β-Cell Cre Expression and Reduced Ins1 Gene Dosage Protect Mice From Type 1 Diabetes

Abstract

A central goal of physiological research is the understanding of cell-specific roles of disease-associated genes. Cre-mediated recombineering is the tool of choice for cell type–specific analysis of gene function in preclinical models. In the type 1 diabetes (T1D) research field, multiple lines of nonobese diabetic (NOD) mice have been engineered to express Cre recombinase in pancreatic β cells using insulin promoter fragments, but tissue promiscuity remains a concern. Constitutive Ins1^{tm1.1(cre)Thor} (Ins1^Cre) mice on the C57/bl6-J background have high β-cell specificity with no reported off-target effects. We explored whether NOD:Ins1^Cre mice could be used to investigate β-cell gene deletion in T1D disease modeling. We studied wild-type (Ins1^WT/WT), Ins1 heterozygous (Ins1^Cre/WT or Ins1^Neo/WT), and Ins1 null (Ins1^Cre/Neo) littermates on a NOD background. Female Ins1^Neo/WT mice exhibited significant protection from diabetes, with further near-complete protection in Ins1^Cre/WT mice. The effects of combined neomycin and Cre knockin in Ins1^Neo/Cre mice were not additive to the Cre knockin alone. In Ins1^Neo/Cre mice, protection from diabetes was associated with reduced insulitis at age 12 weeks. Collectively, these data confirm previous reports that loss of Ins1 alleles protects NOD mice from diabetes development and demonstrates, for the first time, that Cre itself may have additional protective effects. This has important implications for the experimental design and interpretation of preclinical T1D studies using β-cell-selective Cre in NOD mice.

Type 1 diabetes (T1D) is a chronic disorder precipitated by immune-mediated pancreatic β-cell destruction and associated with the presence of autoantibodies against β-cell proteins (1, 2). Owing to the progressive loss of β cells and consequent insulin deficiency, individuals living with T1D have lifelong dependency on exogenous insulin (3). Higher levels of endogenous insulin secretion are associated with better short- and long-term outcomes in people living with T1D. Preservation of residual β cells and their function is therefore imperative (4).

Exogenous insulin administration is not a cure for T1D. Research using preclinical animal models continue to produce new therapeutic possibilities. Initially developed as a model for spontaneous onset of cataracts, the female nonobese diabetic (NOD/ShiLt) mouse (commonly known as NOD) is the most well established and extensively used pre-clinical model of T1D (5). The polygenic NOD mouse strain recapitulates multiple pathophysiological features of human T1D, including the development of autoantibodies in the prediabetic state (6), circulating autoreactive T cells (7), and subsequent onset of hyperglycemia as β-cell loss progresses (8). While the emergence of hyperglycemia can occur as early as 12–15 weeks of age, islet immune infiltration across the pancreas, known as insulitis, is established earlier (e.g. 8 weeks) (9). The cleanliness of housing facilities and the animal breeding approach are just 2 factors that affect the age at which these features present.

Mice express 2 nonallelic insulin genes. The insulin 2 gene (Ins2) is the murine homolog of the human insulin gene and is located on chromosome 7 (10). The insulin 1 gene (Ins1) is the result of an RNA-mediated gene duplication event. Ins1 has a simpler gene structure lacking the second intron present in Ins2, and is found on the murine chromosome 19 (10). Ins1 is expressed specifically in pancreatic β cells, whereas Ins2 is expressed predominantly in β cells, with trace expression in other tissues including the thymus and brain (11). Previous studies have examined the effects of insulin gene knockout in NOD mice. Complete knockout of Ins2 on the NOD background accelerated diabetes onset (2), a phenomenon attributed to a failure in central tolerization to insulin. Thymus-specific deletion of Ins2 was reported to be sufficient to cause spontaneous diabetes, even outside the NOD background (12), although we have not observed autoimmune diabetes in globally deficient Ins2^−/− mice (13, 14). It has previously been reported that male, but not female, NOD:Ins1^Neo/WT; Ins2^−/− mice, with a single remaining Ins1 allele, have been shown to succumb to insulin insufficiency (15), a finding we confirmed on other backgrounds (16). These observations were dependent on housing conditions (11). In contrast, NOD:Ins1^Neo/Neo mice have previously been shown to be protected from insulitis diabetes (10), which was proposed to be due to the loss of autoantigenic Ins1-derived peptides despite the presence of insulin autoantibodies (IAAs). Insulin is a primary autoantigen both in murine and human T1D pathogenesis (1). Replacing Ins1 with nonantigenic human insulin protects NOD mice from the onset of diabetes (17). Together, these previous observations indicate that insulin 1 gene dosage is a key driver of diabetes in NOD mice.

