G protein–coupled receptors differentially regulate glycosylation and activity of the inwardly rectifying potassium channel Kir7.1

Kir7.1 is an inwardly rectifying potassium channel with important roles in the regulation of the membrane potential in retinal pigment epithelium, uterine smooth muscle, and hypothalamic neurons. Regulation of G protein–coupled inwardly rectifying potassium (GIRK) channels by G protein–coupled receptors (GPCRs) via the G protein βγ subunits has been well characterized. However, how Kir channels are regulated is incompletely understood. We report here that Kir7.1 is also regulated by GPCRs, but through a different mechanism. Using Western blotting analysis, we observed that multiple GPCRs tested caused a striking reduction in the complex glycosylation of Kir7.1. Further, GPCR-mediated reduction of Kir7.1 glycosylation in HEK293T cells did not alter its expression at the cell surface but decreased channel activity. Of note, mutagenesis of the sole Kir7.1 glycosylation site reduced conductance and open probability, as indicated by single-channel recording. Additionally, we report that the L241P mutation of Kir7.1 associated with Lebers congenital amaurosis (LCA), an inherited retinal degenerative disease, has significantly reduced complex glycosylation. Collectively, these results suggest that Kir7.1 channel glycosylation is essential for function, and this activity within cells is suppressed by most GPCRs. The melanocortin-4 receptor (MC4R), a GPCR previously reported to induce ligand-regulated activity of this channel, is the only GPCR tested that does not have this effect on Kir7.1.

Kir7.1 is an inwardly rectifying potassium channel with important roles in the regulation of the membrane potential in retinal pigment epithelium, uterine smooth muscle, and hypothalamic neurons. Regulation of G protein-coupled inwardly rectifying potassium (GIRK) channels by G protein-coupled receptors (GPCRs) via the G protein ␤␥ subunits has been well characterized. However, how Kir channels are regulated is incompletely understood. We report here that Kir7.1 is also regulated by GPCRs, but through a different mechanism. Using Western blotting analysis, we observed that multiple GPCRs tested caused a striking reduction in the complex glycosylation of Kir7.1. Further, GPCR-mediated reduction of Kir7.1 glycosylation in HEK293T cells did not alter its expression at the cell surface but decreased channel activity. Of note, mutagenesis of the sole Kir7.1 glycosylation site reduced conductance and open probability, as indicated by single-channel recording. Additionally, we report that the L241P mutation of Kir7.1 associated with Lebers congenital amaurosis (LCA), an inherited retinal degenerative disease, has significantly reduced complex glycosylation. Collectively, these results suggest that Kir7.1 channel glycosylation is essential for function, and this activity within cells is suppressed by most GPCRs. The melanocortin-4 receptor (MC4R), a GPCR previously reported to induce ligand-regulated activity of this channel, is the only GPCR tested that does not have this effect on Kir7.1.
Kir7.1, encoded by the Kcnj13 gene, is a two-transmembrane domain potassium channel, closer in homology to Kir channels associated with potassium transport such as Kir1. 1, 4.x, and 5.1 (1). Compared with other channels, Kir7.1 exhibits a small unitary conductance and low dependence on external potassium (2). This was shown to be due to the presence of a methionine at position 125 in the pore, where other Kir channels have an arginine. Mutation of this residue to arginine was found to mimic/restore the conductance and potassium dependence observed in like-family channels. Kir7.1 is widely expressed, but particularly high expression has been reported in the retinal pigment epithelium (RPE), 2 thyroid, uterine smooth muscle, small intestine, and choroid plexus of the brain (3,4).
Until recently, Kir7.1 was primarily studied within the RPE. In RPE cells, Kir7.1 is found in the apical membrane close to photoreceptor neurons, where it is thought to contribute to ion homeostasis (5). Mutations in Kir7.1 lead to snowflake vitreoretinal degeneration and Leber congenital amaurosis (LCA), which are both retinal dystrophies (6 -10). In jaguar/obelix zebrafish, mutations in Kir7.1 affect pigment patterns, specifically as a result of the failure of melanosomes to respond to changes in light (11). These functional mutations of Kir7.1 in RPE cells and melanocytes have been characterized and shown to cause alterations in channel currents resulting in the described pathologies.
Recent studies have also identified a role for Kir7.1 in the regulation of uterine smooth muscle excitability (12). Specifically, Kir7.1 expression is elevated by 30-fold midgestation, hyperpolarizing muscle cells and reducing uterine excitability. This expression falls off toward the end of pregnancy, leading to parturition. In the mouse, Kir7.1-null mutations cause perinatal lethality, and this was determined to be due to a role for Kir7.1 in tracheal tubulogenesis (13). Kir7.1 was also revealed to be a potential regulator of neuronal excitability in the paraventricular nucleus of the hypothalamus, through modulation of channel activity by the G protein-coupled melanocortin-4 receptor (MC4R) (14). Activation of MC4R led to robust neuronal depolarization, which was found to be mediated by Kir7.1 channel closure. Conversely, inhibition of MC4R by an endogenous inverse agonist, AgRP, led to channel opening and hyperpolarization of the neuron. These results are interesting in that Kir7.1 is not a canonical G proteincoupled inwardly rectifying potassium (GIRK) channel such as Kir3.x. Furthermore, in HEK293T cells, Kir7.1 was found to exist in a complex with MC4R (14).
Mechanisms by which non-GIRK channels interact with GPCRs vary, with some channels behaving like GIRKs, such as Kir2.x channels. Specifically, unlike Kir2.1, it has been reported that Kir2.3 channels can be inhibited directly by G ␤␥ complexes (15). Moreover, within dendritic spines of cholinergic neurons, M1 receptor activation has been shown to lead to inhibition of Kir2.x channels through G q activation of phospholipase C, which depletes the second messenger phosphatidylinositol 4,5bisphosphate (PIP 2 ) that is needed for Kir channel function (16). Kir7.1 likewise has a PIP 2 dependence for activity and was recently shown in RPE cells to be regulated by the oxytocin G q -coupled GPCR, through phospholipase C-mediated inhibition (17). However, the MC4R-mediated closure of Kir7.1 was found to be independent of G proteins (14). The mechanism by which this occurs is unknown, and examples exist of receptors interacting with Kir channels in the absence of G proteins and other second messengers. Specifically, Kir6.2 and the non-GPCR sulfonylurea receptor (SUR), a target of sulfonylurea drugs used to treat diabetes, form the K ATP channel (18). This heterooctamer composed of Kir6.2 subunits in 1:1 stoichiometry with SUR1/2 is expressed in the pancreas, and ATP binding to the SUR subunits increases channel open probability (19). SUR communicates with Kir6.2 through direct interaction between the C-terminal tail of Kir6.2 and an intracellular loop of SUR (20,21).
In our initial investigations of potential molecular mechanisms of interaction for Kir7.1 with MC4R, we observed that co-expression of several GPCRs with Kir7.1 in HEK293T cells led to a dramatic reduction in complex glycosylated forms of Kir7.1 present in the cell. A review of the role of glycosylation within this family of inwardly rectifying potassium channels revealed that glycosylation of Kir1.1, a close family member, reduces open channel probability (22,23). In patients with multiple sclerosis, autoantibodies against glycosylated forms of Kir4.1 have been reported, but the role for this in the pathophysiology of the disease is unclear (24,25). A role for glycosylation of Kir7.1 has not been investigated. We report here on a novel functional role of GPCRs in the complex glycosylation and function of Kir7.1.

