CXCR4-dependent migratory differences between peripheral Tregs and Tcons preserve GVL while limiting GVHD after HSCT

Key Points

• •In hematopoietic cell transplants, infused donor Tregs prevent GVHD without impairing GVL effect, which is mediated by donor Tcons.
• •Peripheral blood Tregs express CXCR4 at lower intensity than Tcons; therefore, Tregs do not home to the BM, leaving Tcons unleashed.

Visual Abstract

Abstract

Although studies on HLA-haploidentical hematopoietic stem cell transplant (HSCT) for high-risk acute myeloid leukemia (AML) showed that regulatory T-lymphocyte (Treg) and conventional T-lymphocyte (Tcon) immunotherapy prevents graft-versus-host disease (GVHD) while exerting T-cell–dependent graft-versus-leukemia (GVL) effect, its mechanism is not well known. The present study clarifies the mechanisms underlying this finding. In a xenogenic immunotherapy model, nonobese diabetic-scidIl2rgtm mice were engrafted with human primary AML and were treated with allogeneic human peripheral blood Tregs followed by Tcons. Treg/Tcon immunotherapy cleared leukemia without causing GVHD. Peripheral blood Tregs are largely CD45RO⁺ and express low levels of CXCR4 bone marrow (BM) homing receptor. They localized in peripheral tissues (ie, liver, gut), but not in the BM. Consequently, in the BM, Tcon effector function was not downregulated and indeed Tcons killed leukemic cells. In contrast, in peripheral tissues, Tcons showed no alloreactivity suggesting that their effector function had been downregulated by the Tregs. Thus, the GVL effect without GVHD was caused by the migratory features of peripheral blood Tregs. These data were confirmed in a major histocompatibility complex–mismatched BM transplant model using H-2b mice as recipients and H-2d mice as donors. Remarkably, in patients receiving T-cell–depleted haploidentical HSCT with Treg/Tcon immunotherapy, the infused Tregs were not found in the BM in the first month after transplant. In contrast, donor Tcons homed to the BM where they exerted leukemic-cell killing. In conclusion, to our knowledge, for the first time, we show the mechanism whereby immunotherapy with Tregs/Tcons prevents GVHD while preserving GVL in murine transplant models and in patients with leukemia who underwent transplantation.

Introduction

Post-transplant relapse is still a major cause of treatment failure of T-cell–replete hematopoietic stem cell transplantation (HSCT) for high-risk acute leukemia regardless of the donor or the transplantation strategy.1, 2, 3

In this setting, alloreactive T lymphocytes in the graft contribute to eradicating residual disease by exploiting histocompatibility differences between donor and recipient.4, 5, 6 However, donor alloreactive T cells may also attack and damage host tissues. Post-transplant pharmacological immunosuppression for graft-versus-host disease (GVHD) prophylaxis may impair the graft-versus-leukemia (GVL) effect mediated by donor T cells.7, 8, 9, 10

Attempts to ensure that alloreactive T cells do not damage normal tissue while simultaneously killing leukemic cells have so far been largely unsuccessful.

Studies in animal models of mismatched transplantation showed that adoptive immunotherapy with FOXP3⁺ regulatory T lymphocytes (Tregs) and conventional T lymphocytes (Tcons) prevented GVHD while retaining a GVL effect.11, 12, 13 Edinger et al¹² found that Tregs inhibited the expansion of Tcons and, therefore, suppressed their potential to cause GVHD. However, Tcon activation and their capacity for in vivo and in vitro lysis of leukemic and lymphoma cells were not blocked. In a xenogeneic mouse model engrafted with human primary acute myeloid leukemia (AML), we found that mice treated with human Tregs and Tcons cleared leukemia without developing GVHD.¹⁴ These insights from animal models provided the conceptual basis for clinical trials in which Treg/Tcon adoptive immunotherapy was given in the context of T-cell–depleted HLA-haploidentical (haplo) HSCTs without post-transplant pharmacological immunosuppression. In patients with high-risk acute leukemia, we showed that such a strategy was associated with almost total prevention of leukemia relapse with low incidence of GVHD.^14,15

More recently, Pierini et al¹⁶ reported a 7% cumulative incidence of relapse and a 64% chronic GVHD (cGVHD) and relapse-free survival in patients with AML with minimal residual disease or active disease at the time of transplant. None of the 20 patients with adverse genetics (most of them with complex karyotypes or monosomy/deletion of chromosome 7) relapsed, and their probability of chronic relapse-free survival was 72%.

