Female Balb/c mice were obtained from Janvier Laboratories (Le Genest Saint Isle, France) and maintained under acidified water upon arrival. Donor cells were from specific pathogen free (SPF) C57BL/6J mice (from Janvier laboratories), congenic CD45.1⁺ C57BL/6J, Actin-GFP knock-in (KI) C57BL/6J, IFN-γ deficient C57BL/6J mice, all raised in our accredited animal facility at Institut Necker Enfants Malades under pathogen-free conditions. All mice were backcrossed for at least ten generations.

Balb/c mice (female, 10 wk-old) were irradiated at 5.8 Gy in a Faxitron X-Ray irradiator at day 0 and reconstituted at day+1 by i.v. retro-orbital injection with 5 x 10⁶ T- and B-cell-depleted BM cells as well as 1 x 10⁶ splenocytes from C57BL/6J donors. Clinical evolution of GVHD was scored over 60-80 days, for survival, diarrhea, weight, posture, mobility and skin damage (18).

Donor T- and B-cell-depleted (TBCD) BM cells were isolated by flushing femurs and tibias from donor mice with RPMI 1640. After centrifugation, cells were stained for 15 min with anti-CD3-PE and anti-CD19-PE in PBS, 2% FCS and rat anti-mouse IgG and sheep anti-mouse IgM were added. Depletion was completed with anti-rat and anti-sheep beads, respectively (ThermoFisher Scientific) after 3 passages over a magnet in 5ml tubes. The TBCD-BM fraction contains mainly myeloid, precursor and stem cells.

CpG-proB cells were isolated from C57BL/6J BM cell cultures activated with 1 μM CpG-1668 (CpG-B) (Eurogentec, Angers, France) for 17h in low endotoxin-RPMI medium (Fisher Scientific, Illkirch, France) supplemented with 10% (vol/vol) FCS and 1% antibiotics (penicillin and streptomycin). c-kit⁺ cells were magnetically sorted using the Robosep automaton (StemCell Technologies, Grenoble, France) and thereafter stained with appropriately labeled mAbs and sorted by flow cytometry on a BD FACS Aria IIIu cell-sorter as c-kit⁺Sca-1⁺B220⁺PDCA-1^-IgM^- cells. Electronically sorted B-cell progenitors were cultured on plates at 20,000 cells/ml over OP-9 stromal cells in OPTIMEM medium (Gibco) supplemented with 10% FCS, 1% antibiotics, 0.1% β-mercaptoethanol and 20 ng/ml Flt3L, SCF (Immunotools, Frisoythe, Germany) and IL-7 (Peprotech France, Neuilly-sur-Seine, France), achieving on average a 10-fold expansion of sorted CpG-proBs over 6 days. Expanded CpG-proBs were further stained and electronically sorted as c-kit^low/- Sca-1⁺B220⁺PDCA-1^-IgM^- cells, routinely assessed as >95% pure, before i.v. injection through the retro-orbital sinus.

Inguinal (n=2), axillary and brachial (n=4), cervical (n=2) and mesenteric (n=3) lymph nodes were collected from GVHD controls and CpG-proB recipients, yielding equivalent cell counts in both groups. Skin samples were harvested and digested in RPMI medium (Fisher Scientific, Illkirch, France) supplemented with 1% (vol/vol) FCS, 1% antibiotics (penicillin and streptomycin), 1mg/mL collagenase D (Roche, Sigma COLLD-RO) and 1,000 IU DNase (Sigma-Aldrich, Fleury-Mérogis, France) for 45 min at 37°C [adapted from (19)].

