A thymic biopsy was obtained from a 23-month-old infant with MHCII deficiency upon informed consent in accordance with the Research Ethics Board at The Hospital for Sick Children in Toronto (Canada). Patient’s data were compiled prospectively and retrospectively from medical records and entered into the Primary Immunodeficiency Registry and Tissue Bank (REB protocol no. 1000005598). The patient presented at 18 months of life with a history of recurrent respiratory tract infections, chronic diarrhea, and CMV hepatitis. She had a family history of MHCII deficiency and was found to be homozygous for a mutation in the CIITA gene. Immunological data are reported in Table 1 . The patient received a matched related HSCT, but engraftment was poor.

A human thymic sample from a healthy control used for comparison was analyzed retrospectively in compliance with Declaration of Helsinki and policies approved by Ethics Board of Spedali Civili in Brescia. Specifically, for retrospective and exclusively observational study on archival material obtained for diagnostic purposes, patient consent was not needed (Delibera del Garante n. 52 del 24/7/2008 and DL 193/2003).

Blood samples were collected from 11 patients with MHCII deficiency [peripheral mononuclear cells (PBMC), n = 4; serum or plasma, n = 9]. Each patient was attributed a numerical code (MHCII_xx). Questionnaires to gather patient’s relevant clinical data were sent to referring clinicians, who were responsible for the collection of informed consent for biological samples’ collection and anonymized biological sample/data sharing from their own patients, according to local research protocols, reviewed and approved by local ethics committees or institutional review board (IRB) [for NIH patient’s samples, protocols 18-I-0041 and 18-I-0128, approved by the NIH IRB]. Data about clinical history, immunological features and HSCT of this cohort of MHCII-D patients are reported in Tables 2 – 4 respectively.

Peripheral blood (PB) from healthy controls (HD) was obtained in accordance with the 1964 Declaration of Helsinki and its subsequent amendments, under biological material collection protocols approved by the Institutional Ethical Committee of San Raffaele Hospital (Tiget07, Tiget09). Informed consent was signed directly by the subject or by parents or legal representatives in case of minors.

Aβ^0/0 mice were kindly provided by Dr. Matteo Iannacone (San Raffaele Scientific Institute, Milan, Italy). Rag1 ^−/− mice were purchased from The Jackson Laboratories. C57/black (BL) 6 control wild-type (WT) mice were purchased from Charles River Laboratories Inc. All mice were housed in specific pathogen-free conditions and treated according to protocols approved by the Animal Care and Use Committee of the San Raffaele Scientific Institute (Institutional Animal Care and Use Committee protocol no. 710 and 712).

Following sacrifice, thymus and gut tissue isolated from mice were formalin-fixed and paraffin-embedded. Hematoxylin and Eosin (H&E) staining was used to assay basic histopathological changes. Paraffin sections were de-waxed, rehydrated, endogenous peroxidase activity was blocked by 0.1% H₂O₂, and nonspecific background reduced with Rodent Block M (Biocare Medical, Concord, CA, USA) before heat-based antigen-retrieval treatment and incubation with antibodies. Depending on the primary antibodies used, sections were incubated with Rat-on-Mouse HRP-Polymer (Biocare Medical) or MACH 1™ Universal HRP Polymer Kit (Biocare Medical) or 4plus Streptavidin HRP label (Biocare Medical); reactions were developed in Biocare's Betazoid DAB and nuclei counterstained with Hematoxylin. Digital images were acquired with an Olympus XC50 camera mounted on a BX51 microscope (Olympus, Tokyo, Japan), with CellF Imaging software (Soft Imaging System GmbH, Münster, Germany) or with a Nikon Eclipse E600 Microscope (Nikon) with NIS Elements Software (Nikon). The following primary antibodies were used on thymic sections: rabbit anti-cytokeratin 5 (CK5) (1:200; Covance), rat anti-cytokeratin 8 (CK8) (1:200; Developmental Studies Hybridoma Bank); rat anti-mouse AIRE (1:300; Millipore); rat anti-mouse FOXP3 (1:100; eBioscience); biotin-UEA (Biotinylated Ulex Europaeus Agglutinin I) (1:800; Vector Laboratories). Rabbit anti-CD3 primary antibody (1:100; ThermoFisher Scientific) was used on gut sections.

