Targeted Disruption of the IA-2β Gene Causes Glucose Intolerance and Impairs Insulin Secretion but Does Not Prevent the Development of Diabetes in NOD Mice

Various restriction fragments isolated from mouse IA-2β cDNA clones (8,14) were used to screen a mouse 129–genomic cosmid library in vector pTCF (a gift of Dr. F. Grosveld) using conditions described previously (5). P1 genomic clones were identified in a gridded mouse 129 genomic library (Leiden Genome Technology Center, Wassenaarseweg, the Netherlands) by PCR. Positive clones were isolated, and appropriate fragments were subcloned and used for automated dideoxy sequence determination. In addition, the basic local alignment search tool (BLAST) program (17,18) was used to identify exon-containing genomic segments in public and commercial databases. Mouse strain C57BL/6 genomic sequences were determined through PCR amplification and sequencing of the relevant fragments. Resulting sequences were analyzed using Vector NTI software (Informax, Oxford, U.K.). Exon-intron junctions were determined by comparison with cDNA sequences, complying maximally to the “cagGT” rule for 3′ splice acceptor sites. Intron sizes were estimated from the sequence data, assuming that ambiguous stretches (n repeats) in the genomic entries were deposited proportionally to their length. Genomic sequences were compared lastly to version 14.30.1 of the mouse genome database (released 6 May 2003; http://www.ensembl.org), and the sequence contig annotated for mouse C57BL/6 Ptprn2 (resulting from the current work) was submitted to the database (accession no. AJ583055).

The IA-2β targeting construct was prepared by ligating genomic IA-2β fragments (2.4 and 6.7 kb) to the 5′ and 3′ ends of the vector, respectively. Homologous recombination resulted from the replacement of a portion of the 5′ upstream region, exon 1, and part of intron 1 of the IA-2β gene with a neomycin-resistant gene cassette into the EcoRI-KpnI site of the targeted locus. Jl ES cells were transfected with the linearized targeting vector by electroporation and grown under double-selection conditions (350 μg/ml G418, 2 μmol/l gancyclovir). G418/gancyclovir-resistant ES cell clones were screened by Southern blot analysis. ES cells carrying the disrupted IA-2β allele were then injected into blastocysts to prepare chimeric mice. Chimeric mice were bred to C57BL/6Nci mice, and offspring that carried the disrupted IA-2β allele were backcrossed to C57BL/6Nci for five generations. Offspring also were backcrossed to NOD/LtJ mice (The Jackson Laboratories, Bar Harbor, ME) for four generations, and littermates were used. All protocols were approved by our institutional animal care and use committee.

Total RNA and proteins from brains were prepared and analyzed using specific probes or antibodies. Antibodies against mouse IA-2β (pol.IA-2β) were raised in rabbits by immunization with the recombinant intracellular domain of mouse IA-2β and absorbed with brain lysate from IA-2β^−/− mice.

Blood glucose was measured with a One Touch II Glucometer (Lifescan, Milpitas, CA). Serum insulin levels were determined with an enzyme-linked immunosorbent assay kit using a mouse insulin standard (Crystal Chem, Downers Grove, IL). For glucagon, blood samples were collected in tubes containing aprotinin, and the serum was stored at −80°C until radioimmunoassay (Linco Research, St. Charles, MO).

All of the major organs and tissues were collected in 10% neutral-buffer formalin and processed for paraffin embedding. Sections were stained with hematoxylin and eosin. Pancreatic sections were incubated with antibodies to insulin, glucagon, somatostatin, and pancreatic polypeptide (Dako, Carpinteria, CA), followed by biotin-conjugated second antibody and streptavidin horseradish peroxidase.

Mice (3 months and 6–10 months of age) were fasted overnight, followed by intraperitoneal glucose injection (2 g/kg body wt). Blood was drawn at 0, 15, 30, 60, and 120 min after the injection, and blood glucose was measured using the One Touch II Glucometer. Insulin secretion was measured by injecting glucose (3 g/kg body wt i.p.) into male and female mice at 3 months and 6–10 months of age. Venous blood was collected at 0, 2, 5, and 20 min in heparinized tubes. After centrifugation, plasma was stored at −20°C. Insulin levels were measured with an enzyme immunoassay kit using a mouse insulin standard (Alpco Diagnostics, Windham, NH).