To block nonspecific Fc receptor binding, cells were pre-incubated for 10 min at room temperature with FcR blocker 2.4G2 mAb. Cells were then stained with appropriately labeled mAbs against CD4, B220, MHC II, PDCA-1, PDL-1, PDL-2, CD21, IgM, CD93, CD23, CD11b, F4/80 (eBioscience, ThermoFisher Scientific, Montigny-le-Bretonneux, France), c-Kit (CD117) (BioLegend, San Diego, CA), Sca-1 (anti-Ly6A/E), CD40, CD80, CD86, CD11c, CD8 (BD Bioscience/Pharmingen, Le Pont-de-Claix, France), CXCR5 (Sony, Weybridge, Surrey, UK) or GFP (ThermoFisher Scientific). Nuclear Foxp3 expression was measured by FACS analysis as per the manufacturer’s instructions (eBioscience, ThermoFisher Scientific). Positive cells were defined using an isotype control antibody. Intra-cytoplasmic cytokine expression was assessed after a 4-h stimulation with PMA (10 ng/ml) plus ionomycin (500 ng/ml) in the presence of Brefeldin A (2 mg/ml), followed by fixation/permeabilization with PFA/saponin and subsequent staining with specific antibodies including PE-labeled anti-TGF-β, PE-labeled anti-IL-27p28, PE-labeled anti-GM-CSF, APC-labeled anti-IL-10, APC-labeled anti–IFN-γ, APC-labelled anti-IL-21 (eBioscience), APC-labeled anti-IL-17 (BD bioscience), FITC-labeled anti-IL-6, PE-labeled anti-IL-13, APC-labeled anti-IL-4 (Sony) and FITC-labeled anti-TNF-α (Biolegend). Positive cells were defined using isotype Ab-stained controls (BD Biosciences and eBioscience). Membrane and intracellular antigen expression was analyzed in a FACS Canto II cytometer (BD Biosciences) using FlowJo software (Treestar, Ashland, OR).

Skin samples (2cm²) were collected from the back of GVHD controls or CpG-proB recipients at day+70, frozen in liquid nitrogen and stored at -80°C. Frozen tissues were then placed in Qiagen lysis buffer and dissociated using GentleMACS dissociator (Miltenyi Biotec, Paris, France). RNA was extracted with RNeasy Plus Universal mini-Kit (Qiagen, Courtaboeuf, France) following the manufacturer’s instructions. The A260/A280 values of all RNA samples ranged from 2.06-2.1. Production of cDNA from 1ng of total extracted RNA was performed using random primers (Invitrogen, ThermoFisher Scientific, Montigny-le Bretonneux, France) and reverse transcriptase superscript II (Life Technologies, Villebon-sur-Yvette, France). qRT-PCR array for measuring the expression of 80 genes of interest (and 8 house-keeping genes), targeting cytokines and fibrosis-related genes, was performed on a custom-made plate (Anygenes, Paris, France) with SYBRGreen, using a qTower2 thermal cycler (Analytic Jena, Jena, Germany). See Supplementary Table 1 for information on primers used in the qRT-PCR array.

Analysis was performed with Qlucore software (Lund, Sweden). Results are expressed as 2-(delta delta Ct) and gene expression was normalized using the geometrical mean of 6 housekeeping genes. The threshold for the selection of differentially expressed genes was an expression fold-change ≥1.4 and a p ≤ 0.05.

Skin sections (4 μm thick) recovered from the back of mice at day+70 were fixed in 4% paraformaldehyde, embedded in paraffin and stained with H&E. Epidermal thickness was measured on scanned images with NDP.view software (Hamamatsu City, Japan).

Statistical analysis was performed using GraphPad Prism (GraphPad Software, La Jolla, CA). Normality and variance equality were assessed for every data set with Shapiro-Wilk test (for samples with n>5) or D’Agostino-Pearson (for samples with n ≤ 5) and F Test respectively. Survival curves were analyzed with Kaplan-Meier estimates. Disease curves and multiple cytokine production were analyzed using a two-way ANOVA test, with Bonferroni multiple comparison post-test. Cell proportions were analyzed using two-way ANOVA with Bonferroni multiple comparison, Student’s t-test or one-way ANOVA. Data are shown as mean ± SEM. P ≤ 0.05 was considered statistically significant.