To advance the understanding of the underlying mechanisms of T1D pathogenesis, powerful genetic engineering tools are employed by researchers across the world. Ins1^Cre mice take advantage of the tissue specificity of Ins1 expression to excise loxP site-flanked (floxed) DNA segments. As the Cre DNA recombinase allele is inserted into exon 2 of Ins1 in these mice, floxed target genes are removed specifically in β cells at the cost of the loss of one Ins1 allele. Both Ins1^Cre and Ins1^Neo mice have reduced Ins1 gene dosages with 50% or 100% in their heterozygous and homozygous states, respectively. For Ins1^Neo mice, the neomycin cassette obliterates the entire Ins1 promotor and gene sequence in addition to deletion of 7 to 9 kb upstream and downstream of Ins1 (Fig. 1A).

Figure 1.

Ins1 replacement with Cre and Neo protects female nonobese diabetic (NOD) mice from type 1 diabetes. A, Structure of the wild-type (WT) Ins1 locus, recombinant alleles result of the neomycin (neo) targeting vector, or the Ins1 Cre locus. B, Overview of breeding strategy. Created with Biorender.com. C, Kaplan-Meier plot denoting diabetes incidence in NOD mice by Ins1 genotype, including Ins1^WT/WT, Ins1^Cre/WT, Ins1^Neo/WT, and Ins1^Neo/Cre. Survival analysis was performed using log-rank (Mantel-Cox) test (P value shown). D and E, Individual and mean random blood glucose of female mice. The mean random blood glucose of the Ins1^Neo/Cre mice was statistically significantly lower than that of the Ins1^WT/WT mice, with an adjusted P value of .002. Mean blood glucose of the Ins1^Neo/WT mice was higher than those of the Ins1^WT/WT and Ins1^Cre/WT mice, with adjusted P values of .001 and .005, respectively. F and G, Individual and mean body mass traces female mice. The mean body mass of the Ins1^Neo/Cre mice was significantly lower compared to the Ins1^Cre/WT colony (adjusted P value < .001). The mean blood glucose of the Ins1^Cre/WT was also higher than the Ins1^WT/WT and the Ins1^Cre/WT colonies, both with an adjusted P value less than .001. Error bars represent SEM.

Since the early days of Cre/loxP system use, concerns have been raised about side effects such as Cre toxicity. More than 2 decades ago, Loonstra et al (18) observed increased sister chromatid exchange frequency, leading to a cellular halt in the G₂/M phase and a reduction in the proliferation of mouse embryonic fibroblasts (MEFs) transduced with a bicitronic retroviral vector encoding Cre. Silver and Livingston (19) demonstrated similar results including ceased proliferation in 293xLac cells, NIH 3T3 cells, and MEFs transfected with Cre-expressing retroviral vectors. The same study revealed chromosomal abnormalities in Cre-expressing MEFs. Lentiviral Cre administration and expression have likewise been shown to reduce proliferation through cellular accumulation in the G₂M phase of Cre-expressing CV-1 and COS cells, 2 kidney cell lines derived from monkeys (20). Cre toxicity has been shown to be dose dependent, even when expressed transiently, when using an adenovirus vector (21). Past studies have confirmed Cre toxicity and apoptosis in mouse cardiac tissue (22, 23) and p53^−/− thymic lymphoma cells (24). Collectively, these findings emphasize the necessity of using Cre-only (no flox) controls for in vivo research study designs since Cre recombinase expression may affect cells beyond the intended target gene deletion.

