Recipient Myeloid-Derived Immunomodulatory Cells Induce PD-1 Ligand–Dependent Donor CD4⁺Foxp3⁺ Regulatory T Cell Proliferation and Donor–Recipient Immune Tolerance after Murine Nonmyeloablative Bone Marrow Transplantation

This article is featured in In This Issue, p.5323

Prevention of graft-versus-host disease (GVHD) while maintaining graft-versus-tumor (GVT) activity remains the “holy grail” of allogeneic hematopoietic cell transplantation (1, 2). To minimize transplant-associated toxicities, including GVHD, regimens of reduced-intensity conditioning have been applied successfully to prepare the transplant recipient to immunologically accept an allogeneic bone marrow graft while allowing maintenance of GVT (3). We and other investigators showed that a recipient reduced-intensity conditioning regimen using total lymphoid irradiation (TLI) and antithymocyte globulin (ATG) results in durable engraftment, profound GVHD protection in both children (4) and adults (5, 6), and maintenance of GVT in patients whose disease features rendered them at high risk for relapse (5, 6).

In the murine MHC-mismatched [C57BL/6 (H-2^b) → BALB/c (H-2^d)] preclinical model of TLI and antithymocyte serum (ATS), we and other investigators showed that GVHD protection is driven by IL-4–secreting Th2-polarized recipient invariant NKT (iNKT) cells (7–9). Specifically, recipient iNKT cells that preferentially survive TLI conditioning as a result of their relative radioresistance (7, 8, 10) secrete IL-4, which, in turn, facilitates the potent in vivo expansion of donor-type naturally occurring regulatory CD4⁺CD25⁺Foxp3⁺ T cells (nTregs) (11). nTregs expanded in vivo then regulate the donor effector CD8⁺ T cell–driven lethal acute GVHD seen when identical transplants are performed into conventional total body irradiation (TBI)-conditioned recipients. Our previous studies established that TLI/ATS results in post–bone marrow transplantation (BMT) expansion of Foxp3⁺ nTregs and not merely peripheral expansion of induced Tregs, because CD25 depletion of the graft prior to BMT was confirmed to result in loss of all expanding CD4⁺Foxp3⁺ cells at day 6 after BMT (11). Although earlier publications (12, 13) suggested that IL-4–driven STAT6 signaling could downregulate FOXP3 gene expression in induced Tregs, more recent studies (14, 15) support our findings by demonstrating that GATA3 may actually stabilize Foxp3 protein expression in nTregs. We sought to determine the specific mechanisms by which recipient iNKT cell–derived IL-4 signaling could induce nTreg proliferation in vivo after TLI/ATS and allogeneic BMT. Defining the specific mechanism by which iNKT cells and Th2-polarizing conditioning in the recipient generate donor-type nTreg proliferation in this model would lay the foundation for future conditioning strategies designed to augment nTreg maintenance and expansion in vivo after allogeneic BMT. In this study, we demonstrate that the effect of recipient IL-4 on donor nTreg expansion in vivo early after TLI/ATS and BMT is not direct; rather, it occurs via a critical recipient B220⁻CD11b⁺Gr-1^lowCD11c⁺ regulatory dendritic cell (DC) subset fitting the immune phenotype of myeloid-derived immunomodulatory cells, the maintenance and expansion of which after TLI/ATS + BMT are STAT6 and iNKT cell dependent. Donor-type nTreg proliferation occurs independent of common regulatory pathways described in other CD11b⁺Gr-1^low populations, including CD40/CD154 (CD40L), TGF-β, STAT6 signaling, arginase 1 (Arg1), or inducible NO synthase (iNOS), but it requires contact-dependent signaling through PD-1 ligands. These recipient DCs induce potent proliferation of donor-type nTregs with stable expression of Foxp3, and blockade of the PD-1 ligand axis using mAb treatment of recipients abrogates donor nTreg expansion after TLI/ATS and allogeneic BMT. To our knowledge, our studies link, for the first time, this regulatory TNF-α and iNOS-producing DC population with expansion of Foxp3⁺ nTregs both in vitro and in vivo and identify a novel means by which nonmyeloablative Th2-polarizing recipient conditioning may maintain durable donor–recipient immune tolerance after allogeneic BMT.

Wild-type (WT) (CD45.2⁺), CD45 congenic (CD45.1⁺), Arg1^flox/flox (ARG1^fl/fl), STAT6^−/− BALB/c (H-2^d), Foxp3-IRES-mRFP (FIR), and STAT6^−/− C57BL/6 (H-2^b) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). iNKT-deficient Jα18^−/− BALB/c mice were kind gifts from Dr. D. Umetsu (Harvard University, Boston, MA) (16) and were bred in our facility. Only male mice aged 8–12 wk were used for experiments (minimum starting weight 25 g for recipients of TLI). A BALB/c LysM-cre breeder mouse was a kind gift from Dr. P. Murray (St. Jude’s Children’s Research Hospital). LysM-cre × ARG1^fl/fl BALB/c mice were bred in the St. Jude Animal Resource Facility. DNTGF-βRII × Foxp3-IRES-GFP (Foxp3^GFP) C57BL/6 mice and Foxp3^GFP mice (C57BL/6) were gifts from Dr. H. Chi (St. Jude’s Children’s Research Hospital) and Dr. A. Rudensky (Memorial Sloan-Kettering Cancer Center, New York, NY), respectively. All animals were housed, monitored, and euthanized on a preapproved protocol reviewed annually by the St. Jude Institutional Animal Care and Use Committee.

TLI was delivered to the lymph nodes, thymus, and spleen with shielding of the skull, lungs, limbs, pelvis, and tail, as previously described (8–11, 17). TLI was administered in 17 doses of 240 cGy each, beginning on day −24 prior to transplantation. TBI was delivered as a single dose (myeloablative, 800 cGy; nonmyeloablative, 400 cGy) 24 h before transplantation. Irradiation was performed with a Gulmay X-ray unit (Gulmay Medical, Suwanee, GA) (300 kV, 10 mA) at a rate of 100 mU/min with a 0.75-mm Cu filter.