To quantify abnormalities of colon pathology, a histological score was used. The degree of colic alterations was blindly graded using combined scores including: the grade of inflammation (grade 0, no evidence of inflammation; grade 1, low; grade 2, moderate; grade 3, high; grade 4, intense and diffuse level of inflammation), the structural changes of intestinal layers (grade 0, absence; grade 1, focal; grade 2, partial; grade 3, diffuse level of structural changes), and the gland secretion alterations (grade 0, normal; grade 1, slight gland secretion alterations; grade 2, moderate; grade 3, diffuse gland secretion alterations). The cumulative total scores ranged from 0 to 10.

Human thymic tissue sample was formalin-fixed and paraffin-embedded. Sections were used for routine H&E staining and treated as described above. Nonspecific background was reduced with Background Sniper (Biocare Medical, Concord, CA, USA) before heat-based antigen-retrieval treatment and incubated with primary antibodies. The following primary antibodies were used: rat anti-human FOXP3 (1:100; eBioscience), mouse anti-human AIRE (1:3000; kindly provided by Prof P. Peterson, University of Tartu, Tartu, Estonia), biotin-UEA (Biotinylated Ulex Europaeus Agglutinin I) (1:800; Vector Laboratories) and Claudin-4 (used according to local standards of Pathological Anatomy Unit of Spedali Civili, Brescia, Italy). Sections were processed as described above for the murine experiments. Morphometric analysis was performed using Olympus Slide Scanner VS120-L100 (Olympus, Tokyo, Japan) and Image-pro software (Olympus) was used to analyze them. Digital images were acquired with the same instruments and software as described above.

Murine TEC isolation was performed on age-matched WT and Aβ^0/0 mice. Briefly, mice were sacrificed by decapitation to avoid excessive bleeding during surgery. Murine thymus was cleaned from fat and stromal tissues, and then digested at 37°C with an enzymatic solution containing Liberase TL and DNAse I (Roche). Digested tissues were collected in DMEM (Lonza) supplemented with 10% fetal bovine serum (FBS), 1% glutamine and 1% penicillin and streptomycin. Single thymic cell suspensions were then incubated with anti-CD45 micro-beads (Miltenyi Biotec) and processed with the AutoMACS Pro Separator (Miltenyi Biotec). The CD45 negative fraction was retrieved and then tested by multicolor FACS analyses for the expression of TEC markers.

Thymocytes, splenocytes and lymphocytes were freshly isolated by mechanically disrupting the thymus, the spleen and lymph nodes of age-matched WT and Aβ^0/0 mice. In addition, splenocytes were lysed with Red Cell Lysis Buffer (RCLB) (150mM M NH4Cl, 10 mM KHCO3, and 0.1 mM EDTA). PB samples underwent red blood cell (RBC) lysis twice with RCLB, before further processing. Recovered cells were re-suspended in D-PBS (EuroClone, Pero, Italy) with 2% FBS and subsequently stained for flow cytometric analysis. CD4⁺ T cells were isolated by pooled spleens and mesenteric lymph nodes (MLNs) from 2–3 mice/pool of age-matched WT and Aβ^0/0 mice using CD4-specific magnetic beads via negative selection according to the manufacturer’s instructions (CD4⁺ isolation kit Miltenyi Biotec).

The list of monoclonal antibodies used for the staining of murine TEC, thymocytes, splenocytes and lymphocytes derived from lymph nodes are reported in the Materials and Methods section in this article’s Supplementary Materials . Cells were acquired on a FACS CANTO (BD Pharmingen) and analyzed with FlowJo software.

