak fractions were concentrated to 6-8 mg/mL. Sample preparation and data collection A volume of 3.5 μL of protein sample at a concentration between 6-8 mg/mL was applied to glow- discharged C-flat 1.2/1.3 400 mesh Au grids, blotted for 3.5 s at room temperature and 100% humidity and plunge-frozen in liquid ethane using a Vitrobot Mark IV (FEI). Micrographs were collected on a Titan Krios G4i microscope operated at 300 kV and equipped with a K3 direct electron detector. Data collection was automated with EPU and SerialEM. Data processing and refinement Motion-corrected and dose-weighting was performed using MotionCor240. All subsequent processing was carried out in CryoSparc41. Defocus was estimated from the motion corrected micrographs using CTFFIND442. Particle picking was first performed using blob picker, and extracted particles were subjected to one round of 2D classification. Selected 2D classes were then used for particle picking using Template picker in CryoSparc. The extracted particles were then passed through multiple rounds of 2D classification and selection. Particles from selected classes were then used to generate ab initio models without applied symmetry. Heterogenous refinement of ab initio classes with C4 symmetry applied was performed, followed by homogenous and non-uniform refinement in Cryosparc. For the apo-WT dataset after heterogenous refinement classes with well-defined transmembrane regions were combined and additional 3D classification in C1 symmetry performed. Classes with transmembrane regions displaying characteristics of the open conformer and refining to below 5 Å were combined, subject to non-uniform refinement. Particle sets were then symmetry expanded and subject to 3D variability analysis. An initial model for a single subunit of R162Q mutant Kir7.1 was generated using AlphaFold and fitted into cryo-EM maps using COOT43. Symmetric copies of the fitted monomer were placed in cryo-EM maps, followed by rounds of real space refinement implemented in Phenix44. For ML418 bound R162Q transmembrane and cytosolic domains were fitted independently before merging in COOT. For apo R162Q the final model included residues 14-325 and excluding residues 31-36 from the N-terminal region due to insufficient density. For ML418 bound structure residues 14-325 and excluding residues from 30-39 and 161-165 were modeled. Autodock VINA45 was employed to generate initial binding poses for ML418 ligand. A top scoring pose consistent with density observed in central pore was then subsequently merged with fitted tetramer in COOT before rounds of real space refinement in Phenix. Generation of linked MC4R-Kir7.1fusion constructs Linked MC4R–Kir7.1 fusion constructs were generated with four linker variants of increasing length. Linker 0 (L0) contained no additional flexible spacer beyond the cloning junction, whereas linkers L1, L2 and L3 incorporated progressively longer glycine–serine–rich linkers derived from (Gly–Ser)4–Ala–Ala–(Gly–Ser)5–Gly–Ala–Ala (see Supplementary for sequences), each flanked by EcoRI and XhoI sites for modular exchange. A TEV protease recognition sequence was inserted between the linker and the N-terminus of Kir7.1. Constructs were first assembled in the pFastBac vector harboring the MC4R coding sequence inserted between AscI and EcoRI sites. The TEV–Kir7.1 cassette was inserted between EcoRI and HindIII sites. Linker sequences were synthesized as gene fragments in a carrier plasmid (GenScript, Piscataway, NJ) and subsequently subcloned between MC4R and Kir7.1 at the EcoRI/XhoI junction. The complete tagged fusion cassette (MC4R–[linker]–TEV–Kir7.1) was then amplified by PCR from each pFastBac construct and ligated into pcDNA3 between AscI/NotI sites for mammalian expression. All pcDNA3-based fusion constructs were verified by full-insert Sanger sequencing. Automated patch clamp recordings Automated whole-cell patch clamp recordings were performed using the SyncroPatch384PE system (Nanion Technologies, Munich, Germany) as previously described28,46. HEK293T Cells (ATCC CRL-11268) were maintained in Dulbecco’s Modified Eagle’s Medium (Invitrogen), supplemented with 10% fetal bovine serum (Life Technologies, Carlsbad, CA, USA) and 10% Antibiotic-Antimycotic (Gibco). They were cultured at 37 °C in a humidified incubator with 5% CO2. For the electrophysiology experiments, 2.5 × 105 cells were seeded in 100 mm culture dishes and transfected with 4 μg of DNA constructs, including MC4R, wild-type Kir7.1, mutant Kir7.1R162Q, or MC4R-Kir7.1 tethered constructs, using the X-tremeGENE HP DNA transfection reagent (Roche Diagnostics, Basel, Switzerland) according to the manufacturer’s instructions. Recordings were conducted 48 h post-transfection. After incubating the cells with TrypLE (Thermo Fisher Scientific) for 5 