A Reg Family Protein Is Overexpressed in Islets From a Patient With New-Onset Type 1 Diabetes and Acts as T-Cell Autoantigen in NOD Mice

RNA from the islets of a patient with type 1 diabetes who died soon after disease onset (10) was generously provided by Dr. Massimo Trucco (Department of Pathology, University of Pittsburgh, PA). Normal human islets were obtained from the islet isolation core facility at the Washington University School of Medicine (St. Louis, MO). Total RNA was isolated from these islets using guanidinium thiocyanate and a cesium chloride gradient. The first-strand cDNA was synthesized using oligo dT primer T (primer sequences are listed below), with a 5′ tail containing restriction endonuclease sites. The cDNA amplicon was synthesized by one cycle of PCR with a 30°C annealing temperature using primer X, followed by 40 cycles of standard PCR amplification with primers XN and TN, corresponding to the 5′ end of primers X and N.

Subtractive hybridization was performed with driver amplicons from cDNA derived from islets of nondiabetic subjects and with tester amplicons from cDNA derived from the islets of the diabetic patient. The driver amplicon was biotinylated by direct incorporation of Biotin-16-dCTP in each PCR. After amplification, primers XN and TN were removed from driver strands by EcoRI/HindIII double-digest. Driver amplicon was mixed with tester amplicon at a ratio of 15:1, the mixture was heated and coprecipitated, and the pellet was redissolved in 10 μl of hybridization buffer (0.5 mmol/l EDTA, 0.5 mol/l sodium chloride, and 50 mmol/l HEPES). The hybridization mixture was covered by mineral oil, heat-denatured, and hybridized at 65°C for 72 h (1st round) or 36 h (2nd and 3rd rounds). After hybridization, strands were removed using streptavidin followed by phenol extraction. The extraction was repeated three more times and then the aqueous material was ethanol-precipitated and dissolved in 100 μl water. The subtracted material was PCR-amplified using primers XN and TN. The amplified subtracted material enriched in differentially expressed sequences served as tester for the next round of subtraction. The subtraction procedure was repeated three times.

T: 5′ GAT ATC CGG CCG GAA GCT TTT TTT TTT TT 3′

TN: 5′ GAT ATC CGG CCG GA 3′

X: 5′ GCC GGA TCC CTG CAG AAT TCN NNN NN 3′

XN: 5′ GCC GGA TCC ATG CAG 3′

Two million λgt 11 phage plaques of a human pancreatic islet library (generously provided by Dr. Alan Permutt, Washington University) were plated on NCZYM-agar plates (density 30,000–40,000 per plate) and lifted onto Nytran Plus membranes. Filters were probed directly with the PCR products from the 3rd subtraction. Radioactive probes were generated by PCR amplification of the subtracted material using ³²P-dCTP (Amersham, Piscataway, NJ). Filters were hybridized for 24 h in 100 ml hybridization solution at 65°C. Filters were then washed in 2% sodium chloride-sodium citrate (SSC) and 0.1% SDS, followed by a second wash in 0.1% SSC and 1% SDS, both at 65°C. Positive plaques were picked after the 3rd screening. Phage inserts were amplified with λgt 11 primers (Promega, Madison, WI). The PCR products of phage inserts were cut from low-melt agarose gels and used as probes to hybridize to the original tester and driver amplicons. Inserts confirmed in this way to be overexpressed in diabetic islets were PCR-amplified using λgt 11 primers. PCR products were purified from agarose gels and submitted for sequencing.

