Materno-Fetal Transfer of Preproinsulin Through the Neonatal Fc Receptor Prevents Autoimmune Diabetes

Islet destruction by autoreactive T cells is the hallmark of type 1 diabetes (T1D). Intense research efforts are therefore ongoing to develop immunotherapies aimed at blunting islet autoimmunity. Antigen (Ag)-specific immunotherapies are particularly attractive due to their selectivity and safety (1) but have met with limited success. Several attempts have focused on tolerogenic vaccination with β-cell Ags derived from preproinsulin (PPI) (2), which is the target initiating the autoimmune cascade in nonobese diabetic (NOD) mice (3) and likely also in humans (2). A recent clinical trial employing intranasal insulin in slow-onset T1D patients did not result in C-peptide preservation, despite successful induction of insulin-specific immune tolerance (4). These results suggest that the timing of intervention may be too late and that the Ag spreading that follows early β-cell destruction leads to a diversification of autoimmune reactions beyond insulin, thus making tolerance restoration to this sole Ag insufficient. The same problem is encountered in prevention trials. Despite absence of clinical disease, selection of at-risk patients based on positivity for multiple autoantibodies (auto-Abs) underscores the presence of an autoimmune reaction that has already spread to several Ags (5). Prospective cohorts of genetically at-risk children further highlighted that β-cell autoimmunity initiates very early, possibly already during fetal life, as the median age at auto-Ab seroconversion is only 9–18 months (6,7).

The corollary to these observations is that prevention strategies should be implemented much earlier, i.e., in children carrying a high HLA-associated genetic risk but with no sign of active autoimmunity (i.e., auto-Ab⁻) (8). The perinatal period offers such opportunities not only in terms of timing but also because it is characterized by immune responses to introduced Ags that are biased toward tolerogenic outcomes. Indeed, Ag introduction during fetal life results in Ag-specific immune tolerance persisting during adulthood (9,10). A key role in this process is played by central tolerance, since thymic negative selection of autoreactive effector T cells (Teffs) and positive selection of regulatory T cells (Tregs) is very active during this period and defines the immunological self that later imprints peripheral immune responses (11).

The richly vascularized placental interface is well suited for translocating appropriately modified Ags from the mother to the fetus by active receptor-mediated transcytosis across the syncytiotrophoblast (12). One ferrying system is provided by the neonatal Fc receptor (FcRn), which delivers maternal IgGs (13). We therefore asked whether maternal administration of PPI through a chimeric Fc-fused PPI protein (PPI-Fc) could be used to transplacentally deliver PPI in utero in order to induce tolerance and subsequently protect from diabetes development. We show that PPI-Fc administered to pregnant mice is transferred to fetuses and ferried to the thymus by migratory dendritic cells (DCs), preventing subsequent diabetes development.

Sequences encoding PPI1 and PPI2 were PCR amplified from pancreatic and thymic cDNA, respectively, obtained from an 8-week-old nondiabetic NOD mouse (see Supplementary Table 1 for primer sequences and cloning, expression, and purification strategies). The anti-CD20 rituximab monoclonal Ab (mAb) (Roche) was used as IgG1 control.

Kinetics constants of interactions between mouse or human FcRn and PPI1-Fc, PPI2-Fc, or IgG1 were determined using Biacore 2000 (GE Healthcare) as detailed in Supplementary Fig. 1.

G9Cα^−/−.NOD (G9C8) and NOD 8- to 15-week-old primiparous pregnant mice, housed in specific pathogen-free conditions, were retro-orbitally injected with 100 µg PPI-Fc (an equimolar mixture of PPI1-Fc and PPI2-Fc), with equimolar amounts of IgG1 or PPI, or with PBS vehicle alone at embryonic day (E)16. After birth, 3.5-week-old G9C8 newborns were immunized with 50 µg PPI_B15–23 peptide and 100 µg CpG (14) and boosted 2 weeks later. Diabetes was monitored by glycosuria and confirmed by hyperglycemia when positive. For FcRn and vascular cell adhesion molecule (VCAM)-1 blocking experiments, PPI-Fc treatment was performed 24 h after intravenous injection of 100 µg IgG (rituximab) or anti–VCAM-1 mAb (clone M/K2.7, produced in-house). For transfer experiments, 15 × 10⁶ splenocytes from the 14-week-old offspring of treated NOD mice were adoptively transferred into 4- to 6-week-old NOD.scid recipients and their pancreata were recovered for insulitis scoring as described (15). The study was approved by Comité d’Ethique pour l’Expérimentation Animale (P2.RM.117.09, CEEA34.SC.158.12).