For this study, we generated and characterized a new NOD Ins1^Cre knockin mouse line. Our intention was to establish an ideal in vivo tool for the study of specific β cell genes in T1D pathogenesis. The goal of this study was to investigate the effect of 1) deleting one Ins1 allele and 2) introducing β-cell–specific Cre expression on the spontaneous diabetes development for which NOD mice are well known. As previous studies by the Eisenbarth group (10, 15)provided evidence that Ins1^Neo/WT mice, with only one intact Ins1 allele, exhibit decreased and delayed diabetes incidence, we were expecting similar results for Ins1^Cre/WT mice. We incorporated both NOD:Ins1^Neo/WT and NOD:Ins1^WT/WT littermate controls in our study design, to control for both the loss of one Ins1 allele and the introduction of Cre expression in NOD Ins1^Cre/WT mice. We compared diabetes incidence, insulitis severity, and immune activation between groups within each sex. Our data demonstrate that Ins1-driven Cre expression has further protective effects beyond the loss of a single Ins1 allele deleted via Ins1^Neo. Carefully chosen proper controls and caution are therefore required when interpreting experiments including β-cell–specific Cre expression in mice.

Materials and Methods

Mice

All animal procedures and ethical standards were in accordance with the Canadian Council for Animal Care guidelines. All animal studies and protocols were approved by the University of British Columbia (UBC) Animal Care Committee and Institutional Care and Use Committee at the University of Michigan. At UBC, mice were housed in the Centre for Disease Modelling specific pathogen-free facility on a standard 12-hour light/12-hour dark cycle with ad libitum access to chow diet (PicoLab, Mouse Diet 20-5058). To generate NOD:Ins1^Cre and NOD:Ins1^Neo mice, we contracted Jackson Laboratories to backcross (> 12 times) B6(Cg)-Ins1^{tm1.1(cre)Thor}/J (Jackson Laboratory, No. 026801) (Ins1^Cre) and NOD.129S2(B6)-Ins1^tm1Jja/GseJ (Jackson Laboratory, No. 0005035) Ins1^Neo mice onto a NOD/ShiLtJ (Jackson Laboratory, No. 001976) (NOD) background. The Ins1^Cre and Ins1^Neo mice were originally on a mixed, largely C57Bl/6J, background. Subsequently, we designed a strict breeding strategy for our study. An Ins1^Cre maternal parent colony as well as an Ins1^Neo paternal parent colony was established (Fig. 1B). Each parent colony was backcrossed every 5 generations. Female Ins1^Cre/WT and male Ins1^Neo/WT mice were set up as breeders, at age 7 to 8 weeks, to generate experimental mice. Mice from the parental colonies were included only once as breeders to eliminate the risk of onset of hyperglycemia during pregnancy and weaning at later ages. This breeding strategy generated littermates of four genotypes: wild-type (WT) NOD:Ins1^WT/WT mice with both insulin 1 alleles, heterozygous NOD:Ins1^Cre/WT mice with an Ins1 replaced with Cre-recombinase, heterozygous NOD:Ins1^Neo/WT mice with one Ins1 allele replaced with a neomycin cassette, and full Ins1 null mice with both Ins1 alleles replaced: NOD:Ins1^Neo/Cre. Specific cohorts were monitored twice per week for the sole purpose of determining hyperglycemia onset incidence and body mass changes. Any mice that developed diabetes, defined as 2 consecutive blood glucose measurements greater than or equal to 16 mmol/L or 1 measurement greater than or equal to 22 mmol/L, were euthanized. Specific cohorts generated for tissue analysis were terminated at age 12 weeks in a prediabetic phase (for the Vancouver housing facility) or at age 1 year. All animals were monitored for diabetes before euthanasia.

An additional colony of NOD:Ins1^Cre mice was generated independently through in-house backcrossing at a second site (University of Michigan) via the speed congenic approach in consultation with Charles River Laboratories. Following each backcross, Ins1^Cre/WT offspring with allelic profiles most closely matching the NOD strain (determined by MAX-BAX mouse 384 SNP panel screening), were selected as breeders for the subsequent backcross. Following 8 generations of backcrossing, animals with an allelic profile percentage match greater than 99.9% were used to generate Ins1^WT/WT and Ins1^Cre/WT experimental mice. Animals were housed in a specific pathogen-free facility on a standard 12-hour light/12-hour dark cycle with access to ad libitum chow diet (LabDiets, Rodent Diet 5L0D) and water. Drinking water was provided at a pH level of 2.5 to 3 on advice from Jackson Laboratories that acidified water supports the diabetes frequency in many NOD colonies (25). Blood glucose measurements were taken once or twice per week. Incidence of diabetes, defined as blood glucose levels greater than 16 mM for 5 consecutive measurements, was recorded. Diabetic mice were euthanized and excluded from future analysis.

