Peripherally-induced regulatory T cells contribute to the control of autoimmune diabetes in the NOD mouse model

Abstract

Type 1 diabetes (T1D) results from the autoimmune destruction of pancreatic beta cells and is partly caused by deficiencies in the Foxp3⁺ regulatory T-cell (Treg) compartment. Conversely, therapies that increase Treg function can prevent autoimmune diabetes in animal models. The majority of Tregs develop in the thymus (tTregs), but a proportion of Foxp3⁺ Tregs is generated in the periphery (pTregs) from Foxp3⁻CD4⁺ T-cell precursors. Whether pTregs play a distinct role in T1D has not yet been explored. We report here that pTregs are a key modifier of disease in the nonobese diabetic (NOD) mouse model for T1D. We generated NOD mice deficient for the Foxp3 enhancer CNS1 involved in pTreg induction. We show that CNS1 knockout decreased the frequency of pTregs and increased the risk of diabetes. Our results show that pTregs fulfill an important non-redundant function in the prevention of beta cell autoimmunity that causes T1D.

Graphical abstract

NOD mice, susceptible to autoimmune diabetes, harbor regulatory T cells both of thymic origin (tTregs) and induced in the periphery (pTregs). We deleted the Foxp3 enhancer CNS1 in NOD mice to decrease the frequency of pTregs but not tTregs, and found that this increased the risk of diabetes.

Introduction

Type 1 diabetes (T1D) is caused by autoreactive T cells that destroy pancreatic beta cells, leading to insulin deficiency. The T cells that drive pathogenesis can be held in check by Foxp3⁺ regulatory T cells (Tregs) [1,2]. Several of the gene variants that increase the risk of T1D do so by diminishing the functionality of the Treg compartment. For example, gene variants in the IL-2 pathway that controls the frequency of circulating Tregs have been shown to associate with disease in mouse and human [3,4]. Therapeutic administration of low-dose IL-2 increases the frequency of Tregs and prevents autoimmune diabetes in mouse models [5]. This approach is now in clinical trial for the prevention of T1D [6]. The success of low-dose IL-2 therapy in mice illustrates how manipulating the Treg compartment is a promising approach for the treatment of T1D. Yet, we do not fully understand the involvement of different Treg populations in the control of autoimmunity. The majority of Tregs emerge as a distinct population during thymic T-cell maturation. Defects in thymic Treg development are a possible cause for the immune dysregulation that leads to autoimmune diabetes. However, a proportion of mature Tregs forms not in the thymus but in the periphery [7,8]. These Tregs, termed peripherally-induced or pTregs, derive from mature Foxp3⁻CD4⁺ T cells following the encounter with tolerogenic stimuli, including TGF-β. The highest frequency of pTregs is found in the gut, likely owing to the capacity of gut microbes to promote pTreg induction [9–11]. pTregs are also present in other organs, but to date, a role for pTregs has only been demonstrated in the gut [12–14], in the lungs [12] and in fetal-maternal tolerance [15]. Whether pTregs participate in immune regulation in other organs is unclear. We set out to ask if pTregs have any role in pancreatic autoimmunity in the context of the nonobese diabetic (NOD) mouse model for T1D. Here, we report that the ability to generate pTregs is a critical modifier of autoimmune diabetes.

Results and discussion

pTregs are present in the pancreas of NOD mice

We started by measuring the frequency of thymus-derived Foxp3⁺CD4⁺ Tregs (tTregs) and pTregs in different organs in NOD mice. We used helios and neuropilin-1 (Nrp1) to distinguish tTregs from pTregs, as both markers have been shown to be primarily expressed in tTregs [16–18]. As reported previously, gut-associated lymphocytes harbored a high frequency of pTregs (25-30% of all colonic CD4⁺ T cells, Fig. 1). In contrast, pTregs represented a minority of Tregs in the spleen and mesenteric lymph nodes (mLN). Notably, pTregs were also detectable in the pancreatic lymph node (pLN) and in the pancreas (Fig. 1). The pLN is a major site of activation for autoreactive T cells in autoimmune diabetes. The presence of pTregs in the pLN and in the pancreas, although at low absolute numbers, suggested that pTregs could affect the autoimmune response that underlies T1D.

