Identification and structural determination of the M(3) muscarinic acetylcholine receptor basolateral sorting signal.

Muscarinic acetylcholine receptors comprise a family of G-protein-coupled receptors that display differential localization in polarized epithelial cells. We identify a seven-residue sequence, Ala(275)-Val(281), in the third intracellular loop of the M(3) muscarinic receptor that mediates dominant, position-independent basolateral targeting in Madin-Darby canine kidney cells. Mutational analyses identify Glu(276), Phe(280), and Val(281) as critical residues within this sorting motif. Phe(280) and Val(281) comprise a novel dihydrophobic sorting signal as mutations of either residue singly or together with leucine do not disrupt basolateral targeting. Conversely, Glu(276) is required and cannot be substituted with alanine or aspartic acid. A 19-amino acid peptide representing the M(3) sorting signal and surrounding sequence was analyzed via two-dimensional nuclear magnetic resonance spectroscopy. Solution structures show that Glu(276) resides in a type IV beta-turn and the dihydrophobic sequence Phe(280)Val(281) adopts either a type I or IV beta-turn.

From the ‡Departments of Pharmacology and ¶Biochemistry, University of Washington, Seattle, Washington 98195 Muscarinic acetylcholine receptors comprise a family of G-protein-coupled receptors that display differential localization in polarized epithelial cells. We identify a seven-residue sequence, Ala 275 -Val 281 , in the third intracellular loop of the M 3 muscarinic receptor that mediates dominant, position-independent basolateral targeting in Madin-Darby canine kidney cells. Mutational analyses identify Glu 276 , Phe 280 , and Val 281 as critical residues within this sorting motif. Phe 280 and Val 281 comprise a novel dihydrophobic sorting signal as mutations of either residue singly or together with leucine do not disrupt basolateral targeting. Conversely, Glu 276 is required and cannot be substituted with alanine or aspartic acid. A 19-amino acid peptide representing the M 3 sorting signal and surrounding sequence was analyzed via two-dimensional nuclear magnetic resonance spectroscopy. Solution structures show that Glu 276 resides in a type IV ␤-turn and the dihydrophobic sequence Phe 280 Val 281 adopts either a type I or IV ␤-turn.
Targeting of proteins to their proper subcellular domains is critical to the functioning of polarized cells. Targeting of proteins and lipids to the apical or basolateral plasma membrane of polarized epithelial cells allows for differentiation of the membrane into functionally and biochemically distinct domains. Understanding the mechanisms responsible for this highly organized cellular process is a fundamental question in cell biology. Many membrane proteins contain sorting signals that direct them to a particular membrane domain. Some signals can be recognized by cellular machinery within the biosynthetic or recycling pathways and then directed to the proper domain. Other types of signals function by partitioning the membrane protein into a particular microenvironment. Apical sorting signals can consist of N-or O-glycans, a glycosylphophatidylinositol anchor or protein sequences in extracellular, transmembrane, or cytoplasmic domains (1). The apical sorting of membrane proteins can also be mediated by lipid rafts (2,3). Basolateral sorting signals reside in the cytoplasmic domains of membrane proteins, and they can consist of tyrosine-based motifs, dihydrophobic motifs, acidic residues, or other, atypical motifs (4 -8).
Little is known about the three-dimensional structure of basolateral targeting signals. The atypical basolateral sorting signal of the polyimmunoglobulin receptor exists as a ␤-turn and a nascent helix, with a critical valine residue in the nascent helix (8). The two dihydrophobic sorting motifs of the major histocompatibility complex-associated invariant chain reside in a nascent helix and a turn with leucine-isoleucine in the nascent helix and methionine-leucine part of a turn (9). Some basolateral sorting signals have functional overlap with tyrosine-based endocytosis signals, and the sequence of a number of basolateral sorting signals closely resembles some endocytosis signals (10). The tyrosine-based endocytosis signals of both the low density lipoprotein receptor and lysosomal acid phosphatase were shown to adopt a ␤-turn conformation (11,12).