Rabbit ATS was purchased from Accurate Laboratories (New York, NY) and complement adsorbed with BALB/c erythrocytes prior to use. Recipient mice were injected i.p. with 0.05 ml ATS in 0.5 ml sterile normal saline on days 12, 10, and 8 before BMT (9, 10).

Animals were sacrificed at day 6 after BMT for specific studies or when moribund as per St. Jude Animal Welfare protocol guidelines. Tissue specimens were obtained at the time of sacrifice from the skin, liver, spleen, mesenteric lymph nodes (MLNs), and terminal 1 cm of descending colon measuring from the anal verge. Tissues were fixed in 10% formalin and embedded in paraffin blocks, and 4–5-μm sections were cut and stained with H&E. Microscopic images were obtained, as described previously in detail (9, 11). At the time of histopathologic analysis of H&E-stained sections, the skin, liver, and colon were assigned scores assessing the severity of GVHD, as per previously published criteria (9, 11). The evaluating pathologist was blinded to the experimental groups. The cumulative and colonic GVHD scores represent the mean ± SEM in each group of animals.

All cells were incubated with anti-CD16/CD32 (2.4G2; Becton Dickinson, San Diego, CA) prior to Ab staining to block FcRγ II/III. The following conjugated mAbs were used: FITC anti–H-2K^b (clone AF6-88.5), PE-Cy7 anti-CD4 (clone GK 1.5), allophycocyanin-Cy7 anti-CD8 (clone 53-6.7), allophycocyanin-Cy7 anti-B220 (clone RA3-6B2), PerCP-Cy5.5 anti-CD11b (clone M1/70), eFluor 450 anti–Gr-1 (clone RB6-8C5), allophycocyanin anti-CD11c (clone HL3), PE anti-CD103 (clone M290), PE anti-CD80 (clone 16-10A1), PE anti-CD86 (clone GL1), PE anti-CD124 (anti–IL-4Rα) (clone mIL-4R–MI), PE anti-CD1d (clone 1B1), PE anti-CD54 (ICAM-1; clone 3E2), PE anti-CD252 (OX40L; clone RM134L), PE anti–H-2K^d (clone SF1-1.1), FITC anti-Ly6C (clone AL21), PE anti-CD8α (clone 53-6.7), FITC anti–TNF-α (clone MP6-XT22) (all from BD Biosciences); PerCP-Cy5.5 anti-TCRαβ (clone Η57-597), eFluor 450 anti-CD25 (clone eBio307), allophycocyanin anti-Foxp3 (clone FJK-165), eFluor 450 anti-F4/80 (clone BM8), allophycocyanin anti-CD115 (clone AFS98), biotin anti–PD-L1 (clone 1-111A), biotin anti–PD-L2 (clone 122), biotin anti-PD1 (clone J43), PE anti-CD40 (clone 1C10), allophycocyanin anti-IA^d (clone AMS-32.1) (all from eBioscience); Pacific Blue anti-Helios (clone 22F6, BioLegend); rabbit anti-iNOS (clone M19; Santa Cruz Biotechnology, Santa Cruz, CA), goat anti-rabbit PE (Southern Biotech, Birmingham, AL); and LIVE/DEAD Aqua (Invitrogen, Carlsbad, CA).

At day 6 after BMT, single-cell suspensions from recipient spleen, MLNs, and mononuclear cells from the liver and colon were prepared as described previously (9, 17). Cells were stained, samples were analyzed with a BD LSR II instrument (BD Biosystems), and data were analyzed using FlowJo software (TreeStar).

Organs were harvested and single-cell suspensions were prepared on day 6, as described previously (9, 11). Cells were surface stained, fixed, permeabilized, and stained with allophycocyanin anti-mouse Foxp3 (clone FJK-16s) or allophycocyanin rat IgG2a isotype control (both from BD Biosciences). To determine the proportion of proliferating CD4⁺Foxp3⁺ nTregs, cells were counterstained with Pacific Blue anti-Helios (clone 22F6) or Pacific Blue hamster IgG isotype control (BioLegend) following fixation and permeabilization. Samples were analyzed on a BD LSR II instrument, and data were analyzed using FlowJo software.

WT BALB/c recipients were conditioned with TLI and ATS. Donor splenocytes in all experiments were labeled with Cell Proliferation dye eFluor 450 (eBioscience), per the manufacturer’s instructions, prior to infusion. On day 0, 60 × 10⁶ labeled splenocytes and 50 × 10⁶ bone marrow cells from donor mice (WT or STAT6^−/− C57BL/6) were injected per conditioned recipient. Spleens of transplanted recipients were harvested on day 6 after BMT. Spleens from two or three recipient mice/analysis were pooled, stained, analyzed using a four-laser BD LSR II, setting the eFluor 450 voltage threshold based on control C57BL/6 splenocytes labeled with Cell Proliferation dye eFluor 450 (10 μM), and fixed on day 0.

Day-6 single-cell suspensions from recipient spleen, MLNs, and liver and intraepithelial mononuclear cells from colon were isolated as described previously (9, 11). Cells were stained, samples were analyzed on a four-laser BD LSR II instrument, and data were analyzed using FlowJo software.

CD11b⁺Gr-1^highCD11c⁻, CD11b⁺Gr-1^intCD11c⁻, and CD11b⁺Gr-1^lowCD11c⁺ cells were sorted from gated H-2K^b−B220⁻ cells at day 6 after BMT from pooled spleens of TLI/ATS-conditioned WT recipients receiving WT C57BL/6 BMT. Cells were cultured in triplicate wells at a concentration of 4 × 10⁴ cells/well in a 96-well plate in RPMI 1640 (Cellgro, Manassas, VA) supplemented with 10% FBS (Atlanta Biological, Lawrenceville, GA), penicillin (100 U/ml), streptomycin (100 μg/ml), and l-glutamine (2 nM) (all from HyClone, Logan, UT), as well as 2-ME (Sigma-Aldrich, St Louis, MO). Sorted populations were stimulated with 1 μg/ml Escherichia coli LPS (L26390; Sigma-Aldrich) for 72 h. Supernatant cytokine concentrations were analyzed using the mouse Milliplex MAP (Millipore). For assays of intracellular cytokine expression by FACS, the above sorted cell populations were stimulated for 12 h with 1 μg/ml E. coli LPS, with GolgiPlug (BD Biosciences) added after 7 h of culture. Cells were fixed, permeabilized (Fixation/Permeabilization kit; eBioscience), and stained with unlabeled rabbit iNOS (clone M-19; Santa Cruz Biotechnologies), PE-conjugated anti-rabbit IgG (Southern Biotech), and FITC-conjugated TNF-α (clone MP6-XT22; BD Biosystems).

