Characterization of the Extra-large G Protein α-Subunit XLαs

Our group previously described a new type of G protein, the 78-kDa XLαs (extra large αs) (Kehlenbach, R. H., Matthey, J., and Huttner, W. B. (1994) Nature 372, 804–809 and (1995) Nature375, 253). Upon subcellular fractionation, XLαs labeled by ADP-ribosylation with cholera toxin was previously mainly detected in the bottom fractions of a velocity sucrose gradient that contained trans-Golgi network and was differentially distributed to Gαs, which also peaked in the top fractions containing plasma membrane. Here, we investigate, using a new antibody specific for the XL domain, the tissue distribution and subcellular localization of XLαs and novel splice variants referred to as XLN1. Upon immunoblotting and immunofluorescence analysis of various adult rat tissues, XLαs and XLN1 were found to be enriched in neuroendocrine tissues, with a particularly high level of expression in the pituitary. By both immunofluorescence and immunogold electron microscopy, endogenous as well as transfected XLαs and XLN1 were found to be predominantly associated with the plasma membrane, with only little immunoreactivity on internal, perinuclear membranes. Upon subcellular fractionation, immunoreactive XLαs behaved similarly to Gαs but was differentially distributed to ADP-ribosylated XLαs. Moreover, the bottom fractions of the velocity sucrose gradient were found to contain not only trans-Golgi network membranes but also certain subdomains of the plasma membrane, which reconciles the present with the previous observations. To further investigate the molecular basis of the association of XLαs with the plasma membrane, chimeric proteins consisting of the XL domain or portions thereof fused to green fluorescent protein were analyzed by fluorescence and subcellular fractionation. In both neuroendocrine and non-neuroendocrine cells, a fusion protein containing the entire XL domain, in contrast to one containing only the proline-rich and cysteine-rich regions, was exclusively localized at the plasma membrane. We conclude that the physiological role of XLαs is at the plasma membrane, where it presumably is involved in signal transduction processes characteristic of neuroendocrine cells.

Heterotrimeric G proteins, which consist of an ␣-subunit and a ␤␥ complex, transduce extracellular signals detected by heptahelical receptors to intracellular effectors (1)(2)(3)(4)(5). Our laboratory previously identified a new type of G protein, XL␣s (extra large ␣s), that is characterized by a bipartite structure (6). The C-terminal half of XL␣s is encoded by exons 2-13 of the G␣s gene and, hence, contains the entire G␣s sequence except for the N-terminal 47 amino acid residues encoded by exon 1 of the G␣s gene. In XL␣s, the latter residues are replaced by a novel sequence, referred to as the XL domain, yielding a protein with a molecular mass of 78 kDa and an electrophoretic mobility corresponding to 94,000 (rat XL␣s) (6,7). Thus, XL␣s is the largest known variant of a G protein ␣-subunit.
Little is known about the tissue distribution and subcellular localization of XL␣s, and no studies on the possible interaction of XL␣s with ␤␥ subunits, heptahelical receptors, and G protein effectors have been reported. Resolving these issues is important for elucidating the as yet unknown function of XL␣s. Using in situ hybridization, the XL␣s mRNA was previously detected in neuroendocrine but not other tissues (6), but it remains to be established whether or not a neuroendocrinespecific expression holds true for the XL␣s protein. Using subcellular fractionation, the ADP-ribosylated form of XL␣s has mainly been found in fractions containing trans-Golgi network (TGN) 1 (6), but it is unclear whether this reflects the subcellular localization of XL␣s under physiological conditions.
In the present study, we have generated antibodies specific to the XL domain of XL␣s. We have used these antibodies to investigate by immunocytochemistry and immunoblotting of adult rat tissues the cellular and tissue distribution of XL␣s and novel splice variants referred to as XLN1. In addition, we have studied by (immuno)fluorescence, immunoelectron microscopy, and subcellular fractionation the subcellular localization of immunoreactive XL␣s and XLN1 and green fluorescent protein (GFP) fusion proteins containing all or portions of the XL domain. In the accompanying paper (32), we investigate the possible interaction of XL␣s with ␤␥ subunits, heptahelical receptors, and the classical effector of G␣s, adenylyl cyclase.

MATERIALS AND METHODS
Antibodies-The peptide NH 2 -EPAAEPAAEPAAEPA-CONH 2 , corresponding to amino acid residues 45-59 of the XL␣s sequence (corrected translational start, see Kehlenbach et al. (7); amino acid residues 176 -190 of the originally published sequence (6)), was coupled to keyhole limpet hemocyanin via glutaraldehyde and used to raise the rabbit antiserum RK5. RK5 antibodies were affinity-purified using the above peptide coupled to Affi-Gel 15 (Bio-Rad). The rabbit antiserum against the C-terminal decapeptide of G␣s and XL␣s was the same as described previously (6). Other antibodies used were monoclonal anti-insulin antibody (mouse ascites fluid clone K36aC10), monoclonal anti-␤-tubulin antibody (clone TUB 2.1), both purchased from Sigma, and rabbit polyclonal anti-GFP affinity-purified antibody (CLONTECH).
cDNAs-The plasmid CDM8-XL␣s, originally called CDM8-XL (6), contains a Ϸ2.6-kilobase insert starting at nucleotide position 380 of the originally published sequence (6) and encodes the entire XL␣s protein sequence (see correction of translational start (7)) under the control of the cytomegalovirus promoter.
