Inbred female C57BL/6, B6.SJL-Ptprc^aPeprc^b (PTPrc^a, CD45.1⁺) and B6D2F1 mice were purchased from the Animal Resource Centre (Canning Vale, WA, Australia). B6.Foxp3.DTR (Stranges et al., 2007), B6.FoxP3-GFP (Fontenot et al., 2003), RAG1^–/– and B6.BDCA2-DTR (Lynch et al., 2018) mice were bred and maintained in-house at QIMR Berghofer Medical Research Institute. B6.IL-10-GFP, B6.FoxP3-RFP were originally provided by Alexander Rudensky (Memorial Sloan Kettering Cancer Institute), and B6.FoxP3-RFPxIL10-GFP mice (Edwards et al., under review) were intercrossed and kindly provided by Christian Engwerda (QIMR Berghofer Medical Research Institute). Atg7^fl/fl (Kuma et al., 2004) mice were provided by Masaaki Komatsu (Tokyo Metropolitan Institute of Medical Science). FoxP3^{Cre–IRES–YFP} mice (Rubtsov et al., 2008) were bred with Atg7^fl/fl to generate Atg7^fl/fl-FoxP3cre⁺ mice. Female mice were used between 8 and 12 weeks, unless otherwise indicated. All mice were maintained in-house under pathogen-free conditions. All animal procedures were conducted with the approval of the QIMR Animal Ethics Committee under the animal ethics number A1503-603M and in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes (National Health and Medical Research Council, Canberra, Australia).

On day 0, recipient mice received 1,100 cGy (B6D2F1) or 1,000 cGy (C57BL/6, B6) total body irradiation split into two doses separated by 3 h. T cell depleted bone marrow (TCD-BM) and splenic T cells were purified using magnetic bead depletion (purity > 85%) as described (Burman et al., 2007). Recipients were transplanted with 5 × 10⁶ (B6) TCD-BM from C57BL/6, B6.FoxP3-GFP.DTR, B6.WT.FoxP3cre-YFP⁺, B6.Atg7^fl/fl. FoxP3cre-YFP⁺, or BDCA2-DTR donor mice as indicated, with or without splenic T cells from B6.FoxP3-GFP, B6.FoxP3-RFP, or C57BL/6 donors. Transplanted mice were monitored daily to evaluate clinical GVHD scores as previously published (Hill et al., 1997). Briefly, weight loss, posture, activity, fur texture, and skin integrity were measured. Mice with a score greater than or equal to 6 were culled, and the date of the following day is recorded as the death date. Where indicated, recipient mice received intraperitoneal injections IL-2 (0.5 μg)/JES6-1 anti-IL-2 mAb (25 μg) complexes on days 21–34. In studies where Treg or plasmacytoid dendritic cells (pDC) were selectively depleted from BM grafts during acute GVHD, mice were injected intraperitoneally with diptheria toxin (DT) (1 μg/animal, in 200 μl total volume) twice weekly from day 21 post-transplant, as indicated.

For adoptive transfer experiments, on day 0, B6.RAG1^–/– recipient mice received 0.5 × 10⁶ sort purified B6.FoxP3.GFP⁺ Treg isolated from the spleens of B6.FoxP3.GFP mice. BM and SP were harvested from recipient mice 4 weeks following adoptive transfer for flow cytometry analysis.

To assess T cell responses to in vivo cytokine stimulation B6.FoxP3.GFP mice were administered with either saline or recombinant murine cytokines over 4 days. Control animals received saline intraperitoneal injections daily for 4 days. IL-2 complex was administered daily for 4 days: 0.5 μg recombinant mouse IL-2 (eBioscience^TM) with 25 μg anti-IL-2 (BioXcell). IL-15 complex was administered on days 1 and 3: 7 μg IL-15Rα-Fc (R&D) with 0.75 μg of recombinant mouse IL-15 (eBioscience^TM). IL-33 was administered daily for 4 days: 1 μg of recombinant mouse IL-33 (Peprotech). IL-9 daily for 4 days: 50 ng of recombinant mouse IL-9 (In Vitro Technologies TY Ltd.). IL-7 complex daily for 4 days: 1 μg of recombinant mouse IL-7 with 5 μg of anti-IL-7 (Jomar Life Research). The absolute number of CD4⁺FoxP3⁺ Treg and CD4⁺FoxP3^neg Tcon in either the spleen and BM of control and treated mice was then assessed on day 6 following the cytokine treatment regime.

