Prohormone Convertase 1/3 Is Essential for Processing of the Glucose-dependent Insulinotropic Polypeptide Precursor*

The physiology of the incretin hormones, glucagon-like peptide 1 (GLP-1) and glucose-dependent insulinotropic polypeptide (GIP), and their role in type 2 diabetes currently attract great interest. Recently we reported an essential role for prohormone convertase (PC) 1/3 in the cleavage of intestinal proglucagon, resulting in formation of GLP-1, as demonstrated in PC1/3-deficient mice. However, little is known about the endoproteolytic processing of the GIP precursor. This study investigates the processing of proGIP in PC1/3 and PC2 null mice and in cell lines using adenovirus-mediated overexpression. Supporting a role for PC1/3 in proGIP processing, we found co-localization of GIP and PC1/3 but not PC2 in intestinal sections by immunohistochemistry, and analysis of intestinal extracts from PC1/3-deficient animals demonstrated severely impaired processing to GIP, whereas processing to GIP was unaltered in PC2-deficient mice. Accordingly, overexpression of preproGIP in the neuroendocrine AtT-20 cell line that expresses high levels of endogenous PC1/3 and negligible levels of PC2 resulted in production of GIP. Similar results were obtained after co-expression of preproGIP and PC1/3 in GH4 cells that express no PC2 and only low levels of PC1/3. In addition, studies in GH4 cells and the α-TC1.9 cell line, expressing PC2 but not PC1/3, indicate that PC2 can mediate processing to GIP but also to other fragments not found in intestinal extracts. Taken together, our data indicate that PC1/3 is essential and sufficient for the production of the intestinal incretin hormone GIP, whereas PC2, although capable of cleaving proGIP, does not participate in intestinal proGIP processing and is not found in intestinal GIP-expressing cells.

The endoproteolytic processing of many protein precursors depends on the expression of members of the family of subtilisin-like endoproteases also known as prohormone convertases (PCs). 2 Three members of the family, PC1/3, PC2, and PC5/6A, are believed to be the principal convertases localized to the granules of the regulated secretory pathway of neuroendocrine cells (1,2). Despite their similar functional and biochemical characteristics, PCs have been demonstrated to cleave motifs of dibasic amino acids with different efficiencies in vitro (3)(4)(5)(6)(7)(8)(9). The finding of specific defects in the processing of prototypic peptide hormones in PC1/3-and PC2-deficient mice seems to have validated this concept (3, 9 -16). That the expression of a specific PC ultimately determines the peptide product becomes especially apparent regarding the processing of proglucagon, a prohormone precursor co-expressed with PC1/3 in the intestinal L-cell and with PC2 in the pancreatic ␣-cell, resulting in entirely different precursor products (11,12,16). Thus, in the islet ␣-cell, the main peptide product is glucagon, and the principal intestinal peptides are the incretin hormone, glucagon-like peptide 1 (GLP-1), and the intestinotrophic hormone GLP-2.
However, processing of the other main incretin hormone precursor pro-glucose-dependent insulinotropic polypeptide (proGIP) has so far been unexplored. ProGIP is expressed in the intestinal mucosal K-cell throughout the intestine with the highest density in the proximal jejunum (17). ProGIP consists of three domains of which the middle one, corresponding to GIP, is thought to be the only biologically active peptide from the precursor. Despite this, most antisera raised against GIP have been found to cross-react with another larger molecule, designated GIP8000 (18,19). This has been suggested to be a precursor product, although other experiments have concluded differently (20). We, therefore, decided to define the processing requirements of proGIP using for this PC1/3-and PC2-deficient mice as well as cell lines manipulated with an adenovirus vector overexpression system.

EXPERIMENTAL PROCEDURES
Animals-The PC1/3 and PC2 null mutant mice were generated by introducing a targeted mutation in 129sv embryonic stem cells as previously described (12,15). 3-5-Month-old PC1/3 and 5-9month-old PC2 null mice and age-matched controls on a C57BL/6j x 129sv F1 background were used in the study. The mice were housed in a specific pathogen-free environment at the Howard Hughes Medical Institute, and all experiments were performed in accordance with institutional guidelines. The animals were euthanized, and intestinal tissue samples were quickly removed, rinsed in PBS, and immediately frozen.