TECs were isolated and enriched with the AutoMACS Pro Separator after digestion of thymi from pool of 5–10 age-matched WT and Aβ^0/0 mice of 4–6 weeks of age, as previously described. To sort mTEC and cTEC, isolated TECs were stained with anti-CD45 (30-F11), anti-CD326 (EpCam; G8.8), anti-MHCII (M5 114.15.2) (all Biolegend), anti-Ulex-1 (FL 1061, Vector) and anti-Ly51 (6C3, Miltenyi) monoclonal antibodies and sorted with the MoFlo Legacy cell sorter (Becton Dickinson), with 100 μl noozle (FRACTAL facility of San Raffaele Hospital, Milan, Italy). Non-viable cells were excluded from analyses using 7-AAD (BD Pharmingen). In order to preserve RNA quality for further analyses, ProtectRNA^TM RNase Inhibitor (Sigma-Aldrich) was added to CD45-negative fractions according to manufacturer’s instruction, before sorting procedure. Moreover, TEC subsets were sorted directly in RNAlater (Sigma-Aldrich), kept at +4°C overnight and then at −20°C until use.

RNA extraction from sorted TEC from WT and Aβ^0/0 mice was performed with RNeasy microkit (QIAGEN, Hilden, Germany) according to manufacturer’s instructions. RNA was then stored at −80°C until use. Reverse transcription of murine-sorted TEC mRNA, synthesis and amplification of cDNA were performed with the Ovation PicoSL WTA System V2 (Nugen), according to the manufacturer’s instructions. Purification of amplified cDNA was achieved using the QIAquick PCR Purification Kit (QIAGEN). Quantitative Real-time (RT) PCR was performed using Fast SYBR green master Mix (Thermo Fisher Scientific) and the ViiA7 Real-Time PCR System (Thermo Fisher Scientific). Each sample was analyzed in duplicate. The relative level of expression was determined by normalization to β-actin (Actb) ribosomal RNA. The primers used are listed in Supplementary Table 1 .

The cDNA libraries of sorted cTEC and mTEC from WT and Aβ^0/0 mice were constructed using the Takara SMART-Seq v4 Ultra low input RNA Kit. PCR library products were quantified using the Agilent DNA 1000 assay (Agilent#5067-1504) on an Agilent Technologies 2100 bioanalyzer. Pooled libraries were loaded on a Paired End Flow Cell using the cBot System (Illumina) and the HiSeq 3000 platform. At the end of the run, around 30 M of 50 bp paired-end reads per sample were generated. Transcript abundance was estimated with Kallisto (29) and differentially expressed (DE) genes were identified using DeSeq2 (30) R package and a FDR corrected p-value <0.05. Gene set enrichment analysis (GSEA) was applied in order to identify gene sets and pathways that were significantly perturbed across conditions. Collections of gene sets were downloaded from MiSigDB (http://www.broadinstitute.org/gsea/msigdb/). We used the Benjamini–Hochberg method to adjust gene set p-values and set 0.1 as the significant threshold. Splicing entropy was calculated on TRA transcripts as described in (31). RNA-Seq data are available under accession number GSE166463.

Murine TEC for proteomic analyses were obtained from three pools of WT or Aβ^0/0 mice, three mice per pool (5–6 weeks-old). TEC were isolated, enriched and processed through CD45⁺ cell-depletion as described above. CD45^- cell-samples were pelleted and stored at −80°C until use. Detailed descriptions of proteomic analysis, data processing, and interaction network reconstructions are described in the Materials and Methods section in this article’s Supplementary Materials .

This analysis was performed at the Genomics & Microarray Core Facility at UT Southwestern Medical Center (USA) on human and murine serum or plasma samples. The Autoantigen array contained 95 autoantigens and eight internal control antigens. Profiling of both IgG and IgM autoantibodies was performed. The autoantibodies binding to the antigens on the array were detected with Cy3 labeled anti-IgG and Cy5 labeled anti-IgM and the arrays were scanned with GenePix® 4400A Microarray Scanner. The images were analyzed using GenePix 7.0 software to generate GPR files. The averaged net fluorescent intensity (NFI) of each autoantigen was normalized to internal controls (IgG or IgM).