Tissue Processing and Histology

Immediately following euthanasia, pancreata were collected according to a preestablished protocol with the exception that extracted pancreases were not further treated to obtain isolated islets and were instead processed as a whole (26). The dissected pancreata were fixed in 4% paraformaldehyde for 24 hours before storage in 70% ethanol at 4 °C. Paraffin-embedded sections were prepared, stained with hematoxylin and eosin (H&E), and imaged by WaxIT Histology Services Inc.

Islet Infiltration Scoring

Images from H&E-stained pancreatic sections were analyzed with QuPath software (27). Islets in pancreatic sections were scored blindly for pancreatic islet infiltration of mononuclear immune cells according to the previously established 4-point scale (28). In brief, 0 equals no insulitis, 1 equals peri-insulitis marked by less than 25% peripheral immune-islet infiltration, 2 equals insulitis marked by 25% to 75% immune cell infiltration, and 3 equals severe insulitis marked by greater than 75% immune-islet infiltration. Random samples were scored by a second person to ensure consistency at 25 to 30 islets per mouse.

Insulin Autoantibodies and Plasma Insulin

Blood for plasma insulin and autoantibodies measurements was collected at study end point by cardiac puncture from anesthetized animals (2.5% isoflurane) and put directly on ice. Insulin plasma samples were measured using a Mouse Insulin enzyme-linked immunosorbent assay (ELISA) kit (No. 80-INSMS-E10, RRID: AB_2923075; ALPCO). Samples were shipped to the Insulin Antibody Core Laboratory at the University of Colorado, Barbara Davis Diabetes Center, for autoantibody analysis. The IAA data are expressed as an index against a standard positive and negative controls as per the procedure at the Insulin Antibody Core Laboratory.

Flow Cytometry

Immediately following euthanasia, lymph nodes and spleen were isolated and splenic single-cell suspensions were counted and stained with fluorescently conjugated monoclonal antibodies for cell-surface markers (Table 1 lists antibodies used). Fixable Viability Dye eFluor 506 was used as viability dye (No. 65-0866-14; ThermoFisher). Following staining, cells were analyzed by flow cytometry and Flow Jo software (Tree Star Inc).

Table 1.

Summary of antibodies used for flow cytometry

Antibody	Manufacturer	Color	Laser, nM	Filter	Catalog No.	RRID
CD3	ThermoFisher	e-Flour 540	405	450/45	48-003382	AB_2016704
CD4	ThermoFisher	BV650	405	660/10	64-0042-82	AB_2662401
CD8	BioLegend	PE-TR	561	610/20	100762	AB_2564027
CD11c	ThermoFisher	PE	561	585/42	12-0114-81	AB_465551
CD19	ThermoFisher	SB780	405	763/43	78-0193-82	AB_2722936
CD44	ThermoFisher	APC	633	660/10	17-0441-82	AB_469390
CD62L	ThermoFisher	PerCP-Cy5.5	488	690/50	45-0621-82	AB_996667
CD69	ThermoFisher	FITC	488	525/40	11-0692-82	AB_11069282
Foxp3	ThermoFisher	AF-700	633	712/25	565773-82	AB_1210557

Statistical Analysis

Statistical significance was assessed using 2-way analysis of variance analysis with the Tukey multiple comparisons test, at a threshold of P less than .05. Plasma insulin, insulitis, and IAA significance were quantified by 1-way analysis of variance analyses. We used the Mantel-Cox log-rank test, corrected with the Benjamini-Hochberg procedure, to analyze the Kaplan-Meier survival plots. Repeated-measures, mixed-effect models were applied for random blood glucose and body mass analyses.

Prism 9 (GraphPad Software Inc) was used for statistical analyses and generation of most figure panels. Data are expressed as mean ± SEM unless otherwise specified.

