Epithelial cytoprotection sustains ectopic expression of tissue-restricted antigens in the thymus during murine acute GVHD

Self-tolerance of the nascent T-cell receptor repertoire is attained through negative selection in the thymus.¹  Essential for clonal deletion is the exposure of thymocytes to self-antigens, including those with highly restricted tissue expression. Within the thymus, the ectopic expression of many tissue-restricted peripheral self-antigens (TRAs) is a distinct property of mature medullary thymic epithelial cells (mTEC^high).^2,3  TRA expression is controlled partly by the transcription factor autoimmune regulator (Aire).³  Deficits in Aire and/or TRA expression (independent of their causes) impair negative selection and can consequently cause autoimmune disease.^2-5 

Acute and chronic graft-versus-host disease (aGVHD and cGVHD, respectively) remain the primary complications of allogeneic hematopoietic stem cell transplantation (allo-HSCT).⁶  aGVHD is initiated by alloreactive donor T cells and targets a restricted set of tissues, including the thymus.^7,8  The development of human aGVHD predisposes to cGVHD, whose autoimmune manifestations are integral components of disease.⁹  It remains uncertain, however, whether and how autoimmunity is linked to antecedent alloimmunity. A hallmark of murine aGVHD is the de novo generation of autoreactive T cells from donor HSC^10,11  which can mediate the evolution from acute to chronic GVHD.^12,13  Because the thymic epithelium is a target of donor T-cell alloimmunity,^7,8,14  we hypothesized that thymic aGVHD interferes with the mTEC^high capacity to sustain TRA diversity, thus decreasing the platform for central tolerance. In this context, interventions directed at epithelial cytoprotection are expected to maintain posttransplantation mTEC^high integrity and function. To test these 2 interrelated hypotheses, we used murine allo-HSCT models and investigated the effects of fibroblast growth factor-7 (Fgf7), which boosts thymopoiesis through its mitogenic action on TEC.^15,16 

Female C57BL/6 (B6;H-2^b), (C57BL/6xDBA/2)F₁ (BDF₁;H-2^bd), Balb/c (H-2^d), CBy.PL(B6)-Thy1^a/ScrJ (Balb/c-Thy1.1;H-2^d), 129/Sv (H-2^b), and B6.SJL-Ptprc^aPep3^b/BoyJ (B6.CD45.1;H-2^b) mice were kept in accordance with federal regulations. Thymic aGVHD was induced in unirradiated or total body–irradiated recipients¹⁴  (supplemental Figure 1). Recombinant human Fgf7 (palifermin, Kepivance, Biovitrum, Sweden) was injected intraperitoneally from day −3 to day +3 after allo-HSCT at a dose of 5 mg/kg per day.^15,16 

Confocal microscopy of thymic sections was performed using a Zeiss LSM510 (Carl-Zeiss AG, Switzerland).¹⁷  The TEC compartment was analyzed by flow cytometry (FACSAria, Becton Dickinson, Mountain View, CA). mTEC^high were identified as cells with a CD45^-EpCam⁺Ly51^-UEA1⁺MHCII^high phenotype^17-19  (supplemental Figure 2). Global gene expression profiling was used to establish TRA diversity in transplant recipients. Microarray data of triplicate samples of pure mTEC^high preparations (Mouse Gene 1.0 ST Array, Affymetrix, Santa Clara, CA; data available under accession number E-MEXP-3745 at http://www.ebi.ac.uk/arrayexpress) were verified by quantitative polymerase chaine reaction. Gene expression data from multiple published sources^20,21  and from MOE430 and GNF1M gene atlases (http://biogps.org) were combined for identification of Aire-dependent and Aire-independent TRA (supplemental Methods). Bioinformatics was implemented using software from “The R Project for Statistical Computing” (http://www.r-project.org).

Total numbers of mTEC^high declined over time in untransplanted adult BDF₁ and B6 mice, confirming age-dependent changes of the thymic stroma²²  (Figure 1A-B). We then investigated 4 mouse models of age-matched allo-HSCT. In unconditioned recipients of haploidentical HSCT (H-2^b→H-2^bd; b→bd), the induction of thymic aGVHD^14,15  (supplemental Figures 1 and 2) progressively reduced total mTEC^high numbers to ≤10³ cells/mouse in animals surviving aGVHD (Figure 1A). This observation was mirrored in allo-HSCT that included total body–irradiated before MHC-identical, MHC-disparate, or haploidentical transplantation (Figure 1B). In these models, the conditioning initially reduced mTEC^high numbers independently of aGVHD, but cell loss was either more pronounced (b→bd) or more protracted (d→b; b→b) in the presence of disease. Hence, reduction in mTEC^high compartment size is a universal manifestation of thymic aGVHD,⁷  and radiation injury is not mandatory. Contraction of the total mTEC^high pool in unconditioned b→bd recipients corresponded to progressive decreases in the subset that expresses Aire^5,23  (Figure 1C-E). In 12-week-old recipients, residual Aire⁺mTEC^high were low in numbers (<300 cells/thymus) and later became virtually undetectable, owing to the smaller than normal frequencies of Aire-expressing cells among declining mTEC^high numbers.