In this report we have further characterized the M 3 BLSS using alanine mutagenesis, conservative residue substitution, and two-dimensional nuclear magnetic resonance (NMR). We have determined the individual amino acid residues required for the functioning of this signal as well as its tertiary structure. The M 3 BLSS signal consists of seven amino acids including a single glutamic acid upstream of a novel phenylalaninevaline (FV) dihydrophobic motif. The novel sorting signal is sufficient for basolateral targeting, as it can be appended to either M 2 or IL-2R␣ and redirect these apical proteins to the basolateral membrane of MDCK cells. Phe 280 and Val 281 can be mutated either individually or in combination to leucine and maintain basolateral sorting function, constituting a novel dihydrophobic motif. Glu 276 is essential for the M 3 BLSS, as a receptor chimera with a mutation of Glu 276 either to alanine or to aspartic acid displayed apical sorting. The structure of this novel sorting signal was determined by two-dimensional NMR. Glu 276 resides in a type IV ␤-turn, and the dihydrophobic motif, Phe 280 and Val 281 , exists in either a type I or IV ␤-turn.

EXPERIMENTAL PROCEDURES
Construction of Chimeric mAChRs-N-terminal FLAG epitopetagged pFM 3 and pFM 2 mAChRs in the expression vector pCD-PS were described previously (23). pFM 2 and pFM 3 were used as templates in sequential PCR (24) Table I). All PCRamplified constructs were engineered with MscI and EcoRI sites at their 5Ј-and 3Ј-ends, respectively, and cloned into the MscI and EcoRI sites of the parental pFM 2 construct.
A unique BspEI restriction site was created in M 2 ϩM 3 -[C-term. 271-291] at nucleotides 811-816 of the M 3 sequence by mutating two sequential wobble bases, nucleotides 813 and 816 from T to C and G to A, respectively. These mutations did not alter the sequence of the protein product. This BspE1 site and an EcoRI site were then digested and used as insertion sites for eight sets of annealed complimentary oligonucleotide pairs containing the M 3 sequence with each amino acid mutated singly to alanine. The M 3 sequences of the oligo pairs were flanked by a 5Ј BspE1 and 3Ј EcoR1 site. Two oligo pairs that changed alanine 275 either to methionine or glycine were also made (Table I).
For annealing of oligo pairs, equimolar amounts of complementary oligos were mixed together on ice and were then placed in a 72°C heat block for 2 min. The heat block was then turned off and allowed to cool to room temperature, at which time the annealed oligos were removed and used in a ligation reaction with appropriately digested vector.
Chimeric receptors were generated in which the M 3 sequence, Ala 275 -Val 281 , or M 3 sequence containing a point mutation, F280A, was added to the C terminus of M 2 by PCR using pFM 2 as template. Primers for the M 2 /M 3 fusion protein had MscI and EcoRI sites at their 5Ј-and 3Ј-ends, respectively, and the PCR products were cloned into the MscI and EcoRI sites of the parental pFM 2 plasmid. IL-2R␣ receptor/M 3 fusion proteins were generated by PCR using IL-2R␣/pCD as template (23). PCR products were engineered to contain a 5Ј BglII and a 3Ј EcoRI site and were cloned back into the BglII and EcoRI sites of the parental pCD plasmid.
PCR was used to generate a set of receptor chimeras with mutated M 3 sequence appended at the C terminus of wild-type (WT) M 2 . PCR products were engineered with MscI and EcoRI sites at their 5Ј-and 3Ј-ends, respectively, and were cloned into to the MscI and EcoRI sites of the pFM 2 plasmid.
Cell Culture-MDCK cells (strain II) were provided by Dr. Keith Mostov (University of California at San Francisco). MDCK cells were maintained in Dulbecco's modified Eagle's medium 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-MDCK cells were seeded on 2-well glass chamber slides (4.2 cm 2 /well; Nalge Nunc International, Naperville, IL). Cells were plated at subconfluency (3.5 ϫ 10 5 cells/well) before transfection via calcium phosphate or at 1.0 ϫ 10 6 cells/well before transfection via the Lipofectamine (Invitrogen) method. Cells were transiently transfected the following day with 2-4 g of cDNA/well. For transfections with Lipofectamine reagent, 4 g of cDNA plus 8 l of reagent was used. Twenty-four hours after achieving confluency and 48 h after transfection the cells were fixed with 4% paraformaldehyde. The cells were permeabilized, blocked, and then incubated with monoclonal anti-FLAG M2 (1.2 g/ml) (Sigma) or anti-IL2R␣ (1:100, Upstate Biotechnology, Inc., Lake Placid, NY) followed by incubation with fluorescein isothiocyanate-or ALEXA 548-conjugated goat anti-mouse antibody, respectively (Cappel Research Products, Durham, NC and Molecular Probes, Eugene, OR). Fluorescent images were collected in both the x-y and x-z planes on a Leica TCS SP1/NT confocal microscope (Leica Microsystems, Inc., Exton, PA) using 100ϫ, 1.4 N.A. oil immersion lens. Images of the x-y plane are a projected z-series generated from ϳ16 -25 optical sections taken at ϳ0.5-m steps from the apical to basolateral borders of the cells.
