Proinsulin 2 Knockout NOD Mice: A Model for Genetic Variation of Insulin Gene Expression in Type 1 Diabetes

Proinsulin 2-deficient 129 mice and NOD mice were bred in our own facilities under specific pathogen-free conditions and allowed free access to food. The prevalence of spontaneous diabetes in our NOD colony reached 75% in female mice by 6 months of age.

Proinsulin 2-deficient 129 mice, in which a proinsulin 2-deficient allele was functionally disrupted by insertion of LacZ and a neomycine resistance (Neo) gene, have been previously reported (11). Female proinsulin 2-deficient 129 mice were crossed with male NOD mice. Female (129/Pro2^−/− × NOD) F1 mice were then backcrossed onto the NOD background for eight generations. Heterozygote carriers of the proinsulin 2-deficient allele were genotyped for the Neo and LacZ genes, using previously described primers (5′-GTG TTC CGG CTG TCA GCG CA-3′ and 5′-GTC CTG ATA GCG GTC CGC CA-3′; and 5′-ACG GCA CGC TGA TTG AAG CA-3′ and 5′-CCA GCG ACC AGA TGA TCA CA-3′, respectively) and were used as breeders at each backcross (BC) generation. Segregants from the second to the fourth BC generation were genotyped for microsatellite markers linked to 16 Idd loci (Idd1 to Idd-16), as previously described (13). At the fourth BC generation (BC4), proinsulin 2^+/− heterozygote mice were intercrossed to generate proinsulin 2^−/− homozygote mice. Proinsulin 2^+/− heterozygote mice and proinsulin 2^−/− homozygote mice were distinguished by amplification of the proinsulin 2 gene, using 5′ and 3′ primers (11), providing a 676-bp fragment only in heterozygote animals.

Detection of mRNA in tissues was performed in 2- and 3-week-old mice. Pancreatic islets of Langerhans were isolated as previously described with slight modifications. RNA were extracted from fresh tissues using guanidium thiocyanate and treated with DNase before reverse transcription to avoid contamination with genomic DNA. Proinsulin 1 and 2 were amplified using two pair of primers: 5′-CCA TCA GCA AGC AGG TTA TTG TTT C-3′ and 5′-CAG CTC CAG TTG TGC CAC TTG TG-3′, which were specific for the proinsulin 2 and provided a 232-bp fragment, and 5′-CCA TCA GCA AGC AGG TCA TTG TTT-3′ and 5′-GCG GGA CTT GGG TGT GTA GAA GAA-3′, which amplified the proinsulin 1 and provided a 193-bp fragment.

NOD backcrossed mice were monitored weekly for glucosuria. Blood glucose was measured when glucosuria was detected. Mice were considered diabetic after two consecutive glycemic values >300 mg/dl. Glycemia was assessed using test strips and a Euroflash glucometer (LifeScan). Plasma insulin was assessed by radioimmunoassay (Linco Research).

Incidence of diabetes in proinsulin 2^+/− and wild-type mice was described using Kaplan-Meier estimates and compared using the log- rank test.

Female (129/Pro2^−/− × NOD) F1 mice were backcrossed onto the NOD background, and BC mice carrying the proinsulin 2-deficient allele were then backcrossed onto the NOD background for eight successive generations. Segregants from the second to the fourth BC were genotyped for 16 NOD microsatellite markers reported as associated with diabetes in the NOD mouse, i.e., Idd1 to Idd-16. After two generations of backcrossing (BC2), four diabetes susceptibility genes were fixed, including Idd1 (I-A^g7). At the fourth BC generation (BC4) and in subsequent BC generations, all proinsulin 2^+/− heterozygotes were homozygous for all 16 NOD Idd alleles.

An increased incidence of diabetes was observed in proinsulin 2^+/− as compared with wild-type mice at each BC generation analyzed (i.e., BC2, BC7, and BC8; P < 0.02, P < 0.001, and P < 0.005, respectively). The prevalence of diabetes at BC2 and BC8 generation is shown in Fig. 1. Noticeably, the incidence of diabetes was low in wild-type BC2 mice, presumably because of insufficient backcrossing and presence of only 87.5% NOD genes. Despite the presence of 12.5% 129 genes at BC2, the presence of the proinsulin 2 null mutation allowed an unexpectedly high incidence of diabetes, reaching 54.0% at 32 weeks of age in female mice as compared with 18.5% in wild-type mice. In further BC generations, the incidence of diabetes increased to reach 52.0% at BC8 in wild-type mice but remained significantly increased in proinsulin 2^+/− mice (85.2%, P < 0.005). Despite the high diabetes incidence observed, female prediabetic proinsulin 2^+/− mice were comparable to their wild-type counterparts for weight, glycemia, and pancreatic mass at 4 and 8 weeks of age. They also showed identical plasma insulin levels at 4 weeks of age, but they had higher plasma insulin levels at 8 weeks of age (Fig. 2).

