Glucose-dependent insulinotropic polypeptide (GIP) is a hormone released from enteroendocrine K cells in response to meals. Posttranslational processing of the precursor protein pro-GIP at residue 65 by proprotein convertase subtilisin/kexin type 1 (PC1/3) in gut K cells gives rise to the established 42-amino-acid form of GIP (GIP1–42). However, the pro-GIP peptide sequence contains a consensus cleavage site for PC2 at residues 52–55 and we identified PC2 immunoreactivity in a subset of K cells, suggesting the potential existence of a COOH-terminal truncated GIP isoform, GIP1–30. Indeed a subset of mouse and human K cells display GIP immunoreactivity with GIP antibodies directed to the mid portion of the peptide, but not with a COOH-terminal-directed GIP antibody, indicative of the presence of a truncated form of GIP. This population of cells represents ∼5–15% of the total GIP-immunoreactive cells in mice, depending on the region of intestine, and is virtually absent in mice lacking PC2. Amidated GIP1–30 and GIP1–42 have comparable potency at stimulating somatostatin release in the perfused mouse stomach. Therefore, GIP1–30 represents a naturally occurring, biologically active form of GIP.
- perfused stomach
glucose-dependent insulinotropic polypeptide, also known as gastric inhibitory polypeptide or GIP, is a gastrointestinal hormone released in a meal-dependent manner by enteroendocrine K cells in the gut mucosa (1, 3). As the names imply, GIP secretion following meals serves both to reduce gastric acid secretion, via the release of the inhibitory peptide somatostatin (26), and to promote brisk release of insulin to dispense incoming nutrients (5). The ability of GIP to promote glucose-dependent secretion of insulin, while also augmenting insulin gene expression (40) and protecting β-cells from apoptosis (7, 20, 37–38), has generated considerable interest in the potential utility of GIP agonists for the treatment of Type 2 diabetes mellitus (27). However, there is increasing awareness that GIP has additional important metabolic actions, including promoting fat deposition in adipocytes (6, 19, 21, 33). Although GIP receptor knockout mice display glucose intolerance as a result of delayed insulin secretion (29), they are also protected from diet-induced obesity (28). Thus both vaccination against GIP and treatment with GIP antagonists are being explored as novel approaches to treat obesity (14).
GIP is commonly referred to as a 42-amino-acid polypeptide hormone (GIP1–42), generated by cleavage of the larger precursor protein pro-GIP at arginine residues by proprotein convertase subtilisin/kexin type 1 (PC1/3) (39). Interestingly however, an analysis of the GIP gene in different vertebrates suggests that certain species such as the zebrafish (Danio rerio) produces GIP1–31 instead of GIP1–42 (13). Although GIP1–31 is not currently recognized as a natural form of GIP in mammals, mammalian pro-GIP contains a highly conserved PC2 cleavage site (KGKK) at residues 52–55 (13). A recent analysis of the GIP fragments following coinfection of GH4 cells with adenoviral vectors expressing pro-GIP and PC2 confirmed that GIP1–31 could be produced (39). Although this same study failed to detect any PC2 in K cells and thus concluded that GIP1–31 is not naturally produced, it is possible that the GIP antisera used were unable to detect these cells. For instance, presumably K cells expressing GIP1–31 will not be identified with antibodies recognizing COOH-terminal epitope of GIP (i.e., GIP34–42). Therefore, we decided to reevaluate the possibility of differing forms of GIP in mammalian enteroendocrine cells using several different GIP antibodies. Our studies reveal that, in addition to K cells producing GIP1–42, there exists a significant population of K cells that produce a COOH-terminal truncated form of GIP, likely GIP1–31 and/or amidated GIP1–30. We also demonstrate that amidated GIP1–30 and GIP1–42 are equipotent in stimulating the release of somatostatin in the perfused mouse stomach.
MATERIALS AND METHODS
Animal and human tissues.
All animal experiments were approved by the University of British Columbia Committee on Animal Care in accordance with guidelines set by the Canadian Council on Animal Care. For tissue extraction, mice were anesthetized via inhalation of 2% isoflurane (AErrane, Baxter, Mississauga, ON, Canada). Human duodenum was provided by the Irving K. Barber Human Islet Isolation Laboratory (Vancouver, BC, Canada) with consent for use for research purposes. All peptides were purchased from American Peptide (Sunnyvale, CA).