Peptide Synthesis-A 19-amino acid peptide representing the Nterminal portion of the i3 loop of M 3 (residues Ser 271 -Ser 289 ) (modified with N-terminal acetylation and C-terminal amidation) was synthesized (Invitrogen). The peptide purity was Ͼ95% as assessed by high performance liquid chromatography and mass spectroscopy.
NMR Data Collection-Lyophilized M 3 peptide was dissolved in 50 mM sodium phosphate buffer at pH 6.4 containing 0.1 mM EDTA, 1 mM NaN 3 , and 10% D 2 O. For NMR spectra collected in D 2 O, the lyophilized peptide was dissolved in the same buffer with 99.9% D 2 O. Sample concentrations were between 1 and 2 mM, as determined by weight. NMR spectra were acquired on a Bruker DMX 500 MHz spectrometer equipped with a triple-resonance, triple-axis gradient probe. Two-dimensional spectra collected include 1) total correlation spectroscopy collected in H 2 O and D 2 O (25), 2) nuclear Overhauser spectroscopy (NOESY) (26), and 3) double-quantum-filtered correlated spectroscopy (COSY) (27). Total correlation spectroscopy spectra were used to assign intra-residue protons by observation of the intra-residue through-bond connections. The NOESY spectrum was used to observe inter-residue through-space connections. Sequential NH and C␣H assignments were obtained by analysis of cross-peaks in the NH-C␣H region in the NOESY spectrum.
NH Chemical Shift Temperature Gradients-A series of NOESY spectra was obtained at 5, 15, 25, 35, and 45°C. NH chemical shift values were determined from the NOESY spectra for each residue (except Pro 283 ) at the respective temperatures. The measured values were subtracted from previously published NH random coil chemical shift values and plotted as the difference in chemical shift (expressed in ppb) versus temperature (expressed in°C) (28,29). Linear regression was applied to each residue across all temperatures, and the resulting slope is reported as the temperature gradient. His 282 and Arg 288 were not included in the final analysis because fewer than three of the five possible temperature data points were observed.
Calculation of Structures-58 inter-residue NOE distance restraints derived from a NOESY spectrum collected at 12°C with a mixing time of 300 ms and 5 dihedral angle restraints derived from a doublequantum-filtered-COSY spectrum were used as input for CNSsolvebased simulated-annealing distance geometry structure calculations (30). Integrated cross-peak volumes and intensities were evaluated and used to separate NOEs into bins with overlapping distance restraints: strong 2.0 -3.5 Å, medium 2.0 -4.5 Å, and weak 2.0 -5.5 Å. 200 structures were calculated and ranked by total energy. The top 10 structures were evaluated by VMD-XPLOR and PROCHECK_NMR (31,32). Peptide ensemble images were created using MOLMOL (33). The electrostatic surface potential image was generated using PyMOL (34).

The M 3 Basolateral Sorting Signal Resides between Residues
Ala 275 and His 282 -Previous work demonstrated that mAChRs are asymmetrically distributed in MDCK cells. M 3 was localized to the basolateral domain, whereas M 2 was localized to the apical domain. The reciprocal polarity of these two related receptors was utilized to generate receptor chimeras of M 2 and M 3 . M 3 was found to contain a 21-residue BLSS in the Nterminal portion of the i3 loop that, when substituted into or appended onto M 2 , could redirect M 2 from the apical domain to the basolateral domain of MDCK cells (23).
