Identification of a basolateral sorting signal for the M3 muscarinic acetylcholine receptor in Madin-Darby canine kidney cells.

Muscarinic acetylcholine receptors (mAChRs) can be differentially localized in polarized cells. To identify potential sorting signals that mediate mAChR targeting, we examined the sorting of mAChRs in Madin-Darby canine kidney cells, a widely used model system. Expression of FLAG-tagged mAChRs in polarized Madin-Darby canine kidney cells demonstrated that the M(2) subtype is sorted apically, whereas M(3) is targeted basolaterally. Expression of M(2)/M(3) receptor chimeras revealed that a 21-residue sequence, Ser(271)-Ser(291), from the M(3) third intracellular loop contains a basolateral sorting signal. Substitution of sequences containing the M(3) sorting signal into the homologous regions of M(2) was sufficient to confer basolateral localization to this apical receptor. Sequences containing the M(3) sorting signal also conferred basolateral targeting to M(2) when added to either the third intracellular loop or the C-terminal cytoplasmic tail. Furthermore, addition of a sequence containing the M(3) basolateral sorting signal to the cytoplasmic tail of the interleukin-2 receptor alpha-chain caused significant basolateral targeting of this heterologous apical protein. The results indicate that the M(3) basolateral sorting signal is dominant over apical signals in M(2) and acts in a position-independent manner. The M(3) sorting signal represents a novel basolateral targeting motif for G protein-coupled receptors.

Targeting of newly synthesized proteins to their correct subcellular locales is essential for cell function. Protein sorting is particularly important in polarized cells such as neurons and epithelia, where cell-surface proteins must be specifically routed to distinct plasma membrane subdomains. The mechanisms responsible for the correct targeting of membrane proteins in polarized cells remain a fundamental question in cell biology. Madin-Darby canine kidney (MDCK) 1 epithelial cells provide a widely used and well characterized model system for studies of protein targeting (1). Polarized MDCK cells establish apical and basolateral plasma membrane domains with dis-tinct protein and lipid compositions. Many cell-surface proteins contain sorting signals that direct them to the apical or basolateral domain. Apical sorting signals can consist of a glycosylphosphatidylinositol anchor (2), N-glycans (3,4), or protein sequences in the extracellular, transmembrane, and/or cytoplasmic domains (5)(6)(7)(8)(9). In contrast, basolateral sorting signals are almost always found in the cytoplasmic domain of transmembrane proteins and frequently contain a critical tyrosine residue, a dihydrophobic motif, a cluster of acidic residues, or a combination of these elements (10 -13). Although much has been learned about the sorting of single-pass transmembrane proteins, little is known about signals that mediate the targeting of proteins with multiple membrane-spanning domains.
Muscarinic acetylcholine receptors (mAChRs) are a family of seven-transmembrane domain, G protein-coupled receptors composed of five distinct subtypes (M 1 -M 5 ). The M 1 , M 3 , and M 5 receptors preferentially couple to activation of phospholipase C via the G q/11 family of G proteins, whereas the M 2 and M 4 receptors preferentially couple to inhibition of adenylyl cyclase via the G i/o family (14). In addition to their biochemical specificities, mAChR subtypes have unique cellular and subcellular distributions (15). Muscarinic receptors are asymmetrically distributed in polarized cells such as pancreatic and lacrimal acinar cells (16,17), lingual epithelial cells (18), Xenopus oocytes (19,20), and MDCK epithelial cells (21). Furthermore, mAChR subtypes are differentially localized in a variety of neuronal cells. For example, the M 1 receptor is expressed in the cell bodies and dendrites of hippocampal pyramidal neurons and granule cells in the dentate gyrus, where it mediates postsynaptic responses to acetylcholine (22). In contrast, M 2 is found mainly in the axon terminals of cholinergic and non-cholinergic septohippocampal projection neurons and hippocampal interneurons, where it modulates neurotransmitter release (23,24). The M 3 receptor is found both on cell bodies and dendrites of hippocampal granule and pyramidal neurons and on axon terminals in the hippocampal molecular layer and striatum (25,26). Despite the differential localization of mAChR subtypes in a variety of polarized cells, little is known about the mechanisms by which their precise subcellular distributions are achieved.
To begin to elucidate the signals and mechanisms that govern mAChR targeting, we have utilized the MDCK cell system to identify sorting determinants for mAChR subtypes. Although the follicle-stimulating hormone receptor possesses a basolateral sorting signal in its C-terminal cytoplasmic tail (27), and basolateral targeting information for the ␣ 2A -adrenergic receptor appears to be in a domain composed of multiple transmembrane sequences (28), sorting information for G protein-coupled receptors in polarized cells remains largely unknown. In this report, we used chimeric receptor constructs in a gain-of-function approach to identify a basolateral sorting signal for the M 3 mAChR in MDCK cells. The M 3 basolateral sorting signal lies in a 21-amino acid sequence from the Nterminal portion of the third intracellular (i3) loop, is dominant over apical signals in the M 2 receptor, and can act in a positionindependent manner. This M 3 sequence represents a novel basolateral sorting motif for G protein-coupled receptors.