PPI-Fc and PPI proteins were conjugated with Alexa Fluor (AF)680 using a SAIVI Rapid Antibody/Protein labeling kit (Invitrogen). G9C8 and β₂m^−/− primiparous pregnant mice (E18) were retro-orbitally injected with 100 µg PPI-Fc, equimolar amounts of PPI, or PBS vehicle. Fluorescence was detected using the Fluobeam imaging system (Fluoptics) at a 690-nm excitation and >700 nm emission wavelengths, with 50–100 ms exposures. After imaging, blood and urine were collected for ELISA quantification, with standard curves obtained by sequential dilutions of PPI-Fc and PPI proteins. Both PPI-Fc and PPI were captured with plate-coated H-86 anti-insulin Ab (Santa Cruz Biotechnology). PPI-Fc was detected with a horseradish peroxidase–labeled goat anti-human Fc antibody (SouthernBiotech). PPI was revealed with an anti-proinsulin mAb (KL-1; kindly provided by Dr. L. Harrison, Walter and Eliza Hall Institute of Medical Research, Parkville, Australia).

Bone marrow–derived DCs prepared from 6-week-old G9C8 mice were pulsed for 8 h with 26 μmol/L PPI_B15–23, PPI-Fc, or PPI. After maturation with 100 ng/mL lipopolysaccharide, they were cocultured for 5 days with carboxyfluorescein succinimidyl ester (CFSE)-labeled splenocytes from the 7-week-old offspring of untreated G9C8 mice. Upon staining with PerCP-eFluor 710–labeled anti-Vβ6, AF700-labeled anti-CD8a, and APC-eFluor 780–labeled anti-CD3ε mAbs (eBioscience); Brilliant Violet (BV)605-labeled anti-CD4 mAb (BioLegend); and Live/Dead Red (Invitrogen), cells were analyzed on a 16-color BD LSR Fortessa. Real-time cytotoxicity assays were performed with the xCELLigence system (ACEA Biosciences). Briefly, mouse fibroblast L cells were plated on 96-well E-plates, irradiated (5,000 rad), and rested for 2 h. FACS-sorted CD8⁺ T cells were added at a 10:1 effector:target ratio in the presence of 10 nmol/L PPI_B15–23 peptide, and impedance was recorded every 5 min for 2 h and then every 15 min for an additional 3 h.

The following mAbs were used: phycoerythrin-labeled anti-Foxp3 and APC-eFluor 780–labeled anti-CD3ε, APC-labeled anti–neuropilin 1 (NRP1) (R&D Systems), and BV421-labeled anti-CD62L and anti–transforming growth factor (TGF)-β latency-associated peptide (LAP) (clone TW7–16B4), BV570-labeled anti-CD44, BV605-labeled anti-CD4, and BV711-labeled anti-CD8a (BioLegend). Cells were additionally stained with Live/Dead Red and BV650-labeled K^d multimers loaded with PPI_B15–23 (LYLVCGERG) or control tumor-derived (TUM) peptide (KYQAVTTTL) as previously described (16).

For quantitative real-time PCR, G9C8 pregnant mice were retro-orbitally injected with 100 µg PPI-Fc or PBS vehicle at E16. After birth, G9C8 offspring were prime boosted with 50 µg PPI_B15–23 and 100 µg CpG at 3.5 and 5.5 weeks. Blood was collected either before immunization or at days 0, 5, 19, and 30 after priming. Peripheral blood mononuclear cells were stained with APC-eFluor 780–labeled anti-CD3ε, BV605-labeled anti-CD4, and BV711-labeled anti-CD8a and sorted on a BD FACSAria III at 10 CD4⁺ or CD8⁺ cells/well into PCR plates. RNA was extracted by direct lysis for 2 min at 65°C, and multiple genes were coamplified as previously described (17) by seminested PCR with the primers listed in Supplementary Table 2. mRNA expression was normalized to Cd3e.