CD4⁺ T cells were isolated from pooled spleens and MLNs of 9–10 wk-old WT or Aβ^0/0 mice as described above. Enriched CD4⁺ T-cell samples were subsequently sorted into naive CD4⁺CD25⁻CD45RB^hi and regulatory CD4⁺CD25^hiCD45RB⁻ cell populations (>98% purity) using a FACS Aria Fusion (Becton Dickinson) cell sorter (FRACTAL facility of San Raffaele Hospital, Milan, Italy). For colitis induction, 4 × 10⁵ WT CD4⁺CD25⁻CD45RB^hi naive T cells were transferred by intraperitoneal injection (i.p.) into 8–9 week-old Rag1^−/− mice alone or with 1.5 × 10⁵ CD4⁺CD25^hi Treg cells from WT or Aβ^0/0 mice. Recipient mice were weighted twice a week and sacrificed at week 4 or 8 after the transfer, when colitis was diagnosed by severe weight loss and/or wasting diarrhea. Gut tissues were isolated and analyzed as described and the colon length was measured at time of sacrifice.

PBMC were isolated from PB of HD and patients by density gradient centrifugation using LymphoprepTM (density: 1.077 g/ml; STEMCELL Technologies, Vancouver, Canada). PBMC were maintained in RPMI medium (CORNING) with 2% FBS, 2% L-glutamine, 1% Penicillin/Streptomycin at 4 °C until use or live frozen in FBS + dimethyl sulfoxide (DMSO) 10%. Multi-color immunophenotype of T-lymphocyte subsets on human PBMC was performed by flow cytometry. Details about the monoclonal antibodies used for each staining are reported in the Materials and Methods section in this article’s Supplementary Materials .

Levels of B-cell activating factor (BAFF) were measured in duplicate in serum or plasma samples of MHCII-D patients and age-matched HD using a Quantikine Human BAFF/BLyS/TNFSF13B Immunoassay kit (R&D Systems, Minneapolis, USA). The assay was performed according to manufacturer’s instructions and the optical density (OD) was determined using a microplate reader.

All results are expressed as median and interquartile range if not stated otherwise. In the present study a descriptive statistical analysis has been performed. No formal inference was performed due to the small sample size. Statistical significance was assessed using a two-tailed Mann–Whitney test to compare continuous outcomes between groups. T test was used for splicing entropy result analysis. Levels of significance were defined as p ≤0.05 (*), p <0.01 (**), p <0.001 (***), and p <0.0001 (****). Statistical testing was performed using Prism GraphPad (Version 5.0f, La Jolla, CA). Graphs were created using the same software.

Data about human thymic histology in patients with MHCII deficiency are limited because of the difficulties in accessing these tissues for technical and ethical reasons. In this study, we had the unique opportunity to perform histological analysis of a thymic biopsy performed in a 23-month-old MHCII-D female patient before HSCT (see Table 1 for immunological data). The patient carried a homozygous mutation in the CIITA gene [c.3317 + 2dup], resulting in an intronic splice variant.

The thymic architecture was perturbed, with reduced representation of thymic medulla ( Figure 1A ), as compared to a control thymus analyzed in parallel. This finding was also supported by staining for Ulex Europaeus Agglutinin I (UEA1) and FOXP3 ( Figure 1B ). Moreover, the MHCII-D thymus presented a reduced frequency of AIRE⁺ and CLAUDIN 4 (Cldn4)-expressing TEC, suggesting possible defects of central tolerance ( Figures 1C, D ).

