Alternative splicing of GAD67 results in the synthesis of a third form of glutamic-acid decarboxylase in human islets and other non-neural tissues.

Two forms of glutamic-acid decarboxylase (GAD) have been identified in mammalian tissues: a 65-kDa form (GAD65) and a 67-kDa form (GAD67). Alternate splicing produces one or two smaller variants of GAD67 in the brain of embryonic mice and rats. Additionally, a short, heretofore unidentified transcript homologous to GAD67 has been detected in human testis RNA. Because GAD, the enzyme responsible for gamma-aminobutyric acid production and a key autoantigen in type I diabetes, has unclear function in non-neural tissue, it is important to understand its pattern of expression. Unlike GAD65, GAD67 is not produced in human pancreatic islets. Here, we describe a novel splice variant of GAD67 that is produced in human islets, testis, adrenal cortex, and perhaps other endocrine tissues, but not in brain. This transcript directs the synthesis of a protein without GAD enzymatic activity: GAD25. A unique peptide sequence at the carboxyl terminus of GAD25 is highly conserved between mice, rats, and humans. We conclude that humans produce a third form of GAD in non-neural tissues and that human islets, although they do not synthesize full-length GAD67, do express this shortened variant.

some 2, is ϳ67 kDa (GAD67). Both forms of the enzyme catalyze the formation of ␥-aminobutyric acid from glutamate, although GAD67 has a markedly higher affinity for the cofactor pyridoxal 5Ј-phosphate, which is necessary for the activity of both forms (2,3).
GAD expression is greatest in two tissues: brain and the pancreatic islets of Langerhans. Regardless of species, GAD65 and GAD67 are both abundant in brain, where ␥-aminobutyric acid is a major inhibitory neurotransmitter (2,4). In islets, the relative abundance of the two forms differs between species. In human islets, GAD65 is abundant, but immunocytochemistry, in situ hybridization, immunoprecipitation, and Western blotting have not detected GAD67. GAD67 message is not detectable in human islets by Northern blotting, although RT-PCR and RNase protection experiments suggest that that there may be a low level of transcription (4 -7). Monkey islets, like human islets, produce only GAD65 (8). In contrast, rat islets synthesize both isoforms, and mouse islets, which produce less GAD overall, produce predominantly GAD67 (5,6,9,10).
Alternative splicing of GAD67 has been described in embryonic rodent brains. In mice and rats, an exon, which is itself alternatively spliced to either 80 or 86 base pairs (bp), is inserted into the full-length GAD67 message upstream of the pyridoxal 5Ј-phosphate-binding site in embryonic and fetal animals, but not in adult animals (11)(12)(13). The embryonic exon harbors an in-frame stop codon, resulting in the synthesis of a 25-kDa variant of GAD67, GAD25. If the exon is spliced to its 80-bp rather than its 86-bp form, another stop codon at the 3Ј-end of the exon is removed, potentially enabling translation of a 44-kDa variant from a start codon within the embryonic exon (12).
GAD is also synthesized in testis. Here, although the 3.7-kb GAD67 message is detectable, the most abundant GAD transcript in humans is shorter, previously estimated to be ϳ2.5 kb (14). This small GAD transcript has not heretofore been identified, and it has not been certain whether it represents the product of a third GAD gene or a splice variant of one of the other two.
Here, we report the synthesis of a novel GAD transcript (a splice variant of GAD67) in human islet cells and testis. This transcript, which is also present in human adrenal cortex, is likely the previously unidentified short testis transcript. We show that the encoded protein, which lacks GAD enzymatic activity, is present in human islets and testis. This is the first report of a third variant of GAD in any human or other mammalian (non-embryonic) tissue and of the expression by human islets of a form of GAD other than GAD65.  1 The abbreviations used are: GAD, glutamic-acid decarboxylase; RT-PCR, reverse transcription-polymerase chain reaction; bp, base pair(s); kb, kilobase pair(s); RACE, rapid amplification of cDNA ends; EST, expressed sequence tag.