Results

Type 1 Diabetes Incidence

It is known that a greater proportion of female NOD mice develop diabetes and the timing of the development of the disease is more consistent compared with male NOD mice. We observed the highest diabetes incidence in female NOD:Ins1^WT/WT mice, with 65% diabetes incidence at 1 year of age (Fig. 1C). The majority of these mice (7/9) were diabetic before age 20 weeks. The diabetes incidence and disease time-course of this study was comparable with previous NOD:Ins1^WT/WT cohorts in our facilities (28), although slightly delayed relative to some other facilities (discussed below). Female NOD:Ins1^Cre/WT exhibited both a delay in diabetes onset and a diabetes incidence by the end of the study of only 27%, which was statistically significantly different when compared to NOD:Ins1^WT/WT mice at age 1 year (P = .033; P_adjust = .078; see Fig. 1C). We also observed a delay in the timing of the onset between female NOD:Ins1^Neo/WT mice (see Fig. 1C), in agreement with previous studies of NOD:Ins1^Neo/WT mice (10), but there was no difference in the final diabetes incidence between female littermate NOD:Ins1^Neo/WT mice (64%) and NOD:Ins1^WT/WT mice (64%) (P = .40; P_adjust = .48; see Fig. 1C). Double-mutant female NOD:Ins1^Neo/Cre mice (lacking both WT alleles of Ins1) were protected from diabetes (29% diabetes incidence) when compared to NOD:Ins1^WT/WT mice (64%), (P = .019; P_adjust = .078), to a similar extent as NOD:Ins1^Cre/WT (27%) mice (P = .98; P_adjust = .98; see Fig. 1C). NOD:Ins1^Neo/WT mice had a statistically significantly lower diabetes incidence than NOD:Ins1^Cre/Neo mice (P = .038; P_adjust = .078; see Fig. 1C), further suggesting Cre expression rather than loss of one Ins1 allele protects against diabetes onset in NOD mice. No differences were observed in random blood glucose (Fig. 1D and 1E) and body mass (Fig. 1F and 1G) before diabetes onset in female mice. Together, these data confirm previous findings that reduced Ins1 gene dosage protects NOD mice from diabetes, but also reveal a further, additive protective effect of β-cell Cre expression.

Male NOD:Ins1^WT/WT mice, as expected, demonstrated a low diabetes incidence (20% incidence). We observed no cases of hyperglycemia in any of the male mice with reduced Ins1 gene dosage (Fig. 2A). There were no differences in random blood glucose (Fig. 2B and 2C) or body mass (Fig. 2D and 2E) between any of the groups. These results confirm the strong sex bias in this NOD model and suggest that the protection conferred by reducing Ins1 gene dosage may not be sex specific.

Figure 2.

Effects of Ins1 replacement with Cre and Neo in male nonobese diabetic (NOD) mice. A, Kaplan-Meier plot denoting diabetes incidence in NOD mice by Ins1 genotype. B and C, Individual and mean random blood glucose in male mice. D and E, Individual and mean body mass traces in male mice. Error bars represent SEM.

Insulitis and Insulin Autoantibodies

Next, we examined the effects of reduced Ins1 gene dosage and Cre expression on insulitis in female mice, the pathological evidence of islet-directed autoimmunity. We found no statistically significant difference in plasma insulin or insulin autoantibodies between any of the genotypes (Fig. 3A and 3B). H&E-stained pancreas sections (Fig. 3C) were blindly scored for immune islet infiltration in a prediabetic cohort of littermates euthanized at age 12 weeks, and also in the mice that survived to 1 year. Insulitis scores from 12-week-old mice with reduced Ins1 gene dosages were not statistically significantly different from NOD:Ins1^WT/WT littermates. We noticed the least amount of immune islet infiltration in double-mutant NOD:Ins1^Neo/Cre mice (Fig. 3D), consistent with their more complete protection from diabetes. The outcomes of the samples gathered at 1 year are unfortunately inconclusive because of low sample size (NOD:Ins1^WT/WT, n = 3) (Fig. 3E). The lower survival rate may therefore mask any potential statistically significant changes for both insulitis scoring (see Fig. 3E) and IAAs (see Fig. 3B). Unfortunately, we did not collect blood for this analysis at the 12-week time point for all cohorts because of challenges brought about by the COVID-19 pandemic.

Figure 3.

Insulitis scoring in female nonobese diabetic (NOD) mice with Ins1 replacement. A, Plasma insulin concentration nanomolar (nM) collected at 14 weeks by cardiac puncture. B, Mouse insulin autoantibodies (IAAs) collected at 14 weeks by cardiac puncture; result of IAA is expressed as an index, against internal standard positive and negative controls. C, Representative images of hematoxylin and eosin–stained pancreata used for insulitis scoring. Scale bars are 100 μm. D, Mean percentage insulitis scores at 12 weeks old of each genotype, categorized by score, 0—no insulitis, 1—peri-insulitis (< 25%), 2—insulitis (25%-75%), and 3—severe insulitis (> 75%), respectively, at E, 1-year-old time points. Error bars represent SEM.

