Cellular and Subcellular Localization of an M, 64,000 Protein Autoantigen in Insulin-dependent Diabetes*

the intensity of the M, 64,000 band on autoradiograms by densitometry. The densities of equivalent fractions from a similar gradient prepared in parallel without islet material were determined by refractometry.


Antibodies
to an M, 64,000 protein from human or rat islets have been detected at high frequency in newly diagnosed insulin-dependent diabetic patients.
In this study, we show that the antigenic and amphiphilic properties of the rat islet M, 64,000 protein resemble that of the human protein.
We have analyzed the expression of the M, 64,000 protein in populations of pancreatic p and non-B cells and in selected rat tissues by immunoprecipitation of [3SS]methionine-radiolabeled proteins with sera from diabetic patients or from healthy control individuals.
When islet cell populations enriched in @ or non-@ cells were tested for the expression of the M, 64,000 antigen, the protein was primarily observed in the /3 cells. On analyzing preparations of islets, liver, kidney, thyroid, adrenal, pituitary, spleen, and thymus, the protein could only be detected in islets. The protein was also characterized in terms of its subcellular localization by Percoll density gradient centrifugation and was recovered in a fraction enriched in the plasma membrane marker, 5'-nucleotidase. These results are consistent with a ,!3 cell-restricted plasma membrane expression of the protein and support the hypothesis that this protein is a target antigen of fi cell-specific autoimmunity in insulin-dependent diabetes.
Organ-specific autoimmune diseases are frequently associated with the presence of autoantibodies directed against components expressed specifically by the target tissue (l-3). ported to be expressed specifically by the fi cells of the islet (7), whereas antigens recognized by islet cell cytoplasmic antibodies may be expressed by al1 islet cell types (8). In studies to determine the nature of diabetes-associated autoantigens, we identified an islet protein of M, 64,000 to which antibodies were detected in 70-80% of insulin-dependent diabetic children at the clinical onset of the disease (9-11). This protein has been shown to have amphiphilic properties typical of integral membrane proteins (10, ll), but the subcellular compartment with which the protein is primarily associated has not been identified.
Furthermore, it remains to be established whether the protein is expressed specifically in the @ cells of the islet or is present in other tissues. The aims of the present study were to investigate the cellular and subcellular distribution of the M, 64,000 protein in islets and to analyze the expression of the protein in normal tissues. onine and once in 5 ml of 10 mM Hepes (pH 7.4), 150 mM NaCl, 0.5 mM L-methionine, 10 mM benzamidine. Samples were stored at -80 "C.
Triton X-114-extracted tissues were subjected to a modification of the detergent phase separation procedure (16) as described (11). Triton X-lOO-solubilized proteins or Triton X-114 phase purified proteins were precleared by incubation for 2 h at 4 "C with pooled normal human serum (25 rl/lOO ~1 of tissue extract) and for 90 min at 4 "C with protein A-Sepharose (Pharmacia LKB Biotechnology Inc., 100 pl/lOO ~1 of tissue extract). Immunoprecipitation and analysis by sodium dodecyl sulfate-polyacrylamide gel electrophoresis were performed as described previously (11).  (17), except that a Percoll suspension of density 1.064 g/ml (22.5%, v/v) was used, and the density gradient was generated by centrifugation at 17,000 X g for 20 min at 4 "C in a Beckman type 40 rotor. Aliquots (5-10 ~1) of fractions from the gradient were analyzed for NADH-cytochrome c reductase (18), cytochrome oxidase (19), and 5'-nucleotidase (20). Radiolabeled insulin in fractions was determined by immunoprecipitation with a guinea pig anti-insulin serum, subtracting cpm precipitated by a normal guinea pig serum. The protein content of the fractions was determined using the BCA protein assay reagent (Pierce Chemical Co.). Fractions (50 ~1) from the gradient were analyzed for M, 64,000 protein expression. After extraction in 2% Triton X-114 for 2 h, Percoll and insoluble islet material were removed by centrifugation for 1 h at 150,000 x g on a Beckman Airfuge. Detergent extracts were warmed to 30 "C for 3 min, and precipitated detergent aggregates sedimented at 1,000 x g for 3 min. The aqueous phase was reextracted in 1% Triton X-114 for 10 min and the phase separation repeated. The pooled detergent pellets from each fraction were used for immunoprecipitation with sera from normal or diabetic individuals as described above. A quantitative estimate of M, 64,000 protein in gradient fractions was obtained by measuring the intensity of the M, 64,000 band on autoradiograms by densitometry. The densities of equivalent fractions from a similar gradient prepared in parallel without islet material were determined by refractometry.