AF647-conjugated PPI-Fc (100 μg i.v.) was injected into pregnant G9C8 mice at E19. Newborns were killed 24 h later and thymi, spleens, and blood from two to four mice were pooled together. For thymi, single-cell suspensions were obtained by enzymatic digestion (18). PPI-Fc⁺ events were identified by gating on live cells of each subset. For evaluation of the migratory capacity of different DC subsets to the thymus, hemolyzed total blood cells from 1-day-old G9C8 newborns were transferred into 5-week-old NOD.scid mice. After 24 h, mice were killed and isolated thymic cells stained for enumeration of DC subsets.

Data from separate experiments are depicted as means ± SEM. Statistical significance (P ≤ 0.05) was assigned with the two-tailed tests detailed in each figure legend using GraphPad Prism 5.

Unlike humans, mice harbor two Ins genes: Ins1 is predominantly expressed in the pancreas, while Ins2 is expressed in the thymus (2). Ins1 and Ins2 were therefore fused with the N-terminus of the CH2–CH3 Fc domain from human IgG1 to obtain PPI1-Fc and PPI2-Fc fusion proteins (Supplementary Fig. 1A and B). Since FcRn is crucial for Fcγ-dependent transcytosis across the placenta, we evaluated the binding affinities of PPI1-Fc and PPI2-Fc on immobilized murine and human FcRn using surface plasmon resonance (Supplementary Fig. 1C). Both proteins displayed efficient binding (Table 1), with a slightly higher affinity for mouse (K_D ∼6 nmol/L) than for human (K_D ∼15 nmol/L) FcRn.

For in vivo studies, we employed the G9C8 mouse (14), which expresses a transgenic T-cell receptor (TCR) derived from the diabetogenic G9C8 CD8⁺ T-cell clone (19) recognizing the H-2K^d–restricted PPI_B15–23 epitope. These mice develop diabetes rapidly (4–8 days) after PPI_B15–23 peptide immunization with CpG adjuvant (14). Since the PPI_B15–23 epitope is shared between PPI1-Fc and PPI2-Fc, a 1:1 mix of both proteins (hereafter designated PPI-Fc) was used for subsequent experiments. For assessment of the efficiency of placental transfer, 100 µg i.v. AF680-labeled PPI-Fc were injected into pregnant G9C8 mice at E18. In vivo imaging demonstrated selective PPI-Fc accumulation in the uterine horns (Fig. 1A) and fetuses (Fig. 1B) 24 h after injection. This transfer was Fc dependent, as injection of Fc-devoid PPI into pregnant G9C8 mice led to its rapid (within 1 min) renal accumulation without detectable placental transfer (Fig. 1A and B). Furthermore, interaction with the FcRn was also required, since PPI-Fc administration to β₂m^−/− mice devoid of functional FcRn expression (13) did not result in any detectable transfer (Fig. 1A and B), as previously observed with FcRn^−/− mice (20). Interestingly however, PPI-Fc was detectable at the vascularized placental interface (Fig. 1B), suggesting that fusion to the Fc domain stabilizes PPI and increases its half-life. Indeed, PPI-Fc fluorescence was still detectable in 7-day-old newborn G9C8 mice, i.e., 9 days after administration to their pregnant mothers at E18 (Fig. 1C).

