Absence of functional fetal regulatory T cells in humans causes in utero organ-specific autoimmunity

Capsule Summary

Regulatory T cells play a critical role in preventing fetal organ-specific autoimmunity in humans. Autopsies of neonatal IPEX patients shortly after birth demonstrate chronic exocrine-dominant pancreatitis with tertiary lymphoid structures containing expanded oligoclonal T/B lymphocytes.

To the Editor:

Fetal T regulatory cells (Tregs) are present by 13 weeks gestation, but their role during the fetal period is unclear. Maternal Tregs clearly are critical for fetal tolerance. Human fetal Tregs promote tolerance to non-inherited maternal antigens (NIMA) in utero¹, but whether tissue specific self-tolerance is needed in utero is unknown. During pregnancy, fetuses with the genetic disorder IPEX (Immune dysregulation Polyendocrinopathy Enteropathy X-linked) syndrome lack functional Tregs, but maternal Treg cells remain functional. IPEX patients often appear healthy at birth, but develop early systemic autoimmunity including early-onset diabetes, enteropathy, thyroiditis and dermatitis. Timing of the initial organ-specific inflammation remains unclear. Early fetal or perinatal IPEX presentations are reported^2-4, but evidence for in utero organ-specific autoimmunity is lacking. In this study, we report two IPEX patients that died shortly after birth with histological evidence for tertiary lymphoid structures, chronic inflammatory changes and targeted exocrine pancreas autoimmunity in the absence of clinical diabetes. Repertoire analysis demonstrated clonal enrichment within the pancreas consistent with an antigen-driven germinal center reaction. A murine model with inducible inactivation of Tregs demonstrated similar exocrine-dominant lymphocytic infiltrates in the pancreas. Thus, absence of functional Tregs promotes organ-specific, exocrine pancreas autoimmunity in utero.

Patient 1 was prenatally diagnosed with IPEX syndrome secondary to a family history of a known pathogenic missense mutation in FOXP3 (c.1087A>G, p.I363V) (see Fig E1 in this article’s Online Repository at www.jacionline.com). Maternal polyhydramnios developed in the absence of fetal hydrops or growth restriction. Prenatal ultrasound detected hyperechoic skin, but was otherwise reassuring. Labor was induced at 39 weeks and immediately the infant developed unexpected respiratory failure requiring intubation. Blood glucose was normal (57mg/dL). Hematocrit was 43%, platelets 194 thousand/uL and WBC 10.34 thousand/uL, with only 1% immature granulocytes. Despite oscillatory ventilation, the child lived only 29 hours and died from respiratory failure. Blood cultures and viral infectious screening were negative and placental pathology was normal. A complete autopsy was performed and revealed severe pulmonary hypoplasia as the etiology for respiratory failure. Regulatory T cell phenotyping performed on cord blood confirmed a lack of CD4⁺FOXP3^hi CD25^hi T cells. Histologic evaluation revealed prominent lymphocytic infiltrates in the pancreas, gastric mucosa and thyroid glands but a lack of inflammation in the testes, adrenal and pituitary glands, a pattern typical for IPEX syndrome (see Fig E2 in this article’s Online Repository at www.jacionline.com). The pancreas infiltrate showed tertiary lymphoid-like structures with extensive T cell (CD3+) zones, including mixed CD4+ and CD8+ cells, surrounding distinct B cell (CD20+) aggregates (Figure 1. Chronic inflammatory changes were present including squamous ductal metaplasia, acinar atrophy and stromal fibrosis. We screened inflamed pancreas sections using interphase fluorescence in situ hybridization (iFISH) for X and Y chromosomes (200 cells analyzed) and found an exclusively male (XY) karyotype with no maternal (XX) cells. The histologic findings indicate chronic in utero inflammation, rarely described in fetal tissues.

Figure 1. Exocrine-predominant pancreatitis in neonatal IPEX patient.

Pancreas tissue from Patient 1 at age 29 hours (columns 1-2), Patient 2 at 19 days (column 3) and representative control patient (column 4) were serial sectioned and stained with H&E (row 1) and antibodies against CD3 (row 2), CD20 (row 3) and insulin (row 4, columns 2-4). Islets of Langerhans (*) were essentially devoid of lymphocytic infiltrates, best appreciated on high power images (CD3/CD20 – column 1, row 4 [400X]). Insulin staining highlights equivalent quantities of endocrine elements in all three samples (row 4, columns 2-4). Dense infiltrates permeated acinar compartments in IPEX cases, where tertiary lymphoid structures were evident, while lymphocytes were rare in control samples. [(*) Islet of Langerhans; (arrowhead) ectopic germinal center].