Immunoreactivity
of the M, 64,000 Protein from Human and Rat Islets to Sera from Healthy and Diabetic Individuals-In the absence of specific antisera raised to the diabetesassociated M, 64,000 protein antigen, our characterization of this protein has been dependent on the use of sera from diabetic patients. During previous studies, we have been able to identify patients having antibodies reacting strongly to the M, 64,000 protein. Sera from five of such patients, together with sera from five healthy individuals, were used to immunoprecipitate proteins in Triton X-lOO-solubilized particulate material from L-[35S]methionine-labeled human and rat islets. The M, 64,000 protein was the major protein from both rat and human islet extracts immunoprecipitated by sera from diabetic patients (Fig. 1). The protein was not immunoprecipitated by the five control sera. The relative intensity of the band corresponding to the protein for each of the five diabetic patients was similar when either rat or human islets were used, indicating that the protein has similar antigenic properties in the two species. The three sera reacting most strongly Homogenates of [35S]methionine-labeled human or neonatal rat islets were centrifugaied-at 36,000 X g and the pelleted oarticulate material extracted in Triton X-100. Detergent-solubilized proteins were precleared with protein A-Sepharose, and the extract from 100 islets was incubated with each of five control sera and five sera from diabetic patients. Immune complexes were isolated on protein A-Sepharose, eluted, and run on 10% sodium dodecyl sulfatepolyacrylamide gels under reducing conditions. The immunoprecipitated islet proteins were visualized by autoradiography. The figure shows proteins immunoprecipitated from human islets (top panel) or rat islets (lower panel) by the five control or five sera from diabetic patients. The arrow marks the position of the M, 64,000 protein.
to the protein (diabetic patients 1,3, and 5 in Fig. 1) together with sera from control individuals were used in subsequent analyses.

Zmmunoprecipitation of Proteins from Rat Zslet Extracts Prepared by Temperature-induced
Phase Separation in Triton X-114-Immunoprecipitation of the M, 64,000 protein from rat islet extracts was consistently associated with a higher background of nonspecifically bound protein than when human islets were used as source of antigen. In an earlier study employing human islets, we showed that the background proteins can be removed by purification of islet amphiphilic proteins by detergent phase partitioning in Triton X-114 prior to immunoprecipitation (10). A modification of the phase separation procedure was developed to facilitate detection of the protein in rat tissues (11). By this method, particulate material from [35S]methionine-labeled rat islet homogenates was sedimented by ultracentrifugation, extracted in 2% Triton X-114, and subjected to the phase separation procedure of Bordier (16). Proteins in the supernatant after centrifugation (soluble fraction) and in the detergent phase and aqueous phase after detergent phase partitioning in Triton X-114 were immunoprecipitated by sera from the diabetic or control individuals. The M, 64,000 protein was readily detected in the detergent phase fraction, and immunoprecipitates were rela-tively free of nonspecifically bound protein (Fig. 2). Lesser quantities of the M, 64,000 protein were present in the soluble and aqueous phase fractions, but detection was hampered by nonspecific binding.
These results show that the rat and human islet M, 64,000 proteins share amphiphilic properties. Subcellular Localization of the M, 64,000 Protein-The subcellular distribution of the M, 64,000 protein was analyzed by fractionation of islets by Percoll density gradient centrifugation prior to immunoprecipitation.
