Activation of Islet Autoreactive Naïve T Cells in Infants Is Influenced by Homeostatic Mechanisms and Antigen-Presenting Capacity

Cord blood (CB) from 20 healthy, full-term newborns, acquired immediately after delivery from the clamped umbilical cord, was collected in citrate-phosphate-dextrose. Samples were provided through the DKMS Cord Blood Bank of the University Hospital Dresden (Germany) with informed consent and ethics committee approval. For the comparison of antigen presentation efficiency at birth versus infancy, a second group of nine healthy newborns was recruited at the Institute of Diabetes Research (Klinikum rechts der Isar, University of Technology Munich) and prospectively followed until 1 year of age, with informed consent and ethics committee approval. CB follow-up of peripheral venous blood at a median infant age of 8.1 months (range, 7–10 months) was obtained. Sodium-heparinized peripheral venous blood samples from 10 healthy nondiabetic adults >18 years of age were provided by the German Red Cross GmbH Dresden (Germany).

CB mononuclear cells (CBMCs) and peripheral blood mononuclear cells (PBMCs) were isolated by density centrifugation over Ficoll-Hypaque. CD4⁺ T cells (>96% purity) and CD14⁺ monocytes (>95% purity) were enriched by negative (CD4) or positive isolation (CD14) using magnetic beads according to the manufacturer’s instructions (Miltenyi-Biotech, Inc., Auburn, CA). Subsequently, CD4⁺CD25⁺ T cells were depleted from the CD4⁺ T-cell fraction by positive selection (resultant purity >98% CD25⁻ of CD4⁺). For the newborn versus infant autologous mixed leukocyte assay, CD14⁺ monocytes (>95% purity) and CD3⁺ T cells (>97% purity) were obtained from CBMCs and PBMCs by negative magnetic bead isolation.

The following monoclonal antibodies were used for FACS staining: BD Biosciences: anti-CD45 APC (HI30), anti-CD4 APC (SK3), anti-CD4 Pacific Blue (RPA-T4), anti-CD3 APC (HIT3a), anti-CD127 PE (M21), anti-CD25 PE, APC-H7 (M-A251), anti-CD45RA APC (HI100), anti-CD45RO PE-Cy7 (UCHL1), anti-Helios PE (22F6), anti-CD14 FITC (M5E2), anti–HLA-DR PE (G46-6), anti-CD86 Pacific Blue (2331), and the Annexin V FITC apoptosis detection kit II; Bioscience: anti-CD127 eFlour450 (eBioRDR5); BioLegend (San Diego, CA): anti-CD40 AF647 (5C3) and anti-FOXP3 Alexa Fluor 488 (259D). After staining, cells were washed twice with FACS buffer and fixed with 1% paraformaldehyde in PBS. To detect intracellular protein expression, cells were fixed and permeabilized after surface staining using the FOXP3 Fix/Perm buffer set (BioLegend). Cells were acquired on a Becton Dickinson LSR-II flow cytometer with FACS Diva. Doublets and clumps were excluded based on SSC-A versus SSC-W plots. Live cell populations were gated as 7-AAD (BD Biosciences)–negative cells. At least 50,000 gated events were acquired for each sample and analyzed using FlowJo software version 7.6.1 (TreeStar, Inc., Ashland, OR).

Monocytes from CBMCs were cultured in six-well plates at 10⁶ cells/well in 3 mL dendritic cell (DC) medium (CellGenix, Freiburg, Germany) supplemented with 3% heat-inactivated FCS, 50 ng/mL GM-CSF, and 10 ng/mL IL-4 (R&D Systems, Minneapolis, MN) at 37°C. After 6 days, immature DCs were pulsed without antigen or with 100 μg/mL GAD65 (Diamyd, Stockholm, Sweden), proinsulin (Lilly, Indianapolis, IN), keyhole limpet hemocyanin (Calbiochem, Darmstadt, Germany), thyroglobulin (Lee Biosolutions, St. Louis, MO), or 10 μL/mL tetanus toxoid and diphtheria toxoid (TT/DPT; Sanofi-Pasteur MSD, Leimen, Germany) for 6 h and subsequently matured overnight with 100 units/mL IFN-γ (R&D Systems) and 10 ng/mL lipopolysaccharide (LPS; Sigma-Aldrich, St. Louis, MO). The mature antigen-loaded DCs were washed three times with PBS. DC phenotype was determined via FACS analysis to ensure maturation.

