PPI-Fc and PPI constructs (containing the murine preproinsulin-1 sequence) were previously described (39). An OVA-Fc construct covering the OVA_243-350 sequence was further developed ( Supplementary Figure 1 ). It comprises the OVA_257-264 and OVA_323-339 epitopes recognized by K^b-restricted OT-I and I-A^d-restricted OT-II T-cell receptor (TCR)-transgenic T cells, respectively. The Fc domains used were from human IgG1. Recombinant proteins were produced using a Baculovirus expression system and purified on protein G columns (for PPI-Fc and OVA-Fc) and nickel columns (for PPI). Peptides OVA_257-264 (SIINFEKL), OVA_323-339 (ISQAVHAAHAEINEAGR), PPI_B15-23 (LYLVCGERG) and PPI_B9-23 (SHLVEALYLVCGERG) were chemically synthesized at 85% purity (Chinapeptides). Herceptin (Trastuzumab anti-HER2/neu monoclonal antibody; Genentech) was used as IgG1 control. Trastuzumab F(ab’)2 was generated using the F(ab’)2 Preparation Kit (Pierce #44988, ThermoFisher).

PPI_B15-23 TCR-transgenic G9Cα^-/-.NOD (G9C8.NOD) (40) and Ins2 ^-/-.NOD mice (25) were kindly provided by F.S. Wong (University of Cardiff, UK) and C. Boitard (Cochin Institute, Paris, France), respectively. Fcgrt ^-/-.C57BL/6 (FcRn^-/-.B6) mice were from the Jackson Laboratory. All mice were housed in specific pathogen-free conditions. The study was approved by the French Ministry of Research (#4873-2015113018321568).

G9C8.NOD and Ins2 ^-/-.NOD mice were force-fed at day 1, days 1, 4 and 8 or days 1, 4, 7 and 10 of life with 30 µl PPI-Fc (50 µg) or equimolar amounts of control proteins (Fc-devoid PPI, OVA-Fc, IgG1) or phosphate-buffered saline (PBS), using a flexible plastic feeding tube (Instech Laboratories). For diabetes induction, 4-week-old G9C8.NOD mice were primed with 50 µg PPI_B15-23 peptide and 100 µg CpG (Eurogentec), followed by a second identical immunization 15 days later. Diabetes development was monitored by glycosuria and confirmed by glycemia when positive.

Protocols are described in Supplementary Methods.

Pellets were lysed in Radio Immuno Precipitation Assay buffer [150 mM NaCl, 1.0% IGEPAL^® CA-630, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate (SDS), 50 mM Tris, pH 8.0] and supernatants retrieved after centrifugation at 16,000g at 4°C for 15 min. Total protein lysates (5 µg/each; Pierce BCA Protein Assay Kit, ThermoFisher) were resolved on a 12% SDS–polyacrylamide gels (SDS-PAGE), transferred to polyvinylidene fluoride membranes and blocked overnight at 4°C in 5% milk/Tris-buffered saline-Tween (TBS-T; 20mM Tris pH 7.5, 150mM NaCl, 0.05% Tween-20). Membranes were then washed thrice in TBS-T and probed with primary antibodies: goat anti-mouse FcRn (RRID : AB_10972971; 1/1,000) and mouse anti-heat shock protein 70 (RRID : AB_627761; 1/1,000) overnight at 4°C, then washed thrice in TBS-T. Membranes were subsequently incubated for 30 min at room temperature with horseradish peroxidase-conjugated goat anti-mouse IgG (RRID : AB_2728714; 1/3,000) or rabbit anti-goat IgG (RRID : AB_562588; 1/5,000), and washed thrice in TBS-T. Bound IgG was detected with the ECL Prime Western Blotting System (GE Healthcare).