Kir7.1 channel glycosylation is altered by the ␤2-adrenergic receptor (␤2AR)
The ␣-MSH neuropeptide regulates food intake and energy expenditure through the MC4R, a G protein-coupled receptor. Studies of MC4R neurons suggested this receptor regulates neuronal depolarization and activation by ligand-mediated regulation of the conductance of Kir7.1 (14). Further, these studies supported a G protein-independent mechanism of regulation of Kir7.1. To investigate this unusual mode of Kir regulation, we first co-expressed a C-terminal FLAG-tagged Kir7.1 either with 3ϫ HA-MC4R or as a control with the widely studied ␤2-adrenergic receptor (3ϫ HA-␤2AR) in HEK293T cells in the absence of ligand. We selected the ␤2AR as a control because like the MC4R this receptor also couples to G␣ S . Unexpectedly, we observed a significant reduction in the 50-kDa migrating doublet of Kir7.1 when co-expressed with ␤2AR, whereas MC4R caused a small increase in this form of the channel in some experiments (Fig. 1A). Treatment of cells with melanocortin agonists, shown to reduce Kir7.1 currents in MC4R neurons of the paraventricular nucleus of the hypothalamus (14), had no effect on the typical amount or pattern of Kir7.1 bands observed (not shown). Using endoglycosidases, we confirmed the upper 50-kDa bands as the Endo H-resistant mature/complex glycosylated form of the channel, indicating that these bands are Kir7.1 that include glycans added in the Golgi complex (Fig. 1B). With Endo H treatment, the 37-kDa doublet migrated as a single band, indicating Endo H sensitivity. As Endo H-sensitive glycans are added in the endoplasmic reticulum, we identified this upper 37-kDa band as the core glycosylated form and classified this lower band as the unglycosylated form of Kir7.1. Additionally, treatment with the glycosidase PNGase F, which cleaves all N-linked glycans of proteins, resulted in a single migrating band at 37 kDa, confirming that Kir7.1 glycosylation is N-linked (Fig. 1B). Furthermore, we established a metric for the proportion of Kir7.1 complex glycosylation, by determining the ratio of mature/complex glycosylation, to immature core glycosylated plus unglycosylated forms of the protein. Although some variability in this ratio was seen, a value of ϳ1 was typically observed across multiple Kir7.1 expression trials in HEK293T cells in the absence of GPCR co-expression. Co-expression with ␤2AR reduced the ratio of glycosylation to less than 0.5, whereas MC4R exhibited a trend toward increasing this ratio, indicating opposing effects of these GPCRs (Fig. 1C).