After induction, GVHD went through an initial phase accompanied by diarrhea between day+2 and day+18 followed by a chronic stage from day+20 onwards, characterized by a second bout of diarrhea together with cutaneous manifestations. CpG-proBs were sorted as c-kit⁺Sca-1⁺B220^lowPDCA-1^-IgM^- cells, as reported before (16) (Supplementary Figure 1A). A dose of 10⁵ CpG-proBs, previously shown to be effective in autoimmune settings, did not significantly reduce the severity of GVHD, when the adoptive transfer took place the day following reconstitution (Supplementary Figure 1B). To increase the amount of progenitors available for transfer, CpG-proBs were co-cultured with OP-9 stromal cells for 6 days. After a 10-fold expansion, on average, these progenitors were electronically sorted to routinely >95% purity. They shared a similar phenotype with CpG-proBs that had not been expanded, except for the loss of c-kit expression, presumably resulting from the presence of its ligand SCF in the expansion medium (Supplementary Figure 1C). When 7.5 x 10⁵ CpG-proBs per recipient were injected on day+2 post-irradiation (DPI), they provided significant protection, as assessed by reduced diarrhea and less skin damage but no significant increase in survival compared to controls with GVHD (Figure 1). By contrast, the same number of non-activated pro-B cell progenitors freshly sorted from the bone marrow as c-kit⁺Sca-1^-B220⁺CD24^hiCD43^hi cells (Supplementary Figure 1C) and expanded in the same conditions had no such effect (Figure 1). The same number of CpG-proBs adoptively transferred on day+9 conserved a reduced but still significant protection against disease symptoms, which was lost when injected on day+23 (Figure 1).

CpG induced a strong upregulation of MHC class II, together with the co-stimulatory molecule CD80, as well as high CD40 expression on proB cell progenitors, thereby improving their capacity to interact with T-cells. There was no significant difference between CpG-proBs and their unstimulated counterpart, in terms of FasL expression, while PDL-1 was upregulated, compared with unstimulated controls, which did not display this molecule at significant levels (Supplementary Figure 2A). However, the difference between CpG-proBs and proBs became less pronounced after expansion on the OP-9 cell layer. Finally, FACS analysis of PMA+ionomycin-activated proBs and expanded CpG-proBs revealed no significant difference between their cytokine expression profiles (GM-CSF, TNF-α, IL-10 and IFN-γ) (Supplementary Figure 2B).

We took advantage of CpG-proBs derived from actin-GFP-knock-in (KI) mice to track their migration in recipients. On day+15, B220⁺GFP⁺ cells, gated as in Figure 2A, represented 20-30% of all B cells analyzed and were detected exclusively in CpG-proB recipients, in mesenteric (mLN) and peripheral lymph nodes (pLN) as well as in the skin (Figure 2B). Using a gating strategy based on relative expression of IgM, CD21, CD23 and CD93 (20–22) in all tissues examined, approximately 40% B220⁺GFP⁺ cells displayed a CD21^lowCD23⁺CD93^-IgM⁺ phenotype (Figures 2C, D), similar to follicular B (FoB) cells, previously identified as the major CpG-proB progeny in NOD mice (15).

Twenty to 80% B220⁺GFP⁺ cells expressed various cytokines, including IL-10, TGF-β, IFN-γ, GM-CSF, TNF-α and IL-27, compared with only 10-25% positive cells among the non-CpG-proB-derived B220⁺GFP^- population. These observations suggest that the CpG-proB cell progeny is highly activated, especially in mLN, in which B220⁺GFP⁺ cells expressing these cytokines, notably IL-10 and TGF-β, were more frequent than in their pLN and skin counterpart (Figures 2E, F).

We investigated whether the two diarrhea phases occurring in this GVHD model, the first one between day+2 and day+18, the second one starting at day+20, concomitant with the onset of skin damage, corresponded to an initial acute inflammatory cytokine storm followed by a chronic phase characterized by a more systemic autoimmune disease associated with the alteration of regulatory mechanisms due to the alloreactive conflict. To this aim, we investigated the CD4⁺ T-cell intracellular expression of cytokines by flow cytometry, in the mLN of control mice with GVHD at day+15 and day+25. Percentages of CD4⁺ cells (Figure 3A) expressing TNF-α, IL-6, IL-17 and to a lesser extent IL-21 were already high at day+15 while IL-6 and IL-4 were statistically reduced at day+25. IL-17 expression also tended to be reduced between day+15 and day+25 but without reaching statistical significance, while TNF-α remained highly expressed at day+25. Conversely, low levels of GM-CSF, IFN-γ, IL-13 and IL-27 expression with nearly no detectable expression of TGF-β by CD4⁺ T-cells were observed at day+15 while their expression was enhanced at day+25, with statistical significance for TGF-β and IL-27. IL-10 expression remained low and unmodified at day+15 and day+25. Such clear-cut shift in the cytokine expression pattern determining two distinct phases of the disease was even more conspicuous in a heatmap representation (Figure 3B). Therefore, the CD4⁺ T-cell response is significantly distinct at day+15 and day+25 in this GVHD model, with an initial inflammatory phase at day+15 followed by a pro-fibrotic cytokine production at day+25, characteristic of the chronic phase of GVHD. Consequently, it was interesting to evaluate the effect of the adoptive transfer of CpG-proBs at these two phases of the GVHD model, i.e. day+15 and day+25.

