Introduction

Rodent models have been widely used to study the genetics of human complex diseases1,2. Except for more complicated breeding methods, experimental strategies typically involve a cross between two inbred strains that are respectively resistant and susceptible to a trait to map genetic loci regulating the phenotype. Initial localization of the chromosomal regions is then followed by generation of congenic strains to physically map the underlying genes. While these approaches are effective to define a relatively small region harboring phenotype-modulating genetic variants, identification of the causal genes remains difficult. When several genes are closely linked within a small genetic region, it is not practical to further test them by a traditional backcross breeding scheme due to limited recombination. The underlying gene is often inferred by its expression level and function or the presence of deleterious polymorphisms. The causal variant could be conclusively determined if coisogenic strains respectively express only one of the two parental alleles of a candidate gene. However, this is practically challenging because multiple polymorphisms often exist in the candidate gene and precise genetic manipulation requires prior knowledge of the credible variant.

Nonobese diabetic (NOD) mice develop spontaneous type 1 diabetes (T1D) mimicking the human disease and have been an effective model for identifying pathogenic genetic determinants3,4. By outcrossing to other strains, more than 30 autoimmune T1D susceptibility (Idd) loci have been identified in NOD mice3. Among those, the Idd9.3 locus has been mapped to the distal end of chromosome 4 containing the CD137-encoding Tnfrsf9 gene5,6. Interestingly, NOD-derived CD137 is hypofunctional compared to the C57BL/10Sn (B10) counterpart, and three nonsynonymous polymorphisms are present between NOD and B10 Tnfrsf9 alleles5,6. NOD mice congenic for the B10-derived Idd9.3 region (NOD.Idd9.3B10) are more resistant to T1D7. Compared to NOD mice, the NOD.Idd9.3B10 congenic strain has significantly higher levels of CD137+ FOXP3+ regulatory CD4 T cells (Tregs) and serum soluble CD1378. CD137+ Tregs are more suppressive than the CD137- subset in vitro and are the primary cellular source of soluble CD137 among T cells8. Importantly, recombinant soluble CD137 prevents NOD mice from developing T1D9. These observations provide a possible mechanism of B10 Idd9.3-mediated T1D resistance and support Tnfrsf9 as its underlying gene.

The purpose of this study was to develop a strategy that allows us to selectively express only the NOD or B10 Tnfrsf9 allelic variant in an identical genetic background. With the recent advance in nuclease-based genetic engineering10, we directly targeted Tnfrsf9 in both NOD and NOD.Idd9.3B10 mice and generated a pair of F1 hybrids mimicking coisogenic strains. We conclusively demonstrate that Tnfrsf9 allelic difference directly controls the frequency of CD137+ Tregs and the level of serum soluble CD137. We also provide a general strategy for testing candidate genes in rodent genetic mapping studies.

Results

Generation of F1 hybrids that selectively express only NOD or B10 Tnfrsf9

We previously targeted the second coding exon of the Tnfrsf9 gene directly in NOD mice using a pair of zinc-finger nucleases (ZFNs) to generate the NOD.Tnfrsf9−/− strain lacking CD137 expression11. Using the same ZFNs, we also successfully targeted Tnfrsf9 directly in the NOD.Idd9.3B10 strain to create NOD.Idd9.3B10.Tnfrsf9−/− mice. The mutations introduced in NOD.Tnfrsf9−/− and NOD.Idd9.3B10.Tnfrsf9−/− strains respectively deleted and inserted two base-pairs (Fig. 1a). While not identical, both mutations result in a premature stop codon at the target site (Fig. 1a). CD137 is upregulated on activated T cells, in particular CD8 T cells12. Thus, we stimulated total splenocytes isolated from NOD.Idd9.3B10 and NOD.Idd9.3B10.Tnfrsf9−/− mice with anti-CD3 overnight to compare CD137 expression. As shown in Fig. 1b, CD137 protein was expressed on CD8 T cells isolated from NOD.Idd9.3B10 but not NOD.Idd9.3B10.Tnfrsf9−/− mice, confirming the knockout phenotype. To determine the allelic effect of Tnfrsf9, we generated two types of F1 hybrids, (NOD.Idd9.3B10.Tnfrsf9−/− x NOD)F1 and (NOD.Idd9.3B10 x NOD.Tnfrsf9−/−)F1, to respectively express NOD and B10 Tnfrsf9 in the “identical” genetic background mimicking coisogenic strains (hereafter Tnfrsf9NOD and Tnfrsf9B10 respectively) for subsequent analyses (Fig. 1c).