To further corroborate these observations, we analyzed thymic tissue isolated from the MHCII knock out mouse model, Aβ^0/0 (17, 18). Histological analysis showed similar abnormalities of thymic architecture, with reduced representation of TECs, especially in the medulla, resulting in a significantly increased cortico-medullary (C/M) ratio in Aβ^0/0 mice, as compared to WT mice (p-value 0.0087) ( Figures 2A, B ). However, CK5, a medulla specific marker, and CK8 staining indicated a correct compartmentalization of cortical and medullary areas in the thymus of Aβ^0/0 mice ( Supplementary Figure 1A ). Overall reduced total Epcam⁺ TEC and mTEC frequency and absolute count in Aβ^0/0 thymi, as compared to WT, were confirmed by flow cytometric analysis ( Figure 2C and Supplementary Figure 1B ). Finally, we detected a decreased frequency of AIRE⁺ TEC and FOXP3⁺ cells in the medullary area of Aβ^0/0 mice thymus, as compared to WT mice ( Figures 2D, E and Supplementary Figure 1A ).

In accordance with previous reports (17, 18, 24), we observed a reduction in the frequency and in the absolute count of CD69^hiTCRβ^hi thymocytes ( Figures 3A, B and Supplementary Figure 2A ), and severe defects of CD4⁺ SP thymocyte generation and maturation ( Figures 3C, D and Supplementary Figure 2B ) in Aβ^0/0 mice, in line with the known fundamental role of MHCII molecules for the progression from the DP stage to the SP CD4⁺ T cell stage (18). Moreover, we confirmed that that majority of these residual SP CD4⁺ thymocytes in Aβ^0/0 mice corresponds to CD1-restricted NKT cells, as previously described (20, 21) ( Supplementary Figure 2C ). This resulted in peripheral CD4⁺ T cell lymphopenia, with significant reduction in the frequency and absolute count of CD4⁺ naïve T cells and increased frequency of activated CD4⁺ T cells in secondary lymphoid organs (spleen and lymph nodes), as compared to WT CD4⁺ SP cells (SP4) cells ( Supplementary Figures 2D, E ).

Since we could not get access to thymocytes from MHCII-D patients, we investigated TRAV1 (Vα7.2) expression on patients’ PB CD3⁺ lymphocytes as a surrogate marker of thymocyte maturation and survival (32). Interestingly, we found that all four patients analyzed (three before and one after HSCT) showed a trend towards reduced expression of Vα7.2 on their CD3⁺ T cells as compared to normal controls tested in parallel (median: MHCII-D 2.1%; HD 6.8%) ( Figures 3E, F ).

In order to evaluate thymic output in MHCII deficiency, we analyzed the proportion of naïve T cells and recent thymic emigrants (RTE) among PBMC from three patients: two untransplanted [of whom one infant (MHCII_11) and one adult (MHCII_12)], and one transplanted patient (MHCII_08). The percentage of naïve CD45RA⁺CD27⁺CD4⁺ T cells was reduced in the two untreated patients, as compared to the normal donor tested in parallel. Conversely, naïve CD4⁺ T cells in the transplanted patient tended to normal levels ( Figure 3G ). No significant differences emerged in the proportion of naïve CD8⁺ T cells (data not shown). We identified RTE as CD31⁺ cells within naïve CD27⁺CD45RA⁺CD4⁺ T cells. In line with data on naïve CD4⁺ cells, the frequency of RTE in the two untransplanted patients was also reduced ( Figure 3G ). This finding was particularly striking for the infant patient. RTE in MHCII_08, who underwent transplant, were comparable to the adult control tested in parallel ( Figure 3G ).

Based on our observations of TEC abnormalities in histologic analysis of thymic samples, we evaluated gene expression profile of Aβ^0/0 mouse TEC subsets by performing bulk RNA-Seq on sorted WT and Aβ^0/0 cTEC and mTEC. In order to obtain a sufficient amount of sorted TEC, we pooled 5–10 WT or Aβ^0/0 mice of 4–6 weeks of age for each biological replicate (n = 3, Supplementary Table 2 ). The gating strategy used to sort cTEC (Epcam⁺CD45⁻Ly51⁺UEA1⁻ cells) and mTEC (Epcam⁺CD45⁻Ly51⁻UEA1⁺ cells) is shown in Supplementary Figure 3A . Principal component analysis (PCA) showed that samples were properly grouped according to genetic background and tissue of origin ( Supplementary Figure 3B ).