Dose-responsive inhibition of Kir7.1 complex glycosylation with increased ␤2AR expression
With the striking observation of reduced Kir7.1 glycosylation following ␤2AR but not MC4R co-expression, we were curious whether differences in receptor expression levels could account for the proportion of Kir7.1 complex glycosylation observed. To examine this, we co-expressed Kir7.1 with increasing amounts of either MC4R or ␤2AR protein and assessed glycosylation by Western blotting analysis. We observed that increasing the amount of MC4R expressed had no effect on proportion of Kir7.1 complex glycosylation, whereas increasing amounts of ␤2AR potentiated the loss of complex glycosylation (Fig. 2, A and B). Moreover, co-expression with ␤2AR appeared to be increasing the amounts of core-glycosylated/immature Kir7.1 expressed. We determined the change in ratio of Kir7.1 glycosylation with changes in receptor expression, and ␤2AR mediated a progressive reduction in the ratio of complex to immature forms (unglycosylated and core glycosylated) of Kir7.1 (Fig. 2C). Increases in the amount of MC4R protein expressed did not have an effect on the ratio of mature/ immature forms of Kir7.1. As a control, and to determine GPCRs differentially regulate Kir7.1 glycosylation whether the effect on glycosylation was specific to Kir7.1, we assessed whether overexpression of either GPCR would alter the glycosylation of the widely studied vesicular stomatitis virus glycoprotein (VSVG) tagged with GFP. VSVG-GFP is used to elucidate protein trafficking, where forms of the protein that have mature/complex glycosylation are Endo H-resistant. This can be determined by observing a band shift when comparing Endo H-treated versus untreated cell lysates of VSVG-GFP with either GPCR. We first treated cells transiently expressing VSVG-GFP only with a transport inhibitor (brefeldin A/monesin) to prevent complex glycosylation and observed the band shift compared with untreated samples as a control (Fig. 2D). Keeping expression levels of Kir7.1 constant throughout (0.25 g of expression vector) and varying ␤2AR expression 5-fold (from 0.05 to 0.25 g of expression vector), the high expression levels of ␤2AR that altered Kir7.1 glycosylation in Fig. 1A (0.25 g of expression vector) had no effect on the glycosylation of the VSVG-GFP protein (Fig. 2E, seventh and eighth lanes). Only when ␤2AR was overexpressed 3-fold relative to Kir7.1 (0.25 g of Kir7.1, 0.75 g of ␤2AR) was VSVG-GFP glycosylation affected, most likely because of squelching of glycosylation machinery. With the exception of the overexpression shown in Fig. 2E (ninth and tenth lanes), the Western blotting analyses reported here never exceeded the 1:1 transfection of 0.25 g each of Kir7.1 and ␤2AR expression vectors. MC4R also did not alter VSVG-GFP glycosylation at levels used throughout this experiment (Fig. 2F). These data further indicated a unique effect of ␤2AR expression, in contrast with MC4R, on suppression of Kir7.1 complex glycosylation, at expression levels that have no effect on the glycosylation of VSVG protein.

␤2AR-reduced Kir7.1 mediated whole-cell currents
Given the specific effect of ␤2AR on Kir7.1 glycosylation, we investigated whether this had an effect on the function of Kir7.1. To study the functional effect of the ␤2AR-mediated loss of Kir7.1 complex glycosylation, Kir7.1 whole-cell recordings were performed in the absence of any ligand in transfected HEK293T cells. Whole-cell patch clamp recordings were performed following Kir7.1 transfection and overexpression of either ␤2AR or MC4R. Because the Kir7.1 WT channel was found to have a small unitary conductance (2), we used a mutant form of the channel in which the methionine within the pore was replaced with an arginine typically found in the pore of other two-transmembrane domain Kir channels (Kir7.1-M125R). This mutant has been widely validated and used in multiple studies of Kir7.1 activity (2,26,27). Kir7.1-M125R expressed alone compared with co-expression with MC4R yielded no differences in whole-cell current, which aligned with previous observations (14) (Fig. 3, A and B). However, ␤2AR caused a Ͼ50% reduction in Kir7.1 whole cell-mediated currents (Fig. 3, C and D). Because Kir7.1-M125R channels exhibit barium sensitivity, activity of expressed Kir7.1 was taken as the total amount of current inhibited by barium at the end of the experiment. In general,

GPCRs differentially regulate Kir7.1 glycosylation
reduced responses to changes in voltage were observed when Kir7.1 was co-expressed with ␤2AR (Fig. 3D). Because these experiments were performed in the absence of ligand, this suggested that glycosylation either reduces the amount of functional channel at the surface of the cell or directly alters channel function.

␤2AR-mediated loss of complex glycosylation does not alter total Kir7.1 surface expression
We next investigated whether co-expression with ␤2AR altered the amount of Kir7.1 trafficked to the cell surface. Cells co-expressing ␤2AR or MC4R with Kir7.1 were biotin-labeled . Whole-cell lysates were analyzed by Western blotting analysis (WB). To assess receptor expression levels, the samples were rerun on a separate blot and imaged with anti-HA. This blot was stripped and reprobed for the loading control GAPDH to confirm plasmid concentration-based increase in receptor protein levels. C, densitometric analysis of the anti-FLAG blot to determine the ratio of mature to complex glycosylation. Densitometric analysis was also performed on the receptor blot and normalized to the GAPDH loading control for both ␤2AR expression and MC4R expression. The ratio of mature to immature Kir7.1 glycosylation was plotted against Log2 receptor intensity values, and a linear regression line was fitted to the data points. The data are representative of at least three experiments. D-F, comparable increases in receptor expression do not alter VSVG protein glycosylation. Whole-cell HEK293T lysates transiently expressing VSVG-GFP (0.25 g) with Kir7.1 (0.25 g) or increasing concentrations of ␤2AR (0.08, 0.125, or 0.25 g) or MC4R (0.8, 1.25, or 2.5 g) were treated Ϯ Endo H. As a control, VSVG-GFP only expressing cells were treated with a transport inhibitor (brefeldin A (BFA)/monesin) and compared with untreated cell lysate. Anti-HA blots show receptor expression levels. **, Endo H-resistant; ***, Endo H-sensitive. VSVG-GFPϩ3ϫ HA-␤2AR OE wells show that a 3-fold overexpression of ␤2AR (0.75 g) can affect VSVG-GFP glycosylation. The data are representative of three independent experiments. M.W, molecular mass.