The mechanisms underlying Treg suppression of GVHD with no loss of T-cell–dependent GVL effect are still obscure. Our current studies were aimed at unveiling the underlying mechanism and were based on the following information: (1) studies in mice on the migration patterns of infused Tregs,¹⁷ and (2) the notion that human CD45RO⁺ Tregs are localized in the skin whereas CD45RA⁺ Tregs preferentially home to the bone marrow (BM) because they express the CXCR4 BM homing receptor.¹⁸

Methods

Human cells

Human samples were collected after informed consent was obtained to participate in the study and to access public data. The study was conducted according to the revised Declaration of Helsinki and was approved and registered by Umbria Regional Hospital Institutional Review Board with identification code 02/14 and public institutional review board registry number 2384/14.

Two primary human AML samples (NPM1 and FLT3 mutated) and BM samples were obtained from patients before chemotherapy or transplant and after transplant.

To infuse nonobese diabetic-scidIl2rgtm (NSG) mice with human Tregs and Tcons, we used human donor Tregs and Tcons. The human Treg/Tcon selection procedure has been previously reported.14, 15, 16 Briefly, after leukapheresis, donor CD4⁺CD25⁺ Tregs were selected by immunomagnetic depletion of CD8⁺/CD19⁺ cells and subsequent positive selection of CD25⁺ cells (using CliniMACS device; Miltenyi Biotec, Bergisch Gladbach, Germany). On average, the final Treg product was composed of 71% ± 8.5% CD4⁺CD25⁺CD127⁻FOXP3⁺ cells and 27% ± 7.6% CD127⁺ cells. To obtain donor Tcon product, lymphomononuclear cells were separated from the leukapheresis product by Ficoll gradient, and CD3⁺ T cells were calculated and infused after CD3⁺ staining. CXCR4 expression on the Tcon product was assessed using immunophenotypic analysis. Lymphocytes were first identified based on forward- and side-scatter parameters, after which Tcons were gated as CD3⁺ cells and analyzed for CXCR4 expression. The final Treg product was stained with antihuman CD3, CD4, CD25, and CXCR4 antibodies. After fixation and permeabilization, intracellular FOXP3 staining was performed and CXCR4 expression was measured by gating within CD4⁺/CD25⁺/FOXP3⁺ cells. CXCR4 expression was evaluated using fluorescence minus one control for both Tcon and Treg products. The naïve and memory phenotypes of both Tcons and Tregs were evaluated using CD45RA and CD45RO antibodies, respectively. The naïve fraction of Treg population was enriched by immunomagnetic positive selection using antihuman CD45RA beads.

Mice

Colonies of NSG mice were bred at the University of Perugia Animal House. Eight- to 12-week-old C57BL/6 and Balb/c mice were provided by Envigo (Indianapolis, IN) or Charles River (Calco, Italy) and were used as recipients or donors of major histocompatibility complex (MHC)–mismatched hematopoietic transplants, respectively. All experiments were performed in accordance with the National Ethics Approval Document for animal experimentation (no. 44/2020 PR).

Murine models of Treg/Tcon immunotherapy

In the murine transplant models, H-2b C57BL/6 mice were irradiated with 9 Gy total body irradiation administered in 2 fractions using an X-ray machine (Precision MultiRad 225, Precision X-Ray Irradiation, Madison, CT). The day after irradiation, C57BL/6 mice were infused IV with 3 × 10⁶ Tregs from Balb/c mice. Two days later, they were infused with 3 × 10⁶ Tcons (CD3⁺ T cells) and 10 × 10⁶ T-cell–depleted BM cells from donor H-2d Balb/c mice. Control mice were infused with 3 × 10⁶ Tregs or Tcons and 10 × 10⁶ T-cell–depleted BM from Balb/c mice. T-cell–depleted BM cells were collected by flushing the femur and tibia shafts of H-2d mice. Murine Tcons were obtained from donor spleen suspension after splenocyte Ficoll and CD3 immunostaining to calculate 3 × 10⁶ T CD3⁺ (Tcons) cells per mouse. Murine Tregs were purified from donor spleen by the use murine regulatory T-cell isolation kit according to the manufacturer’s instruction (Miltenyi Biotec). BM cells were T cell depleted by negative immunomagnetic selection using anti-CD5 microbeads (Miltenyi Biotec), as previously described.¹⁴ GVHD onset and severity were assessed based on various clinical parameters, including reduced cage activity, altered posture, weight loss, fur condition, and diarrhea.¹⁹ A standardized scoring system was used to categorize mice according to GVHD severity (supplemental Table 1). Animals were euthanized upon reaching mild or severe clinical end points, as defined in supplemental Table 2.