Immunohistochemistry-Specimens of upper and lower jejunum and ileum from three wild-type mice and four PC1/3 null mice were fixed in 4% buffered paraformaldehyde and embedded in paraffin. The tissue sections were dewaxed and subjected to antigen-retrieval by microwave irradiation in citrate buffer (pH 6.0) before the immunohistochemical staining. For immunohistochemistry, we used the rabbit PC1/3 antiserum RS-18B7 or PC2 antiserum PC2Pep4, both produced as described in Scopsi et al. (21), and the monoclonal mouse GIP antiserum 3.65H (a generous gift from Dr. C. H. McIntosh, University of British Columbia, Vancouver, Canada). In double staining experiments, PC1/3 antiserum (diluted 1:1000) or PC2 antiserum (diluted 1:10,000) was visualized using biotin anti-rabbit antibody (diluted 1:200, DakoCytomation, Glostrup, Denmark) and Texas Red streptavidin (red fluorescence, diluted 1:200, Amersham Biosciences). Simultaneously, the GIP antibody was diluted 1:500 and visualized using digoxigenin-labeled anti mouse antibody (diluted 1:25, Roche Applied Science) and fluoresceinlabeled anti-digoxigenin antibody (diluted 1:25, green fluorescence, Roche Applied Science). For the double staining, we used the Mouse on Mouse (MOM) kit (Vector Laboratories, Burlingame, CA) for dilution of antibodies and reagents to allow the use of a mouse primary antibody in mouse tissue.
Immunocytochemistry-GH 4 cells (4 ϫ 10 6 cells/10-cm dish) were grown on polylysine (Sigma-Aldrich)-coated 18 ϫ 18-mm glass plates and co-infected with combinations of adenovirus expressing preproGIP and adenovirus expressing PC1/3 or PC2 as described below (infections of cell lines). Cells were then fixed in 4% buffered paraformaldehyde. Before applying the primary antibodies, cells were incubated in Image IT FX Signal Enhancer according to the manufacturer's instructions to block background staining (Molecular Probes Inc, Eugene, OR). The primary antibodies were the rabbit PC1/3 antiserum RS-18B7 diluted 1:2000, the rabbit antiserum PC2Pep diluted 1:2000 (21), and the monoclonal mouse GIP antiserum 3.65H diluted 1:400. The PC1/3 and PC2 immunoreactive cells were incubated with biotinylated swine anti-rabbit antibody as above, and the GIP immunoreactive cells were incubated with biotin-labeled anti-mouse antibody (DAKO). All immune reactions were then visualized using the PerkinElmer Life Sciences tyramide signal amplification-indirect kit as described by the manufacturer and finished by streptavidin coupled to Texas Red (red fluorescence, Amersham Biosciences).
Extraction-Mucosa was scraped of frozen intestinal pieces (ϳ0.1 g), suspended in ice-cold PBS containing protease inhibitor mixture (1:100, Sigma Aldrich, catalog number P8340), and homogenized in 1% trifluoroacetic acid (10 ml/g of tissue). An equal volume of 12 mol/liter urea was added, and the homogenate was centrifuged. The supernatant was retained for gel filtration. Furthermore, small intestines from PC1/3 null mice and controls were extracted using a boiling method or acidic ethanol previously described (11,22). When measuring intestinal GIP content in PC2 Ϫ/Ϫ mice and controls, frozen small intestines were extracted in acidic ethanol as previously described in Holst and Bersani (22), modified to exclude the urea and ether step. The supernatant was retained for protein measurement.
Construction of Adenovirus-An adenovirus construct expressing PC1/3 (AdPC1/3) or PC2 (AdPC2) was previously described in Irminger et al. (23). Adenovirus expressing green fluorescent protein (GFP) from the pACCMV vector (24) was a gift from Ole H. Mortensen (the Panum Institute). cDNA encoding rat preproGIP was amplified by PCR using Pfu polymerase. The amplified fragment was ligated to the pAC-TRECMV vector using the EcoRI and HindIII sites. The resulting plasmid was co-transfected with the PBGH plasmid into HEK293 cells using Superfect transfection reagent (Qiagen). The PBGH vector (Microbix, Inc.) is a modified adenovirus vector expressing the reverse Tet repressor from the deleted E3 region and contains homologous regions for recombination with the pACTRECMV plasmid. The pACTRECMV plasmid (a generous gift from Prof. C.B. Newgard, Duke University Medical Center) was derived from pACCMV (25) by insertion of a Tet-responsive element at the 3Ј end of the CMV promoter. An adenovirus not encoding a protein (AdNil) was made from co-transfection of the pACCMV and pJM17 plasmid as described in Becker et al. (24). After ϳ3 weeks, cell lysates were collected, and recombinant E1/E3 deleted virus was cloned by plaque assay. DNA was isolated from the recombinant clone, and the inserted sequence was amplified with the flanking primers: sense, TGGGAGGTCTATATAAGCAG; antisense, TCTCTGTAGGTAGTTTGTCC. The PCR product was purified, and sequence and orientation were verified on an ABI310 sequencer (PerkinElmer Life Sciences). The virus stocks were amplified in HEK293 cells and purified on a CsCl gradient. All procedures and the pACCMV vector are described in Becker et al. (24). The infection titer of the viral stocks was determined using the Adeno-X rapid titer kit (Clontech).