The isolation of the CDM8-XLN1a and CDM8-XLN1b cDNAs will be described elsewhere. 2 The plasmid CDM8-XLN1a contains the XL-exon spliced to exon 2, exon 3, and exon N1 (8) and encodes a C-terminaltruncated version of the XL␣s protein because of the presence of the stop codon-containing exon N1 (Fig. 1A). The plasmid CDM8-XLN1b is identical to CDM8-XLN1a except for a 95-base pair insertion between the XL-exon and exon 2, which in case of human XL␣s has been shown to be due to the additional exon A20 in the XL␣s/G␣s gene (9).
Cell Culture, Metabolic Labeling, and Transfections-PC12 cells were grown as described (10). [ 35 S]Sulfate labeling and chase of PC12 cells were performed as described (10). For [ 35 S]cysteine-methionine labeling, PC12 cells on 15-cm dishes were rinsed once with Met/Cys-free Dulbecco's modified Eagle's medium and incubated for 30 min at 37°C in this medium. The cells were then pulse-labeled for 5 min at 37°C 2 Y. Wang and W. B. Huttner, manuscript in preparation. EPAA, region containing the EPAA repeats; AARA, region containing the AARA repeats; P, proline-rich region; C, cysteine-rich region; ␤␥, region containing the putative ␤␥ binding site. Numbers refer to the corrected translational start of XL␣s (7).

FIG. 2.
Expression of rat XL␣s mRNA and protein in neuroendocrine tissues. A, immunoblot analysis, using the anti-XL antibody, of a total membrane fraction of PC12 cell PNS and of whole cell lysate of HeLa cells either untransfected (wt) or transfected with XLN1a or XLN1b cDNA. B, immunoblot analysis of whole PC12 cell lysate using the anti-XL antibody in the absence (Ϫ) or presence (ϩ) of EPAA peptide used as the antigen. Note the compression of the XLN1 bands due to the use of a minigel. C, Northern blot analysis of total RNA (10 g) from various adult rat tissues probed with DNA corresponding to exons 2-9 of G␣s. Only the region of the XL␣s mRNA is shown. D, immunoblot analysis of total membranes (200 g of protein) from various adult rat tissues using affinity-purified anti-XL antibody. Pituitary (A), anterior pituitary. Pituitary (IϩN), pars intermedia plus neurohypophysis; this lane belongs to a different immunoblot (30 g of protein) with a shorter exposure; the signal for XL␣s and XLN1 is ϳ80-fold and ϳ120-fold greater than that in the adrenal gland.
HeLa and COS-7 cells were grown in Dulbecco's modified Eagle's medium containing 10% fetal calf serum. HeLa cells on 15-cm dishes were transiently transfected using the calcium phosphate protocol (11) using 40 g/dish of circular plasmid DNA. COS-7 cells were transfected in the same way (for the GFP constructs) or (in the case of XL␣s) by using the FuGENE transfection reagent (Roche Molecular Biochemicals) and 20 g of circular plasmid DNA/15-cm dish.
Wild type and transfected cells were plated on coverslips (polylysinecoated in the case of PC12 cells) for immunofluorescence or on 15-cm dishes (for subcellular fractionation). Cells were used 2 days after transfection, with 10 mM sodium butyrate added in the case of XL␣stransfected PC12 cells during the last 16 h to increase the expression of the transgene (12).
Immunofluorescence-Pancreas and pituitary of adult rats were fixed overnight in 4% paraformaldehyde in PBS at 4°C, infiltrated with 0.5 M sucrose in PBS at 4°C, frozen in Tissue-Tek on liquid nitrogen and stored at Ϫ20°C. Frozen sections (10 m) were cut in a Leica Frigocut 2800N cryostat, air-dried on gelatin-coated slides, and processed for immunofluorescence as described (13), except that bovine serum albumin was used instead of fetal calf serum for blocking the sections, with the primary antibodies as follows: affinity-purified anti-XL antibody at 0.1-0.5 g IgG/ml and monoclonal anti-insulin antibody diluted 1/6000. Indirect immunofluorescence of paraformaldehyde-fixed PC12, HeLa, and COS-7 cells was performed as described (14), using 0.1-0.5 g of IgG/ml (final concentration) of affinity-purified anti-XL antibody. To test its specificity, the affinity-purified anti-XL antibody after dilu-tion to its final concentration was incubated before immunofluorescence overnight at 4°C with 10 g/ml EPAA peptide used as antigen. GFP fusion proteins were detected by GFP autofluorescence. For double immunofluorescence, anti-␤-tubulin was used at 15 g of IgG/ml. Wheat germ agglutinin coupled to tetramethylrhodamine B isothiocyanate was used at 50 g/ml. Secondary antibodies used were fluoresceinor rhodamine-conjugated goat anti-rabbit IgG, fluorescein-conjugated donkey anti-mouse IgG, and Cy2-conjugated goat anti-mouse IgG. Samples were examined by either conventional fluorescence microscopy (Zeiss Axiophot) or confocal laser scanning microscopy (Leica TCS 4D ).
