The JI
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     
 

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow Request Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Myers, M. A.
Right arrow Articles by Rowley, M. J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Myers, M. A.
Right arrow Articles by Rowley, M. J.
The Journal of Immunology, 2000, 165: 3830-3838.
Copyright © 2000 by The American Association of Immunologists

Conformational Epitopes on the Diabetes Autoantigen GAD65 Identified by Peptide Phage Display and Molecular Modeling

Mark A. Myers, Janet M. Davies1, Jonathan C. Tong, James Whisstock, Marita Scealy, Ian R. Mackay and Merrill J. Rowley2

Department of Biochemistry and Molecular Biology, Monash University, Clayton, Victoria, Australia


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
The major diabetes autoantigen, glutamic acid decarboxylase (GAD65), contains a region of sequence similarity, including six identical residues PEVKEK, to the P2C protein of coxsackie B virus, suggesting that cross-reactivity between coxsackie B virus and GAD65 can initiate autoimmune diabetes. We used the human islet cell mAbs MICA3 and MICA4 to identify the Ab epitopes of GAD65 by screening phage-displayed random peptide libraries. The identified peptide sequences could be mapped to a homology model of the pyridoxal phosphate (PLP) binding domain of GAD65. For MICA3, a surface loop containing the sequence PEVKEK and two adjacent exposed helixes were identified in the PLP binding domain as well as a region of the C terminus of GAD65 that has previously been identified as critical for MICA3 binding. To confirm that the loop containing the PEVKEK sequence contributes to the MICA3 epitope, this loop was deleted by mutagenesis. This reduced binding of MICA3 by 70%. Peptide sequences selected using MICA4 were rich in basic or hydroxyl-containing amino acids, and the surface of the GAD65 PLP-binding domain surrounding Lys358, which is known to be critical for MICA4 binding, was likewise rich in these amino acids. Also, the two phage most reactive with MICA4 encoded the motif VALxG, and the reverse of this sequence, LAV, was located in this same region. Thus, we have defined the MICA3 and MICA4 epitopes on GAD65 using the combination of phage display, molecular modeling, and mutagenesis and have provided compelling evidence for the involvement of the PEVKEK loop in the MICA3 epitope.


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
Glutamic acid decarboxylase (GAD)3 is a major auto-antigen in type 1 diabetes and the biosynthetic enzyme responsible for the production of the inhibitory neurotransmitter {gamma}-amino butyric acid from glutamic acid, using pyridoxal 5'-phosphate (PLP) as a cofactor. Abs to glutamic acid decarboxylase (anti-GAD) occur in 70–80% of subjects with newly diagnosed type 1 diabetes and predominantly recognize the smaller, 65-kDa isoform, GAD65 (1, 2, 3, 4), while autoantibodies to the 67-kDa isoform, GAD67, occur in a minority of cases (1). Anti-GAD65 can be detected up to 10 yr before clinical onset and has proven to be a useful marker for identifying individuals at risk of developing type 1 diabetes (5). The pathogenic significance of the humoral autoimmune response to islet ß-cell components remains unclear, but in the nonobese diabetic mouse, which spontaneously develops autoimmune diabetes, B lymphocytes contribute as APCs to the progression of islet ß-cell destruction (6, 7, 8).

It is uncertain what initiates autoimmune type 1 diabetes. Infection with coxsackie B virus (CBV) is suspected due to the coincidence of CBV infection with the disease (9, 10, 11), with one explanation being the activation of autoreactive T lymphocytes by inflammatory events (bystander activation) (12). However, sequence similarity between a region of GAD65 and the P2-C protein of CBV, including a stretch of six identical amino acids (PEVKEK) (13), implicates molecular mimicry of GAD65. If mimicry were an initiating event, a major autoantigenic epitope of GAD65 should correspond to the region of similarity shared by CBV and GAD65. This appears to hold for T lymphocyte epitopes of GAD65, because peptides corresponding to the PEVKEK region from both GAD and CBV can bind to the permissive HLA molecule DR3 (14) and can stimulate T lymphocytes from subjects with newly diagnosed or presymptomatic type 1 diabetes (15). Furthermore, antisera from rabbits immunized with synthetic peptides encompassing the PEVKEK region of CBV can cross-react with GAD65 by ELISA (16). Importantly, however, type 1 diabetes-associated autoantibodies to GAD65 are not cross-reactive with CBV (17).

The autoepitopes of GAD65 that react with type 1 diabetes sera are dependent on conformation and reside within aa 244–570 of GAD65 (18, 19, 20, 21, 22). Islet cell mAbs (MICAs), derived from blood lymphocytes of subjects with type 1 diabetes, react with conformational epitopes of GAD65 (20, 21), and these can be divided into three clusters. These are the MICA4 cluster, comprising an epitope in the middle region (aa 244–450); the MICA3 cluster, comprising an epitope formed by the middle and C-terminal regions (aa 244–585); and a mixed epitope cluster at the C-terminal region (aa 450–585) (23). These epitope clusters appear to correspond to epitopes defined serologically by Abs to GAD in type 1 diabetes. Abs to the MICA4-type epitope appear first, with later epitope spreading to the other clusters (23, 24).

Techniques currently in use for mapping Ab epitopes are inadequate for defining conformational epitopes. We have combined two technological advances that give information on conformational epitopes despite the lack of structural knowledge about the Ag. The first is the identification of Ab-binding peptides through the screening of phage-displayed random peptide libraries. The sequence of the Ab-binding peptides identifies amino acids critical for binding. The second is sequence alignment and secondary structure prediction of PLP-dependent decarboxylases and aminotransferases, which have shown that the PLP binding domain of ornithine decarboxylase (OrnDC) and aspartate aminotransferase from bacteria share a protein fold that is common to other PLP-dependant proteins, including GAD65 (25, 26). This has enabled the derivation of structural models of the GAD PLP binding domain (27), and such models can be used to formulate specific hypotheses regarding epitope structure. Accordingly we screened phage-displayed random peptide libraries with the GAD65-reactive mAbs, MICA3 and MICA4, and mapped the phage peptide sequences to a homology model of the PLP binding domain of GAD65.


