Autoimmune amelogenesis imperfecta in patients with APS-1 and coeliac disease

a, Illustration of the main amelogenesis developmental stages. The inductive (pre-secretory) stage (1) gives rise to functional secretory ameloblasts (2), which secrete various enamel matrix proteins, thereby providing a scaffold for hydroxyapatite crystal formation. Next, in the maturation stage (3), the secreted enamel matrix proteins undergo proteolytic degradation by enamel matrix specific enzymes. This consequently gives rise to the hardest mineralized tissue in the body, the enamel. Created with BioRender.com; b, Illustration of two alternative mechanisms responsible for enamel dysplasia in AIRE-deficient individuals. Created with BioRender.com; c,d, Reanalysis of published microarray data sets²⁵ showing differential gene expression of mouse developing teeth relative to other mouse tissues. Differential gene expression was compared to gene’s mean expression (c), or to the corresponding p-value of genes (d), (Fc>2.0, Log2ME > 4, p-value < 0.05). Amelogenesis imperfecta linked genes (red), tooth-specific genes (blue) and Aire (black). e, Reanalysis of published scRNA-seq dataset²⁹ showing expression of the indicated genes in dental epithelium; high expression (green), no expression (black). f, Heatmap showing normalized expression values (as log2) of amelogenesis imperfecta linked genes and Aire in tooth germ and ameloblast (determined by RNA-seq analyses^26,27,28). g, Relative gene expression of Aire and Amelx in ameloblast cells (at E15.5, 5 days, 3 weeks and 16 weeks) and in mTEC^hi (at 4 weeks) as assayed by qPCR. h,i, Representative immunofluorescence microscopy images of NOD WT (h) or NOD Aire^−/− (i) jaw and thymus sections stained with AIRE-specific antibodies (green), amelogenin-specific antibodies (red) and DAPI (blue). Scale bar 100 µm. j, Bulk RNA-seq results of normalized expression of amelogenesis imperfecta linked genes (red), tooth specific genes (blue) and AIRE-depended positive control gene (black) in NOD WT (n = 3) or NOD Aire^−/− (n = 3) mTECs. Raw P values were adjusted for multiple testing, using the procedure of Benjamini and Hochberg⁶⁵. k,l, Volcano plots of bulk RNA-seq of ameloblast-specific genes (blue), amelogenesis imperfecta linked genes (red) or AIRE-depended positive control gene (black) in mTEC^hi cells isolated from either NOD (k) or B6³⁰ (l) Aire^−/− versus WT mice. In bold are AIRE-dependent genes depicted in mice of both backgrounds. n refers to the number of biologically independent mice per genotype or group. Data are representative of at least two independent experiments. Data shown as mean ± s.e.m. Data were analysed using one-way ANOVA following Tukey’s multiple comparisons test (g).

Source Data

a–e, UMAP of published³¹ scRNA-seq of thymic epithelial cells (TEC; EpCAM⁺CD45^–) from B6 WT mice. Each dot represents a single cell; a, UMAP highlighting the presence of five major colour-coded TEC clusters. b, UMAP of Aire gene expression per TEC subset as assayed by scRNA-seq. c,d, UMAPs of ameloblast-specific and amelogenesis imperfecta related genes whose expression is primarily restricted (c) or not restricted (d) to the Aire⁺ TEC subset. e, UMAPs of ameloblast-specific and amelogenesis imperfecta related genes with undetectable expression levels within the TEC compartment as assayed by scRNA-seq. f, Representative immunofluorescence microscopy images of Rag1^−/− mouse jaw sections stained with sera (green) from NOD Aire^−/− mice (upper panel) or WT mice (lower panel), amelogenin-specific antibodies (red) and DAPI (blue). Scale bar 100 µm. g, Slot-blot analysis of primary ameloblast tissue lysate probed with commercial antibodies to the indicated ameloblast-specific proteins. h, Slot-blot analysis measuring the reactivity of sera from NOD Aire^−/− (n ≥ 9) or WT mice (n ≥ 5) to the indicated ameloblast-specific recombinant proteins. Upper panel shows mouse IgG/M sera reactivity against ameloblast-specific protein. Lower panel shows the same membrane blotted with commercial antibodies specific to the corresponding recombinant protein, or ponceau staining. Each lane represents an individual mouse. i, ELISA analyses measuring the reactivity of IgG/M specific to the indicated ameloblast-specific proteins in sera from NOD Aire^−/− (n = 14, n = 12 for ODAM and FAM20A) or NOD WT mice (n = 14, n = 13 for FAM20A, n = 9 for ODAM). j, ELISA analyses measuring the reactivity of IgA specific to the indicated ameloblast-specific proteins in sera from NOD Aire^−/− (n = 9) or NOD WT mice (n = 6). k, ELISA analyses showing distribution of IgG/M sera reactivity of female vs male in NOD Aire^−/− (n = 14; M = 7, F = 7) or WT mice (n = 14; M = 7, F = 7). n refers to the number of biologically independent mice per genotype or group. Data are representative of at least three independent experiments. Data shown as mean ± s.e.m. Data were analysed using two-sided Fisher exact test (i,j) and two-way ANOVA following Bonferroni post-tests (k).