Quantitative ELISA measurements of intact PPI-Fc were subsequently performed. While serum PPI-Fc concentrations became barely detectable within 24 h after injection in G9C8 pregnant mice (Fig. 1D), they remained stable for 48 h in their fetuses, reaching concentrations of ∼0.75 ng/µL and documenting PPI-Fc integrity after transfer. No serum PPI-Fc accumulation was observed in either β₂m^−/− pregnant mice or their fetuses. Analyses of urine PPI-Fc from pregnant females gave symmetrical results (Fig. 1E): G9C8 mice excreted limited amounts, mostly during the first hours after injection, while β₂m^−/− mice continued to excrete PPI-Fc even 24 h postinjection. Similarly, an ELISA for PPI detected rapid and steady PPI urinary excretion in PPI- treated but not PPI-Fc–treated G9C8 pregnant mice (Fig. 1F).

Taken together, these data indicate that efficient PPI-Fc transplacental transfer is dependent on Fc-FcRn binding. Since TCR expression is first detected in the thymus at E17 (21) and given that maternally administered PPI-Fc persisted in the fetal circulation for at least 48 h and remained detectable in newborn mice, PPI-Fc was injected into pregnant G9C8 mice with a single 100 µg dose at E16 for subsequent experiments.

G9C8 mice harbor increased proportions of splenic CD8⁺ T cells and reduced CD4⁺ T cells compared with nontransgenic NOD mice (14) (Supplementary Fig. 2). As in NOD mice (22), ∼15% of CD4⁺ T cells are Foxp3⁺ Tregs but with higher NRP1⁺ thymic-derived Treg fractions (∼90% of total Tregs vs. ∼70% in NOD mice) (23). Both CD4⁺ and CD8⁺ T cells express the transgenic Vβ6 chain, but only CD8⁺ T cells stain with PPI_B15–23–loaded K^d multimers (Supplementary Fig. 2). In vitro CFSE proliferation assays showed that G9C8 CD8⁺ T cells are stimulated by PPI-Fc but not by Fc-devoid PPI (Fig. 2A), hence demonstrating efficient PPI_B15–23 cross-presentation. CD4⁺ T cells proliferated upon stimulation with both PPI-Fc and PPI (Fig. 2B).

Since PPI-Fc is transferred through the placenta and cross-primes G9C8 TCR-transgenic T cells in vitro, pregnant G9C8 mice were treated with 100 µg PPI-Fc at E16. After delivery, their offspring were immunized with PPI_B15–23 peptide and CpG at 3.5 and 5.5 weeks of age to induce diabetes and prospectively followed. For controls, equimolar amounts of recombinant IgG1 (i.e., irrelevant protein with preserved FcRn binding), PPI (i.e., cognate Ag with no FcRn binding), or PBS vehicle were injected. Diabetes development was rapid and synchronous in the offspring of control-treated mice, mostly within 1 week after prime immunization (Fig. 2C). In contrast, the offspring of PPI-Fc–treated mice were significantly protected, showing reduced and delayed diabetes incidence (70% diabetes-free mice vs. 22–27% at the end of the 30-day follow-up; P < 0.0001). Since no difference was observed for IgG1, PPI, and PBS groups, PBS vehicle alone was used as control for subsequent experiments.

We next asked whether PPI-Fc priming of G9C8 TCR-transgenic T cells also occurred in vivo after transplacental transfer and diabetes induction by PPI_B15–23 prime-boost immunization. Indeed, increased frequencies of splenic CD8⁺ T cells were observed in the 7-week-old offspring of PPI-Fc–treated mice (Fig. 2D) (8.1 ± 0.5% vs. 6.4 ± 0.3% in control-treated animals; P = 0.01), which were limited to the memory (CD44⁺) subset (Fig. 2E) (10.4 ± 2.1% vs. 5.1 ± 0.8%; P = 0.02), while naïve (CD62L⁺CD44⁻) fractions were similar irrespective of treatment. The limited size of this memory CD8⁺ fraction (5–10% of total CD8⁺ T cells) suggests that PPI_B15–23 prime-boost immunization is relatively inefficient at recruiting G9C8 TCR-transgenic CD8⁺ T cells, probably because of their low avidity (24) and that prior PPI-Fc maternal treatment enhances such recruitment.