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
Phage display

Two random nonapeptide phagemid libraries in pC89, in which peptides were inserted into the amino terminal of the pVIII coat protein of bacteriophage f1, were obtained from Dr. Alessandro Luzzago, Istituto di Richerche Biologia Molecolare (Rome, Italy). The libraries contain random sequences of nine amino acids inserted into the amino terminal of the pVIII coat protein of bacteriophage f1. One library contained cysteine residues flanking the nine-amino acid insert so that formation of a disulfide bridge would constrain the conformation of the insert while the other library was unconstrained. The libraries were screened by biopanning as described previously (28). Briefly, a 1-nM solution of MICA3 or 1 µM MICA4 was mixed overnight at 4°C with 1 µl of each phagemid library (2 µl total) in 500 µl of PBS, pH 7.3, containing 1 mg/ml BSA (PBS-BSA). Phage that had bound were isolated with paramagnetic beads coated with anti-human Ig (Chemicon International, Temecula, CA), washed 10 times in PBS-BSA, eluted from the magnetic beads with 1 mg/ml BSA, in glycine-HCl, pH 2.2, and then neutralized with 1 M Tris-HCl, pH 9.1. After this first round of positive selection, there was a negative selection step in which phage not specifically reactive with the mAbs were removed using magnetic beads coated with anti-human Ig in the absence of primary Ab. Following negative selection, the remaining phage were amplified and further enriched by four additional rounds of positive selection using a 1-nM solution of the mAbs with an amplification step between each round. Phage were purified from single colonies after the third and fifth rounds of biopanning and were tested for reactivity with the mAbs by capture ELISA.

DNA inserts were sequenced using the M13–40 sequencing primer, [{alpha}-35S]dATP Redivue (Amersham International, Aylesbury, U.K.), and the Sequenase version 2.0 T7 DNA polymerase sequencing kit (U.S. Biochemical, Cleveland, OH) according to the manufacturer’s instructions. Ambiguous sequences were resequenced using the nucleotide analogue dITP in place of dGTP.

Antibodies

MICA3 and MICA4 were donated by Dr. Joseph Endl (Roche, Mannheim, Germany). The B lymphocyte lines that secreted these Abs were derived from a patient with newly diagnosed type 1 diabetes (20). Both Abs are of the IgG class, and both are reported to have high avidity for GAD65 relative to other GAD65-binding MICAs (29). GAD1 and GAD6 are mouse mAbs to GAD65 (30, 31) and were prepared as described previously (28).

Capture ELISA

Recombinant phage selected by the mAbs were tested for reactivity with the selecting mAb using a capture ELISA. Capture was achieved by coating wells of a 96-well microtiter plate (Maxisorp, Nunc, Naperville, IL) with 100 µl of the mAb diluted to a concentration of 10 µg/ml in sterile PBS. After the microtiter plate was incubated at 4°C overnight in a humidified chamber, wells were washed and blocked with 200 µl/well of 1% skim milk and 0.05% Tween-20 in PBS, pH 7.3 (SM/PBS/Twn), for 2 h on the shaker at room temperature. The blocking solution was removed, and the plate was washed with 1x TBS/0.05% Tween-20. Purified phage were diluted 1/20 in sterile PBS, and 100 µl was added to each mAb-coated well and incubated at 4°C overnight in a humidified chamber. Wells were washed with 1x TBS/0.05% Tween-20, then blocked again with 200 µl/well of SM/PBS/Twn for 1 h on the shaker at room temperature. Sheep anti-M13 Abs (100 µg/ml; Pharmacia Biotech, Piscataway, NJ) were added to the plate at a concentration of 1/2000 diluted in SM/PBS/Twn, to which was added Escherichia coli K91 lysate (32) at 10 µl/ml. IgG was detected using HRP-conjugated anti-sheep/goat Ig (Silenus, Hawthorn, Australia), with 0.5 mg/ml 2,2-azino-di-[3-ethyl-benzthiazoline sulfonate] (Roche) in 0.03 M citric acid, 0.04 M Na2HPO4, and 0.003% H2O2, pH 4, as substrate. Phage containing no insert or inserts with stop codons were used as background controls. Phage that gave OD readings greater than the mean + 3 SD for the background controls were considered to be specifically reactive with the capturing Ab.

Sequence analysis

The sequences of the peptides selected by each mAb that were reactive in the capture ELISA were aligned with each other using the multiple sequence alignment algorithm PILEUP (33) in conjunction with the Tudos matrix for scoring amino acid substitutions on the basis of physicochemical properties (34), as previously described (35). Peptides that were selected more than once were included as many times as they occurred. A low penalty of 1 was applied for the introduction of gaps between aligned sequences, because the peptides expressed by the phage are near the amino terminus of the pVIII protein and as such are structurally flexible (36, 37). The low road alignment option was chosen such that sequences were aligned to the subsequent sequence in the growing guide tree that depicts the clustering of related peptides. Each cluster of similar peptides was then aligned as a group with the amino acid sequence of human GAD65 to identify regions with amino acid composition similar to that of the peptides. Peptides selected using MICA3 were also aligned individually with the sequence of GAD65 using Clustal W (38).

Molecular modeling

The program MODELLER (39) was used to generate a homology model of the PLP binding region of GAD65. The x-ray crystal structure of OrnDC (pdb identifier 1ord) (40) and the primary sequence alignment derived by Momany (25) was used as a template for the model. The model was further refined using the programs CHARMm (Micron Separations, San Diego, CA), with dihedral restraints applied where required. Examination of a Ramachrandran plot of the final model revealed that all residues were in allowed conformations.

Mutagenesis

The codons for the PEVKEK loop of GAD65, aa 258–270, were deleted using a QuickChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) and appropriate oligonucleotides to produce the mutant GAD{Delta}258–270, which was confirmed by sequencing. The template cDNA was a hybrid construct encoding aa 1–101 of GAD67 fused to aa 96–585 of GAD65 (GAD67/65), which reacts similarly to GAD65 in radioimmunoprecipitation (RIP) assays with diabetic sera (41).

RIP and inhibition with phagotopes

Reactivity with hybrid GAD67/65 and with the mutant GAD{Delta}258–270 was tested by RIP using 35S-labeled rabbit reticulocyte lysate (RRL)-expressed Ag. Hybrid GAD67/65 encodes aa 1–101 of GAD67 fused to aa 96–585 of GAD65 (GAD67/65), and this Ag reacts similarly to GAD65 in RIP assays with diabetic sera (41). The assay procedure has been described previously (42). Briefly, 40,000 dpm of RRL-expressed GAD67/65 or mutant GAD{Delta}258–270 per reaction was left with Ab overnight at 4°C, after which the immune complexes were precipitated using protein A-Sepharose. The protein A-Sepharose and immune complexes were harvested by centrifugation and washed, and the amount of bound Ag was determined by scintillation counting. The RRL-expressed Ags were diluted in PBS containing 5% BSA to preserve Ag conformation. RIP assays were performed using MICA3, MICA4, the mouse mAb GAD6 and GAD1, and sera from 45 type 1 diabetes sera. The total amount of Ag present in each translation was assessed by RIP using the mouse mAb GAD6, which binds a continuous epitope of GAD65 located in the C-terminal domain (21). GAD1 recognizes a highly conformational epitope that is destroyed by most truncations and mutations in GAD65 and hence was used to evaluate the conformation of the mutant (21, 43). For RIP with the type 1 diabetes sera, 5 µl of serum was tested in a 30-µl final volume.