Isolation of Pancreatic Islets-Mouse and rat islets of Langerhans were isolated after collagenase digestion of pancreatic tissues and prepared as previously described (26).
Secretion Experiment-GH 4 cells (1 ϫ 10 6 ) were grown in 35-mm polylysine-coated culture dishes and incubated for 48 h at 37°C with 7 m.o.i. AdpreproGIP and 3 m.o.i. AdPC1/3, AdPC2, or AdNil as described above. Cells were washed 4 times with PBS and incubated for 10 min in serum-free medium (1 ml/dish) containing 0.1% human serum albumin with ionomycin (10 mol/liter) and 12-O-tetradecanoylphorbol-13-acetate (0.15 mol/liter) as secretagogues or without secretagogues. PBS and media were prewarmed to 37°C. Triplicate dishes were used for all incubation conditions. After incubation, cell culture dishes were cooled on ice, medium was withdrawn, and cells were washed 4 times in cold PBS. Cells were scraped off in PBS containing 0.1% human serum albumin and protease inhibitor mixture, centrifuged, and sonicated as described above. Media and cell lysates were analyzed by radioimmunoassay (RIA) using the GIP antibody R65.
Flow Cytometry-To obtain GIP antibody for flow cytometry, 100 g of 3.65H monoclonal antibody peritoneal exudate was conjugated using Alexa Fluor 647 monoclonal antibody labeling kit (Molecular Probes). For analysis, GH 4 cells (4 ϫ 10 6 cells/10 cm dish) were infected with 7 m.o.i. AdpreproGIP and 3 m.o.i. AdGFP or with 10 m.o.i. AdPC1/3 as control. Doxycyclin (10 g/ml) was added as an inducer of preproGIP expression. The protocol for intracellular staining has been described previously and modified to exclude a peptide incubation step (28). Briefly, cells were incubated for 48 h at 37°C, loosened with EDTA, fixed in 1% paraformaldehyde, permeabilized using 0.5% saponin, and stained using Alexa Fluor 647-conjugated monoclonal antibody 3.65H at 1:100 dilution. Cells were then washed and run on a FACSCalibur system (BD Biosciences), and results were analyzed using Cell Quest software. For analysis, cells were gated on GIP-positive cells. AdPC1/3-infected cells were used to define cut-off values.
GIP Receptor Signaling-Receptor signaling experiments were carried out using COS-7 cells transfected with the human GIP receptor. Transfection of the COS-7 cells was performed by the calcium phosphate precipitation method (29). For the cAMP accumulation assay, the transiently transfected cells (2.5 ϫ 10 5 cells/well) were incubated for 24 h with 2 Ci/ml of [ 3 H]adenine in 0.5 ml of growth medium per well. Cells were washed twice in HBS buffer (25 mM Hepes (pH 7.2), supplemented with 0.75 mM NaH 2 PO 4 , 140 mM NaCl and 0.05% (w/v) bovine serum albumin, and then 0.5 ml of HBS buffer supplemented with 1 mM phosphodiesterase inhibitor 3-isobutyl-1-methylxanthine (Sigma) was added with increasing concentrations of the native porcine GIP (Bachem) or two different concentrations (30 and 90 pM, as determined by the N-terminal assay (98171)) of GIP immunoreactivity obtained from gel filtration of GH 4 cells co-transfected with preproGIP and PC1/3 or PC2. The gel filtration fractions used for further analysis represented the GIP standard position. After 15 min of incubation at 37°C, the cells were placed on ice, the medium was removed, and the cells were lysed in 1 ml of 5% (w/v) trichloroacetic acid supplemented with 0.1 mM cAMP and 0.1 mM ATP for 30 min. The lysate mixtures were loaded onto Dowex columns (Bio-Rad), which were washed with 2 ml of water and placed onto the top of alumina columns (Sigma) and washed again with 10 ml of water. The alumina columns were eluted with 6 ml of 0.1 M imidazole into 15 ml of scintillation fluid (Highsafe III) and counted. Columns were re-used up to 15 times. Dowex columns were regenerated by adding 10 ml of 2 N HCl followed by 10 ml of water; the alumina columns were regenerated by adding 2 ml of 1 M imidazole, 10 ml of 0.1 M imidazole, and finally 5 ml of water. Determinations were made in duplicate.