GPCRs differentially regulate Kir7.1 glycosylation
and assessed for cell surface expression by Western blotting analysis. This method allowed visualization of the different glycosylated forms of Kir7.1 reaching the cell surface. Surprisingly, Kir7.1 co-expression with ␤2AR and loss of complex glycosylation did not cause a reduction in the percentage of Kir7.1 protein on the cell surface (Fig. 4B). MC4R likewise had no effect on Kir7.1 surface expression. Additionally, we assessed whether the ratio of complex to immature forms of Kir7.1 at the surface changed in the presence of these GPCRs (Fig. 4C). This remained unchanged, with ␤2AR maintaining a Ͻ0.5 ratio of complex Kir7.1 glycosylation compared with the total glycosylated Kir7.1 expressed in the cell. A slight reduction in the ratio of surface mature/immature Kir7.1 following MC4R expression was observed, but this trend was not significant over multiple experiments. These results therefore indicate that complex glycosylation is not required for surface expression of Kir7.1.

Lack of glycosylation of Kir7.1 alters channel gating
Within the Kir family, the functional role of glycosylation has been studied for ROMK (Kir1.1) and GIRK1 (Kir3.1) (22). Although loss of glycosylation at Asn-119 in GIRK1 was reported not to be required for function, loss of glycosylation at Asn-117 in ROMK1 reduced open channel probability and whole-cell currents (23). Additionally, although Kir3.4 is reported to have an asparagine accessible for glycosylation, the site was found to be inactive (22). Thus, divergent roles for glycosylation exist within the family. Based on these studies, we aligned the Kir7.1 protein sequence with ROMK and GIRK1 and determined that Kir7.1 also shared a highly conserved glycosylation site (Fig. 5A). We mutated this site to a glutamine (Kir7.1-N95Q) and confirmed by Western blotting analysis the loss of both core and complex forms of glycosylation (Fig. 5B). Furthermore, we generated a model of Kir7.1 based on the Kir2.2 crystal structure using SWISS-MODEL, which revealed the likely glycosylation site to be in the outer loops gating the pore (Fig. 5C). This was suggestive of a functional role for glycosylation in Kir7.1. To assess this, we performed single-channel recordings using the M125R mutant of Kir7.1 to compare the WT fully glycosylated channel (Kir7.1-M125R) to the unglycosylated mutant channel (Kir7.1-M125-N95Q).
WT Kir7.1 displayed clusters of openings with a main amplitude level of ϳ1 pA (Fig. 6, A and D). Lack of glycosylation at the Asn-95 position of Kir7.1 channels caused openings to occur at an amplitude level of ϳ0.6 pA (N95Q ϭ 0.65 Ϯ 0.03 pA, n ϭ 6; WT ϭ 1.11 Ϯ 0.07 pA, n ϭ 7; p ϭ 0.0001) with a reduced open probability (N95Q ϭ 0.23 Ϯ 0.05, n ϭ 6; WT ϭ 0.41 Ϯ 0.06, n ϭ 7; p ϭ 0.0484) (Fig. 6, B and E). Both channels opened to at least three open states (O1, O2, and O3), with no changes in the relative occurrence of opening events (N95Q ϭ 9.09 Ϯ 3.8 ms, n ϭ 6; WT ϭ 17.9 Ϯ 3.5 ms, n ϭ 7; p ϭ 0.1188). We tested whether␤2ARmaybedirectlysuppressingchannelactivityinde-  5). B, a representative current-voltage (I-V) relationship is shown. The Kir7.1-M125R mutant has a methionine in the pore of the channel mutated to an arginine typically found at the same site in most other Kir channels. This mutation significantly increases the amount of detectable currents from the channel, with increased sensitivity to barium. Therefore, the amount of current inhibited by barium at the end of the experiment was recorded as the total current of Kir7.1-M125R from HEK293T cells. Mock-transfected cells did not exhibit barium-sensitive currents (data not shown). C, Kir7.1-M125R co-expressed with ␤2AR showed a significant reduction in whole-cell current in pA/pF, p0.0001 (n ϭ 7 cells). The error bars represent S.E. D, representative current-voltage (I-V) relationship of Kir7.1-M125R (black) in the presence or absence of ␤2AR (red). NS, not significant.

Other ␤-adrenergic receptors also alter Kir7.1 glycosylation
We next sought to determine whether suppression of complex glycosylation and function of Kir7.1 was unique to the ␤2AR. For this, we first tested whether the ␤1 and ␤3 adrenergic receptors also mediated alterations in Kir7.1 glycosylation. We saw a similar reduction in the proportion of complex glycosylation when co-expressing these other ␤-adrenergic receptors with Kir7.1 in HEK293T cells (Fig. 7A). Special care was taken to express the receptors at similar levels to allow adequate comparison for changes in glycosylation. ␤1AR and ␤3AR similarly reduced the ratio of Kir7.1 glycosylation as seen with ␤2AR, with ␤1AR having a slightly lower value (Fig. 7B). This, however, may be due to receptor expression of ␤1AR being slightly higher than ␤2AR (Fig. 7A). These results indicate that our original observation was not specific to the ␤2AR receptor but shared by others GPCRs tested within the family. We also attempted to screen a larger panel of GPCRs for effects on glycosylation (MC1R, MC3R, and OTXR) and did indeed see a reduction in Kir7.1 glycosylation ratio (Fig. S1). Although we were unable to attain consistent receptor expression levels for these receptors, the results suggest that many GPCRs suppress the complex glycosylation of Kir7.1