Source Data

a, Slot-blot analyses measuring the reactivity of sera from Aire^−/− (n ≥ 8) or WT rats (n ≥ 5) to the indicated ameloblast-specific recombinant proteins. Upper panel shows rat IgG sera reactivity against ameloblast-specific protein. Lower panel shows the same membrane blotted with commercial antibodies specific to the corresponding recombinant protein, or ponceau staining. Each lane represents an individual rat. b,c, ELISA analyses measuring the reactivity of IgG (b) or IgA (c) specific to the indicated ameloblast-specific recombinant proteins in sera from Aire^−/− (n ≥ 15) or WT (n ≥ 5) rats. d, Representative immunofluorescence microscopy images of Rag1^−/− mouse jaw sections stained with sera (green) from paediatric patients with APS-1 (upper panel) or healthy control (lower panel), with LAMB3-specific antibodies (red), and DAPI (blue). Scale bar 50 µm. e–g, ELISA analyses measuring the reactivity of IgA/IgG1 specific to the indicated ameloblast-specific proteins in sera from adult (e and g) or paediatric (f) patients with APS-1 (adults n = 22, paediatric n = 17) or healthy controls (adults n = 20, paediatric n = 18). n refers to the number of biologically independent rats/patients per genotype or group. Data are representative of at least three independent experiments. Data shown as mean ± s.e.m. Data were analysed using two-sided Fisher exact test (b,c, e–g).

Source Data

a, ELISA analyses measuring the reactivity of IgG2, IgG3 and IgG4 antibodies in sera isolated from adult patients with APS-1 (n = 10) or healthy controls (n = 5) to the indicated ameloblast-specific recombinant proteins. b, ELISA analyses showing distribution of IgA/IgG1 sera reactivity of adult females versus males in patients with APS-1 (Females n = 11; Males n = 12, n = 11 for ENAM). c, Violin plots showing enamel defects score distribution of each tooth type (Maxillary teeth) of 4–29-year-old patients with APS-1 (n = 14). Red line represents median value. d, Representation of enamel calcification duration (in years) for each tooth type. e, Correlation between the duration of enamel calcification in years (Molars=3 y, Incisors=4.7 y, Canines=6.5 y) versus the mean enamel defects score for each tooth type (Molars=1.25, Incisors=1.8, Canines=2.43) (in Maxillary teeth) of paediatric patients with APS-1 (n = 14). f, Comparison of gross teeth appearance in 14–24-week-old NOD Aire^−/− and WT mice. Representative images of ~45 mice teeth examined. g, Heatmap showing distribution of enamel area (mm²) within the left and right (L&R) incisors of the individual NOD Aire^−/− (n = 12) and WT (n = 11) mice. The length of the individual incisors was normalized, and the specific relative position on the tooth displayed in percentiles (y axis). The heatmap range was defined individually per tooth to present the largest (red), median (white) and smallest (blue) enamel area per tooth. h, Presentation of enamel mineralization (a.u.) along the relative (%) incisor location (from lateral to apical part) in representative NOD Aire^−/− (n = 12) and WT (n = 11) mice. Vertical dotted lines indicate a region corresponding to the apical part (that is the 75th−95th percentile of the incisor’s length). n refers to the number of biologically independent mice/patients per genotype or group. Data are representative of at least two independent experiments. Data shown as mean ± s.e.m. Data were analysed using two-sided Fisher exact test (a), two-sided unpaired t-test (b), one-way ANOVA following Tukey’s multiple comparisons test (c) and two-sided Pearson R correlation test (e).