Figure 1.

pTregs are present in the pancreas of NOD mice. Lymphocytes isolated from the colonic lamina propria (Col. LP in black), mesenteric lymph nodes (mLN in purple), spleen (in orange), pancreatic lymph nodes (pLN in blue) and the pancreas (in green) of NOD WT mice were analyzed by flow cytometry. The full gating strategy is reported in Supporting Information Figure 3. Total Tregs were characterized as Foxp3⁺ cells (A). pTregs were characterized as Foxp3⁺Helios⁻ (B) or Foxp3⁺Nrp1^lo cells (C). All samples were gated on live CD4⁺CD8⁻ lymphocytes. Representative FACS plots are shown on the left, cell frequencies (mean ± SEM) on the right. Data in the histograms are pooled from 3-5 combined individual experiments with with n=11-24 per group (all mice were 6-week old). Exact P-values are shown (unpaired t-test).

CNS1 deletion in NOD mice decreases the frequency of pTregs in the pancreas

The differentiation of pTregs involves tolerogenic stimuli that include TGFβ [19]. TGFβ has a direct effect on expression of Foxp3 that characterizes CD4⁺ Tregs. De novo expression of Foxp3 in peripheral CD4⁺ T cells involves the conserved non-coding sequence 1 (CNS1) enhancer in the Foxp3 promoter [19,20], a binding site for Smad3 downstream of TGFβ signaling. CNS1 knockout (KO) decreases the frequency of pTregs without affecting tTregs [12,21]. We used CRISPR-Cas9 genome editing to generate CNS1 KO NOD mice (supporting information Fig. 1). The CNS1 KO allele was inherited with Mendelian frequency and had no deleterious effects on development or fertility. CNS1 deficiency decreased the frequency of pTregs in all organs examined, including blood, colon, spleen, lymph nodes and pancreas (Fig. 2A-E). Of note, CNS1 KO did not affect the frequency of tTregs or the frequency of total Foxp3⁺ Tregs, except in the colon where the contribution of pTregs to the total Treg population is greatest (Fig. 2D). Frequency and absolute numbers of other lymphocyte populations were comparable in WT and CNS1 KO mice (supporting information Fig. 2). As reported previously [12,21], CNS1 deficiency impaired both the in vitro and in vivo differentiation of naïve CD4⁺ T cells into Foxp3⁺ Tregs following tolerogenic stimulation (Fig. 2F,G).

Figure 2.

CNS1 deletion impairs pTreg formation in NOD mice. (A-C) Blood from founder mice (F0) as well as WT and CNS1 KO littermates in the third off-spring generation (F3) was analyzed by flow cytometry. Total Tregs were characterized as Foxp3⁺ cells (A) and pTregs were characterized as Foxp3⁺Helios⁻ (B) or Foxp3⁺Nrp1^lo (C) cells, all within live CD4⁺CD8⁻ lymphocytes. Representative plots (left panels) and cell frequencies (mean ± SEM, right panels) are shown. All mice were 10-12 weeks old. n=12 (WT), n=2 (CNS1 KO) for F0 and n=25 (WT), n=18 (CNS1 KO) for F3 mice. (D, E) Lymphocytes from NOD WT and CNS1 KO colonic lamina propria (Col. LP in black), mesenteric lymph nodes (mLN in purple), spleen (in orange), pancreatic lymph nodes (pLN in blue) and the pancreas (in green) were analyzed by flow cytometry. Total Tregs were characterized as Foxp3⁺ cells (D). pTregs were characterized as Foxp3⁺Helios⁻ cells (E). Data are pooled from 3-5 combined individual experiments. All mice were 6 weeks old, n=9-22 mice per group (). (F) Naïve CD4⁺ T cells from NOD WT or CNS1 KO mice were stimulated in vitro with irradiated splenocytes, anti-CD3, IL-2 and TGFβ. CD4⁺Foxp3⁺ Treg frequency was measured by flow cytometry 3 days later. Representative plots (left panels) and cell frequencies (mean ± SEM, right panel) are shown. n=2 mice/group, the value for each mouse was averaged from 3 technical replicates, and results are representative of three similar experiments. (G) Treg-depleted (CD4⁺CD25⁻) T cells were isolated from BDC2.5 TCR transgenic mice, labelled and transplanted into NOD mice. T cell recipients were injected i.v. with BDC2.5 mimetope peptide, and Foxp3 expression in labelled BDC2.5 T cells was measured by flow cytometry 7 days later. n=4 mice/group, results are representative of two similar experiments. Exact P-values are shown for all experiments (two-tailed unpaired t-test).