We took advantage of the possibility to generate proinsulin 2^−/− mice to determine whether proinsulin 1 is expressed in the thymus (Fig. 3). Discrepant results have been reported for tissue distribution of proinsulin 1 expression in the mouse (10,15,16). Proinsulin 2^+/− progeny resulting from BC4 was intercrossed, and thymus was obtained from proinsulin 2^−/− mice. We verified that proinsulin 2 mRNA was not detected in the islets and thymus of proinsulin 2^−/− mice, but it was detected in wild-type and in heterozygous proinsulin 2^+/− mice (not shown). However, by using primers that amplify a 193-bp proinsulin 1 fragment, proinsulin mRNA was detected in the thymus of proinsulin 2^−/− mice as well as in the islets. No proinsulin was detected in proinsulin 2^−/− control tissues.

Insulin exemplifies a key β-cell function through its secretion in response to glucose and is a key autoantigen in the development of type 1 diabetes (15–20). By generating β-cell-deprived NOD mice using alloxan, we previously showed that β-cells were mandatory for the activation of autoreactive lymphocytes against β-cell antigens in type 1 diabetes (21). High levels of plasma insulin in the fasting state and in response to glucose have further been observed at the early stage of development of insulitis in the NOD mouse, suggesting that abnormal β-cell function may play a role in the autoimmune process that leads to diabetes (15). Prevention of diabetes by insulin therapy in prediabetic NOD mice (22) and BB rats has been interpreted as resulting from β-cell rest and as further evidence for the importance of β-cell function in type 1 diabetes. In humans, genetic susceptibility associated with IDDM2 has been considered to link β-cell function to diabetes development (1–4). An alternative hypothesis is that insulin modulates lymphocyte activation by interacting with lymphocyte insulin receptors on activated T-cells or the T-cell receptor on proinsulin-specific lymphocytes (15). Insulin also possibly influences thymic selection of insulin-specific lymphocytes through its expression by bone marrow-derived cells in the thymus (3). Genetic variation of insulin gene expression may thus influence either β-cell function or selection or activation of T-cells, respectively, in the thymus and the periphery.

To establish a model to study these different possibilities, we introduced genetic variation of insulin gene expression by transferring a null mutation of the proinsulin 2 gene onto the NOD genetic background. We observed an increased incidence of diabetes in heterozygote NOD mice carrying a null proinsulin 2 allele at successive BC generations. Our data indicate that genetic alteration of proinsulin 2 expression directly influences diabetes outcome in the NOD mouse. Furthermore, we have evidence that altered expression of the proinsulin 2 gene has direct consequences on lymphocyte recognition of proinsulin. Anti-insulin autoantibodies were detected with a higher frequency in proinsulin 2^+/− mice than in wild-type mice (data not shown). This does not exclude, however, that introducing variation in expression of the proinsulin 2 gene also influences diabetes development by modifying islet cell function. On the 129 genetic background, an increase in insulin 1 transcripts and in β-cell mass has been reported in proinsulin 2^−/− mutants. This compensatory response is reflected by normal insulin content, normal insulin levels, and normal glucose tolerance tests in mutant mice (12). On the NOD background, glucose tolerance tests were normal in proinsulin 2^+/− heterozygotes as compared with wild-type NOD mice. Plasma insulin levels were, however, increased at 8 weeks of age in proinsulin 2^+/− mice, possibly in relation with the diabetes autoimmune process and increased incidence of diabetes in female mutant mice. Increased plasma insulin levels have previously been reported in 8-week-old NOD mice as compared with control strains (15).

Finally, beyond expression of proinsulin 2 in the thymus (10,23,and our unpublished data), our data showing expression of proinsulin in the thymus of proinsulin 2^−/− mice demonstrates that proinsulin 1 can also be expressed in the mouse thymus. Variation in expression of genes encoding for β-cell autoantigens in the thymus possibly influence the pool of autoantigen-specific T-cells present in the peripheral T-cell pool and contribute to development of autoimmunity. A proinsulin 2 knockout model will allow us to precisely correlate expression level of proinsulin 1 and 2 in the thymus with susceptibility to type 1 diabetes. It will help to determine the contribution of proinsulin 1- and 2-specific lymphocytes to autoimmunity and the contribution of proinsulin 1 and 2 secretion by β-cells to diabetes development.

This work was supported by INSERM, Action Concertée Initiative grant no. 42 from the French Ministry of Research, and Juvenile Diabetes Foundation Research International Grant 1-2001-751. K.T. was supported by a grant from Fondation de la Recherche Médicale.