PPI-Fc and PPI proteins were conjugated with Alexa Fluor (AF)680 using a SAIVI Rapid Antibody/Protein labeling kit (Invitrogen) and orally administered to 1-day-old G9C8.NOD or FcRn^-/-.B6 mice. Fluorescence (whole body and isolated digestive tract and thymus) was detected using Photon Imager RT (Biospacelab) at a 690-nm excitation and 700-nm emission wavelengths, with 50–100 ms exposures. After imaging, organs (digestive tract and thymus) were included in O.C.T. Compound Orange (TissueTek, Sakura) and frozen at ‒80°C. Sections (7 µm) were air-dried, fixed for 10 min with 4% paraformaldehyde solution and permeabilized with 0.5% Triton X-100 solution. After blocking with 10% goat serum, sections were stained with the following antibodies: rat anti-EpCAM (RRID : AB_1089027; 1/750), rabbit anti-Cytokeratin 5 (RRID : AB_10979451; 1/750), hamster anti-CD11c (RRID : AB_467115; 1/1,000), and mouse anti-insulin (RRID : AB_260137; 1/1,000). For secondary detection, sections were incubated with Cy5-labeled goat anti-rat (RRID : AB_2338264) or anti-rabbit IgG (RRID : AB_2338013), an AF488-labeled goat anti-hamster IgG (RRID : AB_2338997) or an AF594-labeled goat anti-mouse IgG antibody (RRID : AB_2338883). Images were acquired with a Leica DMI6000 confocal microscope and analyzed with the Image J software.

At sacrifice, blood was collected from the same mice. Standard curves were obtained by sequential dilutions of PPI-Fc protein. PPI-Fc was captured on plates coated with an anti-insulin antibody (RRID : AB_2126540; 1/600) and detected with a peroxidase-labeled goat anti-human Fc antibody (RRID : AB_2687484; 1/4,000).

Five-day-old G9C8.NOD mice were force-fed with 50 µg of AF647-labeled PPI-Fc (or IgG1) or PPI (or IgG1 F(ab’)₂). Spleen, thymus and lamina propria (LP) were harvested 4, 8 or 24 h later.

Detection of AF647-labeled proteins was analyzed in LP, spleen and thymus preparations using an antigen-presenting cell (APC) panel ( Supplementary Table 1 ). Thymic EpCAM⁺ cells were analyzed using a thymic epithelial cell (TEC) panel ( Supplementary Table 2 ). AF647 median fluorescence was calculated on the total positive cell population.

Spleen, mesenteric (MLNs) and pancreatic lymph nodes (PLNs) were harvested from 7-day-old or 4-week-old G9C8.NOD mice orally treated on day 1 of life with PPI-Fc, PPI, OVA-Fc or PBS. Single-cell suspensions were surface-stained with anti-mouse antibodies (T-cell panel, Supplementary Table 3 ), washed in PBS, fixed/permeabilized with Foxp3 Fix/Perm Buffer (BioLegend) and stained for Foxp3.

Cells were acquired on a 16-color BD LSRII Fortessa and analyzed with FlowJo (v10.6.0).

Splenocytes were incubated with increasing concentrations of PPI_B15-23 peptide along with an APC-coupled anti-CD107a antibody to assess the cytotoxic activity. Brefeldin-A (20 µg/ml) and monensin (4 µM) were added after 1 h. Cells were further incubated for 5 h before extracellular staining. Cells were then washed in PBS, fixed/permeabilized with the BD Cytofix/Cytoperm buffer, washed twice in PBS 0.1% saponin and stained for interferon (IFN)-γ and tumor necrosis factor (TNF)-α in PBS/0.1% saponin. This recall antibody panel is listed in Supplementary Table 4 .

All statistical tests were two-tailed and performed using GraphPad Prism 7, as detailed in the legends of each figure. P values <0.05 were considered significant.

We previously demonstrated that intravenous vaccination of pregnant G9C8.NOD mice with a single 100-µg dose of PPI-Fc protected the offspring from subsequent diabetes development (39). We therefore tested whether a similar protection was afforded by oral PPI-Fc vaccination directly in newborn mice. To this end, we treated 1-day-old G9C8.NOD mice with a single 50-µg dose of PPI-Fc. This dose was selected by considering the maximal volume compatible with neonatal gavage (30 µl, 1.7 µg/µl), delivery to single rather than multiple pups (as is the case with the 100-µg transplacental delivery), and some degree of gastrointestinal degradation. We then activated the TCR-transgenic T cells of these mice by prime-boost immunization with PPI_B15–23 peptide and CpG at 4 and 6 weeks of age to induce diabetes (39, 40). As controls, equimolar amounts of recombinant OVA-Fc (i.e., irrelevant protein with preserved FcRn binding), PPI (i.e., cognate antigen with no FcRn binding), or PBS vehicle were orally administered. This oral PPI-Fc treatment did not prevent nor delayed diabetes onset ( Figure 1A ). Similar results were obtained by force-feeding newborn mice with a three-dose 50-µg regimen at day 1, 4 and 8 of life (data not shown).