Gel Filtration-Acidic intestinal extracts were applied to a K16/100 Sephadex G50, fine-grade column (Amersham Biosciences) equilibrated and eluted with 0.5 mol/liter acetic acid. Cell lysates were applied to a similar column equilibrated and eluted with assay buffer (described above). For both columns, flow rates were 16 -21 ml/h at 4°C, and sample size never exceeded 2% of total column volume. 125 I-Labeled albumin and 22 NaCl were added in trace amounts for internal calibration. The coefficient of distribution (K d ) was calculated using the is the elution position of the peak in question, V 0 is the elution volume of 125 I-labeled albumin, and V i is the inner volume of the column, determined as the difference between the elution volumes of 22 NaCl and V 0 . When the processing profile could not be foreseen, as in extracts from PC1/3-deficient mice and cell lysates, only 22 NaCl was added, and the coefficient of distribution (K d ) was calculated from this tracer and previous calibrations. Intestinal murine and rat GIP-(1-42) were used as standards, extracted similarly to the extracts and cell lysates investigated. Furthermore, synthetic rat GIP-(34 -42) was applied as a standard (Schafer-N). Fractions eluted with acetic acid were dried in a vacuum centrifuge (Heto-Holten A/S, Alleroed, Denmark) and reconstituted in assay buffer. All fractions were analyzed by RIA.
RIA-Three different RIAs for GIP immunoreactivity were used. Briefly, antiserum RIC34 (30) (kindly provided by Dr. Linda M. Morgan, University of Surrey, Guildford, UK) is raised to natural porcine GIP and recognizes N-terminal truncated forms and is here shown to also recognize extended forms (see Fig. 4). It is referred to as "side-viewing." Antiserum 98171 is directed toward the N-terminal GIP-(1-10) and binds N-and C-terminal-extended forms (Fig. 4) and require an intact N terminus (31). Antiserum R65 (32) recognizes the C terminus, crossreacts with synthetic C-terminal fragment GIP-(34 -42) (proGIP-(56 -64)) (Schafer-N) but not with synthetic proGIP-(56 -68) (Schafer-N), a C-terminally extended form of GIP, and is therefore, considered to be specific for the correctly processed C terminus of GIP. Furthermore, it does not recognize the so-called GIP8000. Human GIP was used in all assays for standard and tracer preparation.
Statistical Analysis-Results are presented as the mean Ϯ S.E. unless indicated otherwise. Statistical significance was analyzed by Student's t test for unpaired values of equal variance or by Mann-Whitney U test when analyzing the secretion ratios within a group of transfected cells. Comparison of three samples was analyzed by one-way analysis of variance. Values were considered significantly different when p Ͻ 0.05.

Structure of the GIP Precursor, GIP and PC1/3 Co-localization and
PreproGIP mRNA Expression-The GIP precursor is a protein of 144 (rodent) or 153 (human) amino acids (33)(34)(35) expressed in the endocrine intestinal K cells that exhibits the highest density in the proximal jejunum but is found in the entire intestinal mucosa (17). Normal processing is believed to depend on cleavage at the conserved sites 19 RGPR 22 2 at the N terminus and 65 R2EAR 68 at the C terminus, resulting in the biologically active peptide GIP (proGIP-(23-64)) and an N-terminal and C-terminal fragment (Fig. 1A).