For CB T-cell responses, T cells (CD4⁺ or CD4⁺CD25⁻) were labeled with CFSE (Invitrogen, Carlsbad, CA) as previously described (9). CFSE-labeled T cells (1.5 × 10⁵) were added to autologous mature DCs loaded with or without antigen at a ratio of 1 DC:40 T cells in round-bottom 96-well plates. IL-7 (0.1 ng/mL; R&D systems) was added to the culture medium to increase the survival of T cells. For comparison of infant peripheral blood (PB) versus CB T-cell responses, 2 × 10⁵ CFSE-labeled CD3⁺ T cells were added to 2 × 10⁴ autologous monocytes previously pulsed without antigen or with 10 μg/mL GAD65, 10 μg/mL proinsulin, or 1 μL/mL TT for 3 h. In all assays, each condition was performed in triplicate. Cells were cultured in RPMI 1640 supplemented with 2 mmol/L glutamine, penicillin (50 units/mL), streptomycin (50 μg/mL), and 5% (volume/volume) heat-inactivated human AB serum (Invitrogen). After 6 days, the cell cultures were stained for CD4, CD25, and CD45RO and processed for FACS analysis.

To determine specificity of responses in the CFSE dilution assay, antigen-responsive CD4⁺CFSE^dimCD25⁺CD45RO⁺ and nonproliferating CD4⁺CFSE^hiCD25⁻ T cells were sorted on a FACSAriaII cell sorter and subsequently stimulated with autologous antigen-pulsed or unpulsed matured DCs (ratio 1:20 DC/T cell) in the presence of 0.1 ng/mL IL-7 for 3 days. [³H]thymidine (1 μCi/mL) was added for the last 16 h. Cells were subsequently harvested and collected onto a 96-well 1.2-μm pore size filter plate. [³H]thymidine incorporation was measured as counts per minute (cpm) using a Top Count Microplate Scintillation β-particle counter.

Plasma IL-7 was quantified by enzyme-linked immunosorbent assay using the commercial Duoset ELISA system (R&D Systems). The assay detection range was between 3.5 and 2,000 pg/mL.

HLA typing was performed by the DKMS Life Science Laboratory GmbH (Dresden, Germany) using sequence-based typing. INS gene typing was performed by HphI digestion of PCR products amplified from the INS promoter region of interest as previously described (10).

Results are presented as mean and SD or median and interquartile range (IQR) as indicated. Expression of cell surface molecules on cell subsets was assessed as median fluorescence intensity (MFI). For normally distributed data, Student t test for unpaired values was used to compare means between groups and Student t test for paired values was used to compare values for the same sample or subject tested under different conditions. Comparisons of other nonnormally distributed variables between paired groups were made using the nonparametric Wilcoxon matched pairs test. Multivariate analysis of covariates determining responsiveness of CD4⁺ T cells to antigen was performed using linear regression. For all tests, a two-tailed P value of <0.05 was considered to be significant. Statistical significance is shown as *P < 0.05, **P < 0.01, ***P < 0.001, or not significant (ns) P > 0.05. Analyses were performed using the programs GraphPad Prism 5 (La Jolla, CA) and the Statistical Package for the Social Sciences (SPSS 19.0; SPSS Inc., Chicago, IL).