EXPERIMENTAL PROCEDURES
Construction of Epitope-tagged and Chimeric mAChRs-A modified FLAG epitope (DYKDDDDA) was added to the extracellular N termini of the M 1 -M 5 mAChR coding sequences immediately after the initiator methionines using PCR to generate pFM 1 , pFM 2 , pFM 3 , pFM 4 , and pFM 5 . The mouse M 1 (29), porcine M 2 (clone Mc7) (30), human M 3 (31), human M 4 (32), and human M 5 (31) mAChR cDNAs in the mammalian expression vector pCDPS (31) were used as templates. For M 1 , M 2 , and M 4 , the forward primer encoded the FLAG epitope, and the forward and reverse primers contained unique restriction sites to facilitate subcloning into pCDPS. PCR fragments were as follows: M 1 , KpnI-NheI, nt 1-676 of M 1 coding sequence; M 2 , KpnI-MscI, nt 1-689 of M 2 coding sequence; and M 4 , SacI-NheI, nt 1-1229 of M 4 coding sequence. The M 3 and M 5 receptors were FLAG-tagged using a sequential PCR approach as described (33), with the FLAG epitope encoded by internal primers. The M 3 PCR product (NcoI-SnaBI) contained 306 nt of pCDPS vector sequence and nt 1-462 of M 3 coding sequence. The M 5 PCR product (NcoI-EcoRI) contained 367 base pairs of pCDPS vector sequence and nt 1-1021 of M 5 coding sequence. PCR products were subcloned into the parental plasmids to generate epitope-tagged mAChRs. The ability of the FLAG-tagged receptors to bind the muscarinic antagonist [ 3 H]quinuclidinyl benzilate (47 Ci/mmol; Amersham Pharmacia Biotech) was verified by transient expression in COS-7 or JEG-3 cells. The presence of the FLAG epitope was then verified by immunoprecipitation from transfected cell membranes using the anti-FLAG M2 monoclonal antibody (Sigma). Studies of FLAG-M 2 mAChR-mediated inhibition of adenylyl cyclase, receptor desensitization, and sequestration have been reported previously (34). M 2 /M 3 chimeric mAChRs were constructed using sequential PCR as described (33) to replace parts of the M 2 coding sequence with the homologous regions of M 3 coding sequence as aligned in Ref. 31. pFM 2 , pFM 3 , or M 2 /M 3 chimeric constructs were used as PCR templates for subsequent chimeras. All PCR-amplified constructs were engineered with BglII and EcoRI sites at their 5Ј-and 3Ј-ends, respectively, and cloned into the BglII and EcoRI sites of pCDPS. The sequences comprising the M 2 /M 3 chimeras are as follows, with the numbers in parentheses representing the amino acid residues of M 3 that were substituted into Fusion proteins in which M 3 sequences were added to either the i3 loop or the C terminus of M 2 were generated by sequential PCR using pFM 2 and either M 2 /M 3 -(266 -296) or pFM 3 as templates, respectively. The M 2 ϩM 3 -(i3:266 -296) PCR product was cloned into the BglII and EcoRI sites of pCDPS, whereas the M 2 ϩM 3 C-terminal fusion constructs were subcloned into the MscI and EcoRI sites of the parental The interleukin-2 receptor/M 3 fusion protein was generated by sequential PCR using the human interleukin-2 receptor ␣-chain (IL-2R␣; Tac antigen) cDNA (pIL2R3; kindly provided by Dr. Warren J. Leonard, National Institutes of Health, Bethesda, MD) (35) and pFM 3 as templates. This fusion protein consists of coding nt 796 -888 (Ala 266 -Gln 296 ) of M 3 fused to the C terminus of IL-2R␣. Both the fusion protein and wild-type IL-2R␣ were cloned into the BglII and EcoRI sites of pCDPS. PCR-amplified DNA sequences were verified using an Applied Biosystems Model 373A automated sequencing system.
Cell Culture-MDCK (strain II) cells were obtained from Dr. Keith Mostov (University of California, San Francisco, CA). JEG-3 human choriocarcinoma and COS-7 cells were obtained from American Type Culture Collection (Manassas, VA). All cell lines were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin G, and 0.1 mg/ml streptomycin sulfate at 37°C in a humidified 10% CO 2 environment.