We used the full-length M 2 receptor with this 21-amino acid region appended to its C terminus to further define the M 3 BLSS. We generated mutations within the receptor chimera, M 2 ϩM 3 [C-term 271-291], mutating sections of the 21-amino acid M 3 basolateral signal to alanines (Table I). Five receptors were generated by mutating groups of five adjacent and overlapping amino acids within the M 3 sequence of M 2 ϩM 3 [Cterm. 271-291] to alanine residues (Fig. 1A, schematic). Each cDNA was transiently transfected in MDCK cells, and the steady state localization of the protein product was determined via immunofluorescence using confocal microscopy. Glu 276 , Phe 280 , and Val 281 Are the Critical Amino Acids of the M 3 BLSS-To identify the specific residues critical to the M 3 basolateral sorting signal, amino acids Ala 275 -His 282 were individually mutated to alanine by site-directed mutagenesis (Fig. 2). Ala 275 was changed to methionine, since in the previous experiment this WT alanine had not been altered by mutagenesis. The point mutations were made within the context of the same parent receptor construct described above, M 2 ϩM 3 [C-term 271-291], and the WT M 3 sequence Ser 271 -Glu 274 and Pro 283 -Ser 291 remained unchanged. Each of the eight cDNAs was transiently transfected in MDCK cells, and the localization of the protein product was determined as described previously. Five of the eight mutant proteins, A275M, T277A, E278A, N279A, and H282A, were localized to the basolateral membrane ( (Fig. 2, A, C, D, E, and H), indicating that these residues are not critical to the function of the BLSS. The remaining three mutant proteins, E276A, F280A, and V281A, (Fig. 2, B, F, and G) were detected at the apical membrane of MDCK cells along with a dramatic loss of targeting of these proteins to the basolateral membrane. Non-conservative amino acid changes at either Glu 276 , Phe 280 , or Val 281 abrogates the BLSS function, and thus, these residues are critical to the function of the M 3 BLSS.
The Seven-amino Acid BLSS of M 3 Confers Basolateral Targeting to Both M 2 and IL-2R␣-In the experiment above three residues were identified as critically important to the function of the M 3 BLSS. Experiments involving chimeric receptors harboring the five sequential and overlapping alanine mutations showed that amino acids N-terminal to Ala 275 and Cterminal to His 282 are not necessary for the function of the BLSS. We next asked if just seven amino acids, Ala 275 -Val 281 , when appended to the C terminus of M 2 , would still function as a minimal BLSS and be able to redirect M 2 from its WT apical distribution to the basolateral membrane. This cDNA, M 2 ϩM 3 [C-term 275-281], was transiently transfected in MDCK cells, and the steady state localization of the protein was determined. As seen in Fig. 3B, the chimeric receptor, M 2 ϩM 3 [C-term 275-281], had a predominantly basolateral distribution opposite of the apically distributed WT M 2 (Fig. 3A). Another chimera, M 2 ϩM 3 [C-term 275-281] F280A, harboring an alanine point mutation at Phe 280 , was also generated, and its membrane localization was determined. As expected from   (Table I)  the results seen with mutations of the 21-amino acid sequence, this receptor chimera was mistargeted to the apical domain when expressed in MDCK cells (Fig. 3C), its distribution strongly resembling that of both M 2 ϩM 3 [C-term 271-291]F280A (Fig. 2F) and WT M 2 (Fig. 3A).
To test whether this minimal M 3 BLSS can act on a completely unrelated protein, we added M 3 275-281 to the C terminus of IL-2R␣, a heterologous, single-pass transmembrane protein with an apical distribution in MDCK cells (35,36). IL-2R␣ and IL-2R␣ϩM 3 [C-term 275-281] were transfected in MDCK cells, and their localization was imaged. As expected, IL-2R␣ was targeted to the apical membrane (Fig. 3D), similar to WT M 2 . In contrast, the majority of IL-2R␣ ϩ M 3 [C-term 275-281] protein was localized at the basolateral membrane with some intracellular staining also evident (Fig. 3E). This same receptor chimera, with a single alanine mutation at Phe 280 , IL-2R␣ ϩ M 3 [C-term 275-281] F280A, was clearly targeted to the apical membrane (Fig. 3F).
These data demonstrate that a seven-residue sequence, Ala 275 -Val 281 , from the i3 loop of M 3 contains basolateral sorting information that can redirect both M 2 and the heterologous IL-2R␣ from the apical to basolateral membrane of polarized epithelial cells. The signal is dominant over apical targeting information present in either M 2 or IL-2R␣. These data also show that Phe 280 is a critical residue within the signal as the 7-mer fails to redirect either M 2 or IL-2R␣ when it harbors a non-conservative mutation at this position.