Source Data

a, Illustration of two alternative autoreactive mechanisms responsible for defective enamel matrix formation. The response mediated by self-reactive T-cells (left) would result in the destruction of the ameloblast layer (in an analogous manner to the autoimmune destruction of β-cells in type 1 diabetes or retinal degeneration in Aire^−/− mice⁸⁰), while the autoantibody-mediated response (right) would interfere with enamel matrix formation (e.g. through blocking functional epitopes on enamel matrix proteins), rather than cause destruction of ameloblasts. Created with BioRender.com. b, Representative H&E staining images of jaws isolated from NOD Aire^−/− and WT mice showing enamel matrix and ameloblast cells in secretory stage at a proximal and distal orientation. Red arrows indicate immune cell infiltrates. Left and middle images scale bar 50 µm, right images scale bar 20 µm. c, Representative immunohistochemical analysis images of CD8⁺ T-cells, F4.80⁺ macrophages, CD11c⁺ Dendritic cells and CD20⁺ B-cells within the ameloblast layer, in teeth sections from NOD Aire^−/− and Aire^+/− mice. Scale bar 50 µm. d, Representative immunofluorescence microscopy image of CX3CR1-GFP mouse jaw section stained with amelogenin-specific antibodies (red) and DAPI (blue). Scale bar 20 µm. e, Representative immunofluorescence microscopy image of NOD Aire^−/− mouse jaw section stained with MHC-II (RT1B)-specific antibodies (green), amelogenin-specific antibodies (red) and DAPI (blue). Scale bar 100 µm. f, Representative immunofluorescence microscopy images of CX3CR1-GFP mouse jaw section stained with amelogenin-specific antibodies (red), MHC-II (I-A/I-E)-specific antibodies (purple) and DAPI (blue). Scale bar 20 µm. g, Illustration of the serum transfer experimental design shown in h–k. Created with BioRender.com. h, Percentage of mean variability in enamel mineralization at the indicated apical part between left and right incisors of WT mice transferred with sera isolated either from aged NOD Aire^−/− (red) (n = 6) or WT (blue) (n = 5) mice in two independent experimental cohorts. i, Presentation of enamel mineralization (a.u.) along the relative (%) incisor location (from lateral to apical part) in WT mice transferred with sera isolated either from aged NOD Aire^−/− (n = 6) (red and orange) or WT (n = 5) (dark and light blue) mice. Vertical dotted lines indicate a region corresponding to the apical part (that is the 75th−95th percentile of the incisor’s length). Red arrows highlight areas with high variability between left and right teeth. Presentation of two independent transfer experiments. j,k, Representative SEM images showing ultrastructure of enamel matrix of teeth isolated from WT mice that were injected with sera isolated from NOD WT (j) or NOD Aire^−/− (k) mice, as illustrated in g. Scale bar 20 µm. l, Illustration of the proposed model for ameloblast-specific autoantibody production in patients with APS-1: as a result of AIRE deficiency and break of central tolerance, autoreactive T-cells escape from the thymus to the periphery (1). The cells circulate the body and enter lymph nodes (2). In parallel, during the formation of the enamel matrix, antigen presenting cells (APC) sample enamel antigens (3) and migrate to the closest draining lymph node (4), where they present the enamel self-antigens to autoreactive CD4⁺ T-cells that later activate autoreactive B-cells to produce antibodies against enamel self-antigens (5). The plasma cells and/or autoantibodies are then transferred via the bloodstream to the dental area, where the enamel matrix is developing, bind the enamel matrix proteins and thereby interfere with their capacity to aid in hydroxyapatite mineralization and deposition and enamel formation (6). Illustration was created with BioRender.com. n refers to the number of biologically independent mice per genotype or group. Data are representative of at least two independent experiments. Data shown as mean ± s.e.m. Data were analysed using two-sided Mann Whitney test (h).