The three most reactive phagotopes selected with each mAb were used to inhibit MICA3 or MICA4 binding to GAD67/65 in the RIP assay. Twenty-five microliters of phagotope suspension was preincubated with 0.4 ng of mAb in a final volume of 35 µl at room temperature for 2 h. 35S-labeled RRL expressed GAD67/65 (40,000 dpm; 5 µl) was then added, and RIP was performed as previously described. Each inhibition was performed in quadruplicate, and results were compared with the results from inhibition with a phage containing no insert. The assays were performed twice, with similar results.

Statistics

Student’s t tests and {chi}2 tests were performed, and SDs were calculated using Microsoft Excel software (Redmond, WA).


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
MICA3 binding peptides

Individual phage selected after three or five rounds of biopanning were measured for their ability to bind to the selecting Ab by capture ELISA. Phage that produced an OD greater than the mean + 3 SD of phage with inserts containing stop codons or no inserts (OD ± SD = 0.147 ± 0.038) were taken to represent phagotopes that bound the Ab. Twenty-eight of 35 phagotopes from the fifth round of positive selection and 0 of 32 phage from the third round of positive selection were reactive by capture ELISA. The inserts contained in MICA3-binding phagotopes were sequenced, and the ELISA results and the sequences encoded by their DNA inserts are shown in Table IGo. The sequence QKRAKGLSA was selected eight times, and seven of these eight identical phagotopes were among the 12 most reactive by ELISA (range of OD values, 0.414–0.505). Additional sequences, ARKVKGAAG, MRKANSPPT, and YRKKSSAEL, were each selected twice. Twenty-seven of the 28 sequences contained at least two (4 of 28), but usually three (23 of 28), basic amino acid residues, with the motifs RK or KR the most common. The proportion of basic amino acids was significantly greater than that expected to occur in the peptides according to codon frequency (p < 0.0001, by {chi}2 test), suggesting that basic residues are important for MICA3 binding.


View this table:
[in this window]
[in a new window]
  		Table I. MICA3-binding peptides selected from the nonameric phage-displayed random peptide library1

 
Inhibition of MICA3 binding to GAD67/65 was performed using the three most reactive phagotopes selected with MICA3, QKRAKGLSA, YRKKSSAEL, and MRKANSPPT, compared with a phage with no insert. The phagotope encoding the peptide MRKANSPPT inhibited MICA3 binding to GAD67/65 by 11% (p = 0.02, by Student’s t test), and the phagotopes encoding the peptides QKRAKGLSA and YRKKSSAEL showed no significant inhibition in this assay format.

Sequence alignments of MICA3 binding peptides

The MICA3 binding peptides were aligned with each other and grouped according to their degree of similarity. These groups of sequences were then aligned with GAD65 to identify regions of GAD65 that contain similar amino acid residues (Fig. 1Go). This procedure identified five different regions of GAD65. These were residues 262–270 where three peptides aligned, including two with the sequence YRKKSSAEL; residues 285–296 where five peptides aligned; residues 315–334 where 11 peptides aligned, including eight with the sequence QKRAKGLSA and two with the sequence ARKVKGAAG; residues 373–395 where six peptides aligned, including two with the sequence MRKANSPPT; and residues 511–541 where three peptides aligned. Each of these regions lies within the minimum region of GAD65 required for MICA3 binding (residues 244–570) as determined by deletion analysis (21, 22).



View larger version (44K):
[in this window]
[in a new window]
  		FIGURE 1. Alignment of phage-displayed peptides that are reactive with MICA3 mAb to GAD with the amino acid sequence of human GAD65. The PLP binding domain of GAD from aa 218–470 is shaded, and the minimal region of GAD65 required for MICA3 binding, from aa 244–570, is underlined. The phage-displayed peptides were aligned with the amino acid sequence of GAD65 as groups of similar peptides. The number of times a peptide was selected by MICA3 is given, except where the peptide was selected once. Residues present in the peptides that are identical with amino acids in the region of alignment with GAD65 are in bold, and physicochemically similar amino acids are in italics.

 
Mapping of the MICA3 epitope

A model of the three-dimensional structure of the GAD65 PLP-binding pocket was derived, and a representation of the model is shown in Fig. 2Goa. The model predicts that the first three regions of GAD65 identified by the sequence alignments for MICA3 represent a surface-exposed patch, these being residues 262–270 that constitute a surface-exposed loop and residues 285–296 and 315–334 that form surface-exposed {alpha}-helixes (Fig. 2Goa). These three regions lie within 25 Å of each other in the structure. The fourth region of alignment, aa 373–394, includes surface-exposed residues but lies more than 30 Å distant from the other regions. Furthermore, the alignment to aa 373–394 required the introduction of many gaps into the peptide sequences (Fig. 1Go), suggesting a low degree of similarity, and all seven of these peptides contained sequence similarities to the group of peptides that aligned with aa 262–270. Hence, a contribution of the region of alignment at aa 373–394 to the MICA3 epitope was considered unlikely. Interestingly, when peptides that aligned to region 4 (but not other peptides) were realigned with GAD65 individually using Clustal W, only one of the peptides (HKKSLSSPS) realigned in the same position. Depending on the gap penalty used, each of the other sequences aligned to region 2, 3, or 5: the sequence SRKKTFTGA realigned to region 2, LRKKGYDPG realigned to region 3, and the sequences MRKANSPPT (isolated twice) and LSRKKTLTT realigned, with gaps, to region 5 (residues 511–531). Region 5 lies outside the model of the PLP domain, but within a region known to be important for MICA3 binding (see Discussion).



View larger version (68K):
[in this window]
[in a new window]
  		FIGURE 2. a, Regions of GAD65 that aligned with the selected phage-displayed peptides form a presumptive epitope for MICA3. A ribbon diagram of the homology model of the PLP binding domain of GAD65 is shown highlighting the regions to which MICA3 binding peptides aligned. Residues 262–270, which constitutes the PEVKEK loop, is colored yellow, residues 285–296 and 315–334 of GAD65 are colored red, and residues 373–395 are colored blue. The red and yellow regions form a 25 Å diameter surface patch, which is the usual area occupied by an Ab paratope. The blue region (residues 373–395) lies outside this area. The PLP-binding lysine 396 is colored pink. b, Lys358 and Glu321 (yellow) are mostly buried and form a salt bridge. The GAD65 PLP binding domain polypeptide chain is represented as a ribbon. Residues that may contribute to the MICA4 epitope (Tyr357, Thr335, Lys356, Arg318, Leu347, Ala349, and Val350) are shown in red. c, Representation of the GAD65 PLP binding domain with the presumptive MICA3 (blue) and MICA4 (red) epitopes highlighted. The epitopes encompass nonoverlapping regions of <25 Å diameter. d, The structure of OrnDC showing the dimeric configuration and the folding of the C terminus around the PLP binding domain of the partner subunit. One subunit is colored yellow with the C-terminal in blue, the partner subunit is colored red with the C-terminal in magenta. By analogy, in GAD65 the presumptive MICA3 epitope may lie proximal to the C terminus of the partner subunit.

