Human gut microbiota-reactive DP8α Tregs prevent acute graft-versus-host disease in a CD73-dependent manner

Abstract

Graft-versus-host disease (GvHD) is a life-threatening complication frequently occurring following allogeneic hematopoietic stem cell transplantation (allo-HSCT). Since gut microbiota and regulatory T cells (Tregs) are believed to play roles in GvHD prevention, we investigated whether DP8α Tregs, which we have previously described to harbor a T cell receptor specificity for the gut commensal Faecalibacterium prausnitzii, could protect against GvHD, thereby linking the microbiota and its effect on GvHD. We observed a decrease in CD73+ DP8α Treg frequency in allo-HSCT patients 1 month after transplantation, which was associated with acute GvHD (aGvHD) development at 1 month after transplantation, as compared with aGvHD-free patients, without being correlated to hematological disease relapse. Importantly, CD73 activity was shown to be critical for DP8α Treg suppressive function. Moreover, the frequency of host-reactive DP8α Tregs was also lower in aGvHD patients, as compared with aGvHD-free patients, which could embody a protective mechanism responsible for the maintenance of this cell subset in GvHD-free patients. We also showed that human DP8α Tregs protected mice against xenogeneic GvHD through limiting deleterious inflammation and preserving gut integrity. Altogether, these results demonstrated that human DP8α Tregs mediate aGvHD prevention in a CD73-dependent manner, likely through host reactivity, advocating for the use of these cells for the development of innovative therapeutic strategies to preclude aGvHD-related inflammation.

Keywords: Immunology; Immunotherapy; T cells; Translation; Transplantation.

Figure 1. CD73 expression and, to a lower extent, DP8α Treg frequency, is specifically decreased in aGvHD patients 1 month after transplantation.

(A) The gating strategy to study DP8α Tregs is shown. (B and G) Blood samples from healthy donors (HD) and from patients with hematological malignancies were collected before receiving allo-HSCT (pre-Tx, d–7) and at 1 month after transplantation (post-Tx). Samples were analyzed by flow cytometry to assess CD3⁺, CD4⁺, CD8α^lo, CCR6⁺, and CXCR6⁺ DP8α Treg–related characteristics, including their CD73 expression pattern shown here at indicated time points. All patients tested (B) and representative examples at d30 (C) are shown. One-way ANOVA (Kruskal-Wallis test) followed by Dunn’s multiple-comparison test to obtain adjusted P values was used. (D) DP8α Treg frequency among total CD3⁺ T cells is shown at indicated time points. (E) CD39 expression on DP8α Tregs from allo-HSCT patients is shown at indicated time points. (F) CD73 expression, on d30 after transplantation, by indicated T cell subset: total single-positive (SP) CD4⁺ T cells, total SP CD8⁺ T cells, and DP8α non-Treg cells, annotated DP8α CCR-neg (i.e., DP8α cells comprising CCR6^–CXCR6^–, CCR6⁺CXCR6^–, and CCR6^–CXCR6⁺ fractions) is shown. Mann-Whitney tests were performed for single comparisons. (G) Cell frequencies of indicated subsets on d30 after transplantation in aGvHD patients are shown. Mann-Whitney tests were performed for single comparisons.

Figure 2. CD73⁺ DP8α Treg abundance is associated with aGvHD occurrence and severity, but not with cGvHD.