CNS1 deficiency increases the risk of autoimmune diabetes in NOD mice

We generated cohorts of CNS1 KO NOD mice using a breeding scheme that yielded control littermates (WT and heterozygous). Mice were not separated by genotype and were co-housed throughout the study. CNS1 KO causes gastrointestinal pathology in B6 mice [12], but not in a mixed 129/B6 background [21]. Unlike B6 mice, CNS1 KO NOD mice showed no signs of morbidity or weight loss during the 7 months observation period. However, we found that CNS1 deficiency increased the frequency of diabetes in both male and female mice (Fig. 3A). Insulitis was also more severe in pre-diabetic CNS1 KO mice compared with control littermates (Fig. 3B). These data show that pTreg deficiency increases the risk of autoimmune diabetes.

Figure 3.

CNS1 KO increases the frequency of diabetes in NOD mice. (A) Spontaneous diabetes frequency in WT or heterozygous (Control, n = 28) and CNS1 KO (KO, n = 14) female mice (left) and WT (n=23) and CNS1 KO (n=19) male mice (right). Disease frequency between groups was compared using the Log-rank test, exact P values are shown. (B) Insulitis in control and CNS1 KO female mice at 6 weeks (214 and 137 islets from 3 mice/group), 10 weeks (493 and 361 islets from 3-4 mice/group) and 14 weeks of age (725 and 391 islets from 4-5 mice/group). Islets were scored as free of insulitis, or as having moderate or severe infiltration, see representative images (40x magnification, scale bars indicate 50μm). Insulitis between genotypes was compared using Fisher’s exact test (comparing the proportion of infiltrated islets in each age group).

Concluding remarks

Treg therapy is a promising avenue for the treatment of autoimmunity. In this context, it is important that we better understand which Tregs can be leveraged for therapy. pTregs constitute a feasible target population, owing to the fact that these cells can be induced within the mature T-cell compartment by various agents, including microbial metabolites [22–24]. However, the extent to which pTregs contribute to the control of pancreatic autoimmunity had not yet been established. We showed here that impairing pTreg formation in NOD mice increased the risk of diabetes. The role of pTregs in islet autoimmunity may relate to the proximity of the pancreas to the gut. The pLN is the primary site of activation for beta cell-reactive T cells in autoimmune diabetes [25]. At the same time, the pLN is readily accessible to antigen and to cells from the gastrointestinal tract [26], providing a direct route by which gut microbe-induced pTregs could impact pancreatic autoimmunity. This putative link could help explain how the gut microbiota impacts the risk of T1D [27–29].

Materials and Methods

Mice

CNS1 KO NOD mice were generated using gRNAs composed of 5’-CACCGAAGACATACACCACCACGG-3’ annealed with 5’- AAACCCGTGGTGGTGTATGTCTTC -3’ and 5’- CACCGCATCAGTCCTCCAGCCAG -3’ annealed with 5’- AAACCTGGCTGGAGGACTGATGC -3’ cloned into the pX330 vector (Addgene), amplified by PCR and transcribed using the Megashortscript T7 transcription kit (Life Technologies). Cas9 mRNA (Trilink Technologies) and gRNAs were purified using the Megaclear clean-up kit (Life Technologies) and injected into the pronucleus of NOD zygotes that were reimplanted into pseudo-pregnant Swiss-Webster mice. Genotyping was performed using the PCR primers 5’-GGCGCTTATGTGGCTTCTTTC-3’, 5’-GAGGTAGCTTCTCATTTTCAAGTGG-3’ and 5’-GGAAGCCAACATGGGGTGAA-3’. See Supporting Information Fig. 1 for a graphical representation of the location of the gRNA sequences and primers relative to the CNS1 region. All experiments were performed with age- and sex-matched mice and approved by the Institutional Animal Care and Use Committee at the Joslin Diabetes Center.

