Expression of the Vesicular Inhibitory Amino Acid Transporter in Pancreatic Islet Cells : Distribution of the Transporter Within Rat Islets

A candidate human VIAAT gene was identified by using BLAST (National Center for Biotechnology Information) to search for human genomic sequences homologous to the rat gene (16). Genomic data were submitted to GENSCAN (Massachusetts Institute of Technology, Boston, MA) for gene prediction. VIAAT was amplified from human brain cDNA (OriGene Technologies, Rockville, MD) by PCR using flanking primers derived from the candidate sequences: 5′-CTCGGGTCCTTCTGTCCTT and 3′-AGAAGGGAGAGAGCGCAGA. Internal primers were used for a second round of amplification: 5′-GCCGCCATGGCCACCTTGCTC and 3′-CGGGATCCTTGCGCCCTAGTCCTC. PCR products were analyzed by agarose gel electrophoresis and sequenced using the ABI PRISM BigDye Primer Cycle Sequencing Kit (Applied Biosystems, Foster City, CA). NetPhos 2.0 (17) and the protein secondary structure prediction software PHD (18) were used for prediction of protein phosphorylation sites and transmembrane regions.

For in vitro expression, the human VIAAT (hVIAAT) cDNA was cloned into the plasmid pCRII-TOPO (Invitrogen, Carlsbad, CA) downstream of the plasmid’s Sp6 promoter. GAD65 was expressed as previously described (19). A rabbit reticulocyte lysate system (Promega, Madison, WI) and canine pancreatic microsomal membranes (Promega) were used for in vitro transcription and translation experiments, following the manufacturer’s instructions. The quality of the microsomal membrane preparation was verified by assaying glycosylation and signal peptidase activities using transcripts and a protocol provided by the manufacturer.

An affinity-purified peptide antibody to the COOH-terminal 17-amino acid residues of rat VIAAT (COOH-terminal antibody) and a sample of the immunizing peptide were purchased from Chemicon (Temecula, CA). An antibody generated against the NH₂-terminal 127 residues of mouse VIAAT (NH₂-terminal antibody) was a gift from Dr. Bruno Gasnier (20). Antibody vesicular GABA transporter (VGAT)/1 (21) to rat VIAAT residues 75–87 was from Synaptic Systems (Göttingen, Germany). Antibody 7309 is a peptide antibody against the NH₂-terminus of GAD65 (22). The antibody to the NH₂-terminus of GAD67, 9886, was generated in parallel with the previously described antibody, 11616, using the same peptide and methodology and exhibits the same specificity for GAD67 (22). A polyclonal antibody to glucagon was purchased from Zymed Laboratories (South San Francisco, CA). Monoclonal antibodies to glyceraldehyde-3-phospate dehydrogenase, to the COOH-terminus of GAD65 and GAD67 (GC-3108), and to GABA (GB-69) were obtained from Chemicon, from Affiniti Research (Exeter, U.K.), and from Sigma (St. Louis, MO), respectively.

Pancreata were taken from either healthy diabetes-resistant (DR) BB rats or diabetic diabetes-prone (DP) BB rats and paraformaldehyde-fixed (23). Rats were 150 days old; DP rats became diabetic at 63–70 days and were insulin-treated thereafter. Normal adult human pancreas tissue was provided by the Cooperative Human Tissue Network (Cleveland, OH). Tissue was sectioned (5 μmol/l), deparaffinized, rehydrated, and blocked in PBS with 1% BSA and 2% normal goat serum (Vector Laboratories, Burlingame, CA) for 30 min at room temperature. The serum was heat-treated, which prevents nonspecific complement-mediated antibody binding to islet cells (24). An avidin/biotin blocking kit (Vector Laboratories) was used per the manufacturer’s instructions. The primary antibody was diluted in blocking solution and applied to the sections overnight at 4^οC. Primary antibody dilutions were as follows: COOH-terminal VIAAT, 10 μg/ml; anti-GAD67, 1:200; anti-GAD65, 1:200; NH₂-terminal VIAAT, 1:200; VGAT/1, 1:1,000; anti-glucagon, 1:100; and anti-GABA, 1:100. A 30-min room temperature incubation in a 1:500 dilution of biotinylated goat anti-rabbit IgG followed by a 30-min incubation in a 1:200 dilution of alkaline phosphatase streptavidin and development with Vector Red or NBT/BCIP substrate (all from Vector Laboratories) were used to visualize the binding of the primary antibodies. To help test specificity, before use, the COOH-terminal antibody was sometimes diluted to 10 μg/ml in PBS with 1% BSA and an ∼80-fold molar excess of the immunizing peptide or an irrelevant control peptide. The University of Washington Diabetes and Endocrinology Research Center (DERC) Immunohistochemistry Core performed the immunostaining shown in Fig. 7.