Immunogold Electron Microscopy-Posterior lobes of adult rat pituitaries were dissected and fixed by immersion in 4% paraformaldehyde, 0.1% glutaraldehyde in 0.1 M cacodylate buffer, dehydrated in ethanol, and embedded in the acrylic resin LRWhite (London Resin Corp.). Semithin sections were obtained and stained with 1% toluidine blue, 1% sodium tetraborate to localize the pars intermedia. Ultrathin sections were labeled using affinity-purified anti-XL antibody at 1-5 g of IgG/ml followed by protein A gold (9 nm). Sections were stained with saturated uranyl acetate in water.
Subcellular Fractionation-PC12 and HeLa whole cell lysates were obtained by the addition of Laemmli sample buffer to the dishes followed by immediate boiling of the samples. Postnuclear supernatant (PNS) from PC12 cells was prepared as described (10). A total membrane fraction from the PNS was obtained by centrifugation at 100,000 ϫ g for 1 h at 4°C. Velocity sucrose gradient centrifugation of the PNS and equilibrium sucrose gradient centrifugation of selected fractions of the velocity sucrose gradient were performed as described (10).
For the analysis of membrane association of GFP fusion proteins transfected into PC12 cells, equal aliquots of the respective PNS were subjected to centrifugation at 100,000 ϫ g for 1 h at 4°C. Particulate and soluble fractions were analyzed by SDS-PAGE and immunoblotting.
For the analysis of the subcellular localization of newly synthesized G␣s and XL␣s, velocity sucrose gradient fractions from PC12 cells pulse-labeled for 5 min with [ 35 S]cysteine-methionine and chased for 15 or 120 min were mixed with the respective fractions containing membranes from unlabeled PC12 cells (used as carrier) and diluted with 10 mM Hepes-KOH, pH 7.2. Membranes were collected by centrifugation (100,000 ϫ g at 4°C, 1 h). G␣s and XL␣s were immunoprecipitated from the membranes using the antiserum against the C-terminal decapeptide of G␣s and XL␣s at 1:100 dilution, as described (6).
For the analysis of membrane association of newly synthesized G␣s and XL␣s, total membranes and cytosol were prepared by centrifugation (100,000 ϫ g at 4°C, 1 h) of the PNS of PC12 cells pulse-labeled for 5 min with [ 35 S]cysteine-methionine and chased for 0, 15, or 120 min. An aliquot of the total membrane preparation was subjected to carbonate stripping by incubation with 0.1 M Na 2 CO 3, 0.025% saponin, 2 mM EDTA, 0.25 mM phenylmethylsulfonyl fluoride as described (6). G␣s and XL␣s were immunoprecipitated from cytosol, total membranes, and carbonate-stripped membranes using the antiserum against the Cterminal decapeptide of G␣s and XL␣s, as described (6).
Immunoblotting-Adult rat tissues were dissected and immediately frozen in liquid nitrogen followed by storage at Ϫ80°C until use. Tissue was homogenized in 0.3 M sucrose, 10 mM Hepes-KOH, pH 7.2, 1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride. In the case of the pancreas, homogenization was performed in the presence of a protease inhibitor mixture (pepstatin, leupeptin, chymostatin, antipain, soybean trypsin inhibitor). Homogenates were centrifuged at 1000 ϫ g for 10 min at 4°C, and the supernatants were collected and centrifuged at 100,000 ϫ g for 1 h at 4°C to obtain the total membranes. Membranes were resuspended in 10 mM Hepes-KOH, pH 7.2, and then boiled in Laemmli sample buffer.
Immunoblotting after SDS-PAGE was carried out according to standard procedures. For immunoblotting of velocity sucrose gradient fractions, we used 3% bovine serum albumin in PBS to block the nitrocellulose membranes, the antiserum against the C-terminal decapeptide of G␣s and XL␣s at 1:800 -1:1000 dilution, and 125 I-protein A. For immunoblotting of total membranes from various tissues and of velocity sucrose gradient fractions from PC12 cells expressing various GFP fusion proteins, the nitrocellulose membranes were blocked with 5% lowfat milk powder in PBS and incubated (i) with either affinitypurified anti-XL antibody at 0.1 g of IgG/ml or anti-GFP antibody at 1 g of IgG/ml followed by horseradish peroxidase-conjugated secondary antibody and the ECL system (Amersham Pharmacia Biotech) or (ii) with anti-XL antiserum (1:200 to 1:600 dilution) followed by 125 I-protein A. To test the specificity of the anti-XL antibody, blots of whole PC12 cell lysate were incubated with the antibody solution containing 10 g/ml of the EPAA peptide.