Transcriptome analysis revealed an altered gene expression profile in sorted TEC subsets from Aβ^0/0 mice, as compared to age-matched WT mice. These differences were especially prominent in mTEC ( Figure 4A and Supplementary Table 3 ), on which we subsequently focused our analyses. In particular, we identified almost one thousand (n = 929) transcripts that were differentially expressed (DE) in mTEC from WT versus Aβ^0/0 mice. The majority of these transcripts (n = 679, in red in Supplementary Figure 4A ) were upregulated in mTEC from WT mice ( Supplementary Table 3 ). Among DE genes, transcripts that were enriched in Aβ^0/0 mTEC (n = 250) were mainly relative to RNA splicing, histone modifications and transcription factors (most of which with repressive function) ( Figure 4B ).

Similarly, even if to a lesser extent, cTEC transcriptome resulted also altered ( Figure 4A and Supplementary Table 3 ) in Aβ^0/0 mice. Differentially expressed transcripts upregulated in Aβ^0/0 cTEC were mainly relative to RNA splicing and translation, carbohydrate metabolism and cytoskeleton. Conversely, in WT cTEC enriched DE transcripts resulted relative to lipid metabolism, peptidases, apoptosis, transmembrane transport, transcriptional regulation, blood vessels/extracellular matrix, cytokine signaling pathway, inflammatory and immune response ( Supplementary Figure 5 ).

To further define the cellular features of TEC in Aβ^0/0 mice, we performed proteomic analysis of bulk CD45-depleted TEC fractions from pools of 4–6-week-old WT and Aβ^0/0 mice (n = 3, three mice/pool). This analysis could not be performed on sorted TEC subsets due to technical limitations. In these cell subsets, we found 373 DE proteins, 111 of which with high confidence ( Supplementary Tables 4 and 5 ). Most of these proteins were upregulated in WT mice (in red in Supplementary Figure 4B ). Among DE proteins, the only ones that were enriched in Aβ^0/0 TEC-enriched subsets were relative to chromatin assembly, in particular histones, as indicated by combined cluster-network analysis ( Figure 4C ). We cannot exclude the presence of some fibroblasts and endothelial cells in the analyzed samples.

Next, we performed an integrated network analysis on DE transcripts and proteins in WT and Aβ^0/0 TEC in order to identify common specific pathways dysregulated in TEC in the absence of MHCII. The only common variation observed in both cTEC and mTEC transcripts and in CD45-depleted TEC fractions was an increased representation of transmembrane transporter activity in WT mice thymi. However, if restricting the analysis to mTEC only, a concordant pattern of variations in both transcripts and proteins within the same cluster between WT and Aβ^0/0 mice was observed for four clusters ( Figure 4D ). In particular, in WT mice, increased expression was observed for genes and proteins involved in cell metabolism, energy production and cell cycle.