Pancreatic islets from IA-2β^+/+ or IA-2β^−/− mice were isolated using percoll density gradient separation and hand picked under a stereomicroscope (7). Insulin secretion was measured by placing 10 islets in each well (4–10 wells/group). Islets then were incubated in 1.0 ml of Krebs-Ringer bicarbonate buffer supplemented with 3.3 mmol/l glucose for 1 h. Subsequently, the islets were transferred to media containing 27.7 mmol/l glucose and incubated for 1 h. Aliquots were removed and stored at −20°C for quantitation of insulin by enzyme-linked immunosorbent assay.

Statistical analysis was performed using the Student’s t test for unpaired comparisons. Values are presented as means ± SE. P < 0.05 was considered significant.

To accelerate the development of diabetes, 8-week-old IA-2β^+/+ and IA-2β^−/− (male and female) mice received one intraperitoneal injection of cyclophosphamide (250 mg/kg) (Sigma, St. Louis, MO). Two weeks later, mice remaining nondiabetic were given a second injection of cyclophosphamide and followed for an additional 3 weeks. Blood glucose levels were measured with the Elite XL Blood Glucose Meter (Bayer, West Haven, CT). Diabetes was defined as glucose levels of >250 mg/dl on two consecutive readings.

To determine the structure of the mouse IA-2β gene, which has been physically mapped to chromosome region 12F (14,15), we assembled information from the mouse genomic database and used mouse IA-2β cDNA sequences to screen genomic DNA libraries. The IA-2β gene (Ptprn2), similar to the closely related mouse IA-2 gene (Ptprn), consists of 23 exons (Fig. 1A and Table 1) (19), but the distance between the exons of the two genes differs enormously. Mouse Ptprn2 spans ∼800 kb, whereas Ptprn covers only 20 kb. The Ptprn2 introns vary in size from 751 bp (intron 18) to 146.7 kb (intron 2). We also determined the structure of the human IA-2β gene (PTPRN2), located on chromosome 7q36, and found that it too consists of 23 exons distributed over some 1,000 kb (Table 1).

The first mouse IA-2β exon harbors the AUG translation start codon, followed by 106 additional protein-coding nucleotides and a 5′ untranslated region of at least 145 nucleotides. We chose this region of the mouse IA-2β gene as the target for disruption by homologous recombination. In brief, a targeting construct was prepared by replacing a portion of the 5′ upstream promoter region, exon 1, and a portion of intron 1 with a neomycin cassette (Fig. 1B). ES cells then were transfected with the targeting construct, and successful homologous recombination was determined by Southern blot analysis (data not shown). IA-2β^+/− ES cells were then injected into blastocysts, resulting in chimeric mice that transmitted the disrupted IA-2β allele to their offspring. Wild-type (IA-2β^+/+), heterozygous (IA-2β^+/−), and homozygous (IA-2β^−/−) mice were identified by tail DNA PCR with appropriate primers (data not shown) and confirmed by Southern blot analysis using the 1.1-kb EcoRI-BamHI fragment (outside probe) (Fig. 1C). Further evidence that the homologous recombination was successful and that the IA-2β gene had been disrupted was obtained by Northern and Western blot analysis. Hybridization of total brain RNA with probes corresponding to the extracellular region (nucleotides 141-1540) and 3′ noncoding region (nucleotides 3040–3143) of IA-2β resulted in a strong 5.5-kb band with IA-2β^+/+ RNA, a markedly reduced signal with IA-2β^+/− RNA, and no signal with IA-2β^−/− RNA (Fig. 1D and E). Western blot analysis, using rabbit anti-mouse IA-2β antibody, recognized three strong bands (55–75 kDa) in brain lysates of IA-2β^+/+ mice that are weaker in IA-2β^+/− lysates and absent in IA-2β^−/− lysates (Fig. 1F). To see if the deletion of IA-2β had any effect on the level of IA-2, we analyzed brain homogenates from IA-2β^−/− and IA-2β^+/+ mice by Northern and Western blotting. Little if any change in IA-2 mRNA (data not shown) or protein was found in IA-2β^−/− mice (Fig. 1G).