Northern Blot-Tissues from adult rats were directly homogenized in guanidine isothiocyanate, and total RNA was isolated as described (15,16). Northern blot analysis (10 g of total RNA) was carried out using a [ 32 P]dCTP-labeled cDNA fragment corresponding to exons 2-9 of a G␣s cDNA as probe.

Expression of XL␣s and XLN1 in Neuroendocrine
Tissues-Previous results obtained by in situ hybridization showed that the XL␣s mRNA occurs predominantly in certain endocrine cells and neurons (6). The C-terminal decapeptides of XL␣s and G␣s are identical, and hence, use of the corresponding antibody (6) in immunocytochemistry does not allow one to distinguish between these two G proteins. Moreover, in immunofluorescence experiments, no staining was observed with the previously raised antibody (6) against an epitope of the XL domain adjacent to the putative ␤␥ binding site of XL␣s (data not shown). To investigate the tissue distribution of XL␣s by immunocytochemistry, we therefore raised a new antibody against the EPAA repeats of the XL domain (6), referred to as anti-XL antibody. In immunoblots of a total membrane fraction from PC12 cell PNS, this antibody recognized only XL␣s and a group of three proteins of M r 50,000 -65,000 ( Fig. 2A). The latter correspond to novel splice variants derived from the rat XL␣s/G␣s gene 2 and will be referred to as the XLN1 proteins (see Fig. 1A). These contain the entire XL domain followed by C-terminal-truncated ␣s domains, due to the presence of additional exons in the corresponding mRNAs: in the case of XLN1a, the stop codon-containing N1 exon (8), and in the case of XLN1b, the frameshifting 95-base pair insertion (Fig. 1A). Whereas immunoblots using the anti-XL antibody of untransfected HeLa cells showed no immunoreactive bands, immunoblots of HeLa cells transfected with the XLN1a or XLN1b cDNA showed major immunoreactive bands very similar, if not identical, to two of the M r 50,000 -65,000 proteins detected in PC12 cells ( Fig. 2A), corroborating their identity as XLN1 proteins. In immunoblots of whole PC12 cell lysate, the immunoreactive bands corresponding to XL␣s and the XLN1 proteins were abolished when the anti-XL antibody was used in the presence of the peptide used as antigen (Fig. 2B), which further documents the specificity of this antibody. Essentially similar results were obtained when immunoblots were probed with an antibody raised against a glutathione S-transferase fusion protein containing the XL domain (data not shown).
We used the anti-XL antibody to examine the tissue distribution of the XL␣s protein in comparison with that of the XL␣s mRNA. Northern blot analysis of various rat tissues revealed the presence of the XL␣s mRNA in the adrenal gland, brain, cerebellum, heart, and pituitary gland, with the levels in the brain and heart at the limit of detectability and the levels in the pituitary much higher than in the other neuroendocrine tissues (Fig. 2C). No XL␣s mRNA was detected in kidney, liver, and spleen. These observations are consistent with the neuroendocrine-specific tissue distribution of the XL␣s mRNA previously observed by in situ hybridization (6). Immunoblot analysis revealed the presence of the XL␣s protein and the XLN1 proteins in rat brain, cerebellum, adrenal gland, pancreas (at the limit of detection) and, at very high levels, the pituitary gland but not in liver and kidney (Fig. 2D). Within the pituitary gland, XL␣s and XLN1 immunoreactivity per total protein was much higher in the posterior lobe (containing the pars intermedia plus neurohypophysis) than in the anterior lobe (see legend to Fig. 2D). The level of XL␣s and XLN1 immunoreactivity per total protein in the pituitary posterior lobe was ϳ80 times (XL␣s) to ϳ120 times (XLN1) higher than that in the adrenal.
Adult rat pituitary cryosections were examined by immunofluorescence using the anti-XL antibody. Strong immunostaining for XL␣s/XLN1 was observed in most, if not all, cells of the pars intermedia, the melanotrophs ( Fig. 3a and b). The level of immunostaining was lower in the anterior pituitary and restricted to a subset of cells (Fig. 3, a and c). The neurohypophysis was not immunostained above background (Fig.  3a). In both anterior pituitary and pars intermedia, the staining was predominantly observed at the cell periphery, delineating the shape of the cells.
Immunofluorescence of cryosections of adult rat pancreas showed that the immunoreactivity for XL␣s/XLN1 was present in the islets of Langerhans, the endocrine part of the organ, but not in the surrounding exocrine part (Fig. 3d). Comparison of the staining with that for insulin by double immunofluorescence revealed that XL␣s/XLN1 was expressed in most (Fig. 3, d and e, larger arrows) but not all (Fig. 3, d and e, arrowheads) ␤ cells. XL␣s/XLN1 expression was not restricted to ␤ cells but also observed in other endocrine cells of the pancreas, as indicated by the presence of cells immunoreactive for XL␣s/XLN1 but negative for insulin (Fig. 3, d and e, small arrows). Taken together, XL␣s and XLN1 are neuroendocrine-specific proteins expressed in many, but not all, neuroendocrine cells in the adult. Essentially similar results were obtained when immunofluorescence of pituitary and endocrine pancreas was performed using an antibody raised against a glutathione S-transferase fusion protein containing the XL domain (data not shown).