Sorted CD11b⁺ population subsets were stained for morphological assessment using Protocol Hema 3 Giemsa Stain (Fisher Healthcare, Thermo Fisher Scientific, Waltham, MA), according to the manufacturer’s protocol. Photomicrographs were acquired with a 100 X Plan APO 1.4/NA lens and a Nikon DXM 1200 camera. Images were prepared using NIS Elements AR software (Nikon Instruments, Melville, NY).

Recipient BALB/c mice were conditioned with TLI and ATS. Ab clone RB6-8C5 (18) or isotype negative control Ab (Rat IgG2b; both from Bio X Cell, West Lebanon, NH) was diluted in PBS to a final concentration of 200 μg/ml, and recipient mice were injected i.p. with 500 μl (100 μg/dose/mouse) on days 10, 8, 6, and 4 prior to BMT, with WT C57BL/6 bone marrow cells (50 × 10⁶) and spleen cells (60 × 10⁶) injected via lateral tail vein on day 0. On day 6 after BMT, recipients were euthanized; tissue specimens were harvested from the skin, liver, and terminal 1 cm of descending colon; and H&E-stained sections were scored for GVHD. The cumulative and colonic GVHD scores represent the mean ± SEM in each experimental group.

Responder splenocytes from C57BL/6 congenic (CD45.1⁺), FIR, Foxp3^GFP, or DNTGF-βRII × Foxp3-IRES-GFP C57BL/6 male mice were labeled with Cell Proliferation dye eFluor 450, per the manufacturer’s instructions. At day 6 after BMT, stimulator cells were sorted according to CD11b, Gr-1, and CD11c expression from pooled spleens of TLI/ATS-conditioned WT BALB/c hosts receiving BMT from WT C57BL/6 donors. Single-cell suspensions of splenocytes were enriched by CD11b positive selection (cat. #18770; STEMCELL Technologies, Vancouver, BC, Canada), and the following populations were sorted to >97% purity: H-2K^b−B220⁻CD11b⁺Gr-1^high, H-2K^b−B220⁻CD11b⁺Gr-1^intCD11c⁻, and H-2K^b−B220⁻CD11b⁺Gr-1^lowCD11c⁺. Sorting was performed on a four-laser FACSAria II (BD Biosystems). Responder cells (1 × 10⁵) were cultured with each sorted population (1 × 10⁵) in a U-bottom 96-well plate in 5% CO₂ at 37°C. At 72 h of coculture, cells were pooled from six wells; stained for H-2K^b, CD4, and Foxp3; and analyzed by an LSR II. Voltage threshold was defined using eFluor 450–labeled responder splenocytes fixed at day 0 of incubation, and proliferation was calculated using FlowJo software.

In specific proliferation experiments, coculture wells were treated with the Arg1 inhibitor N^ω-hydroxy-nor-l-arginine (nor-NOHA; 50 μM) or the iNOS inhibitor N^G-monomethyl-l-arginine (L-NMMA; 5 μM) (both from EMD Biosciences, Darmstadt, Germany) or with the STAT3/5-blocking Jak2 inhibitor AG490 (25 μM; Cayman Chemical, Ann Arbor, MI). All inhibitors were added at the beginning of cultures to both experimental cocultures and control wells containing only responder cells.

Responder cells were splenocyte single-cell suspensions from FIR C57BL/6 mice (1 × 10⁵/well), and stimulator cells were sorted CD11b⁺Gr1^lowCD11c⁺ cells (1 × 10⁵/well) from WT BALB/c recipient spleens at day 6 after TLI/ATS and BMT from WT C57BL/6 donors. Anti-CD40 (clone FGK45.5; Miltenyi Biotec) or anti–PD-L1 (clone 10F.9G2; BioLegend) and anti–PD-L2 (clone TY25; BioLegend)-blocking Abs were added (5 μg/ml) to sorted CD11b⁺Gr1^lowCD11c⁺ cells just prior to coculture with responders. Rat IgG2a (5 μg/ml; catalog no. 553926; BD Biosciences) was used as isotype control.

WT BALB/c mice were treated with blocking Abs against both PD ligand 1 (PD-L1; clone 10F.9G2; BioLegend) and PD ligand 2 (PD-L2; clone TY25; BioLegend) or isotype control Abs (Rat IgG2a and Rat IgG2b; BioLegend) at 200 μg/dose/mouse in 400 μl PBS injected i.p. on days −2, 0, +2, and +5. BMT consisted of WT C57BL/6 bone marrow cells (50 × 10⁶) and CD45.1 congenic splenocytes (60 × 10⁶) labeled with eFluor 450 proliferation dye (10 μM) and infused i.v. via lateral tail vein on day 0. In separate experiments, iNKT-deficient Jα18^−/− BALB/c recipients received TLI/ATS and adoptive transfer of 1 × 10⁵ CD11b⁺Gr1^lowCD11c⁺ cells sorted at day 6 from spleens of WT BALB/c recipients of TLI/ATS + BMT (details in Fig. 3E). At day 6 after BMT, animals were euthanized; spleen, liver, colon, and MLNs were isolated; and mononuclear cells were prepared for FACS analysis. Voltage threshold for analysis was set using stained eFluor 450 dye-labeled donor splenocytes fixed at day 0, and proliferation was assessed on specific gated subsets using FlowJo software version 9.4.10.

Statistical significance in survival between experimental groups was assessed using the log-rank test. Statistical significance in mean GVHD scores and mean absolute cell numbers between groups was assessed using the Mann–Whitney U test. For all tests, p < 0.05 was considered significant.