(A) CD73⁺ DP8α Treg frequencies are shown in HD and allo-HSCT patients, both before and after transplantation in aGvHD and aGvHD-free patients. One-way ANOVA (Kruskal-Wallis tests) followed by Dunn’s multiple-comparison test to obtain adjusted P values was used. Results are represented as mean ± SEM. (B) Cumulative aGvHD incidence over time was plotted for low versus high CD73⁺ DP8α Treg abundance. Cutoff was determined using the median of CD73⁺ DP8α Treg frequency among total T cells (= 0.0050%) from all patients. The Fine-Gray method, with relapse or death as competing risks, was used (shaded area: 95% CI). (C) Cumulative aGvHD incidence over time was plotted for low versus high CD73⁺ DP8α Treg abundance in the aGvHD-positive subgroup, using the median of this group of patients (= 0.00115%) as a cutoff. Log-rank (Mantel-Cox) test was used; shaded area is 95% CI. (D) Absolute numbers of CD73⁺ DP8α Tregs in 30 mL samples in aGvHD versus aGvHD-free patients were calculated using CBC clinical data. Mann-Whitney test was used to compare both groups (mean ± SEM). (E) CD73⁺ DP8α Treg frequencies are shown in HD and allo-HSCT patients, in cGvHD and cGvHD-free patients (1-way ANOVA [Kruskal-Wallis test] followed by Dunn’s multiple-comparison test, mean ± SEM). (F) Cumulative cGvHD incidence, over approximatively 1500 days (corresponding to the 52-week median follow-up for this cohort, calculated using reverse Kaplan-Meier), was plotted for low versus high CD73⁺ DP8α Treg abundance in all patients (same cutoff as above, Fine-Gray test with relapse or death as competing risks; shaded area is 95% CI). (G) Cumulative cGvHD incidence was plotted for low versus high CD73⁺ DP8α Treg abundance in the cGvHD-positive subgroup, using the median of this group of patients (= 0.0045%) as a cutoff (Fine-Gray test with relapse or death as competing risks; shaded area is 95% CI). (H) CD73⁺ DP8α Treg frequencies are shown in indicated patients’ groups (1-way ANOVA). (I) Absolute numbers of CD73⁺ DP8α Tregs in 30 mL samples in cGvHD versus aGvHD-free patients. Mann-Whitney test was used to compare both groups (mean ± SEM). (J) CD73⁺ DP8α Treg frequencies are shown in HD and allo-HSCT patients, in patients who died of GvHD or not (1-way ANOVA [Kruskal-Wallis test] followed by Dunn’s multiple-comparison test, mean ± SEM). (K) Cumulative death by aGvHD incidence was plotted for low versus high CD73⁺ DP8α Treg abundance in all patients (same cutoff as above, Fine-Gray test with death unrelated to aGVHD as a competing risk). (L) Absolute numbers of CD73⁺ DP8α Tregs in 30 mL samples in patients who died of aGvHD or not. Mann-Whitney test was used to compare both groups (mean ± SEM).

Figure 3. CD73⁺ DP8α Treg frequency does not significantly affect disease relapse.

(A) CD73⁺ DP8α Treg frequencies are shown in HD and allo-HSCT patients, in patients who relapsed or not (1-way ANOVA [Kruskal-Wallis test] followed by Dunn’s multiple-comparison test, mean ± SEM). (B) Cumulative relapse incidence was plotted for low versus high CD73⁺ DP8α Treg abundance in all patients. Cutoff was determined using the median of CD73⁺ DP8α Treg frequency among total T cells (= 0.0050%) from all patients. Fine-Gray test with NRM as a competing risk was used (shaded area: 95% CI). (C) Cumulative relapse incidence was plotted for low versus high CD73⁺ DP8α Treg abundance in the relapse subgroup, using the median of this group of patients (= 0.0152%) as a cutoff (Fine-Gray test with NRM as a competing risk was used; shaded area is 95% CI). (D) Absolute numbers of CD73⁺ DP8α Tregs in patients who relapsed or not were calculated using CBC clinical data. Mann-Whitney test was used to compare both groups (mean ± SEM).

Figure 4. Human DP8α Tregs protect against xeno-GvHD in vivo.

(A) CD39 and CD73 expression by the DP8α Treg clone used in vivo. (B–D) NSG mice were irradiated at 1.5 Gy at least 6 hours prior to being i.v. injected with 10 million freshly purified PBMCs from healthy individuals to induce xeno-GvHD (red). Another group of mice was also i.v. injected with 30 million DP8α Treg clonal cells (CD3/CD28-activated 48 hours prior infusion) on d0, d7, d14, d21, and d28 (green). Experiments were repeated using PBMCs from 4 different healthy volunteers with 3 mice per group. (B) Schematic describing experiment setup created with BioRender. (C and D) Mice were weighed almost daily, starting on d7. Mice from the PBMC (C) and PBMC+DP8α (D) groups had to be sacrificed (†) when they lost 20% of their initial weight. (E) Mice survival is shown. Data were analyzed with a log-rank (Mantel-Cox) test. (F) Human chimerism was assessed weekly, as described in the Methods section, in the blood of each mouse injected i.v. on d0 with healthy donors’ PBMCs (●, HD1; ×, HD2; Δ, HD3; □, HD4). Wilcoxon’s matched-pairs log-rank test was used.

Figure 5. Human DP8α Tregs protect against systemic GvHD-related inflammation in vivo.

Clotted blood and organs were harvested at sacrifice. (A–C) Sera were tested for human IL-10 (A), human IL-6 (B), and human TNF-α (C) cytokines by ELISA. For mice that received either PBMCs only or PBMCs+DP8α Tregs, clotted blood was collected at sacrifice when their weight loss reached 20% of their initial weight, i.e., on day 21.9 ± 2.03 and 34.4 ± 0.61, respectively. The other 2 control groups (naive mice that did not receive any human cells and mice that only received activated DP8α Tregs) were all sacrificed on d14. Results are represented as mean ± SEM. (D–G) Organs were mashed and filtered before red blood cells were lysed as described in the Methods section. Human chimerism was assessed by flow cytometry in the colon (D), liver (E), lungs (F), bone marrow (G), and spleen (H) of all animals. One-way ANOVA (Kruskal-Wallis test) followed by Dunn’s multiple-comparison test to obtain adjusted P values was used.