Human islets were incubated in serum-free CMRL medium (Gibco, Rockville, MD) supplemented with insulin (10 μg/ml), transferrin (5.5 μg/ml), and sodium selenite (6.7 ng/ml) contained in a premixed formulation (ITS supplements; Gibco) and increasing amounts of IL-6 (PharMingen, San Diego, CA) for 24, 48, and 96 h. The supernatant was collected, dialyzed overnight against 50 mmol/l ammonium bicarbonate, and lyophilized. Separation was performed in a 12% polyacrylamide gel. Proteins were electroblotted onto a nitrocellulose membrane (Schleicher-Schuell, Dassel, Germany). The membrane was blocked with 1.5% nonfat milk in PBS with 0.05% Tween 20 (PBS-T) and incubated overnight with an anti-HIP/PAP antibody in PBS-T, supplemented with 1% FCS and 0.02% sodium azide. After incubation with a horseradish peroxidase (HRP)-labeled anti-rabbit antibody (Sigma, St. Louis, MO) for 1 h, the blot was developed with a chemiluminescent substrate for HRP (Pierce, Rockford, IL).

The cloning of HIP/PAP cDNA into an expression vector containing the His-tag has been described elsewhere (24). The recombinant protein contained a 6 Histidine tag at the NH₂-terminal, followed by the putative signal sequence, the putative propeptide, and the COOH-terminal carbohydrate recognition domain (CRD) of HIP/PAP. Escherichia coli bacteria expressing HIP/PAP were lysed, and the cleared lysate was loaded onto a metal chelate affinity matrix (Quiagen, Valencia, CA). Nontagged proteins were removed by extensive washing, and the tagged recombinant HIP was eluted by stepwise reduction of the pH to 3. Fractions were collected and analyzed by SDS-PAGE. The fractions, which contained HIP with negligible contaminating bands, were pooled and dialyzed against 50 mmol/l sodium phosphate buffer at pH 7.5, with stepwise reduction of the urea content.

HIP was removed from the preparation by adding metal chelating groups coupled to magnetic beads (Quiagen). After incubation with gentle agitation for 1 h at room temperature, the beads and the bound HIP were removed by exposing the mixture to a magnetic field. The remaining HIP-depleted preparation was used as negative control in T-cell proliferation assays.

NOD, NOD-SCID, NOD CIITA knockout, and C57BL/6 mice were obtained from the Yale NOD Mouse Core and the Yale Animal Core Facility. They were kept under specific pathogen-free conditions in filter cages on a 12:12-h day/night cycle and fed with autoclaved standard chow.

Spleens from NOD and C57BL/6 mice were homogenized, and erythrocytes were lysed. Cells were resuspended in Bruff’s medium (Gibco) supplemented with 3% FCS (Bruff’s/FCS) at 2.5 × 10⁶ cells/ml. HIP was titrated in Bruff’s/FCS in 96-well plates in triplicate. Next, 100 μl spleen cell suspension was added to each well, and the plates were incubated for 72 h at 37°C in an atmosphere containing 5% CO₂. During the last 16 h, tritiated thymidine (Amersham) was added (1 μCi/well). Cells were then harvested, and light emission was measured in a β counter (Wallac, Gaithersburg, MD).

Islets and pancreatic lymph nodes from four 7- to 10-week-old nondiabetic female NOD mice were collected. The pancreatic lymph node cells and islets were added to the wells of a microtiter plate in Bruff’s/FCS (containing 4 units/ml IL-2 and 5 μg/ml HIP/PAP) and expanded for 5–7 days before they were split. Then, 5 weeks later, the T-cells were pooled and restimulated with irradiated (3,000 rads) NOD spleen cells (2.5 × 10⁶/ml) and HIP/PAP (5 μg/ml). T-cells were restimulated every 3–4 weeks and maintained in Buff’s/FCS plus IL-2 between restimulations.

To characterize the T-cell receptor usage and the surface phenotype of the generated line, denoted WY1, cells were stained with the following antibodies: anti-CD8-flourescein isothiocyanate (FITC); anti-CD4-phycoerythrin (PE); and anti-Vβ2, -3, -4, -6, -7, -8, -10, -11, -12, and -13, all FITC-labeled (PharMingen). Cells were then analyzed in a fluorescence-activated cell sorter (FACS; Beckton Dickinson, Franklin Lakes, NJ).