Single-cell suspensions of spleen and bone marrow cells were stained with fluorochrome- or biotin-conjugated antibodies against antigens as listed in Table 1. Intracellular staining for Foxp3 and MCL-1 was performed after fixation and permeabilization using the eBiosciences Foxp3 staining kit. Sample data were acquired on a LSRFortessa flow cytometer (BD Biosciences) and analyzed using FlowJo software (TreeStar) version 10. Cell sorting was performed using an ARIA III Cell Sorter (BD Biosciences).

For in vitro suppression assays, CFSE-labeled sort purified splenic CD4⁺ T cells from iVac-cre.BFP⁺ mice, purified using previously described magnetic bead depletion, were seeded at 1 × 10⁴ cells/well in 96-well round-bottom plates, with DCs (B6) at 1,000 cells/well, in the absence or presence of Tregs (B6.Foxp3-GFP) at 1,250, 2,500, or 5,000 cells/well (8:1, 4:1, or 2:1 Tcon:Treg, respectively), and supplemented with anti-CD3 (2C11, 1 μg/ml). Cells were harvested for CFSE dilution analysis at 72 h of culture.

RNA was extracted with Qiagen’s RNeasy micro kit from >100,000 FACS isolated cells as per instructions. For generation of sequencing libraries, 15 ng of RNA (RIN value > 7) were submitted to SPIA amplification (NuGen). Four biological replicates per condition were sequenced using the HiSeq 3,000 sequencing platform (Illumina, San Diego, CA, United States). Each library was sequenced single-end with a 80 nucleotide read length. The targeted number of sequencing reads per sample was ∼30–50 million.

Spleen and BM cells from B6.FoxP3-GFP mice were harvested and immunolabelled with surface antibody cocktail containing anti-CD3 and anti-CD4-biotin labeling antibodies. The cells were then stained for anti-Foxp3-AF647, anti-Ki67-BV786, and Hoechst 33342 (Invitrogen; H3570) using the eBioscience^TM Foxp3/Transcription Factor Staining kit (ThermoFisher) following the manufacturer’s instructions. Spleen and BM CD3⁺CD4⁺FoxP3⁺ Treg were analyzed by flow cytometry for Ki67 and Hoechst (Hö) staining, with different stages of cell cycle defined as G0 (Ki67^negHö^low), G1 (Ki67⁺Hö^low), and S-M (Ki67⁺Hö^high).

Total RNA was extracted from sorted cells using the RNeasy Micro kit (QIAGEN), and reverse transcription was performed using SuperScript III Reverse Transcriptase (Life Technologies). RNA relative expression was determined using Taqman gene expression assays (Life Technologies). The housekeeping gene HPRT was used to normalize the gene expression to the starting quantity of RNA. PPARγ mRNA expression was calculated using the 2^–ΔΔCt method (Livak and Schmittgen, 2001).

CD4⁺ T cells were isolated from the spleen and BM of B6.FoxP3-GFP mice and incubated for 15 min at 37°C with or without recombinant murine IL-2 (100 ng/mL), IL-7 (50 ng/mL), IL-9 (20 ng/mL), or IL-33 (25 ng/mL). Cells were processed according to manufacturer’s protocol (BD Phosflow Perm Buffer III).

The long bones (femur, tibia, and iliac crest) from C57BL/6 mice were cleaned of muscle and tissue and then flushed with cold PBS (250 μl) to isolate the BM cells and extracellular fluid. Spleens from C57BL/6 mice were harvested, mashed between two microscope slides until homogenized, placed in an Eppendorf vial and then cold PBS (200 μl) was added and the mixture vortexed vigorously. The cell suspensions from both bone marrow and spleen were centrifuged at 400 × g for 5 min and the supernatant was then transferred to a clean Eppendorf vial. The supernatant was centrifuged at 10,000 × g for 10 min and stored frozen at −80°C. The resultant cell pellets from BM and spleen were resuspended in PBS (2% FBS) and enumerated for subsequent data normalization. The supernatant samples were analyzed using the Quantibody^® assay at Crux Biolabs and protein concentration normalized relative to cell number.