 
The first three sequence alignments, to aa 262–270, 285–296, and 315–334, identified at least three amino acids in each of these regions of GAD65 that were either identical or physicochemically similar to the aligned amino acids in the peptides (Fig. 1Go). These were Lys263, Lys265, Leu270, Lys286, Lys287, Ala291, Gly295, Thr296, Ala322, and Lys323, all of which are surface exposed according to our model. These amino acid residues are proximal to each other in the model, are similar to amino acids in the MICA3 binding peptides, and so may contribute to the conformation-dependent MICA3 epitope.

MICA4 binding peptides

Phage selected after three or five rounds of biopanning were measured for their ability to bind to the selecting Ab by capture ELISA. Phage that produced an OD greater than the mean + 3 SD of phage with either no insert or inserts containing stop codons (OD ± SD = 0.213 ± 0.021) were taken to represent phagotopes containing peptides that bound the Ab. Twelve of 14 phage from the fifth round of positive selection and 14 of 21 phage from the third round of positive selection reacted by capture ELISA. The inserts contained in 26 phagotopes that expressed peptides reactive with MICA4 were sequenced, and the ELISA results and peptide sequences are shown in Table IIGo. The sequence MRKSTGTAS was selected twice, and all other phagotopes contained a unique insert. The two most reactive sequences by capture ELISA, QKKMVALSG and SRKVALQGG, shared the motif Val-Ala-Leu-Polar-Gly.


View this table:
[in this window]
[in a new window]
  		Table II. MICA4-binding peptides selected from the nonameric phage-displayed random peptide library1

 
Like the MICA3 binding peptides, the MICA4 binding peptides were rich in basic amino acid residues. Thirteen of the 26 peptides contained at least three basic residues, 11 contained two basic residues, and two contained a single basic residue. Overall, basic amino acids constituted 29% of the amino acids in the peptides, which was significantly greater than expected by chance based on codon frequency (p < 0.0001, by {chi}2 test). In addition, serine and threonine tended to be over-represented, constituting 20% of the amino acids in the peptides (p = 0.076, by {chi}2 test).

Inhibition of MICA4 binding to GAD67/65 was tested using the three most reactive phagotopes selected by biopanning with MICA4. The MICA4-selected phagotopes encoding the peptides SRKVALQGG and RRKLRTNAG inhibited MICA4 binding to GAD67/65 by 9% (p = 0.02, by Student’s t test) and 15% (p = 0.01, by Student’s t test), respectively, and the phagotope encoding the peptide QKKMVALSG showed no significant inhibition (4%) in this assay format.

Mapping of the MICA4 epitope

In contrast to the peptides selected with MICA3, the sequence alignment approach gave no clear result for MICA4-selected peptides as individual sequences or pairs of sequences aligned throughout the GAD65 sequence (data not shown). This appeared to be due to the great diversity of sequences selected with MICA4. However, examination of the sequence of GAD65 reveals the sequence Leu-Leu-Ala-Val at aa 347–350, which is the reverse of the motif Val-Ala-Leu contained in the two most reactive MICA4 peptides. Lys358 has previously been shown to be critical for MICA4 binding to GAD65 (27), and aa 347–350 are adjacent to this residue (Fig. 2Gob). The MICA4 binding peptides were rich in hydroxyl-containing amino acids, and consistent with this is the presence of Ser313, Tyr357, and Thr335 near Lys358. Other amino acids in this region are the hydrophobic Phe327 and basic Lys356 and Arg318. Thus, amino acids most likely to contribute to the MICA4 epitope are Val350, Ala349, Leu347, Tyr357, Thr335, Lys356, and Arg318 (Fig. 2Gob). This presumptive MICA4 epitope lies 27 Å from the PEVKEK loop and does not overlap the MICA3 epitope (Fig. 2Goc).

Immunoreactivity of GAD{Delta}258–270

Our prediction that the PEVKEK loop comprising aa 262–270 forms a central part of the MICA3 epitope prompted the testing of a deletion derivative of GAD65 lacking these residues. The homology model and alignments of GAD65 with OrnDC indicated that aa 258–270 of GAD65 represented a sequence insertion relative to ornithine decarboxylase. Assuming conservation of the major fold of the PLP binding domain of GAD65 and ornithine decaboxylase, deletion of the PEVKEK loop should cause minimal disruption of the conformation of the PLP binding domain. Maintenance of conformation was tested by the ability of the mouse mAb, GAD1, to bind GAD{Delta}258–270. GAD1 recognizes a highly conformational epitope of GAD65 that has previously been shown to require the entire protein (21, 43). The binding of GAD1, MICA3, and MICA4 to GAD{Delta}258–270 was measured by RIP assay. In each case, dilutions of Ab were used that precipitated 70% of the available GAD67/65 Ag (Fig. 3Go). Binding of MICA3 and MICA4 was reduced compared with that of GAD1 by 70 and 40%, respectively.



View larger version (37K):
[in this window]
[in a new window]
  		FIGURE 3. Reactivity of MICA3 and MICA4 with GAD is reduced by deletion of residues 258–270, the PEVKEK loop of GAD65. Gray, GAD1; white, MICA3; black, MICA4. The mutation was introduced into GAD67/65, a derivative of GAD65 that has the first 96 aa replaced with the first 101 aa of GAD67 and reacts similarly with diabetic sera and mAbs to GAD65 (see Materials and Methods). Results are expressed as the percentage precipitated of the total precipitable Ag, as measured using the conformation-independent mouse mAb, GAD6. Six nanograms of MICA3 or MICA4 were used in the RIP assay. At this concentration, binding to the mutant is reduced relative to the highly conformation dependent Ab GAD1 by 71% for MICA3 and by 38% for MICA4. Results are the mean of four replicates, and bars show SDs. Similar results were obtained in three independent experiments.

 
We determined whether the epitope disrupted by deletion of aa 258–270 of GAD represents a frequent disease-associated epitope by testing 45 type 1 diabetes sera (Fig. 4Go) along with GAD1. Because the reactivity of GAD1, which recognizes a highly conformational epitope, with the mutant was 54% of that with GAD67/65, this figure was taken as the proportion of mutant GAD molecules that were in a correctly folded conformation in the GAD{Delta}258–270 preparation. Of the 45 type 1 diabetes sera, 33 (73%) had greater than 54% reduction in reactivity with GAD{Delta}258–270 compared with GAD67/65, indicating that aa 258–270 do contribute to the formation of an epitope recognized by most type 1 diabetes sera. The 12 that retained >54% reactivity had relatively high levels of anti-GAD, and hence were likely to contain Abs to both linear and conformational epitopes (19).