Transfection and Immunocytochemical Analysis of Chimeric mAChR Constructs-To analyze the targeting of mAChR constructs, MDCK cells seeded at near-confluency (3.5 ϫ 10 5 cells/well) on 2-well glass chamber slides (4.2 cm 2 /well; Nalge Nunc International, Naperville, IL) were transfected the following day using the calcium phosphate precipitation method (36) with 4 g of receptor cDNA/well. Cells were fixed at confluence (36 -48 h post-transfection) with paraformaldehyde solution (4% (w/v) paraformaldehyde and 4% (w/v) sucrose in phosphate-buffered saline (PBS; 137 mM NaCl, 2.7 mM KCl, 4.3 mM Na 2 HPO 4 , and 1.5 mM KH 2 PO 4 ), pH 7.4) for 30 min at room temperature and processed for immunocytochemistry. Fixed cells were rinsed twice with PBST (PBS containing 0.1% (v/v) Tween 20), permeabilized with 0.25% (v/v) Triton X-100 (in PBS) for 5 min at room temperature, and blocked with 10% (w/v) bovine serum albumin in PBST containing 0.25% Triton X-100 for 2 h at room temperature. After blocking, cells were incubated with anti-FLAG M2 (1.2 g/ml), anti-IL-2R␣ (1:100; Upstate Biotechnology, Inc., Lake Placid, NY), or anti-␤-catenin (1:100; Transduction Laboratories, Lexington, KY) monoclonal antibody in PBST containing 3% bovine serum albumin and 0.25% Triton X-100 overnight at 4°C in a humid chamber. Following four washes with PBST, cells were incubated with fluorescein isothiocyanate-conjugated goat anti-mouse secondary antibody (1:250; Cappel Research Products, Durham, NC) in PBST containing 3% bovine serum albumin and 0.25% Triton X-100 for 2-3 h at room temperature. After four more washes with PBST, slides were coverslipped with Vectashield (Vector Labs, Inc., Burlingame, CA). Staining was visualized using a Nikon Optiphot 2 microscope equipped with a 60ϫ Nikon oil immersion objective. Fluorescent images were collected in both the x-y and x-z planes using a Bio-Rad MRC600 laser scanning confocal microscope. For each x-y image, a z-series of ϳ20 optical sections was taken at 0.7-m intervals from the apical to the basolateral regions of the cells. Images were projected and analyzed using Adobe Photoshop.
Quantitation of the apical/basolateral distributions of mAChR constructs was performed using the public domain NIH Image program (developed at the National Institutes of Health). The mean pixel intensity/unit area (pixel values 0 -255) of staining in the apical and basolateral domains was determined by manually outlining the areas of interest in the raw (unprocessed) x-z images. Data were processed using Microsoft Excel.
Functional Assays-Muscarinic receptor-mediated changes in forskolin-stimulated cAMP levels in transiently transfected JEG-3 cells were analyzed as described previously (37). Transfection mixtures contained (per well) 30 ng of receptor cDNA, 25 ng of ␣168-CRE-luciferase plasmid (38), 40 ng of Rous sarcoma virus-␤-galactosidase plasmid (39), 100 ng of G␣ i2 (40) in pCDPS, and 55 ng of pCDPS carrier to achieve a total of 250 ng of DNA/well. The medium was changed 20 -24 h after transfection; cells were treated with 0.4 M forskolin and various concentrations of carbamylcholine (carbachol) an additional 20 -24 h later as described (41) and lysed; and assays of luciferase and ␤-galactosidase activities were performed (37). Muscarinic receptor-mediated stimulation of phosphatidylinositol hydrolysis was determined in COS-7 cells as previously described (41) using 5 g of receptor DNA/100-mm dish for transfection.

N-[ 3 H]Methylscopolamine
Binding Assays-Cell-surface expression of mAChR constructs in transfected JEG-3 cells was determined by the binding of N-[ 3 H]methylscopolamine, a membrane-impermeable muscarinic antagonist, to intact cells as previously described (41) with the following modifications. Transfection mixtures contained (per 100-mm culture dish) 1.2 g of receptor cDNA, 1.0 g of ␣168 CRE-luciferase plasmid, 1.6 g of Rous sarcoma virus-␤-galactosidase plasmid, 4.0 g of G␣ i2 , and 2.2 g of pCDPS carrier to achieve a total of 10.0 g of DNA/dish. Cells from each dish were subcultured onto one 6-well plate 20 -24 h after transfection and allowed to attach for an additional 24 h. N-[ 3 H]Methylscopolamine binding assays were performed as described (41), except that protein content was determined by the method of Lowry et al. (42).