An additional neonatal IPEX autopsy case, Patient 2, revealed similar findings. Patient 2 was the proband with several family cases (FOXP3 c.1189C>T, p.R397W) (see Fig E1 in this article’s Online Repository at www.jacionline.com). He died at day 19 from peritonitis. Pancreas histology also showed a lymphocyte-rich mononuclear infiltrate within the pancreas, but with more advanced fibrosis (Figure 1A). Focal clusters of CD20+ B cells were surrounded by dense CD3+ T cell zones, with mixed CD4+ and CD8+ staining, as seen in Patient 1. In contrast to the two IPEX patients, control neonatal pancreas tissue obtained from age-matched autopsy cases (n=5) showed no lymphocytic infiltrates regardless of the cause of death (Figure 1). RNA-seq analysis from Patient 1 pancreas tissue showed increased expression of CCL19, CCL21, CCL22 and LTB (lymphotoxin beta) transcripts known to be involved in tertiary lymphoid organization, type 1 diabetes and in interferon-driven pro-inflammatory chemokines (CXCL9, CXCL10, CXCL11) known to recruit CXCR3+ activated T cells (see Fig E3 in this article’s Online Repository at www.jacionline.com). Overall, the chronic and organized nature of the pancreatic inflammation in both IPEX subjects suggests a role for Treg cells in restraining selfreactive T cell responsiveness prior to microbial colonization.

Interestingly, islets were structurally intact and there were no significant inflammatory infiltrates in, or directly surrounding islets despite the presence of extensive inflammation within surrounding exocrine tissue (Figure 1, columns 1-2>). Indeed, there was preferential loss of exocrine tissue compared to age-matched control tissue. Quantitative measurement of insulin-staining islet tissue in both IPEX patients revealed sparing of the islets with an insulin-staining area similar to control autopsy pancreas (see Fig E4 in this article’s Online Repository at www.jacionline.com). Even in the 19-day old infant (Patient 2) with more advanced fibrosis, many of the islets were intact and the area of insulin staining was only mildly reduced compared to controls. Although the ratio of exocrine to endocrine gland mass in the developing human pancreas is endocrine biased compared to adult tissue, the lack of exocrine tissue in the early IPEX pancreas tissue was more pronounced than age-matched control samples.

Exocrine-dominant pancreatitis has been previously described in FoxP3-deficient Scurfy mice⁵. To directly assess whether the exocrine-dominant early inflammation was a general feature of Treg dysfunction versus a feature of loss of tolerance at a distinct stage in pancreatic development, we employed an inducible IPEX-like mouse model of Treg inactivation to analyze early pancreas inflammation following normal lymphoid development. Inducible Treg depletion has been demonstrated to rapidly result in a scurfy like phenotype⁶, but a transient lymphopenia has also been associated with the Foxp3-DTR model. Our group bred mice with homozygous floxed Cdc42 alleles⁷ to mice with Foxp3^Cre-ERT2 allele⁸ that experience normal development (data not shown). Treatment of this strain with tamoxifen leads to inducible Treg-specific cre-recombinase deletion of the Cdc42 gene rendering Tregs dysfunctional. These mice quickly succumb to IPEX-like pathology similar to the inducible Foxp3^DTR-GFP mice⁶.

Cdc42^flox/flox Foxp3^Cre-ERT2 mice treated with peritoneal tamoxifen injections demonstrated perivascular and exocrine-predominant lymphocytic inflammatory infiltrates in pancreatic tissue similar to the human pancreas tissue from Patient 1 (Figure 2A,B). In contrast, treatment of C57Bl/6 control mice led to mild serosal inflammation at the edges of tissue flanking the peritoneum (arrows) but no significant pancreas tissue inflammation and no lymphocytic infiltrate into glandular tissue (Figure 2C,D). Non-injected C57BI/6 control mice had no inflammation (Figure 2E,F). In the mice with dysfunctional Tregs, most inflammatory infiltrates were perivascular and involved exocrine acini sparing islet tissue. Minimal endocrine inflammation was noted (islets marked *) where infiltrates were more intense. These observations closely parallel the exocrine focused inflammation observed in the two IPEX subjects in this study.

Figure 2. Tamoxifen-inducible exocrine-dominant pancreatitis in the absence of functional regulatory T cells in mouse pancreas.