Tissues-Human
Stevens (Department of Surgery, University of Washington Medical Center and the Human Islet Isolation and Cell Processing Facility, Puget Sound Blood Center/Northwest Tissue Center, Seattle, WA). Monkey islets (from Macaca nemestrina) and additional human islets were acquired from the University of Washington Diabetes and Endocrinology Research Center Islet Core. RNA was extracted using Trizol reagent (Life Technologies, Inc.), and poly(A) RNA was isolated using oligo(dT)-cellulose (Life Technologies, Inc.) following the manufacturer's instructions. RNA extracted from other human tissues was purchased from CLONTECH (Palo Alto, California), as was a detergent extract of healthy human testis tissue.
Cloning and Amplification of cDNA-3Ј-RACE was performed as described previously (19). To create a primer for reverse transcription, a (dT) 17 tail was added to the 3Ј-end of the 3Ј-amplification primer: GGCCACGCGTCGACTAGTAC. Two nested 5Ј-primers specific for GAD67 upstream of the putative splice site were used sequentially for amplification: first CTCCTGGAAGTGGTGGACAT and then AGA-CATTTGATCGCTCCACC. PCR products were sequenced using the ABI PRISM BigDye Terminator Cycle Sequencing kit (PE Biosystems, Foster City, CA) following the manufacturer's instructions. PCR products intended for use in sequencing were amplified using a high fidelity polymerase (Life Technologies, Inc.). For qualitative PCR to detect the tissue-specific expression of transcripts and also for cloning and PCR radioactive probe generation, the following 3Ј-primers were used along with the second GAD67 5Ј-primer listed above: CAGCCCCAGCTT-TCTTTATG (GAD67) and TGGAAACCATGTGTGCAGTT (GAD67S; a HindIII linker was added to the 5Ј-end of this primer for cloning: GCGCTAAGCT). The previously described plasmid construct pEx12 (20) served as the template for GAD67 amplification and synthesis (see below). To clone GAD67S, the region of the pEx12 GAD67 insert 3Ј to the single AvaI site was removed by digestion with AvaI and HindIII and then replaced with the AvaI-HindIII fragment of the GAD67S PCR product.
Sequence homology searches in the GenBank TM Data Bank were performed using the most recent version of BLAST software at the National Center for Biotechnology Information (21). ClustalW and BLAST were used for sequence alignment (22).
Northern Blot Analysis-Northern blots were prepared using standard techniques (25) and also purchased from CLONTECH. Poly(A)selected RNA was used for all blots except those of islet RNA. Randomprimed probes for GAD65 and GAD67 were generated from cDNAs encompassing the entire coding regions. These cDNAs were excised from the plasmid constructs pEx9 and pEx12 (20), respectively, and gel-purified. The probe for GAD67S was generated by PCR using the primers described above. Hybridization was with ExpressHyb (CLON-TECH) following the manufacturer's instructions, except that denatured fish sperm DNA (100 g/ml) was added to reduce the background. Washes were performed under increasingly stringent conditions; final washing was done at 55-65°C in 0.1ϫ SSC and 0.1% SDS.
Immunoblotting-Tissues were extracted in SDS buffer with ␤-mercaptoethanol and analyzed by Western blotting using standard methodologies (25). Tween 20 (0.04%) together with powdered milk (4%, w/v) were used as blocking reagents. Two different antisera raised against a peptide composed of the 18 amino acids beginning with the second residue (Ala) of GAD67 were utilized: antisera 9886 and 10266 (9). Both detected GAD25, although the latter was used preferentially as it produced a lower background. The secondary antibody was horseradish peroxidase-conjugated goat anti-rabbit polyclonal antibody preabsorbed with human and mouse IgG (Santa Cruz Biotechnology, Santa Cruz, CA). Detection was by chemiluminescence (25). In some experiments, the primary antibody was preincubated for 30 min at 4°C in blotting buffer with 20 ng/ml (final concentration) GAD67 (residues 2-19) or GAD65 (residues 4 -21) amino-terminal peptide.