We analyzed the effect of adoptively transferred CpG-proBs on various recipient cell populations. On day+15, incidence and cell counts of CD4⁺ T cells or CD4⁺Foxp3⁺ Treg cells were neither significantly different from controls nor did the cytokine expression by CD4⁺ T-cells in mLN and pLN change (Supplementary Figure 3A, B). On day+25, once the chronic phase initiated, percentages of CD4⁺, CD4⁺Foxp3⁺ Treg and CD8⁺ T-cells as well as cell counts were not significantly modified (Figures 4A, B). However the proportion of CD4⁺ T cells generating cytokines, such as TNF-α, TGF-β, IL-21 and IL-13, which are critically involved in chronic GVHD (23), was significantly reduced in mLN from CpG-proB recipients (Figure 4C), while only IL-13-expressing CD4⁺ T cells were diminished in pLN (Figure 4D). No significant difference was noted for IL-10 expression in CD4⁺ cells (Figures 4C, D and Supplementary Figure 3), while it was slightly but non-significantly enhanced in mLN but not in pLN CD8⁺ T-cells (Figures 4E, F).

T follicular helper (Tfh) cells, counterbalanced by T follicular regulatory (Tfr) cells, are known to play a key role in the CD4⁺ T-B cell interaction (24). In addition, Bregs have been reported for interacting with both Tfh and Tfr subsets (25, 26). This led us to examine how CpG-proBs and their progeny affected the balance between these two populations. Tfh evaluation on day+15 disclosed no difference between GVHD controls and CpG-proB recipients (Supplementary Figure 3C). Conversely, on day+25, the ratio between CD4⁺CXCR5⁺Foxp3⁺ follicular T regulatory cells (Tfr) and CD4⁺CXCR5⁺Foxp3^- follicular helper T (Tfh) cells was markedly increased in both mLNs (Figure 5A) and pLNs (Figure 5B) of CpG-proB recipients relative to their counterpart in control mice undergoing GVHD. Moreover, the percentage of Tfh cells expressing IL-10 was increased in mLN, while Tfh cells expressing IL-21 were diminished in pLN of mice having received CpG-proBs relative to untreated GVHD controls (Figures 5C, D). Finally, percentages of CD19⁺GL7⁺CD38^low GC B cells did not differ significantly in spleen and mLN (not shown). Altogether, these data show that CpG-proBs switch the follicular T-cell compartment towards regulation, by favoring the accumulation of Tfr over Tfh cells and by promoting their production of the immunoregulatory cytokine IL-10 over IL-21.

IFN-γ plays a key role in the protective effect of CpG-proBs in autoimmune T1D (15) and EAE (16). In GVHD mice, their migrated B220⁺GFP⁺ progeny expressed IFN-γ at similar levels, whatever the target tissue (Figure 2F), which prompted us to evaluate its role in the GVHD model. Using CpG-proBs isolated from IFN-γ-deficient mice, we found that graft recipients displayed exacerbated diarrhea and skin damage, compared with those having received WT CpG-proBs (Figure 6A). This finding proved the importance of IFN-γ in the protection against GVHD by CpG-proBs. The progeny of IFN-γ deficient CpG-proBs having migrated to the mLN did not express IFN-γ as expected, but also generated less IL-10, compared to its WT counterpart (Figure 6B). Moreover, co-culturing peripheral and mesenteric lymph node cells isolated from naive mice with CpG-proBs significantly enhanced IL-10 expression in gated CD4⁺CXCR5⁺PD1⁺ Tfh cells, only when the progenitors were competent IFN-γ producers (Figure 6C and Supplementary Figure 5).