The M 3 BLSS Consists of an Acidic Residue, Glu 276 , and a Dihydrophobic Motif, Phe 280 -Val 281 -The importance of dihydrophobic motifs in basolateral sorting is well documented. To date dihydrophobic motifs have been shown to be composed of LL, LI, LV, FI, and ML (7,9,35,(37)(38)(39). It is also known that some basolateral sorting signals consist of a dihydrophobic motif in conjunction with a nearby, usually N-terminal, acidic cluster or acidic residue (9,35). Our data suggest FV may be a previously unidentified dihydrophobic basolateral sorting signal that may work in conjunction with the upstream glutamic acid residue at position 276. To determine whether FV represents a novel dihydrophobic motif and if glutamic acid at position 276 is critical, we generated chimeric M 2 /M 3 receptors harboring various combinations of conservative amino acid mutations at residues Glu 276 , Phe 280 , and Val 281 . For these experiments, aspartic acid was substituted for Glu 276 , and leucine was substituted for Phe 280 and Val 281 .
Five chimeras were generated in all: M 2 ϩM 3 (E276D), M 2 ϩM 3 (FV-LL), M 2 ϩM 3 (F280L), M 2 ϩM 3 (V281L), and M 2 ϩM 3 (EFV-DLL). After transfection, the localization of these proteins was determined as described above. Only M 2 ϩM 3 (E276D) displayed an apical distribution (Fig. 4B), whereas the other four constructs displayed a basolateral distribution (Fig. 4, C-F). When the dihydrophobic residues were mutated individually or in combination to leucine, the distribution of the receptors harboring these conservative mutations was unchanged and resembled the distribution of M 2 ϩM 3 [C-term 275-281], a receptor with an intact M 3 WT sequence (Fig. 4A). Therefore, leucine residues can substitute for either phenylalanine or valine within this region of the M 3 BLSS. In contrast, aspartic acid does not substitute for Glu 276 despite the conservation of charge. The receptor chimera, M 2 ϩM 3 [C-term 275-281] (EFV-DLL) has all three critical residues of the BLSS mutated to the conservative amino acids stipulated above. Despite the presence of the glutamic to aspartic acid mutation, this receptor was still targeted to the basolateral membrane (Fig. 4F).
Two-dimensional NMR Structure Analysis of the M 3 mAChR BLSS-To determine the structural features that contribute to  the basolateral sorting function, the structure of a 19-residue peptide that corresponds to residues Ser 271 -Ser 289 of the i3 loop of M 3 was analyzed by NMR spectroscopy. This peptide was chosen because it contained most of the residues in the 21amino acid sequence originally identified by Nadler et al. (23) to contain the BLSS. NH-NH inter-residue cross-peaks in a NOESY spectrum gave the first indication that the peptide was structured in solution (Fig. 5A). Highly flexible, unstructured peptides do not exhibit NH-NH interactions (40). To extract and interpret the structural information contained within the NOESY spectrum, assignment of each NMR resonance to a specific proton in the peptide was required. This was accomplished using conventional homonuclear two-dimensional NMR spectra as described under "Experimental Procedures." Most backbone and side-chain protons were unambiguously assigned; however, some side-chain spectral overlap did occur.
With resonance assignments in hand, the NH-NH NOE peaks observed in the NOESY spectrum could be assigned to specific unique pairs of amide groups within the polypeptide. A total of eight pairs were observed, each of which reflects a short distance between two neighboring amide groups within the sequence. Sequential NH-NH cross-peaks occur for residues starting with Gly 272 and ending with His 282 . The stretch of consecutive NH-NH NOEs beginning with Asn 279 can be extended through Thr 284 by inclusion of NOE cross-peaks attributed to the C␦H protons of Pro 283 that have cross-peaks with NH protons of His 282 and Thr 284 (Fig. 5B). Typically, a string of sequential NH-NH connections is a strong indicator of ␣-helical structure. However, there are breaks in the consecutive NH-NH connections; no peaks are observed for Ala 275 and Glu 276 or for Glu 278 and Asn 279 (40).
Additional inter-residue cross-peaks were assigned in the NOESY spectrum ( Fig. 5B and Table II). A majority of the assigned cross-peaks consist of side-chain/side-chain, sidechain/backbone, and backbone/backbone proton NOEs of consecutive residues. Extensive contacts, observed by the density of NOEs, exist between Ala 275 and Glu 276 and between Phe 280 and Val 281 . These observations indicate structured regions of the peptide. Our ability to observe additional NOEs for His 282 was limited by the overlap of its C␣H resonance with the resonance of H 2 O. Several long-range i to iϩ2 and iϩ3 are also observed (Fig. 5B) Hatched bars indicate NOEs observed between i and iϩ3 residues. White bars indicate NOEs observed between residues greater than three residues apart. Panel C, correlation of NH chemical shift with temperature. NH chemical shift temperature gradients (⌬␦/⌬T) are plotted versus residues in the peptide. ⌬␦/⌬T values were derived by plotting NH chemical shifts (ppm) versus temperature (°C). Linear regression was applied to the data set for each residue, and the resulting slope equaled the ⌬␦/⌬T value expressed in ppb/ o C. Slopes could not be obtained for His 282 , Pro 283 , and Arg 288 . a Distances are categorized by upper boundary distances (x) set during structure calculations: short (x Յ 3.5 Å), medium (3.5 Ͻ x Յ 4.5 Å), and long (x Ͼ 4.5 Å).