WT BALB/c recipients of nonmyeloablative TLI/ATS conditioning were protected from acute GVHD after transplantation of WT C57BL/6 bone marrow (50 × 10⁶) and splenocytes (60 × 10⁶) (BMT), whereas mice receiving 800 cGy (myeloablative) TBI + ATS (TBI 800/ATS) or control 400 cGy (nonmyeloablative) TBI + ATS (TBI 400/ATS) developed lethal acute GVHD, as demonstrated by survival curves (Supplemental Fig. 1A), histopathologic GVHD scoring at day 6 after BMT (Supplemental Fig. 1B), and mean weight curves (data not shown). The doses of bone marrow and splenocytes were the same as used in prior murine studies (9, 11), which were chosen to recapitulate the cumulative CD34⁺ hematopoietic stem cell dose and peripheral CD3⁺ T cell dose (2–3 × 10⁸/kg) administered in clinical trials using TLI/ATG conditioning in adult patients with hematologic malignancies (5, 6). The difference in both cumulative and colonic GVHD scores was highly significant between either myeloablative or submyeloablative TBI/ATS and TLI/ATS groups (p < 0.001, TLI/ATS versus TBI 800/ATS; p < 0.001, TLI/ATS versus TBI 400/ATS). Representative photomicrographs of colon sections (×200) from each group and controls are shown in Supplemental Fig. 1C.

TLI/ATS-conditioned recipients had dramatically reduced day-6 donor TCRαβ⁺CD8⁺ T cell accumulation in spleen, MLNs, and colon compared with either the TBI 800/ATS or TBI 400/ATS group (Fig. 1A, 1B).

Both the percentage among total donor CD4⁺ T cells (Fig. 1C) and the absolute number (Fig. 1D) of H-2K^b+TCRαβ⁺CD4⁺Foxp3⁺ donor Tregs recovered from recipient spleen and MLNs at day 6 were significantly increased after TLI/ATS + BMT versus after 800 cGy TBI/ATS or 400 cGy TBI/ATS conditioning. The same trend was seen for donor Treg accumulation in the recipient colon and liver at day 6 after BMT (data not shown).

Analysis of T cell subset proliferation at day 6 demonstrated no loss of Treg proliferation (Fig. 1E) and stable Foxp3 protein expression with ongoing Treg proliferation in vivo (Fig. 1F) and low donor CD8⁺ T effector cell proliferation (data not shown) when WT BALB/c recipients received TLI/ATS and BMT with 50 × 10⁶ bone marrow cells and 60 × 10⁶ eFluor 450–labeled splenocytes from STAT6^−/− C57BL/6 donors compared with BMT from WT donors. Notably, there was no increase in GVHD at day 6 after BMT from STAT6^−/− donors compared with WT donors (data not shown). The donor Tregs were confirmed to be naturally occurring thymically derived Tregs by their uniform (>90% gated H-2K^b+CD4⁺Foxp3⁺ cells) expression of Helios (Fig. 1G), a surface marker shown to differentiate thymically derived murine nTregs from Foxp3⁺ Tregs induced in the periphery (19).

Because nTregs, as well as all donor cells, were IL-4–signaling incompetent in BMT from STAT6^−/− donors, these data generated the hypothesis that recipient iNKT cell–derived IL-4 effects on donor nTreg proliferation are not direct and that an IL-4 signaling–dependent recipient cell population is responsible for inducing donor nTreg expansion in vivo after TLI/ATS + BMT.

After examining B and T lymphocytes, monocytes, macrophages, and DC subpopulations, including plasmacytoid DCs (pDCs) and myeloid DCs, in spleen, liver, MLNs, and colon, we identified three H-2K^b−B220⁻CD11b⁺ populations consistently increased at both day 0 pre-BMT (data not shown) and day 6 after BMT (Fig. 2) in TLI/ATS-conditioned recipients compared with TBI/ATS-conditioned recipients: CD11b⁺Gr-1^high, CD11b⁺Gr-1^int, and CD11b⁺Gr CD11b⁺Gr-1^low. Fig. 2A shows representative FACS plots of gated H-2K^b−B220⁻ cells in recipient spleens at day 6. Although these cells are rare in untreated WT BALB/c mice (Fig. 2A, left panel), their recovery is markedly increased both in relative fraction (Fig. 2A, right panel) and in absolute number in the spleen (Fig. 2B) in WT recipients of TLI/ATS versus 800 cGy or 400 cGy TBI/ATS + WT C57BL/6 donor BMT. Similar quantitative comparisons were obtained for MLNs, liver, and colon (Fig. 2C).

At day 6 after TLI/ATS + BMT in recipient spleen, the gated H-2K^b−B220⁻CD11b⁺Gr-1^low, but not the CD11b⁺Gr-1^high or CD11b⁺Gr-1^int, subpopulation expressed CD11c (Fig. 3A). The mean percentage (± SEM) of CD11c⁺ cells among gated CD11b⁺Gr-1^low cells was 70.0% ± 5.6% (range, 60.3–79.9%; n = 5 experiments, n = 8 mice). After CD11b enrichment by MACS selection, the CD11b⁺Gr-1^int and CD11b⁺Gr-1^low populations overlap in mean fluorescence intensity for Gr-1 expression by FACS analysis (Fig. 4B, left panel). The CD11b⁺Gr-1^low and the CD11b⁺Gr-1^int populations could be differentiated based upon surface CD11c expression (Fig. 3B, right panel), resulting in three distinguishable populations at sort: H-2K^b−B220⁻CD11b⁺Gr-1^highCD11c⁻ (Gr-1^high), H-2K^b−B220⁻CD11b⁺Gr-1^intCD11c⁻ (Gr-1^int), and H-2K^b−B220⁻CD11b⁺Gr-1^lowCD11c⁺ (Gr-1^lowCD11c⁺).