A single-cell suspension of splenocytes from recently diabetic NOD mice was incubated for 30 min on ice with anti-CD4 monoclonal antibody (GK1.5, rat IgG). After washing, antibodies against mouse IgG, IgM, and rat IgG, coupled to magnetic beads (PerSeptive Biosystems, Farmingham, MA), were added. After gentle agitation on ice, the cell suspension was exposed to a magnetic field, displacing cells toward the wall of the incubation vessel. An aliquot of the purified cells was retained for FACS analysis. Cells were stained with anti-CD3-PE, anti-CD8-Cy-chrome, and anti-B220-FITC (PharMingen). Altogether, >95% of the B220⁺ cells and >90% of the CD3⁺/CD8⁻ cells could be removed.

In addition, an insulin-specific diabetogenic Vβ6⁺, CD8⁺ T-cell clone, denoted 6426 (unpublished), was also used in transfer experiments.

Mice were anesthetized by inhalation of methoxyflurane (Shering-Plough, Kenilworth, NJ), and T-cells were injected intravenously. Animals received a maximum volume of 200 μl per injection. Glucosuria was monitored with dipsticks twice weekly, and the presence of diabetes was confirmed by blood glucose testing (>250 mg/dl). The following amounts of cells per animal were transferred: 10⁷ cells of WY1 alone or 10⁷ purified CD8⁺ cells alone. The following were coinjected: 1) 0.7 × 10⁷ purified CD8⁺ cells with 0.7 × 10⁷ cells of WY1, and 2) 0.15 × 10⁷ cells of clone 6426 with 0.3 × 10⁷ cells of WY1.

Pancreata destined for paraffin-embedding were fixed in 10% neutral buffered formaline solution (Sigma). Pancreatic Reg expression in NOD mice was studied with two different affinity-purified rabbit-derived polyclonal anti-Reg primary antibodies. One was produced against a fusion protein of the CRD of HIP/PAP (HIP-CRD) with glutathion-S-transferase (24), and the other was produced against recombinant rat PAP1, the homologue to mouse RegIIIβ. Guinea pig anti-insulin antibodies, HRP-labeled anti-guinea pig antibodies, and HRP-labeled anti-rabbit antibodies were purchased (Zymed, Sigma). Sections were incubated with primary antibody for 1 h at 37°C and with secondary antibody for 45 min at 37°C. The HRP substrate used for color development was 3-amino-9-ethylcarbazol (AEC) (Sigma). Sections were incubated for 5–10 min in substrate solution, washed in distilled water, counterstained with Mayer’s hematoxylin, and mounted with a permanent aqueous mountant (Serotec, Kidlington, U.K.). Negative controls were secondary antibody alone and primary antibody that had been preincubated with immobilized HIP/PAP.

Pancreata destined for cryo-sectioning to detect islet-infiltrating cells were fixed overnight at 4°C in PLP (1% paraformaldehyde, 1.4% l-lysine, and 0.21% sodium m-periodate in 0.1 mol/l sodium phosphate buffer at pH 7.3). To preserve tissue morphology, pancreata were soaked sequentially for 20 min each in a 10% and a 20% sucrose solution in 0.1 mol/l phosphate buffer at pH 7.3 and then embedded in O.C.T. medium (Sakura Finetek, Torrance, CA). Sections of 5-μm thickness were incubated for 1 h at room temperature with biotinylated anti-Vβ8 and -Vβ11 antibodies (PharMingen), the latter serving as a negative control. HRP-labeled Streptavidin (Sigma) was used (45 min incubation), and the substrate for HRP was again AEC. Double-staining was performed by staining first for insulin (see above), followed by staining for Vβ8, with a second layer consisting of alkaline phosphatase-labeled straptavidin (Sigma). Alkaline phosphatase substrate was 5-bromo-4-chloro-3-indolylphosphate/nitro blue tetrazolium (BCIP/NBT) (Sigma).