b Defined using residues listed with respect to the lowest energy structure.
and Gly 272 and from Gly 285 through Ser 289 , indicating unstructured regions of the peptide. NOEs between the i C␣H and the iϩ3 C␤H proton(s) are signatures for ␣-helical structures (40). Many of these NOEs could not be unambiguously identified due to overlap within the spectra. Nevertheless, where overlap was not an issue, NOEs were absent. Double-quantum-filtered-COSY spectra were collected to obtain 3 J HN␣ coupling constants that provide information about backbone dihedral angles for each residue (40). 3 J HN␣ coupling constants that are greater than 6 -8 Hz indicate dihedral angles that correspond to an extended conformation as seen in ␤-strands. 3 J HN␣ coupling constants that are less than 6 -8 Hz indicate dihedral angles consistent with a kinked conformation as seen in ␣-helices (3-4 Hz). Two residues, Ala 275 and Glu 276 , have 3 J HN␣ coupling values around 6 Hz, and three residues, Glu 274 , Val 281 , and Gly 285 , have 3 J HN␣ coupling values above 9 Hz (Table II).
To further investigate the structural properties of the peptide, the effect of temperature on the degree of protection of the backbone amide protons from exchange with solvent was determined. The change in chemical shift of the backbone amide protons with the change in temperature, denoted as ⌬␦/⌬T, specifies exchange-protection provided by exclusion from solvent in structured peptides (29). ⌬␦/⌬T values more negative than Ϫ7 ppb/°C indicate backbone amide protons present in random coils with reduced protection from exchange with solvent. ⌬␦/⌬T values less negative than Ϫ7 ppb/°C are indicative of backbone amides that are protected from exchange with solvent. Values less negative than Ϫ4 ppb/°C indicate hydrogen-bonding. A series of total correlation spectroscopy spectra obtained at 5, 15, 25, 35, and 45°C enabled us to follow the chemical shifts of individual NH protons within the peptide. The resulting data revealed a general trend (Fig. 5C). Residues N-terminal to Val 281 have an average ⌬␦/⌬T value between Ϫ4 and Ϫ7 ppb/°C, indicating a moderate degree of backbone amide exchange protection. In particular, Glu 274 and Glu 276 have ⌬␦/⌬T values of Ϫ5.1 ppb/°C, and Phe 280 has a ⌬␦/⌬T value of Ϫ4.4 ppb/°C, suggesting that these residues experience a larger degree of backbone amide exchange protection. In contrast, residues C-terminal to Val 281 have an average ⌬␦/⌬T value more negative than Ϫ7 ppb/°C. The temperature gradient values observed are consistent with the density of NOEs associated with the residues. Again, information for His 282 could not be obtained due to its resonance being obscured by H 2 O.
NMR Solution Structure-Taken together, the NMR data indicate that the peptide does not adopt regular secondary structure that can be inferred directly. Thus, it is unlikely that the M 3 BLSS adopts an ␣-helical or ␤-sheet secondary structure. Similarly, circular dichroism spectra also indicated a lack of ␣-helix or ␤-sheet structures (data not shown). Nevertheless, the data are consistent with the peptide adopting a stable structure that can be defined by the NMR parameters. Distance geometry and simulated annealing calculations were performed using distance restraints generated from NOESY data and dihedral angle restraints generated from COSY data. Two hundred structures were calculated and ranked according to their total energy. The top 10 low energy models were evaluated by VMD-XPLOR for violations of NOE restraints and by PROCHECK for violations of / angles. Two distinct structured regions of the peptide are observed in the ensemble of accepted structures (Fig. 6A). Region 1 consists of residues Thr 273 through Glu 276 ; region 2 consists of residues Asn 279 through Thr 284 . Overlays of the ensemble with respect to region 1 or region 2 demonstrate the conformational flexibility between the two structured regions. Thr 277 and Glu 278 are positioned between the two structured regions and sample the greatest range of / angles (Fig. 7A). Residues adjacent to Thr 277 and Glu 278 demonstrate reduced conformational flexibility (Fig. 7, B-D). The backbone trace of region 1 shows that all structures in the ensemble adopt a similar backbone conformation whereas side-chain atoms of residues within the region experience greater variations (Table II and data not shown). Analysis by PROMOTIF of possible motifs adopted in region 1 indicate that the region can adopt a Type IV ␤-turn involving different residues. Backbone and side-chain atoms within region 2 have the highest number of restraints and the least variation among solution structures (Fig. 6, A-B, and Table II). Analysis by Promotif indicates that a type IV ␤-turn is adopted for residues Phe 280 (i) through Pro 283 (iϩ3), and a type I ␤-turn is predominantly adopted for residues Val 281 (i) through Thr 284 (iϩ3).