Differential Localization of mAChR Subtypes in MDCK
Cells-Previous studies demonstrated that mAChRs are asymmetrically distributed in a variety of polarized cells (16 -22). Despite many observations of mAChR localization, little is known concerning the cellular mechanisms and molecular signals that underlie the sorting of mAChRs to specific subcellular domains. To identify sorting signals for mAChR subtypes, we examined their targeting in MDCK epithelial cells, a widely used and well characterized model system for protein sorting. For these studies, recombinant mAChRs were FLAG-tagged at their N termini to enable immunochemical detection. The FLAG-tagged M 1 , M 2 , and M 3 receptors were expressed at levels similar to their non-tagged counterparts when transfected into COS-7 or JEG-3 cells, whereas the expression of FLAG-M 4 and FLAG-M 5 was significantly lower than that of the non-tagged receptors. 2 For receptor targeting studies, the steady-state distributions of recombinant mAChRs were analyzed in confluent MDCK cells by immunocytochemistry and confocal microscopy. The M 2 and M 3 receptors displayed reciprocal polarized distributions. Although M 2 was highly enriched on the apical membrane, M 3 was localized to the basolateral domain (Fig. 1, B and C). Although some basal M 3 staining was evident, most M 3 immunoreactivity was restricted to the lateral subdomain. In contrast, the M 1 , M 4 , and M 5 receptors exhibited non-polarized distributions, with labeling apparent throughout the cells (Fig.  1, A, D, and E). MDCK cell polarity was verified by examining the distribution of the endogenous E-cadherin-associated protein ␤-catenin, a basolateral marker (4,43). ␤-Catenin was exclusively localized to the lateral subdomain (Fig. 1F), indicating that the cells are correctly polarized under our experimental conditions. These results demonstrate that the M 2 and M 3 mAChRs are targeted to opposite domains of MDCK cells at steady state and suggest that they possess sorting signals that direct them to distinct subcellular locations.
The N-terminal Portion of the M 3 Third Intracellular Loop Contains a Basolateral Sorting Signal-The differential targeting of the M 2 and M 3 mAChRs allowed us to test the feasibility of using receptor chimeras to identify regions of the receptors important for either apical sorting of M 2 or basolateral targeting of M 3 . Since basolateral sorting signals can be dominant over apical signals when present in the same molecule (4, 10, 44), we analyzed M 2 /M 3 receptor chimeras in a gain-of-function approach to identify regions of M 3 sequence that would confer basolateral targeting to the otherwise apical M 2 receptor. Schematic diagrams of the initial set of chimeric constructs are presented in Fig. 2. Fig. 3 shows the steady-state localizations of these hybrid receptors. The first construct, M 2 /M 3 -(240 -590), contains M 3 Phe 240 -Leu 590 , encompassing the C-terminal half of the fifth transmembrane domain (TM5), the i3 loop, the sixth and seventh transmembrane domains (TM6 and TM7, respectively), and the C-terminal tail in the context of the M 2 receptor. The M 2 /M 3 -(240 -590) chimera displayed a primarily basolateral localization in MDCK cells similar to wild-type M 3 (Fig. 3, B and C). This result suggests that a region of sequence in the C-terminal half of M 3 is sufficient for basolateral targeting.
We next sought to identify the sequence containing the basolateral sorting signal by substitution of smaller regions of M 3 into the homologous positions of M 2 . M 3 Glu 486 -Leu 590 did not confer basolateral targeting to M 2, with the chimera having an apical distribution very similar to wild-type M 2 (Fig. 3, A and  D). This indicates that the M 3 basolateral sorting signal does not lie in TM6 or TM7 or in the C-terminal tail. M 2 /M 3 -(240 -485), which contains the C-terminal half of TM5 and the i3 loop of M 3 in the context of M 2 , did exhibit a primarily basolateral distribution that was very similar to wild-type M 3 (Fig. 3, B  and E), suggesting that the basolateral sorting signal lies within the M 3 i3 loop. Further dissection of the M 3 sequence confirmed this possibility. Substitution of M 3 Phe 240 -Leu 383 , containing the N-terminal half of the i3 loop, into M 2 conferred a mainly basolateral localization to the receptor molecule, although minor apical staining was also apparent (Fig. 3G). In contrast, substitution of M 3 Pro 384 -Lys 485 , comprising the Cterminal half of the i3 loop, did not confer basolateral localization, and this chimera displayed an apical distribution similar to wild-type M 2 (Fig. 3, A and F). This suggests that the M 3 basolateral sorting signal lies in the N-terminal half of the i3 loop. Furthermore, when M 3 Phe 240 -Gly 309 was substituted into M 2 , the receptor exhibited a basolateral distribution virtually indistinguishable from wild-type M 3 (Fig. 3, B and H).
These data indicate that a 70-amino acid region from TM5 and the i3 loop of M 3 contains a basolateral sorting signal that is sufficient to redirect the M 2 receptor to the basolateral domain of MDCK cells.