Islets from a patient with recent-onset type 1 diabetes and normal islets were used to isolate cDNA sequences overexpressed in diabetic islets. An islet cDNA library derived from nondiabetic donors was then screened with these sequences. Positive phage inserts were used to probe Southern blots of amplicons from normal islets and from those of the diabetic patient to confirm overexpression (Fig. 1A). Sequence analysis of clone 23 D1/1 identified its insert as part of the cDNA of HIP/PAP corresponding to most of exon 3 and all of exons 4 and 5 of HIP/PAP.

Increased expression of HIP/PAP in human diabetic islets may have been a consequence of inflammation. Because the promoter region of HIP/PAP contains a putative IL-6 response element, we incubated human islets with increasing doses of IL-6 and tested the supernatant for HIP/PAP protein at 24, 48, and 96 h. As shown in Fig. 1B, IL-6 induced a strong dose-dependent expression/secretion of HIP/PAP that was visible after 24 and 48 h. After 96 h, however, the amount of HIP/PAP produced from islets cultured without exogenous IL-6 was similar to the amount of HIP/PAP produced with the addition of IL-6. Thus, HIP/PAP is released at a basal level from human islets, and exogenous IL-6 increases the rate of HIP/PAP expression.

To determine the expression pattern of Reg in the pancreata of NOD mice, we collected pancreata from 3-, 7-, and 10-week-old nondiabetic mice as well as from diabetic mice, and we stained sections with antibodies against Reg and insulin. Figure 2A–D and F show the expression pattern of Reg in the pancreas, using an antibody produced against the CRD of HIP/PAP. Reg-staining was detected in the islets as well as in the cells of the ductal epithelium (inset of Fig. 2C). There was strong Reg-staining in the islets at all observed ages, including diabetic mice, where most of the islets cells had been replaced by infiltrating lymphocytes (Fig. 2D, inset of 2D). This staining appeared to be colocalized with staining for insulin but did not seem to be restricted uniquely to the β-cells (Fig. 2E and F). The expression pattern appeared unchanged throughout, although the signal in the ductal epithelia within the pancreas tended to become more intense with progression toward overt diabetes. Staining was lost after adsorption of the primary antibody with immobilized recombinant HIP/PAP, confirming the specificity of the signal. A second antibody, which had been produced against rat PAP1, resulted in the same staining pattern, although the overall staining intensity was somewhat weaker (data not shown). These findings provide strong evidence that certain members of the Reg family are expressed in NOD islets and the ductal epithelium. Reg was also detected by Western blot in the supernatant of hand-picked isolated islets from NOD-SCID mice using these anti-HIP/PAP antibodies (data not shown).

Spontaneous splenic T-cell responses against HIP/PAP in NOD mice of different ages are shown in Fig. 3A. Splenocytes from C57BL/6 mice did not respond spontaneously when stimulated with HIP/PAP (data not shown). Because maximal responses were detected after 7 weeks, we used islet-infiltrating lymphocytes and lymphocytes from pancreatic lymph nodes of 7- and 10-week-old female NOD mice to generate a T-cell line specific for HIP/PAP. The T-cell line obtained (WY1) showed a strong dose-dependent proliferative response to HIP/PAP. No response was detected when recombinant HIP/PAP was adsorbed out of the preparation or a control antigen with the same His-tag was used (Fig. 3B). FACS analysis at restimulation round 14 showed this line to be CD4⁺/CD8-/Vβ8⁺. No other Vβ families were detected. The response of WY1 against recombinant HIP/PAP could be blocked by an antibody against I-A^g7, indicating that the response was antigen specific (Fig. 2C). WY1 also responded to islets from NOD mice, and this response could be blocked by anti-I-A^g7 antibodies as well. Although both NOD-SCID and NOD CIITA knockout mice have the same Reg staining pattern in the pancreas as normal NOD mice (data not shown), only islets from NOD-SCID animals, but not those from NOD CIITA knockout mice (which do not express I-A^g7), could stimulate WY1 (Fig. 3C). Therefore, the T-cell line WY1, specific for recombinant HIP/PAP, cross-reacts with one or more of the Reg-proteins present in mouse islets, and this response is I-A^g7-restricted.