CD11b⁺ populations with variable Gr-1 expression and immunoregulatory properties have been described, including myeloid-derived suppressor cells (MDSCs) and regulatory DC subsets (20–25). We analyzed the three CD11b⁺ populations for surface markers associated with these and other regulatory myeloid cell subsets (Fig. 3C). All three populations expressed MHC class I (H-2^d), but only the Gr-1^lowCD11c⁺ population demonstrated significant MHC class II (I-A^d) expression. Surface expression of CD1d, the canonical MHC class I–like Ag-presenting ligand for iNKT cells, was increased significantly on Gr-1^lowCD11c⁺ cells compared with the Gr-1^high or Gr-1^intCD11c⁻ cells. CD80 was highly expressed on Gr-1^high cells, whereas CD86 expression was highest on the Gr-1^lowCD11c⁺ population. CD40, but not OX40L, was expressed on the Gr-1^lowCD11c⁺ population. CD40 was reported previously to mediate MDSC- and DC- induced contact-dependent expansion of Foxp3⁺ nTregs via activation of CD40L (CD154) on nTregs (26, 27). Gr-1^lowCD11c⁺ cells did not express CD8α but did express F4/80, supporting their derivation from macrophages. All three populations also expressed the M-CSF receptor (CD115) but lacked CD103 expression, indicating that these cells derive from monocyte progenitors and not lamina propria DCs (28). Ly6C-specific Ab (clone AL-21, BD Biosciences) significantly stained all three CD11b⁺ subsets. ICAM-1 (CD54), a critical adhesion molecule (29), was highly expressed on the Gr-1^intCD11c⁻ and Gr-1^lowCD11c⁺ subsets. Notably, all three CD11b⁺ subsets uniformly expressed IL-4Rα (CD124).

By light microscopy, the Gr-1^high, Gr-1^int, and Gr-1^lowCD11c⁺ populations demonstrated neutrophil, monocyte/macrophage, and DC morphology, respectively (Fig. 3D).

Serbina et al. (30) were the first to describe a population of Ly6C⁺CD11b⁺CD11c⁺ monocytes that egress from the bone marrow after Listeria monocytogenes infection and produce TNF-α and iNOS that are measurable by intracellular protein staining upon specific Ag restimulation. These cells, termed “TNF and iNOS-producing DCs” (Tip-DCs) have since been described in other settings (31). When stimulated in vitro with LPS and brefeldin A for 12 h, the Gr-1^lowCD11c⁺ cells (Fig. 3E), but not the Gr-1^high or Gr1^intCD11c⁻ cells, showed a dramatic increase in intracellular iNOS and TNF-α expression.

Gr-1^high, Gr-1^intCD11c⁻, and Gr-1^lowCD11c⁺ populations were sorted from gated H-2K^b−B220⁻ cells among pooled recipient splenocytes prepared on day 6 after TLI/ATS + WT C57BL/6 BMT. The Gr-1^intCD11c⁻ and Gr-1^lowCD11c⁺ cells were differentially sorted, as indicated in Fig. 3B. TLR4 stimulation (E. coli LPS) was provided for 72 h prior to supernatant harvest, and supernatant cytokine and chemokine levels were assessed by 22-plex Luminex assay. Gr-1^high and Gr-1^intCD11c⁻ populations secreted insignificant levels of all cytokines tested after LPS stimulation, whereas Gr-1^lowCD11c⁺ cells stimulated with LPS secreted significant amounts of IFN-γ (500 pg/ml), CCL3 (500 pg/ml), CCL5 (500 pg/ml), and TNF-α (200 pg/ml) (Fig. 3F). Notably, none of these subsets secreted IL-4, IL-10 (Fig. 3F), IL-5, or IL-13 (data not shown). Thus, none of these CD11b⁺ populations secreted Th2 cytokines, which we (9, 11) and other investigators (7, 8, 10) showed derive mainly from recipient Th2-polarized relatively radioresistant iNKT cells enriched by TLI conditioning and required for graft-versus-host tolerance after TLI/ATS + BMT.

Because prior work defined that IL-4–induced Th2 polarization and recipient iNKT cells are both critical to durable transplantation tolerance after nonmyeloablative TLI/ATS and allogeneic BMT (7–11), we investigated whether the induction of these recipient CD11b⁺ subsets is Th2 or iNKT dependent. We assessed the absolute number of cells of each induced CD11b⁺ population recovered from the colon of WT, STAT6^−/−, and Jα18^−/− BALB/c recipients of TLI/ATS + WT C57BL/6 BMT. In each mouse, the entire colon was isolated from 1 cm proximal to the anal verge to the ileocecal valve, as previously described (9, 11).

As shown in Fig. 4A, there was a significant reduction in the Gr-1^lowCD11c⁺ subset (p < 0.05) but no difference in the Gr-1^high (p = 0.5) and Gr-1^intCD11c⁻ (p = 1.0) subsets in STAT6^−/− recipients compared with WT recipients (Fig. 4A), supporting a central role for Th2 (specifically IL-4 or IL-13) signaling in generating these regulatory DCs.

As in STAT6^−/− recipients, the absolute number of Gr-1^lowCD11c⁺ cells at day 6 was dramatically reduced in iNKT-deficient Jα18^−/− BALB/c recipients of TLI/ATS + WT C57BL/6 donor BMT compared with WT BALB/c recipients (p < 0.01), indicating a requirement for recipient iNKT cells in the enhanced recovery of Gr-1^lowCD11c⁺ cells (Fig. 4B).

Fig. 4C and 4D show mean cumulative and colonic GVHD scores and representative photomicrographs of colon (×200) at day 6 after BMT for the two groups of knockout recipients of WT C57BL/6 BMT. Loss of GVHD protection was seen in both STAT6^−/− and Jα18^−/− recipients compared with WT BALB/c recipients of TLI/ATS + BMT (Fig. 4C). There was no significant difference in GVHD severity between STAT6^−/− and Jα18^−/− BALB/c recipients.

The Ab clone RB6-8C5 recognizes both Ly6C and Ly6G epitopes and has been used to deplete Ly6-expressing CD11b⁺ populations in vivo (18). Because of the lack of suitable CD11b-diphtheria toxin receptor or CD11c-diphtheria toxin receptor murine models on BALB/c background, we performed experiments using depletive Ab treatment of WT BALB/c recipients of TLI/ATS + BMT. At day 6, there was a dramatic and significant increase in histopathologic acute GVHD between recipients of RB6-8C5–depletive versus isotype-control Ab (cumulative GVHD score, p < 0.01; colon GVHD score, p = 0.01) (Fig. 5A). Fig. 5B shows representative photomicrographs of colon sections (×200) obtained at day 6.

