Role of the thymus in spontaneous development of a multi-organ autoimmune disease in human immune system mice

Abstract

We evaluated the role of the thymus in development of multi-organ autoimmunity in human immune system (HIS) mice. T cells were essential for disease development and the same T cell clones with varying phenotypes infiltrated multiple tissues. De novo-generated hematopoietic stem cell (HSC)-derived T cells were the major disease drivers, though thymocytes pre-existing in grafted human thymi contributed if not first depleted. HIS mice with a native mouse thymus developed disease earlier than thymectomized mice with a thymocyte-depleted human thymus graft. Defective structure in the native mouse thymus was associated with impaired negative selection of thymocytes expressing a transgenic TCR recognizing a self-antigen. Disease developed without direct recognition of antigens on recipient mouse MHC. While human thymus grafts had normal structure and negative selection, failure to tolerize human T cells recognizing mouse antigens presented on HLA molecules may explain eventual disease development. These new insights have implications for human autoimmunity and suggest methods of avoiding autoimmunity in next-generation HIS mice.

1. Introduction.

Human immune system (HIS) mice, defined as immunodeficient mice engrafted with human immune cells, are widely-used tools for the study of human hematopoiesis, immunity, infectious diseases, autoimmunity, cancer and transplantation biology. These models allow experimental manipulation of human immune cells in vivo [1–3].

HIS mice spontaneously develop a poorly-understood syndrome resembling autoimmunity or Graft-versus-Host Disease (GVHD) at varying times post-transplantation, depending on the model used. This condition reportedly occurs around 20 weeks post-transplantation in NOD-scid common gamma chain knockout (NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ) (NSG) mice receiving fetal human HSCs and thymus tissue, with 100% mortality by 30-35 weeks post transplantation [4]. Although a window of many weeks is available for studies before the mice develop disease, it presents a significant limitation. This disease is characterized by hair loss, weight loss, enlargement of lymphoid organs and infiltration of T cells in non-lymphoid tissues. It has been reported in HIS mice generated with human fetal CD34⁺ hematopoietic stem cells (HSCs) in adult NSG mice, with and without transplantation of a human fetal thymus graft [4–7].

Although different mechanisms have been suggested for development of this disease, its main cause is unknown. Lockridge et al. suggested that the pathology in this model is driven largely by factors associated with the engrafted human thymus and described an inverse correlation of disease severity with the frequency of thymic regulatory T cells (Tregs) [7]. On the other hand, Covassin et al. did not find any association between the disease onset and a loss of human Tregs or absence of intrathymic mouse CD45⁺ cells, which contribute to negative selection of mouse-reactive T cells [4]. We have previously shown that FOXP3⁺ Helios⁺ natural Tregs develop normally in human fetal thymic grafts, are present in the periphery of HIS mice and demonstrate normal suppressive function [8]. We have also shown that human Tregs in grafted human thymus include a highly diverse TCR repertoire [9].

It has not been determined whether the T cells driving this disease are residual T cells originally present in the human thymic graft or are newly-generated HSC-derived T cells. Sonntag et al. have shown that the disease occurs even without a grafted human thymus, where the HSC-derived T cells develop in a native mouse thymus [5]. We have now investigated the contribution of pre-existing thymus graft-derived T cells and HSC-derived T cells that develop in grafted thymus compared to native mouse thymus in the development of disease.

Despite lower proportion of Tregs in native mouse thymus compared to grafted human thymus, peripheral Treg levels were similar between affected and non-affected mice with native and grafted thymi. Through high-throughput TCRβ sequencing, we discovered abnormalities in the repertoire of T cells that develop in native NSG mouse thymus and found that abnormal structure of the native thymus, probably due to the lack of common gamma chain and hence IL-7 signaling, impairs negative selection of autoreactive TCRs and possibly contributes to faster disease development in HIS mice with a native mouse thymus compared to HIS mice with a grafted human thymus. Early-life injection of HSCs delays disease development in HIS mice with a native mouse thymus, but fails to rescue generation of a structurally normal thymus. By comparing disease development in NSG mice lacking mouse MHC I and II, we demonstrate the ability of human T cells recognizing antigens presented on “self” HLA molecules to promote autoimmune disease in the absence of direct antigen recognition.

2. Materials and Methods

2.1. Animals and human tissues and cells

NOD-scid common gamma chain knockout (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) (NSG) and NSG-(K^b D^b)^null (IA)^nul1 (NSG MHC KO) mice were obtained from the Jackson Laboratory or bred in-house and housed in a specific pathogen-free microisolator environment. They were used at 6–10 wk of age. Discarded human fetal thymus and liver tissues of gestational age 17–20 wk were obtained without identifiers from Advanced Bioscience Resources (ABR, Alameda, CA). Informed consent from women choosing to donate fetal remains for research was obtained by ABR. Fetal thymus fragments were cryopreserved in 10% dimethyl sulfoxide (DMSO) and 90% human AB serum (Atlanta Biologicals). Fetal livers (FL) were cut into small pieces and incubated in 5ml of 0.1mg/ml liberase (Liberase TM Research Grade, 05401119001, Roche) for 20 minutes in a 37°C CO2 incubator. Every 5-7 minutes, the samples were vortexed and pipetted up and down to facilitate the digestion process. Digested cells were washed and CD34⁺ cells (referred to as HSCs) were positively sorted with magnetic cell sorting (MACS) (Miltenyi Biotec). From each FL, we harvest 10-30 million CD34⁺ cells with a purity of 80-90%. CD34⁺ HSCs were also cryopreserved in 10% DMSO and 90% human AB serum. Protocols involving the use of human tissues and animals were approved by the Columbia University Medical Center Institutional Animal Care and Use Committee and all of the experiments were performed in accordance with these protocols.

2.2. Generation of human immune system (HIS) mice

Four types of HIS mice were generated with human FL-derived CD34⁺ HSCs: HIS mice with no thymus, with a native mouse thymus, with a human thymus graft, or with both human and mouse thymi. To generate HIS mice with no thymus, NSG mice were thymectomized (as we previously described [10]) and 2 weeks later, sublethaly irradiated (100cGy) and injected i.v. with 1-2x10⁵ human FL-derived CD34⁺ HSCs (Figure S1A). HIS mice with a native mouse thymus were generated similarly, except they were not thymectomized (Figure S1B). Similar to the first group, HIS mice with a human thymus graft were thymectomized. Two weeks later, the mice were injected i.v. with HSCs and transplanted with autologous (to human HSCs) human fetal thymus fragments measuring about 1 mm³ under the kidney capsule (Figure S1C). HIS mice with both human and mouse thymi were generated similar to the HIS mice with a human thymus, except they were not thymectomized (Figure S1D). To ensure that pre-existing T cells in the transplanted donor human thymus were not able to persist, we froze and thawed the thymus tissues as described [11] and also pipetted them up and down several times before transplantation to release as many thymocytes as possible. To further deplete passenger thymocytes that might migrate to the periphery, an anti-human CD2 antibody was injected to the mice in 2 weekly doses (400μg/mouse, intraperitoneally) as we have described [11]. In some studies, human fetal thymus was used fresh with either autologous or allogeneic FL HSCs. In one experiment, HIS mice with native mouse thymus were generated by injecting FL HSCs into neonates 1-2 days after birth (intra-liver, 10⁵ HSCs in 50μl PBS injected using a 28-gauge needle) and, in another group, into adult mice (10⁵ HSCs i.v./ 7-10 weeks after birth). Neonatal mice received half of the irradiation dose (50cGy) that the adult mice received (100cGy).

To investigate the disease development in Personalized Immune (PI) HIS mice (as we previously described [11]) with a human thymus graft, thymectomized NSG mice were sublethally irradiated (100cGy) and injected i.v. with 2x10⁵ bone marrow (BM)-derived CD34⁺ HSCs from an adult donor. Two weeks after thymectomy, the mice were transplanted with human fetal thymus fragments (matched to the BM HSCs for a class I and a class II HLA allele) measuring about 1 mm³ under the kidney capsule (Figure S1E). Similar to generation of mice with fetal HSCs, measures were applied to deplete the pre-existing thymus graft-derived T cells.

2.3. Assessing human immune cell reconstitution using flow cytometry (FCM).

For analysis of human immune cell reconstitution, mice were bled at regular intervals for FCM analysis of human T cells, B cells, conventional dendritic cells (cDCs), and monocytes. Starting with 25 μl of blood at each time point, ACK lysis buffer (Gibco) was used to lyse the red blood cells (RBCs) and counting beads (123count eBeads™ Counting Beads, ThermoFisher Scientific) were used to quantify different cell subsets per μl of blood. Analytic FCM was performed on a Fortessa (BD Biosciences, Mountain View, CA). Cells were stained with the following antibodies directed at: human CD3 (clone OKT3; catalog 317324; BioLegend), human CD45 (HI30, 562279, BD Biosciences), mouse CD45 (30-F11, 563709, BD Biosciences), human CD4 (SK3; 563550, BD Biosciences), human CD8 (SK1; 557834; BD Biosciences), human CD25 (M-A251; 565096; BD Biosciences), human Foxp3 (236A/37; 12-4777-42; eBioscience), human CD14 (61D3, 17-0149-42, ThermoFisher Scientific), human CD11c (B-ly6, 551077, BD Biosciences), human CD19 (HIB19, 302207, Biolegend), human CD31 (WM59, 562855, BD Biosciences), human CD45RA (HI100, 304141, Biolegend), human CD45RO (UCHL1,563722, BD Biosciences), HLA-DR (L234, 307632, Biolegend), Ki67 (B56, 558615, BD Biosciences) and human CTLA-4 (14D3, 46-1529-42, eBioscience). For intracellular staining, eBioscience™ Foxp3 / Transcription Factor Staining Buffer Set (catalog 00-5523-00; ThermoFisher Scientific) was used. Analyses were performed with FlowJo software (Tree Star, San Carlos, CA).

At different times post-transplantation (18-22 weeks for non-affected mice and 26-60 weeks for “affected” mice i.e. mice with clinical GVHD/autoimmunity), HIS mice were euthanized. Grafted thymi (for HIS mice with grafted human thymus), native mouse thymus (for HIS mice with native mouse thymus), spleen, pooled lymph nodes (cervical, axillary, brachial and mesenteric LNs), BM and skin were harvested. Thymocytes, spleen and LN cells were isolated by physical force (crushing the thymus tissue between two slides and crushing the spleen and LNs through a 70μm cell strainer using a syringe plunger). ACK lysis buffer (Gibco) was used to lyse RBCs in spleen cells. BM cells were isolated by flushing the femur and tibia bones (both legs) with PBS using a 28-gauge needle. To isolate skin cells, the trunk area was shaved, a 1cm² piece of skin was removed and washed in PBS containing 2% penicillin/streptomycin. The subdermal fat tissue was removed with a blade. The skin was cut into small pieces using the blade and incubated in 3ml of 0.1mg/ml liberase in a petri dish at 37°C CO₂ incubator for 1 hour. After pipetting up and down several times, the cells were washed and passed through a 70μm cell strainer. After counting the total number of cells, 0.2-1 million cells from each organ were stained with the above antibodies for FCM analysis.

2.4. Histology and Immunofluorescent (IF) staining.

For studies addressing histology of different organs of affected and non-affected mice, tissues were fixed in 4% formaldehyde overnight. After embedding in paraffin, 5 μm-thick sections were cut and stained with hematoxylin and eosin dyes, as described [12]. IF staining was performed as previously described [12], using the following primary antibodies: anti-human CD3 Alexa Fluor 488 (UCHT1, 300454, Biolegend, 1:50 dilution), purified rabbit anti-human CD68 (EPR20545, ab213363, Abcam, 1:50 dilution) and anti-mouse F4/80 Alexa Fluor 488 (F4/80, ab204266, Abcam,1:50 dilution). Donkey anti-rabbit IgG Alexa Fluor 555 antibody (A31572, Thermo Fisher Scientific, 1:400 dilution) was used as the secondary antibody. For IF studies involving Vectra microscopy, the following antibodies were used: CD8 (C8/144B, MA5-13473, Invitrogen), CD19 (SP110, MA5-16414, Invitrogen), CD68 (PG-M1, M0876, DAKO), cytokeratins 8&18 (AE1/AE3, NCL-L-AE1/AE3-601, Leica), CD3 (F7.2.38, ab17143, Abcam), CD4 (4B12, MA5-12259, Invitrogen), and stained with OPAL fluorochromes according to the manufacturer’s method. (https://www.akoyabio.com/phenopticstm/opal-multiplex-immunohistochemistry)

For IF studies involving native vs grafted thymus, cryopreservation was performed using OCT compound (Sakura). As previously described [13], 6 μm-thick sections were cut and fixed with acetone. For staining of grafted human thymus, the following primary antibodies were used: mouse anti-human HLA-DR (purified, mouse IgG2a, L243, 307602, Biolegend, 1:40 dilution), rabbit anti-human cytokeratin 8 (Alexa Fluor 647 conjugated, rabbit IgG, EP1628Y, ab192468, Abcam, 1:50 dilution), mouse anti-human cytokeratin 14 (biotinylated, LL002, MS-115-B0, NeoMarkers, 1:100 dilution). After washing, the following secondary reagents were used: rat anti-mouse IgG2a (Alexa Fluor 488 conjugated, SBB4a, ab172324, Abcam, 1:500 dilution), streptavidin (Alexa Fluor 555 conjugated, S32355, Life technologies, 1:500 dilution).

For staining of native mouse thymus, the following primary antibodies were used: anti-mouse I-A^k (Aβ^k) (cross-reactive to IA^g7) (purified, mouse IgG2a, 10-3.6, 109902, Biolegend, 1:50 dilution), rabbit anti-human cytokeratin 8 (Alexa Fluor 647 conjugated, rabbit IgG, EP1628Y, ab192468, Abcam, 1:50 dilution), Ulex Europaeus Agglutinin I (UEAI) (biotinylated, B-1065, Vector Laboratories, 1:50 dilution). After washing, the following secondary reagents were used subsequently: rat anti-mouse IgG2a (Alexa Fluor 488 conjugated, SBB4a, ab172324, Abcam, 1:500 diluted), streptavidin (Alexa Fluor 555 conjugated, S32355, Life technologies, 1:500 dilution).

2.5. Bead-based immunoassay to quantify serum cytokines.

After euthanizing affected and non-affected mice, serum was isolated from their peripheral blood. Diluted serum (1:4) was used to measure human IL-17A, TNF-α and IFN-γ using a bead-based immunoassay, according to manufacturer’s instructions (https://www.biolegend.com/en-us/legendplex).

2.6. FACS sorting of different cell subsets from grafted and native thymus.

At week 20 after transplantation, HIS mice with grafted vs native thymus were euthanized and single cell suspensions of their thymi were prepared as described above. After counting the total number of cells, they were stained with the following antibodies for FACS sorting: anti-human CD45 (HI30, 562279, BD Biosciences), anti-human CD4 (SK3; 563550, BD Biosciences) and anti-human CD8 (SK1; 557834; BD Biosciences). After gating out the dead cells and doublets, double positive (DP) CD4 CD8 cells, single positive (SP)-CD8 cells and SP-CD4 cells were sorted within a human CD45⁺ gate.

2.7. DNA isolation and high throughput CDR3β TCR sequencing

Genomic DNA was isolated from sorted cell populations using the Qiagen DNeasy Blood and Tissue Kit. DNA was frozen at −20°C and shipped on dry ice to Adaptive Biotechnologies for high-throughput TCRβ CDR3 sequencing. The TCR sequencing data were retrieved from Adaptive’s ImmunoSEQ software. Computational and statistical analysis was performed using R software, as we previously detailed [9].

2.8. Single cell TCR sequencing combined with phenotypic profiling of T cells.

For this experiment, an affected PI HIS mouse with both human and mouse thymi was used. At week 24, when the disease developed, CD3⁺ T cells from skin, liver, spleen, native thymus and pancreatic islets of an affected mouse were sorted using FACS. SP-CD4 and SP-CD8 T cells from grafted human thymus of the same mouse were also sorted. Cells were sorted directly into RT-PCR buffer (Qiagen) using an Influx cell sorter (Becton Dickinson). Index sorting was performed to collect marker information for each single cell. The sequences of expressed TCRα and TCRβ chains from single cells were obtained by a series of three nested PCR reactions with multiple internally nested TCRVα, Vβ, Cα, and Cβ primers as previously described [14]. Reaction products were barcoded and sequencing was performed on a MiSeq System (Illumina). In addition to TCR sequencing, gene expression level for multiple cytokines and transcription factors that influence T-cell function and define certain T-cell subsets were obtained simultaneously [14].

2.9. Generation of HIS mice with Melanoma Antigen Recognized by T cells 1 (MART1) TCR-expressing and MART1 peptide-expressing HSCs.

Non-thymectomized NSG mice were conditioned with sublethal (1.5 Gy) total body irradiation and injected intravenously with HLA-A2⁺ human CD34⁺ HSCs (2 × 10⁵/mouse). Mice were then implanted with a cryopreserved / thawed fetal thymus fragment (~1 mm³) from the same fetal donor (HLA-A2⁺) under the recipient kidney capsule. One group of mice (“Peptide group”) was injected with a mixture of 1 × 10⁵ MART1 TCR-expressing HSCs and 1 × 10⁵ MART1 peptide-expressing HSCs, while another group of mice (“Control group”) was injected with a mixture of 1 × 10⁵ MART1 TCR-expressing HSCs and 1 × 10⁵ mock-transduced HSCs, as previously described [15]. Half of the mice in the Peptide and Control groups were HLA-A2 Tg NSGs and the other half were non-Tg NSGs. At the endpoint (24–29 weeks after reconstitution), grafted and native thymi were harvested and analyzed for the presence of MART1 TCR-expressing T cells identified using the HLA-A*0201/MART-1 (ELAGIGILTV) tetramer (MBL). Co-staining with the following antibodies was performed: anti-human CD45 (HI30, Biolegend), human CD3 (SK7, Biolegend), human CD4 (OKT4, Biolegend), human CD8 (HIT8a, Biolegend), human CD11c (B-ly6, BD Biosciences) and human PD-1 (EH12.1, BD Biosciences).

3. Results.

3.1. Development of autoimmune-like syndrome is T cell-dependent and its onset is associated with the type of thymus in which human T cells develop.

In order to evaluate the role of the thymus in the development of the disease, we generated 4 different groups of humanized mice. All groups received human FL CD34⁺ HSCs. In Group 1, the native mouse thymus was removed and no human thymus was transplanted, so no T cells were formed [10] (Figure S1A). In Group 2, human T cells developed in mouse thymi [10] (Figure S1B). In Group 3, a human fetal thymus, autologous to the HSCs, was transplanted under the kidney capsule of mice in which the native mouse thymus had been removed, so all T cells developed in the human thymus [10] (Figure S1C). In Group 4, which had both a native mouse thymus and a grafted human thymus, human T cells developed in both mouse and human thymi (Figure S1D). All mice were followed for the presence of human immune cells in peripheral blood every 3-4 weeks. They were examined weekly for signs of disease, including weight loss, hair loss and changes in posture and activity. As soon as any of these signs were evident, the mice were considered “affected”. In most cases, hair loss, especially in the face and abdomen, was the earliest visible sign. As shown in Figure 1A, humanized mice with no thymus, which had no T cells, did not develop the autoimmune-like syndrome, indicating that T cells are essential for disease development. All mice with human T cells eventually developed the disease. However, mice with only a human thymus had a delayed onset of the disease compared to those with a mouse thymus or both thymi. Mice with both thymi had a faster disease onset compared to the mice with only a human thymus. Within the group of mice with an intact native thymus, disease onset was similar in mice receiving different fetal tissues with different HLAs (Figure S1F) and in male versus female recipient mice (Figure S1G). The disease also occurred in “personalized immune” (PI) HIS mice generated with human adult BM-derived CD34⁺ HSCs and HLA-matched human fetal thymus (with no mouse thymus) (Figure S1H).

Figure 1. Autoimmune disease onset, histology of different tissues and serum cytokines.

A: Percentage of mice free of autoimmune disease over time post-transplantation in HIS mice generated with human FL CD34⁺ HSCs and different thymi (autologous human fetal thymus vs native mouse thymus vs both thymi vs no thymus). Log-rank (Mantel-Cox) test was used for statistical analysis between different groups. B: Representative H&E staining of liver, lung, salivary gland, skin and pancreas of 8 affected and 3 non-affected HIS NSG mice. C: IF staining of liver, lung, skin and spleen from a representative affected HIS mouse. D: measurement of human IL-17A, TNF-α and IFN-γ in the serum of affected and non-affected HIS mice. Each dot represents one mouse. One-way ANOVA was used for comparisons between different groups. Mean±SEM is shown for each group. Statistically significant differences are shown with * (*0.01<p-value<0.05, **0.001<p-value<0.01, ***p-value<0.001).

3.2. Histologic and phenotypic description of the disease.

One of the first disease manifestations is hair loss, which usually starts from the face and spreads to the abdomen, flank and back. In parallel, the affected mice lose weight and become lethargic. Diarrhea is observed in a subset of mice. H&E staining of sections from different organs of affected and non-affected mice showed involvement of different organs, including liver, small and large intestine, spleen, lung, salivary gland, pancreas and skin (Figure 1B and S2A). The detailed histological findings are listed in Table S1. Lymphocytic and neutrophilic infiltration, epithelial apoptosis, hair follicle atrophy, presence of multinucleated giant cells and extramedullary hematopoiesis (in spleen and lung) are among the major histological findings. There were no differences between the affected mice with mouse or human thymus regarding the histological findings in different tissues (Table S1). As shown in Figure 1C, T cells and macrophages comprised the majority of human immune cells infiltrating the target organs. To confirm that the multinucleated giant cells are of human origin, we performed co-staining of human CD68 and mouse F4/80. Although mouse macrophages (mF4/80⁺ cells) were present in different organs of both affected and non-affected mice, the multinucleated giant cells were of human origin and more prevalent in the affected mice (Figure S2B). We found significantly higher levels of IFN-γ in the serum of affected mice compared to non-affected mice (ie animals sacrificed at 18-22 weeks post-transplantation), while the levels of TNF-α and IL-17A were not significantly different (Figure 1D).

The pooled LN cell count was higher in affected mice compared to non-affected animals (Figure 2A). In affected mice, skin, as one of the target organs, was highly infiltrated with human CD45⁺ immune cells that were mainly T cells (Figure 2B), while non-affected mice had very low numbers of immune cells in skin (not shown). The proportion of human CD45⁺ immune cells was comparable in spleen, while higher in LNs and lower in BM of affected mice (Figure 2B). There was a significant increase in affected mice in the proportion of T cells within human CD45⁺ cells in spleen, BM and LNs (Figure 2B) and also within total spleen cells (Figure S3A). The proportions of B cells within human CD45⁺ cells (Figure 2C) and also within the total cells (Figure S3B) were significantly decreased in spleen and BM of affected mice. CD11c⁺ cells were significantly increased within human CD45⁺ cells (Figure 2C) and also within the total cells (Figure S3C) in spleen and BM. We did not observe a significant difference in proportions of CD14⁺ monocytes within human CD45⁺ cells between affected and non-affected mice (Figure 2C). Except for a decreased proportion of CD4 T cells in LN of affected mice, the CD4 and CD8 proportions within T cells were similar across different tissues of affected and non-affected mice (Figure 2D and 2E). Affected mice had increased memory phenotype (mainly effector memory (CD45RA⁻ CCR7⁻)) and decreased naïve phenotype (CD45RA⁺ CCR7⁺) CD4 and CD8 T cells, which was a consistent finding across different lymphoid tissues (Figure 2D and 2E).

Figure 2. Immune cells in different organs of affected and non-affected mice.

A: Total cell count of spleen, bone marrow, pooled lymph nodes and skin in affected and non-affected HIS mice. A 1cm² piece of skin was removed from the trunk area and digested with liberase. B: Proportion of human CD45⁺ cells among total cells and the proportion of T cells within human CD45⁺ cells in the four different organs of affected and non-affected HIS mice. C: proportion of human CD19⁺ B cells, CD11c⁺ cells and CD14⁺ monocytes within human CD45⁺ cells in the four different organs of affected and non-affected HIS mice. D and E: Proportion of human CD4⁺ and CD8⁺ cells among total CD3⁺ T cells and naïve and effector memory (EM) CD4 and CD8 T cells in the four different organs of affected and non-affected HIS mice. F: Proportions of Tregs (FOXP3⁺, CD25⁺) among peripheral (in spleen, bone marrow and skin) CD4 cells for both affected and non-affected mice. Proportions of CTLA-4⁺ and CD45RO⁺ cells among peripheral Tregs of affected and non-affected mice are also shown. Each dot represents one mouse. Mean±SEM is shown for each group. Unpaired 2-tailed t-tests with nonparametric Mann Whitney tests were used for comparisons between the two groups. Statistically significant differences are shown with * (* 0.01<p-value<0.05, **0.001<p-value<0.01, ***p-value<0.001).

There was a higher proportion of activated T cells (HLA-DR⁺) in affected mice compared to non-affected mice (Figure S3E). CD4 and CD8 cells infiltrating the skin of affected mice expressed very high levels of CTLA-4 (Figure S3F). High proportions of CTLA-4⁺ and HLA-DR⁺ cells in skin CD4 and CD8 T cells suggest that these were activated T cells. However, the high proportions of CTLA-4⁺ cells might also be a sign of exhaustion.

Similar levels of peripheral Tregs were detected in different tissues of affected and non-affected mice (Figure 2F).The proportion of Tregs expressing CTLA-4 was higher in the spleens of affected compared to non-affected mice. Additionally, spleen and bone marrow Tregs had a more activated phenotype (CD45RO⁺) in affected compared to non-affected mice (Figure 2F).

3.3. Clonal expansion and sharing in multiple affected tissues.

To determine whether the same or distinct groups of T cells attack different organs in affected mice, we performed single cell paired αβ sequencing of TCRs combined with phenotypic profiling of T cells infiltrating different tissues of an affected mouse, as previously described [14]. With less than 100 T cells sequenced from each organ, we found substantial expansion of certain clones of T cells that were widely shared between different affected organs, including pancreas, skin, liver, spleen and native thymus (Figure 3A). In contrast to the native thymus, SP-CD4 and SP-CD8 T cells sorted from the grafted human thymus of the same mouse were highly diverse and, with the exception of a few clones, devoid of expanded clones found in target tissues (Figure 3A). Some highly expanded T cell clones were found with both CD4 and CD8 phenotypes and also with Th1 (TBET⁺) and Th2 (GATA3⁺) phenotypes in different tissues. The amino acid sequence of one of these expanded clones is shown in Figure 3B. Simultaneous transcriptomic analysis of certain cytokines and transcription factors (TFs) indicated high proportions of IFN-γ⁺, TBET⁺ and GATA3⁺ T cells in affected organs (Figure 3C). Expression levels of interleukins 2, 5, 10, 13, 17 and 21 were very low and undetectable in most cells (Figure 3C). Circos plot connecting the cytokines and TFs that T cells express indicated that there is overlap between GATA3⁺ (Th2), RORC⁺ (Th17) and TBET⁺ (Th1) T cells (Figure 3D). IFNG⁺ T cells were mainly TBET⁺, and to a lesser degree GATA3⁺ and RORC⁺ (Figure 3D). Some Foxp3⁺ T cells expressed RORC, GATA3 or TBET (Figure 3D). Circos plot connecting the tissues of origin and the TFs that T cells express indicated that GATA3⁺ T cells constituted the highest proportion of T cells infiltrating the liver, spleen and pancreatic islets, followed by TBET⁺ and RORC⁺ cells (Figure 3E). TBET⁺ T cells were more prevalent in skin and native thymus than the other T cell phenotypes. There were a very small proportion of infiltrating T cells expressing Foxp3 and BCL6 in different organs (Figure 3E). In all the affected organs, the majority of infiltrating T cells were of CD4 phenotype (Figure 3F).

Figure 3. Single cell TCR sequencing combined with phenotypic profiling of T cells.

CD3⁺ T cells from skin, liver, spleen, native thymus and isolated pancreatic islets (contaminated with some exocrine tissues) of an affected HIS mouse (generated with adult HSCs and both human and mouse thymi) were single cell sorted using FACS. SP-CD4 and SP-CD8 T cells from grafted human thymus of the same mouse were also sorted. Single cell sequencing of TCRs combined with phenotypic profiling of T cells was performed. A: Each pie chart shows the results of TCR α and β sequencing for each organ. SP-CD4 and SP-CD8 cells from the grafted thymus are shown in separate charts. Each wedge represents one T cell clone with a unique TCR α and β. Colored wedges represent T cell clones that are either shared (found in more than one organ) or expanded (more than 1 T cell in an organ). Each shared or expanded T cell is shown in a unique color, specific for that T cell across different organs. The size of each wedge represents the frequency of that T cell in each organ. B: Characteristics of a highly expanded T cell clone (CDR3β:CASSLGRANTGELFF, CDR3α: CAASLNSMRF) that was found with different transcriptional patterns and CD4/CD8 phenotypes in different tissues. C: Simultaneous transcriptomic analysis of certain cytokines and transcription factors for the same T cells in the mentioned organs. The frequency of T cells that express a certain cytokine or TF is shown in a heat map graph. D: Circos plot showing the number and frequency of T cells that are that are positive for the indicated cytokines and transcription factors. Connecting lines show the level of overlap between these cytokines/ transcription factors. E and F: Circos plots showing the number and frequency of T cells in each tissue that are positive for the indicated transcription factors and CD4/CD8 phenotypes, respectively.

3.4. Role of pre-existing T cells in the thymus graft.

One possible mechanism for development of disease in humanized mice with grafted human thymus tissue could be xenoreactivity of pre-existing grafted thymus-derived T cells against mouse antigens. As pre-existing cells developed in the donor human fetal thymus in the absence of intrathymic mouse antigens, they should not be tolerant of mouse antigens. In order to investigate the role of pre-existing grafted thymus-derived T cells, we designed three separate experiments. In Experiment 1, we generated a group of humanized mice with HLA-A3⁺/HLA-B12⁺ FL HSCs and an allogeneic human fetal thymus that was HLA-A3⁻/HLA-B12⁻. We used our standard procedure to eliminate thymic T cells by cryopreservation/thawing and pipetting up and down several times to release as many thymocytes as possible before transplantation. To further deplete the thymocytes that might develop from progenitors in the graft or pre-exist and then exit to the periphery, an anti-human CD2 antibody was injected to the mice in 2 weekly doses, as described [11]. While up to 50% of peripheral T cells were initially thymus-derived in some but not all animals grafted with these procedures, the level of grafted thymus-derived CD4 and CD8 T cells declined over time and by week 24 was close to zero (Figure 4A). Conversely, the level of HSC-derived T cells reached around 100% of total CD4 and CD8 T cells by around week 24 (Figure 4A). We euthanized these mice at week 24 and looked for grafted thymus-derived T cells in different organs. Similar to peripheral blood, almost all CD4 and CD8 T cells in BM, liver, LNs and spleen were of HSC origin (Figure 4A).

Figure 4. Role of thymus graft-derived T cells in development of the disease.

Three separate experiments were designed to address the role of thymus graft-derived T cells in development of the disease in HIS mice. A: In Experiment 1, we investigated the dynamics of persistence of thymus graft-derived T cells in peripheral blood of recipient HIS mice. A group of thymectomized humanized mice was generated with HLA-A3⁺/HLA-B12⁺ FL HSCs and an allogeneic human fetal thymus that was HLA-A3⁻HLA-B12⁻. Measures to deplete thymus graft-derived T cells, including cryopreservation and pipetting of fetal thymus and injection of 3 weekly doses of anti-CD2 antibody, were taken. Using antibodies against these two HLAs, the level of thymus graft-derived and HSC-derived CD4 and CD8 T cells in peripheral blood was recorded at different time points until 24 weeks post-transplantation. At week 24, these mice were euthanized and the level of these T cells in BM, liver, LN and spleen was measured. B: In Experiment 2, in which the grafted human thymi were similarly treated to deplete the pre-existing thymocytes, three HIS mice were generated with HLA-A2⁻ FL HSCs and an allogeneic human fetal thymus that was HLA-A2⁺. At 25 weeks post transplantation, two of these mice had developed the disease and the other one was still healthy. We euthanized the mice and used an HLA-A2 antibody to track the grafted thymus-derived T cells. C: In Experiment 3, we generated a group of HIS mice with a fresh fetal human thymus without the usual measures to deplete thymus-derived T cells. Autologous FL HSCs were injected to half of the mice, while the other half received allogeneic FL HSCs. The level of human CD45⁺ cells and T cells in total white blood cells (WBCs) and their absolute counts, the proportion of total, naïve and effector memory CD4 and CD8 T cells and the level of HLA-A2⁺ (thymus origin) T cells, B cells and monocytes were recorded at the indicated time points post transplantation. Each dot represents one mouse. Mean±SEM is shown for each group. D: Kaplan Meier plot shows the proportion of disease-free mice in both groups of HIS mice (with autologous vs allogeneic HSCs) in Experiment 3. Log-rank (Mantel-Cox) test was used for statistical analysis. E: Correlation between human APC level (combined B cell and monocyte count) at weeks 12 and 16 post-transplantation and week of disease onset are shown. Pearson correlation coefficients were analyzed and the P-value and correlation coefficient (r) for each time point is indicated.

In Experiment 2, in which the grafted human thymus and recipients were similarly treated to deplete the pre-existing thymocytes, three humanized mice were generated with HLA-A2⁻ FL HSCs and an allogeneic human fetal thymus that was HLA-A2⁺. At 25 weeks post transplantation, 2 of these mice had developed the disease and the other one appeared healthy. We euthanized the mice and used an HLA-A2 antibody to track the grafted thymus-derived T cells. As shown in Figure 4B, one affected mouse had 10-16% thymus-derived T cells, while the non-affected mouse had 3-5% HLA-A2⁺ T cells in blood, spleen and skin and affected mouse #2 had no thymus-derived cells. The majority of T cells in both affected mice were of HSC origin. Although it is still possible that some mouse-reactive T cells in these mice originated from the pre-existing T cells in grafted thymus, most T cells were of HSC origin.

To further investigate the role of pre-existing thymus-derived T cells, in Experiment 3 we generated HIS mice without the usual measures that we take to deplete thymus-derived T cells. Fresh fetal human thymi were transplanted without cryopreservation or pipetting measures and the anti-CD2 antibody was not injected after transplantation. Autologous FL HSCs were injected to half of the mice, while the other half received allogeneic FL HSCs. Unlike the HIS mice that are generated with our regular treatment, which began to show human T cells in their peripheral blood at around 10 weeks post-transplantation, the mice with freshly-transplanted grafted thymi showed a low but visible human T cell population in their peripheral blood at the earliest time point examined (week 4 post transplantation) (Figure 4C). In the group of mice that received allo-HSCs, the majority of peripheral T cells were thymus graft-derived (HLA-A2⁺) until around week 16-20, after which HSC-derived T cells (HLA-A2⁻) developed. B cells and monocytes were mainly of HSC origin, although some thymus-derived B cells and monocytes were detectable in the periphery up to 12 weeks post-transplantation. After 12 weeks, when HSC-derived T cells started to develop, the total level of T cells, B cells and monocytes declined and human chimerism disappeared only in the group with allogeneic thymus and HSCs, likely reflecting HSC destruction by allogeneic human T cells from the thymus graft. At around week 16 post-transplantation, mice started developing the disease and by around 30 weeks post-transplantation, the majority of mice in both auto and allo groups had developed the disease (Figure 4D). The mice that had higher levels of human antigen-presenting cell (APC) (B cell and monocyte) reconstitution showed an earlier naïve to memory T cell conversion in the periphery and also developed the disease more rapidly (Figure 4E). The appearance of disease at 16-20 weeks in these animals, when most T cells were of thymic origin, indicates that thymocytes in untreated thymic grafts are capable of causing disease.

3.5. Thymopoiesis, thymic structure and Treg development in grafted and native thymi.

In view of the more rapid development of disease in mice with a native mouse thymus compared to those with a grafted human thymus, we compared the tissue structure and the cell subsets in each type of thymus. At different times post-transplantation (18-22 weeks for non-affected mice and 26-60 weeks for affected mice), HIS mice were euthanized. Thymocyte counts were similar between grafted thymi of affected and non-affected mice (Figure 5A). Thymocyte counts were significantly lower in native thymi than in grafted human thymi. Furthermore, the numbers were significantly lower in native thymi of affected mice compared to non-affected mice (Figure 5A).

Figure 5. Thymopoiesis and Treg development in grafted and native thymi.

At different times post-transplantation (18-22 weeks for non-affected mice and 26-60 weeks for affected mice), HIS mice were euthanized. Grafted thymi (for HIS mice with grafted human thymus) and native mouse thymus (for HIS mice with native mouse thymus) were harvested and thymocytes were counted and analyzed by FCM. A: Thymocyte counts for grafted and native thymi of affected vs non-affected mice. Each dot represents one mouse. B and C: Proportions of DP, SP-CD4 and SP-CD8 cells among human CD45⁺ cells and the proportions of CD45RO⁺ and Ki67⁺ cells within these three cell subsets are shown for grafted and native thymi of affected vs non-affected mice. D: Proportions of Tregs (FOXP3⁺, CD25⁺) among thymic (grafted vs native) CD4 cells for both affected and non-affected mice. E: Proportions of CTLA-4⁺, CD45RO⁺ and Ki67⁺ cells among thymic Tregs of affected and non-affected mice. The animals with human thymus were thymectomized and did not have a native thymus. Mean±SEM is shown for each group. Unpaired 2-tailed t-tests with nonparametric Mann Whitney tests and one-way ANOVA were used for comparisons between two and more than two groups, respectively. Statistically significant differences are shown with * (* 0.01<p-value<0.05, **0.001<p-value<0.01, ***p-value<0.001).

In grafted thymi of both affected and non-affected mice, the majority of thymocytes were DP and both groups had similar proportions of DP and SP thymocytes, showing a normal pattern of thymopoiesis (Figure 5B). Also, the proportions of DP and SP thymocytes that were positive for CD45RO and Ki67 were not different between affected and non-affected mice (Figure 5B). The majority of human thymocytes in grafted human thymi were CD45RO⁺, which is similar to the proportions reported for thymocytes in human thymus [16]. In both groups, DP cells were the most proliferative (Ki67⁺) thymocyte subsets, followed by SP-CD8 cells and SP-CD4 (Figure 5B). While non-affected mice had normal proportions of DP and SP thymocytes in their native thymus, the proportion of DP thymocytes was significantly lower in affected mice, pointing to impaired thymopoiesis (Figure 5C). These findings suggest that the native thymus may be a target of autoimmune attack and are consistent with the finding of the same expanded T cell clones in native thymus and target tissues in the affected mouse (Figure 3A).

Treg levels in both grafted and native thymi were similar between affected and non-affected mice. However, Treg proportions were lower in native compared to grafted thymi in non-affected mice (Figure 5D). Similar proportions of Tregs in native and grafted thymi in affected and non-affected mice were positive for CTLA-4, CD45RO and Ki67 (Figure 5E).

While grafted thymi had a well-organized structure with defined cortical and medullary areas and detection of CK14⁺ medullary thymic epithelial cells (mTECs) and CK8⁺ cortical TECs (cTECs) in both affected and non-affected mice, native thymi did not have a well-organized structure and lacked mTECs (UEAI⁺) in both non-affected (Figure 6A) and affected mice (not shown). In contrast, the cortical and medullary areas are well-defined in NOD mice (Figure 6A). HLA-DR⁺ CK⁻ cells in both native mouse and grafted human thymi represent the human BM-derived APCs that have migrated to these thymi.

Figure 6. Histology and high-throughput TCRβ sequencing of thymocytes developed in grafted human vs NSG mouse thymus.

A: H&E and IF staining of representative grafted (representative of 10 H&E and 5 IF stainings) and native thymi (representative of 7 H&E and IF stainings) from non-affected HIS mice euthanized at 22 weeks post-transplantation. Grafted human thymus was stained for CK14, CK8 and HLA-DR, while the native mouse thymus was stained for UEAI, CK8 and HLA-DR in IF images. Additionally, an IF image of a NOD mouse thymus stained for mouse MHC II I-A^g7, CK8 and UEAI is shown. B: Results of high-throughput TCRβ sequencing of DP, SP-CD4 and SP-CD8 thymocytes in native thymi compared to grafted thymi. Clonality and [abundance plot slope], as measures of TCR diversity, for the three cell subsets in grafted and native thymi are shown.

3.6. High-throughput TCRβ sequencing reveals abnormalities in T cell repertoire of native thymi.

To determine how the unorganized structure of the native thymus might affect the human TCR repertoire, we performed high-throughput TCRβ sequencing of DP, SP-CD4 and SP-CD8 thymocytes in native thymi compared to grafted human thymi. Number of sorted cells, total and productive templates and unique CDR3βs are listed in Table S2. We used clonality and abundance plot slope as measures to quantify diversity of the TCR repertoire [17]. Both measures showed that the TCR repertoire was less diverse in native thymi compared to grafted thymi, especially in DP cells (Figure 6B). An increased diversity of TCR repertoire as thymic selection progresses from DP to SP subsets was noted in native thymi, which was contrary to what normally occurs in grafted human thymi [9] (Figure 6B). Similar levels of CDR3β sharing were observed comparing native to native thymi, grafted to grafted thymi and native to grafted thymi (Figure S4). Defective thymic structure could explain abnormalities in TCR repertoire selection in native mouse thymus.

3.7. Impaired negative selection in native thymus.

In order to directly measure the effect of thymus structure on selection of autoreactive T cells, we generated humanized mice with human HLA-A2⁺ fetal HSCs transduced with a MART1-reactive TCR restricted by HLA-A2. Half of the mice also received autologous HLA-A2⁺ HSCs transduced with MART1 peptide so that APCs would express the target antigen. The recipient mice were not thymectomized and were transplanted with human fetal thymus (HLA-A2⁺) autologous to HSCs, and therefore had both a native mouse thymus and a grafted human thymus. Similar levels of human HSC-derived CD11c⁺ cells were present in both native and grafted thymi of recipient mice, regardless of whether or not some HSCs were transduced with MART1 peptide (Figure 7A). Using a GFP reporter indicating the expression of MART1 peptide, we confirmed that MART1⁺ CD11c⁺ cells were detected in both native and grafted thymi of “peptide⁺” groups with similar frequencies (Figure 7B). While MART1 DP and SP thymocytes were completely deleted in grafted human thymi in the presence of MART1⁺ DCs, negative selection of these T cells did not occur in native mouse thymi (Figure 7C). Half of the mice in each group were HLA-A2 Tg and there was no difference between the HLA-A2 Tg and non-Tg NSG mice with regard to negative selection of MART1-specific T cells in native mouse thymus (data not shown). Interestingly, the small number of tetramer-positive human thymocytes in grafted human thymi and the larger number in native thymi from mice with peptide⁺ APCs upregulated PD-1 (Figure 7D), suggesting that MART1-specific T cells had encountered MART1-presenting APCs in both groups, but were only deleted in grafted human thymi (Figure 7C). In the absence of MART1 peptide in grafted thymi, the proportion of MART1-specific T cells increased in SP-CD8 T cells as selection progressed from DP cells, which points to positive selection of these T cells in human grafted thymi (Figure 7C). This pattern was not seen in native thymi (Figure 7C), suggesting impairment of positive selection in native thymi. IF staining showed that both human and mouse HSC-derived APCs were present in mouse thymi (Figure 7E). Overall, these data suggest that failure of normal negative selection of autoreactive T cells in native mouse thymus due to abnormal thymic architecture likely contributes to the faster development of autoimmunity in HIS mice with a native mouse thymus compared to those with a grafted human thymus.

Figure 7. Negative selection of T cells expressing MART1-reactive TCR in presence and absence of MART1 peptide in grafted human vs NSG mouse thymus.

We generated HIS mice with human fetal HSCs transduced with MART1-reactive TCR (n=8). Half of the mice also received HSCs transduced with MART1 peptide. The recipient mice were not thymectomized and were transplanted with human fetal thymus autologous to HSCs. A: The proportion of CD11c⁺ cells in native and grafted thymi in groups with and without MART1 peptide. B: The proportion of CD11c⁺ cells that express GFP as an indicator of MART1 peptide expression, in native and grafted thymi in groups with and without MART1 peptide (n=4 per group). C: Proportion of MART1-TCR tetramer⁺ T cells among DP, SP-CD4 and SP-CD8 thymocytes in native and grafted thymi in groups with and without MART1 peptide. Representative histograms plotting tetramer⁺ cells among DP, SP-CD4 and SP-CD8 thymocytes are shown on the left. D: Proportion of PD-1⁺ cells among MART1-TCR tetramer⁺ T cells in DP, SP-CD4 and SP-CD8 thymocytes in native and grafted thymi in groups with and without MART1 peptide. E: IF staining of a native mouse thymus stained with HLA-DR and mouse CD11c, markers of human and mouse HSC-derived APCs, respectively.

3.8. Early-life injection of HSCs delays disease onset.

One possible reason for disorganized organ structure and defective negative selection in native mouse thymus could be delayed interaction between TECs and T cell progenitors in native mouse thymus, as such interactions are needed for normal TEC development [18–20]. We hypothesized that injection of HSCs immediately after birth, instead of in adulthood, might rescue the structure of native mouse thymus and prevent or delay the autoimmune disease. To address this possibility, we generated two groups of humanized mice by injecting the same HSCs into neonates 1-2 days after birth (intra-liver) and also into adult mice (i.v./ 7-10 weeks after birth). Both groups of mice relied on their native thymus to generate T cells. T cells appeared faster in the peripheral blood of HIS mice in the “neonate” group (Figure S5A). However, the mice in the “adult” group reached the same level of T cell reconstitution at around week 15 post-transplantation (Figure S5A). Intraliver injection of HSCs right after birth significantly delayed the disease, but did not prevent it from developing (Figure 8A). Although the thymocyte count in neonatally-injected mice did not change significantly (Figure 8B), IF and H&E staining suggested a slight improvement in the structure of native thymi in neonatally-injected mice (Figure 8C and 8D). However, mouse mTECs were still rarely seen in native thymus of neonatally-injected mice (Figure 8D).

Figure 8. Disease and structure of native thymus in NSG mice receiving HSCs at neonatal vs adult stage.

Two groups of NSG HIS mice received the same HSCs as the neonates (1-2 days after birth (intra-liver)) or as adults (i.v./ 7–10 weeks after birth). A: Kaplan-Meier plot showing the proportion of disease-free mice receiving HSCs as neonates vs adults. Log-rank (Mantel-Cox) test was used for statistical analysis. B: Native thymus cell count in “neonate” and “adult” groups of HIS mice euthanized about 40 weeks post-transplantation. C: H&E staining showing the structure of native mouse thymi in both “neonate” and “adult” groups of HIS mice euthanized at 11 and 45 weeks post-transplantation, in addition to a representative non-humanized NSG mouse thymus. D: IF staining of native mouse thymi in both “neonate” and “adult” groups of HIS mice euthanized at 11 and 45 weeks post-transplantation, stained with mAbs specific for mouse MHC II I-A^g7, CK8 and UEAI (representative of 4 and 7 stainings in neonate and adult groups, respectively).

3.9. Direct recognition of murine antigens is not required to induce disease.

It has recently been reported that the absence of murine MHC in NSG mice prevents the development of GVHD following transfer of mature human peripheral T cells [21]. To determine whether or not direct recognition of murine antigens was similarly required for disease development in HIS mice with de novo T cell development in a human thymus graft, we generated HIS mice with human fetal thymus and autologous HSCs in recipient NSG mice expressing or not expressing mouse MHC I and II (NSG MHC KO). While a slight, but not statistically significant delay in disease development was observed in NSG mice lacking murine MHC, all mice eventually developed the disease (Figure 9A). The kinetics of development of human CD45⁺ cells, T cells and B cells were very similar between the two groups of mice. However, naïve to memory conversion of CD8+ T cells was slower in MHC KO compared to wild-type NSG mice (Figure 9C). The development of disease in mice lacking murine MHC molecules is consistent with a failure to tolerize human T cells recognizing mouse antigens indirectly (ie presented by HLA molecules). In view of the abundance of murine APCs in human thymus grafts, this failure of tolerance is most likely specific for mouse tissue-restricted antigens, which would not be present in human thymic tissue.

Figure 9. Disease development in NSG vs NSG MHC KO HIS mice.

HIS mice were generated by transplantation of human fetal thymus and HSCs into NSG mice or NSG mice lacking mouse MHC class I and II (NSG MHC KO). A: Kaplan-Meier plot showing the proportion of disease-free mice in NSG vs NSG MHC KO HIS mice. Log-rank (Mantel-Cox) test was used for statistical analysis. B: Kinetics of development of human immune cells. C. Kinetics of peripheral appearance of T cells and B cells, in addition to the proportions of naïve (CD45RA⁺) and memory (CD45RO⁺) CD4 and CD8 T cells are shown over time after transplantation. Mean±SEM is shown for each group. Two Way ANOVA with multiple comparisons was used for statistical analysis. Statistically significant differences are shown with * (* 0.01<p-value<0.05).

4. Discussion.

Here we have demonstrated a role for the thymus in development of a multi-organ autoimmune-like disease that spontaneously occurs in HIS mice. By removing the native mouse thymus, we demonstrated that T cells are essential for the disease to occur. Activated CD4 T cells were the predominant T cell subset infiltrating the target tissues, though CD8 cells were also present. Disease development in mice lacking murine MHC molecules indicates that direct recognition of murine antigens is not required for disease development. Thus, indirect presentation of antigens on human APCs is capable of causing end organ target tissue damage through indirect effector mechanisms, likely involving cytokines. This observation has important implications for human GVHD induction and suggests a parallel to murine GVHD, in which indirect effector mechanisms have been implicated [22].

Single cell TCR sequencing revealed that the same clones of T cells are expanded and detected in different target tissues, including the native mouse thymus, with high frequencies. In contrast to the mouse thymus, the human thymus did not contain these putative autoreactive T cell expansions, preserved a normal architecture, and generated diverse T cell repertoires. The appearance of a few of these clones in grafted thymus might be due to recirculation of mature T cells back to the thymus [23]. Remarkably, in some instances, the same T cell clone present in different tissues expressed both a CD4 and/or CD8 phenotype, raising the possibility that autoreactive TCRs can be cross-reactive and recognize both class I and II-restricted antigens or may recognize antigen in a coreceptor-independent fashion. The same clones also showed different functional phenotypes in different tissues, or sometimes even within the same tissue.

Not only the target organs, but also the lymphoid organs of affected mice had a very high percentage of memory T cells (mainly effector memory). Although T cells constitute the majority of immune cells infiltrating the target tissues, multinucleated human macrophages were widely detected in affected tissues, suggesting that interactions between human T cells and macrophages may drive injury to mouse tissues. We found a significant increase in the level of human IFN-γ in the serum of affected mice, likely produced by activated T cells, as indicated by single cell transcriptomic analysis of T cells in target tissues. IFN-γ⁺ T cells were mainly detected among TBET⁺ (Th1) and GATA3⁺ (Th2) cells and to a lesser degree among RORC⁺ (Th17) CD4 T cells. A Th1 response is likely to activate macrophages. A similar disease with prominent infiltration of T cells and multinucleated macrophages in multiple organs has been described in humanized NSG mice carrying human SCF/GM-CSF/IL-3 transgenes (SGM3) [24] and was equated to human hemophagocytic lymphohistiocytosis (HLH), also known as hemophagocytic syndrome (HPS) or macrophage activation syndrome (MAS). MAS is an important complication after HSC transplantation, viral infections or autoimmune diseases and is triggered by hypercytokinemia, mainly IFN-γ [25]. In humanized NSG-SGM3 mice, the disease was associated with elevated myeloid cytokines and mutual T cell and macrophage activation [24].

One possible mechanism for development of this disease could be xenoreactivity of pre-existing T cells in grafted human thymi against mouse antigens. As these cells are not exposed to mouse antigens during development, they could attack recipient mouse cells. While our data show that the activated/memory T cells that infiltrate different target tissues are almost all derived from HSCs, this does not exclude an initial role for pre-existing thymus graft-derived T cells, which led to early peripheral T cell reconstitution. Disease development was rapid in mice that received fresh grafted thymi without measures to deplete the thymus graft-derived T cells, suggesting that it is important to deplete the grafted thymus of T cells before transplantation to delay or prevent disease. To further deplete passenger thymocytes that might migrate to the periphery, an anti-human CD2 antibody is injected to the mice in 2 weekly doses [11]. These measures may explain why GVHD/autoimmunity in our mice occurs much later (around week 30-60) compared to that reported following transplantation of fresh thymus tissues (disease onset around week 10-30) [7]. When we generated HIS mice with fresh thymus and did not apply our measures to deplete thymus-derived T cells, the majority of peripheral T cells were of thymus origin.