Kir7.1 has a conserved Golgi export site, essential for complex glycosylation
As complex glycosylation occurs in the Golgi network, we sought to identify mechanisms by which GPCRs might suppress complex glycosylation of Kir7.1 described above. With recent reports of a Golgi export site required for surface expression of Kir channels (28,29), we found that Kir7.1 shared a site similar to that in the N-and C-terminal regions of Kir2.1, 4.1, and 5.1 that has been identified as essential for Golgi export (Fig. 8, A and B) (28). Within the channels mentioned, deletion GPCRs differentially regulate Kir7.1 glycosylation of the serine and tyrosine (SY) residues within the C-terminal site was found to be sufficient to prevent Kir export from the cis-Golgi and trans-Golgi compartments of the cell, where complex glycosylation takes place (28,29). Additionally, although WT Kir2.1 is not glycosylated, it was reported that deletion of SY in Kir2.1 also reduced complex glycosylation of an artificially placed asparagine residue (29). Given the high conservation of this site in Kir7.1, we tested whether deletion of these residues would alter WT Kir7.1 glycosylation. Deletion of the SY motif in Kir7.1 led to a significant loss of complex glycosylation with a ratio of Ͻ0.1 in Cos1 cells (Fig. 8, C and D). Cos1 cells were used because published studies of the export site in other Kir channels were performed in Cos cell lines (30). We further confirmed the loss of complex glycosylation by endoglycosidase digestion using Endo H and PNGase F (Fig. 8E). This indicates that the Golgi export residues SY in Kir7.1 are essential for complex glycosylation of Kir7.1. Although the SY mutant did have a significantly low glycosylation ratio, we also observed a glycosylation ratio of Ͻ0.8 in the WT channel in Cos1 cells (Fig. 8E), lower than the typical ratio of 1 observed in HEK293T cells. Because differential glycosylation of Kir channels have been reported across different tissue subtypes (31), we tested whether ␤2AR would likewise mediate a loss of glycosylation in the Cos1 cell line. The striking loss of Kir7.1 glycosylation observed within HEK293T cells when co-expressed with ␤2AR was repeated within Cos1 cells compared with the WT channel, and co-expression with MC4R again lacked the same effect (Fig. 8, F and G). This further indicated the strong specificity of the ␤2AR-Kir7.1 interaction.

Kir7.1 forms heterotetramers of immature, unglycosylated, and fully glycosylated subunits
Having established that complex glycosylation of Kir7.1 is required for function, we tested whether the channel exists as homo-or heterotetramers of subunits with no glycosylation, core glycosylation, and complex glycosylation. We co-expressed FLAG-tagged Kir7.1-N95Q or Kir7.1⌬SY with an HA-tagged WT Kir7.1 channel. HA-tagged WT Kir7.1 co-immunoprecipitated with both the unglycosylated (N95Q) and immature glycosylated (⌬SY) forms of the channel, in addition to the fully glycosylated WT control (Fig. 9). This suggests that nonfunctional Kir7.1 subunits may reduce channel function through subunit poisoning.

Retinopathy-associated mutations of Kir7.1 also alter channel glycosylation
Genetic studies have linked retinopathies of the eye to mutations in Kir7.1, demonstrating that Kir7.1 plays a key role in the function of the retinal pigment epithelium. Snowflake vitreoretinal degeneration and LCA are two Kir7.1-associated hereditary eye disorders, where known variants lead to loss of channel function (32). We wanted to determine whether glycosylation is also affected by any of these disease mutations, and a few known variants of Kir7.1, namely R162W, R162Q, Q117R, and L241P, were tested. These are all considered to be loss-of-function mutations except R162Q, which is linked to blindness but has no reported phenotype (32). Notably, none of these mutations are in known sites required for complex glycosylation. Interestingly, we observed a significant reduction in glycosylation in the L241P mutation (Fig 10). The Q117R mutation appeared also to have a modest reduction in the proportion of complex glycosylation.

Discussion
GIRKs are Kir family members best understood to be regulated by GPCR signaling (1,33). Our study reveals another mechanism by which GPCRs appear to regulate the function of another inward rectifier, Kir7.1, prior to cell surface expression and regulation by ligand binding. Our results demonstrate a highly specific effect of GPCRs on Kir7.1 glycosylation, which in GPCRs differentially regulate Kir7.1 glycosylation turn directly regulates channel function. Additionally, although glycosylation of Kir7.1 has been mentioned in other publications (2,32), no other study has characterized the glycosylation of the channel and its effect on function. Specifically, we have shown that a loss in the proportion of receptor subunits undergoing complex glycosylation is mediated by expression of a wide variety of GPCRs. Further, we observed that the reduction in Kir7.1 whole-cell currents when co-expressed with unliganded ␤2AR (Figs. 3 and 6) was indeed due to loss of complex glycans, an event occurring prior to trafficking of the channel to the cell surface. The MC4R has previously been reported to regulate Kir7.1, with agonists inhibiting and antagonists stimulating channel function; however, expression of the MC4R did not reduce complex glycosylation of the channel, with or without the presence of ligand. Nonetheless, it is intriguing to note that the MC4R is one of the few GPCRs we examined that leaves channel glycosylation intact.
Glycosylation has been previously reported for channels of the same family, Kir1.1 (ROMK) and Kir3.1 (22). Loss of glycosylation in Kir1.1 was shown to decrease whole-cell currents because of a significant reduction in open channel probability (23). We observed similar results in our N95Q glycosylation resistant mutant by single-channel recordings. Interestingly, Kir3.1, although having an active glycosylation site, is not reported to require glycosylation for function. Although Kir3.1 and Kir3.4 form complexes, Kir3.4 has a consensus glycosyla-