Subcellular Localization of XL␣s: Subcellular Fractionation-Upon subcellular fractionation of PC12 cells, we previously observed a differential distribution of XL␣s and G␣s, [ 32 P]ADP-ribosylated by cholera toxin, across a velocity sucrose gradient (6). The majority of the [ 32 P]ADP-ribosylated G␣s, but at most a quarter of the [ 32 P]ADP-ribosylated XL␣s, was recovered in the top fractions of the gradient (6), which are known to contain plasma membrane (17). By contrast, only about one-third of the [ 32 P]ADP-ribosylated G␣s, but almost two-third of the [ 32 P]ADP-ribosylated XL␣s, was recovered in the bottom fractions of the gradient (6), which are known to contain TGN membranes (10). Fig. 4a shows the results of another such subcellular fractionation experiment, which confirms the previously observed differential distribution of [ 32 P]ADP-ribosylated G␣s and XL␣s upon velocity sucrose gradient centrifugation.
Surprisingly, however, no significant difference in the distribution upon velocity sucrose gradient centrifugation was observed when immunoreactive rather than [ 32 P]ADP-ribosylated G␣s and XL␣s were compared (Fig. 4b). Immunoblotting using the antibody against the C-terminal decapeptide of G␣s/ XL␣s showed that for either G protein, the majority was found in the top fractions of the gradient and, at most, one-third in the bottom fractions. Equilibrium sucrose gradient analysis provided further evidence that the XL␣s found in the top fractions of the velocity sucrose gradient, i.e. the majority of the immunoreactive XL␣s (Fig. 4b) but only about one-third of the [ 32 P]ADP-ribosylated XL␣s (Fig. 4a), was indeed associated with the plasma membrane. Both the [ 32 P]ADP-ribosylated (Fig. 4c) and the immunoreactive (Fig. 4d) XL␣s of fractions 2-5 of the velocity sucrose gradient cofractionated upon equilibrium sucrose gradient centrifugation with [ 35 S]sulfate-labeled heparan sulfate proteoglycan chased from the TGN to constitutive secretory vesicles and the plasma membrane (18,19) and exhibited a distinct distribution from [ 35 S]sulfate-labeled secretogranin II chased into immature secretory granules (18,20).
Immunoblotting of velocity sucrose gradient fractions using the anti XL-antibody showed that the distribution of XLN1 was very similar to that of XL␣s, with the majority found in the top fractions containing plasma membrane (Fig. 5, A and B). Remarkably, the distribution of XL␣s and XLN1 in the bottom half of the gradient was distinct from that of secretogranin II pulse-labeled with [ 35 S]sulfate in the TGN (Fig. 5B). This suggests that even the XL␣s and XLN1 found in the bottom half of the velocity gradient is associated with membranes other than the TGN. Subcellular Localization of XL␣s: Immunofluorescence and Immunoelectron Microscopy-Immunofluorescence analysis of PC12 cells stained with the anti-XL antibody showed that most immunoreactivity was associated with the plasma membrane, with only occasional staining in the perinuclear region of the cells (Fig. 6, a-c). The plasma membrane-associated immunoreactivity extended into plasmalemmal protrusions and processes emerging from the cell body (Fig. 7). PC12 cells overexpressing XL␣s upon transfection showed an increased immunoreactivity, the pattern of which was similar to that of mock-transfected PC12 cells (Fig. 6, d-f). Preincubation of the anti-XL antibody with the EPAA peptide used as antigen abol- ished the immunostaining (Fig. 6, g and h).
XL␣s was transfected into HeLa and COS-7 cells, two nonneuroendocrine cell types that normally do not show XL-immunoreactivity (see the untransfected cell indicated by the arrow in Fig. 8a, which is a counterstaining for ␤-tubulin, and data not shown). Immunofluorescence using the anti XL antibody showed that transfected XL␣s was exclusively localized at the plasma membrane, with clear labeling of membrane protrusions (Fig. 8b). The same was observed for HeLa cells transfected with XLN1 (data not shown). An exclusive plasma membrane localization of XL␣s was also observed for the vast majority of transfected COS-7 cells (Fig. 8c). In a minority of transfected COS-7 cells (Ͻ10%), which showed very high levels of XL␣s expression, XL immunoreactivity was not only seen on the plasma membrane but also in a perinuclear location (Fig.  8d, arrowhead). These observations show that transfected XL␣s is primarily targeted to the plasma membrane in nonneuroendocrine cells and imply that at least part of the plasma membrane-associated endogenous immunoreactivity observed with the anti-XL antibody in PC12 cells was due to XL␣s (rather than XLN1 only). Essentially similar results were ob-tained when immunofluorescence of transfected HeLa cells was performed using the antibody raised against a glutathione S-transferase fusion protein containing the XL domain (data not shown).