The increased frequency of CD8⁺ T cells in the offspring of PPI-Fc–treated mice was opposite what was expected in light of the protective effect of PPI-Fc on diabetes development. We therefore analyzed the phenotype of circulating CD8⁺ T cells in the progeny of PPI-Fc– and PBS-treated mice at different time points before and after PPI_B15–23 immunization by quantitative real-time PCR (Fig. 3A). While undetectable before PPI_B15–23 immunization, the expression of granzyme A (Gzma), perforin (Prf1), and Fas ligand (Fasl) was increased after immunization and more so in PBS-treated than in PPI-Fc–treated mice. Conversely, TGF-β receptor 2 (Tgfbr2) expression was increased in the PPI-Fc–treated group but not in the PBS-treated group. In vitro cytotoxicity assays under limiting (10 nmol/L) PPI_B15–23 peptide concentrations confirmed that CD8⁺ T cells from PPI-Fc–treated mice were less cytotoxic (Fig. 3B). Taken together, these results show that prior maternal PPI-Fc treatment imprints the phenotype of later CD8⁺ T-cell responses, making them less cytotoxic and more prone to TGF-β–mediated regulation (25).

To identify potential sources of TGF-β, we analyzed splenic CD4⁺ T cells. Total CD4⁺ T-cell numbers were not significantly different between treatment groups (Fig. 2D). However, Foxp3⁺ Tregs were more abundant in the offspring of PPI-Fc–treated mothers (Fig. 3C) (25.7 ± 3.6% vs. 17.0 ± 2.5% in control-treated animals; P = 0.05), without significant differences in Foxp3⁻CD4⁺ T cells. This Treg increase was exclusively made up by thymic-derived (NRP1⁺) Foxp3⁺ Tregs (Fig. 3D) (18.8 ± 3.3% vs. 13.7 ± 3.5%; P = 0.0003), while the percentage of peripheral (NRP1⁻) Tregs was similar in both treatment groups (5.3 ± 3.9% vs. 4.4 ± 2.8%), as was expression of surface TGF-β LAP in FoxP3⁺CD4⁺ Tregs after in vitro activation (Fig. 3E and data not shown). Finally, quantitative real-time PCR analysis on circulating CD4⁺ T cells showed a higher TGF-β1 (Tgfb1) expression in the progeny of PPI-Fc–treated mice (Fig. 3F) (0.21 ± 0.09 vs. 0.05 ± 0.00; P = 0.03), which, in light of the Treg-specific TGF-β LAP expression (Fig. 3E), can be attributed to the increased Treg numbers observed in PPI-Fc–treated mice. Taken together, these data show that maternal PPI-Fc treatment protects G9C8 newborns from diabetes development and that such protection is associated with increased priming of CD8⁺ T cells that are less cytotoxic and with an enrichment in thymic-derived Tregs expressing TGF-β.

Given the observed effect of PPI-Fc on thymic-derived Tregs, we investigated whether fluorescence-labeled PPI-Fc was capable of reaching the thymus. Twenty-four hours after injection into pregnant mice at E18, PPI-Fc was readily detected in fetal thymi, whereas PPI-treated mice showed no signal (Fig. 4A). No fluorescence was detected in the spleen. In line with the in vivo imaging data (Fig. 1C), PPI-Fc was still weakly detectable in the thymi of 5-day newborn mice, i.e., 7 days after PPI-Fc maternal treatment (Fig. 4A).

Next, we asked whether Ag-presenting cells were responsible for ferrying PPI-Fc to the thymus. A population of migratory CD8^lowCD11b⁺SIRPα⁺ conventional (c)DCs is known to transport blood-borne Ags to the thymus and promote central tolerance via negative selection of Ag-specific Teffs and Treg positive selection (26,27). CD11c^intermediateB220⁺PDCA-1⁺ plasmacytoid (p)DCs have also been suggested to ferry peripherally acquired Ags and participate in central tolerance (28). Hence, we first determined the DC subsets capable of migrating to the thymus. Total blood cells from neonatal G9C8 mice were injected into 6-week-old NOD.scid mice. Thymi were removed 24 h later and DC subsets analyzed. (See gating strategy in Supplementary Fig. 3.) Migratory SIRPα⁺ cDCs were significantly enriched in the thymi of adoptively transferred mice (Fig. 4B and C) (0.67 ± 0.54% vs. 0.02 ± 0.02% in control mice; P = 0.05), while thymic resident (CD8^highCD11b⁻Sirpα⁻) cDCs and pDCs were not.