We searched for unprimed islet autoreactive naïve T cells in CB using autoantigen-stimulated T-cell proliferation. To increase sensitivity, assays were performed using CD4⁺CD25⁺-depleted CD4⁺ T cells as responder cells (Supplementary Fig. 1A) and autologous DCs matured from CB monocytes as APCs. CD45RO⁺ memory T cells were not depleted since they were infrequent in CB (Supplementary Fig. 1B). Antigen-responsive cells (Fig. 1A) were identified as proliferative (CD4⁺CFSE^dim), activated (upregulation of CD25), and memory-like (CD45RO⁺) phenotype. Significant T-cell responses against GAD65 (stimulation index [SI]: median 3.4 [IQR 1.7–6.2], P = 0.0002) and proinsulin (2.1 [1.2–2.7], P = 0.002) as compared with medium were observed (Fig. 1B and Table 1). Responses to GAD65 and proinsulin were also increased when excluding the two highest responders for GAD65 or proinsulin (P = 0.0005 and P = 0.005, respectively). Little or no response was observed against the exogenous antigens TT/DPT (0.8 [0.3–2.1], P = 0.8). No response was detected against the vehicle in which the antigen GAD65 was solubilized (Supplementary Fig. 1C). Consistent with previous findings (11), significant proliferative responses in CB CD4⁺CD25⁻ T cells were observed to the highly immunogenic large antigen keyhole limpet hemocyanin (3.2 [1.5–8.4], P = 0.01). Some newborns also had responses to the thyroid autoantigen human thyroglobulin (Supplementary Fig. 1C), and CB T cells had vigorous responses to the superantigen Staphylococcal enterotoxin B (Supplementary Fig. 1D).

Some newborns had robust responses to GAD65 and/or proinsulin with up to 3% of cells having undergone proliferation at the end of culture. Although this is not the precursor frequency, it is substantial and likely includes bystander-type responsive cells as shown for the influenza vaccine (12,13). Therefore, to confirm that proliferating cells were enriched for antigen-specific T cells, GAD65-responding CFSE^dimCD25⁺CD45RO⁺CD4⁺ T cells were FACS sorted and rechallenged with antigen in a 3-day proliferation assay (Fig. 1C). The sorted cells responded specifically to GAD65 (mean ± SD, 15,737 ± 7,335 cpm) but not proinsulin (254 ± 267 cpm), and initially nonresponding T cells remained nonresponsive to autoantigen.

Naïve antigen-specific T cells are rare within the T-cell repertoire (11). We expected that the range of responses observed between children may partially reflect the naïve precursor frequency of antigen-specific T cells. Responses to the type 1 diabetes-associated autoantigens were >20-fold over background in three of the children tested (Table 1), suggesting a higher precursor frequency of T cells specific for these antigens in those children. These three children uniquely had two type 1 diabetes high-risk HLA-DRB1*HLA-DQB1* haplotypes, and responses to both GAD65 and proinsulin were higher in these neonates as compared with the remaining neonates (P = 0.006 for GAD65; P = 0.0004 for proinsulin).

Polymorphisms in the insulin gene (INS) promoter region are associated with inefficient deletion of proinsulin-reactive T cells in the thymus (14) and associated with responses to insulin (15). Consistent with this, all CB with an SI >3.0 for proinsulin (CB 1, 3, and 17) had the type 1 diabetes-susceptible INS VNTR I/I genotype. Multivariate analysis that included the HLA genotype (children with two susceptible haplotypes vs. others), INS VNTR genotype (children with I/I vs. others), and background proliferation in medium confirmed that the HLA genotype was a determinant of CB CD4⁺ T-cell responses to GAD65 (P = 0.006) and proinsulin (P = 0.002). Responses to GAD65 or proinsulin were not associated with the proportion of CD4⁺CD25⁺CD127^hi T cells or the proportion of CD4⁺CD25⁺CD127^low T regulatory cells (Tregs) present in the original CB sample (Supplementary Fig. 2), or the MFI of CD127 on either population (data not shown).

Having established that the newborn is not devoid of autoreactive T cells, we examined factors that may modify the susceptibility of these T cells to be activated at ages relevant to low (birth) and high (8 months of age) islet autoimmunity seroconversion periods. T-cell expansion in infants is high and is maintained by homeostatic mechanisms, including IL-7 (8,16). Consistent with this, serum concentrations of IL-7 were increased in CB (mean 6.4 ± 5.0 pg/mL) and PB of 8-month-old infants (8.6 ± 6.1 pg/mL) as compared with concentrations in adults (0.3 ± 0.7 pg/mL; P = 0.007 vs. CB; P = 0.003 vs. 8 months) (Fig. 2A). Furthermore, IL-7R α-chain surface expression was increased on CB CD4⁺ T cells (MFI 2,491 ± 162) as compared with adult naïve T cells (MFI 1,509 ± 244, P < 0.0001) (Fig. 2B).