Detection of GAD Variants in Pancreatic Islets-As
Northern blot analysis of islet RNA with a probe for GAD65 had previously revealed a 2.5-kb band (the putative transcript was referred to as "GAD3") in addition to the expected 5.6-kb band (26,27), we tested the hypothesis that a variant of GAD65 is synthesized in human islets. We initially probed Northern blots of monkey and human islet and pancreas RNAs with random-primed DNA probes and riboprobes for GAD65; washes were performed at various stringencies. Although detectable in some blots, the 2.5-kb band was not consistently reproducible, even under conditions in which there was crosshybridization with GAD67 (data not shown).
As there is some evidence of autoreactivity to GAD67 in type 1 diabetes mellitus (20, 28 -30), we speculated that a GAD67like protein may be synthesized in human pancreatic islets. Because of this, we next utilized a probe specific for GAD67 for Northern blot analysis. The expected 3.7-kb band was seen in brain, but two unexpected, shorter bands (one ϳ1.5 kb and the other ϳ1 kb) were reproducibly observed in pancreatic RNA (Fig. 1).
Identification and Sequencing of a GAD67 Splice Variant (GAD67S)-These bands, representing shorter transcripts, pointed toward the existence of splice variants of GAD67 or possibly a third GAD gene. To identify possible splice variants or genes with high homology to GAD67, we searched EST sequences deposited in the GenBank TM Data Bank. Five human ESTs were identified that were homologous to GAD67 upstream of the codon for amino acid 213 (Met), but that diverged 3Ј of this site. Two of the ESTs were from a parathyroid tumor library (IMAGE clones 1405787 and 1341987), and the others were from testis, colon tumor, and breast tumor libraries (IMAGE clones 1644588, 1148313, and 1071440, respectively). Comparison of these ESTs to sequences in Gen-Bank TM suggested that they derived from a single, novel transcript, most likely a splice variant of GAD67.
Primers were designed specific for this putative splice variant based on a consensus sequence derived from the ESTs. RT-PCR using these primers revealed that this transcript, which will be referred to as GAD67S, is produced in human testis and islets (Fig. 2). We were also able to amplify the expected GAD67S PCR product from monkey testis RNA, but not from RNA prepared from human breast tissue, human breast carcinoma tissue, or monkey brain ( Fig. 2 and data not  shown). The lengths of the RT-PCR products resulting from amplification using two different 5Ј-primers specific for GAD67 upstream of the putative splice site and a 3Ј-primer specific for GAD67S were consistent with our hypothesis that GAD67S represents a splice variant of GAD67 rather than a novel gene.
We sequenced GAD67S through the splice site to the poly(A) tail using 3Ј-RACE. The cDNA sequence, which was identical whether we sequenced 3Ј-RACE products derived from islets or testis, has been deposited in the GenBank TM Data Bank (accession number AF178853). The site at which GAD67S diverges from GAD67 is the human homologue of the 5Ј-splice site utilized for insertion of the rodent brain embryonic exon (11)(12)(13). As shown in Fig. 3, homology between GAD67S and the rodent embryonic transcript continues downstream to the 3Ј-splice site of the embryonic exon (GenBank TM accession numbers M38351 and Z49977). Here, GAD67S becomes homologous to genomic mouse DNA that was previously sequenced 3Ј of the embryonic exon (GenBank TM accession number Z49977). In contrast to the mouse genomic sequence, however, the transcribed human sequence encodes a polyadenylation signal (Fig.  3). The predicted protein (Fig. 3) is the human form of GAD25 (12). Of note, the carboxyl-terminal 11 amino acids (where GAD25 diverges from GAD67) are identical in humans, mice, and rats. A search of the data bases at the National Center for Biotechnology Information for other homologous DNA or protein sequences downstream of the alternative splice site yielded no matches, suggesting that that this 11-residue peptide sequence is unique to GAD25.