GVHD recipients of CpG-proBs developed less alopecia and skin damage (Figure 7A right) compared with GVHD controls at day+70 (Figure 7A left). Histological analysis of H&E-stained skin sections recovered on day+70 revealed 50% reduced epidermal thickness (Figure 7B), consistent with diminished skin fibrosis. In addition, hair follicles that are a target of GVHD (27) are preserved in the skin of CpG-proB recipients, correlating with the observed reduced alopecia (Figure 7A). qRT-PCR microarray expression profiles, established at day+70, of genes involved in fibrosis and cytokine production (Figure 7C) disclosed that Col3a1 (Figure 7D) as well as of Pdgfa, a Col3a1 inducer implicated in fibrosis were downregulated in samples from CpG-proB recipients. The expression of Pdgfa, a known inducer of CXCR4 (28), which attracts fibrocytes to fibrotic tissues (29, 30) was likewise reduced in the skin of CpG-proB recipients. By contrast, thrombospondin-2 (thsb2, TSP-2), an anti-angiogenic matricellular protein that improves wound healing (31) was upregulated in CpG-proB recipients. The same applied to MMP9, which behaves like a collagenase (32) and can further regulate leukocyte infiltration into inflammatory tissues (33) by inactivating a number of chemoattractants. However, neither total immune cell nor T-cell infiltration was significantly different between GVHD controls and CpG-proB recipients on day+15 or day+42 (Figure 7E). The two-fold reduction in the total cell counts infiltrated at day+42 relative to day+15 (Figure 7E) observed in both control and CpG-proB-treated groups may reflect an initial transient wave of infiltration followed by a gradual inactivation of chemoattractants ligands or receptors occurring in the second phase of the model with profibrotic events taking over, thus contributing to enhance the cutaneous score observed in Figure 1. The enhanced IL12rb expression suggested a proTh1 effect of CpG-proBs on skin infiltrates, possibly controlling the deleterious Th2-driven fibrotic process. This conclusion was in keeping with the observed decrease in IL-13 expression by CD4⁺ T-cells in the lymph nodes. Increased Stat6 expression in CpG-proB recipients (Figure 7C) was intriguing, knowing that this signal transducer can mediate skin fibrosis (34). However, this upregulation might result from increased expression of IL-33, which occurs upstream of IL-13 (35). Of note, IL-33 can substitute for IL-2 as an inducer of tissue ST2⁺ Treg expansion (36). Although the proportions of CD4⁺Foxp3⁺ Tregs and CD4⁺IL-10⁺ Tr1 cells were not significantly increased in skin infiltrates, as measured by FACS analysis (Figure 8), IL-10-expressing CD8⁺ T cells, reported for their ST-2 expression and responsiveness to IL-33 (37), markedly accumulated in the skin of CpG-proB recipients, both on day+15 and day+42, while total CD8⁺ T-cell counts and percentages remained unchanged (Figure 8).

The proportion of IL-10 producers increased also among the B220⁺PDCA-1^- B subset as early as day+15, while on day+42, both B220⁺PDCA-1^- B cells and CD11c⁺CD11b⁺ dendritic cells expressing IL-10 accumulated (Figure 8). During GVHD, macrophages stimulated by Fc immunoglobulin fragments contribute to fibrosis by releasing TGF-β. Csf1r was enhanced in the microarray analysis of skin samples from CpG-proB recipients. However, FACS analysis of the skin cell infiltrate revealed that cell counts, percentages as well as IL-10 production by F4/80⁺CD11b⁺ macrophages remained unchanged on day+42 (Supplementary Figure 4). Moreover, microarray analysis detected no significant difference between Arg and iNOS expression. In mice, CSF1R is expressed by monocytes/macrophages, but also by conventional and plasmacytoid dendritic cells. However, the observed incremental increase in conventional (Figure 8) and plasmacytoid dendritic cell percentages and IL-10 expression (Supplementary Figure 6) did not reach statistical significance. A late accumulation of csf1r⁺ cells in the skin analyzed on day+70, compared to the flow cytometry analysis performed on day+42, cannot be excluded.

Collectively, the analysis of skin samples and infiltrates revealed the histological effects of CpG-proBs resulting in reduced skin damage, including fibrosis, epidermal thickness and collagen accumulation. These findings correlated with immune tolerance evidenced by enhanced infiltration by IL-10-expressing DCs, CD8⁺ T cells and B cells. The two latter populations were first to accumulate in the skin.