Given the intimate associations among Aire-deficiency, TRA repression, and autoimmunity,³  we determined whether aGVHD interfered with TRA transcription. Development of thymic GVHD (b→bd) altered global gene expression profiles in total residual mTEC^high cell pools isolated at 2 to 4 weeks after transplantation (Figure 2A, supplemental Table 1). Using a special algorithm (supplemental Methods: “Bioinformatics”), we discriminated between TRA and ubiquitously expressed transcripts (Ub) in mTEC^high. Several hundred transcripts were expressed at reduced levels (>3-fold change) in mice with aGVHD (Figure 2A). Notably, the TRA/Ub ratios for repressed transcripts were significantly higher than would have been expected to be observed by chance (O/E ratio >1.0 TRA; <1.0 Ub; P < .001; Figure 2A, right panel)). A low correlation coefficient between TRA expression in b→bd vs bd→bd mice (Pearson’s r = 0.700 and r = 0.625 in 10- and 12-week-old recipients, respectively) indicated a sizable number of TRAs being reduced during lethal aGVHD (Figure 2B, supplemental Figure 3). We interpreted these data such that a contraction in TRA diversity during aGVHD was caused by cellular loss of mTEC^high, in particular Aire-positive mTEC. This explanation was based on the fact that ectopic TRA expression is a stochastic process in which only a limited number of mTEC^high express a given TRA.^2,18  A comprehensive coverage of TRA expression is therefore only achieved by a sufficiently large mTEC^high pool, which is, however, missing under conditions of aGVHD. The most substantially reduced TRAs were found to be enriched for genes that are characteristically expressed in tissues known to be targets of cGVHD⁹  (Figure 2C). Many of these genes belonged to the subset of Aire-dependent TRAs²⁰  (Figure 2D,F, supplemental Figure 3).

Because efficient central tolerance requires the full scope of TRA, we next asked whether mTEC^high compartment size and heterogeneity is sustained by using strategies for epithelial cytoprotection. Peritransplant administration of the TEC mitogen Fgf7^15,16  failed to prevent initial mTEC^high cell loss in b→bd recipients (Figure 1A). However, Fgf7 maintained stable numbers of mTEC^high, including the subset expressing Aire (∼3000 cells/mouse during aGVHD; Figure 1C-E). Consistent with reports that mTEC are continuously replaced in adulthood,^18,22  ∼5% of mTECs underwent cell division within 1 week in the b→bd and bd→bd transplant groups (Figure 2E). This frequency increased to ∼15% in mice with aGVHD but that treated with Fgf7. Label retention analyses indicated high turnover of mTEC in response to Fgf7 in b→bd recipients (Figure 2E). Rescue of mTEC^high numbers was therefore the consequence of enhanced proliferation within the entire mTEC compartment. During early aGVHD, the TRA spectrum was not preserved by Fgf7 (Pearson’s r = 0.635 vs bd→bd mice; Figure 2B, supplemental Figure 3). However, a broader array of TRA was expressed by 7 weeks after allo-HSCT as indicated by an expression profile that was closer to that of age-matched bd→bd controls (r = 0.807). Hence, peritransplant administration of Fgf7 sustained a more diverse TRA transcriptome during aGVHD. Individual TRA expression patterns were differentially affected by aGVHD and Fgf7 as some but not all Aire-dependent TRA returned to nearly normal expression in the observation period (Figure 2F, supplemental Figures 3 and 4). However, the molecular mechanism for this biased profile remains to date unknown.

Taken together, our data provide a mechanism for how autoimmunity may develop in the context of aGVHD (supplemental Figure 5). Because thymic-negative selection is sensitive to small changes in TRA expression,^4,5  approaches for mTEC cytoprotection may prevent the emergence of thymus-dependent autoreactive T cells¹²  and thus alter autoimmune outcome. The identification in cGVHD of the yet undefined specificities of autoreactive effector T cells²⁴  will allow to test this argument directly in experimental systems and ultimately in clinical allo-HSCT.

The authors thank Philippe Demougin (Life Sciences Training Facility, LSTF, Biocenter, University of Basel) for DNA microarray analyses, and Dr Gabor Szinnai (University Children’s Hospital, Basel) for critically reviewing our manuscript.

This work was supported by Swiss National Science Foundation grants 310030-129838 (W.K.) and 310010-122558 (G.A.H.), and by a grant from the Hematology Research Foundation, Basel, Switzerland (W.K.).

Contribution: S.D., G.N., and M.M.H.-H. designed and performed work; R.I. implemented bioinformatics and statistics; and W.K. and G.A.H. share senior authorship and both designed the work and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Werner Krenger, Department of Biomedicine, University of Basel, Mattenstrasse 28, 4058 Basel, Switzerland; e-mail: werner.krenger@unibas.ch.