Because PC1/3 is expressed throughout the intestine (11) and known to cleave the single Arg site in GLP-1-(1-37) (7, 36), co-localization of PC1/3 and GIP was analyzed by immunohistochemistry in intestinal sections from PC1/3 null and wild-type mice. In agreement with previous studies, the density of the GIP-immunoreactive cells was highest in the upper part of the small intestine, and GIP immunoreactivity was found to be co-localized with PC1/3 immunoreactivity (Fig. 1B), whereas PC2 immunoreactive cells was not found (data not shown). Furthermore, visual levels of GIP immunoreactivity in intestinal sections were similar in PC1/3 null and wild-type mice (data not shown). To investigate if preproGIP expression was normal in PC1/3 null mice, preproGIP mRNA expression in the proximal jejunum was analyzed by Q-PCR. In agreement with the immunohistochemical findings, similar levels of preproGIP mRNA were found in PC1/3 null mice and controls (see Fig. 3C).
Processing of ProGIP in PC1/3 and PC2 Null Mice-ProGIP processing profile was investigated in PC1/3 and PC2 Ϫ/Ϫ mice and matched controls after gel filtration of acidic intestinal extracts and RIA using antisera directed to different regions of the GIP sequence (Fig. 1A). We found efficient processing of proGIP to GIP in wild-type mice, as evidenced by a single peak eluting in the standard position, whereas in extracts from PC1/3 null mice nearly no mature GIP could be detected ( Fig. 2A). Similar findings were made in neutral extracts of whole small intestines (data not shown). Surprisingly, in neither of the extracts was a peak corresponding to proGIP detected despite the apparent block in processing of the precursor. In contrast, when comparing the proGIP FIGURE 1. Structure and expression of rodent prepro-glucose-dependent insulinotropic polypeptide (preproGIP). A, schematic representation of the structure of rodent preproGIP. Black half circles represent the antisera used in this study, and their positions indicate their suggested sequence specificity. Arrows indicate the suggested cleavage site. SP, signal peptide. B, co-localization of GIP and PC1/3 in immunohistochemical staining of wild-type mouse small intestine was studied using monoclonal GIP antibody (green) and polyclonal PC1/3 antibody (red). Using a double filter, the red and green fluorescence signals are shown simultaneously (middle picture) and appear yellowish. Double-stained cells are indicated by arrowheads. C, relative expression levels of preproGIP mRNA investigated by Q-PCR demonstrate similar levels in PC1/3 wild-type (ϩ/ϩ, black bar) and null (Ϫ/Ϫ, white bar) mice. Data are shown as mean Ϯ S.D., n ϭ 3 in both groups. processing profile in PC2 Ϫ/Ϫ mice and controls, no difference in processing profile could be detected (Fig. 2B). In addition, measurements of GIP tissue concentrations in acidic extracts using the C-terminal-specific antibody demonstrated similar levels of GIP in PC2 Ϫ/Ϫ mice and controls (wild type, 172.3 pmol/g (157.1-208.2); PC2 Ϫ/Ϫ, 164.7 pmol/g (141.5-180.6); median and range, n ϭ 3 in each group, p ϭ 0.42).
Expression of Cell Line PCs and Infection of Cell Lines-Cell line studies were performed to determine whether PC1/3 alone is able to mimic the processing profile of proGIP detected in vivo or whether other convertases like PC2 have the specificity required. Here we focused on the AtT-20, ␣-TC1.9, and GH 4 cell lines because several studies have demonstrated that they can be efficiently used for prohormone and prohormone convertase overexpression studies and, thus, possess the cofactors needed for prohormone convertase-mediated processing (6 -9, 37, 38). First, to thoroughly examine the expression of PC1/3 and PC2 in AtT-20, ␣-TC1.9, and GH 4 cell lines, we performed Q-PCR after isolation of mRNA (Table 1). Isolated islets were cultured overnight and analyzed in parallel as rough indications of relevant expression levels of the convertases in vivo. In accordance with previous reports, we found that AtT-20 cells express large amounts of PC1/3 but low levels of PC2, the ␣-TC1.9 cell line expresses large amounts of PC2 but no PC1/3, and the GH 4 cells express no PC2 and low levels of PC1/3. AtT-20 and ␣-TC1.9 cells were infected with recombinant adenovirus expressing preproGIP (AdpreproGIP), and GH 4 cells were co-infected with Adpre-proGIP and an adenovirus expressing either GFP, PC1/3, or PC2. To confirm that the recombinant adenoviruses were efficiently expressed in the GH 4 cells, we performed immunocytochemistry and found high levels of positively stained cells (Fig. 3A). Furthermore, to investigate whether the adenovirally encoded proteins were sorted to the regulated secretory pathway, stimulation of secretion with ionomycin and 12-Otetradecanoylphorbol-13-acetate as secretagogues was performed in GH 4 cells after infection with AdpreproGIP and AdPC1/3 or AdPC2 or AdNil (Fig. 3B). Cells stimulated for 10 min were found to secrete significantly higher levels of GIP compared to unstimulated cells (p Ͻ 0.05), indicating a correct subcellular sorting of adenovirally encoded proteins. Furthermore, no difference in secretion ratio could be detected between the three groups of stimulated cells (p ϭ 0.957). Finally, as the accurate analysis of proGIP processing in GH 4 cells required efficient co-transfection, we analyzed co-infection efficiencies by flow cytometry. The analysis indicated that with the viral dosage used, 91% of the AdpreproGIP (7 m.o.i) infected cells would also be infected with the low dosage virus (3 m.o.i) (Fig. 3C).