The labeling of islets with [""Slmethionine was followed by a l-h chase with unlabeled methionine to allow labeled protein to mature. Islet particulate material, depleted of nuclei and cell debris, was found to sediment as three distinct bands on the gradient. Analysis of fractions from the gradient indicated that a band of particulate material sedimenting at a density of 1.04 g/ml was enriched in the plasma membrane marker, 5'-nucleotidase, whereas a broad band at density 1.07 g/ml contained cytochrome oxidase and NADH-cytochrome c reductase, markers of mitochondria and endoplasmic reticulum, respectively (Fig.  3A). Insulin was recovered at the top of the gradient, probably released from damaged granules, and in a fraction of density 1.10 g/ml (Fig. 3B) sera from a control or diabetic individual. The M, 64,000 protein was found to be enriched in the band of material containing predominantly 5'-nucleotidase activity, consistent with a plasma membrane localization of the protein (Fig. 3B). Expression of the M, 64,000 Protein in Islet Cell Populations-To determine whether expression of the M, 64,000 protein is restricted to particular cell types within the pancreatic islet, islet cell populations enriched in /3 cells or nonfl cells were prepared by sorting on the basis of light scatter activity and flavin adenine dinucleotide autofluorescence (15). Analysis by electron microscopy indicated that the /I cell population comprised >95% /3 cells, whereas the non-@ cell population contained ~5% /3 cells, consistent with previous analyses (15). The cell populations were reaggregated, cultured for 18 h, and then labeled with L-[35S]methionine. The mean incorporation of label was 9.8 x lo7 and 6.7 X lo7 cpm/ lo6 cells for the p and non-p cell populations, respectively. Detergent phase purified amphiphilic proteins representing 5 x lo5 cells of each population (two experiments) or 3 X lo6 cpm (two experiments) were immunoprecipitated by control sera or sera reactive to the M, 64,000 protein.
The latter sera were found to immunoprecipitate a heavy band corresponding to the n/l, 64,000 protein in the p cell-enriched population (Fig. 4). In contrast, only a weak band was detected by the M, 64,000 protein-reactive sera in the /3 cell-depleted fraction (Fig. 4).
Expression of the M, 64,000 Protein in Adult Rat Tissues-Organs from normal adult rats were tested for tissue-specific expression of the M, 64,000 protein. Isolated islets and cells from liver, kidney, thyroid, adrenal, pituitary, spleen, and thymus were prepared by mechanical or enzymatic disruption. After a l-h period of recovery in tissue culture, the tissue fragments were labeled by incubation for 5 h with L-[~'S] methionine, and Triton X-114 phase purified amphiphilic proteins were prepared from each tissue. The mean incorporation of label into detergent phase purified proteins from each tissue was 1.2 X 10' (islets), 4 x lo7 (liver), 2 X lo7 (kidney), 2.3 x lo7 (thyroid), 2.4 x lo7 (adrenal), 3 X lo7 (pituitary), 1.2 X lOa (spleen) and 1.3 X 10' (thymus) cpm/ 100 ~1 of packed cells, respectively (n = 2). Portions of each tissue extract representing 3 x lo6 cpm of detergent phase purified proteins were used for immunoprecipitation by control sera or sera reactive to the M, 64,000 protein. The M, 64,000 protein was readily detected in the adult rat islets, but the diabetic serum failed to immunoprecipitate specifically the M, 64,000 protein from the other seven tissues tested (Fig.  5). Similar results were obtained with the sera from the other two diabetic patients.
The failure to detect the M, 64,000 protein in non-islet tissue could be attributable to degradation of the protein in Populations of p and non-0 cells were prepared from dissociated adult rat islets by autofluorescenceactivated cell sorting and labeled with [%]methionine as described under "Materials and Methods." Detergent phase purified proteins from 5 X lo6 cells of each population were immunoprecipitated with serum from a control (A; control 4 in Fig. 1) or a diabetic (B; patient 5 in Fig. 1) individual, separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, and visualized by autoradiography. A typical autoradiogram is shown in the figure. The arrow marks the position of the M, 64,000 protein.
these tissues during processing for immunoprecipitation or to a slower turnover of the protein in these tissues resulting in a low specific activity of the protein after radiolabeling.