Human and rat brain protein extracts were purchased from Chemicon and Clontech Laboratories (Palo Alto, CA). Human islets (∼70–80% pure) were provided by the Human Islet Transplantation in Seattle program (Seattle, WA). Islets from healthy DR BB rats were provided by the Islet Satellite of the University of Washington DERC Cell and Tissue Core. Human and rodent islets were harvested after Liberase-mediated pancreas digestion (Roche Molecular Biochemicals, Indianapolis, IN). Liberase was injected into the pancreatic duct and used per the manufacturer’s instructions. Islets were then purified in a gradient solution of Optiprep (Nycomed, Olso, Norway) (25). Human islets were sampled and assessed by dithizone staining (26), pelleted, rinsed with cold PBS, and stored at −80°C. Rat islets were placed in Hank’s solution and handpicked before being pelleted, rinsed, and frozen at −80°C. Islet pellets were lysed in Novex sample buffer (Invitrogen) with freshly added protease inhibitor cocktail (P8340; Sigma) and boiled for 4 min. To help verify quality, in preliminary experiments, extracts were immunoblotted for the brain- and islet-specific protein GAD65. The bands formed by the brain extracts comigrated with the islet bands but differed in curvature because of the properties of the buffers in which the extracts were supplied.

Samples (15 μg/lane) were run out under reducing conditions on Novex bis-tris 10% or 4–12% gels (Invitrogen). Proteins were transferred to polyvinylidene fluoride (PVDF) membrane using NuPage transfer buffer (Invitrogen). Using the method of O’Farrell (27), two-dimensional electrophoresis of human islet protein extract (190 μg/gel) and of ³⁵S-labeled hVIAAT protein and subsequent Western transfer was performed by Kendrick Laboratories (Madison, WI). The Columbia University Howard Huges Medical Institute (HHMI) Protein Chemistry Core (New York) performed mass spectrometer (ms/ms) sequencing of protein spots from a gel run in parallel. Membrane blocking and antibody incubations were done in PBS with 5% nonfat dried milk and 0.05% Tween-20. Antibody dilutions were as follows: NH₂-terminal VIAAT antibody, 1:4,000; COOH-terminal VIAAT antibody, 0.5 μg/ml; anti-glyceraldehyde-3-phosphate dehydrogenase, 0.5 μg/ml; and GC-3108, 1:8,000. Peptide competition was carried out with the COOH-terminal VIAAT peptide or an irrelevant peptide (control), as described for immunostaining. Protein bands were visualized with horseradish-peroxidase-coupled secondary antibody (Chemicon) and enhanced chemiluminescent reagents (Amersham Pharmacia Biotech, Piscataway, NJ).