GAD25 lacks the binding site for the cofactor pyridoxal 5Јphosphate, suggesting a lack of glutamate decarboxylase activity by the protein. Consistent with this, in an in vitro assay for GAD activity, GAD25 failed to catalyze the release of 14 CO 2 from [1-14 C]glutamate (Fig. 4).
Tissue Distribution of GAD67S Expression-We utilized Northern blot analysis to ascertain which human tissues produce GAD67S. Consistent with our RT-PCR results (Fig. 2), the transcript (determined to be ϳ1.5 kb) was detected in testis (Fig. 5A). Brain, the organ in which GAD67 is most abundant, did not synthesize GAD67S ( Fig. 5A; see also Fig. 1), although the probe did detect GAD67 (4). Adrenal cortex, a site of low level GAD67 transcription, produced GAD67S in greater abundance than testis (Fig. 5A) (4). Northern blot analysis of human pancreatic islet RNA (Fig. 5B) confirmed that the message was transcribed in these cells. Based on immunoblotting results (see below), we did not expect to find the message in monkey islets. However, it was detectable, although at levels ϳ4-fold less than in human islets. The message was not detected in the other tissues tested.
The Protein Product of GAD67S, GAD25, Is Present in Human Islets and Testis-Western blot analysis was employed to test for the presence of the GAD67S protein product in different tissues. To detect the protein, we used antisera raised against a synthetic peptide consisting of the amino-terminal 18 amino acid residues of GAD67/GAD25 (excepting the initiating me-thionine) (9). We detected GAD25 in human islet and testis extracts, but not in rat brain or monkey or rat islets (Fig. 6). GAD67 was present in rat brain extract, but consistent with previously published results (5, 7), not in human islets (Fig.  6C). As the primary and secondary antibodies were both polyclonal, immunoblotting produced a significant number of nonspecific bands. To ensure that the ϳ25and ϳ67-kDa bands represented GAD25 and GAD67, we demonstrated that we could specifically block antibody binding and detection of these bands by blocking the primary antibody with the GAD67/ GAD25 amino-terminal peptide. A control peptide with amino- FIG. 2. RT-PCR confirms that GAD67S is transcribed in human tissues. We utilized primers specific for GAD67 or GAD67S (GAD67S primers were based on a consensus of the EST sequences) for RT-PCR. The source of the starting RNA sample is indicated above each lane (H, human; M, monkey; control, PCR with no reverse transcription product added). The bands on the right are from a 100-bp size ladder; the brightest band is at 500 bp. The sizes of the amplified fragments were consistent with our sequence data (Fig. 3). That GAD67 message can be amplified by PCR from human islet RNA has been previously reported (7). RT-PCR also revealed GAD67S in monkey testis, but not in monkey brain or human breast, although GAD67 message was detected in all three tissues (data not shown) .   FIG. 3. Sequence of GAD67S. A, the rat and mouse embryonic GAD67 sequences (11,12) are shown in the first two lines, with the embryonic exon (boldface) and the 5Ј-splice site (2) indicated. The site of the overlapping embryonic stop/start codons is indicated (*****), as is the stop codon found only in the longer variant of the embryonic exon (###). 3Ј of the embryonic exon, the rodent GAD67 transcript continues unaltered from the adult form (3). The third line is human GAD67S, which is homologous to the rodent embryonic sequences up to the 3Ј-splice site of the embryonic exon. The fourth line is mouse genomic sequence from the beginning of the embryonic exon past the 3Ј-splice site. The GAD67S poly(A) addition signal is in boldface and underlined; the aligned mouse genomic sequence lacks this signal sequence. Sequences were aligned with the program ClustalW (22). B, shown is the sequence of human GAD25 amino acids 200 -224. The underlined residues, encoded by the alternatively spliced exon, are unique to GAD25 and are identical in mouse, rat, and human.