Data are shown as mean ± SEM. Statistical significance was determined using an unpaired 2-tailed Mann-Whitney U test, Wilcoxon matched-paired test, log-rank (Mantel-Cox) test, or paired t test when appropriate (^∗P < 0.05; ^∗∗P < 0.01; ^∗∗∗P < 0.001; ^****P < 0.0001). Statistical analyses were performed using GraphPad Prism software version 6.01.

We have previously reported that the Treg population that resides in the BM is enriched with an autophagy-dependent TIGIT⁺ Treg population critical for establishing tolerance and controlling GVHD following SCT (Le Texier et al., 2016). Consistent with previous reports (Zou et al., 2004; Wei et al., 2006; Camacho et al., 2020), characterization of the BM resident Treg (BM-Treg) in naïve B6 mice aged 8–12 weeks demonstrated that Foxp3⁺ Treg are highly enriched in the CD4⁺ T cell compartment within the BM as compared to the spleen (SP) (BM: ∼40% vs SP: ∼10%). Further, BM-Treg expressed reduced FoxP3 levels compared to their splenic counterparts (Figures 1A,B). As FoxP3 is increased upon Treg activation and is critical for Treg suppressive function (Hori et al., 2003), reduced FoxP3 expression by BM-Treg suggests these Treg are in a less activated or functional state than pTreg (Liu et al., 2013; Rothstein and Camirand, 2015). Consistent with this, we show that compared to SP Treg, BM-Treg have a reduced capacity to suppress splenic Tcon proliferation in vitro (Figure 1C). However, the BM-Treg compartment is enriched in TIGIT-expressing Treg (BM: ∼36%, vs SP: ∼28%) which identifies a highly functional and activated memory Treg population (Figure 1D; Le Texier et al., 2016). While the TIGIT⁺ Treg within the BM and SP exhibited higher FoxP3 expression than TIGIT^neg Treg, FoxP3 expression was reduced in both TIGIT⁺ and TIGIT^neg Treg subsets in the BM as compared to the SP (Figure 1E). Comparing FoxP3 expression in CD4⁺ Treg across multiple lymphoid organs demonstrated that BM-Treg expressed the lowest FoxP3 levels irrespective of their TIGIT status (Supplementary Figure 1). To investigate this further, we adoptively transferred FoxP3⁺GFP⁺ Treg isolated from the SP of B6.FoxP3-GFP mice into B6.RAG1^–/– recipients (Figure 1F). Four weeks following adoptive transfer, we observed a marked enrichment of FoxP3⁺ Treg within the BM compartment of recipient mice (Figure 1G) and, despite originating from the SP, Treg isolated from the BM of recipient mice acquired a BM-Treg phenotype, including reduced expression of FoxP3 and enriched TIGIT expression (Figure 1H). Taken together, these data suggest that the BM microenvironment influences the phenotype and function of locally residing Treg and may play a central role in development and/or maintenance of the BM-Treg signature.

A failure in Treg homeostasis, with an associated reduction in pTreg numbers, contributes to cGVHD pathogenesis in both patients and mice (Zorn et al., 2005; Koyama et al., 2016; McDonald-Hyman et al., 2016). To assess the impact of cGVHD on the BM Treg compartment, we transferred B6 BM (5 × 10⁶) supplemented with splenic T cells (0.5 × 10⁶) from B6.FoxP3-GFP donors into lethally irradiated allogeneic B6D2F1 or syngeneic B6 recipients (Figure 2A). Late post-transplant, the Treg pool can be comprised of Treg derived from either the mature T cells contained in the graft (Tgraft Treg) or those emerging from the BM compartment (BMgraft Treg). The relative impact on these Treg populations was assessed using B6.FoxP3-GFP T cells. On day 28 after transplant, intracellular FoxP3 staining used to quantify all CD4⁺ Treg revealed a significant reduction in Treg frequency within the CD4 T cell compartment of the SP and to an even greater extent in the BM of mice receiving GVHD-inducing grafts as compared to syngeneic graft recipients (Figure 2B). Notably, in syngeneic recipients, similar to that seen in naïve mice, there was an increased proportion of Treg in the BM compartment compared to the SP, however, Treg failed to enrich in the BM of GVHD recipients. Overall, there was a significant reduction in CD4⁺FoxP3⁺ Treg numbers and FoxP3 expression in SP and BM of GVHD mice with lower FoxP3 expression in BM than SP residing Treg (Figures 2C,D). In syngeneic recipients, the majority of Treg in both the SP (∼85%) and BM (∼70%) were derived from the BMgraft and not the Tgraft GFP⁺CD4 Treg (Figure 2E). While GVHD impaired the reconstitution of both Tgraft and BMgraft Treg in the SP and BM, the relative impact of GVHD was greatest on the BMgraft Treg numbers.