GPCRs differentially regulate Kir7.1 glycosylation
tion site that is reported to be inactive (22). Thus, within the two-transmembrane domain family of Kir channels, a divergent role for channel glycosylation exists. Additionally, the Kir4.1 channel has been shown to be differentially glycosylated in different tissues. Although higher amounts of complex glycosylation of Kir4.1 have been observed in kidney versus brain lysates, a functional role for glycosylation has not been investigated (31). We were curious whether GPCRs could affect the glycosylation of these closely related channels Kir1.1 and Kir4.1, but we were unable to observe consistent complex glycosylation of these channels in transiently transfected HEK293T cells.
Kir1.1 or ROMK has the highest similarity to inward rectifier Kir7.1, with 38% sequence homology. As mentioned previously, Kir1.1 is the only inward rectifier with a clear reported role for glycosylation. The Asn-95 site in Kir7.1 appears structurally related to the Asn-117 site in Kir1.1, which suggested a similar role for glycosylation in Kir7.1, as validated here. From sequence analysis, the glycosylation site of both these channels is present in the turret region of the channel, an extracellular loop before the pore helix between the two transmembrane domains (34). This region has high variability across the large potassium channel family, with a potential function for sensitivity to toxin binding (34). Based on our homology model of Kir7.1 (Fig. 5C), these extracellular loops would house the relatively large glycans at Asn-95, which by our estimate from Western blotting analysis are over 13 kDa in size. However, the exact mechanism by which the presence of these glycans is essential for a fully functional channel is not clear. Structural comparison of the Kir7.1 model to the Kir2.2 crystal structure also shows that the turret region of Kir7.1 is longer than that in Kir2.2 with additional prolines present in the loop. These proline residues may introduce additional turns in the loop either toward or away from the pore. However, we were unable to predict these conformations with SWISS-MODEL. A crystal structure of this extracellular region of Kir7.1 would indeed allow us to deduce the potential mechanisms by which glycosylation aids channel function.
It was interesting to observe that the L241P LCA associated variant of Kir7.1 exhibited a nearly complete loss of complex glycosylation. This mutation has not been extensively studied, but the loss of channel function was predicted because of its association with LCA, a severe form of retinal degeneration (7). This mutation is thought to be in a ␤ sheet of the protein, near the C-terminal region and is predicted to affect protein folding. It remains unknown whether the change in structure is responsible for the loss of glycosylation and thus a loss of function, a reduction in protein stability, or a loss of activity caused by another structural defect (10). The protein did not express well following transfection of HEK293 cells (Fig. 10), but data are not available to assess whether or not the mutation reduces expression levels in RPE cells in vivo.
Of other non-GIRK channels that interact with GPCRs, Kir4.1 is reported to interact with the calcium-sensing receptor (CaR) GPCR within the kidney (35). Co-expression of CaR and Kir4.1 in HEK293 cells was found to reduce surface expression of the channel, thereby limiting its activity. This was also reported to hold true for Kir4.2 and was found to be mediated by CaR Gq signaling and internalization by caveolin-1 (35,36). This was not the case with ␤2AR's effect on Kir7.1, wherein co-expression with ␤2AR did not reduce Kir7.1 surface expression. In the case of CaR and Kir4.1, it was indeed reported that mutant nonfunctional CaRs were unable to have an effect on Kir4.1 surface expression or activity (36). This has not been tested for GPCRs that we have shown to alter glycosylation of Kir7.1.
Although glycosylation is indeed not required for surface expression of ion channels such as Kir2.1, which is seemingly unglycosylated, we were surprised to see expression of immature forms of Kir7.1 at the cell surface. Particularly, the use of cell surface biotinylation allowed us to profile the different forms of the channel present at the surface and obtain a ratio of complex to immature plus unglycosylated forms. In co-expression with ␤2AR, the ratio of complex glycosylation remained unchanged at the surface, indicating that GPCR expression does not appear to alter trafficking of the channel. Furthermore, we were able to show that these channels can exist as heterotetramers of unglycosylated, partially glycosylated, and fully glycosylated subunits. Therefore, a change in the ratio of glycosylation most likely reflects a change in channel subunit composition. This would thereby contribute to subunit "poisoning" of Kir7.1 tetramers, further influencing the reduction in channel activity observed (Fig. 3, C and D). We also observed complexes of ␤2AR with the immature forms of Kir7.1 by coimmunoprecipitation in cells (Fig. S2). After cross-linking, only GPCRs differentially regulate Kir7.1 glycosylation the immature forms of Kir7.1 co-immunoprecipitated with ␤2AR, although small amounts of complex Kir7.1 were present in the cell as observed in the input lane (Fig. S2). This suggests that immature channels may be the predominant form in the cells when co-expressed with certain GPCRs.
With ␤2AR forming complexes with the immature form of the channel, this suggests a possible mechanism for the GPCRmediated loss of glycosylation. The protein titration experi-ment showed a progressive increase in immature forms of the protein, with increased expression of ␤2AR. The amount of complex glycosylated Kir7.1, however, was generally reduced. A possible hypothesis is that ␤2AR forms complexes with Kir7.1 early in the endoplasmic reticulum and sterically blocks Kir complex glycosylation in the Golgi. Alternatively, the ␤2AR could also differentially stabilize immature forms of the protein. The potential role of G proteins in the formation of these proposed receptor-channel complexes should be examined in future experiments.
We have observed that deletion of the serine and tyrosine resides within the Golgi export site of Kir7.1 also inhibits complex glycosylation but leaves core glycosylation intact. The SY mutant of Kir7.1 described here may thus be a useful tool in future experiments to further validate the necessity of complex glycosylation for full channel function, suggested by the experiments in Fig. 2. Complex glycosylation is reported to occur at different stages during passage from the cis-medial-trans-Golgi before export to the surface (37). Within other Kir channels, the Golgi export site has been identified as a binding site of the AP-1 ␥-adaptin protein, which is required for export to the cell surface (28,29). In Kir2.1, deletion of these two residues reduced complex glycosylation of an artificial site placed within the channel by mutagenesis, with an intermediate glycosylation pattern observed (29). They argue that because complex glycosylation occurs in the trans-Golgi network, loss of the export site traps Kir2.1 in both the cis-and trans-Golgi networks, preventing further transport leading to incomplete complex glycosylation. However, in the work of Li et al. (28), the authors comment that compared with other Kir channels, Kir7.1 is the only channel that does not seem to share the full export site. Although we have confirmed the export site to be essential for complex glycosylation, these two residues do not appear to be required for Golgi export (data not shown); however, this was expected because we did see immature forms of the WT channel at the surface in biotinylation experiments (Fig. 4).
The AP-1 ␥-adaptin protein is the only protein reported in literature to bind to the Golgi export site in Kir channels. The role of this protein in glycosylation has not been investigated, nor any other proteins that may bind to this site to mediate complex glycosylation. Our data suggest that GPCRs such as ␤2AR are indeed preventing this crucial step of glycosylation from occurring though the mechanism is unclear. It is plausible that GPCRs may indeed be blocking proteins from binding to this protein export site, which we have shown is required for glycosylation.

