Human putative AGA sera and normal control sera were obtained from the laboratory serum bank and Advanced Diagnostics Laboratory at the University of Calgary, Canada. Some AGA sera were also provided by Drs R L Humbel (Luxembourg), Kiyomitsu Miyachi (Keigu Clinic, Yokohama, Japan), and Carlos A von Mühlen (Pontifícia Universidade Católica do Rio Grande do Sul, Porto Alegre, Brazil). All sera were provided as anonymous samples and were stored at -80°C until use. The reactivity to Golgi complex in all AGA sera was confirmed by indirect immunofluorescence (IIF) microscopy on HEp-2 cells (Immuno Concepts Inc., Sacramento, CA, USA). Double staining was performed using the human sera (1 : 100 dilution) and rabbit antigiantin antiserum (1 : 500 dilution) as a marker of the Golgi complex 19. The secondary antibodies were Alexa Fluor^® 488 conjugated goat antihuman IgG reagents and Alexa Fluor^® 568 conjugated goat antirabbit IgG reagents (Molecular Probes, Eugene, OR, USA) used at a dilution of 1 : 400. Nuclei were counterstained with 4',6-diamidino-2-phenylindole. By using this approach, a total of 80 sera exhibited specific staining of the Golgi complex.

Recombinant human Golgi autoantigens were produced using the expression plasmid pET28 system in Escherichia coli BL21 (DE3; Novagen, Madison, WI, USA) as previously described 19. Recombinant proteins of golgin-245 (amino acids 811–2083) 11, golgin-160 (amino acids 787–1348) 10, golgin-95/GM130 (amino acids 370–990) 10, and golgin-97 (amino acids 1–767) 12 were subcloned into pET28 vectors for the expression of recombinant bacterial proteins. Six overlapping fragments P1–P6 representing the full-length giantin cDNA (GenBank accession number NM_004487) 7 were generated for epitope mapping analysis. Two fragments (P1 and P2) were obtained by expression cloning from a random-primed lambda phage cDNA library generated from human T24 cells using an antigiantin-specific human serum. Three fragments were obtained from an available expression sequence tag clone (P3, GenBank accession number N_76853; P4, GenBank accession number BG_567238; P5, GenBank accession number AI_458639). One fragment (P6) was cloned from reverse transcription polymerase chain reaction synthesis using total RNA purified from HeLa cells. All six fragments of overlapping recombinant proteins cDNAs were inserted into pET28 expression vector and introduced into Escherichia coli BL21 (DE3). Sequencing was conducted in both directions using custom primers. Bacterial pellets were suspended in 6 M guanidinium hydrochloride containing buffer, and the recombinant proteins were purified by nickel column chromatography according to manufacturer's instructions (Qiagen, Valencia, CA, USA). The concentration of the purified recombinant proteins was measured by a Protein DC Assay Kit (Bio-Rad, Hercules, CA, USA) and these samples were stored at -80°C until they were required for subsequent experiments.

The ELISA protocol described by Rubin 22 was used with some modifications. In brief, Ni column affinity purified recombinant proteins were diluted in phosphate-buffered saline to a final concentration of 1 μg/ml and then coated on Immulon 2 microtiter plates (Dynatech Laboratories, Alexandria, VA, USA). Human sera were diluted 1 : 1000 and then incubated in the antigen-coated wells. Horseradish peroxidase-conjugated goat antihuman IgG (CALTAG Laboratories, San Francisco, CA, USA) was used at 1 : 5000 dilution and the substrate 2,2'-azinobis (3-ethylbenzthiazoline) sulfonic acid was added as the detection reagent. Each sample was analyzed in duplicate and the average optical density (OD) at 405 nm with a substrate development time of 15–45 min was used for data analysis. The cutoff value designating a positive reaction was the mean OD of 12 normal sera +3 standard deviations (SDs).

HeLa cells (ATCC, Rockville, MD, USA) were metabolically labeled overnight with [³⁵S]-methionine (Trans ³⁵S-label; ICN), as described previously 1112. Cell extracts were harvested in a lysis buffer containing 1% NP-40, 50 mmol/l Tris.HCl, pH 7.5 and 150 mmol/l NaCl, and supplemented with Complete™ protease inhibitor cocktail (Boehringer Mannheim, Indianapolis, IN, USA). Soluble fractions were used as substrate for immunoprecipitation reactions by combining 100 μl 10% protein A-Sepharose beads (Sigma, St. Louis, MO, USA), 10 μl human serum, 500 μl NET2 buffer (50 mmol/l Tris-HCl, 150 mmol/l NaCl, 5 mmol/l EDTA, 0.5% Nonidet P-40, 0.5% deoxycholic acid, 0.1% SDS, 0.02% sodium azide, pH 7.4), and 50–100 μl labeled cell extract. After 1 hour of incubation at 8°C, the Sepharose beads were washed five times in NET2. Proteins were eluted in 20 μl sample buffer and analyzed by 10% gel SDS-PAGE 23, followed by autoradiography.