View larger version (12K):
[in this window]
[in a new window]
  		FIGURE 4. Reactivity of the majority of diabetic sera with GAD is reduced by deletion of residues 258–270, the PEVKEK loop of GAD65. The GAD{Delta}258–270 mutant and GAD67/65 were expressed and used as Ag in RIP assays as described in Fig. 3Go. Results were expressed as unitsGAD{Delta}258–270 or unitsGad67/65 calculated as a percentage of the counts precipitated by GAD6. The x-axis shows the level of anti-GAD in samples measured using GAD67/65. The y-axis represents the percent residual reactivity with GAD{Delta}258–270, calculated according to the formula (1 - unitsGAD{Delta} (258–270)/unitsGad67/65) x 100. Results for diabetic sera ({diamondsuit}) and GAD1 ({square}) are shown. The dotted line shows the percentage of reactivity remaining for GAD1.

 

   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
Ab epitopes are often formed by amino acids that are distant from each other in the primary sequence of a protein but are brought together as a reactive site on the surface of the folded molecule. This is true of GAD65, for which the autoantibody epitopes are highly dependent on conformation (18, 20, 21, 22). Consequently, the utility of truncated molecules in epitope mapping has been limited, as even small deletions have had considerable impact on the structure of the molecule. The infrequent reactivity with autoantibodies of the closely related isoform, GAD67, has prompted the use of hybrids of GAD65 and GAD67 or swapping of individual amino acids that differ between the isoforms (27, 44). While more informative, these approaches are also limited because amino acids important for epitope formation that are shared by the isoforms will not be detected. In the study by Schwartz et al. (27) the different reactivities of GAD65 and GAD67 with various monoclonal autoantibodies could not readily be explained by single amino acid differences. Rather, a difference in conformation of the GAD isoforms appears to be responsible for the particular autoantigenicity of GAD65.

Ab screening of random peptide libraries overcomes the limitations of epitope mapping with truncated molecules and swap mutants. In the present study we have identified phagotopes that bind to mAbs derived from a subject with type 1 diabetes, MICA3 and MICA4, and have shown that several of the most reactive phagotopes inhibit reactivity with GAD in RIP assays. Analysis of the sequences of phage-displayed peptides that bound to MICA3 and MICA4 identified several short regions of sequence identity with the GAD65 amino acid sequence, and a model of the PLP binding domain of GAD was used to locate these regions on the surface of the GAD molecule.

For MICA3-selected phagotopes, the regions of sequence identity with the GAD PLP binding domain were located on a surface-exposed loop and in two helixes that form a hydrophilic patch on the surface of the folded protein. The existence of this presumptive epitope is consistent with studies using truncated GAD65 molecules, which showed that aa 245–570 bind MICA3, but aa 295–570, which lacks the PEVKEK loop, do not (21). We have demonstrated that deletion of aa 258–270 of GAD greatly reduces MICA3 binding. One explanation for this may be that the PEVKEK loop is essential for correct folding of the PLP domain, and we have observed that deletion of the PEVKEK loop causes conformational instability of the protein (data not shown). However, the binding of a highly conformation-dependent mAb, GAD1, was reduced to a lesser degree than the binding of MICA3, indicating that conformation was mostly maintained under our particular assay conditions. Hence, the interpretation is that the PEVKEK loop region is necessary for the complete formation of the epitope recognized by MICA3. Testing of type 1 diabetic sera suggests that the PEVKEK loop forms an important element of a common type 1 diabetes-associated epitope of GAD65. Given the abundance of basic amino acid residues in peptides selected with MICA3, we propose that Lys263 and/or Lys265 in the PEVKEK loop may be directly involved in binding of MICA3 to GAD65.

The alignment of the phage peptide inserts with GAD65 indicated that a region of the C terminus may also be important for binding of MICA3, because a group of MICA3 binding peptides aligned with amino acids between residues 511 and 531 of GAD65. The C-terminal domain of GAD65 is known to be essential for the GAD65 specificity of MICA3 binding, as swapping aa 511–585 with those of GAD67 eliminates MICA3 binding (27). A recent study delineated this region more precisely and found that aa 513–527 of GAD65 were necessary for MICA3 binding (44), consistent with our alignment of MICA3-selected peptides with aa 511–531. This region of GAD65 contains nine differences between GAD65 and GAD67, including three (Thr515, Leu516, and Ser527) that were identical or physicochemically conserved in the aligned peptide sequences (Fig. 1Go). It is noteworthy that in the OrnDC dimer, the C-terminal domain of one subunit wraps around the partner subunit, bringing part of a C-terminal domain near to the partner subunit PLP binding domain (Fig. 2God). We have previously found that a dimeric configuration of GAD65 appears to favor the binding of diabetes-associated autoantibodies (45). Thus, we propose that the entire MICA3 epitope consists of two structural components, one on the surface of the PLP binding domain and the other in the C-terminal domain between aa 511 and 531. By analogy with OrnDC, the C-terminal domain of GAD65 would wrap around the PLP binding domain of the partner subunit proximal to the PEVKEK loop 262–270 and {alpha}-helixes 284–296 and 315–334 so that in the quaternary structure of the GAD65 dimer the two structural components of the MICA3 conformational epitope are brought together. Knowledge of the three-dimensional structure of dimers of GAD65 derived from x-ray crystallography will be necessary to test this prediction.

In regard to the MICA4 epitope, a previous study (27) has identified a single amino acid substitution, Lys358 to Asn, that abrogates MICA4 binding to GAD65. The MICA4 binding peptides that we derived by phage display indicated that in addition to basic residues, hydroxyl groups and hydrophobic amino acids were important for MICA4 binding. Examination of the homology model of the PLP binding domain identifies Ser313, Tyr357, Thr335, Lys356, and Arg318 near Lys358, so that the side chains of these residues may interact with the paratope of MICA4. Furthermore, the sequence Leu347, Leu348, Ala349, and Val350, which was present in reverse in the two peptides most reactive with MICA4, is also near Lys358. Because there were two mutants derived by Schwartz et al., (27), Ser313 to Ala and Leu348 to Gln, that did not affect MICA4 binding, these residues could be excluded as potential contributors to MICA4 binding. Hence, as for the MICA3 epitope, residues common to both GAD isoforms appear to contribute to the MICA4 epitope. Thus, the surface patch formed by residues Arg318, Thr335, Leu347, Ala349, Val350, Lys356, and Tyr357 could constitute the MICA4 epitope.