In the xenogenic transplants, NSG mice were irradiated with 1.5 Gy total body irradiation and infused IV with 10 × 10⁶ human primary AML cells. Leukemia engraftment was defined by detection of 10% to 15% BM infiltration by leukemic cells. Human leukemia engrafted mice were infused with 3 × 10⁶ human Tregs and 2 days later with 3 × 10⁶ human Tcons. Control mice were infused with leukemic cells alone or with leukemic cells and 3 × 10⁶ Tcons or with leukemic cells and 3 × 10⁶ Tregs. Tregs and Tcons used for immunotherapy were obtained from the same donor and were HLA-mismatched with human leukemia. Tcons were composed of variable CD8⁺ and CD4⁺ T-cell ratio in accordance with the healthy donor T-cell composition that we used in each experiment. In some experiments, NSG mice were infused with purified human CD45RA⁺ Tregs obtained by positive selection with antihuman-CD45RA⁺ microbeads (Miltenyi Biotec). Such cells express CXCR4 at higher intensity than CD45RO⁺ Tregs. In such experiments, a group of mice was infused with anti-CXCR4 antibody treated Tregs: CD45RA⁺ Tregs were incubated with human anti-CXCR4 antibody (500 μg for 1 × 10⁶ cells per mL; R&D Systems, Minneapolis, MN) for 1 hour at 5°C before infusion.

Each experimental group comprised 5 mice. All experiments were repeated in triplicate. Results of triplicate experiments were combined. At different time points after transplants (+7, +14, +21, +30, +60 days), recipient mice were evaluated for GVHD by observation score and histopathology. T-cell migratory properties were evaluated by flow cytometry analyses with a combination of antimouse CD25, FOXP3, CXCR4, CD62L, CD44, CD4, CD8, CD3, and H-2D^d monoclonal antibodies (eBioscience, San Diego, CA) in the BM, liver, gut, and lymph nodes. In particular, live lymphocytes were first identified based on forward- and side-scatter properties, followed by selection of CD3⁺ cells within the H-2D^d-positive population. CD4, CD25, FOXP3, and CD8 T-cell subsets were subsequently analyzed within this gated population. CXCR4 expression was evaluated using a fluorescence minus one control for both Tcon and Treg products. At different time points after human Treg/Tcon immunotherapy (+7, +14, +21, 30, +60 days), NSG mice were evaluated for GVHD using observation score and histopathology. Leukemia infiltration and T-cell homing were evaluated by flow cytometry analyses with a combination of antihuman CD45, CD25, FOXP3, CD34, CD13, CXCR4, CD45RA, CD45RO (eBioscience), CD3, CD56, CD8, CD4, CD14 (Miltenyi Biotec), CD33, CD19, CD20, and CD10 monoclonal antibodies (Beckman Coulter, Marseilles, France) in the BM, liver, and gut. Human T cells were isolated from the BM, liver, and gut of NSG mice by the use of antihuman CD4 or CD8-conjugated microbeads (Miltenyi Biotec) 7 days after infusion. Mouse T cells were isolated from the BM, liver, gut, and lymph nodes of C57BL/6 recipient mice by the use of antimouse CD4, CD8, or CD5-conjugated microbeads (Miltenyi Biotec) 7 days after infusion. T cells were then assessed for cytotoxicity in a ⁵¹chromium release assay at a 10:1 effector-to-target ratio (Microbeta, PerkinElmer, Waltham, MA). Human leukemia or phytohemagglutinin (PHA; Invitrogen, Waltham, MA) blasts or splenocyte suspension were used as targets of human T cells isolated from NSG mice. Concanavalin A (ConA; Sigma-Aldrich, St. Louis, MO) murine blasts were used as targets of T cells isolated from C57BL/6 recipient mice. Percentage of lysis in the ⁵¹chromium release assay was calculated using the following formula: % lysis = ([experimental release − spontaneous release]/[maximum release − spontaneous release]) × 100. Mixed lymphocyte reaction (MLR) was assayed by incubating donor murine T cells (2 × 10⁵ cells per well, responder) from Balb/c spleen with 20 Gy–irradiated murine recipient C57 ConA blasts (1 × 10⁵, stimulator). Purified T cells from BM or liver of transplanted mice were added in a 2:1 or 1:1 ratio with responder cells to evaluate their ability to inhibit the MLR. After 7 days, cultures were pulsed overnight with 1 μCi of ³H-thymidine per well (PerkinElmer), and DNA synthesis was determined as counts per minute using a Betacounter (Microbeta, PerkinElmer). Similarly, in the humanized model, donor human T cells were incubated for 7 days in 96-well plates (2 × 10⁵ cells per well, responder) with irradiated (20 Gy) NSG spleen mononuclear cells, allogeneic human PHA T blasts, or leukemic cells (1 × 10⁵ cells per well, autologous to PHA T cells) as stimulators in RPMI culture medium supplemented with 10% donor serum. Purified T cells from BM or liver of humanized NSG mice were added in a 2:1 or 1:1 ratio with responder cells to evaluate their ability to inhibit the MLR. MLR against autologous to responder cells was used as a negative control; in both models and in humans, it resulted in counts per minute of <700.