Aliquots of ["'Slmethionine-labeled islet homogenates were therefore mixed with homogenates of unlabeled tissues (20fold excess by cell volume) prior to processing for immunoprecipitation. In these experiments, a reduction in the intensity of the radiolabeled, islet-derived M, 64,000 protein on autoradiograms would provide evidence of degradation by the tissue extract or of competition for antibody by unlabeled antigen. The addition of unlabeled homogenates from each of the non-islet tissues tested previously did reduce the intensity of a number of bands on the autoradiograms, for example the M, 35,000, 37,000 and 45,000 proteins by most tissues or the M, 67,000 protein by kidney. In contrast, the recovery of the M, 64,000 protein was not affected (Fig. 6).

DISCUSSION
Most studies to date on the diabetes-associated M, 64,000 islet cells antigen have involved immunoprecipitation of the protein from detergent extracts of [35S]methionine-labeled human or rat islets by sera from insulin-dependent diabetic patients. For the present study, we have selected sera from diabetic patients known to react strongly to the M, 64,000 islet cell antigen, and, in view of the paucity of viable human tissue for this analysis, tissues employed were of rat origin.
Our studies indicated that the protein from rat islets had an immunoreactivity similar to that of human origin since the relative intensity of the protein immunoprecipitated by sera from individual diabetic patients was similar for both rat and human islet extracts. However, the use of rat islet material was associated with higher nonspecific binding of labeled protein which could hamper detection of the M, 64,000 protein. To facilitate unambiguous detection of the rat protein, advantage was taken of its known amphiphilic properties (10) to include a preliminary purification step of phase separation in Triton X-114. When rat islet particulate material was subjected to this procedure, the diabetes-associated antigen could be readily detected in the detergent phase, relatively free of nonspecifically bound protein (Fig. 2). Proteins of similar M, but immunoprecipitated by both control and diabetic sera and distinct from the M, 64,000 antigen by twodimensional electrophoresis (22)  the islet homogenate (Fig. 2). A portion of the M, 64,000 protein was recovered in the aqueous phase; very little was found in the soluble fraction. The partitioning of proteins during the phase separation procedure is a function of the hydrophobicity of the protein (23). Depending on the extent of the hydrophobic domain, residual detergent in the aqueous phase may be sufficient to retain a significant proportion of the protein in this fraction. Nevertheless, the yield of the M, 64,000 protein in the detergent phase was sufficient for phase separation to be employed as a purification step in the analysis of tissue specificity and subcellular localization. In many diabetic patients, islet cell surface antibodies have been shown to react specifically with the islet p cells (7). The amphiphilic nature of the M, 64,000 protein seen in this and earlier studies (10, 11) implies a membrane location, but the subcellular compartment in which the protein is predominantly expressed is not known. The failure to detect the M, 64,000 protein in immunoprecipitates of human islets surface labeled with iz51 was taken as evidence against a cell surface location of the protein (24). However, other factors such as the absence of exposed tyrosine or histidine residues, loss of the antibody-binding sites, scarcity of the protein, and failure to subject the iodinated cell preparation to detergent phase separation may explain the negative data. Although potential cell surface proteins are unlikely to be associated only with the plasma membrane but are likely to be also found in intracellular membranes, certainly during biogenesis, we tested whether immunoreactive M, 64,000 protein would cosediment with marker enzymes for different subcellular organelles. After a 7-h labeling period and a l-h cold chase, radiolabeled M, 64,000 protein was recovered in fractions containing predominantly 5'-nucleotidase activity and depleted in markers of other subcellular organelles. Although a shortage of labeled islet material precluded a more extensive analysis of the purity of gradient fractions, our data are consistent with a plasma membrane localization of the M, 64,000 protein. The results call for further investigations to determine whether the protein is located on the outer leaflet of the plasma membrane and may thus be a potential protein as antigen, to define the role of the protein in the pathogenesis of insulin-dependent diabetes. 12.