Immune Cell Characterization

To assess immune cell populations in the pancreatic lymph nodes and spleen at age 14 weeks, we used a panel of validated antibodies for flow cytometry. While we were able to confidently identify many key immune cell populations (Fig. 4), there were no statistically significant differences between groups (Fig. 5). These observations demonstrate that β-cell–specific insulin gene manipulations alter T1D incidence without robust effects on the lymphocytes found in the pancreatic lymph nodes and spleen.

Figure 4.

Gating strategy for flow cytometry of pancreatic lymph node and spleen cells. A, Singlets were obtained with use of FSC-A × FSC-H parameters and viable cells were identified by selecting viability dye-negative cells for subsequent analysis. The populations were subsequently split into 3 groups of interest with dendritic cells identified by CD11c, B cells identified by CD19, and T cells identified by CD3. T cells were further categorized into cytotoxic and helper phenotypes with use of CD4 and CD8 markers, and their respective single marker populations were assessed for activation and priming status (naive, effector, memory) with the use of CD69, CD44, and CD62L. Regulatory T-cell populations were further selected for with the use of a Foxp3 marker.

Figure 5.

Immune profiling in female nonobese diabetic (NOD) mice with Ins1 replacement. Flow cytometric analysis of cell populations within the pancreatic lymph node and spleen at age 14 weeks. A, Percentage of CD3+-positive T cells and CD19+ B cells from the pancreatic lymph node. B, Percentage of CD11+ dendritic cells from the pancreatic lymph node. C, Percentage of CD8 + CD4- cytotoxic T cells, CD8 + CD4– helper T cells and CD4 + Foxp3+ Treg cells from the pancreatic lymph node. D, Percentage of CD8 + CD4 + immature cells and CD8-CD4– double negative (DN) T cells from the pancreatic lymph node. E, Percentage of CD69+ activated cytotoxic T cells from the pancreatic lymph node. F, CD44loCD62L + naive cytotoxic T cells from the pancreatic lymph node. G, Percentage of CD44hiCD62—memory cytotoxic T cells from the pancreatic lymph node. H, Percentage of CD44midCD62—effector cytotoxic T cells from the pancreatic lymph node. I, Percentage of CD69+ activated helper T cells from the pancreatic lymph node. J, CD44loCD62L + naive helper T cells from the pancreatic lymph node. K, Percentage of CD44hiCD62 + memory helper T cells from the pancreatic lymph node. L, Percentage of CD44midCD62—effector helper T cells from the pancreatic lymph node. M, Percentage of CD3+-positive T cells and CD19+ B cells from the spleen. N, Percentage of CD11+ dendritic cells from the spleen. O, Percentage of CD8 + CD4− cytotoxic T cells, CD8 + CD4− helper T cells, and CD4 + Foxp3+ Treg cells from the spleen. P, Percentage of CD8 + CD4 + immature cells and CD8-CD4– DN T cells from the spleen. Q, Percentage of CD69+ activated cytotoxic T cells from the spleen. R, CD44loCD62L + naive cytotoxic T cells from the spleen. S, Percentage of CD44hiCD62—memory cytotoxic T cells from the spleen. T, Percentage of CD44midCD62—effector cytotoxic T cells from the spleen. U, Percentage of CD69+ activated helper T cells from the spleen. V, CD44loCD62L + naive helper T cells from the spleen. W, Percentage of CD44hiCD62 + memory helper T cells from the spleen. X, Percentage of CD44midCD62—effector helper T cells from the spleen. Error bars represent SEM.

Independent Validation Cohort

To ensure the protective effects of β-cell Cre expression were not solely limited to a single animal housing facility, we also studied female NOD:Ins1^Cre/WT mice that were independently generated at a separate site (Fig. 6) in parallel to the cohorts studied in Fig. 1. The overall incidence of diabetes development in female NOD mice in this second animal colony was as expected (∼ 65%-80% by age 25 weeks; Fig. 6A). Similar to our data shown in Fig. 1, we observed that female NOD:Ins1^Cre/WT animals were protected from diabetes incidence (25% by 1 year) and had statistically significantly improved mean blood glucose (Fig. 6C) when compared to NOD:Ins1^WT/WT littermates (75% incidence by 1 year). Again, these studies confirm the protective effects of reduced Ins1 gene dosage and β-cell Cre expression in NOD mice and suggest that these findings are not a consequence of environment or housing.