min, they were centrifuged at 200 g for 5 min, resuspended in an external recording solution at a density of 400,000 cells/mL, and transferred to the instrument cell hotel, which was maintained at 10 °C with 200 rpm agitation until the start of the experiment. Recordings were conducted in a whole-cell voltage-clamp configuration using 4-hole, 384-well recording chips with medium resistance (2–4 MΩ). The external solution contained 140 mM NaCl, 2 mM CaCl2, 4 mM KCl, 1 mM MgCl2, 5 mM glucose, and 10 mM HEPES (pH 7.4, adjusted with NaOH, 298 mOsm). Cell sealing was achieved using a modified external solution that included 6 mM CaCl2. The internal electrode solution comprised (in mM): KF 110, KCl 10, NaCl 10, HEPES 10, and EGTA 10, with pH adjusted to 7.2 (using KOH, ~280 mOsm). Pulse generation and data collection were carried out with PatchController384 V.3.2.0 software (Nanion Technologies). Parameters such as seal resistance, capacitance, and series resistance were determined for each well after applying a test pulse and monitored throughout the experiment. To evaluate the MC4R-dependent channel closures or activations induced by AgRP or α-MSH, a hyperpolarization step to −140 mV for 200 ms at a holding potential of −60 mV was applied and repeated every 5 seconds. The same voltage protocol was employed to analyze the effect of ML418. The peak inward current amplitude of Kir7.1 during the hyperpolarizing step was measured and normalized to the remnant currents observed after introducing 10 mM BaCl2 (which fully blocks the current). Peptides were delivered at various concentrations across the 4-hole, 384-well recording chip. Responses were further normalized by calculating the ratios of peak currents in the presence or absence of the test peptides (Icompound/I0). Concentration–response profiles were generated to identify the respective EC50 or IC50 values for the activating or inhibiting responses of the peptides. Data analysis and plotting were conducted using DataController384 V.3.2.0 software (Nanion Technologies) and GraphPad Prism V.9.4.0 software. Concentration–response curves were fitted to a four-parameter sigmoid model. NanoBiT protein–protein interaction assay Full-length human Kir 7.1 (KCNJ13, NM_002242.4) and MC4R (NM_005912.3) cDNAs were subcloned into NanoBiT expression vectors (Promega) to create fusion constructs for protein-protein interaction assays. Kir 7.1 was cloned into the pBiT 1.1-C [LgBiT] vector to produce a C-terminal LgBiT fusion (Kir 7.1-C-LgBiT) and into the pBiT 1.1-C [SmBiT] vector for a C-terminal SmBiT fusion (Kir 7.1-C-SmBiT). Additionally, Kir 7.1 was cloned into the pBiT 1.1-N [LgBiT] vector for an N-terminal LgBiT fusion (Kir 7.1-N-LgBiT), and into the pBiT 1.1-N [SmBiT] vector for an N-terminal SmBiT fusion (Kir 7.1-N-SmBiT). MC4R was cloned into the pBiT 2.1-C [SmBiT] vector to generate a C-terminal SmBiT fusion (MC4R-C-SmBiT) and into the pBiT 2.1-C [LgBiT] vector to produce a C-terminal LgBiT fusion (MC4R-C-LgBiT). All constructs were verified by Sanger sequencing. Control plasmids included an empty vector (either SmBiT or LgBiT alone) and NanoBiT positive control proteins provided by Promega. HEK 293 T cells (ATCC CRL-11268) were cultured in Dulbecco’s Modified Eagle Medium (DMEM, Thermo Fisher), supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin-streptomycin, at 37 °C with 5% CO₂. For NanoBiT assays, cells were seeded at 2 × 10⁴ cells per well in white 96-well plates (Corning, opaque flat-bottom) 24 h prior to transfection. Transient transfection was performed using Lipofectamine 3000 (Invitrogen) according to the manufacturer’s instructions. Cells were co-transfected with 100 ng of various combinations of Kir 7.1-N/C-LgBiT/SmBiT and 100 ng of MC4R-C-LgBiT/SmBiT per well. Six h after transfection, the media were replaced with fresh complete medium. Twenty-four h post-transfection, NanoBiT luminescence was measured using the Nano-Glo Live Cell Assay System (Promega, N2011). The substrate was prepared following the manufacturer’s protocol and added directly to the cell culture medium (25 μL per well of Nano-Glo Live Cell substrate diluted in 100 μL Opti-MEM). Luminescence was measured after a 10-min incubation at room temperature using a BMG PHERAStar multi-mode plate reader with an integration time of 0.5 seconds per well. Luminescence values were normalized to background signals from negative control wells transfected with single constructs plus empty vectors. cAMP response assays We used cAMP responses as a surrogate for Gs protein-dependent receptor coupling, as described previously46,47. In brief, to evaluate the cAMP responses of the MC4R and MC4R-Kir7.1 tethered constructs, HEK293GScAMP22F cells stably expressing a split-luciferase PKA regulatory subunit-based sensor48 were seeded at