Human chromosome 20 contains sequences homologous to rodent VIAAT. Three possible exons are situated in the region encoding the putative human VIAAT gene: exon 1 is 390 bp, exon 2 is 957 bp downstream from exon 1 and 231 bp in length, and exon 3, 1,888 bp, is 1,151 bp downstream of exon 2. Exons 1 and 3 are highly homologous to rodent VIAAT. Primers were designed using the genomic data. Amplification of cDNA from human brain yielded a single product consistent with the length of a transcript containing exons 1 and 3; there was no evidence of alternative splicing. The sequence of the PCR product confirmed that human VIAAT is encoded by two exons on chromosome 20. The cDNA sequence was submitted to GenBank (accession AY044836).

hVIAAT, like rat, is a protein of 525 amino acids (Fig. 1). Rat and hVIAAT differ by seven amino acids, with six of the substitutions occurring NH₂-terminal to residue 115. In contrast, the mouse and rat proteins differ by only one residue before amino acid 514. Past residue 514, the mouse sequence lacks homology to the COOH-terminal 11 residues of the human and rat proteins. Computer analysis predicts that hVIAAT, like rodent VIAAT, has 10 transmembrane domains and that, within the cytoplasmic domains, there are nine conserved possible serine phosphorylation sites, including five within the COOH- and NH₂-terminal domains (residues 15, 17, 20, 25, and 511; Fig. 1).

Immunoblot analysis was used to test for the presence of VIAAT in rat and human islets. The COOH-termini of rat and human VIAAT are identical, so we probed blots of islet and brain tissue extract with an antibody to this region (Fig. 2). As expected, the ∼57-kDa VIAAT band was detected in the lane with rat brain extract. Also, a faster-migrating ∼52-kDa VIAAT band described in two earlier reports was observed (20,21). Both VIAAT bands were also detected in rat islet extract. In contrast to rat brain, in islet, the ∼52-kDa form was the most abundant. In human brain extract, only the ∼52-kDa band was detected. Consistent with past results (12), VIAAT was not detected in other tissues, including kidney (not shown), liver, and ovary. Competition with the immunizing VIAAT COOH-terminal peptide prevented detection of the bands, indicating that binding was specific (Fig. 2). In contrast, an ∼67-kDa human islet band detected by the COOH-terminal antibody was not efficiently competed. VIAAT was not detected in any of the five human islet extracts tested (two shown).

Transcription and translation of the hVIAAT cDNA in a rabbit reticulocyte lysate system resulted in the synthesis of both the 57- and 52-kDa forms of the protein (Fig. 3). Addition of microsomal membranes to the reticulocyte lysate enabled membrane-dependent post-translational processing such as signal peptide cleavage (data not shown) and core glycosylation. Without the addition of exogenous membranes, these modifications did not occur (e.g., compare lanes 3 and 4 in Fig. 3A). Both VIAAT forms, however, were synthesized with or without the addition of membranes to the reaction, indicating that formation of a second VIAAT band was not a result of processing by membrane-associated proteins. Also, both bands persisted after treatment of the expressed hVIAAT with calf intestinal phosphatase, as described previously (28) (not shown). Both the 57- and 52-kDa proteins were precipitated by the COOH-terminal antibody and, with greater efficiency, by an antibody (20) raised against the NH₂-terminal 127 residues of VIAAT (the NH₂-terminal antibody; Figs. 3B and data not shown).

Fig. 4A shows that both the 57- and 52-kDa forms of VIAAT are also detected by the NH₂-terminal antibody. The two hVIAAT bands comigrate with the corresponding rat brain bands. A third higher-molecular weight brain band was also observed that likely represents the phosphorylated form of the protein previously detected with the NH₂-terminal, but not the COOH-terminal, antibody (14,28). The 52- and 57-kDa VIAAT bands were also detected in rat islets, confirming that rat islets contain VIAAT (Fig. 4B). As observed with the COOH-terminal antibody, brain and islet differ in that, in islet, the 52-kDa form is more abundant. Also, the higher-molecular weight, possibly phosphorylated form is absent in rat islets (Fig. 4B).