Figure 6.

Ins1 replacement with Cre protects female nonobese diabetic (NOD) mice from type 1 diabetes in an independent facility. A, Kaplan-Meier plot denoting diabetes incidence. B and C, Individual and mean random blood glucose of female NOD colonies from a second, independent site. The mean blood glucose of the Ins1^Neo/Cre (green) was significantly lower than that of the Ins1^WT/WT (blue) littermates, with an adjusted P value less than .05. Female NOD nonlittermate controls (red) were used to track overall diabetes incidence in the colony. Error bars represent SEM.

Discussion

In this study, we used a rigorous littermate control study design to examine the effects of replacing one Ins1 allele with Cre-recombinase on the onset of diabetes in mice of a NOD background. To investigate whether effects were due to the loss of Ins1 or the introduction of Cre, we included Ins1^Neo/WT mice in our study. Both Ins1^Cre and Ins1^Neo mice have reduced Ins1 gene dosages with 50% or 100% in their heterozygous and homozygous states, respectively. We found a similar reduction in the diabetes incidence in female Ins1^Cre/WT mice and Ins1^Neo/Cre mice when compared to littermate control Ins1^WT/WT mice. This work has implications for the understanding of the pathogenesis of T1D, as well as for the use of Cre recombinase as a tool for in vivo genome engineering in pre-clinical models of disease.

Our findings concur with previous work that showed a similar reduction in diabetes incidence in female NOD mice lacking 1 or 2 alleles of Ins1 (10, 15), although we did not observe additional protection with the double Ins1 knockout beyond what was seen with the Cre-replacement allele. Similarly, our data are in alignment with a previous study showing replacing the murine Ins1 gene with the human INS gene was found to protect female NOD mice from diabetes both in heterozygous and homozygous states (17). We were unable to detect differences in diabetes incidence in male Ins1^Cre/WT, Ins1^Neo/WT, and Ins1^Neo/Cre NOD mice, perhaps because of an insufficient study time period. A previous study found that removal of a single Ins1 allele is sufficient to abolish spontaneous diabetes in 50-week-old male NOD mice (10). Together with the work of others, our experiments support the contention that proinsulin 1 is a key player in the development of autoimmunity in mice. While there are several autoantigens targeted by autoreactive T cells in T1D (2), insulin and proinsulin are particularly common autoantibody targets in prediabetic humans (29, 30). Our experiments were underpowered to detect subtle differences in the levels of IAAs because we were limited to examining only a single time point. However, we did examine insulitis at 2 time points, and insulin antibodies are often correlated with insulitis (31). In our hands, there was a qualitative difference in the number of islets that did not exhibit insulitis at age 50 weeks in mice with at least 1 Ins1 allele replaced. At age 12 weeks, there was a slight trend toward more insulitis-free islets in the Ins1^Neo/Cre mice, consistent with the greater protection from T1D incidence. These observations are consistent with previous studies showing that Ins1 knockout in NOD mice is protective against the development and severity of insulitis (10). Thus, our results show that reducing the Ins1 gene dosage lowers the threshold required for diabetes onset, likely by removing the source of primary autoantigens and suppressing insulitis.