Tissues collected from wild-type C57BL/6 mice (Jackson Laboratories, Bar Harbor, ME) and PC2 knockout mice, kindly provided by Dr. Donald Steiner (University of Chicago, Chicago, IL) were fixed in 4% paraformaldehyde, embedded in paraffin, and sectioned (Wax-it Histology Services, Vancouver, BC, Canada). After dewaxing and rehydration, heat-induced epitope retrieval was performed in 10 mM citrate buffer (0.05% Tween 20; pH 6) for 10 min at 95°C by use of EZ-Retriever Microwave (BioGenex, San Roman, CA). Slides were then incubated overnight at 4°C in primary antibodies against GIP, PC2, PC1/3, and furin (Table 1). For preincubation studies, 10 μg of GIP1–42 or GIP1–30 peptide, prepared in double-distilled H2O, was added to anti-GIP antibodies (Ab#1 and Ab#3) and incubated for 1 h at room temperature and then added to the slides and incubated overnight at 4°C. Slides were then washed 3 × 10 min in PBS and then sequentially stained with GIP antibodies from different species for 1 h at room temperature. Secondary incubations were then performed for 1 h at room temperature in Alexa Fluor-conjugated secondary antibody (Alexa Fluor 488 or Alexa Fluor 594; Molecular Probes, Eugene, OR; 1:1,000). Slides were mounted in Vectashield Mounting Medium (Vector Laboratories, Burlingame, CA). For quantification analysis, longitudinal sections of whole intestine of three C57BL/6 wild-type and two PC2 knockout mice were coimmunostained for GIP using Ab#3 and either Ab#1 or Ab#2. On average ∼1,000 K cells were counted per animal and colocalization of two antibodies was examined. For C57BL/6 mice (n = 3), results for each intestinal segment from all mice are averaged and expressed as means ± SE. For PC2 knockout mice (n = 2), the average result for each intestinal segment for the two mice are shown. Images were collected using an Axiovert 200 microscope (Carl Zeiss, Toronto, ON, Canada) connected to a digital camera (Retiga 2000R, QImaging, Surrey, BC, Canada) controlled with Openlab 5.2 software (Perkin Elmer, Waltham, MA).
The peptides used for perfusion, porcine GIP1–42, and amidated GIP1–30 were dissolved in 0.01 M acetic acid containing 0.2% BSA and stored lyophilized in 1 μg aliquots at −20°C. A fresh aliquot was used for each experimental day and reconstituted in 1 ml 0.01 M acetic acid and then diluted to 1.0 nM in perfusate for each perfusion. Male C57BL/6 mice (12–15 wk old) were fasted for 12 h and anesthetized with sodium pentobarbital (65 mg/kg; Bimeda-MTC Animal Health, Cambridge, ON, Canada). Surgical procedures were previously described (42). In brief, an abdominal midline incision was made to expose the stomach. Vasculatures supplying the kidneys, adrenal glands, spleen, pancreas, and intestines were tied off or cut between double ligatures. Care was taken to dissect the pancreas, spleen, and intestines away from the stomach and remove them from the preparation. A duodenal cannula was placed into the duodenum into the stomach to drain gastric content. An aortic cannula was placed just below the celiac artery to allow the flow of perfusate into the stomach via the celiac artery. A portal vein cannula was placed to collect the effluents from the gastric vasculatures. The preparation and perfusate were thermostatically controlled at 37°C. Arterial perfusion began with a flush of heparin saline (300 U/ml; Sigma-Aldrich; St. Louis, MO) into the aortic cannula, followed by Krebs-Ringer bicarbonate buffer perfusate containing 3% dextran, 0.2% BSA, and 5.1 mM glucose (preequilibrated with 95% O2-5% CO2) at a rate of 1 ml/min with a peristaltic pump. Following a 30-min equilibration period, portal vein effluent was collected at 2-min intervals. A GIP gradient (0–1.0 nM) was introduced to the perfusate via a 100-ml gradient mixer consisting of two connected, identical perfusion flasks. Fractions were aliquoted, transferred to cold tubes containing 25 μl Trasylol (aprotinin, 10,000 KIU/ml Bayer, Etoblcoke, ON, Canada) and stored at −20°C.
Radioimmunoassay and statistical analysis.