For verification of whether migratory cDCs were capable of uptaking and ferrying PPI-Fc to the thymus, fluorescently labeled PPI-Fc was injected into pregnant G9C8 mice 24 h before delivery (E19). Thymi were then removed from their newborns and analyzed for PPI-Fc fluorescence in different thymic subsets (Fig. 4D). Only SIRPα⁺ cDCs carried PPI-Fc in ∼11% of cells, while neither other DC subsets nor medullary thymic epithelial cells (CD45⁻EpCAM⁺CDR1⁻) displayed any detectable fluorescence. Moreover, SIRPα⁺ cDCs were loaded with PPI-Fc not only in the thymus but also, to a lesser extent, in peripheral blood (13.1% vs. 6.8%) (Fig. 4E), suggesting that PPI-Fc is uptaken in the periphery and subsequently ferried to the thymus. When other thymic, blood, and spleen subsets (Supplementary Fig. 4) were analyzed, SIRPα⁻ cDCs, B cells, and macrophages were also loaded with PPI-Fc in peripheral blood (6.7%, 3.5%, and 1.9%, respectively), but only B cells displayed some fluorescence in the thymus (2.5%). In line with the results of ex vivo whole-organ imaging (Fig. 4A), PPI-Fc uptake was negligible in the spleen.

We then asked whether FcRn-mediated PPI-Fc transplacental transfer and SIRPα⁺ cDC migration were responsible for the protective effect of PPI-Fc on diabetes development. Pregnant mice were intravenously injected 24 h prior to PPI-Fc treatment with either an IgG isotype control, in order to compete with PPI-Fc for FcRn binding (13), or with an anti-VCAM-1 mAb, since SIRPα⁺ cDC migration is dependent on very late Ag-4 (VLA-4)–VCAM-1 interactions (26). As before, diabetes was then induced in the offspring by prime-boost PPI_B15–23 immunization. While PBS pretreatment did not reduce the PPI-Fc protective effect in the offspring (Fig. 4F), the isotype control IgG partially inhibited this protection (49% vs. 71% diabetes-free mice; P = 0.04). More strikingly, anti–VCAM-1 mAb pretreatment completely abolished the PPI-Fc diabetes protection, with only 18% of mice remaining diabetes free (P < 0.0001 and P = 0.01 compared with PBS and isotype mAb pretreatment, respectively). Although VCAM-1 is also essential for lymphocyte homing to inflamed tissues, including islets (29,30), the early single-dose treatment employed is unlikely to retain a blocking effect on islet infiltration, since it was administered 4 weeks before diabetes induction.

Taken together, these results show that PPI-Fc is ferried to the thymus by migratory SIRPα⁺ cDCs and that both transplacental delivery through FcRn and cDC migration are needed for PPI-Fc–mediated diabetes protection.

Finally, we evaluated whether PPI-Fc could prevent diabetes in polyclonal NOD mice. We injected 200 µg i.v. PPI-Fc or PBS into pregnant NOD mice at E16. The prediabetic female progeny of these mice were sacrificed at 14 weeks, and their splenocytes adoptively transferred into 4- to 6-week-old NOD.scid mice. Pancreata from donor NOD mice recovered for insulitis scoring displayed milder islet infiltration in females born from mothers treated with PPI-Fc compared with controls (Fig. 5A) (P = 0.007). This was paralleled by a significantly lower diabetogenic potency of splenocytes from the offspring of PPI-Fc–treated NOD mice (Fig. 5B). While, in line with previous reports (31), 60% of NOD.scid recipients receiving splenocytes from control NOD donors developed diabetes, only 37% of those adoptively transferred from PPI-Fc–treated animals became diabetic (P = 0.04). Taken together, these data show that PPI-Fc transplacental delivery blunts the insulitis and the splenocyte diabetogenic activity of polyclonal NOD mice.