Processing Profiles in Infected Cell Lines-Lysates from infected cells were subjected to gel filtration and analyzed by RIA using antisera directed at different regions of the mature GIP sequence (Fig. 1A). Representative gel filtration processing profiles from cell lines exposed to different combinations of recombinant adenoviruses are shown in Fig. 4. AtT-20 cells were found to fully process the GIP precursor, resulting in a single peak eluting in the standard position of mature GIP detected by RIA with all three antisera. We found no intermediate forms in the cell lysate. The processing profile investigated in ␣-TC1.9 cell line showed increased immunoreactivity at the GIP standard position with all three antisera indicating that PC2 is able to produce GIP. However, levels and appearance of the curves varied between assays, suggesting that other intermediate forms are produced. Furthermore, when using the C-terminal-specific antiserum, we measured an additional peak (only the ascending part is shown in the figure) eluting in the total volume (K d 1) of the column. This was speculated to be a result of cleavage at the K 54 K 55 site present at the C-terminal of GIP, as cleavage at this site would result in the production of a nine-amino acid fragment with  a molecular size smaller than the lowest fractionation range of the column.
These studies indicate different processing of the GIP precursor in PC1/3and PC2-producing cell lines. However, it cannot be excluded that other cellular endogenously expressed convertases may play a role. To investigate the direct role of PC1/3 and PC2 in the processing of proGIP, co-transfections were carried out in GH 4 cells. Analysis of GH 4 cells co-transfected with AdpreproGIP and AdGFP showed accumulation of an immature form believed to be proGIP because of its early elution position measured with the side-viewing and N-terminal assays (Fig. 4). In addition and applying to all assays, lower levels of GIP immunoreactivity were found at the GIP standard position, in agreement with the presence of low, but significant endogenous levels of PC1/3 in GH 4 cells. When GH 4 cells were co-infected with AdpreproGIP and AdPC1/3, the processing profile shifted from a predominantly early elution position to an elution position corresponding to the GIP standard position. No intermediate forms were detected, but low levels of a larger unprocessed form remained after AdPC1/3 infections. GH 4 cells co-infected with AdpreproGIP and AdPC2 largely reproduced the processing profile found in the ␣-TC1.9 cell line (Fig. 4). GIP immunoreactivity at the standard position was detected with all assays, and the finding of a peak eluting at K d 1.0, detected by the C-terminal assay only, confirmed the finding of an additional PC2 mediated C-terminal cleavage.
In addition, when GH 4 cells were co-infected with AdpreproGIP (7 m.o.i), AdPC1/3 (3 m.o.i), and AdPC2 (3 m.o.i), characteristics from both the PC1/ 3-specific and PC2-specific processing profiles were found (data not shown). These data indicate that PC1/3 is sufficient for cleavage of the GIP precursor to mature GIP and that PC2 also is able to produce the mature GIP product from proGIP but in addition mediates a C-terminal cleavage not previously described in in vivo studies (18). To investigate these findings further, GIP immunoreactive material, eluting at the GIP standard position found by gel filtration, was applied to HPLC and analyzed by RIA using the N-and C-terminal assays. A single peak in the standard position was detected by both antisera when analyzing the AtT-20 cell line processing profile and confirmed by HPLC analysis of GH 4 cells infected with Adpre-proGIP and AdPC1/3 (Fig. 5). When analyzing the ␣-TC1.9 cell line, we detected a peak corresponding to mature GIP with the N-and C-terminal antisera (Fig. 5). However, an additional peak was found when investigating N-terminal GIP immunoreactivity, in agreement with a PC2-derived C-terminal cleavage, resulting in a C-terminal-truncated GIP. Similar results were obtained by analysis of GH 4 cells co-infected with AdpreproGIP and AdPC2 (Fig. 5).