The observation that substitution of Lys358 for Asn abrogates MICA4 binding to GAD65 (27) may suggest that there is a direct interaction between Lys358 and the paratope of MICA4. However, our model of the GAD65 PLP binding domain (Fig. 2Gob) predicts that Lys358 is buried and, therefore, inaccessible to Ab. Rather, we predict that Lys358 forms a salt bridge with Glu321 such that substitution of Lys358 would disrupt this salt bridge and cause a local conformational disruption and loss of Ab binding, thus accounting for the earlier observations (27). The possible existence of a salt bridge between Lys358 and Glu321 is intriguing in view of the fact that GAD67 has Asn in place of Lys358, which would be unable to form the salt bridge with Glu321. GAD65 and GAD67 have quite different affinities for PLP and different binding to monoclonal autoantibodies (1, 2, 3, 4), consistent with significant conformational differences between the isoforms. While caution must be used in interpreting the model of the molecular structure of the GAD65 PLP binding domain, the contribution of a salt bridge between Lys358 and Glu321 to the conformational differences between the GAD isoforms warrants further investigation.

Autoantibodies to GAD65 appear very early in the preclinical phase of type 1 diabetes, and B lymphocytes are seen as an important component in the pathogenesis of type 1 diabetes by reason of their ability to efficiently present Ag to T lymphocytes (46). If CBV infection with subsequent cross-reaction with GAD65 were indeed the cause of insulitis, then the major autoantibody epitope of GAD65 would logically be in the PEVKEK region where there is sequence similarity between GAD and the P2C protein of CBV. However, autoantibodies to GAD65 do not cross-react with CBV (17). Our molecular model of the PLP binding domain of GAD65 predicts that the PEVKEK region forms a large surface-exposed loop that should be readily accessible to Ab, and mutagenic deletion of this loop confirmed that it contributes to the MICA3 epitope. However, other regions of GAD65, including the C-terminal domain, are also required for formation of the complete epitope. Thus, the nonreactivity of MICA3 with the P2C protein of CBV is explained by the different structural context of the PEVKEK region in the P2C protein compared with that of GAD65. This along with the involvement of surrounding elements on the surface of GAD65 appear to be necessary for optimal binding of MICA3 to GAD65.

In conclusion, screening of phage-displayed random peptide libraries with human mAbs to GAD65, MICA3, and MICA4 yielded peptide sequences that mimic the conformational epitopes of GAD65. Alignment of these sequences together with molecular modeling and mutagenesis implicate the participation of the PEVKEK loop in the conformational MICA3 epitope and support earlier studies suggesting that aa 511–531 of GAD65 contribute to this epitope. The phage display approach has also further revealed the structure of the MICA4 epitope. Our procedures illustrate novel approaches to the structural characterization of conformational Ab epitopes reactive with autoimmune disease Abs.


   Acknowledgments
 
The MICA3 and MICA4 Abs were kindly provided by Dr. Joseph Endl (Roche, Mannheim, Germany). The phage-displayed random peptide libraries were kindly provided by Dr. Alessandra Luzzago, Istituto di Richerche Biologia Molecolare (Rome, Italy). We also thank Prof. Paul Zimmet (International Diabetes Institute, Melbourne, Australia) for his collaborative scientific interest.


   Footnotes
 
1 Current address: Department of Pathology and Immunology, Monash University, Alfred Hospital, Commercial Road, Prahran, 3181 Victoria, Australia. Back

2 Address correspondence and reprint requests to Dr. Merrill J. Rowley, Department of Biochemistry and Molecular Biology, Monash University, Wellington Road, Clayton, 3800 Victoria, Australia. Back

3 Abbreviations used in this paper: GAD, glutamic acid decarboxylase; PLP, pyridoxal 5'-phosphate; CBV, coxsackie B virus; MICA, islet cell mAbs; OrnDC, ornithine decarboxylase; SM/PBS/Twn, 1% skim milk and 0.05% Tween-20 in PBS, pH 7.3; RIP, radioimmunoprecipitation; RRL, rabbit reticulocyte lysate. Back