Representative H&E stained pancreas tissue from (A,B) Cdc42^flox/floxFoxp3^ER-Cre mouse (n=2) and (C, D) control C57Bl/6 mouse (n=2) receiving tamoxifen intraperitoneal injections for 19 days or from (E,F) an untreated control C57Bl/6 mouse. (A, B) Cdc42^flox/floxFoxp3^ER-Cre animals exhibited perivascular and exocrine inflammation (arrowheads) associated with acinar destruction despite unperturbed islets. (*) Islet of Langerhans. There was no inflammation in the pancreas of (D, F) either control mouse. Only moderately intense mononuclear cell infiltrates were restricted to the mesentery (arrowhead) in the control C57BI/6 mouse. (H&E, 40X, A, C, E; 200X, B, D, F).

Tissue clonal enrichment can indicate antigen-specificity. We compared the frequency of T and B cell clones in pancreatic tissue with those in peripheral cord blood in Patient 1. We identified unique T and B cell clones and examined global repertoire using next-generation repertoire sequencing (see Fig E5 in this article’s Online Repository at www.jacionline.com). There were no dramatic differences in overall V_H gene family or TRB gene family usage comparing lymphocytes from cord blood and pancreas tissue with similar distributions when compared to previous reports on human neonatal or adult repertoires⁹. Despite equivalent V_H gene family or TRB gene family usage, clonal enrichment in the pancreas could be clearly observed at the sequence level when compared to cord blood (see Fig E5-F,G in this article’s Online Repository at www.jacionline.com). Defining a unique clone as identical in nucleotide sequence, we utilized an algorithm to identify differential abundance between the pancreas and cord blood for clones based upon IGH and TCRβ total productive rearrangements. The TCR compartment also showed significant oligoclonal enrichment in the pancreas compared to cord blood. Several of the top B and T cell clones demonstrated more than 100-fold enrichment in the inflamed pancreas compared to the cord blood. The pancreatic B cell compartment showed dramatic enrichment in frequency of clones with hundreds of clones >0.1% with the highest being 0.325% of productive templates (251 of 77,179 templates) with a single CDR3 amino acid sequence and another 0.17% with only one amino acid difference (132 templates) not detected in the cord blood. The top IGH clonal abundance in the cord blood was only 0.038% (25 of 65,025 templates). Importantly, nucleotide mutations in the V genes within IgH clones in the inflamed pancreas tissue occurred (n=133) at a 10-fold greater rate compared to that observed in cord blood consistent with selective pressure from antigen recognition and/or somatic hypermutation. Gene transcript analysis of the tissue also showed increased expression of AICDA, IL21R, IRF4, CXCR5, FDCSP, and numerous TNFRSF members with known roles in germinal center formation (see Fig E3A in this article’s Online Repository at www.jacionline.com). Overall, the observed oligoclonal enrichment strongly supports the presence of antigen-driven autoimmunity within the pancreas prior to birth in IPEX syndrome.

Our work highlights a role for human fetal regulatory T cells in controlling in utero tissue inflammation to fetal self-antigens. Neonatal autopsy results from two unrelated IPEX cases showed inflammation prior to the development of overt diabetes revealing impressive and organized inflammatory infiltrates at birth. Areas of fibrosis and chronic inflammatory changes were present at birth and there was oligoclonal expansion of pancreas T and B cells suggestive of an antigen-targeted autoimmune process in utero. While IPEX subjects are only rarely identified prenatally, our finding suggest that maternal immune suppression using an agent such as tacrolimus that crosses the placenta may be beneficial in such settings. Our combined observations suggest that active fetal Treg-mediated immune suppression occurs in utero to prevent autoimmunity and implies that fetal, not just maternal, suppression and tolerance mechanisms are operative during pregnancy and fetal development. Further work will be needed to elucidate the role of the developing neonatal immune system on the developing fetus.

METHODS:

Human subjects.

The study was approved by the local ethical institutional review board with written informed consent for phenotypic, functional and genetic analysis according to the requirements of the review board. Autopsies were performed according to standard practices.

Mice.

C57BL/6, Cdc42^flox/flox FoxP3^{tm9(EGFP/cre/ERT2)Ayr/J} were generated by crossing Cdc42^flox/flox mice to FoxP3^{tm9(EGFP/cre/ERT2)Ayr/J} mice and intercrossing. Mice were maintained in the specific pathogen-free animal facility of Seattle Children’s Research Institute (Seattle, WA) and handled according to Institutional Animal Care and Use Committee approved protocols.