FIG. 4. GAD25 lacks glutamate decarboxylase activity. GAD67 and GAD25 were synthesized using in vitro transcription and translation in the presence of [ 35 S]methionine. Visualization of the translated proteins by SDS-polyacrylamide gel electrophoresis followed by autoradiography (not shown) confirmed that they were the expected sizes. Equimolar amounts of translation product (as assessed using trichloroacetic acid precipitation to measure incorporated radiolabel and then adjusting for methionine content) were compared in the activity assay, with all samples being added to the assay in equal volumes of the reticulocyte lysate reaction mixture. As a control, an equimolar amount of in vitro translated luciferase was also assayed, as was a coupled transcription/translation reaction to which no DNA template was added (None). GAD activity is represented as cpm 14 CO 2 released from L-[1-14 C]glutamate. Data are representative of results from one experiment; using more of the translation product increased only background counts (from the reticulocyte lysate alone) without increasing GAD25 activity above this background (not shown).
terminal GAD65 sequence did not block detection of the two bands (Fig. 6, B and C). DISCUSSION Unlike rodent islets, human islets are presumed to synthesize only GAD65, not GAD67 (2, 3). Neither GAD65 nor GAD67 variants have been described in the islets of either species. Here, we have shown that a short form of GAD, encoded by the novel transcript GAD67S, is expressed in human islets. GAD67S is a splice variant of GAD67. The encoded protein, GAD25, has been detected previously only in embryonic and fetal mouse brain, though it is likely also synthesized in rat embryonic brain (11,12).
Knowledge of the pattern of GAD gene expression in human islets is essential, as GAD is a key and possibly the triggering autoantigen in type 1 diabetes mellitus, a disease resulting from autoimmune destruction of the insulin-producing beta cells within the islets (2,3,15,16). Although autoantibodies and T cell reactivity to GAD67 have been detected in patients with type 1 diabetes, this evidence of autoimmunity is generally attributed to cross-reactivity by autoantibodies and T cells reactive to GAD65. The reasons that GAD65 and not GAD67 is thought to function as an autoantigen in type I diabetes are 3-fold: first, because GAD65 autoantibodies are much more prevalent in patients with new-onset diabetes (70 -90% versus ϳ10%); second, because GAD67 autoreactivity most often occurs in patients who also exhibit autoreactivity to GAD65; and, finally, because GAD67, unlike GAD65, is thought not to be synthesized in human islet cells (2,3,7,17,18). Our results show, however, that although these cells do indeed lack GAD67, human islets produce a truncated variant: GAD25.
A common feature of the major antigens targeted by autoantibodies in patients with type 1 diabetes is their direct association with the beta cell secretory apparatus (3). It is thus interesting that the GAD67 amino-terminal sequences that may mediate association with GAD65 (and thus with the membrane of the islet synaptic-like microvesicle) are preserved in GAD25 (2,3). In light of evidence from the non-obese diabetic mouse model of autoimmune diabetes that autoreactivity to GAD67 may help propagate or initiate islet cell destruction, it will be important to determine whether GAD25 plays a role in the pathogenesis of the disease in humans (2,3,15,16). Presently, determination of whether there is T cell reactivity to unique GAD25 epitopes is hindered by the fact that reliable T cell assays for human autoreactive T cells have yet to be developed (31). Also, although we have detected a low prevalence of GAD25-specific humoral autoimmunity in patients newly diagnosed with type I diabetes, 2 determination of whether autoantibodies targeted to GAD25 are present early in the disease process, around the time of onset of islet cell autoimmunity, or are involved in the pathogenesis of other autoimmune glandular diseases will require further study.