We verified whether, despite lack of protection, PPI-Fc treatment was able to induce tolerogenic T-cell modifications in the spleen, PLNs and MLNs at 4 weeks of age (i.e. just before PPI_B15-23 peptide immunization). No difference in CD8⁺, CD4⁺, CD4⁺Foxp3^‒ T-cell or CD4⁺Foxp3⁺ Treg fractions ( Figures 1B–D ) or absolute numbers ( Supplementary Figure 2 ) was observed among treatment groups. Functional analysis of diabetogenic CD8⁺ T cells by in vitro peptide recall showed no impairment in the upregulation of activation (CD69) and cytotoxicity (CD107a) markers or in the production of pro-inflammatory cytokines (IFN-γ, TNF-α) ( Figure 1E ). Analysis of the LP, spleen and thymus of force-fed G9C8 mice earlier after oral treatment, at day 7 of life, yielded similar findings ( Supplementary Figure 3 ).

Collectively, these results show that neonatal oral PPI-Fc vaccination does not prevent diabetes nor it modulates T-cell responses in PPI TCR-transgenic G9C8.NOD mice.

Given the absence of any measurable tolerogenic effect, we verified whether orally administered PPI-Fc was able to cross the intestinal epithelium. We first assessed FcRn expression in gut epithelial cells of PPI TCR-transgenic G9C8.NOD mice at different ages by Western blot. FcRn expression was readily detectable at day 1 of life, significantly increased at day 5, 10 and 15, but completely disappeared at day 20 ( Figure 2A ). It remained undetectable in the gut of adult (4-13-week-old) mice, as in 1-day-old FcRn^-/-.B6 control mice.

We then force-fed G9C8.NOD newborns (wild type, WT) with fluorescent PPI-Fc and analyzed trans-epithelial transfer. Ex-vivo gut imaging revealed an intense PPI-Fc signal in both WT and FcRn^-/-.B6 mice 4 h post-administration, which was maintained over time in WT mice but significantly decreased in FcRn^-/-.B6 mice already at 24 h, and was barely detectable at 48 h ( Figure 2B ). Confocal microscopy confirmed retention of PPI-Fc signal (counter-stained with an anti-insulin antibody) below the epithelial layer of the small intestine villi (EpCAM⁺, red; Figure 2C ). Interestingly, PPI-Fc co-localized with the CD11c (green) staining of APCs, and was observed in WT mice but not in FcRn^-/-.B6 mice.

To better identify the cells responsible for PPI-Fc uptake in newborn mice, we used an antibody panel defining APC sub-populations ( Supplementary Figure 4 ). We first compared the uptake of PPI-Fc and Fc-devoid PPI in LP APCs at 4 and 8 h after force feeding ( Figure 2D ). Two subsets of CD11c^hi conventional DCs (cDCs) were initially identified, namely CD103⁺ and CD103^‒. F4/80⁺ macrophages comprised a CX3CR1⁺CD11b^‒ subset and a CX3CR1^‒CD11b⁺ subset (CX3CR1⁺ and CX3CR1^‒ from hereon, respectively). Fluorescent PPI-Fc was detected at both time points in CD103⁺ and CD103^‒ cDCs and, to a larger extent, in macrophages, especially in the CX3CR1^‒ subset. The PPI-Fc signal did not increase between 4 and 8 h, and the 4 h time point was retained for further studies. PPI uptake was instead non-significant in all APC populations.