Lymphocyte Isolation

Single cell suspensions were prepared from spleen and lymph nodes by mechanical disruption of tissue followed by red blood cell lysis using ACK buffer. Colonic lamina propria cells were isolated after removal of intraepithelial lymphocytes using EDTA followed by digestion with Collagenase Type VIII (Sigma Aldrich) and DNAse (Roche). Cells were collected using a Percoll (GE Healthcare) density gradient. Intra-pancreatic lymphocytes were isolated from the whole pancreas by digestion using Collagenase P (Roche) followed by a Histopaque-1077 (Sigma) density gradient.

Flow Cytometry

Cell suspensions were pre-incubated with Fc-block antibody (anti-CD16/32, Biolegend), then stained with cell surface antibodies (listed below). Dead cells were excluded using Zombie Fixable Viability dye (Biolegend). Intracellular staining for Foxp3 and Helios was performed with a FoxP3-labelling kit (eBioscience). Data was acquired on a LSRII or LSR Fortessa instrument (BD Biosciences) and analyzed with the FlowJo software. All data are shown using log-scale axes. The list of the antibodies is reported in Tabel 1 and the full gating strategy is reported in Supporting Information Figure 3.

In vitro regulatory T-cell induction

Naïve CD4 T cells were isolated using the naïve CD4 T-cell isolation kit (Miltenyi Biotech). Cells were stimulated in the presence of irradiated T cell-depleted WT splenocytes, 2 μg/ml anti-CD3 (2C11, Biolegend), 200 U/ml IL-2 and 2 ng/ml TGFβ (both from Preprotech) for 3 days, then analyzed by flow cytometry.

In vivo regulatory T-cell induction

CD4⁺CD25⁻ T cells were isolated from BDC2.5 TCR transgenic WT or CNS1 KO NOD mice by magnetic purification (using the CD4⁺CD25⁺ purification kit from Miltenyi Biotech) and labeled with fixable Cell Proliferation Dye eFluor-670 (eBioscience). 10⁷ purified T cells were injected intravenously (i.v.) into WT NOD mice. The following day, mice were injected i.v. with 5 μg BDC2.5 mimetope peptide (RTRPLWVRME, from Anaspec) [30]. Mice were sacrificed 7 days later to measure the frequency of Foxp3⁺ cells among labelled cells in the spleen by flow cytometry.

Diabetes Frequency Studies

Disease studies were performed with age-matched, contemporary cohorts of mice. Onset of diabetes was monitored by weekly measurements of glycosuria using Diastix (Bayer). Mice with two consecutive readings > 250 mg/dL were considered diabetic.

Insulitis

Pancreatic tissue was frozen in OCT (Fisher Scientific) prior to cryosectioning into 7 μm slices. For insulitis scoring, sections were stained with hematoxylin (Fisher Scientific) and counter-stained with eosin (alcoholic Eosin Y, Fisher Scientific). Pancreatic islets were scored blinded to the genotype as having no infiltration, moderate infiltration or severe infiltration. All sections were acquired on a Olympus BX-60 microscope equipped with an Olympus DP70 camera using the DPManager software.

Statistical Analyses

Data were analysed with the Prism software (Graphpad). Diabetes frequency was compared using the Log-rank test. Insulitis was compared using Fisher’s exact test. All other comparisons were performed using an unpaired t-test, with P < 0.05 considered significant. Exact P values are shown. Sample sizes were approximated in initial experiments, and adjusted to increase power as needed in replicate experiments.

Table 1.

Antibodies used for flow cytometry.

Antibody	Clone	Vendor
CD4-Brilliant Violet 605	RM4-5	Biolegend
CD4- Brilliant Violet 785	RM4-5	Biolegend
CD8-Brilliant Violet 711	53-6.7	Biolegend
CD8a-APC/Cy7	53-6.7	Biolegend
CD8a-APC	53-6.7	Biolegend
Helios-Pacific Blue	22F6	Biolegend
Helios-Alexa Fluor 488	22F6	Biolegend
Armenian Hamster IgG Isotype Control-Pacific Blue	HTK888	Biolegend
Foxp3-PE	FJK-16s	eBioscience
Rat IgG2a, k Isotype Control-PE	RTK2758	Biolegend
Foxp3-eFluor450	FJK-16s	eBioscience
CD19-APC/Cy7	6D5	Biolegend
CD304-PE/Cy7	3E12	Biolegend
CD304-PE/Cy7	3DS304M	eBioscience
Rat IgG2a, k Isotype Control-eFluor450	eBR2a	eBioscience