Having identified a region of M 3 sequence containing a putative basolateral sorting signal, we next wanted to test the role of the transmembrane residues in basolateral targeting. As shown in Fig. 4 (C and D), M 2 /M 3 -(253-296), which contains TM5 from M 2 , displayed a basolateral localization similar to wild-type M 3 , suggesting that transmembrane residues are not necessary for basolateral targeting. This observation allowed us to focus on residues located in the N-terminal portion of the M 3 i3 loop to identify a minimal sequence that provides basolateral targeting information. Since M 3 Arg 253 -Gln 296 confers basolateral targeting, we also tested Gln 297 -Gly 309 for basolateral sorting activity. As shown in Fig. 4 (B and E), M 2 /M 3 -(297-309) exhibited an apical distribution similar to wild-type M 2 , suggesting that the basolateral targeting activity lies within M 3 Arg 253 -Gln 296 . M 3 Arg 253 -Gln 269 did not confer basolateral targeting to M 2 (Fig. 4F), with the chimera having an apical distribution. However, M 2 /M 3 -(266 -296) displayed a primarily basolateral localization very similar to wild-type M 3 (Fig. 4, C  and G). These results strongly suggest that M 3 Ala 266 -Gln 296  contains a basolateral sorting signal that is sufficient to redirect the apical M 2 receptor.

The M 3 Sorting Signal Confers Basolateral Targeting to Both M 2 and IL-2R␣ and Is Position-independent-
The results above show that M 3 Ala 266 -Gln 296 contains a signal that, when substituted into the M 2 receptor, can redirect this apical receptor to the basolateral domain of MDCK cells. However, basolateral targeting of the substitution constructs could be due either to addition of a basolateral signal from M 3 or to removal of an apical signal from M 2 . To distinguish between these possibilities, M 3 Ala 266 -Gln 296 was added either to the homologous position in the i3 loop of M 2 (M 2 ϩM 3 -(i3:266 -296)) or at the C terminus, following the last amino acid of the M 2 coding sequence (M 2 ϩM 3 -(C-term:266 -296)). When analyzed by confocal microscopy, both of these receptors showed a predominantly basolateral distribution similar to wild-type M 3 (Fig. 5,  B-D). Thus, the redirection of M 2 is due to addition of the M 3 basolateral sorting signal rather than elimination of apical targeting information. Furthermore, the data show that the M 3 basolateral sorting signal acts in a position-independent manner and is dominant over any apical targeting information present in the M 2 receptor.
One important property of basolateral sorting signals is that they are autonomous, i.e. they can confer basolateral targeting to an unrelated, heterologous protein. To test whether the M 3 basolateral sorting signal acts in an autonomous fashion, we added M 3 Ala 266 -Gln 296 to the C terminus of IL-2R␣ (Tac antigen), a single-pass transmembrane protein with a short cytoplasmic tail (35). Consistent with previous results (13), wild-type IL-2R␣ had a predominantly apical distribution when expressed in MDCK cells (Fig. 5E), similar to wild-type M 2 (Fig. 5A). By contrast, a substantial fraction of the IL-2R␣/M 3 fusion protein was found in the basolateral domain, with strong immunoreactivity in the lateral subdomain, although a significant portion of the fusion protein was still detected apically (Fig. 5F). Thus, the M 3 basolateral sorting signal can at least partially redirect IL-2R␣ to the basolateral domain, suggesting that it can confer basolateral targeting to a heterologous protein.
The above results demonstrate that a 31-amino acid sequence, Ala 266 -Gln 296 , from the N-terminal portion of the M 3 i3 loop contains a basolateral sorting signal in MDCK cells. In an attempt to further define the basolateral targeting determinant, we created M 2 ϩM 3 fusion proteins in which shorter segments of M 3 sequence were fused to the C-terminal coding residue of M 2 to investigate whether basolateral targeting is lost upon removal of critical amino acids. Analysis of the steady-state distributions of these fusion proteins revealed that addition of just 21 residues of the M 3 i3 loop to the C terminus of M 2 could still redirect it to the basolateral domain (Fig. 6). Three fusion proteins, M 2 ϩM 3 -(C-term:271-296), M 2 ϩM 3 -(Cterm:266 -291), and M 2 ϩM 3 -(C-term:271-291), showed a primarily basolateral localization (Fig. 6, D-F) similar to the wild-type M 3 receptor (Fig. 6C). To exclude the possibility that addition of any 21-amino acid sequence to the C terminus of M 2 would disrupt its apical targeting and result in a basolateral distribution, M 3 Lys 570 -Leu 590 was fused to the C-terminal coding residue of M 2 . This sequence comprises the C-terminal 21 amino acids of M 3 , which did not alter the apical sorting of M 2 when included in a substitution construct (Fig. 3D). M 2 ϩM 3 -(C-term:570 -590) exhibited an apical distribution very similar to wild-type M 2 (Fig. 6, B and G), indicating that addition of a random 21-amino acid sequence does not disrupt the apical targeting of M 2 . These data show that a 21-amino acid peptide, Ser 271 -Ser 291 from the i3 loop of M 3 , provides a basolateral sorting signal capable of rerouting the otherwise apical M 2 receptor.