Received for publication December 30, 1999. Accepted for publication July 6, 2000.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. Hagopian, W. A., B. Michelsen, A. E. Karlsen, F. Larsen, A. Moody, C. E. Grubin, R. Rowe, J. Petersen, R. McEvoy, A. Lernmark. 1993. Autoantibodies in IDDM primarily recognize the 65,000-Mr rather than the 67,000-Mr isoform of glutamic acid decarboxylase. Diabetes 42:631.[Abstract]
  2. Atkinson, M. A., N. K. Maclaren, D. W. Scharp, P. E. Lacy, W. J. Riley. 1990. 64,000 Mr autoantibodies as predictors of insulin-dependent diabetes. Lancet 335:1357.[Medline]
  3. Hagopian, W. A., A. E. Karlsen, A. Gottsater, M. Landin-Olsson, C. E. Grubin, G. Sundkvist, J. S. Petersen, E. Boel, T. Dyrberg, A. Lernmark. 1993. Quantitative assay using recombinant human islet glutamic acid decarboxylase (GAD65) shows that 64K autoantibody positivity at onset predicts diabetes type. J. Clin. Invest. 91:368.
  4. Thivolet, C. H., M. Tappaz, A. Durand, J. Petersen, A. Stefanutti, P. Chatelain, B. Vialettes, W. Scherbaum, J. Orgiazzi. 1992. Glutamic acid decarboxylase (GAD) autoantibodies are additional predictive markers of type 1 (insulin-dependent) diabetes mellitus in high risk individuals. Diabetologia 35:570.[Medline]
  5. Tuomilehto, J., P. Zimmet, I. R. Mackay, P. Koskela, G. Vidgren, L. Toivanen, E. Tuomilehto-Wolf, K. Kohtamaki, J. Stengard, M. J. Rowley. 1994. Antibodies to glutamic acid decarboxylase as predictors of insulin-dependent diabetes mellitus before clinical onset of disease. Lancet 343:1383.[Medline]
  6. Serreze, D. V., H. D. Chapman, D. S. Varnum, M. S. Hanson, P. C. Reifsnyder, S. D. Richard, S. A. Fleming, E. H. Leiter, L. D. Shultz. 1996. B lymphocytes are essential for the initiation of T cell-mediated autoimmune diabetes: analysis of a new "speed congenic" stock of NOD.Ig µ null mice. J. Exp. Med. 184:2049.[Abstract/Free Full Text]
  7. Noorchashm, H., N. Noorchashm, J. Kern, S. Y. Rostami, C. F. Barker, A. Naji. 1997. B-cells are required for the initiation of insulitis and sialitis in nonobese diabetic mice. Diabetes 46:941.[Abstract]
  8. Serreze, D. V., S. A. Fleming, H. D. Chapman, S. D. Richard, E. H. Leiter, R. M. Tisch. 1998. B lymphocytes are critical antigen-presenting cells for the initiation of T cell-mediated autoimmune diabetes in the nonobese diabetic mouse. J. Immunol. 161:3912.[Abstract/Free Full Text]
  9. Clements, G. B., D. N. Galbraith, K. W. Taylor. 1995. Coxsackie B virus infection and onset of childhood diabetes. Lancet 346:221.[Medline]
  10. Hyoty, H., M. Hiltunen, M. Knip, M. Laakkonen, P. Vahasalo, J. Karjalainen, P. Koskela, M. Roivainen, P. Leinikki, T. Hovi, et al 1995. A prospective study of the role of coxsackie B and other enterovirus infections in the pathogenesis of IDDM. Childhood Diabetes in Finland (DiMe) Study Group. Diabetes 44:652.[Abstract]
  11. Szopa, T. M., P. A. Titchener, N. D. Portwood, K. W. Taylor. 1993. Diabetes mellitus due to viruses—some recent developments. Diabetologia 36:687.[Medline]
  12. Horwitz, M. S., L. M. Bradley, J. Harbertson, T. Krahl, J. Lee, N. Sarvetnick. 1998. Diabetes induced by coxsackie virus: initiation by bystander damage and not molecular mimicry. Nat. Med. 4:781.[Medline]
  13. Kaufman, D. L., M. G. Erlander, M. Clare-Salzler, M. A. Atkinson, N. K. Maclaren, A. J. Tobin. 1992. Autoimmunity to two forms of glutamate decarboxylase in insulin-dependent diabetes mellitus. J. Clin. Invest. 89:283.
  14. Vreugdenhil, G. R., A. Geluk, T. H. M. Ottenhoff, W. J. G. Melchers, B. O. Roep, J. M. D. Galama. 1998. Molecular mimicry in diabetes mellitus: the homologous domain in coxsackie B virus protein 2C and islet autoantigen GAD(65) is highly conserved in the coxsackie B-like enteroviruses and binds to the diabetes associated HLA-DR3 molecule. Diabetologia 41:40.[Medline]
  15. Atkinson, M. A., M. A. Bowman, L. Campbell, B. L. Darrow, D. L. Kaufman, N. K. Maclaren. 1994. Cellular immunity to a determinant common to glutamate decarboxylase and coxsackie virus in insulin-dependent diabetes. J. Clin. Invest. 94:2125.
  16. Lonnrot, M., H. Hyoty, M. Knip, M. Roivainen, P. Kulmala, P. Leinikki, H. K. Akerblom. 1996. Antibody cross-reactivity induced by the homologous regions in glutamic acid decarboxylase (GAD65) and 2C protein of coxsackievirus B4. Childhood Diabetes in Finland Study Group. Clin. Exp. Immunol. 104:398.[Medline]
  17. Richter, W., T. Mertens, B. Schoel, P. Muir, A. Ritzkowsky, W. A. Scherbaum, B. O. Boehm. 1994. Sequence homology of the diabetes-associated autoantigen glutamate decarboxylase with coxsackie B4-2C protein and heat shock protein 60 mediates no molecular mimicry of autoantibodies. J. Exp. Med. 180:721.[Abstract/Free Full Text]
  18. Ujihara, N., K. Daw, R. Gianani, E. Boel, L. Yu, A. C. Powers. 1994. Identification of glutamic acid decarboxylase autoantibody heterogeneity and epitope regions in type I diabetes. Diabetes 43:968.[Abstract]
  19. Tuomi, T., M. J. Rowley, W. J. Knowles, Q. Y. Chen, T. McAnally, P. Z. Zimmet, I. R. Mackay. 1994. Autoantigenic properties of native and denatured glutamic acid decarboxylase: evidence for a conformational epitope. Clin. Immunol. Immunopathol. 71:53.[Medline]
  20. Richter, W., J. Endl, T. H. Eiermann, M. Brandt, R. Kientsch-Engel, C. Thivolet, H. Jungfer, W. A. Scherbaum. 1992. Human monoclonal islet cell antibodies from a patient with insulin-dependent diabetes mellitus reveal glutamate decarboxylase as the target antigen. Proc. Natl. Acad. Sci. USA 89:8467.[Abstract/Free Full Text]
  21. Richter, W., Y. Shi, S. Baekkeskov. 1993. Autoreactive epitopes defined by diabetes-associated human monoclonal antibodies are localized in the middle and C-terminal domains of the smaller form of glutamate decarboxylase. Proc. Natl. Acad. Sci. USA 90:2832.[Abstract/Free Full Text]
  22. Syren, K., L. Lindsay, B. Stoehrer, K. Jury, F. Luhder, S. Baekkeskov, W. Richter. 1996. Immune reactivity of diabetes-associated human monoclonal autoantibodies defines multiple epitopes and detects two domain boundaries in glutamate decarboxylase. J. Immunol. 157:5208.[Abstract]
  23. Sohnlein, P., M. Muller, K. Syren, U. Hartmann, B. O. Bohm, H. M. Meinck, M. Knip, H. K. Akerblom, the Childhood Diabetes in Finland Study Group, W. Richter. 2000. Epitope spreading and a varying but not disease-specific GAD65 antibody response in type 1 diabetes. Diabetologia 43:210.[Medline]
  24. Bonifacio, E., V. Lampasona, L. Bernasconi, A. G. Ziegler. 2000. Maturation of the humoural autoimmune response to epitopes of GAD in preclinical childhood type 1 diabetes. Diabetes 49:202.[Abstract]
  25. Momany, C., R. Ghosh, M. L. Hackert. 1995. Structural motifs for pyridoxal-5'-phosphate binding in decarboxylases: an analysis based on the crystal structure of the Lactobacillus 30a ornithine decarboxylase. Protein Sci. 4:849.[Medline]
  26. Qu, K., D. L. Martin, C. E. Lawrence. 1998. Motifs and structural fold of the cofactor binding site of human glutamate decarboxylase. Protein Sci. 7:1092.[Medline]
  27. Schwartz, H. L., J. M. Chandonia, S. F. Kash, J. Kanaani, E. Tunnell, A. Domingo, F. E. Cohen, J. P. Banga, A. M. Madec, W. Richter, et al 1999. High-resolution autoreactive epitope mapping and structural modeling of the 65 kDa form of human glutamic acid decarboxylase. J. Mol. Biol. 287:983.[Medline]
  28. Cook, A. D., J. M. Davies, M. A. Myers, I. R. Mackay, M. J. Rowley. 1998. Mimotopes identified by phage display for the monoclonal antibody CII-C1 to type II collagen. J. Autoimmun. 11:205.[Medline]
  29. Jury, K. M., D. Loeffler, T. H. Eiermann, B. Ziegler, B. O. Boehm, W. Richter. 1996. Evidence for somatic mutation and affinity maturation of diabetes associated human autoantibodies to glutamate decarboxylase. J. Autoimmun. 9:371.[Medline]
  30. Chang, Y. C., D. I. Gottlieb. 1988. Characterization of the proteins purified with monoclonal antibodies to glutamic acid decarboxylase. J. Neurosci. 8:2123.[Abstract]
  31. Gottlieb, D. I., Y. C. Chang, J. E. Schwob. 1986. Monoclonal antibodies to glutamic acid decarboxylase. Proc. Natl. Acad. Sci. USA 83:8808.[Abstract/Free Full Text]
  32. Sambrook, J., E. F. Fritsch, T. Maniatis. 1989. Molecular Cloning: A Laboratory Manual Cold Spring Harbor Laboratory Press, Cold Spring Harbor.
  33. Feng, D.-F., R. F. Doolittle. 1987. Progressive sequence alignment as a prerequisite to correct phylogenetic tress. J. Mol. Evol. 25:351.[Medline]
  34. Tudos, E., M. Cserzo, I. Simon. 1990. Predicting isomorphic residue replacements for protein design. Int. J. Pept. Protein Res. 36:236.[Medline]
  35. Davies, J. M., M. Scealy, Y.-P. Cai, J. Whisstock, I. R. Mackay, M. J. Rowley. 1999. Multiple alignment and sorting of peptides derived from phage-displayed random peptide libraries with polyclonal sera allows discrimination of relevant phagotopes. Mol. Immunol. 36:659.[Medline]
  36. Felici, F., L. Castagnoli, A. Musacchio, R. Jappelli, G. Cesareni. 1991. Selection of antibody ligands from a large library of oligopeptides expressed on a multivalent exposition vector. J. Mol. Biol. 222:301.[Medline]
  37. Makowski, L.. 1993. Structural constraints on the display of foreign peptides on filamentous bacteriophages. Gene 128:5.[Medline]
  38. Thompson, J. D., D. G. Higgins, T. J. Gibson. 1994. CLUSTAL W: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22:4673.[Abstract/Free Full Text]
  39. Sali, A., T. L. Blundell. 1993. Comparative protein modelling by satisfaction of spatial restraints. J. Mol. Biol. 234:779.[Medline]
  40. Momany, C., S. Ernst, R. Ghosh, N. L. Chang, M. L. Hackert. 1995. Pyridoxal-5'-phosphate, ornithine decarboxylase, cofactor, decarboxylase, and crystallographic structure: crystallographic structure of a PLP-dependent ornithine decarboxylase from lactobacillus 30a to 3.0 angstrom resolution. J. Mol. Biol. 252:643.[Medline]
  41. Teoh, K. L., S. Fida, M. J. Rowley, I. R. Mackay. 1998. Autoantigenic reactivity of diabetes sera with a hybrid glutamic acid decarboxylase GAD67–65 molecule GAD67(1–101)/GAD65(96–585). Autoimmunity 28:259.[Medline]
  42. Feeney, S. J., M. A. Myers, I. R. Mackay, P. Z. Zimmet, N. Howard, C. F. Verge, M. J. Rowley. 1997. Evaluation of ICA512As in combination with other islet cell autoantibodies at the onset of IDDM. Diabetes Care 20:1403.[Abstract]
  43. Butler, M. H., M. Solimena, Jr R. Dirkx, A. Hayday, P. De Camilli. 1993. Identification of a dominant epitope of glutamic acid decarboxylase (GAD-65) recognized by autoantibodies in stiff-man syndrome. J. Exp. Med. 178:2097.[Abstract/Free Full Text]
  44. Powers, A. C., K. Bavik, J. Tremble, K. Daw, W. A. Scherbaum, J. P. Banga. 1999. Comparative analysis of epitope recognition of glutamic acid decarboxylase (GAD) by autoantibodies from different autoimmune disorders. Clin. Exp. Immunol. 118:349.[Medline]
  45. Rowley, M. J., Q. Y. Chen, K. L. Teoh, P. Z. Zimmet, T. Tuomi, W. J. Knowles, I. R. Mackay. 1996. Diabetic sera react with the glutamic acid decarboxylase molecule in a dimeric-oligomeric form. Clin. Exp. Immunol. 106:323.[Medline]
  46. Lernmark, A.. 1996. GABA and glutamic acid decarboxylase (GAD): gene to antigen to disease. J. Intern. Med. 240:259.[Medline]