Next, we interrogated RNA-Seq data for genes known to be involved in the establishment and maintenance of central tolerance. GSEA showed a significant enrichment for both Aire-dependent and Aire-independent transcripts known to have a tissue-restricted pattern of expression, also known as Tissue Restricted Antigen (TRA) genes, in WT mTEC ( Figure 5A ). Reduced expression of TRA genes in Aβ^0/0 mTEC emerged also at DE analysis and was confirmed by qRT-PCR analysis for a selected number of them ( Figures 5B, C ). Consistent with these observations, Aire gene expression was significantly reduced in Aβ^0/0 mTEC, as compared to WT (Log2 fold change mTEC WT/Aβ^0/0: 2.12, adjusted p-value: 0.0017) and this result was confirmed by qRT-PCR ( Figures 5B, C ). These results confirm and extend previous observations (24). In addition, we confirmed reduced Aire expression also at the protein level, as shown by intracellular FACS analysis ( Figure 5D ). Analysis of Fezf2 expression did not reveal any relevant differences in Aβ^0/0 and WT mTEC ( Figures 5C, D ). In addition, DE analysis suggested a disruption of the mechanisms controlling Aire expression in Aβ^0/0 mTEC, as indicated by a significantly reduced expression of the transcripts encoding for the protein deacetylase Sirtuin-1 (Sirt1) and Irf8 in Aβ^0/0 mTEC ( Figure 5B ), respectively an essential Aire regulator and an Aire transcriptional activator. Finally, in order to further investigate mTEC functionality (34), we also calculated TRA splicing entropy (31), a measure of the diversity of transcripts isoforms, which resulted decreased in Aβ^0/0 mTEC ( Figure 5E ). This disruption of the complex machinery regulating Aire expression could be due to Aβ^0/0 mTEC impaired maturation. Collectively, these findings suggest that lack of MHCII expression in mTEC is associated with impaired promiscuous gene expression (PGE) and abnormalities of the mechanisms that govern central tolerance.

RNA-Seq results showed a significantly decreased expression of genes involved in mTEC maturation in Aβ^0/0 sorted mTEC. Accordingly, GSEA showed a significant enrichment in genes known to be expressed in mature mTEC in WT mice, while genes known to be expressed in immature mTEC where significantly enriched in Aβ^0/0 mTEC (33) ( Figure 5A ). Moreover, DE gene analysis showed a significantly increased expression of co-stimulatory molecules (mainly CD86 and ICOS ligand, which are involved in TEC-thymocyte cross-talk), in WT mTEC, as compared to their Aβ^0/0 counterpart ( Figure 5B ). A similar pattern was observed for transcripts encoding for molecules involved in NF-κB signaling, and in particular for CD40, which was more abundantly expressed in WT mTEC ( Figure 5B ). These results were confirmed when analyzing numbers of CD40L⁺ and RANKL⁺ thymocytes in Aβ^0/0 mice. Indeed, we detected by flow cytometry a severe reduction of both CD40L⁺ and RANKL⁺ thymocyte absolute counts in Aβ^0/0 mice ( Supplementary Figure 6 ), suggesting a low CD40L and RANKL-mediated stimulation of Aβ^0/0 mTEC. This reduction was particularly evident in mice aged 3 weeks or older ( Supplementary Figure 6 ).

Based on the alterations detected in the thymus of MHCII deficient patients and mice, we hypothesized that impairment in central tolerance mechanisms could have an impact also on peripheral tolerance. In line with published results for another MHCII mouse model, the Aα^−/− mice (25), Treg cells were nearly absent in the thymus of Aβ^0/0 mice, but they appeared relatively enriched in spleen and lymph nodes where they were present at higher frequency than in WT mice ( Figure 6A , left panel). Nonetheless, Treg absolute count was reduced in all lymphoid organs in Aβ^0/0 mice ( Figure 6A , right panel).