Notably, there also was a significant (p < 0.01) reduction in the percentage of donor CD4⁺Foxp3⁺ nTregs recovered among total donor CD4⁺ T cells in the colon at day 6 in RB6-8C5–depletive Ab-treated recipients (mean ± SEM, 4.7% ± 2.2%, n = 13) versus isotype control–treated recipients (mean ± SEM, 13.4% ± 3.7%, n = 9) (Fig. 5C). Fig. 5D shows representative FACS plots of the percentage of Foxp3-expressing cells among gated H-2K^b+CD4⁺ cells from recipient spleen at day 6 in RB6-8C5 Ab–treated versus isotype Ab–treated recipients of TLI/ATS + BMT.

Fig. 5E details the adoptive-transfer strategy used to study the direct effect of Gr-1^lowCD11c⁺ cells on nTregs and effector CD8⁺ T cell recovery in GVHD target organs in iNKT-deficient recipients of TLI/ATS + BMT. At day 6 following adoptive transfer, there was a dramatic increase in the accumulation of donor CD4⁺Foxp3⁺ nTregs (p < 0.01) and a correlative significant decrease in effector CD8⁺ T cell accumulation (p < 0.05) in spleen (Fig. 5F) and colon (p < 0.05) (Fig. 5G) of Jα18^−/− BALB/c recipients of WT Gr-1^lowCD11c⁺ myeloid-derived immunomodulatory cells compared with vehicle-treated controls. Notably, there also was a robust (p < 0.01) increase in donor CD4⁺Foxp3⁺ nTregs and a concomitant reduction in donor CD8⁺ effector T cell proliferation (p < 0.01) (Fig. 5H) and calculated division index (DI) (p < 0.01) (Fig. 5I) in pooled spleens at day 6 in recipients of Gr-1^lowCD11c⁺ cells compared with recipients of vehicle control. These data confirm a mechanistic role for recipient Gr-1^lowCD11c⁺ myeloid-derived immunomodulatory cells in linking iNKT cell–secreted IL-4 and MHC-mismatched donor nTreg proliferation and subsequent donor recipient immune tolerance after TLI/ATS + BMT.

CD11b⁺ cell subsets, sorted to >98% purity based on Gr-1 and CD11c expression, were cocultured for 72 h with eFluor 450 proliferation dye–labeled FIR C57BL/6 splenocytes. Robust proliferation of gated H-2K^b+CD4⁺Foxp3⁺ splenic nTregs was observed in coculture with sorted Gr-1^lowCD11c⁺ cells (24.1% ± 4.5, n = 5) but not in cocultures of responders with Gr-1^high (4.6% ± 0.6, n = 6) or Gr-1^intCD11c⁻ cells (7.8% ± 1.5, n = 6) (n = 6 experiments) (Fig. 6A, top panels, 6B). Proliferation in coculture with Gr-1^lowCD11c⁺ cells was restricted to the CD4⁺Foxp3⁺ subset and was not seen in gated CD4⁺Foxp3⁻ effector cells (Fig. 6A, bottom panels), supporting that the induced proliferation is specific to nTregs rather than a pan–T cell stimulatory function of Gr-1^lowCD11c⁺ cells. Of note, stability of expression of Foxp3 was seen with ongoing cycles of nTreg proliferation (Fig. 6A, middle panels), suggesting that these cells may have a physiologic role in maintenance of nTregs in specific settings. Although the Gr-1^high and Gr-1^intCD11c⁻ populations did not induce nTreg proliferation in vitro compared with responders alone (Gr-1^high: p = 0.5; Gr-1^intCD11c⁻: p = 0.3), coculture with the Gr-1^lowCD11c⁺ population of cells dramatically increased nTreg proliferation from either baseline nTreg culture or control nTreg coculture with Gr-1^intCD11c⁻ cells (p = 0.02) (Fig. 6B). nTreg proliferation was abrogated when the responder and stimulator populations were separated in Transwell assays (Fig. 6B) (5.2% ± 0.7, n = 5; p = 0.03), supporting a dependence upon myeloid-derived immunomodulatory cell–nTreg direct contact for induction of nTreg proliferation. Of note, the induction of proliferation of nTregs in this setting is independent of MHC–TCR interactions, because nTregs in this assay derive from C57BL/6 (H-2^b, I-A^b) and Gr-1^lowCD11c⁺ cells from class I– and class II–mismatched BALB/c recipients (H-2^d, I-A^d). Notably, when CD45.2⁺Gr-1^lowCD11c⁺ cells were sorted at day 6 from spleens of CD45.2⁺ (WT) BALB/c recipients given TLI/ATS + BMT from syngeneic (CD45.1⁺ congenic) BALB/c donors, their capacity to induce FIR C57BL/6–derived Foxp3⁺ nTreg proliferation was maintained (data not shown), excluding the requirement for prior donor cell exposure in priming recipient Gr-1^lowCD11c⁺ cells to induce donor-type nTreg proliferation across MHC barriers.

To exclude a requirement for membrane-bound TGF-β in induction of nTreg proliferation, we cocultured Gr-1^lowCD11c⁺ cells with eFluor 450 proliferation dye–labeled responder splenocytes from Foxp3-IRES-GFP (Foxp3^GFP) C57BL/6 mice bred with homozygous transgenic C57BL/6 mice expressing a truncated TGFβ type II receptor. This receptor acts as a dominant-negative signaling receptor for TGF-β (DNTGF-bRII) and serves as a durable model by which to study TGF-β signaling dependence in murine T cell subsets (32). Notably, nTreg proliferation was maintained when the responder cells were obtained from DNTGF-bRII–transgenic mice (Fig. 6C).

Expression of the enzyme Arg1 has been defined as a major STAT6-dependent regulatory pathway in CD11b⁺Gr-1⁺ MDSCs (33) and was recently ascribed a role in allo-regulatory function and GVHD protection induced by donor-derived monocytoid MDSCs under the influence of the Th2-polarizing cytokine IL-13 (34). MDSC regulation has also been associated with activation of iNOS (21). We blocked Arg1 activity in cultures of C57BL/6 CD4⁺Foxp3⁺ nTregs and CD11b⁺Gr-1^lowCD11c⁺ cells using the Arg1 inhibitor nor-NOHA and saw no effect on coculture-induced nTreg proliferation in vitro (Fig. 6C). Treatment of cocultures with the iNOS inhibitor L-NMMA also did not alter nTreg proliferation induced by Gr-1^lowCD11c⁺ cells (p = 0.1) (Fig. 6C). Granulocytoid MDSCs were shown to suppress T cell responses via reactive oxygen species, driven by STAT3 and STAT5 signaling (21, 35). We found no alteration in nTreg proliferation induced by Gr-1^lowCD11c⁺ cells with the STAT3/STAT5/Jak2 inhibitor AG490 (35) (p = 0.1; Fig. 6C), excluding this pathway of nTreg proliferation induction by Gr-1^lowCD11c⁺ cells. Target inhibition by nor-NOHA, L-NMMA, and AG490 were confirmed by functional and phosphoflow assays (data not shown).

Gr-1^lowCD11c⁺ myeloid-derived immunomodulatory cell–mediated induction of CD4⁺Foxp3⁺ nTreg proliferation requires PD-1 ligand signaling in vitro. CD40 costimulation of CD40L on nTregs was shown to be critically important for regulatory DC- and MDSC-mediated nTreg expansion (27), and blockade of this axis augments T effector responses in tumor-bearing mice (26). Gr-1^lowCD11c⁺ cells induced potent proliferation of Foxp3⁺ nTregs in vitro, independent of CD40 blockade (p = 0.8) (Fig. 6D). PD-L1 and PD-L2 signaling are two other previously described MHC-independent, contact-dependent mechanisms by which regulatory APCs, including pDCs, can induce proliferation of nTregs (36, 37). The induction of nTreg proliferation was significantly inhibited by the addition of either PD-L1– or PD-L2–blocking Abs (p = 0.02 for both versus isotype) (Fig. 6D).

At day 6 in the spleens of WT TLI/ATS-conditioned BALB/c recipients of WT C57BL/6 BMT, gated H-2K^b−B220⁻CD11b⁺Gr-1^lowCD11c⁺ cells showed significant expression of PD-1, PD-L1, and PD-L2 (Fig. 6E). At day 6 after TLI/ATS + BMT, gated splenic H-2K^b+CD4⁺Foxp3⁺ nTregs expressed PD-1 and PD-L1 but insignificant levels of PD-L2 (data not shown).

PD-1 and its ligands (PD-L1 and PD-L2) deliver critical immunomodulatory signals to T cells, thus regulating the balance between T cell activation and immune tolerance (36). In particular, PD-L1 is expressed specifically on multiple subsets of tolerogenic DCs (22, 36–39). At day 6 after TLI/ATS and BMT, followed by treatment with blocking Abs, <2% PD-L1 and PD-L2 could be detected by counter-staining with noncross-reactive Ab clones (Fig. 6F, right panels) compared with counter-staining in isotype Ab–treated control mice (Fig. 6F, left panels), confirming specific therapeutic blockade. Concomitant with this blockade, the absolute numbers of CD4⁺Foxp3⁺ nTregs recovered in the spleen at day 6 post-BMT in TLI/ATS and blocking Ab–treated recipients were significantly lower than in isotype control Ab–treated recipients (1 × 10⁵ ± 0.03 versus 2.6 × 10⁵ ± 0.04, p = 0.02) (Fig. 6G). Significant inhibition of nTreg recovery was not seen when mice were treated with anti–PD-L1 or anti–PD-L2 alone (data not shown; n = 5 mice/experiment, n = 3 experiments), suggesting an overlap of function of these two ligands on Gr-1^lowCD11c⁺ myeloid-derived immunomodulatory cells in inducing nTreg proliferation. The cumulative data indicate that PD-L1 and PD-L2 signaling mediate recipient CD11b⁺Gr-1^lowCD11c⁺ cell–driven in vivo donor nTreg proliferation after TLI/ATS + BMT (Fig. 7).

We demonstrated previously that recipient Th2-polarized iNKT cells induced donor Foxp3⁺ nTreg cell expansion in vivo early after TLI/ATS and BMT, which results in robust protection against GVHD compared with myeloablated (TBI800/ATS) controls (11). However, the underlying mechanism by which iNKT cell–derived IL-4 could drive donor nTreg cell proliferation remained unclear, and no data were available using nonmyeloablated TBI controls. In this study, we demonstrate that both recipient iNKT cells and recipient STAT6 signaling were indispensable for the generation of a subset of recipient regulatory APCs (B220⁻CD11b⁺Gr-1^lowCD11c⁺) following nonmyeloablative TLI/ATS + BMT but not nonmyeloablative TBI/ATS + BMT (TBI400/ATS controls). These regulatory APCs induce proliferation of donor nTregs in a PD-1 ligand–dependent manner without requirement for MHC compatibility between donor and recipient. To investigate their role in the maintenance of donor nTreg expansion and GVHD protection after TLI/ATS and BMT, we depleted them in vivo and found severe acute GVHD along with a significant reduction in donor nTregs in key GVHD target organs. To confirm a cause-and-effect relationship, we adoptively transferred these cells into iNKT-deficient Jα18^−/− recipients following TLI/ATS but before BMT. We found protection from GVHD and potent induction of nTreg cell proliferation in multiple lymphoid organs of the adoptively transferred recipients compared with vehicle-treated controls.