Human regulatory T cell FOXP3 flow cytometry staining:

Peripheral blood mononuclear cells (PBMCs) were isolated from heparinized blood using Ficoll-Paque PLUS (Biosciences AB, Uppsala, Sweden) and phenotyped but multicolor flow cytometry with the following antibodies: Anti-Human Helios Alexa 647 (Biolegend), Anti-Human CD4 Alexa 700 (BD Biosciences), AntiHuman CD25 PE-Cy7 (Biolegend), Anti-Human FOXP3 Alexa 488 (Biolegend), Mouse IgG1, kappa isotype control Alexa 488 (Biolegend). Prior to staining cells were prepared using 4X FoXp3 Fix/Perm solution (Biolegend), 10X FOXP3 Perm Buffer (Biolegend). Cell events were acquired on an LSR II (BD) and analyzed using FlowJo softward (Tree Star).

Genetic analysis:

The FOXP3 gene (OMIM: *300292) was amplified from genomic DNA (gDNA) and cDNA by PCR. Briefly, gDNA was prepared from whole blood by QIAamp DNA Blood kit (QIAGEN, Valencia, California, USA) and amplified gene fragments were sequenced by Sanger sequencing.

TCRB and IGH sequencing and analysis

A total of 2-3ug gDNA from umbilical cord blood or pancreas frozen tissue was processed using DNeasy Blood and Tissue Kit (Qiagen). Repertoire sequencing was using a two-step multiplex PCR to amplify and barcode the CDR3 regions for Illumina sequencing prepared by Adaptive Biotechnologies (Seattle, WA) for immunoSEQ human TCRβ and human IgH survey platform. Raw sequence data were filtered based on the TCRβ V, D, and J gene definitions provided by the IMGT (ImMunoGeneTics) database (http://www.imgt.org) and binned using a modified nearest-neighbor algorithm to merge closely related sequences and remove both PCR and sequencing errors. Raw repertoire data was uploaded to IGMT for IGH and TCRB gene family identification, usage and CDR3 analysis. Clonal pairwise comparison was performed using the Adaptive Biotechnologies immunoSEQ Analyzer (Adaptive Biotechnologies, Seattle, WA). Differential abundance plots were generated using the program software^E1. Briefly, the input data consisted of the absolute abundance for each productive TCRβ or IgH clone respectively determined at the nucleotide level. A minimum total inclusion filter of 10 clones in either sample was used to exclude clones with frequencies too low to make statistical inference. Clones were independently tested for significance by using a two-by-two contingency table with abundances of each clone in each sample and remaining abundances. The Fisher exact test was used to compute a p value for each clone across the two samples with the null hypothesis that the population abundance of the clone is identical. pFDR method was applied for each p value to select an appropriate threshold of significance (p<0.01). The productive frequency equality was based upon the sample size to adjust for different sample sizes. Scatter plots were generated for TCRβ or IgH clones separately. Productive frequency was determined by summing the total number of in-frame templates without stop codons for each rearrangement divided by the sum of the templates for all productive rearrangements in the sample.

Pancreas histology

Human pancreas tissue was fixed in neutral buffered formalin, processed, and embedded in paraffin according to standard practices. Tissue sections were stained with hematoxylin and eosin (H&E) or stained by immunohistochemistry for insulin, CD20, CD3, CD68 (Dako; Carpinteria, CA). Insulin containing surface area was visualized and calculated as a percentage of 100X microscopic fields using Nikon Eclipse 80i microscope with digital camera. Images were analyzed using NIS-Elements Advanced Research Software v4.13 (Nikon Instruments Inc, Melville, NY). Results were averaged per case and/or across multiple blocks when available. Mouse pancreas tissue was fixed in neutral buffered formalin, processed, and embedded in paraffin according to standard practices. Tissue sections were stained with hematoxylin and eosin (H&E) and analyzed in a blinded fashion for inflammatory infiltrates.

RNA-seq method and analysis:

Frozen human autopsy pancreas or testes tissue was embedded in OCT compound and frozen over dry ice and isopropyl alcohol. 20μm sections were cut on a cryostat directly into RLT buffer (QIAGEN) and mRNA was extracted using QIAshredder and gDNA was removed using on-column DNAse digestion. RNA quality was assessed using RNA Nano electrophoresis (Agilent). Sequencing libraries were constructed from total RNA using TruSeq RNA Sample Preparation Kits v2 (Illumina) and clustered onto a flowcell using a cBOT amplification system with a HiSeq SR v4 Cluster Kit (Illumina). Single-read sequencing was carried out on a HiSeq2500 sequencer (Illumina), using a HiSeq SBS v4 Kit to generate 58-base pair reads, with target of approximately 25 million reads per sample. Standard RNA-sequencing QC metrics show uniform library counts of ~15million counts per sample aligned to the genome after removal of duplicates. All samples showed >95% counts aligned to the reference genome and medium CV coverage values <1.