Immunogold electron microscopy using the anti-XL antibody was performed on ultrathin sections of LRWhite-embedded posterior lobe of adult rat pituitary to examine the subcellular localization of XL␣s/XLN1 in the pars intermedia, a tissue with a high content of XL␣s as shown by immunoblotting (Fig. 2D). Gold particles were exclusively detected over the plasma membrane of melanotrophs (Fig. 9). No immunogold labeling was detected in the axon terminals of the neurohypophysis present in the same sections (data not shown). We conclude that by subcellular fractionation, immunofluorescence, and immunoelectron microscopy, the vast majority of XL␣s is associated with the plasma membrane.
The XL Domain Contains a Plasma Membrane-targeting Signal-The finding that both XL␣s and XLN1 are localized on the plasma membrane, as observed by immunofluorescence and subcellular fractionation, suggested that the targeting signal responsible for this localization resides within the XL domain. To investigate this directly, we constructed chimeric proteins containing the XL domain and portions thereof fused to GFP (Fig. 1B). Although the GFP reporter was soluble in the cytosol of transfected PC12 cells (Figs. 10 and 11, b-d), the XL domain indeed mediated the association of GFP with the plasma membrane ( Figs. 10 and 11, f-h). The same was the case for an XL domain lacking the putative ␤␥ binding region (32) (Figs. 10 and 11, j-l), indicating that the plasma membrane association of XL␣s does not depend on ␤␥ interaction (as has been reported for G␣s (21)) but is mediated by another region within the XL domain. In fact, this region appears to be the cysteinerich region because GFP fused to an XL domain lacking both the putative ␤␥ binding region and the cysteine-rich region (6) was found to be soluble in the cytosol (Figs. 10 and 11, n-p), as is wild type GFP. A portion of the XL domain comprising only the proline-rich region and the cysteine-rich region (6) was sufficient to mediate the association of GFP with membranes ( Fig. 10) that, remarkably, were not only plasma membrane but also some internal membranes of PC12 cells (Fig. 11, r-t).
Transfection of the various GFP fusion constructs (Fig. 1B) into COS-7 cells gave essentially the same results (Fig. 11, a, e, i, and m), except that in the case of GFP fused to the prolinerich and cysteine-rich regions of the XL domain, the staining of the internal membranes that was observed in addition to the plasma membrane fluorescence was characterized by a distinct perinuclear pattern (Fig. 11, q, arrow).
We further investigated the differential localization of the GFP fusion proteins containing either the entire XL domain (XL-GFP) or only its proline-rich and cysteine-rich regions ((PϩC)-GFP) by subcellular fractionation. Immunoblotting of velocity sucrose gradient fractions with anti-GFP antibody showed that XL-GFP was exclusively found in the top fractions FIG. 8. Immunofluorescence analysis of HeLa and COS-7 cells transfected with XL␣s. HeLa cells and COS-7 cells were transfected with the CDM8-XL␣s vector. a and b, double immunofluorescence of HeLa cells using a monoclonal anti-␤-tubulin antibody (a) and affinitypurified anti-XL antibody (b) followed by conventional fluorescence microscopy. Immunoreactivity for XL␣s in the transfected cell is located to plasma membrane protrusions (b). Tubulin staining reveals the presence of an additional, untransfected cell that lacks endogenous XL␣s (a, arrow). c and d, immunofluorescence of COS-7 cells using affinity-purified anti-XL antibody followed by confocal microscopy. The pattern of XL␣s immunoreactivity shown in panel c, which is characteristic of its localization on the plasma membrane, is representative of the vast majority of transfected cells, whereas the pattern shown in panel d, which shows staining of both the plasma membrane and internal membranes (arrowhead), is observed only in a minority of transfected cells with a very high level of XL␣s expression. of the gradient known to contain plasma membrane (Fig. 12A), whereas (PϩC)-GFP was found not only in these fractions but also in the bottom half of the gradient with a peak in fraction 9 (Fig. 12B), which also contains the peak of TGN membranes (compare Fig. 5B, open squares).
Dynamics of Membrane Association of XL␣s-We used pulsechase in conjunction with subcellular fractionation of PC12 cells to investigate the kinetics of membrane association of XL␣s in comparison with G␣s. PC12 cells were pulse-labeled with [ 35 S]cysteine-methionine and chased for either 15 min or 2 h. After 15 min, more than half and, after 2 h, nearly all of XL␣s were found to be membrane-associated (Fig. 13, filled  squares). At either chase time, the majority of the membraneassociated XL␣s was resistant to carbonate extraction (Fig. 13,  open squares). The kinetics of membrane association of XL␣s was similar to that of G␣s (Fig. 13, circles).
The membranes to which newly synthesized XL␣s becomes associated were fractionated on the velocity sucrose gradient (compare Figs. 4, a and b, 5, and 12). After 15 min of chase, the majority of newly synthesized and membrane-associated XL␣s was found in the top fractions, and only a minor proportion was found in the bottom fractions of the gradient (Fig. 14A, filled  circles). The latter proportion of XL␣s decreased slightly after 2 h of chase (Fig. 14A, open circles). Similar observations were made for G␣s (Fig. 14B), with the change in its distribution between top and bottom fractions upon long chase more obvious than in the case of XL␣s.