Figure 6. Human DP8α Tregs protect against colonic xeno-GvHD–related inflammation in vivo.

Colons were harvested at sacrifice and measured. Colonic segments were split into 2 for mechanical dissociation and subsequent flow cytometry analyses on the one hand, or histological analyses on the other hand. (A) Colon length measured at sacrifice. (B) Hematoxylin/phloxine/saffron (HPS) staining performed on 5-μm cryostat section of colon for conditioned-control mice and both treated groups. Vessels are also highlighted with arrows. (C) Crypt height was measured on 5 representative colonic regions from each mouse on sections harboring correctly oriented villi. (D) Apoptosis detection was assessed by immunohistochemistry labeling of activated caspase-3. Representative images are shown for each group of mice. (E) Vessel diameter measurements (5 measurements per animal), as highlighted in C. (F) Alcian blue staining obtained for each group of mice. Vessels are also highlighted with arrows. (G) Blue-stained areas, corresponding to secreted mucus and goblet cells producing mucus, and total crypt areas were detected using automatic structure recognition by deep learning in QuPath software. Data are expressed as the percentage of mucus area among the total crypt area from 4 representative regions from each mouse. (H) Immunostainings of human CD4 or CD8α for both groups. (I) Percentages of positive cells for CD4 or CD8α expression among the total number of cells in colonic mucosa. The percentage of positive cells was determined using the positive cell detection function of QuPath from 5 regions of interest for each mouse. Scale bars: 100 μm.

Figure 7. CD73 drives DP8α Treg suppressive activity.

(A and B) Eight different DP8α Treg clones, whose CD73 expression is shown for a representative clone (Figure 2A), were all separately cocultured with sorted and VPD-stained CD4⁺ T cells derived from 4 different healthy donors (1:1 ratio) in the presence or absence of CD73 inhibitors (adenosine 5′-[α,β-methylene]diphosphate sodium salt at 2 mM, 20 mM, and 200 mM; PBS 12379 at 10 nM, 100 nM, and 1 mM; AB-680 at 5 nM, 50 nM, and 500 nM; blocking anti-CD73 antibody at 2 mM, 5 mM, and 20 mM). Proliferation was measured 5–6 days later as the percentage of VPD^lo CD4⁺ T cells. (A) A representative example for the coculture of a DP8α Treg clone with CD4⁺ T cells from 1 donor, with or without the CD73 inhibitor AB-680 at indicated concentrations, is shown. (B) The entire data set from this experiment is presented. One-way ANOVA with post hoc Dunnett’s multiple-comparison test was used to compare indicated conditions to the “no-treatment” data, corresponding to DP8α Treg inhibition of CD4⁺ T cell proliferation in the absence of any inhibitor. Results are represented as mean ± SEM.

Figure 8. Regulatory potential and host reactivity of donor-derived DP8α Tregs tend to discriminate patients who will develop aGvHD from those who will not.

(A and B) Circulating DP8α Treg frequency among indicated T cell subsets (A) and their CD73 expression (B) are shown for allograft donors before inducing mobilization with granulocyte colony–stimulating factor (G-CSF). (C and D) Circulating DP8α Treg frequency among indicated T cell subsets (C) and their CD73 expression (D) are shown on mobilized blood donors and thus corresponds to the actual grafted cells. (E) CD73⁺ DP8α Treg frequencies in bone marrow samples before transplantation. (F and G) CD4⁺ T cells, comprising DP8α Tregs, derived from HSCs’ donors were magnetically sorted and stained with 1 mM VPD before being cocultured in the presence of low-dose IL-2 (20 IU/mL) with either patient-derived magnetically sorted monocytes (ratio 1:1) (obtained before transplantation) previously loaded overnight or not with F. prausnitzii (ratio 1 monocyte:5 bacteria) or with magnetically sorted monocytes from the corresponding donors (ratio 1:1). Five days later, T cell proliferation of gated DP8α cells was measured through VDP dilution assessment by flow cytometry. Patients developing aGvHD or not (G; red and green, respectively) are shown. Results are represented as mean ± SEM. SP, single-positive. (H and I) Cumulative aGvHD incidence was plotted for low versus high allogeneic host-reactive DP8α cells (H) or F. prausnitzii–reactive DP8α cells (I), in all patients. Cutoff was determined using the median of host-reactive cells (= 9.5%, H) or F. prausnitzii–reactive cells (= 14.8%, I) among total DP8α cells. Log-rank (Mantel-Cox) test was used.