We next assessed the contribution of BMgraft Treg to the control of GVHD. Again using the B6 → B6D2F1 cGVHD model, which is characterized by the development of sclerodermatous cGVHD (Koyama et al., 2016), we transferred BM from B6.FoxP3.GFP.DTR donor mice in which diphtheria toxin receptor (DTR) is driven off the FoxP3 promoter to facilitate specific Treg depletion in diphtheria toxin (DT) treated mice. To restrict Treg depletion to the BMgraft Treg compartment, B6.FoxP3.GFP.DTR BM grafts were supplemented with splenic T cells from a B6.FoxP3-RFP donor that is unaffected by DT administration (Figure 3A). On day 50 post-transplant, recipients administered DT twice weekly from day 21, when thymic T cell reconstitution begins, exhibited exacerbated skin cGVHD compared to those receiving saline, confirming a functional role for BMgraft Treg in contributing to the control of cGVHD (Figure 3B).

Our earlier studies using BM and T grafts from Atg7fl/fl-FoxP3cre⁺ mice with disabled autophagy specifically in FoxP3⁺ Treg demonstrated that Treg intrinsic autophagy is required for efficient Treg reconstitution and GVHD control (Le Texier et al., 2016). To examine the requirement for autophagy in BMgraft Treg engraftment and GVHD control, we next utilized grafts comprised of BM from WT-FoxP3cre-YFP⁺ or Atg7^fl/fl-FoxP3cre⁺-YFP⁺ donors supplemented with B6.FoxP3-RFP T cells (Figure 3C). As compared to WT-FoxP3cre-YFP⁺ BM recipients, enumeration of BMgraft and Tgraft Treg in SP and BM on D35 after transplant revealed that selective Treg autophagy deficiency significantly reduced YFP⁺BMgraft Treg numbers in BM but not SP. Irrespective of the BM source, RFP⁺Tgraft Treg numbers were similar in both the SP and BM compartments (Figure 3D). Importantly, this Treg perturbation, which was restricted to the BM compartment, was associated with increased skin cGVHD pathology including increased dermal thickening and collagen deposition in the skin of recipients of autophagy-deficient BM grafts (Figure 3E).

We have previously shown the administration of anti–IL-2 mAb/IL-2 complexes expands pTreg and reduces cGVHD severity (McDonald-Hyman et al., 2016). Whether anti-IL-2mAb/IL-2 complexes improve BM-Treg numbers during cGVHD is unknown. To address this issue, we performed transplants where B6D2F1 recipients received grafts comprised of BM from FoxP3-YFP donors supplemented with T cells from B6.FoxP3-RFP mice that permits discrimination of Tgraft-derived from BM-derived Treg post-transplant (Figure 3F). Anti–IL-2 mAb/IL-2 complexes effectively expanded SP-Treg, including both Tgraft and BMgraft-derived Treg (Figure 3G). In contrast, there was minimal impact on the BM-Treg, suggesting that additional Treg enhancing therapeutic strategies are required to specifically improve BM-Treg numbers for better control of cGVHD.