The NH₂-terminal antibody did not detect VIAAT in human islets by immunoblotting (Fig. 4B) or by immunohistochemical staining (not shown). Long exposures of immunoblots loaded with twofold more human, but not rat, islet protein resulted in detection of a weak ∼55-kDa band by the NH₂-terminal but not the COOH-terminal antibody. Two-dimensional gel electrophoresis, immunoblotting, and subsequent microsequencing showed that this 255 kDa band did not comigrate with in vitro synthesized hVIAAT in either dimension (not shown) and that it represented keratin 8, a component of pancreatic acinar, but not islet, tissue (29). VIAAT was not detected. In contrast, the islet-specific protein GAD65 was detected in all human islet extracts tested.

We used the COOH- and NH₂-terminal antibodies to determine the distribution of VIAAT in rat islets. Immunostaining revealed that VIAAT is most abundant in the mantle (peripheral) region, where glucagon-secreting α-cells and other non-β-cells are located (Fig. 5). Immunostaining with a third antibody, VGAT/1 (21), confirmed this result (not shown). VIAAT is also present in the central β-cell region. The distribution of VIAAT that we observed paralleled that of GAD67. In agreement with a previous study, GAD67 was present in β-cells but was more abundant in the mantle region, whereas GAD65 was β-cell specific (22).

To confirm the presence of both VIAAT and GAD67 in non-β-islet cells, we immunostained islets from diabetic BB rats (23). DP BB rats develop autoimmune diabetes (30). After the onset of glycosuria, they rapidly lose all of their β-cells, resulting in shrunken islets comprised mostly of α-cells with relatively reduced numbers of δ- and PP-cells (31,32). Accordingly, the β-cell-depleted islets studied here stained positive for glucagon but not for GAD65 (Fig. 6) or insulin (Fig. 7). Consistent with expression of both GAD67 and VIAAT in rat islet mantle (non-β) cells, the diabetic islets stained positive for both proteins. The presence of GAD67 and VIAAT in the rat islet mantle predicted that GABA should be found there too. In agreement with a prior report indicating that GABA is present in the rat islet mantle, although at lower levels than in the β-cells (33), diabetic islets stained positive for GABA (Fig. 7).

The existence of an islet cell vesicular GABA transporter is suggested by the presence of GABA within the SLMV, where it is probably stored before exocytosis despite the cytoplasmic localization of GAD (4). We have found that VIAAT, a transmembrane GABA transporter associated with synaptic vesicles in the brain, is expressed in rat islets.

We determined that human VIAAT is 525 amino acid residues in length and shares nearly 99% sequence identity with the rat protein. There was no evidence of splice variants by PCR, despite suggestive genomic sequence data. VIAAT consists of 10 transmembrane domains with a large NH₂-terminal cytoplasmic domain (∼125 residues) and a smaller COOH-terminal cytoplasmic domain (∼14 residues) (12,13). Six of the seven residues that differ between rat and human VIAAT lie in the NH₂-terminal cytoplasmic domain. The COOH-terminal domain of the mouse protein diverges from that of the human and rat proteins. The central transmembrane region then is the most highly conserved. In vitro translation of hVIAAT yields two protein bands that comigrate with the two rat VIAAT forms that we observed in rat islet and brain. Surprisingly, we were unable to detect VIAAT in human islets.

Our results indicate that there are at least two forms of the VIAAT protein; these migrate at ∼57 and ∼52 kDa. Both forms were detected by both an antibody to the COOH-terminus of VIAAT and an antibody raised against the NH₂-terminal 127 residues, and both were formed after in vitro synthesis of VIAAT. The NH₂-terminal antibody also detects an ∼59-kDa band in rat brain that was previously shown, using the same antibody, to be due to serine/threonine phosphorylation of VIAAT (28). The failure of the COOH-terminal antibody to detect this band suggests that one site of phosphorylation is the COOH-terminus. Residue 511, a serine, is a predicted phosphorylation site and is part of the peptide sequence against which the COOH-terminal antibody was raised; phosphorylation here may prevent antibody recognition. The ∼59-kDa band was not detectable in rat islets.