The tolerogenic Ag vaccination strategies explored to date for T1D have targeted peripheral tolerance mechanisms (1). Here, we undertook a different strategy to target the earliest checkpoint in autoimmune progression, namely, the development of central tolerance in the thymus. Previous reports suggest that it is possible to “upgrade” central tolerance by administering Ags either intrathymically (32) or in the periphery (33). In the latter case, a key role is played by migratory DCs that ferry these Ags to the thymus (26–28), with a direct thymic entry of soluble Ags also documented (34,35). We aimed at translating this concept into a therapeutically viable strategy.

Several lines of evidence show that defective central tolerance is involved in T1D development. First, Ins2^−/− NOD mice develop accelerated diabetes (36) owing to absent thymic PPI expression (37). Second, the human INS variable number of tandem repeats polymorphic region, which ranks as the second most powerful T1D susceptibility locus after DQB1 (2), modulates INS expression in the thymus (38). However, this knowledge has not translated into therapeutic strategies aimed at boosting central tolerance ab initio. The notion that autoimmune activation against PPI appears already during the first 9–18 months of life, as witnessed by anti-insulin auto-Abs (6,7), lends further rationale to these strategies.

Transferred through the placental FcRn pathway, which physiologically delivers maternal IgGs (13), PPI-Fc fusion proteins were efficiently delivered to fetuses upon administration to pregnant G9C8 mice. The mechanism was Fc-FcRn dependent, since delivery did not occur in the absence of either and diabetes protection was inhibited with excess IgG. Subsequent ferrying of PPI-Fc to the thymus by migratory SIRPα⁺ cDCs (26,27) was also essential, since diabetes protection was lost when cDC migration was inhibited. Surprisingly, transplacental PPI-Fc delivery resulted in enhanced rather than decreased recruitment of CD8⁺ Teffs in the periphery upon PPI_B15–23 immunization. However, these CD8⁺ Teffs were less cytotoxic. The low-affinity G9C8 TCR and reduced PPI_B15–23 availability due to the requirement for PPI-Fc cross-presentation may favor CD8⁺ Teff expansion and limit the effect on thymic negative selection. Nonetheless, this low TCR affinity was sufficient to promote thymic positive selection of TGF-β1–expressing CD4⁺ Tregs (39), possibly regulating more efficiently CD8⁺ Teffs, which expressed higher Tgfbr2 levels. This latter finding is reminiscent of data in both NOD mice (22) and T1D patients (40), showing that Teff susceptibility to Treg suppression is a key parameter for immune tolerance and is reduced in T1D.

In summary, the therapeutic mechanism was dependent on FcRn-mediated PPI-Fc transfer and cDC migration to the thymus and resulted in impaired Teff cytotoxicity and enhanced selection of thymic Tregs. Moreover, we recently applied a similar strategy of transplacental Ag-Fc administration in the CD4⁺ hemagglutinin (HA)_110–119 TCR-transgenic 6.5 mouse model (41) to fully dissect therapeutic mechanisms (20). HA-Fc ferrying to the thymus was also observed in this model, and three differences were highlighted. First, HA-Fc was uptaken by SIRPα⁺ cDCs and, to a lesser extent, by macrophages. This may be due to the higher molecular weight of HA-Fc (65 vs. 38 kDa for PPI-Fc) and by HA interaction with different cell types through sialic acid moieties expressed on cell membranes, independent of Fc. Second, marginally reduced rather than increased CD4⁺ Teffs were observed in the periphery (but not in the thymus). This discrepancy may be due to peripheral effects mediated by HA-Fc–loaded macrophages and to the higher affinity of the HA_110–119 TCR, which may favor activation-induced apoptosis upon high-dose Ag encounter. Third, both thymic-derived and peripheral Ag-specific Tregs were induced. This points again to additional effects on peripheral tolerance mechanisms (10), which cannot be ruled out for PPI-Fc and will be further investigated.