DISCUSSION
Tissue Distribution-The present study shows that the XL␣s and XLN1 proteins are specifically expressed in neuroendo-crine tissues. For XL␣s, these findings are consistent with the previous observation (6) confirmed here that the XL␣s mRNA is present in neuroendocrine but not other tissues. Within the neuroendocrine system, however, XLϪimmunoreactivity (which reflects XL␣s and XLN1) was not detected in all (neuro)peptide-and peptide hormone-producing cells but, rather, in certain subpopulations of these cells. For example, in the adult rat, XL immunoreactivity was found at high levels in virtually all cells of the intermediate lobe of the pituitary but at lower levels in, and not in all cells of, the anterior pituitary. In the endocrine pancreas, XL immunoreactivity was detected in most but not all insulin-producing ␤-cells and also in other islet cells. From our data, it appears that there is no straightforward correlation between the expression of XL immunoreactivity in a given neuroendocrine cell and the peptide hormone/neuropeptide produced or the extracellular input received by this cell. If any such correlation is to be speculated upon, it would be with the pituitary adenylyl cyclase-activating polypeptide (PACAP) receptor, whose tissue distribution shows some resemblance to that of XL␣s and XLN1 (22)(23)(24).
The neuroendocrine-specific tissue distribution of XL␣s may have implications for the pathophysiological mechanisms underlying the phenotype of patients afflicted by mutations in exons 2-13 of the GNAS1 gene (25)(26)(27)(28). Since these exons encode not only most of the G␣s sequence but also half of the XL␣s sequence (6,9,29,30), any phenotype confined to the neuroendocrine system of these patients may reflect the dysfunction of XL␣s rather than that of the ubiquitously expressed G␣s.
Subcellular Localization-An unexpected finding of the present study was that in both tissues and isolated cells in culture, XL immunoreactivity was almost exclusively localized to the plasma membrane, as observed in immunofluorescence and immunoelectron microscopy. It cannot be argued that the XL immunoreactivity at the plasma membrane reflects solely the subcellular localization of XLN1 rather than that of XL␣s for the following reasons. First, upon subcellular fractionation of PC12 cells, XL␣s and XLN1 show the same distribution (Fig.  5). Second, upon expression from cloned cDNAs, almost exclusive plasma membrane localization of XL immunoreactivity was observed for either XL␣s (Figs. 6, d-f, and 8, b-d) and XLN1 (data not shown).
The almost exclusive plasma membrane localization of immunoreactive XL␣s is in contrast to the previous conclusion (6) that XL␣s is predominantly associated with the TGN, which was based on the distribution of ADP-ribosylated XL␣s in comparison with ADP-ribosylated G␣s upon subcellular fractionation of PC12 cells using an established velocity sucrose gradient centrifugation protocol (18). Specifically, the majority of ADP-ribosylated XL␣s, but only the minority of G␣s, was recovered in the bottom half of the velocity gradient (6), an observation confirmed here (Fig. 4a). The bottom half of the velocity gradient is known to contain the bulk of the TGN membranes of PC12 cells (18) (Fig. 5B). However, unexpectedly, the distribution within the bottom half of the velocity gradient of immunoreactive XL␣s and XLN1 was distinct from that of the TGN (defined by [ 35 S]sulfate pulse-labeled secretogranin II) (Fig. 5B). This finding together with the almost exclusive plasma membrane localization of XL␣s and XLN1 observed in immunofluorescence leads us to conclude that the fractions in the bottom half of the velocity gradient that contain the peak of XL␣s and XLN1 (but not the peak of TGN membranes) are enriched in plasma membrane fragments that represent a specialized domain of the plasma membrane (see below).
The reason why the ADP-ribosylated XL␣s is differentially FIG. 10. Membrane association of GFP fusion proteins containing the XL domain or portions thereof. PNS from PC12 cells transfected with the GFP fusion proteins described in Fig. 1B was subjected to ultracentrifugation, and supernatant (S) and pellet (P) were analyzed by SDS-PAGE followed by immunoblotting using an anti-GFP antibody (A). For each fusion protein, the amount in the pellet is expressed as a percentage of total (sum of supernatant and pellet) in panel B.