To gain a better understanding of the phenotypic and functional differences between Treg of BM and splenic origin, we preformed bulk RNA sequencing (RNAseq) analysis on FoxP3⁺ Treg isolated from the BM or SP of naïve B6.FoxP3-GFP mice (Figure 4A). We identified 1056 differentially expressed genes (DEGs; 526 downregulated and 494 upregulated) in BM-Treg confirming that BM-Treg are transcriptionally distinct from SP Treg (Figure 4B). In line with recent studies, we observed increased expression of several genes [KLRG1, Fgl2, Itga2 (CD49b), and Tsc22d3 (GILZ)] that were consistent with an enrichment of an activated memory Treg phenotype within the BM (Figure 4C; Joller et al., 2014; Camacho et al., 2020). In contrast, and in concordance with our observed reduction in suppressive function in BM-Treg, multiple genes associated with Treg function [Gzmb, Itgea (CD103), Lag3, Lrrc32 (GARP), Nt5e (CD73), and Gpr83] were downregulated in BM-Teg. Relative to SP-Treg, BM-Treg exhibited divergent patterns of cytokine and chemokine receptor gene expression with BM-Treg expressing elevated levels of cytokine receptors for IL-9, IL-33, IFNγ, IL-7, IL-6, IL-10, and chemokine receptors CCR3 and CCRL2. In BM resident Treg, differences in transcription factor expression including the downregulation of FoxP3 interacting transcription factors Ezh2, Irf4, and Tcf7 were noted. Thus, the BM-Treg exhibit a unique transcriptional profile with differential expression of genes associated with Treg function, activation, cytokine responsiveness, and migration.

Using Ingenuity interpretative phenomenological analysis (IPA) of likely upstream regulators (USRs) we observed a signature in BM-Treg consistent with a loss of activity and gain of a quiescent state. MYC was one of the most significantly dysregulated USRs predicated by IPA (Figure 4D). BM-Treg MYC activity was predicted to be significantly downregulated, with a z-score of −5.39. Additionally, 127 of the 1053 DEGs identified by our bulk RNAseq analysis are direct MYC targets, 78 of which were differential expressed in BM-Treg consistent with inhibition of MYC signaling. MYC drives cell cycle and proliferation in response to TCR and IL-2 (Tzachanis et al., 2004; Chapman et al., 2020; Rodriguez et al., 2021) which are also predicted to be downregulated in BM-Treg (Figure 4D) suggesting this subset is in a more quiescent and less proliferative state than pTreg. Supporting this, fewer BM-Treg expressed Ki-67, and cell cycle analysis showed an increased proportion of BM-Treg over SP-Treg in G0, and thus quiescent (Figures 4E,F). Together these data demonstrate that while BM-Treg may be enriched with activated memory Treg subsets, they are simultaneously in a more resting and quiescent state than pTreg.

Extending our phenotypic analysis as guided by the DEG list and literature, we performed flow cytometric profiling to identify unique features of BM-resident Treg. Analysis of a broad panel of functional markers on CD4⁺FoxP3^neg Tcon and CD4⁺FoxP3⁺ Treg from the SP and BM of 12 week old WT B6 mice identified several markers that were specifically increased on BM-Treg compared to SP-Treg (Supplementary Figure 2). Another set of markers (CD69, CD49b, and CD39) were observed to be increased on both Tcon and Treg in the BM, while both GARP and TCF-1 (encoded by Tcf7) were downregulated on BM-Treg (Supplementary Figure 2). Differential expression of functional markers was confirmed using B6.FoxP3-GFP reporter mice (Supplementary Figure 3), and the relative expression of these antigens was further assessed in TIGIT⁺ and TIGIT^neg CD4⁺FoxP3-GFP⁺ Treg sub-populations. As noted for TIGIT expression, the majority of these Treg markers [PD-1, CD103, EpCam (CD326), CD39, CD103, and CD69] were expressed on a BM-Treg subset, rather than globally upregulated. Sca-1 expression was noted to be specifically upregulated within the TIGIT⁺ Treg subset of BM-Treg compared to TIGIT⁺ SP-Treg. In contrast, CD127 (IL-7R), CD73, and INF-γR expression were globally upregulated on BM-Treg population. Thus, through recruitment and/or the local influence of the microenvironment, the BM harbors distinct and heterogeneous Treg subsets.

To further assess the heterogeneity of BM-Treg we surveyed key BM-Treg signature marker expression as related to TIGIT. KLRG1 marked a distinct population of TIGIT⁺ Treg that co-expressed ST2, signature markers of non-lymphoid tissue Treg (Figure 4G). Using gene set enrichment analysis of the RNAseq data, we identified an enrichment of genes related to visceral adipose tissue (VAT) Treg (Feuerer et al., 2009), and PPARγ signaling in Treg which drives VAT Treg accumulation, phenotype and function (Cipolletta et al., 2012; Supplementary Figure 4A). IPA identified PPARγ signaling as a top upregulated canonical pathway in BM-Treg (Supplementary Figure 4B). Although the level of PPARγ did not pass gene expression filters during RNAseq analysis, as GSEA and IPA analysis suggested PPARγ to be active in the BM-Treg, we performed RT-qPCR and confirmed upregulation of PPARγ in BM-Treg relative to SP-Treg (Supplementary Figure 4C), indicative of a VAT-like KLRG1⁺ST2⁺ tissue Treg subset enrichment in BM.

At birth, Treg numbers in VAT are limited, but progressively accumulate over time with peak Treg numbers occurring at ∼25 weeks of age (Feuerer et al., 2009). To assess whether the BM-Treg, and particularly the VAT-like KLRG1⁺ Treg, exhibited a similar accumulation pattern, SP- and BM- Treg compartments were examined in mice aged from 2 weeks to 1 year old. At 2 weeks, there was already a significant FoxP3⁺ Treg enrichment in the BM CD4 T cell compartment (Figure 5A), which was maintained over time (Figure 5B). Similarly, in 2-week-old mice, TIGIT⁺ and KRLG1⁺ Treg subsets were significantly increased in BM-Treg compared to the SP (Figures 5C,E). The proportion of TIGIT⁺ Treg was consistently higher in the BM-Treg compartment; over time, the proportion of TIGIT⁺ Treg increased in both SP and BM (Figure 5D). In contrast, and opposed to Treg accumulation in VAT, the proportion of KLRG1⁺ BM-Treg reduced significantly after 2 weeks, and gradually declined thereafter (Figure 5F). In 2-week-old mice, Foxp3 expression was not reduced in BM-Treg as was seen for BM-Treg in older mice. Rather, we noted increased FoxP3 expression in the TIGIT⁺ BM-Treg (Figures 5G,H). Treg in neonates undergo robust expansion due to a lymphogenic environment (Min et al., 2003). Since proliferating Treg upregulate TIGIT expression, these data suggest that the reduced BM-Treg FoxP3 levels in adult mice may be related to their relative quiescent state.

BM-Treg have been reported as high producers of IL-10 which has direct effects on multiple populations within the BM (Fujisaki et al., 2011; Fischer et al., 2019; Camacho et al., 2020). We previously demonstrated an enrichment of IL-10 producing Treg within the TIGIT⁺ BM-Treg compartment (Le Texier et al., 2016; Yu et al., 2018). Using B6.FoxP3-RFPxIL10-GFP reporter mice, we identified IL-10⁺ Treg as a CD304⁺CD69⁺PD1⁺ memory Treg subset (Supplementary Figure 5). Notably, the KLRG1⁺ Treg subset lacked IL-10 expression. Although increased in proportion in the BM, the IL-10⁺ Treg phenotype was similar in both the SP and BM compartments, with the exception that IL-10⁺BM-Treg had a higher level of Sca1 expression. Together these data further support that BM harbors not only phenotypically, but functionally distinct subsets of Treg and that the BM microenvironment influences multiple aspects of BM-Treg phenotype and function.

Although IL-2 is critical for pTreg homeostasis, the quiescent status of BM-Treg correlated with a predicted reduction in IL-2 signaling consistent with the finding that IL-2mAb/IL-2 complexes failed to expand BM-Treg after transplant and suggesting an alternative cytokine(s) may function to maintain and shape the Treg in the BM niche. Since BM-Treg exhibited upregulated IL-9R, CD127 (IL-7R), and IL-33R (ST2) expression, we first compared phospho-STAT5 (pSTAT5) responses in BM and SP Treg following in vitro cytokine stimulation, and accordingly, observed increased pSTAT5 response in BM-Treg following in vitro IL-2, IL-7, and IL-9 stimulation (Figure 6A). IL-33 failed to elicit pSTAT5 in either population. Only IL-7 elicited pSTAT5 in FoxP3^neg Tcon that was similar in both BM and SP Tcon (Figure 6B). As the local cytokine milieu can influence cytokine receptor expression and cellular responses, and IL-7 is an established abundant cytokine in the BM, we assessed IL-2, IL-9, and IL-33 levels in the SP and BM niche. We found limited amounts of IL-2 in either tissue. IL-9 was significantly increased in the BM while IL-33 was more abundant in the SP than BM (Figure 6C), suggesting a potential role for IL-9 in modulating BM-Treg locally.

We next examined the impact of in vivo IL-7, IL-9, and IL-33 administration on Treg and Tcon populations is SP and BM. In addition to IL-2, IL-15 also acts as a potent inducer of pTreg mediated responses. Therefore, we compared responses to those elicited by IL-2mAb/IL-2 complexes or IL-15mAb/IL-5 complexes. Following 4 days of administration to naïve B6.FoxP3-GFP mice (Figure 6D), IL-2mAb/IL-2 complexes stimulated the greatest expansion of SP-Treg, with no expansion of Tcon. Similar to what we observed in cGVHD mice, there was little to no impact on BM-Treg numbers (Figure 6E). IL-15mAb/IL-15 complexes failed to increase SP-Treg numbers, however, BM-Treg were decreased and there was a statistical trend to increased Tcon in the BM. IL-33 had no impact on Treg in either the SP or BM, but decreased Tcon in both compartments. While IL-7 expanded both Tcon and Treg in the SP and is reduced in the BM. In contrast and in support of IL-9 as a BM-Treg support cytokine, IL-9 treatment significantly increased BM-Treg numbers in parallel with a decrease in BM Tcon in the BM but had no effect on splenic Treg or Tcon.

Dendritic cells (DC) are required for the induction and maintenance of Treg both in steady state and after SCT (Stenger et al., 2012; Yu et al., 2019). Impaired function of conventional DC (cDC) and reconstitution of plasmacytoid DC (pDC) during GVHD have both been demonstrated as mechanisms underpinning failed Treg homeostasis (Banovic et al., 2009; Koyama et al., 2016; Tian et al., 2021). Comparison of the DC subset composition in the SP and BM of naïve mice demonstrated pDC (CD11c^dimCD317⁺) as the predominant DC population, with limited contribution of CD11^hiMHCII⁺ cDC in BM (Figures 7A,B). We confirmed impaired pDC reconstitution in the BM of mice with cGVHD in the B6 → B6D2F1 model (Figure 7C). We therefore reasoned that pDC, rather than cDC, may contribute to the local maintenance of Treg in the BM. To address this we used BDCA2-DTR transgenic mice (hereafter pDC-DTR) (Swiecki et al., 2010) to induce the specific depletion of pDC and examined the impact on Treg in the SP and BM. In naïve mice we confirmed rapid and maintained pDC depletion in SP and BM by DT administration (Figures 7D–F). On day 3 of pDC depletion, SP Treg numbers were reduced, but had rebounded by day 7. In contrast, BM-Treg showed a gradual and persistent reduction, supporting the notion that pDC preferentially contributed to BM-Treg maintenance (Figure 7G). SP Treg numbers increased to ∼ 2-fold over basal levels, indicative of niche filling activity (Pierson et al., 2013). In this regard, in the SP but not the BM, pDC depletion induced a marked increased Treg expression of the anti-apoptotic protein MCL-1, in parallel with a reduction in apoptosis, highlighting divergent homeostatic mechanisms in these tissues (Supplementary Figure 6).

To determine the impact of pDC depletion on the GVHD severity and Treg engraftment during cGVHD, we undertook transplants where BM from pDC-DTR or WT littermate donor mice were transferred with WT B6 T cells. pDC were deleted with DT administration from day 21 to 34 (Figure 7H). Following pDC depletion that was restricted to the BMgraft, we observed an increase in cGVHD severity and associated skin cGVHD pathology (Figures 7I,J). Enumeration of BMgraft and Tgraft Treg in the SP and BM on D35 after transplant revealed that while pDC depletion did not impact the frequency of Tgraft derived T cells (Figure 7K), we observed a significant reduction in BMgraft-derived Treg in both the BM and SP (Figure 7L). Taken together, these data demonstrate that pDCs within the BMgraft are critical for BM-Treg maintenance during GVHD. Thus, the absence of pDCs post-transplant further contributes to loss of Treg mediated control of GVHD leading in an increase in disease severity.