In summary, the NMR data define the structure adopted within the peptide as composed of two ␤-turn structured regions that are flexible around two residues linking the two regions. The orientation of the individual turns relative to one another does not appear to be well defined in this fairly short peptide in the absence of a binding partner. An electrostatic surface potential of the calculated solution structures high- lights charge differences between the two regions (Fig. 6C). A representative structure has been selected for illustration purposes; however, due to the conformational flexibility around Thr 277 and Glu 278 , the relative orientation of the two regions will vary between structures. With this in mind, in the ensemble of structures, the side chains of Glu 274 and Glu 276 in region 1 and Glu 278 are presented along the exterior of the turn-like structure and create a large negatively charged surface. Residues in region 2 are mostly hydrophobic, and the ␤-turn structures create a large exposed hydrophobic surface.

DISCUSSION
In this study we have systematically examined the sequence requirements and structural features of the M 3 mAChR BLSS. We identified a seven-residue signal that consists of a single glutamic acid residue N-terminal to a novel FV dihydrophobic motif. NMR analyses determined that these critical amino acids reside in neighboring type I or type IV ␤-turns.
Experiments involving chimeric muscarinic receptors in combination with alanine mutagenesis identified three critical residues within the sorting signal. We determined that both hydrophobic residues, Phe 280 and Val 281 , contain basolateral sorting information. When either residue was mutated independently to alanine, the integrity of the signal was lost, and chimeric receptors harboring these point mutations were sorted to the apical membrane domain. Similar results were found for the dileucine sorting motif in nucleotide-pyrophosphatase (41), where basolateral sorting was disrupted when either of the two leucine residues was mutated to alanine. Our alanine mutagenesis results also demonstrated that the M 3 BLSS does not contain a mono-valine sorting determinant like pIgR or a single hydrophobic signal like stem cell factor and CD147 (8,20,42,43).
The basolateral sorting signal of CD147 requires the presence of both a mono-leucine and an upstream cluster of acidic residues (43). Also, numerous dihydrophobic-based endocytosis signals require an upstream acidic cluster; however, the dihydrophobic basolateral determinants of E-cadherin and CD44 function without any upstream acidic residues (37,39,44). Our results show that the M 3 BLSS has strict requirements for a single glutamic acid residue at position 276, N-terminal of the FV motif.
To verify that the novel M 3 signal of FV represented a motif within the class of dihydrophobic sorting determinants and to determine whether the M 3 sorting signal could be mimicked by a classical di-leucine motif, we generated receptor chimeras with leucine point mutations at Phe 280 and Val 281 . When either residue was mutated individually or in combination to leucine, the mutant receptors maintain basolateral sorting. Thus, FV represents, to our knowledge, the newest member in the class of dihydrophobic basolateral sorting signals.
We also wanted to determine whether Glu 276 was specifically required for the function of the signal or if aspartic acid could substitute. Interestingly, aspartic acid could not substitute for glutamic acid at this amino acid position, suggesting that structure in addition to charge is important for the proper function of the sorting signal. In contrast, some endocytosis signals of the class (D/E)XXXL(L/I) can tolerate either a glutamic or aspartic acid residue at the Ϫ4 position (10). Together, these findings expand the known dihydrophobic sorting signals and demonstrate that a dileucine can substitute for the WT M 3 dihydrophobic residues, but a glutamic acid at position Ϫ4 to the FV motif is specifically required.