The specific radioimmunoassay used for the measurement of somatostatin-like immunoreactivity (SLI) in samples has been described previously (25, 42). The monoclonal antibody SOMA-3 was used (University of British Columbia Regulatory Peptides Group, Vancouver, BC, Canada; Ref. 25). SLI release data are expressed as a percentage release of the first fraction. Data from mice in the same treatment group were averaged and are expressed as means ± SE. Area under the curve (AUC) was calculated by averaging the AUC of each mouse in the same group and is expressed as means ± SE. Statistical significance was assessed by one-way ANOVA via Prism 5.01 software (GraphPad Software, La Jolla, CA).
There are two distinct forms of GIP expressed in gut.
It is well established that GIP1–42 is expressed in mammalian K cells. Using double immunofluorescence techniques and antibodies that recognize different regions of the GIP peptide, we investigated the existence of shorter GIP isoforms in mouse and human intestine. We used two antibodies raised against the midportion of GIP (Ab#1 or #2) along with a monoclonal antibody (3.65H) directed against the COOH terminus of GIP (Ab#3) (Fig. 1A). As shown in Fig. 1B, there are two different GIP-immunoreactive-positive cell populations in mouse and human intestine. One population is immunoreactive to Ab#3 (COOH terminus) and Ab#1 or #2 (midportion) (Fig. 1B; rows 1 and 3), whereas a second cell population is immunoreactive to Ab#1 or #2 but not Ab#3 (Fig. 1B; rows 2 and 4). We never observed a cell that was only immunoreactive to Ab#3. These observations suggest that the first cell population expresses classical full-length GIP1–42 whereas the second cell population expresses a shorter GIP isoform to which the COOH terminus-specific antibody (Ab#3) does not bind. Thus, aside from the full-length GIP1–42, there is a shorter isoform of GIP with a COOH terminus truncation.
To further investigate the specificity of these GIP antibodies, we preincubated the antibodies with either amidated GIP1–30 or GIP1–42 peptide prior to immunostaining of the tissue sections. As shown in Fig. 2, GIP immunoreactivity was blocked when Ab#3 was preincubated with GIP1–42 but not with amidated GIP1–30 (Fig. 2B). In contrast, preincubation of Ab#1 (Fig. 2C) and Ab#2 (data not shown) with either GIP1–42 or amidated GIP1–30 abrogated GIP immunoreactivity. Controls are shown in Fig. 2A. These results further support that Ab#3 is COOH terminus specific and does not bind to the COOH-terminal truncated GIP isoform. Conversely, Ab#1 and Ab#2 can detect both full-length and the truncated forms of GIP. GIP immunoreactivity for all three antibodies was not blocked by preincubation with either GLP-1 or glucagon peptides (data not shown).
Furin is likely not involved in pro-GIP processing.
Furin is a member of the proprotein convertase subtilisin/kexin (PCSK) family that is reportedly present in the majority of GIP-positive cells in wild-type mouse intestine (11). To further explore whether furin is potentially involved in pro-GIP processing, we examined the colocalization of furin with GIP in wild-type mouse intestine using three different anti-furin antibodies. In agreement with Gagnon et al. (11), we observed ∼80% of GIP immunoreactivity colocalized with furin when we used the same anti-furin antibody (Zymed). However, when we used two other commercially available anti-furin antibodies (Novus Biologicals and R & D Systems) only ∼7% of GIP-immunoreactive-positive cells also contained furin immunoreactivity (representative images are shown in Fig. 3A). In addition, when we costained mouse intestine with the anti-furin antibodies from Novus and R & D, we observed complete colocalization of furin immunoreactivities (data not shown). These observations cast considerable doubt as to whether furin is indeed expressed in the majority of K cells and suggest that the Zymed furin antibody is cross-reacting with another protein in K cells.
PC1/3 is coexpressed with GIP1–42 and PC2 with the COOH-terminal truncated form of GIP.
Previous studies have demonstrated an important and essential role for PC1/3 in pro-GIP processing and that PC2 is capable of processing pro-GIP to biologically active GIP (39). To further examine the potential involvement of PC2 in processing of pro-GIP, we investigated colocalization of PC1/3 and PC2 with GIP in the intestine. Cells that were immunoreactive for Ab#3 were mostly (∼90%) immunoreactive for PC1/3 (Fig. 3B) but never for PC2 (Fig. 3C). This observation supports the idea that PC1/3 is important for processing of pro-GIP to GIP1–42 and that PC2 is not necessary for this process. Although cells that were immunoreactive for Ab#1 were also mostly (∼85%) immunoreactive for PC1/3 (Fig. 3B), some were immunoreactive for PC2 (Fig. 3C). To test the specificity of our two anti-PC1/3 antibodies, we coimmunostained the mouse intestine and observed that they both colocalized in the same cells (data not shown). These results indicate that the full-length GIP1–42, recognizable by both GIP Ab#1 and Ab#3, is coexpressed with PC1/3, whereas the COOH-terminal truncated form of GIP that only Ab#1 can recognize is coexpressed with PC2.