The diabetes protection afforded by transplacental PPI-Fc delivery is noteworthy when considering the challenges posed by the G9C8 model, namely, disease aggressiveness harnessed through PPI_B15–23 immunization and the need for PPI-Fc cross-presentation to exert effects on CD8⁺ Teffs. Moreover, a single 100 µg PPI-Fc dose was sufficient to confer protection. This was likely favored by the Fc moiety conferring enhanced stability (42), since PPI-Fc remained detectable in the offspring as long as 9 days after maternal treatment. Another key issue was whether inducing tolerance to PPI alone would be sufficient to impact a polyclonal autoimmune T-cell repertoire. Adoptive transfer of NOD splenocytes suggests that this is the case. Follow-up studies are needed to explore whether such protection is maintained in NOD mice prospectively observed for diabetes development. Of further note, applications could reach beyond autoimmunity, as this strategy was also effective at promoting neo–Ag-specific tolerance toward clotting factor VIII (FVIII) in FVIII^−/− mice challenged with therapeutic FVIII (20).

The possibility of employing a single PPI-Fc dose to induce long-lasting tolerance is attractive for translation to genetically at-risk children. From this perspective, a puzzling observation is that the risk conferred by T1D mothers is half that transmitted by T1D fathers (3–4% vs. 6–8% at 20 years) (43). This relative protection seems linked to transplacental transfer of maternal auto-Abs (44). We could hypothesize that one mechanism for this protection may be the transfer of Ab-bound islet Ags through placental FcRn, similar to what observed with PPI-Fc.

A combination of several Ag-Fc therapeutics could be used to induce broad immune tolerance (45), and islet Ags displaying defective thymic expression (39) may be particularly suitable to this end. Given its initiating role (2), PPI remains an Ag of choice, and enrollment of newborns based on expression of T1D-susceptible INS variable number of tandem repeat alleles may be considered. However, a less invasive administration route is desirable for clinical translation. In this respect, the high FcRn expression in the gut epithelium may allow exploitation of the oral route directly in newborns. Of further note, intestinal FcRn expression persists throughout life in humans, possibly widening the time window of intervention. This raises another question, i.e., whether thymic Ag “upgrading” would still be effective when applied later in life. Albeit reduced, thymopoiesis remains active during human adulthood (46–48). Experimental autoimmune encephalomyelitis models further show that subcutaneous or oral immunization of 4- to 8-week-old mice with myelin peptides promotes Ag-specific Teff depletion and, to a larger extent, thymic Treg selection, leading to long-lasting tolerance (49,50). Follow-up studies are warranted to explore PPI-Fc oral vaccination.

Funding and Duality of Interest. This work was performed within Département Hospitalo Universitaire (DHU) AUToimmune and HORmonal diseaseS and supported by the INSERM Avenir program and by grants from Agence Nationale de la Recherche (ANR-10-BLAN-1118), the European Foundation for the Study of Diabetes/JDRF/Novo Nordisk European Programme in Type 1 Diabetes Research 2009, Association pour la Recherche sur le Diabète 2012, Aviesan/AstraZeneca “Diabetes and the vessel wall injury” program, and Fondation pour la Recherche Médicale (Equipe FRM DEQ20140329520). No other potential conflicts of interest relevant to this article were reported.

Author Contributions. S.C., N.G., S.L.-D., and R.M. planned the work. S.C., R.B., G.A., M.-C.G., J.D., and M.A. performed experiments. T.O., S.J., S.L., B.K., S.B., and F.S.W. contributed essential material and protocols. S.C., N.G., R.B., J.D., M.A., S.L.-D., and R.M. analyzed data. S.C. and R.M. wrote the manuscript. N.G., B.K., S.B., F.S.W., and S.L.-D. critically reviewed the manuscript. S.C. and R.M. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the 49th Annual Meeting of the European Association for the Study of Diabetes, Barcelona, Spain, 23–27 September 2013; at 5th Rencontres Internationales de la Recherche, Paris, France, 24 October 2013; and at the International Immunology of Diabetes Society Congress, Lorne, Australia, 7–11 December 2013.