GPCRs differentially regulate Kir7.1 glycosylation
Both the oxytocin receptor (17) and the MC4R (14) have been shown to regulate Kir7.1 function at the membrane in a physiologically relevant manner, with the former mediated via a PIP 2 intermediate. The mechanism(s) for regulation of Kir7.1 by MC4R remain to be determined, although the potential for Kir7.1 and MC4R to form complexes was shown by co-immunoprecipitation of tagged proteins from HEK293T cells (14). Although MC4R is a slightly smaller GPCR, its inability to block complex glycosylation relative to other GPCRs we tested is not understood. One hypothesis is that MC4R may only form complexes with Kir7.1 post-Golgi export, but this is unknown.
The data shown here provide evidence for molecular regulation of Kir7.1 structure and function by GPCRs. Within tissues where Kir7.1 is highly expressed, large variations in glycosylation have been recently observed using a knockin mouse expressing an HA-tagged Kir7.1 channel (38). Notably, the study showed predominantly immature forms of the channel in the lung, whereas full complex glycosylation was observed in the trachea (38). No explanation for variation in glycosylation patterns between these two tissues was given, but our data may suggest a possible mechanism by which such changes could occur.

Cell lines and cell culture
HEK293T cells were grown in Dulbecco's modified Eagle's medium, with high glucose, L-glutamine, and phenol red supplemented with 10% fetal bovine serum (Atlanta Biologicals) and 1% antibiotic-antimycotic (Thermo Fisher Scientific). The cultures were maintained in 5% CO 2 environment at 37°C. Cos1 cells were cultured similarly to HEK293T cell lines with medium containing sodium pyruvate. The protein transport inhibitor mixture from BD Biosciences was used at 0.5ϫ, and the cells were incubated for 9 h prior to harvesting.

Transfection
Plasmid DNA constructs were transfected into HEK293T or Cos1 cells at 70 -80% confluency using LipD293 reagent (SignaGen) according to the manufacturer's instructions. The cells were allowed to grow for 24 h before harvesting. If cells were kept for 48 h, the medium was changed after 24 h.

Cell surface biotinylation
Cell surface biotinylation was performed as described in Chandrasekhar et al. (39). Briefly, HEK293T cells were transiently co-transfected at 70 -80% confluency with Kir7.1-3ϫ FLAG and either 3ϫ HA-␤2AR or 3ϫ HA-MC4R. 24 h posttransfection, the cells were washed three times with PBS with calcium and magnesium (PBS 2ϩ ). The cells were then incubated with 1 mg/ml of biotin-SS-sulfo (ApexBio) in PBS 2ϩ twice for 15 min each. Excess biotin was quenched with two short washes followed by two 15-min incubations with 100 mM glycine in PBS 2ϩ . The cells were washed with PBS and then lysed with modified radioimmune precipitation assay buffer (see Western blotting protocol). The protein samples were quantified, and 25 l of Pierce TM high-capacity streptavidin agarose beads (Thermo Scientific) were incubated overnight with 75 g of protein lysate. The supernatant was removed, and the beads were washed three times with lysis buffer. The beads were first eluted with 2ϫ LDS with 200 mM DTT then with 100 mM DTT in lysis buffer. Both elutions were combined and loaded on a 10% Bolt TM Bis-Tris Plus gel (Invitrogen) next to one-fifth of the input (15 g of total cell lysate). Western blotting analysis was continued as described.