In vitro experiments with BM human cells

T-cell homing was assayed in BM aspirates obtained monthly after transplant by flow cytometry analyses using a combination of antihuman CD45, CD25, FOXP3, CD34, CD13, CXCR4, CD45RA, CD45RO (eBioscience), CD3, CD56, CD8, CD4, and CD14 monoclonal antibodies (Miltenyi Biotec). T cells were then assessed for cytotoxicity in a ⁵¹chromium release assay at a 10:1 effector-to-target ratio (Microbeta, PerkinElmer). Recipient leukemia or PHA blasts were used as targets of BM T cells. BM T cells were also used to assess their ability to inhibit MLR between donor T cells used as responders and recipient leukemia or PHA blasts used as stimulators.

Statistical analyses

Bars represent the mean ± standard deviation of at least 3 independent experiments. To compare 2 means ± standard deviations of quantitative variables, the Student t test was performed. P values <.05 were considered statistically significant. Data were analyzed and plotted using the GraphPad software (version 6, San Diego, CA).

Histological analyses

At different time points after infusion of human Tregs and/or Tcons, liver and BM tissue samples from NSG mice were fixed in 10% buffered formalin and embedded in paraffin; 3.5 μm tissue sections were stained with hematoxylin and eosin stain. Immunostaining for human CD3 and FOXP3 was performed with the following antibodies: anti-CD3 (clone LN10, Leica BioSystems, Newcastle, United Kingdom) and anti-FOXP3 (clone SP97, 1:100, Invitrogen). Primary antibodies were developed using a biotin-free polymeric horseradish peroxidase–linker antibody conjugate system (Bond Polymer Refine Detection, Leica BioSystems) with heat-induced epitope retrieval, using the Bond III automated immunostainer (Leica BioSystems). Hematoxylin and eosin– and immunohistochemistry-stained tissue sections were scanned using an Aperio AT2 scanner. CD3 and FOXP3 expressions were evaluated using the Aperio ScanScope system. A semiquantitative cell count was performed in previously identified hot spot areas of T-lymphocyte infiltrates. The selected areas in the liver included at least 1 portal region. A manual count of immunolabeled T cells was then performed. Results were expressed as the number of cells per square millimeter. Human BM biopsies were also evaluated for CD3 and FOXP3 expression at 20 to 25 days after Treg/Tcon adoptive immunotherapy.14, 15, 16

Results

Immunophenotypic analyses demonstrated that mouse spleen and human peripheral blood CD3⁺/CD4⁺/CD25⁺/FOXP3⁺ Tregs expressed the CXCR4 BM homing receptor at lower intensity than CD3⁺ Tcons (Figure 1). Therefore, we hypothesized that Tregs, unlike Tcons, could not home to the BM and, consequently, they could not antagonize the antileukemic effect mediated by Tcons.

Figure 1.