Quantitation of mAChR Apical/Basolateral Distributions-To confirm the basolateral targeting results described above, quantitation of the immunofluorescence signals for selected mAChR constructs in the apical and basolateral domains of MDCK cells was performed using NIH Image. As shown in Fig. 7, ϳ80% of M 3 receptor immunoreactivity was found in the basolateral domain, similar to the value obtained for the basolateral marker ␤-catenin (85%), whereas only 30% of M 2 immunoreactivity was basolateral, reflecting the predominantly apical distribution of M 2 . All M 2 /M 3 chimeras that contain either a substitution or addition of the M 3 basolateral sorting signal, Ser 271 -Ser 291 , displayed a basolateral enrichment of at least 70%, confirming the ability of the sorting signal to redirect M 2 to the basolateral domain. In contrast, M 2 /M 3 chimeras that lack these amino acids showed ϳ30% basolateral immunoreactivity, similar to wild-type M 2 . Additionally, staining for wild-type IL-2R␣ was only ϳ30% basolateral, whereas that for the IL-2R␣/M 3 fusion protein was 57% basolateral. This result confirms that the M 3 basolateral targeting determinant provides a basolateral sorting signal to IL-2R␣, although in this case, the basolateral targeting activity is not sufficient to fully override the endogenous apical signals. These data support the immunocytochemical results discussed above and confirm the ability of the M 3 basolateral sorting signal to cause significant targeting of both M 2 and IL-2R␣ to the basolateral domain of MDCK cells.
Functional Analysis of M 2 /M 3 Chimeras-The sequence containing the M 3 basolateral sorting signal overlaps a region of the i3 loop (Arg 252 -Thr 272 of rat M 3 ) shown to be important for functional coupling to the G q family of G proteins (45,46). Therefore, we tested whether addition or substitution of the M 3 basolateral sorting signal into M 2 would either confer M 3 -like coupling to G q or interfere with M 2 coupling to the G i family of G proteins. Coupling of mAChRs to the G q family of G proteins was assessed by examining their ability to activate phospholipase C in response to the muscarinic agonist carbamylcholine (carbachol). In COS-7 cells transfected with the M 3 mAChR, treatment with a maximal concentration of carbachol (1 mM) led to an ϳ2-fold stimulation of phospholipase C activity relative to untreated controls, whereas carbachol treatment of M 2transfected cells resulted in only a 1.2-fold increase in phospholipase C activity (2.36 Ϯ 0.21-and 1.24 Ϯ 0.06-fold stimulation of phospholipase C activity normalized for receptor expression for M 3 and M 2, respectively; mean Ϯ S.E., n ϭ 4). Three M 2 /M 3 chimeric receptors were tested: M 2 /M 3 -(266 -296) (substitution construct), M 2 ϩM 3 -(i3:266 -296), and M 2 ϩM 3 -(Cterm:266 -296) (addition constructs to either the i3 loop or the C-terminal tail of M 2 , respectively). None of these chimeras stimulated phospholipase C activity to a significant extent following treatment with 1 mM carbachol (1.05 Ϯ 0.02-, 1.14 Ϯ 0.07-, and 0.99 Ϯ 0.04-fold stimulation for M 2 /M 3 -(266 -296), M 2 ϩM 3 -(i3:266 -296), and M 2 ϩM 3 -(C-term:266 -296), respectively; mean Ϯ S.E., n ϭ 4). These data indicate that the M 3 basolateral sorting determinant is not sufficient to confer G q coupling to the M 2 mAChR. M 2 receptor coupling to the G i family of G proteins was assessed in JEG-3 human choriocarcinoma cells by determination of the carbachol-mediated regulation of expression of a luciferase reporter gene under the transcriptional control of a promoter containing a cAMP response element (CRE-luciferase). This system has been used extensively to measure M 2and M 4 -mediated inhibition of forskolin-stimulated adenylyl cyclase activity and cAMP production (33,34,37,41,47). Consistent with previous results (33,34,41), the M 2 receptor showed a concentration-dependent inhibition of forskolin-stimulated CRE-luciferase activity (Fig. 8). In contrast, the M 3 mAChR showed stimulation of CRE-luciferase activity (Fig. 8), presumably due to the inability of M 3 to couple to G i and to ectopic coupling of M 3 to G s . Similar results have been observed previously for the M 1 receptor in this system (33,41). All three chimeric receptors tested inhibited forskolin-stimulated CREluciferase activity in a concentration-dependent manner (Fig.  8). Whereas both substitution and addition of M 3 sequence to the M 2 i3 loop resulted in inhibition of CRE-luciferase activity to a similar extent as wild-type M 2 (51 Ϯ 4, 57 Ϯ 4, and 58 Ϯ 4% inhibition by 10 Ϫ5 M carbachol for M 2 , M 2 /M 3 -(266 -296), and M 2 ϩM 