An important observation of our study is that Cre expression itself has protective effects in NOD mice, beyond the protection afforded by the loss of 1 Ins1 allele. To examine the specific consequences of Cre expression, we compared NOD; Ins1^Cre/WT to a different knockin (neo) Ins1^Neo/WT mouse line and found roughly twice as much final diabetes protection in Ins1^Cre/WT mice compared with Ins1^Neo/WT mice. Though it is possible diabetes could have been delayed further than 1 year, these findings suggest that Cre itself may affect diabetes rates in female NOD mice, with the caveat that the Neo and Cre insertions are different, altering the local gene structure. Mechanistically, we attribute the observed protection to differences in insulitis, suggesting a β-cell autonomous effect of Cre expression. Previous studies highlighted the potential of Cre recombinase to result in toxicity due to DNA damage (32, 33). Mammalian genomes contain hundreds of pseudo-loxP sites, and even though these sequences can deviate considerably from the consensus loxP site, they can still serve as functional recognition sites for Cre (32, 33). The sustained presence of high levels of Cre in fibroblasts can cause growth arrest and chromosomal abnormalities (18–20, 22, 23). Cre-dependent DNA damage and accumulation of cytoplasmic DNA have been shown to initiate a STING-dependent immune response (34). STING is an intracellular adaptor molecule, associated with the endoplasmic reticulum membrane (35), that can play a critical role in detecting pathogen-derived DNA in the cytoplasm (36). There is precedence for diabetes protection in NOD mice with early exposure to coxsackievirus (37). A recent paper reported that STING is required for normal β-cell function in mice (38), although, ironically, the study did not employ Cre-only controls. Theoretically, Cre expressed in β cells could delay the onset of diabetes in a STING-dependent manner. Future studies, will be required to delineate the molecular mechanisms by which Cre expression induces further protection than Ins1 loss in the NOD mouse model. To unequivocally demonstrate an effect of Cre activity on progression to T1D/insulitis, mice wouldideally be generated with enzyme-dead Cre knocked into the same locus.

As with all studies, this work has a number of limitations. For example, the broadly observed phenomenon that diabetes incidences differ between NOD mouse colonies means that we cannot directly compare the results between sites. We do not know the reasons for the apparent difference in diabetes incidence in WT NOD mice between our colonies, but it could be related to many environmental factors, including native microbiome, water, food, or bedding, as well as subtle differences in genotype. Another caveat of our study is that we could not simultaneously address Ins1 gene dosage and the effects of the Cre transgene because generating both homozygous and heterozygous littermates with each of the knockin alleles is impossible. To reduce the potential effects of environment and genotype, mentioned earlier, it was imperative that we prioritized using littermates. Another limitation is the uncertainty as to whether Ins1^Cre/WT mice could have developed diabetes beyond the end of this study. While our study has limitations, we believe it is important to report these results that will help guide those in the field who use T1D mouse models or Cre in any context.

In summary, our observations suggest caution when interpreting experiments that involve Cre recombinase in NOD mice. Our data showed that Cre expression itself has protective effects in NOD mice beyond the protection afforded by the loss of 1 Ins1 allele. Cre-loxP systems are vital tools for research; however, there are multiple caveats that should be considered related to off-target effects and the determination of correct controls. At the bare minimum, Cre-only controls are essential. Additional tools for in vivo genome engineering are required to advance the field. Many studies will need to be reinterpreted.

Abbreviations

IAAs

insulin autoantibodies

H&E

hematoxylin and eosin

MEF

mouse embryonic fibroblast

NOD

nonobese diabetic

T1D

type 1 diabetes

UBC

University of British Columbia

WT

wild-type

Financial Support

This work was primarily supported by the Canadian Institutes for Health Research (operating grant No. PJT-152999 to J.D.J.) and the JDRF Centre of Excellence at UBC (No. 3-COE-2022-1103-M-B). S.A.S. was supported by the JDRF (Nos. CDA-2016-189 and COE-2019-861), the National Institutes of Health (Nos. R01 DK108921 and U01 DK127747), and the Department of Veterans Affairs (No. I01 BX004444).

Author Contributions

S.S. co-conceived experiments, designed and executed in vivo experiments, analyzed data, and co-wrote the manuscript; P.O. executed in vivo experiments, analyzed data, and co-wrote the manuscript; J.M.Z. analyzed data and co-wrote the manuscript; J.L. designed and analyzed flow cytometry experiments; J.C.C.Y. executed in vivo experiments; I.S. designed and executed flow cytometry experiments; V.S. conducted in vivo validation experiments; E.L.D. conducted in vivo validation experiments; J.Z. conducted in vivo validation experiments; S.A.S. oversaw and interpreted in vivo validation experiments; M.H. oversaw and interpreted in vivo experiments; J.D.J. designed the model, co-conceived experiments, analyzed data, co-wrote the manuscript, and is the ultimate guarantor of this work.

Disclosures

The authors have no conflicts of interest to disclose.

Data Availability

Data generated during and/or analyzed during the present study are not publicly available but are available from the corresponding author on request.