Affinity purified recombinant proteins were loaded on 12.5% SDS-PAGE gels (4 μg/lane), separated by electrophoresis, and transferred to nitrocellulose membranes using a Semi-Dry Trans-Blot apparatus (Bio-Rad), as described previously 19. Human sera containing antigiantin antibodies were used at dilutions of 1 : 100 to 1 : 500. Detection of bound antibodies was achieved using horseradish peroxidase-conjugated goat antihuman IgG antibody (CALTAG Laboratories), used at 1 : 5000 dilution, in combination with enhanced chemiluminescence (Super Signal; PIERCE Products, Rockford, IL, USA).

Reactivity to Golgi complex antigens in all sera was confirmed by IIF, and all sera exhibited a specific staining pattern against Golgi complex structures as determined by colocalization with rabbit antibodies to giantin (Fig. 1). This approach yielded 80 human AGA sera that were used to investigate the prevalence of autoantibodies to five individual Golgi autoantigens represented by purified recombinant proteins in an ELISA and by immunoprecipitation using extracts from [³⁵S]-methionine labeled HeLa cells (Fig. 2). The number of positive sera and frequency of reactivity of the 80 human AGA sera are summarized in Table 1. The most common Golgi complex autoantigen target was giantin (40/80 [50%]) and the second most common target was golgin-245 (19/80 [24%]). The lowest frequency reactivity (3.8%) was to golgin-97, and 25 AGA sera (31.3%) did not react with any of the five Golgi autoantigens used in the present study. Interestingly, the frequency of AGA sera reactive with the five Golgi autoantigens was numerically correlated with the molecular masses of the native Golgi autoantigens (Table 1), and the number of positive sera that reacted with giantin was remarkably higher than those for other golgins. The confirmation by immunoprecipitation was important because some of the Golgi autoantigens used as substrate in the ELISA did not represent full-length proteins and we were concerned that reactivity to these five Golgi autoantigens may be underestimated by ELISA alone. Among the 25 AGA sera that were negative for the five Golgi autoantigens, there were no predominant reactivities other than the five Golgi autoantigens described, even though the immunoprecipitation assay showed unidentified bands that were recognized by many of these sera.

We then determined whether there were specific correlations between any of the five specific AGAs with another AGA. None of the sera had AGAs to four or more of these five Golgi autoantigens. There were 6, 15, and 34 sera with antibodies to three, two, and one of the five Golgi autoantigens, respectively. Among the six sera with antibodies to three of the five antigens, four sera had antigiantin, antigolgin-245 and antigolgin-160, which were the three most common antibodies detected; one serum had antigiantin, antigolgin-245 and antigolgin-95/GM130 (Fig. 2, lane 5); and the remaining serum had antigiantin, antigolgin-245, and antigolgin-97. Among the 15 sera with antibodies to two of the five antigens, five had antigiantin and antigolgin-245, two had antigiantin and antigolgin-160, two had antigiantin and anti-GM130, two had antigiantin and antigolgin-97, two had antigolgin-245 and antigolgin-160, and two had antigolgin-245 and anti-GM130 (Table 2). No specific correlations were observed between the two most abundant antibodies, namely antigiantin and antigolgin-245. For example, among the 19 AGA sera positive for antibody to golgin-245, 11 (57.9%) were positive and 8 (42.1%) were negative for antigiantin antibody. Among the 40 AGA sera positive for antibody to giantin, 11 (27.5%) were positive and 29 (72.5%) were negative for antigolgin-245. Although the number of sera that bound golgin-160, GM130, and golgin-97 were relatively small, it was interesting that sera with these three autoantibodies did not overlap. In other words, sera positive for antigolgin-160 were negative for antibody to GM130 and golgin-97, and sera positive for anti-GM130 were negative for antigolgin-97 (Table 2).