Functionally, low concentrations of IL-7 were sufficient to prevent the marked CB T-cell apoptosis normally observed during in vitro culture (live CD4⁺ CB T cells: 60 ± 9.8% with 0.1 ng/mL IL-7 vs. 32 ± 17.9% without IL-7, P = 0.001) (Fig. 2C). Moreover, both CB and 8-month T cells were highly responsive to IL-7 in the absence of antigen as compared with adult T cells and showed strong proliferation already at 1.0 ng/mL (CFSE^dim of CD4⁺ T cells: 17.5 ± 3.3% CB; 13.8 ± 4.1% 8 month vs. 1.1 ± 0.4% adult PB naïve T cells, P < 0.0001) (Fig. 2D). Of note, exposing CB and 8-month T cells to high IL-7 (10 ng/mL) induced proliferation of both CD45RO⁻ (naïve) and CD45RO⁺ (memory-like) T cells, with the naïve proliferating T cells remaining CD45RO⁻ (Fig. 2E). In adult PB, however, proliferation in response to IL-7 was predominantly limited to the CD45RO⁺ T cells (CFSE^dim of CD45RO^-CD4⁺ T cells at 10 ng/mL of IL-7: 5.8 ± 2.5% vs. 24.2 ± 6.6% in CD45RO⁺ CD4⁺ T cells, P < 0.0001).

The immune system at birth is biased toward tolerance (17,18), and IL-7 may affect the Treg pool (19,20). We therefore examined whether culture of naïve CB CD4⁺CD25⁻ T cells with autoantigen presented by DCs with and without IL-7 leads to cells with Treg phenotypes, indicated as CD4⁺CD25⁺CD127^lowFOXP3⁺ and the Helios-negative subset of these cells. Helios negativity was used as a potential marker of peripherally induced Tregs (21), and it is noted that the large majority of uncultured CB Tregs were Helios positive (Supplementary Fig. 3A). We found that the maximum proportion of CD4⁺CD25⁺CD127^lowFOXP3⁺ cells among CD4⁺ T cells present at the end of culture was <2% with no influence by antigen stimulation or IL-7 (Supplementary Fig. 3B). Similarly, the proportion of Helios-negative Tregs among CD4⁺ T cells at the end of culture was <0.2% and not influenced by antigen or IL-7 (Supplementary Fig. 3B, right).

High responsiveness to IL-7 could facilitate islet-reactive T-cell activation (19) but would not explain the difference in low and high islet autoimmunity incidence periods seen in the first year of life. We therefore hypothesized that differences in immune competence may play a role, and separately compared APC competence and T-cell responsiveness in CB and at 8 months of age in healthy children.

Monocytes (APCs) and T cells from the same child isolated at birth and at 8 months of age were studied in mixed leukocyte assays (Fig. 3A). Responses to TT were similar to background when unprimed CB CD4⁺ T cells were stimulated by autologous CB monocytes (median SI, 1.0 [IQR 1.0–1.1]), whereas, as expected, tetanus vaccination–primed 8-month CD4⁺ T cells responded strongly to TT when stimulated by 8-month monocytes (191 [34–650], P = 0.007). The vaccine-primed 8-month CD4⁺ T cells also responded when CB monocytes were used for antigen presentation (132 [31–386]), consistent with an ability of CB monocytes to activate memory CD4⁺ T cells. In contrast, priming of naïve T cells against autoantigen by CB monocytes was impaired (2.7 [1.5–63]; P = 0.04). CB and 8-month CD4⁺ T cells from none of the five children were responsive to GAD65 when CB monocytes were used as APCs. However, both CB and 8-month CD4⁺ T cells from three children consistently responded to GAD65 when monocytes from the 8-month samples were used for presentation, and responses to GAD65 were increased when 8-month monocytes were used for presentation (median SI all five children, 2.95 [1.5–4.3]) as compared with CB monocytes (median SI, 1.05 [1.0–1.6]; P = 0.004). Moreover, variation in response to GAD65 was associated with the CB versus 8-month APCs (P = 0.001) and not the CD4⁺ T-cell source (P = 0.3) in a multivariate analysis. Analogous differences were also observed for proinsulin (data not shown). These data suggest that naïve T cells at birth can be activated by antigen but that the APC compartment at birth may be inefficient in its ability to activate and expand naïve T cells.