Figure 1
figure 1

A general strategy for allele-specific expression of Tnfrsf9 in genetically identical NOD background strains. (a) Comparing the mutant sequences of Tnfrsf9 in our previously reported NOD.Tnfrsf9−/− and the newly generated NOD.Idd9.3B10.Tnfrsf9−/− strains to the wildtype NOD mice. A partial sequence of the Tnfrsf9 second coding exon is shown. The wildtype sequence of the ZFN target site of the Tnfrsf9 gene is shown above. The respectively altered sequences for NOD.Tnfrsf9−/− and NOD.Idd9.3B10.Tnfrsf9−/− are indicated below the wildtype. Each of the ZFN binding sequences on the opposite strands is underlined. The closed and open arrowheads respectively indicate the locations of the 2 base-pair deletion and 2 base-pair insertion (lower case) in the NOD.Tnfrsf9−/− and NOD.Idd9.3B10.Tnfrsf9−/− strains. The bold and italic letters depict the premature stop codons introduced as a result of the mutations at the ZFN target site. (b) Lack of CD137 protein expression in NOD.Idd9.3B10.Tnfrsf9−/− mice confirms the knockout phenotype. Total splenocytes isolated from 15-week-old NOD.Idd9.3B10 (+/+) and NOD.Idd9.3B10.Tnfrsf9−/− (−/−) male mice were stimulated with anti-CD3 (2.5 μg/ml) overnight and analyzed for CD137 expression by flow cytometry the following day. Cells were stained with anti-CD8 (for detecting CD8 T cells) and anti-CD137 or the corresponding isotype control antibody. The percentages of cells in the four quadrants of each plot are indicated. Results from two independent experiments are shown. (c) Schematic illustration of chromosome 4 in the parental NOD.Idd9.3B10.Tnfrsf9−/−, NOD, NOD.Idd9.3B10, and NOD.Tnfrsf9−/− mice as well as the (NOD.Idd9.3B10.Tnfrsf9−/− x NOD)F1 and (NOD.Idd9.3B10 x NOD.Tnfrsf9−/−)F1 mice (respectively indicated as Tnfrsf9NOD and Tnfrsf9B10) that selectively express only the NOD or B10 Tnfrsf9 gene. The B10-derived Idd9.3 congenic region is not drawn to scale.

Full size image

Allelic difference in the Tnfrsf9 gene controls the accumulation of CD137+ Tregs

Except the origin of the functional Tnfrsf9 allele, newly generated Tnfrsf9NOD and Tnfrsf9B10 F1 mice are considered genetically identical. These F1 mice allowed us to determine if selective expression of NOD or B10 Tnfrsf9 alone is sufficient to explain the phenotypic differences previously observed between NOD and NOD.Idd9.3B10 mice. T1D suppression in NOD.Idd9.3B10 mice is associated with a remarkably higher frequency of CD137+ Tregs8. Thus, we asked if this phenotype is controlled by the allelic difference between B10 and NOD Tnfrsf9. We first analyzed the frequencies of total Tregs and those that express CD137 in the spleen and pancreatic lymph node (PLN) of Tnfrsf9NOD and Tnfrsf9B10 F1 mice. We observed similar percentages of total FOXP3+ Tregs in the spleen and PLN of Tnfrsf9NOD and Tnfrsf9B10 F1 mice (Fig. 2a,b). Conversely, the frequency of CD137+ Tregs was higher in the spleen of Tnfrsf9B10 than Tnfrsf9NOD F1 mice (Fig. 2c), although the difference between Tnfrsf9NOD and Tnfrsf9B10 F1 hybrids was diminished relative to the comparison of NOD and NOD.Idd9.3B10 mice (Fig. 2d). This was likely because only one copy of functional Tnfrsf9 was expressed in Tnfrsf9NOD and Tnfrsf9B10 F1 mice. Similarly, the frequency of CD137+ Tregs was also increased in the PLN of Tnfrsf9B10 F1 progeny compared to that of Tnfrsf9NOD F1 mice, but the difference was not as significant as what was observed between NOD and NOD.Idd9.3B10 mice (Fig. 2e,f).