Human peripheral blood and murine splenic Tregs express less CXCR4 BM homing receptor compared with Tcons. (A-D) Flow cytometry gating strategy for the characterization of Tcons starting from forward and side scatter for the identification of lymphocytes (A), CD3⁺ Tcons (B), and CXCR4⁺ Tcons (C-D). (C) A representative histogram of the expression of CXCR4 in human Tcons (purple histogram) compared with control (blue histogram). (D) A representative histogram of the expression of CXCR4 in murine splenic Tcons (purple histogram) compared with control (blue histogram). (E-I) Flow cytometry gating strategy for the characterization of Tregs starting from forward and side scatter for the identification of lymphocytes (E), CD4⁺/CD25⁺ Tregs (F), and CXCR4⁺ Tregs (G-I). (G) A representative histogram of the expression of CXCR4 in human Tregs (purple histogram) compared with control (blue histogram). (H) A representative histogram of the expression of CXCR4 in murine splenic Tregs (purple histogram) compared with control (blue histogram). (I) CXCR4 expression on human CD45RA⁺ Tregs isolated by immunomagnetic positive selection. (J) Column bar graph showing the percentages of human naïve Tregs (red column) and memory Tregs (blue column) in the total peripheral blood Treg population. (K) Column bar graph showing percentages of CXCR4 expression in human and murine Tcons (red columns), Tregs (blue columns), and human naïve Tregs (green column). Data are shown as mean ± SD. FSC, Forward Scatter; SD, standard deviation; SSC, Side Scatter.

Mismatched murine BM transplant with Treg/Tcon adoptive immunotherapy

To track T-cell homing in vivo, we set up an MHC-mismatched T-cell–depleted hematopoietic transplantation model with Treg/Tcon adoptive immunotherapy using H-2b mice (C57BL/6) as recipients and H-2d mice (Balb/c) as donors (Figure 2A). Mice that had been infused with Tregs and Tcons survived without signs of GVHD (Figure 2B; supplemental Figure 1). At different time points (7, 14, 21, and 30 days) after BM transplant and Treg/Tcon immunotherapy, flow cytometry analysis of BM cells showed high numbers of donor CD3⁺/CD8⁺ cells (Figure 2C-D). As expected, such cells were unable to inhibit MLR between donor Tcons and irradiated recipient target cells (ConA blasts) and were able to lyse recipient target cells (Figure 2E-F). In the BM, CD3⁺/CD4⁺/CD25⁺/FOXP3⁺ Treg cells were <1% of T cells in the first 2 weeks. In contrast, the liver (Figure 2G-H), gut, and lymph nodes (supplemental Figure 2A-B) contained CD3⁺ T cells that were mostly CD4⁺. Sixty percent of CD4⁺ cells were FOXP3⁺; the remaining were FOXP3⁻. CD3⁺/CD4⁺ cells exhibited regulatory activity as they inhibited MLR between donor Tcons and recipient target cells (Figure 2I). We isolated the few CD3⁺/CD8⁺ T cells that were present in the liver; they were not able to lyse recipient cells (Figure 2J), presumably because their effector function had been downregulated by the CD3⁺/CD4⁺/CD25⁺/FOXP3⁺ Treg cells.

Figure 2.

In a murine MHC-mismatched transplantation model, donor Tregs did not home to the BM. (A) After 9 Gy TBI, H-2b mice received an MHC-mismatched T-cell–depleted hematopoietic transplant with Treg and/or Tcon adoptive immunotherapy from H-2d donor mice. Mice that had been infused with Tregs and Tcons survived without GVHD. (B) Flow cytometry analyses at day +7 from Tcon infusion showed very low numbers of CD3⁺/CD4⁺ T cells in the BM. (C-D) The T cells in the BM of these mice were CD4⁺ (blue column), conventional (and not regulatory; light blue column), or CD8⁺ cells (green column). (E-F) In fact, such cells did not inhibit MLR between donor Tcons and irradiated recipient ConA blasts (E) and lysed recipient ConA blasts (F). (G-H) The liver of these mice contained mostly CD3⁺/CD4⁺ T cells. Sixty percent of such cells were FOXP3⁺ (light blue column), the remaining being FOXP3⁻ (blue column). (I) CD3⁺/CD4⁺ cells exhibited regulatory activity as they inhibited the MLR. (J) CD8⁺ T cells did not lyse recipient ConA blasts. Panel B also shows that mice that had been infused with Tcons, but not Tregs, died with signs and histology that were typical of GVHD (top). (K-P) Flow cytometry analyses at day +7 from Tcon infusion showed that the BM (K,L) and liver (O,P) both contained 80% to 85% CD3⁺/CD8⁺ (green column) and 15% to 20% CD3⁺/CD4⁺ (blue column). (M-N,Q-R) As expected, such cells did not inhibit the MLR and lysed recipient ConA blasts. Panel B shows that mice that had been infused with Tregs, but not Tcons, survived without GVHD (top). (S-T) The BM contained very low numbers of T cells (such that no functional analysis could be performed). (U-X) In contrast, a few CD4⁺/FOXP3⁺ T cells (light blue column) were detected in the liver (U-V) and were indeed able to inhibit the MLR (W) and did not lyse recipient target cells (X). Dot plots are merely representative of one flow cytometric analysis. Statistical significance was defined as ∗P < .05. cpm, counts per minute; FITC, Fluorescein Isothiocyanate; PE, Phycoerythrin; TBI, total body irradiation.