3 -(i3:266 -296), respectively; mean Ϯ S.E., n ϭ 4), addition of M 3 sequence to the C terminus of M 2 (M 2 ϩM 3 -(C-term: 266 -296)) resulted in 35 Ϯ 6% inhibition by 10 Ϫ5 M carbachol (mean Ϯ S.E., n ϭ 4). This difference could be due either to reduced functional coupling or to lower cell-surface expression of the fusion protein compared with wild-type M 2 . To distinguish between these possibilities, we determined the level of each receptor at the cell surface using N-[ 3 H]methylscopolamine, a membrane-impermeable muscarinic antagonist. Under the same transfection conditions used in the functional assay, the M 2 /M 3 chimeras were expressed at similar levels (358 Ϯ 17, 381 Ϯ 2, and 356 Ϯ 52 fmol/mg protein for M 2 /M 3 -(266 -296), M 2 ϩM 3 -(i3:266 -296), and M 2 ϩM 3 -(C-term:266 -296), respectively; mean Ϯ S.E., n ϭ 3), which were slightly higher than those for wild-type M 2 and M 3 (215 Ϯ 41 and 261 Ϯ 30 fmol/mg protein, respectively; mean Ϯ S.E., n ϭ 3). Therefore, the slightly reduced ability of the M 2 ϩM 3 -(C-term:266 -296) receptor to inhibit forskolin-stimulated CRE-luciferase activity is likely due to a slightly reduced efficiency of coupling to G␣ i2 . Taken together, the data suggest that the M 2 /M 3 chimeric receptors fold correctly, reach the cell surface, and stimulate a functional response similar to the wild-type M 2 mAChR. Thus, addition or substitution of the M 3 basolateral sorting signal does not appear to substantially alter M 2 receptor conformation or function. DISCUSSION The goal of this study was to characterize the molecular mechanisms and signals involved in the polarized targeting of mAChR subtypes. Examination of the steady-state distributions of mAChRs in MDCK cells revealed that the M 2 and M 3 receptors are targeted to opposite domains (Fig. 1). This is the first demonstration of differential sorting of highly homologous members of a single G protein-coupled receptor family in MDCK cells. We utilized a gain-of-function approach to identify a basolateral sorting signal in the M 3 receptor by analysis of M 2 /M 3 chimeric receptor constructs. The use of chimeras between closely related proteins with opposite phenotypes is advantageous over studies using deletion or truncation mutagen- esis because it greatly reduces the possibility that a loss of receptor targeting is due to a generalized effect on protein structure. This consideration is especially important for polytopic membrane proteins such as G protein-coupled receptors.
The M 3 basolateral sorting signal is contained within a 21amino acid sequence, Ser 271 -Ser 291 , from the N-terminal portion of the M 3 i3 loop. Addition of this signal to the M 2 receptor in either the i3 loop or the C-terminal tail caused M 2 to be redirected from the apical domain to the basolateral domain of MDCK cells, whereas addition of an irrelevant 21-residue sequence did not alter the apical distribution of M 2 . Furthermore, substitution of a 70-amino acid region of M 3 containing the basolateral determinant also conferred basolateral targeting to the otherwise non-polarized M 1 mAChR. 3 Together, the data show that M 3 Ser 271 -Ser 291 contains a basolateral sorting signal that acts in a position-independent manner and is dominant over targeting signals in other mAChR subtypes. Interestingly, this sequence conferred partial basolateral targeting when transferred to the apical IL-2R␣; although a substantial fraction of the chimeric molecules were redirected to the basolateral domain, a detectable fraction remained apical. The incomplete basolateral targeting of IL-2R␣ by the M 3 basolateral sorting signal suggests that the targeting activity of this basolateral determinant is not sufficient to completely override the activity of the apical signals in IL-2R␣, although it can completely counteract the apical signals in M 2 . Thus, the apical targeting information in IL-2R␣ may be stronger than that in M 2 , perhaps due either to higher signal strength or to a greater number of apical targeting determinants. This notion is consistent with recent suggestions that basolateral sorting signals may not always be dominant over apical signals and that the overall targeting phenotype of a protein may be determined by the relative strength (9) or valence (48) of multiple sorting signals. Alternatively, we cannot rule out the possibility that incomplete basolateral targeting of the IL-2R␣/M 3 fusion protein is due to saturation of the basolateral targeting pathway and spillover of excess protein into the apical pathway, perhaps resulting from higher expression of IL-2R␣/M 3 as compared with the M 2 /M 3 chimeras.