Figure 2
figure 2

B10 Tnfrsf9 increases the frequency of CD137+ Tregs. (a,b) The frequencies of total FOXP3+ Tregs are comparable in the spleen and PLN of Tnfrsf9NOD and Tnfrsf9B10 F1 mice. (a) Representative flow cytometry profiles of CD25 and FOXP3 staining in splenic CD4 T cells of 9–12 week-old Tnfrsf9NOD and Tnfrsf9B10 F1 female mice are shown on the left. Summarized results from three independent experiments are shown on the right. (b) Representative flow cytometry profiles of CD25 and FOXP3 staining in PLN CD4 T cells of 9–12 week-old Tnfrsf9NOD and Tnfrsf9B10 F1 female mice are shown on the left. Summarized results from three independent experiments are shown on the right. (c) The percentage of CD137+ Tregs is increased in the spleen of Tnfrsf9B10 F1 mice. Representative flow cytometry profiles of CD137 staining in splenic Tregs of 9–11 week-old Tnfrsf9NOD and Tnfrsf9B10 F1 female mice are shown on the left. Cells from a NOD.Tnfrsf9−/− mouse was used as the negative control. Summarized results from three independent experiments are shown on the right. As indicated, the average frequency of CD137+ Tregs in the spleen of Tnfrsf9B10 F1 is 2.6% higher than that of Tnfrsf9NOD F1 mice. *p < 0.05 by unpaired t test. (d) NOD.Idd9.3B10 mice have a higher frequency of splenic CD137+ Tregs than that of NOD mice. The results are summarized from 2 experiments using 9–12 week-old female mice. As indicated, the average frequency of splenic CD137+ Tregs in NOD.Idd9.3B10 is 7.2% higher than that of NOD mice. ***p < 0.0005 by unpaired t test. (e) The percentage of CD137+ Tregs is increased in the PLN of Tnfrsf9B10 F1 mice. Representative flow cytometry profiles of CD137 staining in PLN Tregs of 9–11 week-old Tnfrsf9NOD and Tnfrsf9B10 F1 female mice are shown on the left. Cells from a NOD.Tnfrsf9−/− mouse was used as the negative control. Summarized results from three independent experiments are shown on the right. As indicated, the average frequency of CD137+ Tregs in the PLN of Tnfrsf9B10 F1 is 3.7% higher than that of Tnfrsf9NOD F1 mice. *p < 0.05 by unpaired t test. (f) NOD.Idd9.3B10 mice have a higher frequency of PLN CD137+ Tregs than that of NOD mice. The results are summarized from 2 independent experiments using 9–12 week-old female mice. As indicated, the average frequency of PLN CD137+ Tregs in NOD.Idd9.3B10 is 10% higher than that of NOD mice. ***p < 0.0005 by unpaired t test.

Full size image

Allelic difference in the Tnfrsf9 gene controls the level of serum soluble CD137

CD137 is expressed as a membrane form or a soluble protein due to alternative splicing that removes the transmembrane domain13. Compared to NOD mice, the NOD.Idd9.3B10 congenic strain has a higher level of serum soluble CD1378. Thus, we asked if the level of serum soluble CD137 is also controlled by the allelic difference of Tnfrsf9. Sera were collected from Tnfrsf9NOD and Tnfrsf9B10 F1 mice and analyzed for soluble CD137 by enzyme-linked immunosorbent assay (ELISA). We observed a significant increase of serum soluble CD137 in Tnfrsf9B10 compared to Tnfrsf9NOD mice (Fig. 3a), albeit the difference was considerably reduced compared to that between NOD and NOD.Idd9.3B10 (Fig. 3b).

Figure 3
figure 3

B10 Tnfrsf9 increases the level of serum soluble CD137. (a) Serum soluble CD137 is higher in Tnfrsf9B10 than in Tnfrsf9NOD F1 mice. Sera were collected from 10-week-old female mice and analyzed by ELISA for CD137. As indicated, the average level of serum soluble CD137 in Tnfrsf9B10 F1 is 28.3 pg/ml higher than that of Tnfrsf9NOD F1 mice. *p < 0.05 by unpaired t test. (b) Serum soluble CD137 is higher in NOD.Idd9.3B10 than in NOD mice. Sera were collected from 10-week-old female mice and analyzed by ELISA for CD137. As indicated, the average level of serum soluble CD137 in NOD.Idd9.3B10 is 332.2 pg/ml higher than that of NOD mice. **p < 0.005 by unpaired t test.