Truncated GIP isoform is virtually absent in gut of PC2 knockout mice.
To further assess the role of PC2 in the formation of the truncated GIP isoform, we performed immunohistological studies on intestine of PC2 knockout mice. As shown in Fig. 4A, we observed GIP immunoreactivity in the intestine of PC2 knockout mice with both the midportion specific GIP antibody (Ab#1) and the COOH terminus-specific GIP antibody (Ab#3). Interestingly, unlike the observation in wild-type mouse intestine, almost all the K cells observed in PC2 knockout mice were immunoreactive for both antibodies. We performed quantification analysis on different regions of intestine of three C57BL/6 wild-type and two PC2 knockout mice (Fig. 4B). In wild-type mice, a total of 3,325 K cells were visualized with Ab#1, and, although most of these cells were also immunoreactive for Ab#3, 212 (6.4%) did not display Ab#3 immunoreactivity, indicating these cells express a COOH-terminal truncated form of GIP. Interestingly, the percentage of K cells that were not immunoreactive with Ab#3 varied from 5 to 15%, depending on the region of intestine examined (Fig. 4B). In PC2 knockout mice, 1,733 K cells were visualized with Ab#1, and only 9 (0.5%) of these were not also immunoreactive for Ab#3. These observations support the idea that PC2 is involved in generating a truncated form of GIP that does not contain the COOH-terminal epitope of GIP1–42 that is recognized by Ab#3.
GIP1–42 and GIP1–30 equivalently induce somatostatin release from perfused mouse stomach.
GIP1–42 is known to promote somatostatin release from the stomach (15, 24, 26). To determine whether amidated GIP1–30 has similar potency as GIP1–42, we performed an in situ perfusion study using a vascularly perfused isolated mouse stomach preparation. As shown in Fig. 5, we observed equivalent stimulation of somatostatin release with gradient perfusion of either GIP1–42 or amidated GIP1–30 from 0 to 1.0 nM. The AUCs generated are also comparable.
In previous studies Gagnon et al. (11) reported that the majority of K cells express furin, leading them to conclude that furin might be involved in pro-GIP processing. In contrast, Ugleholdt et al. (39) reported that PC1/3 is essential and sufficient for processing pro-GIP to GIP1–42 and also that pro-GIP processing is impaired in PC1/3-null mice. Our observations favor the idea that furin is expressed in a minority of K cells and is therefore less significant in pro-GIP processing. On the other hand, our results are consistent with the findings by Ugleholdt et al. and suggest that the majority of K cells express GIP1–42 and PC1/3. Like these authors, we also did not detect PC2 immunoreactivity colocalized with GIP using the same monoclonal GIP antibody (Ab#3). However, our blocking studies with GIP1–42 and amidated GIP1–30 and our examination of tissues from PC2 knockout mice clearly indicate that Ab#3 is a COOH-terminal specific antibody, unable to recognize the form of GIP that is generated by PC2. This truncated form of GIP is, however, still recognizable by GIP Ab#1 and Ab#2. Importantly, Ugleholdt et al. demonstrated that PC2 cleaves pro-GIP at the K54K55 site in vitro, resulting in GIP1–31 (39). It is highly probable that GIP1–31 is converted to amidated GIP1–30 in vivo by ubiquitous peptidyl-glycine α-amidating monooxygenase (8). Therefore, we propose that the majority of K cells express PC1/3 but not PC2 and produce GIP1–42, but that there is a population of K cells that express PC2 and produce a COOH-terminal truncated isoform of GIP (GIP1–31), which is likely converted to amidated GIP1–30.