Physical examination of the IA-2β^−/− mice revealed no gross abnormalities. Body weight of the IA-2β^−/− mice did not differ from wild-type controls at 8 weeks of age (IA-2β^+/+ males, 24.0 ± 0.7 g; IA-2β^−/− males, 24.1 ± 0.3 g; IA-2β^+/+ females, 17.5 ± 0.3 g; and IA-2β^−/− females, 18.5 ± 0.8 g). Intercrossing the IA-2β^−/− mice yielded the expected litter size (average 8.6 pups, 14 pairs). Nonfasting blood glucose levels of the IA-2β^−/− male and female mice, measured at 3 and 10 months of age, were slightly elevated as compared with IA-2β^+/+ mice, but did not fall within the diabetic range (Fig. 2A and B). Fasting blood glucose levels also fell within the normal range (Fig. 2C and D). Similarly, nonfasting and fasting insulin (Fig. 2E and F) and nonfasting and fasting glucagon (Fig. 2G and H) levels of IA-2β^−/− male and female mice did not differ significantly from those of IA-2β^+/+ male and female mice.

Histological studies failed to reveal any abnormality in the neuronal and neuroendocrine cells of the IA-2β^−/− mice, including hippocampus (Fig. 3A), anterior pituitary (Fig. 3B), adrenal gland (Fig. 3C), and pancreatic islets (Fig. 3D). Immunohistochemical studies on the IA-2β^−/− mice (Fig. 3E–H) also revealed normal-appearing pancreatic islets, with no difference in the morphology or staining pattern for insulin, glucagon, somatostatin, or pancreatic polypeptide as compared with IA-2β^+/+ mice (data not shown).

In contrast to the fasting and nonfasting blood glucose and insulin levels, an intraperitoneal glucose tolerance test revealed elevated glucose levels and depressed release of insulin in IA-2β^−/− mice. Blood glucose levels were significantly elevated in IA-2β^−/− female mice at both 3 months (Fig. 4B) and 6–10 months (Fig. 4D) of age, but not in male IA-2β^−/− mice (Fig. 4A and C). Insulin levels were mildly depressed (Fig. 4E–H), but only reached statistical significance at 20 min in the female mice (Fig. 4F and H). Statistically significant differences, however, were found when cultured pancreatic islets isolated from IA-2β^+/+ and IA-2β^−/− mice were stimulated with glucose. Male and female IA-2β^−/− mice released 33 and 38% less insulin, respectively, than IA-2β^+/+ mice when switched from basal glucose (3.3 mmol/l) to high glucose (27.7 mmol/l) (Fig. 5).

Chimeric mice were bred to C57BL/6Nci mice, and the offspring that carried the disrupted IA-2β allele were backcrossed to NOD/LtJ mice for four generations. IA-2β^+/− NOD mice then were intercrossed to obtain IA-2β^−/− NOD mice as well as IA-2β^+/+ NOD mice. Heterozygous crosses tend to level any confounding genetic background effect that might still be present from 129-derived sequences, even though the most likely effect of the residual 129 genes would be protection from diabetes. In addition to genetic typing (data not shown), the lack of IA-2β protein expression was confirmed by Western blot analysis (Fig. 6A). Mice then were treated with 250 mg/kg i.p. of cyclophosphamide, a known accelerator of diabetes (20). Approximately 57% of IA-2β^+/+ male, 62% of IA-2β^−/− male, 58% of IA-2β^+/+ female, and 53% of IA-2β^−/− female NOD mice developed diabetes (Fig. 6B). Although the colony is still small, five IA-2β^−/− NOD mice that were not given cyclophosphamide spontaneously developed diabetes between 16 and 36 weeks of age. Although not yet tested, since the knockout of IA-2β results in impaired secretion of insulin, the IA-2β^−/− NOD mice might develop even more severe or accelerated diabetes than the IA-2β^+/+ NOD mice because of the superimposition of impaired insulin secretion on the NOD diabetes background.

Recently, we showed (7) that targeted disruption of IA-2 resulted in elevated glucose tolerance tests and impaired secretion of insulin. The targeted disruption of IA-2β, reported here, also showed mild impairment of glucose tolerance tests and insulin secretion, resembling our findings in the IA-2 knockout mice and arguing that the two proteins have similar functions but act independently. This, in part, is supported by the fact that the knockout of IA-2 did not alter the expression of IA-2β mRNA (7), and, as shown here, the absence of IA-2β had little effect on the expression of IA-2. The possibility that IA-2β and IA-2 act in a complementary fashion is now being explored with double-knockout mice, and preliminary results show that glucose tolerance test abnormalities are considerably more pronounced in the double-knockout than in the single-knockout mice.