Full size image

T1D incidence in Tnfrsf9 NOD and Tnfrsf9 B10 F1 mice

Compared to NOD mice, T1D development is partially suppressed in the NOD.Idd9.3B10 congenic strain5,7. The Tnfrsf9NOD and Tnfrsf9B10 F1 “coisogenic” strains provided an opportunity to directly test if Tnfrsf9 variants are responsible for the T1D modulatory effect of the Idd9.3 locus. To address this question, we monitored Tnfrsf9NOD and Tnfrsf9B10 F1 females for T1D development. While there was a trend of reduced T1D incidence in Tnfrsf9B10 compared to Tnfrsf9NOD F1 mice (p = 0.13 by Log-rank test), the difference did not reach statistical significance (Fig. 4). This could be due to the overall reduction of CD137 expression from a single Tnfrsf9 copy in Tnfrsf9B10 and Tnfrsf9NOD F1 mice that minimized the allele-specific functional differences important for regulating T1D development. This possibility is supported by the comparative analyses for the frequency of CD137+ Tregs and the level of serum soluble CD137 between Tnfrsf9B10 and Tnfrsf9NOD F1 hybrids and between NOD and NOD.Idd9.3B10 mice as shown in Figs 2 and 3.

Figure 4
figure 4

T1D incidence in Tnfrsf9NOD and Tnfrsf9B10 F1 mice. Tnfrsf9B10 and Tnfrsf9NOD F1 female mice were monitored for T1D development for 30 weeks. The incidence of T1D is not significantly different between Tnfrsf9B10 and Tnfrsf9NOD F1 mice (p = 0.13 by Log-rank test).

Full size image

Discussion

In this study we combined the power of congenic mapping and nuclease-based mutagenesis to establish a genetic strategy to directly compare two different alleles of the gene of interest in genetically identical strains. We tested this approach using the Idd9.3 candidate gene Tnfrsf9 in the NOD mouse model of T1D. NOD Tnfrsf9 encodes a hypofunctional CD137 compared to the B10 variant6. The functional difference has been associated with an increased frequency of CD137+ Tregs in the NOD.Idd9.3B10 strain compared to standard NOD mice8. In the present study, we also detected a significant increase in the frequency of CD137+ Tregs in the spleen and PLN of our F1 mice expressing the B10 allele of Tnfrsf9. These results indicate that a more functional CD137 molecule alone leads to the accumulation of CD137+ Tregs, independent of other genes within the Idd9.3 region or the differential state of diabetes progression in NOD and NOD.Idd9.3B10 mice. CD137+ Tregs exhibit superior function than their negative counterpart, possibly due to the production of soluble CD1378. Consistent with the previous study8, we also found a higher level of serum soluble CD137 in NOD.Idd9.3B10 than in NOD mice. We further show that a higher concentration of soluble CD137 was found in the serum of F1 mice expressing the B10 Tnfrsf9 allele compared to those possessing the NOD allele. These results indicate that the Tnfrsf9 allelic difference also directly controls the level of circulating soluble CD137. It is currently not known how alternative splicing of Tnfrsf9 is regulated to give rise to membrane and soluble forms of CD137. Whether an increased level of serum soluble CD137 in mice carrying the B10 Tnfrsf9 allele is a direct consequence of elevated CD137+ Tregs or involves regulation by other non-coding regions remains to be determined.

Tnfrsf9 has been implicated as an Idd9.3 underlying gene6. The Tnfrsf9NOD and Tnfrsf9B10 F1 “coisogenic” strains allowed us to directly test if allelic difference in Tnfrsf9 alone is sufficient to modulate T1D progression. Different from the results of the frequency of CD137+ Tregs and the level of serum soluble CD137, we did not observe significant T1D suppression in Tnfrsf9B10 compared to Tnfrsf9NOD F1 mice. The lack of diabetes inhibition conferred by the B10 Tnfrsf9 allele in our F1 mice could be that the functional difference conferred by one copy of B10 and NOD Tnfrsf9 is not sufficient to impact T1D development. This possibility is supported by the diminished differences in the frequency of CD137+ Tregs and the level of serum soluble CD137 between Tnfrsf9NOD and Tnfrsf9B10 F1 hybrids relative to the comparison between NOD and NOD.Idd9.3B10 mice (Figs 2 and 3). Another non-mutually exclusive explanation to the lack of T1D inhibition in Tnfrsf9B10 F1 mice is that other polymorphic gene(s) in the Idd9.3 region also impact T1D development. One possible candidate is the non-coding micro RNA Mir34a. In a recent study it was found that the NOD.Idd9.3 congenic strain had a significantly lower number of B cells than that in NOD mice14. The reduction of B cells in NOD.Idd9.3 congenic strain correlated with a higher level of Mir34a that was thought to impair normal B cell development by negatively regulating FOXP1 expression14. However, we did not observe a reduction of B cells in the NOD.Idd9.3B10 congenic strain compared to NOD mice (Supplementary Fig. S1). The reason for the discrepancy is currently not known but could be due to differences in the housing environment. Immune cells are under the circadian control15. The circadian gene Per3 represents another possible candidate in the Idd9.3 region. A similar genetic approach reported here can be used to test additional genes within the Idd9.3 region for their role in T1D development.