This article has been cited by other articles:


Home page
 Int ImmunolHome page
G. Yao, W. Chen, H. Luo, Q. Jiang, Z. Xia, L. Zang, J. Zuo, X. Wei, Z. Chen, X. Shen, et al.
Identification of core functional region of murine IL-4 using peptide phage display and molecular modeling
Int. Immunol., January 1, 2006; 18(1): 19 - 29.
[Abstract] [Full Text] [PDF]

Home page
 J. Immunol.Home page
J. A. Dromey, S. M. Weenink, G. H. Peters, J. Endl, P. J. Tighe, I. Todd, and M. R. Christie
Mapping of Epitopes for Autoantibodies to the Type 1 Diabetes Autoantigen IA-2 by Peptide Phage Display and Molecular Modeling: Overlap of Antibody and T Cell Determinants
J. Immunol., April 1, 2004; 172(7): 4084 - 4090.
[Abstract] [Full Text] [PDF]

Home page
 J. Biol. Chem.Home page
K. Vanhoorelbeke, H. Depraetere, R. A. P. Romijn, E. G. Huizinga, M. De Maeyer, and H. Deckmyn
A Consensus Tetrapeptide Selected by Phage Display Adopts the Conformation of a Dominant Discontinuous Epitope of a Monoclonal Anti-VWF Antibody That Inhibits the von Willebrand Factor-Collagen Interaction
J. Biol. Chem., September 26, 2003; 278(39): 37815 - 37821.
[Abstract] [Full Text] [PDF]

Home page
 J. Biol. Chem.Home page
D. Bresson, M. Cerutti, G. Devauchelle, M. Pugniere, F. Roquet, C. Bes, C. Bossard, T. Chardes, and S. Peraldi-Roux
Localization of the Discontinuous Immunodominant Region Recognized by Human Anti-thyroperoxidase Autoantibodies in Autoimmune Thyroid Diseases
J. Biol. Chem., March 7, 2003; 278(11): 9560 - 9569.
[Abstract] [Full Text] [PDF]

Home page
 J. Immunol.Home page
M. Bearzatto, H. Naserke, S. Piquer, K. Koczwara, V. Lampasona, A. Williams, M. R. Christie, P. J. Bingley, A.-G. Ziegler, and E. Bonifacio
Two Distinctly HLA-Associated Contiguous Linear Epitopes Uniquely Expressed Within the Islet Antigen 2 Molecule Are Major Autoantibody Epitopes of the Diabetes-Specific Tyrosine Phosphatase-Like Protein Autoantigens
J. Immunol., April 15, 2002; 168(8): 4202 - 4208.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow Request Permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Myers, M. A.
Right arrow Articles by Rowley, M. J.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Myers, M. A.
Right arrow Articles by Rowley, M. J.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS