Female C57BL/6J (B6; H-2^b, CD45.2, Thy1.2) mice were purchased from CLEA Japan. C57BL/6J.SJL (B6.SJL; Ptprca Pepcb, CD45.1, Thy1.2), C3H.SW-H2b (B6; H-2^b, CD45.2, Thy1.2), and C57BL/6J-Igh^a-Thy1^a-Gpi^a (B6.Thy1a; CD45.2, Thy1.1) mice were purchased from the Jackson Laboratory. B6 background congenic strains were crossed in-house to obtain CD45.1⁺ CD45.2⁺ Thy1.2⁺ and CD45.1⁺ CD45.2⁺ Thy1.1⁺ congenic strains. All animal experiments were conducted in accordance with institutional guidelines with the approval of the Animal Care and Use Committee of the University of Tokyo.

Cell preparation and allo-HSCT were performed as described previously (17, 18) with some modifications. In brief, T-cell-depleted BM (TCD BM) was prepared by depleting Thy1⁺ mature T cells from BM using an autoMACS system (Miltenyi Biotec). Splenic T cells were negatively enriched from splenocytes by autoMACS, using antibodies against CD11b, CD11c, B220, Ter-119, NK1.1, and c-kit. Recipients were lethally irradiated (9 Gy, split into two doses given 3 h apart) on day −1, then injected intravenously with 5 × 10⁶ TCD BM cells with or without 3–4 × 10⁶ splenic T cells on day 0. The development of systemic GVHD was quantified by measuring weight loss and using a clinical GVHD scoring system, as described previously (19).

For the assessment of pathological changes in tissues, 4–6-µm formalin-fixed paraffin sections were stained with H&E and assessed by a pathologist (Teppei Morikawa; blinded to experimental group) using a scoring system described previously (20–22). For immunohistological analyses, the left lobes of the lung were inflated by infusion of 500 µl OCT compound intratracheally before lung tissue was harvested. Acetone-fixed 6- to 8-µm cryosections were incubated sequentially with primary antibodies and the appropriate fluorochrome-labeled secondary antibodies after blocking. Sections were mounted with Prolong Gold Antifade Reagent (Life Technologies) and visualized using an SP-5 confocal microscope (Leica Microsystems).

Single cell suspensions were prepared from the liver, lung, spleen, BM and thymus after systemic transcardial perfusion with PBS. Liver cells were prepared by pressing liver tissue through a 200-µm stainless steel mesh (23). The right lobe of lung tissue was cut into small fragments and digested for 1 h at 37°C with 0.2% collagenase D (Roche, Penzberg, Germany) and 2,000 U/ml DNase I (Calbiochem, La Jolla, CA, USA). BM cells were flushed from femurs using a needle and syringe. Spleen and thymus tissue were pressed through a 70-µm cell strainer. Non-hematopoietic cells and cell debris were removed by 40% Percoll (GE healthcare) phase separation, and erythrocytes were removed using ACK lysing buffer.

Labeled and purified antibodies were purchased from BD Biosciences, BioLegend, or eBioscience (Table 1). Single cell suspensions were incubated sequentially with anti-CD16/32 (to block Fc receptors) then primary antibodies. Data were collected on a Gallios flow cytometer (Beckman Coulter) and analyzed using FlowJo software (Tree Star). For intracellular cytokine analysis, 5–10 × 10⁵ leukocytes were restimulated with ionomycin (1 µg/ml) and PMA (25 ng/ml) in the presence of Brefeldin A (10 µg/ml) for 4.5–5 h at 37°C. Following staining for surface antigens, cells were stained for intracellular cytokines using a Cytofix/Cytoperm kit (BD Bioscience), according to the manufacturer’s instructions. For short-term pulse BrdU labeling, mice were injected intraperitoneally with 1 mg/mouse BrdU (Sigma-Aldrich) in 100 µl PBS 1 h before sacrifice. BrdU incorporation was examined using a BrdU flow kit (BD Biosciences), according to the manufacturer’s instructions. For the intracellular staining of Foxp3 and Ki-67, 1 × 10⁶ leukocytes were stained for surface antigens and then were fixed and permeabilized with Foxp3 Fix/Perm Buffer (eBioscience) according to the manufacturer’s instructions.

Chronic graft-versus-host disease was induced by transferring CD45.1⁺ Thy1.1⁺ TCD BM and CD45.1⁺ Thy1.2⁺ splenic T cells. Single cell suspensions of the liver and lung were prepared on day 49, and total T cells were enriched by negative selection with antibodies against CD11b, CD31, CD140a, CD326, B220, NK1.1, and Ter-119. T_G and T_HSC were enriched from the T cell fraction using positive selection with antibodies against Thy1.2 and Thy1.1, respectively, and used as secondary donors in in vivo proliferation assays, or as responders in the ex vivo proliferation assay. Non-hematopoietic cells were prepared from the lung of untreated C3H.SW-H2b mice or day 49 BMT recipients by negative selection with antibodies against CD45 and Ter-119. CD3⁻ NK1.1⁻ hematopoietic cells were prepared from the spleen of untreated C3H.SW-H2b mice or day 49 BMT recipients by negative selection with antibodies against CD3 and NK1.1. In an ex vivo proliferation assay, T_G and T_HSC were labeled with 5 µM CFSE and were cocultured with the non-hematopoietic cells or CD3⁻ NK1.1⁻ hematopoietic cells at a ratio of 1:10 for 3 days. In an in vivo proliferation assays, CFSE-labeled T_G and T_HSC were equally mixed and transferred intravenously into day 49 BMT recipients reconstituted with CD45.2 TCD BM.

To deplete T_G or T_HSC selectively, recipient mice expressing various combinations of the congenic markers Thy1.1 and Thy1.2 were injected intraperitoneally with purified rat IgG2b anti-mouse Thy1.2 mAb (clone 30H12, BioXcell) on days 21 (200 µg) and 24 (100 µg) after allo-HSCT, followed by weekly administration (200 µg) from days 31–63.

All values are expressed as mean ± SEM. All data are representative of results obtained in at least two independent internally controlled experiments, where intergroup effects observed within each experiment were consistent across all experiments. Survival curves for GVHD were plotted using Kaplan–Meier estimates and compared by log-rank analysis (Prism version 5.0; GraphPad Prism Software). Comparisons between two groups were performed by unpaired two-tail Student’s t-tests. Multiple comparisons were performed by one-way ANOVA with Dunnett’s post-test. P-Values less than 0.05 were considered to be statistically significant.

We first established a B6→C3H.SW minor-mismatched cGVHD model to facilitate identification and manipulation of T_G and T_HSC populations using CD45 and Thy1 congenic markers. Lethally irradiated C3H.SW recipients were transplanted with TCD BM alone (“BMT group”) or TCD BM with splenic T cells (“cGVHD group”) from B6 donors. In the cGVHD group, body weight loss and GVHD score became progressively worse from day 14 onward (Figures 1A,B), and cGVHD mice had a significantly lower survival rate (Figure 1C). At day 63, the histology of the salivary glands (SG), skin, liver, and lung revealed inflammatory cell infiltration and fibrotic changes, which are the most relevant histological features in the diagnosis of human cGVHD (Figures 1D,E) (24, 25). The pathological scores for these organs were higher in cGVHD mice compared to BMT control mice, reaching statistical significance in the SG and liver (Figure 1F). These results demonstrate that cGVHD mice in the B6→C3H.SW minor-mismatched allo-HSCT model develop pathology that recapitulates human cGVHD.

We next examined the kinetics of T_G and T_HSC in our cGVHD model using CD45 congenic markers (Figure 2A). The number of CD4⁺ CD8⁺ thymocytes of CD45.1⁺ CD45.2⁺ TCD BM origin (T_HSC) in cGVHD mice increased markedly between days 21 and 35 (Figure 2B). In peripheral tissues such as the liver, lung, and spleen, detectable numbers of T_HSC appeared from day 21 after HSCT, but within the CD8⁺ T cell populations of these tissues, T_G outnumbered T_HSC approximately 10-fold for the duration of our experiments (Figure 2C). A similar trend was observed in CD4⁺ T cell populations. Interestingly, T_G were more abundant than T_HSC within CD4⁺ Foxp3⁺ regulatory T cells in the liver by day 35; however, there was a trend toward T_HSC dominance by day 63 (Figure 2D). Immunofluorescent staining of the liver showed that both Foxp3⁺ CD4⁺ T_G and Foxp3⁺ CD4⁺ T_HSC were uniformly distributed throughout the T cell pool at day 63 (Figure 2E). These results suggest that T_G rather than T_HSC persist as the dominant CD8⁺ T cell population, while T_HSC gradually become the dominant CD4⁺ T cell population, particularly within the Foxp3⁺ CD4⁺ regulatory T cell populations in cGVHD-affected organs.

We next compared the phenotype and function of T_G and T_HSC in the liver at day 63. Compared to their T_HSC counterparts, CD8⁺ T_G expressed high levels of the exhaustion markers PD-1 and LAG-3, the senescence/terminal differentiation marker KLRG-1 (11, 26), and low levels of IL-7Rα, which is inversely correlated with alloreactivity (27) (Figure 3A). Expression levels of the early activation marker CD69 were comparable between T_G and T_HSC (Figure 3A). However, we did not observe significant differences in surface marker expression between CD4⁺ T_G and CD4⁺ T_HSC (Figure 3B). Upon ex vivo stimulation of liver-infiltrating T cells with PMA and ionomycin, a substantial proportion of CD8⁺ T_G produced IFNγ and TNFα at levels that were equal to those produced by their T_HSC counterparts (Figure 3C). Because T_G far outnumbered T_HSC in cGVHD-affected organs, the majority of cells with potential to produce inflammatory cytokines in cGVHD-affected organs were of T_G origin (Figure 3D). A similar trend was observed for CD4⁺ T_G and T_HSC (Figures 3C,D). These results suggest that a substantial proportion of CD4⁺ and CD8⁺ T_G persist as a functional population in the target organs of cGVHD.

Interestingly, the proportion of PD-1⁻ KLRG-1⁺ cells within the CD8⁺ T_G population increased between day 21 and day 63 (Figure 3E). To examine whether these PD-1⁻ KLRG-1⁺ cells are functionally distinct from the PD-1⁺ KLRG-1⁻ cells that make up the majority of CD8⁺ T_G in early phase, we analyzed the expression of functional molecules and the proliferation marker Ki-67 by these two populations at day 60. The PD-1⁺ KLRG-1⁻ population displayed higher IFNγ production and LAG3 expression than the PD-1⁻ KLRG-1⁺ population, whereas the PD-1⁻ KLRG-1⁺ population displayed higher granzyme B and Ki-67 expression than the PD-1⁺ KLRG-1⁻ population (Figures 3F,G). We also compared the functional profiles of the PD-1⁺ KLRG-1⁻ populations from CD8⁺ T_G and T_HSC. Although expression levels of LAG3 and granzyme B were equivalent between T_G and T_HSC, PD-1⁺ KLRG-1⁻ T_HSC had lower IFNγ production and significantly higher expression of Ki-67 than their T_G counterparts (Figure 3G). These results suggest that among CD8⁺ T_G the PD-1⁺ KLRG-1⁻ and PD-1⁻ KLRG-1⁺ populations are functionally distinct and that CD8⁺ T_HSC have a higher proliferating potential than CD8⁺ T_G even when PD-1 expression is equivalent.

We next investigated the proliferative responses of T_G and T_HSC by 1 h in vivo BrdU labeling on day 63 after allo-HSCT. The proportion of proliferating (BrdU⁺) cells among liver-infiltrating CD4⁺ T cells was significantly lower within the T_G population than in the T_HSC population, whereas there was no significant difference in the proportion of BrdU⁺ CD8⁺ T cells (Figure 4A). However, in terms of cell number, CD4⁺ BrdU⁺ T_G were present in equivalent numbers to CD4⁺ BrdU⁺ T_HSC, while BrdU⁺ CD8⁺ T_G far outnumbered BrdU⁺ CD8⁺ T_HSC (Figure 4B). To further investigate the proliferative potential of T_G and T_HSC, we prepared T_G and T_HSC populations from the liver and lung of cGVHD mice and adoptively transferred them into phase-matched BMT recipients on day 49 after allo-HSCT (Figure 4C). A CFSE-dilution assay demonstrated that a significant proportion of CD4⁺ and CD8⁺ T_G, and an even greater proportion of T_HSC proliferated 3 days after transfer into secondary BMT recipients (Figures 4D,E). These results demonstrate that T_G maintain proliferative potential during the chronic phase.

In the chronic phase of GVHD models, APCs are classified into host radioresistant non-hematopoietic cells or donor BM-derived hematopoietic cells. The former directly present complexes of host-type minor histocompatibility antigen (miHA)-derived peptide and MHC molecule to donor T cells, whereas the latter uptake host-type miHA from host radioresistant non-hematopoietic cells and present miHA-peptide/MHC complexes to donor-T cells through an indirect pathway. To investigate the allo-recognition pathway that contributes to the maintenance of T_G and T_HSC populations after allo-HSCT, we cocultured T_G or T_HSC with BMT recipient-derived APCs 49 days after allo-HSCT. To examine the direct pathway, we used radio-resistant host-derived non-hematopoietic cells that present endogenous host-type antigens (Figures 4F,G). To examine the indirect pathway, we used donor-derived T cell/NK cell-depleted hematopoietic cells containing professional APCs, which uptake and then present host antigens. We also used untreated C3H.SW mouse-derived hematopoietic cells and non-hematopoietic cells to examine the direct pathway that is involved in the early induction of effector T_G. CFSE-dilution assays demonstrated that neither CD4⁺ nor CD8⁺ T_G proliferated in response to coculture with non-hematopoietic cells prepared from BMT recipient or untreated C3H.SW mice. In contrast, both CD4⁺ and CD8⁺ T_G proliferated when cocultured with hematopoietic cells from C3H.SW mice (direct pathway) and to a lesser extent when cocultured with hematopoietic cells from BMT recipients (indirect pathway) (Figure 4G). T_HSC displayed a similar proliferative response except that a small proportion of CD8⁺ T_HSC proliferated in response to coculture with non-hematopoietic cells. Considering that host-type hematopoietic APCs are completely absent in mice suffering cGVHD (data not shown), these results suggest that indirect host antigen presentation by donor-derived hematopoietic cells may be involved in the maintenance of T_G in cGVHD-affected organs.

Since inflammatory factors such as IFNγ modify antigen processing by inducing expression of immunoproteasomes and MHC molecules (28), we performed similar donor T cell and APC ex vivo coculture experiments with APCs that had been pretreated with 10 U/mL IFNγ for 16 h, which induced the expression of MHC class I and MHC class II on both hematopoietic and non-hematopoietic APCs. IFNγ pretreatment did not affect antigen presentation by non-hematopoietic APCs (Figure 4H). However, IFNγ pretreatment of hematopoietic APCs significantly increased the proliferation of CD4⁺ and CD8⁺ T_G in both direct and indirect pathway-dependent settings. In contrast, IFNγ pretreatment of hematopoietic APCs significantly suppressed the direct pathway-dependent proliferation of CD4⁺ T_HSC and indirect pathway-dependent proliferation of CD8⁺ T_HSC. These results suggest that IFNγ expression in cGVHD-affected tissues may cause prolonged dominance of T_G among CD8⁺ T cells.

To determine the involvement of T_G and T_HSC in the pathogenesis of cGVHD, we selectively depleted T_G, T_HSC, or both T_G and T_HSC from cGVHD mice using Thy1.1/1.2 congenic markers and an anti-Thy1.2 depleting mAb (Figure 5A). This protocol resulted in >95% depletion of Thy1.2⁺ T cells from cGVHD-affected organs from day 35 to the end of our experiments (Figure 5B). Unexpectedly, survival rate, fibrotic changes, and inflammatory cell infiltration in cGVHD-affected organs were not ameliorated by the depletion of T_G (Figures 5C–E), despite the predominance of functionally competent T_G in the cGVHD-affected organs of untreated cGVHD mice. These results suggest that T_G alone are not responsible for the development and/or maintenance of the histopathology of cGVHD. On the other hand, survival was markedly reduced in the T_HSC-depleted group, in which all mice died within approximately 3 weeks of initial antibody treatment (Figure 5F). This lethal exacerbation of GVHD was not due to the depletion of Thy1⁺ HSCs by the anti-Thy1.2⁺ antibody, because the number of CD48⁻ CD150⁺ lineage⁻ c-Kit⁺ Sca1⁺ long-term HSCs was equivalent between the cGVHD and T_HSC-depleted groups (Figure 5G). In addition, there was no significant difference in the frequency of Foxp3⁺ liver-infiltrating CD4⁺ T cells between the cGVHD and T_HSC-depleted groups, suggesting that the lethal exacerbation of GVHD in the T_HSC-depleted group could not be attributed to the loss of Foxp3⁺ regulatory T cells (Figure 5H). Importantly, simultaneous depletion of T_G and T_HSC rescued cGVHD mice from the lethal exacerbation of disease induced by the depletion of T_HSC alone (Figure 5F). These results suggest that neither T_G nor T_HSC are solely responsible for cGVHD and that the lethal exacerbation of GVHD induced by T_HSC depletion is mediated by T_G.

The observation that simultaneous depletion of T_G and T_HSC ameliorated cGVHD while selective depletion of T_G or T_HSC failed to ameliorate cGVHD suggests that an interaction exists between these two distinct T cell populations. To investigate this interaction further, we examined the effects of T_HSC depletion on the number and function of T_G. Since T_HSC-depleted mice died within 3 weeks of depletion (day 21), analyses were performed on days 28 and 35. There were no significant differences in liver and lung T_G numbers between the T_HSC-depleted and cGVHD groups (Figure 6A). In addition, the frequency of BrdU⁺ (proliferating) cells within the T_G population and the expression levels of CD44, CD25, and CD69 on T_G cells were equivalent between the T_HSC-depleted and cGVHD groups (Figures 6B,C). However, the proportion of IFNγ⁺ and TNFα⁺ cells among liver-infiltrating CD4⁺ and CD8⁺ T_G cells was higher in the T_HSC-depleted group compared to the cGVHD group on both days 28 and 35 (Figure 6D and data not shown). These results suggest that T_HSC, despite being 10–100 times lesser in number than T_G at the time of depletion, inhibit inflammatory cytokine production by T_G and play a critical role in preventing lethal exacerbations of GVHD.

Finally, we examined the effects of T_G depletion on the number and function of T_HSC. In kinetic studies of CD4⁺ and CD8⁺ T cell number in the liver, T_G rapidly decreased after anti-Thy1.2 mAb treatment in the T_G-depleted group, whereas T_HSC increased to numbers approximately 10-fold higher than the cGVHD group between days 35 and 63 (Figure 7A). By day 63 after allo-HSCT, the number of T_HSC (particularly CD4⁺ T_HSC) in the T_G-depleted group had increased to levels comparable to the sum of T_G and T_HSC numbers in the cGVHD group (Figure 7B). To determine whether this increase in T_HSC number was due to an increase in T_HSC proliferation, we performed BrdU labeling on day 63 (Figure 7C). The proportion of BrdU⁺ cells among liver-infiltrating CD4⁺ and CD8⁺ T_HSC was significantly higher in the T_G-depleted group, outnumbering the sum of BrdU⁺ T_G and T_HSC in the cGVHD group. There was no significant difference in the frequency of Foxp3⁺ cells among CD4⁺ T cells between the cGVHD and T_G-depleted groups (Figure 7D). The majority of CD8⁺ T_HSC in the liver and lung, but not in the spleen, of T_G-depleted group mice were CD44^hi effector T cells and had increased expression of the activation markers CD25 and CD69 (Figure 7E). The total proportion of IFNγ⁺ or TNFα⁺ cells among liver-infiltrating CD4⁺ and CD8⁺ T cells was equivalent between T_G-depleted and cGVHD group mice (Figure 7F). However, the cellular composition of these populations switched from predominately T_G in the cGVHD group to predominately T_HSC in the T_G-depleted group. Collectively, these results demonstrate that the proliferation and activation of T_HSC are augmented by T_G depletion in cGVHD-affected organs. This change appears to compensate in part for the decrease in T_G numbers and is likely to contribute to the histopathology observed in the cGVHD affected organs of the T_G-depleted group.