WY1 cells were adoptively transferred into NOD-SCID mice with and without CD8⁺ T-cells obtained from splenocytes of diabetic NOD mice. Between 18 and 40 days after the T-cell transfer, 5 of 6 animals receiving WY1 plus purified CD8⁺ T-cells developed diabetes, whereas none receiving WY1 cells alone and only 1 of 6 mice receiving CD8⁺ T-cells alone did so (χ² test, P < 0.0053) (Fig. 4). To investigate the conditions that are needed for WY1 to be able to infiltrate pancreatic islets, cells were adoptively transferred, either alone or in combination with cells of the diabetogenic insulin-specific CD8⁺ T-cell clone 6426, into NOD-SCID mice. When WY1 Vβ8⁺ T-cells were injected alone, Vβ8⁺ cells were found in the pancreatic lymph nodes (data not shown), but they were not detected in pancreatic islets (Fig. 5A). However, when WY1 cells were coinjected with small numbers of clone 6426-expressing Vβ6 cells, islet infiltration with Vβ8⁺ WY1 cells became clearly evident. The Vβ8⁺ cells of line WY1 were localized particularly in the periphery of the infiltrate (Fig. 5B). In addition, CD8⁺ cells of the insulin-specific T-cell clone were detected in the islets as well (not shown). As expected, mice that had received clone 6426 alone showed insulitis but no staining for Vβ8 (Fig. 5C).

Subtractive hybridization was used to hybridize and remove homologous cDNA strands from islets of a diabetic patient by excess amounts of a normal islet cDNA. This approach has been successfully used for tissue-specific cDNA and genomic DNA subtraction (25,26) to isolate cancer-related and differentiation stage-specific genes (27–30) and to clone drug-induced genes or pathogens from infected tissues (31–33). As compared with other low-yield subtractive hybridization techniques, the experimental approach used in this study is applicable to small amounts of rare samples, such as RNA extracted from islets derived from patients with recent onset of diabetes. To identify overexpressed genes, we choose to use the subtracted material to screen a phage cDNA library derived from normal nondiabetic islets. This approach was taken because of the lack of a type 1 diabetes islet library. We showed that HIP/PAP is overexpressed in islets from a patient with recent onset of type 1 diabetes; however, the subtractive hybridization method used was based on multiple rounds of amplification followed by subtraction, thus precluding quantitation. The small amount of sample available and the lack of a reliable internal control did not allow us to quantitate the amount of overexpression by Northern blot, RNase protection assay, or quantitative RT-PCR.

Because our goal was to identify novel autoantigens in type 1 diabetes, we tested whether HIP/PAP, a putative islet regeneration factor, could act as an autoantigen in the NOD mouse and could thus play a role in the pathogenesis of type 1 diabetes.

HIP/PAP belongs to the Reg family of proteins. All members of this family consist of a putative signal peptide linked to a short propeptide, which is followed by a single C-type CRD. These CRDs contain a number of highly conserved residues, among them four cysteines involved in the formation of two disulfide bonds (34–36). HIP/PAP shows lactose-binding activity and binds to laminin and fibronectin (23,24). However, it is not clear whether binding to these extracellular matrix proteins involves carbohydrate-protein or protein-protein interaction.