Because the C-terminal processing was speculated to be a result of a cleavage at the dibasic site K 54 K 55 , present in the GIP sequence and resulting in the 9-amino acid peptide, proGIP-(56 -64), the C-terminal immunoreactive material found by gel filtration eluting at K d 1.0 was analyzed by HPLC using synthetic proGIP-(56 -64) as a standard. The immunoreactive material from ␣-TC1.9 and GH 4 cells was found to elute in two peaks, whereas one peak corresponded to the synthetic proGIP-(56 -64) standard, confirming PC2-mediated processing at the K 54 K 55 site (data not shown).
Finally, to investigate whether the GIP immunoreactive material from gel filtration of co-transfected GH 4 cells was indeed biological active, we measured cAMP accumulation in COS-7 cells transfected with the GIP receptor. In accordance with the previously reported intracellular signaling pathway of the receptor (39, 40), we observed dose-  responsive cAMP accumulation when GIP receptor-transfected cells were exposed to synthetic GIP (Fig. 6A). Furthermore, 30 and 90 pmol/ liter of GIP immunoreactive material from GH 4 cells co-transfected with AdpreproGIP and AdPC1/3 or AdPC2 resulted in an increase in cAMP in transfected cells compared with nontransfected cells (Fig. 6B).

DISCUSSION
The results presented above indicate that PC1/3 is both essential and sufficient for the cleavage of proGIP to GIP. In agreement with this, we found by immunohistochemistry PC1/3 to be expressed throughout the murine intestine and all GIP-immunoreactive cells to co-stain for PC1/3. An essential role for PC1/3 in proGIP processing was demonstrated by analysis of both acidic and neutral intestinal extracts applied to gel filtration. This revealed severely impaired processing of the GIP precursor to GIP in PC1/3-deficient mice compared with controls, whereas an unaltered processing profile was demonstrated in PC2 Ϫ/Ϫ mice. That PC1/3 is sufficient for the processing was confirmed in AtT-20 cells after infection with AdpreproGIP and in GH 4 cells coinfected with AdpreproGIP and AdPC1/3. PC1/3-mediated processing likely results in cleavage after the Arg-65 in the proGIP sequence. An additional C-terminal trimming is then performed by another cellular carboxypeptidase. Which carboxypeptidase is responsible for this conversion has not been investigated, but carboxypeptidase E is a likely candidate (41). This has, however, not yet been investigated. Nevertheless, GIP produced in GH 4 cells after co-transfection was demonstrated to be biologically active. Thus, GIP immunoreactive material eluting in the GIP standard position found by gel filtration significantly mediated cAMP accumulation in COS-7 cells transfected with the GIP receptor. Recently, we characterized the intestinal proglucagon processing in PC1/3 null mice and demonstrated an accumulation of proglucagon, in agreement with impaired processing of the precursor (11). A similar proGIP accumulation would be suspected in this case. In agreement with continued transcription of the preproGIP gene, we found at least as high levels of preproGIP mRNA in null mice compared with controls (Fig. 1C). Furthermore, GIP immunoreactive cells were found in PC1/3 wild-type as well as in null mice by immunohistochemistry, in accordance with translation of the sequence (Fig. 1B). Despite these results, we were not able to demonstrate accumulation of proGIP in neither acidic nor neutral intestinal extracts from PC1/3-deficient mice, indicating that precipitation or immediate degradation might occur in these cells. When GH 4 cells not expressing PC1/3 or PC2 in biologically relevant levels were co-infected with AdpreproGIP and AdGFP, we found accumulation of a large immature form believed to represent proGIP, suggesting that the side-viewing and N-terminal-directed antisera employed are, indeed, capable of detecting proGIP.