Source Data

a, ELISA analyses measuring the reactivity of IgA specific to the indicated enamel proteins in sera from paediatric patients with coeliac disease (n = 24, n = 23 for AMBN) or control individuals without coeliac disease (n = 21). b, ELISA analyses measuring the reactivity of IgA/IgG1 specific to the indicated enamel proteins in sera from adult patients with coeliac disease (n = 48, n = 39 for AMBN and AMTN) or healthy controls (n = 14, n = 48 for FAM20A, n = 6 for AMBN and AMTN). c, Correlation between IgA reactivity to TGM2 vs IgA reactivity to the indicated enamel proteins in paediatric patients with coeliac disease (n = 24) (black dots) and control individuals without coeliac disease (n = 21) (blue dots). d, Representative immunofluorescence microscopy images of Rag1^−/− mouse jaw sections stained with sera (green) from patients with coeliac disease (upper panel) or healthy control (lower panel), amelogenin-specific antibodies (red) and DAPI (blue). Scale bar 50 µm. e, ELISA analysis measuring the titre levels of TGM2-specific-sera (IgA) isolated from paediatric patients with coeliac disease (n = 24) or control individuals without coeliac disease (n = 21). f, Representative immunofluorescence microscopy images of Rag1^−/− mouse small intestine sections stained with sera (green) from patients with coeliac disease (upper panel) or healthy control (lower panel), TGM2-specific antibodies (red) and DAPI (blue). Scale bar 50 µm. g, Diaminobenzidine (DAB) staining with anti-TGM2 specific antibody showing localization of TGM2 in the developing tooth (IEE=inner enamel epithelium; PreSecA= presecretory ameloblast; SecA=secretory ameloblast) of WT animal. Scale bar 50 µm (upper panel), 25 µM (middle panel) and 20 µM (lower panel). h, Representative immunofluorescence microscopy images of Rag1^−/− mouse jaw sections stained with sera (green) from patients with coeliac disease (upper panel) or healthy control (lower panel), TGM2-specific antibodies (red) and DAPI (blue). Scale bar 50 µm. i, Presentation of enamel area (mm²; upper panel) or mineralization (a.u.; lower panel) along the relative (%) incisor location (from lateral to apical part) in representative Tgm2^−/− (n = 4) (orange) and WT (n = 4) (blue) mice. Vertical dotted lines indicate a region corresponding to the apical part (that is, the 75th−95th percentile of the incisor’s length). j, Relative expression of LAMB3 gene in different human tissues, based on scRNA-seq datasets publicly available at the human protein atlas database. k, Representative immunofluorescence microscopy images of Rag1^−/− mouse small intestine sections stained with sera (green) isolated from patients with coeliac disease celiac (upper panel) or healthy control (lower panel), laminin beta subunit 3-specific antibodies (red), TGM2-specific antibodies (yellow) and DAPI (blue). Scale bar 50 µm. n refers to the number of biologically independent mice/patients per genotype or group. Data are representative of at least two independent experiments. Data shown as mean ± s.e.m. Data were analysed using two-sided Fisher exact test (a, b and e), two-sided Pearson R correlation test (c).

Source Data

a, Heatmap showing normalized expression values (as log2) of Csn3 in tooth germ and ameloblast (determined by RNA-seq analysis^26,27,28) in comparison to known amelogenesis imperfecta linked genes, other casein genes and Aire. b, Relative gene expression of indicated ameloblast specific genes and Csn3 in neonatal jaw, salivary gland and liver as assayed by qPCR (n = 6 biologically independent animals). c, RNA-ISH analysis of Csn3 mRNA expression and localization in a mouse jaw. Longitudinal sections of mouse incisors where different stages of ameloblast differentiation are visible. Right panel staining with sense probe as a negative control. Left panel, staining with Csn3 probe. Black arrow indicates the position of the ameloblasts layer. Red arrow indicates the position of the odontoblasts layer. Scale bar 200 µm. d, Illustration of the Csn3 locus and the CRISPR design used for deleting Csn3 exon 1. DNA represented as a black line, with exon location indicated as grey frames. The location of the guide RNAs is indicated in pink frames. Primers for DNA sequencing are shown with arrows representing the amplification product of each reaction. Red filled area represent sequenced DNA. Illustration Created with SnapGene software 6.1 (GSL Biotech). Bottom, alignment of Csn3^+/+ and Csn3^−/− mice DNA sequence at the Csn3 gene locus. Exons are marked in yellow frames. e, Illustration of the PCR genotyping strategy of Csn3^−/− mice. Illustration was created with BioRender.com; on the right is PCR analysis of Csn3^+/+ (WT), Csn3^−/+ (het) and Csn3^−/− (KO) mice. Primer A, hybridizing upstream to exon 1 of Csn3, in combination with primer B results in amplification of a 276-bp fragment indicative of the WT Csn3 locus. Combination of primer A with primer C yields a 433-bp fragment specific for the knockout Csn3 locus, or 2436-bp fragment that cannot be seen in the gel and is specific to the WT Csn3 locus. f, Relative gene expression of Csn3 gene in neonatal jaw of Csn3^−/− and WT mice as assayed by qPCR. Data are representative of at least two independent experiments. Data shown as mean ± s.e.m. Data were analysed using one-way ANOVA following Tukey’s multiple comparisons test for each gene separately (b), and two-sided unpaired t-test (f).