Priming of naïve CD4⁺ T cells requires strong costimulatory signals. Therefore, the expression of costimulatory molecules was analyzed on monocytes in CB, at 8 months, and in adults (Fig. 3B). Consistent with the impaired ability to activate naïve T cells, monocytes at birth had significantly lower surface expression of CD40 (MFI: 740 ± 409) and HLA-DR (MFI: 302 ± 154) than monocytes obtained at 8 months of age (MFI CD40: 3,663 ± 805, P = 0.0009; HLA-DR: 1,302 ± 408, P = 0.007) and from adults (MFI CD40: 3,500 ± 1,560, P = 0.0007; HLA-DR: 1,316 ± 549, P = 0.0005).

The presence of autoreactive T-cell clones in the T-cell repertoire is determined in the thymus, but activation and expansion of naïve clonal populations is regulated in the periphery by antigen encounter, homeostatic mechanisms, and the inhibitory network of Tregs. Here, we demonstrate that naïve CD4⁺ T cells responsive to the β-cell autoantigens GAD65 and proinsulin are present in the T-cell repertoire at birth at a frequency that appears to be associated with HLA class II genotype. Neonatal naïve T cells were highly sensitive to IL-7, but antigen presentation was inefficient at birth and rescued by 8 months of age. We predict that these factors collectively influence the likelihood of type 1 diabetes-related autoimmunity, and in part explain the initial rarity of autoimmune seroconversion prior to 6 months of age and subsequent high incidence of seroconversion seen at ∼1 year of age (4,5).

Autoreactive T cells in man are difficult to consistently and reliably demonstrate in vitro (22,23). We examined naïve CB CD4⁺ T cells from healthy newborns and chose conditions that should enhance activation and proliferation of naïve T cells, including the use of matured DCs for antigen presentation and removal of CD4⁺CD25⁺ Tregs. Importantly, we used stringent criteria for T-cell responsiveness. First, only proliferating cells (CD4⁺CFSE^dim) that expressed high levels of the IL-2Rα chain (CD25⁺) and were CD45RO⁺ at the end of culture were considered responsive. Second, we showed that isolated GAD65-responsive CB CD4⁺ T cells could be successfully and specifically rechallenged. Although we cannot be certain that in vitro responsive naïve autoreactive T cells have the potential to be activated by autoantigen in vivo, our findings strongly suggest that naïve T cells that can specifically respond to islet autoantigen are found in the T-cell repertoire of CB. This is not unreasonable since the thymus is active already in fetal life (24), and negative selection appears to allow the passage of low-affinity self-reactive T cells (25,26).

There was substantial heterogeneity between neonates in the responsiveness of their CB CD4⁺ T cells to islet autoantigens. This heterogeneity is informative with respect to mechanisms controlling autoreactive T-cell frequency. By examining the HLA genotype, we could show an association between the presence of high type 1 diabetes risk HLA class II genotypes and responsiveness, consistent with the role of HLA class II in thymic selection of CD4⁺ T cells. The striking response observed in the three neonates with full HLA class II risk genotypes suggests inefficient deletion of islet autoreactive T cells in the thymus and, as a consequence, a relative abundance in the periphery. This was observed for both proinsulin and GAD65. The limitation of these data is the small number of neonates studied, and therefore these data need to be confirmed on a larger set of samples and in neonates who subsequently develop islet autoimmunity. A further limitation is that we are unable to determine precursor frequencies from the in vitro assays since limiting dilution was not performed. Moreover, since some responses were robust, it is possible that the responsive cells also included bystander cells that proliferated to stimuli other than autoantigen.

Our study is potentially informative with respect to the early islet autoantibody seroconversion observed in children who develop type 1 diabetes (4). Homeostatic proliferation is discussed as a mechanism of autoreactive T-cell expansion (27–30). The immune system of an infant must adapt to rapid growth, and, consequently, neonatal T cells are exposed to active homeostatic signals (8,31). These include increased concentrations of circulating IL-7 and, as shown in our study, high sensitivity of naïve infant T cells to IL-7. Increased IL-7 concentrations also promote the activation and expansion of T cells in the presence of antigen (32), as well as Tregs (19,20). We hypothesize, therefore, that the increased homeostatic expansion pressure and the increased sensitivity to IL-7 increase the number of naïve autoreactive T-cell clones in the periphery of the type 1 diabetes-susceptible infants, and favor their activation during events of β-cell antigen exposure. In the current study, we were unable to find in vitro evidence for substantial conversion of naïve CD4⁺ T cells to Tregs in the presence of IL-7. However, it remains possible that this occurs and could be a preferred in vivo pathway in neonates, since IL-7 is essential for peripheral Treg survival and expansion in vivo (33).