a density of 3 × 106 cells per ten-centimeter cell culture dish. Twenty-four h after seeding, the cells were transfected with 4 μg of each plasmid and Lipofectamine™ 3000 (Thermo Fisher Scientific) transfection reagent at a 1:1.5 plasmid DNA-to-reagent ratio, following the standard protocol. A submaximal concentration of forskolin (2 µM) was included as a normalization reference to account for variations in cell number and differences in assay transducer efficiency among receptor clones. Twenty-four h before measuring the cAMP response, HEK293GScAMP22F cells expressing MC4R or the MC4R-Kir7.1 tethered constructs were plated in poly-D-lysine-coated, clear-bottom, black-wall 384-well plates (Corning Inc., Corning, NY) at a density of 20,000 cells per well in complete growth medium. Two h before the experiment, the growth medium was replaced with CO2 Independent Medium™ (Thermo Fisher Scientific), containing 4% D-Luciferin (Promega). The luciferase substrate permeated the cells for 120 min at 37 °C. Intracellular cAMP levels were measured using an FDSS 7000 EX Functional Drug Screening System (Hamamatsu Photonics, Hamamatsu, Japan). This system allowed for the in-line addition of various α-MSH concentrations while simultaneously recording luminescence from live cells. The measurement rate was set to 0.3 Hz, with a 1/10 neutral density filter and 2 × 2 binning. Assay steps included a 2-min baseline acquisition, followed by the in-line addition of 10 µL of various 3× concentrations of the agonist. Luminescence responses to agonizts were normalized to each intra-plate maximum response of α-MSH. Data were fitted to a four-parameter sigmoid model, with the slope constrained to a range of 0–2, using GraphPad Prism version 10.5. Experimental animal models Experiments were performed in male C57BL/6 J mice (wild-type, WT) purchased from The Jackson Laboratory (JAX), transgenic Mc4r-2a-Cre (JAX strain #: 030759, bred in-house), and transgenic Kcnj13flox/flox mice24 bred in-house. All experimental mice were 8–14 weeks old except for the dietary-induced obese (DIO) mice, which were 30-32 weeks old. All mice were single-housed with a 12-h light/12-h dark cycle and fed either regular chow (5LOD from Lab Diet) or high-fat diet (60 kcal% fat, Research Diets, Cat#: D12492). Mice had free access to food and water unless specified. DIO models were developed by exposing mice to a high-fat diet at 4–5 weeks of age for 26 weeks. Data from individual mice are shown in the figures. Drugs and reagents ML41819, A-4166 and VU0134992 (Tocris) were dissolved in sterile DMSO and subsequently in USP-grade sterile saline. Setmelanotide (Vivitide) was dissolved in sterile saline, and working concentrations were made on the day of the experiment using Ringer’s solution in mM: 123 NaCl, 1.5 CaCl2 5 KCl, pH 7.3- 7.4. Slice electrophysiology Acute hypothalamic brain slices were prepared similarly to previously described6. Briefly, adult male and female MC4R-GFP49 mice were deeply anesthetized with isoflurane and decapitated. The brain was entirely removed and submerged in carbogen (95% O2 and 5% CO2) saturated with ice-cold artificial cerebrospinal fluid (aCSF) solution containing (in mM) 126.2 NaCl, 3.1 KCl, 2 CaCl2, 1 MgCl2, 1 NaH2PO4, 26.2 NaHCO3, 10 glucose and 11 sucrose (315-320 mosm/kg, pH 7.39 when gassed with 95% O2, 5% CO2 at room temperature). Brain blocks containing the hypothalamus were made by trimming the whole brains, and slices of 200 μm thickness were cut and transferred to a glass beaker containing oxygenated ACSF at 31 °C for 1 h. To conduct recording, individual slices were placed in a recording chamber perfused with warmed (31-32 °C) ACSF at a flow rate of 2-3 ml/min. GFP-PVH cells were patched using epifluorescence and IR-DIC optics. This study used whole-cell current clamp recordings to obtain information on the membrane potential of PVH MC4R neurons. Recordings were performed using patch pipettes of 3 - 5 MΩ resistance when filled with a solution containing (in mM) 125 K gluconate, 8 KCl, 4 MgCl2, 10 HEPES, 5 NaOH, 4 Na2ATP, 0.4 Na3GTP, 5 Na2-creatine phosphate, 7 sucrose and 7 KOH which resulted in a pH ~7.23 and osmolality of 295–300 mosmol/kg. α-MSH (250 nM) was bath applied to the slice. ML418 on α-MSH were added to aCSF and bath applied to the slice via the perfusion system. To examine the effect of ML418 on α-MSH-induced effects on membrane potential, slices were pre-treated with ML418 (10 µm). In order to quantify the action potential firing and amplitude of depolarization induced by α-MSH, current clamped MC4R neurons were held at membrane potentials between -55 to -70 mV to prevent spontaneous action potential firing. The firing frequency and membrane potential of neurons was measured during a 3 min period before, and for another 3 min period 