Precisely how IA-2 and IA-2β exert their effect on insulin secretion is not known. Recent protein interaction experiments suggest two alternative models. In the first, IA-2, as an intrinsic membrane protein of secretory vesicles, may be linked to cytoskeletal proteins and thereby aid in regulating secretory vesicle mobilization and recruitment in exocytosis (21). In the second, IA-2 and IA-2β may heterodimerize with receptor-type PTPs and consequently influence the signaling pathways involved in regulating neuropeptide or peptide hormone exocytosis (22).

The partial impairment of the insulin secretory process in both IA-2β–and IA-2–deficient animals may reflect functional redundancy. Indeed, IA-2β and IA-2 are almost 75% identical in their intracellular PTP-like segment, even though the underlying gene duplication event preceded vertebrate radiation (23). Although the amino acid sequence identity in the NH₂-terminal luminal segment is only 26%, the overall exon-intron segmentation of the two genes is well conserved. In sharp contrast, the gene sizes differ by no less than 40-fold in mice and by as much as 50-fold in humans. This is primarily due to the presence of five introns in IA-2β that equal or exceed 100 kb. At least six potential coding regions for still-unknown genes are annotated within the 1,000 kb of the human IA-2β gene. To date, reports (24, 25) on intron size divergence mostly relate to interspecies comparisons between orthologous genes, but the mammalian talin and phospholipase D genes represent comparable cases of unusual size differences (of 6- and 20-fold, respectively) between paralogous genes. The underlying reason or mechanism remains unknown, although the impact of repeat sequences on intron length is well documented (26). Furthermore, analysis of the human genome has shown that the average intron size is inversely related to the G-C content of the genomic area (27), and indeed the 40-fold expansion of the IA-2β–encoding gene as compared with that of IA-2 is paralleled by a drop in G-C content from 51 to 45%. Whether the remarkable size difference merely reflects an epiphenomenon, a coevolution of the IA-2β and IA-2 genes (with their distinct genomic surroundings), or the acquisition of unique discriminatory differences at the regulatory level is not known. It is of interest in this context that differences in the gene size of the IA-2 and IA-2β orthologs are also apparent in the Japanese pufferfish (Fugu rubripes) (7 and 63 kb, respectively), which has a genome one-eighth the size of the human genome and diverged from the human lineage about 450 million years ago (28).

Since IA-2β is a major autoantigen in type 1 diabetes and the NOD mouse is the most widely used animal model for human type 1 diabetes (29), the development of IA-2β knockout mice gave us the opportunity to determine whether the absence of IA-2β would influence the development of diabetes in NOD mice. Although none of the known NOD susceptibility loci (i.e., Idd1–Idd21) are found on chromosome 12 (i.e., the location of IA-2β), our knockout experiments clearly show that IA-2β is not required for the development of cyclophosphamide-induced diabetes in these mice. Recent knockout experiments (30) showed that IA-2 is also not required for the development of diabetes in NOD mice. Moreover, by targeted gene disruption, GAD65, another major autoantigen in human type 1 diabetes, has been shown (31) not to be required for diabetes in NOD mice. In humans, the autoantibodies to IA-2, IA-2β, and GAD65 are important diagnostic and predictive markers for type 1 diabetes (32), but not all individuals with type 1 diabetes have these autoantibodies. At least in the case of the NOD mouse, it now appears that neither these proteins nor the humoral immune responses to them (33) are required for the development of diabetes. Taken together, these findings suggest that at the human level the autoimmune response to IA-2, IA-2β, and GAD65 may be a consequence of and/or a contributor to type 1 diabetes rather than being required for the development of the disease.

Part of this work was financially supported by the European Community Research Fund.

We thank Jan Schepens for excellent technical assistance. We thank the Dutch Life Sciences Organization for providing access to the P1 library facility at the Leiden Genome Technology Center, the Netherlands. S.G. and W.H. are members of the EU-TMR (European Union Training and Mobility of Researchers) Network CT1997-00142.