Additional staining for the signal regulatory protein (SIRP)α marker highlighted that the CD103^‒ cDC subset can be positively defined by SIRPα expression, and we therefore further refer to this CD103^‒ subset as SIRPα⁺ cDCs ( Figure 3A ). Moreover, CX3CR1^‒ macrophages could be subdivided into SIRPα^lo and SIRPα^hi. To characterize the 4-h uptake of these refined APC subsets, we used fluorescently-labeled IgG1 and its F(ab’)₂ form as a proxy for PPI-Fc and PPI, respectively, as these proteins were more readily available and labeled more efficiently. We thus confirmed that the IgG1 uptake by LP CD103⁺ and SIRPα⁺ cDCs was not significantly different ( Figure 2E , left), as observed for PPI-Fc. More importantly, the SIRPα^hi subset accounted for the dominant uptake by CX3CR1^‒ macrophages ( Figure 2E , right). Similar to PPI, no significant uptake of F(ab’)₂ was observed. As the 4-h uptake of IgG1 and F(ab’)₂ was similarly distributed when compared to that of PPI-Fc and PPI, these proteins were subsequently used to phenotype the APC subsets responsible for the uptake in other organs. We also asked whether this uptake was stable over a longer 24-h time course ( Figure 2F ). While there was a trend toward increased uptake in DCs, the fluorescent signal remained stable and higher in macrophages. In contrast, F(ab’)₂ uptake remained marginal.

Finally, we investigated the contribution of FcRn to the protein uptake by LP DCs and macrophages of G9C8 newborn mice. In contrast to intestinal epithelial cells, none of these subsets exhibited a significant FcRn expression ( Supplementary Figures 5A, B ), arguing for an FcRn-independent acquisition of PPI-Fc, likely relying on Fcγ receptors.

Collectively, these results show that orally administered PPI-Fc crosses the intestinal epithelium in an Fc-FcRn-dependent fashion. It is subsequently taken up by LP CD11c⁺ APCs, both CD103⁺ and SIRPα⁺ cDCs and macrophages (mostly CX3CR1^‒SIRPα^hi), and this uptake is not observed for Fc-devoid PPI.

Having defined that oral PPI-Fc crosses the gut epithelium and is taken up locally by cDCs and macrophages, we asked whether systemic bioavailability was achieved. To this end, the uptake by splenic APCs was assessed 4 h after gavage. These APCs comprised CD11c^hi cDCs, which were less represented than in the LP but could be similarly subdivided into CD103⁺SIRPα^‒ (CD103⁺) and SIRPα⁺CD103^‒ (SIRPα⁺) ( Figure 3B ; see Supplementary Figure 6 for the detailed gating strategy). Although all F4/80⁺ macrophages were here CD11b⁺, CX3CR1⁺ and CX3CR1^‒ subpopulations were distinguished. The proportion of CX3CR1^‒ macrophages was higher in the spleen, and could be further subdivided into SIRPα^lo and SIRPα^hi as in the LP. Contrary to what observed in the LP, IgG1 fluorescence in cDCs was mostly detected in the SIRPα⁺ rather than in the CD103⁺ subset ( Figure 4A ). Also in the spleen, CX3CR1^‒SIRPα^hi and, to a lesser extent, CX3CR1⁺ macrophages accounted for most of the uptake ( Figure 4B ), while F(ab’)₂ uptake was negligible in all cases. At 24 h, IgG1 fluorescence decreased in CD103⁺ cDCs, and it drastically declined in SIRPα⁺ cDCs and in CX3CR1⁺ and CX3CR1^‒SIRPα^hi macrophages ( Figure 4C ).

Since macrophages are non-migratory APCs and CD103⁺ cDCs (which displayed lower uptake than SIRPα⁺ cDCs and macrophages) are the main subset described as migrating from the gut (41), our observations suggest that most Fc-coupled proteins may travel from the LP to the spleen in soluble form to be taken up locally. We therefore verified whether orally administered PPI-Fc could be detected in the serum of treated mice. Four hours after gavage, serum PPI-Fc was present at a concentration of 2.2 µg/ml in WT G9C8.NOD mice ( Figure 4D ). Serum PPI-Fc concentrations in FcRn^-/-.B6 mice were instead significantly lower (0.6 µg/ml), but not absent, indicating some passive transfer independent of the FcRn pathway. In line with the kinetics of uptake by non-migratory splenic APCs, PPI-Fc concentration dropped at 24 and 48 h, indicating that the distant uptake of soluble PPI-Fc outside the LP is most efficient within the first hours after oral treatment.

Finally, as in the LP, also in the spleen neither DCs nor macrophages expressed significant levels of FcRn ( Supplementary Figure 5C ), suggesting a preferential protein-Fc uptake through Fcγ receptors.