9–15 min after administration of the drugs, and results were compared. Changes in the input resistance of neurons by ML418 were evaluated by examining the responses of MC4R neurons to applications of consistent current pulses in vehicle and in the drug. Data were acquired at 10 kHz using a MultiClamp 700 A amplifier (2000X gain, −3 dB filter freq: 5 kHz) and Clampex 10.0.1 software (Axon Instruments, Union City, CA). Acute feeding experiments Mice were individually housed for at least 7 days and acclimated to handling and saline injection for 5–7 days. Compound/drug injections occurred at least 30 min prior to the onset of the dark cycle. ML418 (5-7.5 mg/kg) and A-4166 (5 mg/kg) were intraperitoneally (ip) injected into mice at the onset of the dark cycle. ML418 and VU0134992 (500pmoles/1uL) were delivered into the left lateral ventricle (icv). Setmelanotide and ML418 (200pmoles/0.5uL/side) were injected into the paraventricular nucleus of the hypothalamus (PVH) of mice. Pre-weighed food pellets were given to mice for all experiments, and food intake was measured over 24 h or at the 24-h mark after compound administration. Body weight and water intake were measured 24 h post- administration. Surgery and cannula placement Mice were housed in groups of 3 or more with ad libitum access to standard chow and water as before surgery. An Intracerebroventricular (icv) cannula was implanted under 3–4% (v/v) isoflurane before being placed in a stereotaxic surgical frame (David Kopf Instruments, Tujuna, CA) and then maintained at 1.5–2% (v/v) isoflurane for the rest of the surgery. Cannula were implanted as previously described50. The skull was exposed by an incision and leveled between lambda and bregma. Guide Cannula (Plastics One, Roanoke, VA) targeted the left lateral ventricle using the following flat skull coordinates (-0.460 mm posterior to Bregma, 1.00 lateral to the midline, and -2.20 ventral to the surface of the skull). The cannula was secured to the skull with dental cement. Following recovery, mice received an icv injection of 20 ng angiotensin II (Sigma- Aldridge) to confirm cannula placement by angiotensin-induced water intake. Cannula placement was also confirmed postmortem under a microscope. For intra-PVH cannulation, bilateral guide cannula (Plastics One, Roanoke, VA) dipped in DiD dye (Invitrogen) targeted the PVH at the coordinates (0.7 mm posterior to bregma, ±0.3 mm from midline, 5.0 mm ventral to the surface of the skull). The proper coordinates were histologically verified post-mortem by visualizing DiD under a fluorescent microscope. Only mice with DiD expression in the PVH were included in the final analysis. Immunofluorescence staining MC4R-GFP mice were handled and peripherally injected with 150 μl sterile daily for 4-5 days before the experiment to minimize background neuronal activation (c-Fos) due to stress. Mice received an injection of 5 mg/kg ML418 (ip) or vehicle (DMSO, 20uL, ip). After 90 mins, mice were given a lethal dose of anesthesia and transcardially perfused with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4). Brains were removed and postfixed in the same fixative overnight. Tissues were rinsed in PBS and cryoprotected in 30% sucrose in PBS until the tissue sank. Brains were embedded in OCT (Tissue-Tek), frozen in a dry ice ethanol bath, and stored at -80 °C. Hypothalamic sections were cut on a cryostat at 50 μm thick. The tissues were immediately washed with PBS, rocking slowly at room temperature (RT) to warm. Free-floating sections were blocked for 1 h with 0.3% triton and 5% normal serum in PBS. Primary c-Fos (1:2000; synaptic systems 226 004) and primary GFP antibody (1:4000; Aves Labs GFP-1020) were diluted in the blocking buffer. Sections were incubated with primary antibodies overnight, rocking at RT, rinsed 3 times in PBS, and incubated with secondary antibodies in blocking buffer (Alexa conjugated antibodies, Thermo Fisher Scientific, 1:500) for 2 h at RT. The primary antibodies were omitted for control experiments, and sections were incubated with the blocking buffer instead. Sections were rinsed 3 times with PBS, counterstained with Hoechst 33342 (Thermo Fisher Scientific, 1/2000 in PBS-0.05% Tween 20), and rinsed 2 times with PBS. Sections were moved and slides and coverslips were mounted with Fluoromount G (Southern Biotechnology). c-Fos was counted on sections on all PVH sections (0.60 -1.20 mm posterior to bregma). Materials availability ML418 is available through Tocris. Setmelanotide was custom synthesized at Vivitide. Further information and requests for strains, resources, and reagents should be directed to Roger Cone (rcone@umich.edu) Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Supplementary Data 1 Reporting Summary Transparent Peer Review file Source data Source Data