Source Data

a,b, Presentation of enamel area (mm²) (a) or mineralization (a.u.) (b) along the relative (%) incisor location (from lateral to apical part) in representative Csn3^−/− (n = 7) (orange) and WT (n = 7) (blue) mice. Vertical dotted lines indicate a region corresponding to the  the apical part (that is, the 75th−95th percentile of the incisor’s length). c,d, Percentage of mean variability in enamel area (mm²) (c) or mineralization (a.u.) (d) at the indicated apical part between left and right incisors of Csn3^−/− (red) (n = 7) or WT (blue) (n = 7) mice. e, Masks of sections from molar teeth roots (M₁, M₂ and M₃) of WT (upper panel) or Csn3^−/− (lower panel) mice. Root M₂ marked by yellow frame. Structure abnormalities marked by red arrows. f, Representative images of µCT analysis on molars of Csn3^−/− or WT mice, assessing dentine root thickness (upper panel) and pulp root thickness (lower panel) in pseudo colours. Images are of sagittal sections. h, ELISA analyses measuring the reactivity of IgA (g) or IgG1 (h) specific to bovine CSN3 in sera from paediatric patients with coeliac disease (n = 24) or control individuals without coeliac disease (n = 21). i, Correlation between reactivity of IgA vs IgG1 antibodies specific to bovine CSN3 in paediatric patients with coeliac disease (n = 24), as measured by ELISA. j, ELISA analysis measuring the reactivity of IgG1 specific to bovine CSN3 in sera isolated from adult patients with coeliac disease (n = 48) or healthy controls (n = 14). k, Correlation between reactivity of IgA vs IgG1 antibodies specific to bovine CSN3 in adult patients with coeliac disease (n = 48), as measured by ELISA. l, Amino-acid alignment of bovine and human CSN3 proteins. Domain structure of CSN3 protein after cleavage: Signal peptide (blue), para-kappa-casein (orange) and macro-peptide (green). Highly conserved epitopes between the bovine and human homologues are marked in red frames. m,n, ELISA analyses measuring the reactivity of IgA (m) or IgG1 (n) specific to human CSN3 in sera from paediatric patients with coeliac disease (n = 24) or control individuals without coeliac disease (n = 21). o, ELISA analysis measuring the reactivity of IgG1 specific to human CSN3 in sera from adult patients with coeliac disease (n = 48) or healthy controls (n = 14). p, Correlation between reactivity of IgA antibodies specific to bovine versus human CSN3 in paediatric patients with coeliac disease (n = 24), as measured by ELISA. q, ELISA analysis measuring the reactivity of IgA sera obtained from adult patients with coeliac disease (n = 48) or healthy controls (n = 14) towards a library of 25 amino-acid long peptides derived from human CSN3. r, Heatmap showing positive (red) or negative (blue) serum reactivity to the indicated proteins in paediatric patients with coeliac disease (n = 24) and control individuals without coeliac disease (n = 21), based on the ELISA results in Fig. 3a and Extended Data Fig. 6a. Each row represents a patient. s,t, ELISA analyses measuring the reactivity of IgA specific to human (s) or bovine (t) CSN3 in sera isolated from paediatric patients with APS-1 (n = 17) or healthy controls (n = 18). u, Normalized expression of Csn3 gene in WT or AIRE-deficient mTECs isolated from NOD mice. Raw P values were adjusted for multiple testing, using the procedure of Benjamini and Hochberg⁶⁵. n refers to the number of biologically independent mice/patients per genotype or group. Data are representative of at least two independent experiments. Data shown as mean ± s.e.m. Data were analysed using two-sided Mann Whitney test (c and d), two-sided Fisher exact test (g, h, j, m, n, o, s and t), two-sided Pearson R correlation test (i, k and p), and two-way ANOVA followed by Bonferroni multiple comparisons test (q).

Source Data

Illustration of the proposed model for ameloblast-specific autoantibody production in patients with coeliac disease. The inflamed environment of the small intestine in patients with coeliac disease disrupts intestinal barrier capacity and increases intestinal permeability (i.e., leaky gut). This results in generation of antibodies to various food antigens with high immunogenicity (e.g., gliadins or caseins). Reactivity to food antigens is often accompanied by epitope and antigen spreading and subsequent production of autoantibodies to various proteins expressed in the intestinal microenvironment (e.g., TGM2 and LAMB3) (1). The plasma cells and/or autoantibodies are then transferred via the bloodstream to the developing enamel matrix where they can bind to their targets expressed in the dental tissues and enamel matrix (2). The opsonized enamel autoantigens can then be internalized by dental macrophages (and possibly other resident phagocytes), which serve as local sentinels and bind these autoantibodies via their FCGR/FCAR receptors expressed on the surface of these cells (3). Such interactions facilitate the internalization of the antibody-self-antigen complexes by the phagocytes (4). The engulfed antigens will then be presented by the phagocytes on their MHCII molecules in the context of inflammation to self-reactive T cell escapees that are normally kept in check by peripheral mechanisms of self-tolerance. This further potentiates the formation of the adaptive immune response, including further antibody formation and epitope spreading to enamel-specific proteins, in a self-perpetuating mechanism. Illustration was created with BioRender.com.