In conclusion, we demonstrate the feasibility to establish a “coisogenic” system by using a pair of F1 hybrids that selectively express an allele of the gene of interest for direct comparison of its functional consequence. The limitation of our approach is that only one copy of the gene is expressed, and the phenotypic difference may be quantitatively reduced if there is a dose-dependent effect. This possibility needs to be considered when testing candidate genes using our approach. Nevertheless, our strategy has the potential to be broadly applied to other experimental models to facilitate genetic mapping studies.

Materials and Methods

Mouse strains

NOD/ShiLtDvs (hereafter NOD) and NOD.B10Sn-Idd9.3C57BL/10SnJ/1106MrkTacJ (hereafter NOD.Idd9.3B10) were obtained from The Jackson Laboratory and subsequently maintained at the Medical College of Wisconsin MCW). NOD.Tnfrsf9−/− mice (line-6), currently maintained at the N4 generation, have been previously reported11. NOD.Idd9.3B10.Tnfrsf9−/− mice were generated using the previously described Tnfrsf9 specific zinc-finger nuclease pairs designed, assembled, and validated by Sigma-Aldrich11. The condition of microinjection and methods of founder screening and genotyping have been described in detail11. A founder (line-24) with a two base-pair insertion at the target site, resulting in a premature stop codon (Fig. 1a), was backcrossed to NOD.Idd9.3B10 for 4 generations followed by intercrossing to fix the mutation to homozygosity. All mouse experimental protocols were carried out in accordance with the MCW Institutional Animal Care and Use Committee guidelines and approved by the committee.

Flow cytometry analysis

Fluorochrome-labeled antibodies specific for CD8 (53-6.72), CD4 (RM4–5), CD3 (145-2C11), CD25 (7D4), CD137 (175B), Foxp3 (FJK-16s) were purchased from BD Biosciences (San Jose, CA), BioLegend (San Diego, CA), or eBioscience (San Diego, CA). Antibody staining and flow cytometry analysis procedures have been described previously16. For intracellular staining with FOXP3, samples were first stained for surface markers and then fixed/permeabilized for 2–4 hours using the FOXP3 staining buffer set from eBioscience. After fixation, the cells were washed twice with permeabilization buffer and then stained with anti-FOXP3 for 30 minutes. Stained cells were washed once with the FACS buffer and acquired using the FACSCalibur or LSRII flow cytometer (BD Biosciences). All flow cytometric data were analyzed with the FlowJo software (Tree Star, Ashland, OR). To analyze CD137 expression in NOD.Idd9.3 and NOD.Idd9.3.Tnfrsf9−/− mice, splenocytes from 15-week-old males were stimulated with 2.5 µg/ml anti-CD3 (145-2C11) overnight. Cells were then harvested and analyzed for CD137 protein expression by flow cytometry.

Soluble CD137 enzyme-linked immunosorbent assay (ELISA)

Sera were collected from 10-week-old females of the indicated strains. The concentration of soluble CD137 was determined using the Mouse 4-1BB/TNFRSF9 DuoSet ELISA kit (R&D Systems). Soluble CD137 was not detectable in serum from NOD.Tnfrsf9−/− or NOD.Idd9.3B10.Tnfrsf9−/− mice, validating the specificity of the ELISA kit.

Assessment of T1D

T1D development was monitored weekly using urine glucose strips (Diastix, Bayer) with onset defined by two consecutive readings of >250 mg/dl.

Statistical analysis

Unpaired t test was used for comparison between two groups. Log-rank test was used for the analysis of T1D incidence. All statistical analyses were performed using the GraphPad Prism 6 software (La Jolla, CA).