We then tested in vivo the function of Aβ^0/0 Treg cells. To this purpose, we made use of a model of colitis induced by adoptive transfer of WT T cells, alone or in combination with WT or Aβ^0/0 Treg cells, into Rag1 ko mice ( Supplementary Table 6 ), and analyzed the capacity of Treg cells to attenuate colitis. Recipient mice were monitored weekly and sacrificed four or eight weeks after the transfer, when signs of colitis (wasting diarrhea and weight loss) became evident in mice that had received WT T naive cells only or in combination with Aβ^0/0 Treg cells ( Supplementary Figure 7A ). At sacrifice, the gut tissue was analyzed macroscopically and colon length, an indirect sign of gut inflammation (35), was measured. As shown in Figure 6B , colon length in mice receiving co-injection of WT Treg cells together with WT T naïve cells, was similar to that of untreated mice. Conversely, colon length was reduced in mice receiving WT naïve T cells alone and in those receiving co-injection of Aβ^0/0 Treg cells together with WT naïve T cells. Moreover, in mice that received either WT naïve T cells only or in combination with Aβ^0/0 Treg cells, the study of intestinal pathology revealed different degrees of spontaneous colitis and wasting diarrhea. Substantial thickening of colonic mucosa, indicative of inflammation, was observed. Histologically, colonic inflammation was characterized by crypt elongation and large inflammatory cell infiltrate, mainly consisting of T lymphocytes, with occasional crypt abscesses ( Figure 6C ). A colitis score was calculated on histological sections of gut tissue based on the sum of the evaluation of architectural alterations, degree of inflammation and muciparous gland activity (see Materials and Methods for details). When compared to untreated animals, the colitis score was significantly increased in mice receiving WT naïve T cells alone or in combination with Aβ^0/0 Treg cells, while it was not statistically different in mice receiving both WT naïve T cells and WT Treg cells ( Figure 6D ). Furthermore, among the four treatment groups, only mice that had received co-transplantation of WT Treg cells showed an increased proportion of Treg cells in the spleen and MLN at sacrifice ( Supplementary Figures 7B, C ). Altogether, these results demonstrate that adoptive transfer of Aβ^0/0 Treg cells failed to block colitis induction in Rag1 ko mice, suggesting an impaired function of these cells.

To confirm the results obtained in the mouse model of MHCII deficiency, we set out to evaluate peripheral tolerance in MHCII-D patients. To this purpose, we collected PB samples from a cohort of patients with MHCII deficiency, whose main clinical features are summarized in Table 2 . RFXANK gene mutations were identified in eight of them. Median age at sampling was 4.7 years (range: 0.4–24.1 years). All samples were collected before HSCT, except in one patient for whom only post-transplant sample was available. They suffered from severe and recurrent infections and the majority of them also presented with chronic diarrhea. Autoimmune manifestations were described in three patients. Available data about patients’ immune phenotype and serum immunoglobulin levels are reported in Table 3 . Most patients were treated with immunoglobulin replacement therapy and anti-infective prophylaxis, as standard supportive care. Seven patients were treated with HSCT, at a median age of 4.5 years (range: 1.1–16.2 years). Characteristics and outcome of HSCT are detailed in Table 4 .

We first investigated the frequency of circulating Treg cells in three MHCII-D patients. Treg cells were defined as CD4⁺CD25^hiCD127^loFOXP3⁺ cells. Moreover, in order to differentiate thymic from peripherally induced Treg cells we also stained cells for the expression of HELIOS. The frequency of Treg cells was slightly reduced in one of the untransplanted patients (MHCII_11), as compared to published reference values (36). However, the other two patients showed a frequency of Treg cells comparable to the normal donor, and within normal range also as compared to published references ( Figure 7A and Supplementary Table 7 ). Interestingly in all patients, most CD4⁺CD25^hiCD127^loFOXP3⁺ cells were also positive for HELIOS, suggesting their likely thymic origin ( Supplementary Table 7 ).

We then investigated possible tolerance breakdown as a result of impaired T-B cell cross-talk in the absence of MHCII molecules. In order to evaluate the presence of autoantibodies in the serum of MHCII-D patients, we performed a protein microarray (37) to screen for a large panel of IgG and IgM autoantibodies. This assay revealed a significantly increased titer of many different autoantibodies, especially of IgG isotype, in the serum of MHCII-D patients ( Figure 7B ), as compared to healthy controls. Furthermore, BAFF levels, which are known to play a role in B-cell homeostasis and peripheral tolerance (38, 39), were increased in the sera of most of the patients analyzed, as compared to healthy controls ( Figure 7C ). Both findings of multiple serum autoantibodies and increased BAFF levels, even if not linked to overt autoimmune manifestations in most patients, are suggestive for the presence of a defective peripheral B-cell tolerance checkpoint in MHCII-D.