distributed across the velocity gradient to the immunoreactive XL␣s is unclear. This difference does not appear to be caused by ADP-ribosylation of XL␣s. Upon immunoblotting of velocity gradient fractions from PC12 cells incubated overnight in the presence of cholera toxin, which is likely to result in stoichiometric ADP-ribosylation of XL␣s (6), the majority of the immunoreactive XL␣s was found in the top fractions of the gradient (data not shown), as was the case for XL␣s from control PC12  -d, f-h, j-l, n-p, and r-t) were transfected with the GFP fusion proteins described in Fig. 1B, fixed, and analyzed by confocal microscopy . Panels a, b, e, f, i, j, m, n, q, and r, GFP fluorescence (green). Panels c, g, k, o, and s, the fixed PC12 cells were permeabilized and stained with tetramethylrhodamine B isothiocyanate-conjugated wheat germ agglutinin (WGA, red), a lectin-staining Golgi and Golgi-derived membranes such as plasma membrane, before confocal analysis; GFP-expressing cells are indicated by asterisks. Panels d, h, l, p, and t, overlay. Note the plasma membrane staining in cells transfected with the XL-GFP (e-h), XL⌬␤␥-GFP (i-l), and (PϩC)-GFP (q-t) and, for (PϩC)-GFP, the additional clustered staining in the perinuclear region in COS-7 cells (q, arrow). cells (Fig. 4b). Rather, the XL␣s associated with the plasma membrane fragments recovered in the upper half of the velocity sucrose gradient appears to undergo in vitro ADP-ribosylation by cholera toxin to a lesser extent than the XL␣s associated with the plasma membrane fragments sedimenting to the lower half of the gradient. Possible explanations for the lesser extent of ADP-ribosylation of the XL␣s recovered in the upper half of the velocity sucrose gradient include (i) a lower probability to be in the trimeric state (see also the accompanying paper by Klemke et al.(32), (ii) a reduced ability to interact with ADP-ribosylation factor, and (iii) an interaction with a hypothetical protein that inhibits ADP-ribosylation. Whatever the explanation, the plasma membrane fragments obtained after homogenization of PC12 cells that contain the majority of the ADP-ribosylatable XL␣s but relatively little ADP-ribosylatable G␣s must be relatively large because they sediment to the bottom half of the velocity gradient. These fragments may well be derived from thin cell surface protrusions, a specialized domain of the plasma membrane, which contain XL␣s (Figs. 6  and 7).
The almost exclusive localization of both XL␣s and XLN1 on the plasma membrane is mediated by the XL domain, as shown by the targeting of a GFP fusion protein (Figs. 11, c-h, and  12A). In contrast to the GFP fusion protein containing the entire XL domain, a fusion protein containing only the prolinerich and cysteine-rich regions of the XL domain ((PϩC)-GFP) was not only targeted to the plasma membrane but also to internal membranes in the perinuclear region (Fig. 11, q-t), most likely Golgi membranes, as suggested by subcellular fractionation (Fig. 12B). The subcellular localization of (PϩC)-GFP) in COS-7 cells (Fig. 11q) is very similar to that observed by Ugur and Jones for an essentially identical fusion protein and reported (31) after submission of the original version of this paper. Our finding that deletion of the cysteine-rich domain abolishes membrane association (Figs. 10 and 11, m-p) also is consistent with the conclusion of these authors that the cysteine residues in this domain, which are palmitoylated, are of critical importance for membrane binding of XL␣s (31). However, the present study clearly demonstrates that the vast majority of XL␣s is not associated with the Golgi complex, as claimed by Ugur and Jones (31), but with the plasma membrane. The discrepancy in conclusion between this and the present study may in part be explained by the fact that Ugur ]cysteine-methionine for 5 min and chased for 0, 15, or 120 min. A PNS was prepared and centrifuged to yield cytosol and total membranes. The total membranes were subjected to carbonate extraction at pH 11. G␣s and XL␣s were immunoprecipitated from the cytosol, total membranes and the pH 11-treated membranes using the antibody against the C-terminal decapeptide of G␣s/XL␣s, followed by SDS-PAGE and phosphoimaging. G␣s (circles) and XL␣s (squares) in the total membranes (filled symbols, total) and the pH 11-treated membranes (open symbols, pH 11-resistant) were quantified and are expressed as percent of that in the PNS (sum of cytosol and total membranes). and Jones (31) did not use an antibody specific to the XL domain as we have done (Fig. 2, A and B) but rather an antibody to the C terminus of G␣s, which also recognizes XL␣s; this poses a certain risk when attributing immunostaining obtained with this antibody to XL␣s rather than G␣s. Our data also show that the targeting of (PϩC)-GFP to the Golgi complex region is at best of secondary physiological relevance and should not be taken as an indication of the targeting of XL␣s, since very little if any localization in the Golgi complex region was observed for a GFP fusion protein containing the entire XL domain and for full-length XL␣s.
A virtually exclusive recovery in subcellular fractions enriched in plasma membrane (rather than TGN, compare Fig.  14A with Fig. 5B) was also observed for newly synthesized, metabolically labeled XL␣s. Tight association of newly synthesized XL␣s with the plasma membrane occurred rapidly, since already after 15 min of chase, the majority of the membraneassociated XL␣s was resistant to extraction at pH 11 (Fig. 13). Hence, in conclusion, our results point to a role of XL␣s at the plasma membrane in neuroendocrine cells, most likely in signal transduction. Consequently, the possible interaction of XL␣s with ␤␥ subunits and receptor/effector systems known to couple to G␣s are addressed in the subsequent paper (Klemke et al. (32).