The potential existence of a minor component of GIP, aside from the predominant GIP1–42 and the dipeptidyl peptidase 4 product GIP3–42 (17), was reported by Jornvall et al. (16) in the pig upper intestine. The authors were uncertain of the sequence of this third component but speculated that it could be an additional processed form of GIP (16). Differential posttranslational processing resulting in multiple products has been described for other peptides such as somatostatin, gastrin, and cholecystokinin (CCK) (4, 23). In addition, the uneven distribution of varying forms of gastrin and CCK throughout the small intestine supports the notion of a nonuniform expression of posttranslational processing enzymes in the gut (11, 23). Our findings of increased expression of the COOH-terminal truncated form of GIP in the distal gut relative to proximal regions also suggests a nonuniform expression of the enzymes involved in pro-GIP processing. Furthermore, although the PC2 cleavage site in pro-GIP is highly conserved in mammalian species, the expression of processing enzymes may differ among species resulting in different proportions of the GIP isoforms. Species-dependent expression of peptide isoforms has been described for other gut peptides, such as CCK (23).
The earliest identified biological action of GIP was the suppression of gastric acid secretion (35). In 1981 McIntosh et al. (26) suggested that gastric somatostatin might mediate the acid-inhibitory action of GIP following their observation that GIP1–42 potently stimulated somatostatin release from the perfused rat stomach. Subsequent studies also documented that GIP1–30 can increase the release of somatostatin in this model (31, 36). Here we found that amidated GIP1–30 is equipotent to GIP1–42 in inducing release of gastric somatostatin in mice. Amidated GIP1–30 and GIP1–31 have been shown to bind the GIP receptor equivalently to that of GIP1–42 (12, 18, 22). GIP receptor-mediated cAMP production was similar when pro-GIP was processed either by PC1/3 or PC2 (39). In addition, like GIP1–42, amidated GIP1–30 stimulated proinsulin gene expression and enhanced insulin release from a β-cell line (9). Recently, we have also demonstrated that GIP1–42 and amidated GIP1–30 are equally potent in stimulating glucose-dependent insulin release from the perfused mouse pancreas (10). Collectively, these findings suggest that amidated GIP1–30 has similar biological activity as GIP1–42 via activation of GIP receptors in the mouse pancreas and stomach. Notably, however, a recent study found that GIP1–30 was less potent than GIP1–42 in inducing lipoprotein lipase activity in cultured 3T3-L1 adipocytes (41). Therefore, it remains possible that there are some distinct physiological activities of GIP1–30 and GIP1–42, perhaps mediated by different receptors. Although it is generally believed that mammalian GIP, GLP-1, and glucagon are selective for their respective receptors, studies in fish have demonstrated that glucagon can activate the goldfish GLP-1 receptor (43) and GIP can activate the putative zebrafish glucagon receptor (32). Whether GIP1–30, the highly conserved and more ancestral form of the peptide (13), has more promiscuous receptor activation than GIP1–42 remains to be determined and should be examined in different tissues and different species.
In conclusion, we have determined that there are two distinct isoforms of GIP in the gut. In most K cells PC1/3 converts pro-GIP to GIP1–42, whereas in other K cells PC2 liberates GIP1–31, which is likely further converted by peptidyl-glycine α-amidating monooxygenase to amidated GIP1–30. Previous studies likely missed this subset of K cells as a result of the use of COOH-terminal directed antibodies. It is also probable that circulating GIP levels have in some cases been underestimated, again owing to the use of antibodies that do not detect COOH-terminal truncated forms of GIP. In this regard, we have determined that some currently available commercial GIP assays do not cross-react with GIP1–30. The development of immunoassays capable of distinguishing between GIP isoforms will be required to establish the relative contributions of amidated GIP1–30 and GIP1–42 to GIP biology. Whereas in most respects GIP1–30 appears to function similarly to GIP1–42, the possibility of divergent actions should be explored further.
Y. Fujita was supported by Postdoctoral Fellowships from the Canadian Diabetes Association and Stem Cell Network (SCN). G. K. Yang is a recipient of the Alexander Graham Bell Canada Graduate Scholarship granted by the Natural Sciences and Engineering Research Council of Canada. T. J. Kieffer is a Michael Smith Foundation for Health Research Senior Scholar and gratefully acknowledges grant support from SCN and the Juvenile Diabetes Research Foundation.
The authors declare that there is no duality of interest associated with this manuscript.
Noncommercial antibodies were graciously provided by Dr. Alison Buchan (University of British Columbia), Dr. Linda Morgan (University of Surrey), Dr. Lakshmi Devi (Mount Sinai Medical Center), and Dr. Gunilla Westermark (Uppsala University). We thank Dr. Donald Steiner (University of Chicago) for providing PC2 knockout mice.
- Copyright © 2010 the American Physiological Society