The pattern of expression in rat islets and brain also differed in that the 57-kDa band was more prevalent in the latter, whereas, in islets, the 52-kDa form was more abundant. The lower abundance in neural tissue may explain why the 52-kDa form is described in only one (21) of several published reports characterizing VIAAT in rodent brain. Transfected COS-7 cells were also found to synthesize the lower-molecular weight molecule (20). It is unclear how the two different forms of the protein differ. In vitro, both forms were synthesized regardless of whether microsomal membranes were added to the reaction. Thus, they are not formed by a post-translational modification, such as core glycosylation, mediated by microsome-bound enzymes. It is also unlikely that the 52-kDa form results from proteolytic cleavage of the larger form because the smaller protein was consistently present in cell extracts prepared in the presence of protease inhibitors, extended incubations of cell extracts or in vitro translation reactions did not result in conversion of the longer form to the smaller, and there was no evidence of similar proteolysis in other proteins studied in parallel. One possibility is that the smaller form results from translation initiation at an alternate start site, such as perhaps the AUG codon encoding residue 31, which, in rat and human, is part (positions +1 to +3) of a stretch of 10 residues (−6 to +4) containing eight matches to the Kozak consensus sequence, including matches at the important −3 (A) and +4 (G) positions (34). Examples of alternate translational start site utilization in mammalian cells have been described (35–37).

The distribution of VIAAT in rat islets parallels that of GAD67. Both were most abundant in the islet mantle, where glucagon-producing α-cells and other non-β-cells are localized. VIAAT was also present in the β-cells but in a lower abundance. Immunostaining of diabetic rat tissue confirmed that both VIAAT and GAD67 were present in non-β-cells. Previous work has demonstrated that, in endocrine cells, as in neurons, VIAAT localizes to microvesicles: in transfected PC-12 cells, VIAAT associated with the endogenous SLMVs (12).

The distribution of GAD67 in rat islets observed in this study matches that described in an earlier report (22). Others, however, have reported that GAD67 is only expressed in rat islet β-cells (38,39). These latter studies relied on in situ nucleic acid hybridization, which did not directly assess GAD protein expression and carried the risk of cross-hybridization with the homologous GAD65 message, on immunoblot analysis using an antibody that recognizes both GAD isoforms and on immunohistochemistry using the same antibody. Rat islets contain significantly more GAD65 than GAD67 (40). As a result, GAD67 may be difficult to detect and distinguish from GAD65 in experiments performed with an antibody that recognizes both isoforms. Another report described GAD67 expression in glucagon-containing α-cells in rat islet monolayers, consistent with the results presented here (39). GAD65 expression was β-cell specific, which is also in agreement with our data. Besides methodological differences, the conflicting results regarding rat islet GAD67 distribution could be due to variations in GAD expression in different rat strains. This study used BB rat pancreata; other studies used Wistar rats (38,39). Similarly, there have been conflicting reports as to the presence of GABA in rat islet mantle cells. Whereas some studies conclude that GABA is absent (41,42), other reports demonstrate that the islet mantle and also α-cells in rat islet cell monolayers contain GABA, although at lower levels than β-cells (33,39). Here, using islets depleted of β-cells, we confirm that rat islet non-β-cells are capable of synthesizing GABA. This result is consistent with our finding that these cells contain GAD67. Whether insulitis and diabetes altered cellular levels of GABA is unknown.

We were able to compete away COOH-terminal antibody-mediated immunostaining with the immunizing peptide. Also, control stainings with normal rabbit serum and other unrelated antibodies did not result in positive islet staining. It is therefore highly unlikely that our immunostaining results are attributable to nonspecific binding. Also, β-cell-depleted diabetic islets were, as expected, not bound by GAD65 or insulin antisera—further evidence that the islet staining shown herein was specific.