MDSCs are known to suppress effector T cell responses (both allogeneic and syngeneic) via STAT6-dependent expression of Arg1 (21, 22, 40). Recently, Highfill et al. (34) also demonstrated that donor-type ex vivo–expanded monocytoid MDSCs, generated in vitro under the influence of IL-13, can regulate murine experimental GVHD in a C57BL/6 → BALB/c system following TBI-based conditioning, in an Arg1-dependent manner. This raised the intriguing possibility that the regulatory nTreg proliferation induced by B220⁻CD11b⁺Gr-1^lowCD11c⁺ cells might be Arg1 dependent. Using nor-NOHA inhibition of Arg1 with WT BALB/c recipient-derived B220⁻CD11b⁺Gr-1^lowCD11c⁺ cells or Arg1-deficient B220⁻CD11b⁺Gr-1^lowCD11c⁺ cells sorted from ARG1^fl/fl × LysM-cre mice (data not shown), we confirmed that B220⁻CD11b⁺Gr-1^lowCD11c⁺ cell–induced nTreg proliferation is not Arg1 dependent. Particularly of note, because these cells express iNOS upon activation, is the finding that selective iNOS inhibition with L-NMMA demonstrated that induction of nTreg proliferation by these APCs is iNOS/STAT1 independent. Cumulatively, these data demonstrate that canonical T cell regulatory pathways defined for MDSCs are not used by recipient B220⁻CD11b⁺Gr-1^lowCD11c⁺ cells in inducing donor nTreg proliferation after TLI/ATS + BMT.

The regulatory APCs that we found in our studies express F4/80, similar to IL-4/IL-13–induced alternatively activated macrophages with regulatory capacity (41). However, alternatively activated macrophages lack surface CD11c (42, 43). Although pDCs were shown to induce proliferation and/or induction of Foxp3⁺ nTregs, as well as allotolerance to vascularized cardiac allografts (44–47), recipient Gr-1^lowCD11c⁺ cells after TLI/ATS + BMT are B220⁻, which excludes them as being pDCs. A recent report describes a CD11c⁺ murine splenic DC subset capable of MHC class II–independent induction of nTreg proliferation (48). However, these regulatory splenic DCs induced nTreg proliferation in a syngeneic system, with a clear requirement for paracrine IL-2 in the maintenance of nTreg proliferation, in contrast to the PD ligand–dependent proliferation of nTregs after MHC class I and II–mismatched allogeneic BMT, which we demonstrated in this study. Consequently, based on phenotypic assessment from our studies and those in the literature, recipient B220⁻CD11b⁺Gr-1^lowCD11c⁺ cells that expanded after TLI/ATS + BMT appear to represent a regulatory monocytoid DC population expressing both CD11b and Gr-1 and secreting TNF-α and iNOS upon activation. This most closely matches the immunophenotype of Tip-DCs (30, 31). However, there has been significant debate as to whether Tip-DCs represent a distinct DC subset or a myeloid-derived subset with TNF-α and iNOS secretion activated under specific conditions of infectious restimulation. Moreover, the cell population that we describe is clearly Gr-1^low and Ly6C isoform expressing, suggesting a divergence from previously described Tip-DCs. Because these cells express iNOS and TNF-α in vitro upon stimulation with the TLR4 agonist LPS (like myeloid immunomodulatory populations, including MDSCs) but also express CD11c (like Tip-DCs and other immunomodulatory DC subsets) we opted to refer to them in this article as myeloid-derived immunomodulatory cells.

To our knowledge, these are the first data suggesting that myeloid-derived immunomodulatory cells can be generated under the influence of iNKT cells and Th2-polarizing (TLI), but not Th1-polarizing (TBI), nonmyeloablative conditioning in a STAT6-dependent manner, as well as the first description of myeloid-derived cells of the described immunophenotype inducing either nTreg proliferation or immune tolerance in allotransplantation. The possibility that these B220⁻CD11b⁺Gr-1^lowCD11c⁺ cells represent immunomodulatory DC precursors generated from bone marrow myeloid precursors spared in the setting of nonmyeloablative lymphoid radiation, the mechanisms of their generation, and why this population is not enriched following nonmyeloablative TBI are areas of active investigation.

In summary, B220⁻CD11b⁺Gr-1^lowCD11c⁺ myeloid-derived immunomodulatory cells are regulatory DCs that expanded preferentially after nonmyeloablative, Th2-polarizing TLI/ATS conditioning, but not after non-Th2–polarizing conditioning, including nonmyeloablative or myeloablative TBI/ATS. We propose that the Th2- and NKT-dependent preservation and enrichment of myeloid-derived immunomodulatory cells after TLI/ATS conditioning explain how recipient Th2-polarizing NKT cells induce the in vivo expansion of donor-type nTregs, which we previously showed regulates CD8-mediated GVHD and maintains long-term donor–recipient immune tolerance after TLI/ATS + BMT (Fig. 7). These data shed critical light on the donor–recipient immunoregulatory networks critical to enhanced development of durable donor–recipient immune tolerance after nonmyeloablative TLI/ATS conditioning and BMT compared with BMT following TBI-based regimens.

Ongoing studies include characterization of molecular mechanisms by which regulatory myeloid-derived immunomodulatory cells may differentially expand under the influence of iNKT cells and Th2-polarizing cytokines, how PD-1/PD-ligand interaction drives Foxp3⁺ nTreg proliferation in the MHC-mismatched setting, and determination of whether B220⁻CD11b⁺Gr-1^lowCD11c⁺ cells play a role in tumor-associated immune suppression via induction of proliferation of autologous nTregs. These studies are expected to delineate targetable pathways of innate immune cell–driven regulation that can be either inhibited to augment cancer immunotherapy or augmented to optimize allogeneic transplant tolerance.

We thank J. Houston, L. He, S. Schwemberger, S. Perry, and R. Ashmun for cell sorting, the Animal Resource Center staff, Drs. V. Frohlich and J. Peters (all from St. Jude Children's Research Hospital) for image acquisition, D. Umetsu (Harvard University) for Jα18^−/− breeders, P. Murray for a LysM-cre breeder and primer information, and H. Chi (both from St. Jude Children's Research Hospital) and R. Flavell (Yale University, New Haven, CT) for Foxp3^GFP × DNTGF-βRII breeders. We thank M. Sommers, S. Woolard, V. Morales-Tirado, and W. Luszczek (all from St. Jude Children's Research Hospital) for technical assistance, and Drs. E. Pamer (Memorial Sloan-Kettering Cancer Center), N. Chao (Duke University, Durham, NC), and D. Zeng (City of Hope Research Center, Duarte, CA) for constructive critiques of the manuscript.

The authors have no financial conflicts of interest.