In addition, we found by gel filtration (Fig. 4) and HPLC (data not shown) GH 4 cells to produce some GIP without supplement of recombinant convertase, indicating that an endogenously expressed convertase is able to mediate the processing of proGIP to GIP. As previously indicated, expression levels of PC1/3 and PC2 in GH 4 cells were lower than expression levels found in rat pancreatic islets. However, detectable levels of PC1/3, but not PC2, were found, and this may be sufficient for the partial cleavage demonstrated. Other convertases like PC5/6A and furin may play a role. Thus, PC5/6A was found to be expressed in GH 4 cells (data not shown). PC5/6A has also been reported to be widely expressed in the small intestine (42,43), suggesting that PC5/6A play a role in the processing of a variety of intestinal propeptides. However, lack of intestinal GIP in the PC1/3 null mice indicates that other convertases cannot substitute for PC1/3 in proGIP processing in vivo. Furthermore, PC1/3 was found to be sufficient for the cleavage of proGIP to GIP in the GH 4 cell line, as co-infection with AdpreproGIP and AdPC1/3 resulted in a shift from an early elution position to an elution position corresponding to GIP (Fig. 4) confirmed by HPLC (Fig. 5). No intermediate forms were detected, indicating efficient processing, and much lower levels of an apparently unprocessed form was found, possibly in agreement with our finding that ϳ9% of cells are predicted to express AdpreproGIP without AdPC1/3 and, therefore, probably produce unprocessed proGIP.
Previous studies have examined gel filtration profiles from porcine intestinal extracts analyzed by RIA using different antisera raised against GIP (18,19). Aside from the known biologically active peptide, GIP, a larger fragment, designated GIP8000, has been identified in human and porcine extracts from proximal jejunum by a variety of cross-reacting antisera raised against porcine GIP. GIP8000 was subsequently suggested to be a precursor product. This has, however, not been confirmed (20). Gel filtration profile analysis of intestinal neutral and acidic extracts from PC1/3 wild-type mice using a side-viewing antiserum (RIC34) previous reported to detect human and porcine GIP8000 do not clearly demonstrate a peak corresponding to GIP8000, indicating that the antibody either does not cross-react with murine GIP8000 or that the peptide is not present in significant levels in mice.
The processing of protein precursors has been demonstrated to depend on the convertase expressed. Accordingly, analysis of gel filtration profile from PC2-producing cell lines indicated that PC2 is able to mediate the cleavage of proGIP to biologically active GIP but, in addition to other fragments, with molecular sizes larger and smaller than GIP. Thus, the observed processing could be argued to reflect a nonspecific degradation of wrongly sorted recombinant proGIP and not a PC2-mediated processing. However, GH 4 cells co-transfected with preproGIP and PC2 or PC1/3 or an empty vector and stimulated with 12-O-tetradecanoylphorbol-13-acetate and ionomycin for 10 min secreted significantly and similarly higher levels of GIP compared with nonstimulated cells, suggesting that all recombinant proteins in the overexpression studies are correctly targeted to the regulated secretory pathway independent of simultaneous prohormone convertase overexpression. In addition, the observed PC2-mediated products have not been described in previous reports on GIP gel filtration profile analysis of intestinal extracts (18,19), indicating that PC2 does not seem to play a role in vivo. This is supported by that fact that we did not find PC2 to be expressed in GIP-producing cells of the murine intestine by immunohistochemistry (data not shown) and the fact that gel filtration profiles of intestinal extracts from PC2 Ϫ/Ϫ mice resemble the profile demonstrated in wild-type mice and that levels of intestinal GIP tissue concentration are similar in PC2 null mice and controls. Beyond this, the PC2-mediated processing profile resulting in the formation of novel GIP fragments could represent a tissue-dependent processing where nonintestinal tissues expressing PC2 could contribute with this component, as has been observed for the processing of proglucagon. However, except from the submandibular gland, proGIP has not yet been reported to be expressed in other tissues than the gastrointestinal tract (34).
Reports on PC2 expression in the small intestine are, however, conflicting. PC2 was demonstrated by double immunohistochemistry to be expressed in the intestinal canine L-cell (44). In addition, genetically engineered K-cells were reported to produce biological effects of insulin when proinsulin was expressed under the control of the GIP promoter in mice; however, proinsulin processing was not evaluated in the study (45).We previously reported that PC2 could not by immunohistochemistry be localized to jejunum or ileum in mice (11). Similar results were obtained by immunohistochemistry in duodenal sections from mice and in rat and porcine intestinal sections. 3 In addition, PC2 could not by RT PCR be detected in any sections of the murine intestine. 4 In conclusion, the results presented here strongly suggest an essential and sufficient role for PC1/3 in processing of proGIP to the biologically active incretin hormone GIP, whereas PC2, although capable of cleaving pro-GIP, is not essential for proGIP processing in vivo.