This study is the first to examine VIAAT expression in cells that express only GAD65 or GAD67. In brain tissue, both isoforms are uniformly expressed, with GAD65 constituting from 61 to 89% of GAD content (43). Rat islet β-cells also contain both GAD isoforms. In contrast, human islets express only GAD65, and rat islet mantle cells, as confirmed here, contain just GAD67 (4,22). Whereas GAD65 is mostly bound to the surface of the SLMV, GAD67 is primarily nonmembrane associated, so its role in producing GABA for the SLMV is less clear. Limited trafficking of GAD67 to microvesicles, however, suggests that GAD67 may also produce GABA for SLMV uptake and secretion (8). The expression of VIAAT by rat islet mantle cells is evidence of such a role for GAD67.

The parallel distributions of VIAAT and GAD67 in rat islets is striking, especially since GAD65 is the most abundant isoform in the rat islet (40). This pattern of distribution could be due to the relative degrees of association with the SLMV of GAD65 and GAD67. Because GAD65 is mostly bound to the SLMV—and is perhaps in direct contact with VIAAT—GAD65 is probably more efficient in delivering GABA to the microvesicle than GAD67. Perhaps, then, to compensate for this and allow for adequate GABA uptake, VIAAT content in the SLMV membrane is increased in cells that use GAD67 to produce some or all of their GABA stores. Because VIAAT was not detected in human islets, cells that express solely GAD65 may require little VIAAT. Alternatively, a different protein—perhaps the same one that binds GAD65 to the surface of the SLMV—mediates the vesicular uptake of GABA produced by GAD65, and cells with only GAD65 contain no VIAAT.

VIAAT is also a transporter for glycine, which is, like GABA, an inhibitory neurotransmitter (13,14). Further work is necessary to determine whether the SLMVs of islet cells store and secrete glycine.

In summary, we have shown that VIAAT is expressed in rat islets and thus likely mediates the proton gradient-driven vesicular GABA transport activity previously described in rodent islets (10). VIAAT could not be detected in human islet β-cells. Our data confirm a previous report that rat islet mantle cells express GAD67 and do so at greater levels than β-cells (22). We have also confirmed that rat islet non-β-cells can produce GABA. VIAAT expression in the rat islet, which is greatest in the islet periphery, parallels the distribution of GAD67 expression. We hypothesize that this is because the localization of GAD65 to the SLMV membrane allows efficient GABA transport that requires little or no VIAAT, whereas cells that contain GAD67 require relatively increased levels of VIAAT expression to facilitate GABA entry into the SLMV. The GAD-GABA-VIAAT signaling machinery in islets exhibits great heterogeneity, ranging from human β-cells, which express high levels of GAD65, no GAD67, and undetectable levels of VIAAT, to rat mantle cells, which contain only GAD67, relatively low levels of GABA, and increased levels of VIAAT. Further work will be necessary to characterize how the 57- and 52-kDa forms of VIAAT differ in structure and function to determine why there are relatively increased levels of the smaller form in rat islets and to determine whether they interact differently with GAD65 and GAD67.

This research was funded by grants to S.D.C. from the Juvenile Diabetes Research Foundation International (JDRF) and the National Institutes of Health (K08-DK02944) and by grant DK26190 to Åke Lernmark. The Human Islet Transplantation in Seattle islet distribution program was funded by JDRF.

We thank Angela Wallen for technical assistance with islet isolation. Dr. Bruno Gasnier graciously provided an antibody to VIAAT. Rodent islet isolations were performed by the Islet Satellite of the University of Washington DERC Cell and Tissue Core (DK17047). The DERC Cytohistochemistry Core performed some of the immunohistochemical staining shown herein; we thank Joyce Murphy for her help in this regard. The DERC Molecular Genetics Core assisted with cDNA cloning and sequencing, for which we thank Dr. Libby Rutledge, Brian Van Yserloo, and Paul Gohlke. We thank Åke Lernmark for his helpful comments regarding this manuscript, and we also thank Bunny Williams for her kind support.