The M 3 basolateral sorting signal identified in this study is sufficient to confer basolateral targeting to other mAChR subtypes. However, an M 3 deletion mutant lacking Ala 266 -Gln 296 displayed a basolateral distribution virtually identical to that of the wild-type M 3 receptor. 2 Thus, the Ser 271 -Ser 291 domain is not the only region of M 3 that can mediate its basolateral targeting. Our results suggest that the sequence identified here is the strongest in terms of conferring basolateral targeting activity to heterologous proteins, but other elements of the M 3 receptor may mediate its basolateral sorting in the absence of this signal. These other M 3 basolateral sorting elements may not be strong enough to counteract the apical signals in M 2 and so would not be detected in our experimental approach, but may mediate basolateral targeting of M 3 in the absence of any additional signals. Consistent with this idea, it has been reported that the polarized sorting of other proteins can be mediated by multiple, independent targeting motifs (8,10,49). FIG. 7. Quantitation of the steadystate distributions of mAChR constructs in MDCK cells. Quantitation of the immunofluorescence signals for selected mAChR constructs and ␤-catenin in the apical and basolateral domains of MDCK cells was performed using NIH Image as described under "Experimental Procedures." Data represent the mean fluorescence intensity per unit area in the basolateral domain and are expressed as the percentage of total fluorescence intensity in the apical and basolateral domains. Data are plotted as the mean of two or the mean Ϯ S.E. of three to seven images for each construct. One common feature of basolateral sorting signals is that they sometimes overlap with sequences involved in endocytosis from the plasma membrane (10 -12, 44, 50). However, the M 3 basolateral sorting signal, which resides in the N-terminal portion of the i3 loop, does not coincide with known internalization motifs for M 3 or the highly homologous M 1 receptor, which are located in the middle and the immediate membraneproximal portions of the loop (51,52). Targeting studies of other proteins have also revealed that basolateral sorting signals can be spatially distinct from endocytosis signals (4,13,27,53,54).
The M 3 basolateral sorting signal has a 3-amino acid overlap (Ser 271 -Thr 273 ) with a membrane-proximal region of the i3 loop implicated in M 3 receptor coupling to the phospholipase C pathway via the G q family of G proteins (45,46). For this reason, we tested whether the basolateral targeting motif could also confer coupling to phospholipase C. M 2 /M 3 chimeras containing the basolateral sorting signal did not stimulate phosphatidylinositol turnover, demonstrating that it is not sufficient for coupling to G q proteins. This is not surprising, as there is minimal overlap between the two motifs, and G protein coupling of mAChRs is thought to be mediated by a multisite domain (46). The basolateral targeting motifs for other G protein-coupled receptors are also distinct from their functional G protein-coupling domains (27,53,55).
Basolateral targeting signals have now been identified in many transmembrane proteins. Although no consensus sequence exists, structural determinants such as tyrosine-based (10,27,49,50,56) or dihydrophobic (11)(12)(13)49) motifs are often found in basolateral sorting signals. However, some basolateral targeting sequences act independently of these motifs (7,54,57,58) or do not contain them at all (4,59). The M 3 basolateral sorting signal (Ser 271 -Ser 291 ) does not contain a critical tyrosine or a dihydrophobic motif. Thus, the sequence identified here may represent a novel basolateral targeting determinant. Either the continuous amino acid sequence itself or a threedimensional epitope formed by noncontiguous elements within the sequence may form the actual signal for basolateral targeting.
Heterologous protein expression studies have suggested that epithelial cells and neurons use common cellular mechanisms to generate a polarized distribution of membrane proteins. Based on the sorting of viral glycoproteins, it was initially proposed that the apical domain of epithelial cells corresponds to neuronal axons, whereas the basolateral domain corresponds to cell bodies and dendrites (60). Although this parallel does not hold true for all proteins (61,62), recent studies have confirmed that some basolateral proteins in MDCK cells are restricted to the somatodendritic domain of cultured hippocampal neurons and that the same signals used for basolateral targeting are also likely to mediate somatodendritic targeting (63). However, proteins that are apical in MDCK cells are not restricted to the axon, but instead are distributed uniformly throughout the axon and dendrites of cultured hippocampal neurons (63). These and other studies have suggested the existence of "axon-including" signals, rather than signals that mediate targeting to the axon exclusively (64). The differential targeting of the M 2 and M 3 mAChRs in MDCK cells (Fig. 1) and neurons in vivo (24,25) suggests that similar signals may operate to achieve polarized sorting of mAChRs in epithelial cells and neurons. While the basolateral sorting signals in M 3 may mediate its localization to the somatodendritic domain, the apical targeting information in M 2 may allow its inclusion in axons. It will be of interest to determine whether the signals that mediate mAChR targeting in MDCK cells are also important for the differential localization of mAChR subtypes in neurons.
In conclusion, we have identified a novel 21-amino acid sequence from the N-terminal portion of the M 3 mAChR i3 loop that mediates basolateral targeting in MDCK cells. This determinant, although not uniquely necessary for the basolateral targeting of M 3 , is dominant over apical sorting signals in the M 2 mAChR and can be transferred to a heterologous protein, IL-2R␣, in an autonomous fashion. The findings reported here add significantly to our knowledge of the signals that underlie the polarized targeting of G protein-coupled receptors. Future work will be aimed at further elucidation of the molecular signals and cellular machinery involved in mAChR sorting in both epithelial cells and neurons.