We have likely identified the heretofore uncharacterized short GAD67 testis transcript. Prior Northern blot analysis of human testis RNA revealed the expected 3.7-kb transcript, but also a more abundant, shorter message. Since the probe employed in these earlier studies included the entire coding region of GAD67, it would have hybridized to GAD67S (14,32). The shorter GAD message was estimated to be ϳ2.5 kb (ϳ1 kb longer than GAD67S), but the basis of that estimate is unclear. If there is a third, 2.5-kb GAD67-like transcript in testis, it is unclear why all three messages were not detected. At the time of writing, there was no evidence in the GenBank TM Data Bank of other human GAD67 splice variants, although one possible FIG. 6. GAD25 protein is present in human islet and testis extracts. Western blot analysis was performed using extracts from the tissues indicated above the lanes (R, rat; H, human; M, monkey). The primary antibody was specific for a peptide derived from the amino terminus (N-term.) of GAD67/GAD25 and detected both GAD25 (arrows; A and B) and GAD67 (C). In A, the primary antibody was preincubated with (؉) or without (Ϫ) the amino-terminal peptide. In B, islets from other animals were similarly tested, although as a control, a GAD65 peptide was employed (like the GAD67 peptide, from the amino terminus of the protein). *, expected position of the GAD25 band. In C, the antibody detected GAD67 in rat brain, but not in human islets, and was displaced by the GAD67 peptide (67), but not by the GAD65 peptide (65). Displacement by the GAD67 competitor peptide (and lack of displacement by the GAD65 peptide) confirms specificity of binding to the GAD67/GAD25 amino-terminal sequence. Note that the background bands were unaffected by the competitor.
FIG. 5. GAD67S is transcribed in human testis, adrenal cortex, and islets, but not in brain. A, a commercially produced multitissue Northern blot made using human poly(A)-selected RNA was hybridized with a probe specific for GAD67S (and separately, one specific for actin). The 1.5-kb GAD67S band was apparent in the adrenal cortex and testis lanes, but not in the other lanes shown, including brain (which was on a separate membrane that was hybridized and washed in tandem). The 5Ј-portion of the probe contained sequence upstream of the splice site, resulting in hybridization to GAD67 message (brain lane). explanation for the ϳ1-kb pancreas band in the Northern blot in Fig. 1 is the existence of another, yet shorter GAD67-related transcript. Although the existence of an islet variant of GAD65 was previously proposed on the basis of Northern blot results, we were unable to consistently reproduce the relevant band (26,27). To the best of our knowledge, only the turtle produces a proven GAD65 splice variant (1).
The function of GAD25 is unclear. It lacks the pyridoxal 5Ј-phosphate-binding site and is enzymatically inactive as a glutamate decarboxylase. Although the unique, carboxyl-terminal 11 amino acids are perfectly conserved between humans, rats, and mice, it is unclear what functionality this short peptide sequence might confer upon the protein; we could find no evidence of homology to other proteins. One may speculate that alternative splicing of the GAD67 message is possibly a means to down-regulate expression of the enzyme, but such alternate splicing, resulting in the synthesis of a 25-kDa protein, would be a surprisingly inefficient way to decrease GAD67 synthesis, and it would not explain the addition of a conserved stretch of amino acids to the truncated protein.
In addition to islets and testis, GAD67S is produced in adrenal cortex. Low level transcription of GAD67 (but not GAD65) has previously been detected in this organ (4). Two of the GAD67S ESTs we found were derived from a parathyroid library, suggesting that GAD67S is produced in that tissue as well. In contrast, the transcript was not detected in brain. It is interesting that synthesis of this transcript may be confined to endocrine organs. An increased incidence of autoreactivity to GAD67 has been noted in association with autoimmune polyglandular syndrome type II, which commonly involves the adrenal cortex, islets (diabetes mellitus), and gonads (18). GAD25 is the only GAD67-like molecule expressed by all three of these tissues.
A key finding that has contributed to our current understanding of GAD function in general and the role of autoreactivity to GAD in diabetes mellitus has been the absence of GAD67 in human islets. In addition, only two forms of GAD protein have heretofore been described in humans. The novel transcript that we have found encodes a third form of human GAD, one that is present in pancreatic islets, testis, and likely other endocrine organs, including adrenal cortex. Why GAD25 is expressed in these tissues as well as in rodent embryonic brain remains uncertain.