Collectively, these results show that orally administered PPI-Fc reaches the spleen mainly in soluble form and, to a lesser degree, via gut-derived CD103⁺ cDCs. It is subsequently taken up locally by SIRPα⁺ cDCs and macrophages.

We next investigated whether orally administered PPI-Fc can reach the thymus, as shown for the transplacental route (39). In the thymus ( Supplementary Figure 7A ), CD11c^hi cDCs can be divided into a SIRPα^lo and a SIRPα^hi subset, corresponding to resident and migratory thymic cDCs, respectively (39, 42, 43). Fluorescent signal was preferentially detected in migratory SIRPα^hi cDCs and was limited to IgG1 (70% AF647-labeled IgG1⁺ cells vs 10% F ( ab ’ ) 2 + ; Figure 5A ), suggesting that ferrying by cDCs is the predominant, although not exclusive, mechanism of Fc-protein delivery to the thymus.

We further analyzed the uptake by medullary (m)TECs (CD45^‒EpCAM⁺UEA-1⁺) and cortical (c)TECs (CD45^‒EpCAM⁺Ly51⁺) ( Supplementary Figure 7B ). Although detectable, their uptake was much more modest, and again limited to IgG1 ( Figure 5B ). As mTECs and cTECs are non-migratory, their IgG1 uptake suggests that Fc-coupled proteins gain access to the thymus also in soluble form, in line with the low-level signal detected in resident SIRPα^lo cDCs.

We verified by imaging whether the ferrying observed for IgG1 occurs also with PPI-Fc. Newborn G9C8.NOD mice analyzed 72 h after gavage displayed fluorescent PPI-Fc but not PPI signal both at the whole-body level and in isolated thymi ( Figure 5C ). Confocal microscopy on thymic tissue sections ( Figure 5D ) confirmed the presence of PPI-Fc fluorescence (white) associated with CD11c⁺ DCs (green) in WT G9C8.NOD mice, but not in their FcRn^-/-.B6 counterparts. These CD11c⁺ DCs were competent for in vitro antigen processing and able to present PPI_B15-23 peptide to G9C8 CD8⁺ T cells, inducing their activation and proliferation ( Supplementary Figure 8 ).

Collectively, these results show that orally administered Fc-coupled proteins, but not Fc-devoid proteins, reach the thymus mostly through ferrying by migratory SIRPα^hi cDCs and, to a lesser extent, in soluble form.

Having defined that oral PPI-Fc crosses the gut epithelium, achieves systemic bioavailability in both soluble and APC-associated form and reaches the thymus, we reasoned that the lack of diabetes protection in G9C8.NOD mice may reflect the aggressive nature of this TCR-transgenic model. We therefore repeated these experiments in Ins2 ^-/-.NOD mice. This model is informative because it develops spontaneous diabetes and harbors a natural polyclonal T-cell repertoire. It thus provides a more physiological setting, in which oral PPI-Fc treatment should impact only the PPI-reactive T-cell fraction present in the natural repertoire. Moreover, Ins2 knock-out leads to absent PPI expression in the thymus and lack of thymic deletion of PPI-reactive T cells, thus accelerating diabetes development (25, 26). The thymic delivery of orally administered PPI-Fc may thus compensate this defect and impact the overall autoimmune cascade downstream of PPI. We applied the same single-dose 50-µg regimen as before in 1-day-old Ins2 ^-/-.NOD mice. Also in this case, no diabetes protection was observed ( Figure 6A ). To verify whether this was due to suboptimal dosing, we attempted a more intensive regimen. Based on the kinetics of intestinal FcRn expression ( Figure 2A ), repeated 50-µg doses were administered at day 1, 4, 7 and 10 of life, in order to cover the time window displaying the highest intestinal FcRn expression. Again, no significant protection was obtained, but diabetes development was delayed by approximately 10 weeks in 33% of PPI-Fc-treated mice compared to control groups ( Figure 6B ).

Collectively, these data show that oral PPI-Fc, administered either in a single-dose or multiple-dose regimen, does not significantly prevent diabetes development in polyclonal Ins2 ^-/-.NOD mice, albeit a partial delay is observed with the multiple-dose regimen.