Bioinformatic analysis

FASTQ files were downloaded from https://basespace.illumina.com. Libraries were processed via Galaxy^E2-E4. Libraries were aligned via TopHat (v1.4.1)^E5 to the human reference genome, Ensembl’s Homo sapiens GRCh38 version 77 (GRCh38.77.gtf). The single-paired flag was set to “single”, while all other TopHat parameters were set to defaults. HTSeq-count was used to generate gene counts with mode as “Intersection (nonempty)” and minimum alignment quality set to 0 and otherwise set to default parameters. Picard tools was used for the median coefficient of variation of coverage of the 1000 most highly expressed transcripts and plotted versus percent aligned for each sample for quality control. Transcripts were filtered by genelists from Human Immunology Project Consortium’s ImmuneSpace (http://www.immuneprofiling.org/innate/geneList/list). “Type One Diabetes” genelist enrichment analysis calculation was based upon total transcripts included with non-zero value for one of the samples and positive inclusion if >5 fold expression change or absolute count change +/− 500. Hypergeometric testing was performed in R. Heatmap plots are of normalized raw counts were displayed from uninflamed “Testes”, “Head of Pancreas” and “Tail of Pancreas” and plotted using Heatplus package.

Public GEO RNA-seq datasets from fetal pancreas and gonad tissues using a similar analysis pipeline demonstrated a lack of inflammatory transcripts in fetal pancreas. Additionally adult pancreas tissue lacked these inflammatory transcripts as well when querying the pancreatic expression database (www.pancreasexpressiondatabase.org).

Extended Data

Figure E1. Pedigree and Clinical Data.

Family pedigree and clinical information for IPEX Patients 1 and 2 (top), as well relevant clinical information regarding the five control patients used for comparison (bottom). (PNA: pneumonia; GBS: group B streptococcal disease; SMA: spinal muscular atrophy). The familial FOXP3 gene mutations are referenced as both are previously published to be associated with IPEX disease^E2,E3.

Figure E2. Inflamed and noninflamed tissues from neonatal IPEX Patient 1.

Mononuclear cell infiltrates were evident in the (A) pancreas, (B) stomach and (C) thyroid while no inflammation was found in other organs commonly affected in IPEX, including the (D)adrenal, (E) pituitary, or (F) uninflamed testis (Hematoxylin and eosin, 100X [A-C, F], 200X [D,E]).

Figure E3. RNA-Seq data from IPEX patient 1.

RNA-seq data comparing mRNA transcripts in head and tail of pancreas compared to the uninflamed testes from the extracted fetal autopsy frozen tissue. Gene counts were filtered for fold change >5 or transcript difference >500 when compared to uninflamed testes sample. Heatmaps were generated for: (A) Genes involved in tertiary lymphoid structure and/or germinal center formation; (B) Genes known to be upregulated in T1D; or (C) Interferon-driven chemokine and cytokine genes. Gene expression levels are shown as row normalized Z-scores with blue reflecting low relative expression and orange representing high relative expression.

Figure E4. Chronic histologic changes and preserved insulin staining.

(A) Pancreas tissue sections from Patient 1 showed acinar dropout and fibrous inflammatory damage of exocrine ducts that was associated with squamous metaplasia (boxes). Islets of Langerhans were numerous (*) and lacked inflammation or cell destruction. Tertiary lymphoid structures including germinal center formation (arrowheads) were present (40X; all images 100X). (B) Percentage of insulin positive surface area compared among pancreatic tissue from Patient 1 (head and tail included), Patient 2 and five age-matched controls. Each mark represents the average of 10 random, non-overlapping 100X fields.

Figure E5. Oligoclonal enrichment of unique TCR and BCR clones in pancreas compared to cord blood sample.

(A-B) Differential abundance scatter plots were generated comparing pancreas to cord blood to illustrate rearrangements that demonstrate expansion or enrichment based upon nucleotide identity. Testing was excluded on infrequent clones (light grey) determined by minimum threshold (n>10) given the sample size. Pair-wise scatter plots for productive frequency for each rearrangement were plotted between pancreas versus cord blood. Dots displayed on the axis were unique to each tissue. Clones with significant (p<0.01) enrichment in pancreas tissue (red) versus clones enriched in cord blood (blue) and clones lacking significance were also determined (dark grey). Analysis was performed separately for (D) IGH productive rearrangements and (E) TCRB productive rearrangements. The Frequency Equality line represents productive frequency equivalence between samples given total sample size of each. Total productive clonal templates analyzed: IGH cord blood (n=65,025), IGH pancreas (n=77,179), TCRB pancreas (n=193,719), and TCRB cord blood (n=70,668).