Western blots and quantitative analysis
Post-transfection, the cells were lysed in a modified radioimmune precipitation assay buffer containing 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1% Triton X-100, 2 mM EDTA, and protease inhibitor mixture (Sigma/Roche). The cells were scraped and incubated on ice with rocking for 30 min followed by a 10,000 rpm centrifugation for 10 min. The supernatant was collected and protein-quantified using a Pierce BCA protein assay kit. Lysates were prepared for denaturing gel electrophoresis by mixing with NuPAGE LDS sample buffer (4ϫ) (Invitrogen) with 400 mM DTT to a final concentration of 1ϫ LDS, 100 mM DTT. Invitrogen Bolt TM 10% Bis-Tris Plus precast polyacrylamide gels were used to run 10 -15 g of protein at 200 V for 40 min using Bolt TM MOPs running buffer. The gels were run with Bolt TM running buffer with Bolt TM antioxidant added to the first chamber of the mini gel tank (Thermo Fisher Scientific). A Bio-Rad protein ladder, Kaleidoscope, Dual Color, or All Blue was also included on each gel. The gels were then transferred to polyvinylidene difluoride membrane (Millipore) using Bolt TM transfer buffer, with the Trans-Blot Turbo TM transfer system from Bio-Rad. The membranes were blotted with 5% nonfat dry milk in PBS with 1% Tween 20 for 30 min, prior to overnight incubation with antibodies. The blots were washed a minimum of three times for 10 min each before imaging with SuperSignal TM West Dura extended detection substrate using the Bio-Rad ChemiDoc TM touch imaging system. Images with samples of interest within the dynamic range were chosen for further quantitative analysis using the Bio-Rad Image Lab TM software. Ratios of mature to immature glycosylation were obtained by dividing the intensity of the upper 50-kDa doublet, by the lower 35-kDa doublet intensity. No normalization was performed on these values because the ratios are independent of protein loading. For comparison of receptor expression levels, intensities recorded were normalized to the loading control GAPDH. The data were analyzed using GraphPad Prism software. A one-or two-way ANOVA statistical test was used to determine significance (p Ͻ 0.05) with Tukey post-test for multiple comparisons.

Glycosidase treatments
Channel glycosylation was investigated using the endoglycosidases Endo H and PNGase F from New England Biolabs. Briefly, cells transiently transfected with Kir7.1-3ϫ FLAG or Kir7.1⌬SY-3ϫ FLAG were harvested, and protein was quantified as previously described. 50 g of lysate was denatured by addition of 10ϫ glycoprotein denaturing buffer (New England Biolabs) followed by heating for 10 min at 55°C. Samples being treated with Endo H were mixed with GlycoBuffer 3 and 2 l of Endo H enzyme. The mixture was incubated for 1.5 h at 37°C in a thermocycler. Samples treated with PNGase F were mixed with GlycoBuffer 2, plus 1 l of the enzyme and also incubated GPCRs differentially regulate Kir7.1 glycosylation for 1.5 h at 37°C. The treated samples were then mixed appropriately with 4ϫ NuPAGE LDS buffer with 400 mM DTT and analyzed by Western blotting analysis as described. The samples were compared with untreated lysates and observed for band shifts in migration.

Expression constructs
Plasmids with full-length cDNA for the GPCRs ␤1AR, ␤2AR, ␤3AR, and MC4R were obtained from the cDNA Resource Center with 3ϫ HA N-terminal tags in the pcDNA3.1ϩ expression vector. The full-length cDNA sequence for KCNJ13 (AJ006128.1) was cloned into the pciNEO expression vector (Promega) with a C-terminal 3ϫ FLAG tag (Sigma). To create the HA-Kir7.1 construct, the full-length cDNA sequence for KCNJ13 was cloned into a pcDNA3.1ϩ vector with a single N-terminal HA tag. To create mutations in Kir7.1, primers were designed using the NEBaseChanger online tool and ordered from Sigma. Mutagenesis was performed using New England Biolabs Q5 site-directed mutagenesis kit (New England Biolabs). For electrophysiology, the cDNA sequence of Kir7.1 was cloned into pcDNA5/TO expression vector (Thermo Fisher Scientific) without tags. The VSVG-GFP construct was provided by the Kenworthy lab at Vanderbilt University.

Whole-cell electrophysiology
Whole-cell patch-clamp electrophysiology was performed as described by Raphemot et al. (18). Briefly, HEK-293T cells were transfected with Kir7.1-M125R-encoding plasmids (0.5 g), MC4R plasmids (3.75 g), or ␤2AR plasmids (3.75 g) and EGFP plasmid (0.05 g) as a marker for transfections. The following day, the cells were dissociated with trypsin and plated on poly-L-lysine-coated glass coverslips. The plated cells were allowed to recover for 1 h prior to experiments. The cover slips were placed in the recording chamber on an inverted microscope stage and perfused with a bath solution containing 135 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 5 mM glucose, 10 mM HEPES, pH 7.4. A Flaming-Brown P-1000 micropipette puller was used to pull electrodes with resistances between 2 and 3 M⍀. The pipettes were filled with 135 mM KCl, 2 mM MgCl 2 , 1 mM EGTA, 10 mM HEPES, pH 7.3. Transfected cells were identified by EGFP fluorescence, and voltage-clamp conditions were used to record whole-cell currents. To obtain a current-voltage curve, the cells were voltage-clamped at a holding potential of Ϫ75 mM and then stepped to Ϫ150 mV for 500 ms before ramping to 150 mV at a rate of 2.4 mv/ms. The cells were superfused with 4 mM BaCl 2 at the end of the experiment to fully block all Kir7.1 channels. This value was used to derive barium inhibitable current.