Although we showed that the majority of sera that react with the Golgi complex in an IIF assay react with known Golgi autoantigens, 25 out of 80 (31.3%) AGA sera did not recognize any of the five Golgi autoantigens examined in this study. The data suggest that these sera react with other Golgi autoantigens. A candidate Golgi autoantigen is GMAP-210, a reported cis-Golgi network associated protein that also contains characteristic coiled-coil domains 26. Among the 80 AGA sera, our immunoprecipitation data revealed three sera with a common band at approximately 210 kDa that might represent GMAP-210; however, because we did not have the cDNA for GMAP-210 and the frequency of this putative anti-GMAP-210 antibody was low, we did not confirm these data using independent methods. Two other Golgi proteins that may be candidate autoantigens include golgin-84, an 84 kDa transmembrane Golgi protein 27; and βI Sigma spectrin, a 220 kDa protein that is associated with Golgi complex and vesicles 28. However, our immunoprecipitation data did not yield any bands consistent with these Golgi complex candidate autoantigens. We did not include other known Golgi autoantigens such as golgin-67 13 and p115 29 because the frequencies of these autoantibodies are known to be low. Thus, our data support the notions that the five selected Golgi autoantigens are the most prevalent in AGA sera and that giantin is the most common Golgi autoantigen recognized in AGA sera.

The Golgi autoantigens identified to date are related because they have similar overall secondary structures, as evidenced by extensive coiled-coil rod domains in the central region and small non-coiled-coil or globular domains at both the carboxyl-terminus and amino-terminus 1. The cumulative length of coiled-coil domains are thus directly proportional to the molecular mass of the Golgi autoantigens (Fig. 3b). For example, giantin clearly has more coiled-coil units than does golgin-97.

The human autoimmune response to Golgi autoantigens appears to be highly specific because many AGA sera react with only one (34/80 [42.5%]) or two (15/80 [18.8%]) of the five autoantigens. The specificity of the autoimmune response is demonstrated in the present study. For example, 23 of the 40 antigiantin positive sera reacted with giantin without coexisting autoantibodies to other five golgins. The lack of correlation with the frequency of antibody, as shown in Table 2, is consistent with the conclusion that it is unlikely that the immune response is merely directed at cross-reactive coiled-coils in these self-proteins. It is interesting to note that large (approximately 100 kDa or greater) coiled-coil rich proteins were noted in many non-Golgi cytoplasmic organelles, including endosomal protein EEA1 30 and CLIP-170 31, and the centrosomal proteins pericentrin 32, ninein 33, and Cep250 and Cep110 34. The mitotic organelles are also known to be associated with large coiled-coil rich autoantigens, including the mitotic apparatus proteins NuMA 3536 and centromere-associated protein CENP-E 37 and CENP-F 38. It is noteworthy that we did not observe coexisting autoantibodies to these other coiled-coil rich organelles in our study of these AGA sera. These endosome, centrosome, and mitotic apparatus associated autoantigens are, like the golgins, proteins with high molecular masses and high content of coiled-coil domains. The combination of these two physical features in autoantigen may promote the induction and production of autoimmune antibody in certain disease states. As discussed above, this may have general significance in other autoantigens other than those associated with the Golgi complex.

Another possible reason why giantin has a high frequency of reactivity among the Golgi autoantigens is that giantin is a somewhat unique Golgi complex autoantigen in that it possesses a transmembrane domain. It is not clear why and how the immune system is able to recognize or target these proteins because it is generally thought that the immune system is not exposed to intact intracellular self-antigens. One possible explanation is that they may be surface structures represented on cytoplasmic organelles that are recognized as foreign by the immune system in aberrant disease states associated with unregulated cell death (apoptosis or necrosis) resulting from injury or infection. A variety of autoantigens are cleaved into signature fragments during apoptosis and necrosis 39. The emerging view is that the modified forms of autoantigens generated during cell death might stimulate autoantibody responses if presented to the immune system in a proinflammatory context 40. We and others previously reported that distinct cleavage fragments of Golgi autoantigens were generated during apoptosis and necrosis, and we also observed that, compared with other golgins, giantin is readily cleaved into multiple fragments during apoptosis 1941. Furthermore, we observed that the Golgi complex itself was also fragmented during apoptosis and necrosis 19. It is interesting to note that, unlike 60 kDa SS-A/Ro and some other autoantigens targeted by autoantibodies from Sjogren's syndrome and SLE sera 42, golgins do not appear to be expressed on membranous apoptotic blebs 19. One explanation for this apparent paradox may be the unique nature of the trans-membrane domain of giantin and the GRIP domain of other golgins that allow the presentation of these Golgi membrane-stabilized antigens to the immune system independently of apoptotic blebs. It is possible that giantin is more stably associated with the remaining Golgi surface membrane than other golgins by virtue of its transmembrane domain when cells undergo cell death. Because the cleaved Golgi autoantigens are antigenic 1941, they may play a role in sustaining autoantibody production in certain autoimmune disease states.