Homeostatic T-cell expansion mechanisms in early childhood could help explain the higher incidence of islet autoimmunity observed in this period. However, such stimuli appear to be strongest in the first months of life where islet autoantibody seroconversion is rare. Thus, mechanisms that prevent very early autoimmunity must be ongoing. Tregs at birth appeared functional as judged by the ability to suppress the response to antigen in vitro in this (Supplementary Fig. 1) and previous studies (34). Equally potent as an autoimmune brake is the state of immune competence of APCs at birth. We demonstrated that the monocyte compartment of CB is functionally impaired to prime naïve T cells. This is consistent with previous findings that mononuclear cells at birth are not readily activated (35), APCs in neonates are immature (36,37), and that early after birth, the immune system is relatively tolerant (38). Relevant to the poor priming, CB monocytes had decreased expression of key molecules involved in antigen presentation as compared with adult monocytes. It is noted that we have not demonstrated that DCs circulating at birth also have similar deficiencies in antigen presentation, but others have reported such defects (39,40). Relevant to the heightened incidence of seroconversion after 6 months of age, monocytes obtained from blood samples at 8 months strongly expressed HLA-DR, CD40, and CD86 and appeared to be fully competent in their ability to prime naïve CD4⁺ T cells. It is also demonstrated that Toll-like receptor signaling in the first months of life favor IL-10 production, potentially reducing the risk to activate autoreactive T cells (18,41). In addition, it is possible that maternal transfer of APCs may further modify the risk of activating autoreactive T cells (42). Although our findings are consistent with a strong role of APCs in compromising T-cell responsiveness in neonates, we cannot exclude additional deficiencies in birth T cells as have been recently reported (43).

In conclusion, we demonstrate that naïve islet autoantigen-responsive CD4⁺ T cells are present from birth in healthy newborns and that their frequency appears to be increased in neonates with type 1 diabetes-susceptible HLA genotypes. Our findings predict that homeostatic mechanisms of T-cell expansion in infants will increase the likelihood of islet autoimmunity, and that inefficient antigen presentation will operate to reduce this risk in the first months of life. We propose that these features are relevant to the timing of islet autoantibody seroconversion in the pathogenesis of type 1 diabetes, and provide an opportunity when designing intervention trials to prevent the development of autoimmunity against β-cells.

This work was supported by funding from the Competence Network for Diabetes Mellitus funded by the Federal Ministry of Education and Research (KFZ 01GI0807 and 01GI1102), the DFG Research Center and Cluster of Excellence, Center for Regenerative Therapies Dresden, (FZ 111), and the Helmholtz Center Munich. A.-K.H. was enrolled in the International PhD program “Metabolism and Endothelium” as part of the Medical Faculty of the Dresden University of Technology. P.M. is supported by an advanced postdoctoral fellowship from the Juvenile Diabetes Research Foundation (10-2009-284).

No potential conflicts of interest relevant to this article were reported.

A.-K.H. designed and performed experiments, analyzed data, conceived ideas, and wrote the manuscript. P.M. conceived ideas, contributed to discussion, and edited and reviewed the manuscript. C.W. provided technical assistance. P.S. organized the CB and 8-month blood samples from healthy children. D.K. performed all experiments documenting the Treg phenotype. A.-G.Z. provided blood sample collection, contributed to discussion, and edited and reviewed the manuscript. E.B. conceived ideas, oversaw research, and helped write the manuscript. E.B. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

The authors thank Dr. A. Platz (DKMS Cord Blood Bank) for providing CB samples, the midwives of the Technical University Hospital Rechts der Isar (Munich, Germany) for sample collection, K. Ruhner (DKMS Life Science Laboratory GmbH) for HLA typing, and all subjects for their blood donation.