We have previously reported that, despite the higher thymic cell counts in grafted human thymus compared to NSG native mouse thymus, distribution of DP and SP T cells is similar between the two types of thymi [10]. Here we showed that, while this distribution is also similar between human thymocytes in grafted human thymi of affected and non-affected mice, the frequency of DP cells and total thymocyte counts were significantly decreased in native mouse thymi of affected mice. Mouse thymus may be a target of autoreactive T cells in HIS mice, while the grafted human thymus is spared, as we did not find any evidence that the disease attacks human thymus grafts or human hematopoietic cells. Therefore, the disease is most likely directed against mouse antigens and not human antigens. The similar pattern of disease development in NSG MHC KO mice compared to NSG mice suggests that presentation of mouse antigens on human HLAs is the major instigator of disease development. Effector T cells recognizing mouse antigens on human HLAs might fail to be deleted and Tregs with such specificities would not be expected to be positively selected in human thymic grafts, thus explaining disease development in these mice.

We showed that HIS mice with a native mouse thymus have a significantly faster onset of disease compared to HIS mice with a grafted human thymus. There are apparent structural defects in native NSG mouse thymus that affect TCR repertoire and negative selection of autoreactive T cells. Native NSG mouse thymus fails to develop cortical and medullary areas after human HSC transplantation and mTECs are barely detectable. In contrast, grafted human thymus has a normal structure and contains human mTECs and cTECs in distinct medullary and cortical areas [9]. Grafted human thymus grows much larger than a native NSG mouse thymus and contains orders of magnitude higher numbers of thymocytes. High-throughput TCRβ sequencing revealed that DP thymocytes in native mouse thymus have a markedly lower diversity than those in grafted human thymus, possibly pointing to a general failure of positive selection in the native mouse thymus. The inability of native mouse thymus to delete MART1-TCR⁺ T cells in the presence of MART1 peptide-presenting human APCs points to a failure of negative selection of autoreactive TCRs recognizing antigens presented by HLA, which could explain the faster development of autoimmune disease in these mice. Interestingly, MART1-TCR⁺ T cells upregulated PD-1 in native mouse thymus in the presence of MART1 peptide-presenting human APCs, indicating that these T cells were able to interact with these APCs and undergo TCR signaling. However, the lack of a normal structure in native mouse thymi may have prevented these T cells from being deleted.

Deficiency of mTECs and organized medullary areas might explain the lower levels of Treg development in native mouse thymus compared to grafted human thymus prior to disease onset. However, the level of Tregs were not different in the periphery of mice with native or grafted thymi. Tregs in affected mice even had a more activated phenotype and higher levels of immunosuppressive molecules such as CTLA-4. However, it is likely that Tregs specific for mouse tissue-restricted antigens may fail to develop in either native or grafted thymi.

NSG mice lack the common cytokine receptor γ chain (γc), which is a shared receptor needed for signaling of IL-2, IL-4, IL-7, IL-9, IL-15 and IL-21 cytokines [26]. Development of lymphoid organs relies on interaction of IL-7 receptor-α (IL-7Rα)-expressing lymphoid tissue inducer (LTi) cells and VCAM-1⁺ ICAM-1⁺ mesenchymal organizer cells [27] and mice lacking the expression of common cytokine receptor γ chain have defective lymphoid development [28], including thymus structure [28]. Since TEC differentiation relies on interaction with developing T cells [18–20], we hypothesized that the delayed interaction between TECs and T cell progenitors may contribute to ongoing structural deficiencies in native mouse thymi of reconstituted NSG mice. Consistently, injection of HSCs at the neonatal stage delayed disease onset compared to that at the adult stage and slightly improved structural organization. However, we did not observe an improvement in cellularity of the native thymus with neonatal injection of HSC, though earlier appearance of T cells in peripheral blood was observed.

Although HIS mice with a grafted human thymus show delayed onset of autoimmune disease development compared to those with a native mouse thymus, all of them eventually develop the disease. Unlike the native NSG mouse thymus, the grafted thymi have a well-organized structure, with distinct cortical and medullary areas and Hassall’s corpuscles mainly in the medullary region. As we showed previously, they select a diverse TCR repertoire and show a normal pattern of TCR repertoire formation [9]. Also, as we showed here and in our previous report [15], autoreactive T cells are efficiently deleted in the human thymus grafts in the presence of corresponding peptides. Considering the seemingly excellent ability of the grafted human thymi to remove autoreactive T cells and the presence of mouse DCs in grafted human thymi, it is expected that T cells with specificities against mouse hematopoietic antigens presented on mouse DCs are deleted in these grafts, while T cells specific for mouse antigens solely expressed by mouse TECs (i.e. tissue-restricted antigens (TRAs)) may be released to the periphery. We have recently demonstrated spontaneous negative selection of an autoreactive human TCR recognizing an mTEC-derived autoantigen in human thymus grafts in humanized mice [29], and the absence of mTECs in the native murine thymus would be expected to impair negative selection in the native murine thymus as well. In both groups, there would also be an expected lack of Tregs recognizing mouse TRAs. Even though there are reports of a contribution of hematopoietic-derived APCs in the positive selection of thymic Tregs [30], studies using TCR transgenic mice with xenogeneic thymus grafts ruled out a role for non-thymic MHC in positive selection [31]. The general consensus from mouse studies is that TECs play a major role in selection of Tregs [32]. Therefore, release of mouse TRA-specific effector T cells and lack of selection of mouse TRA-restricted Tregs could explain the eventual development of autoimmunity in HIS mice with a grafted human thymus. Lack of negative selection for T cells recognizing mouse TRAs presented on human HLA molecules likely contributes to disease development, consistent with the development of disease in mice lacking murine MHC antigens. We propose that injection of wild-type NOD thymic epithelial cells into the fetal human thymus before engraftment could prevent autoimmunity in NSG HIS mice with a grafted human thymus. In fact, our group has previously reported a multi-organ autoimmune disease, similar to that reported here, in BALB/c nude mice grafted with swine fetal thymus [33]. Our data suggested that this autoimmune disease is caused by the lack of mouse TRA-specific Tregs and impaired negative selection of mouse TRA-specific thymocytes and it was corrected by co-implantation of murine TECs in the porcine thymus grafts [34]. A similar approach deserves exploration for the ability to prevent the eventual development of autoimmune disease in thymectomized NSG mice receiving thymocyte-depleted human fetal thymus grafts and HSCs. As intercellular self-antigen transfer in thymic medulla has been established [35,36], it is likely that transfer of mouse antigens from mouse TECs to human APCs in human thymus grafts could result in deletion of T cells recognizing mouse TRAs on human HLA molecules and positive selection of Tregs with such specificities, thereby preventing the disease.

This model has several important biological implications, specifically with regards to human autoimmunity and GVHD. The data provide a direct demonstration of the role of impaired thymic selection of human T cells in autoimmune disease development. We demonstrated that T cells recognizing antigens presented indirectly on human APCs can induce autoimmune disease, a finding that has implications for patients undergoing bone marrow transplantation in addition to those with autoimmune disease. We showed that newly developed human T cells that form in a recipient thymus lacking normal structure fail to undergo normal negative selection, providing a mechanism for initiation of autoimmunity, an observation that is relevant to individuals with primary and secondary thymic abnormalities. In studies to be published elsewhere, we have further dissected the mechanisms of disease, obtaining evidence that the multi-organ autoimmune disease bears some resemblance to systemic lupus erythematosus (SLE).

• T cells are essential for development of multi-organ autoimmunity in humanized mice

• Macrophages and De novo-generated T cells were the major disease drivers

• Defective structure and impaired selection of thymocytes in the NSG mouse thymus

• Lower regulatory T cells in native mouse thymus compared to grafted human thymus

• Disease developed without direct recognition of antigens on recipient mouse MHC