Several studies have reported different Reg expression patterns in the mouse pancreas. A recent study by Sanchez et al. (37), investigating the expression of RegI and -II in the NOD mouse by using RegII peptide-specific antibodies, could find neither RegI nor -II in the endocrine pancreas, but showed localization of Reg in the acinar cells of the exocrine pancreas. However, colocalization of Reg with insulin-staining and localization in ductal epithelial cells have been reported by Anastasi et al. (21) in a model of streptozotocin-induced diabetes in C57BL/6J mice. This latter finding is in agreement with our observations of Reg expression in the pancreas of the NOD mouse. A potential explanation for the seemingly contradictory observations is that our antibodies, which were produced against the CRD of human HIP/PAP or against rat PAP1, do not recognize RegI or -II, but only certain members the RegIII subfamily. This is likely, because HIP/PAP has a higher degree of amino acid identity to RegIII-α, -β, and -γ than to RegI, -II, or -IIIδ (Table 1). If this explanation is correct, it implies that members of the Reg family are differentially expressed in the cells that constitute the mouse pancreas. The use of gene-specific probes and monoclonal antibodies, which recognize specific members of the Reg family, are needed to determine the expression pattern of individual Reg members at the cellular level. Similarly, should pancreas sections from normal and recently diabetic human donors become available, probes for HIP/PAP can be used to confirm overexpression of this protein.

Our data demonstrate that NOD mice display spontaneous T-cell responses to HIP/PAP. As a result, we were able to generate a T-cell line, WY1, derived from pancreatic islets and lymph nodes, that is specific for human HIP/PAP and can be stimulated by NOD islets. Thus, there are T-cells within the pancreas of NOD mice capable of recognizing one or more of the Reg family members present in mouse islets. Once the T-cell epitope of HIP/PAP recognized by WY1 is determined, corresponding peptides of the individual members of the Reg families can be synthesized to investigate the degree of cross-reactivity WY1 cells show toward them.

WY1 cells can home to the pancreatic islets and can transfer disease when coinjected with CD8⁺ T-cells from diabetic NOD mice. It is possible that WY1 cells require the cytokine/chemokine signals produced by islet-infiltrating CD8⁺ cells in order to leave the pancreatic lymph nodes and migrate to and remain in the islets. Alternatively, islet-infiltrating CD8⁺ cells might cause upregulation of production/secretion of Reg from the islets, which then activates cells of WY1 resting in the pancreatic lymph nodes, causing them to accumulate in the islets. We have shown that IL-6 mediates upregulation of Reg production/secretion from isolated human islets. This might be a mechanism by which the cytokine environment in the islets exerts its influence on Reg expression. If such a mechanism indeed plays a role in vivo, as our data suggest, one could imagine a scenario where during the progressive islet inflammatory process, increased amounts of IL-6 are released either from the infiltrating cells or (when triggered by their cytokines, such as γ-interferon) from the islets themselves (38,39). This would then lead to upregulation of Reg expression; potentially, proliferation of T-cells recognizing the autoantigen HIP/PAP; and perhaps a progressive acceleration of the disease process. That islets respond with upregulation of Reg expression/secretion when injured by inflammation would fit the role the Reg family of proteins has been reported to play. Reg protein has been shown to be mitogenic for β-cell and ductal cell lines and has produced an increase in 3-[H]thymidine incorporation into the nuclei of isolated rat islets (40–42). A high rate of β-cell replication and increased Reg transcription were observed when isolated human fetal β-cells were stimulated to proliferate by hepatocyte growth/scatter factor (43). Moreover, although RegI knockout mice do not show any specific phenotype under normal physiological conditions after treatment with aurothioglucose to induce hyperplastic islet growth, the islet cell volume in knockout mice is significantly smaller than that of control mice, implying a role for Reg in islet growth (18). Overall, these findings are consistent with a mitogenic trophic effect of the lectin Reg. Its function in response to islet inflammation might be to support islet regeneration and to protect the islet from inflammatory damage. If this were the case, overexpression of a putative islet regeneration/mitogenic protein that has the potential to act as an autoantigen could create a vicious cycle, accelerating the immune process leading to diabetes. However, regardless of whether it promotes islet growth, any islet protein, such as HIP/PAP, that is overexpressed in association with inflammation could be expected to increase β-cell loss in those individuals/mice capable of generating an autoimmune response against it.

This work was supported by grants from the National Institutes of Health (DK-53015 and DK-45735) and a gift from Camp NYDA (New York Diabetes Association).

We thank Tamara Prus and Marie-Thérèse Simon for their excellent technical assistance.