Intriguingly, the chimeric receptor harboring mutations at all of the critical sites (EFV-DLL) resulted in a receptor that displayed basolateral targeting similar to M 2 ϩM 3 [C-term 275-281]. Despite the glutamic to aspartic acid mutation that renders the receptor apically targeted, the presence of the dileucine created a dominant basolateral sorting signal. It is possible that the dileucine motif is a stronger signal than the native FV and is able to override the otherwise disruptive E276D mutation. Our results could suggest that a hierarchy of dihydrophobic sorting signal strengths exist, with LL stronger than FV and possibly other, related motifs. It is also possible that mutations at all three sites created a distinct and functional acidic/di-leucine basolateral sorting motif. In fact, the DXXXLL sequence we generated does resemble a well characterized endocytosis motif, found in numerous membrane proteins, although nothing is known about its structure (10). Surprisingly, when this motif was appended to the M 2 sequence, it did not result in visible increased endocytosis of the M 2 ϩM 3 chimeric receptor. Instead, the signal functioned to redirect the majority of M 2 from the apical to the basolateral membrane domain. The DXXXLL motif may have been unable to adopt a functional endocytosis structure in this context and instead existed in a structure more closely resembling the M 3 basolat-FIG. 7. Ramachandran plot per region. / angles for all 10 structures are shown. A, linker region, Thr 277 (red) and Glu 278 (green). B, residues in region 1 and region 2 adjacent to the linker residues. Shown are Glu 276 in region 1 (red) and Asn 279 in region 2 (green). C, region 1 residues Glu 274 (red) and Ala 275 (green). D, region 2 residues Phe 280 (red), Val 281 (green), His 282 (blue), and Pro 283 (magenta). eral sorting signal. Alternatively, DXXXLL might have assumed a tertiary epitope mimicking a different yet dominant basolateral signal that was capable of overriding all other apical information within the M 2 mAChR.
Because mutagenesis studies clearly showed that substitution of alanine at Glu 276 , Phe 280 , or Val 281 abrogates the basolateral targeting function of the M 3 signal, we examined the structural role of these residues in the BLSS by solving the solution structures of a 19-residue peptide corresponding to the basolateral sorting signal of the M 3 mAChR. Two-dimensional NMR analysis reveals that structured regions of the peptide contain residues required for basolateral sorting. Two regions of the peptide displayed the strongest tendencies to adopt structure in solution, as observed through NOESY, COSY, and temperature gradient data. NOESY data show that residue pairs Ala 275 /Glu 276 and Phe 280 /Val 281 have the greatest density of NOEs for the entire peptide, indicating that these residues are highly ordered. COSY data show that Glu 276 adopts a kinked conformation, whereas Val 281 adopts an extended conformation. NH chemical shift temperature gradients show that residues N-terminal to Val 281 had greater solvent exchange protection than residues C-terminal to Val 281 . Phe 280 and Glu 276 displayed the strongest exchange protection of all 19 residues, with values of Ϫ4.4 ppb/°C and Ϫ5.1 ppb/°C, respectively, consistent with their adopting stable secondary structure in solution.
Structural calculations using restraints derived from NOESY and COSY data provide solution structures of the 19-residue peptide. Glu 276 is present in a type IV ␤-turn, whereas the dihydrophobic residues Phe 280 and Val 281 are present in either a type I or IV ␤-turn. In addition, the acidic (region 1) and hydrophobic (region 2) regions of the structure, although not constrained in their orientation to one another, are individually well structured for a short peptide. Although the structure of the sorting motif may be altered by interaction with a binding partner, our results show that the individual sequences required for conferring basolateral targeting have the most structure in solution.
This is the first report identifying the solution structure of a G-protein-coupled receptor sorting signal. More importantly, the M 3 sorting motif is also the first example of a basolateral sorting signal that exists as a turn-turn structural motif. We cannot rule out the possibility that the M 3 BLSS exists as a turn-turn epitope in solution but assumes a different secondary structure in the presence of a cellular interactor protein. For example, the epidermal growth factor receptor endocytosis signal was predicted via two-dimensional NMR data to adopt a ␤-turn in solution but in a crystal structure adopted an extended conformation when in a complex with its binding partners (45). Nevertheless, understanding the native protein conformation of the novel sorting signal in the M 3 mAChR independent of an interacting protein is important. Determination of the structure of the M 3 BLSS may aid in the recognition of similar signals in other related or heterologous proteins. Knowledge of the predicted tertiary structure of the sorting signal may facilitate the identification of interactor proteins. Ultimately, determining the protein or proteins that interact with the M 3 BLSS will likely give the greatest insight regarding the cellular processes that guide the basolateral sorting of M 3 and possibly other membrane proteins to their proper cellular destinations.