Mice that had been infused with Tcons (but not Tregs) died of GVHD (Figure 2B; supplemental Figure 1). Flow cytometry analyses of the BM (Figure 2K-L), liver (Figure 2O-P), gut, and lymph node (supplemental Figure 2C-D) showed that they contained CD3⁺ T cells (80% were CD8⁺ and 20% were CD4⁺). T cells harvested from the BM and liver did not inhibit the MLR and lysed recipient target cells (Figure 2M-N,Q-R). Mice that had been infused with Tregs alone survived without GVHD (Figure 2B). In contrast to mice that had received effector T cells, the BM of mice that received Tregs contained very few T cells (Figure 2S-T). CD4⁺ T cells (80% were FOXP3⁺) were detected in the liver (Figure 2U-V), gut, and lymph nodes (supplemental Figure 2E-F). They were able to inhibit the MLR and were unable to lyse recipient target cells (Figure 2W-X).

Humanized model of AML treated with Treg/Tcon immunotherapy in NSG mice

Next, we performed experiments using 1.5 Gy–irradiated NSG mice engrafted with primary human AML and given adoptive immunotherapy with human peripheral blood Tregs and/or Tcons (Figure 3A). Unlike Tcons, peripheral blood Tregs (harvested from a leukapheresis product and used for adoptive immunotherapy in our clinical trials)14, 15, 16 were largely CD45RO⁺ and expressed low levels of the CXCR4 BM homing receptor (Figure 1G,J-K). A minority of peripheral blood Tregs (5%-25%) were CD45RA⁺ and expressed the CXCR4 BM homing receptor at an intensity similar to that of Tcons (Figure 1I-K). Mice that received human leukemic cells and Tcons and Tregs survived without evidence of leukemia in the BM or GVHD (Figure 3B-C; supplemental Figure 3). Human T cells harvested from BM were still predominantly CD8⁺, did not inhibit the MLR between human donor T cells and irradiated human leukemic cells or murine recipient cells, and displayed alloreactivity against human leukemic cells, PHA blasts (Figure 3D-G; supplemental Figure 4A-E), and murine recipient cells (splenocyte suspension, not shown), suggesting that BM Tcons had retained their alloantigen recognition against human and mouse MHC. In contrast, human T cells harvested from the gut, lymph nodes (supplemental Figure 5A-B), and liver were composed of CD8⁺ T cells (40%), CD4⁺/FOXP3⁺ T cells (40%), and CD4⁺/FOXP3⁻ T cells (20%; Figure 3H-I; supplemental Figure 4F-J). We isolated CD4⁺ T cells that displayed regulatory function, as they inhibited the MLR (Figure 3J). The isolated CD8⁺ T cells recovered from the liver displayed no alloreactivity against human leukemic cells, PHA blasts, and murine cells (Figure 3K), suggesting that their effector function had been downregulated by Tregs.

Figure 3.

In NSG mice engrafted with human leukemia and given human Treg/Tcon adoptive immunotherapy, Tregs did not home to the BM and, therefore, did not antagonize the antileukemic effect mediated by Tcons. (A) After 1.5 Gy TBI, NSG mice received human AML and human HLA-mismatched Treg and/or Tcon adoptive immunotherapy. (B) The percentages of leukemic cells detected in the BM of humanized AML mouse models at 45 to 50 days according to the different experimental groups. (C) Mice that received Treg /Tcon immunotherapy survived without leukemia and GVHD. In these mice flow cytometry analyses at day +7 from Tcon infusion showed very low numbers of CD3⁺/CD4⁺ T cells in the BM. (D-E) The T cells in the BM of these mice were CD4⁺ (blue column), conventional (and not regulatory, light blue column), or CD8⁺ (green column) cells. (F-G) In fact, such cells did not inhibit MLR between donor Tcons and irradiated leukemia (F) and/or PHA blasts (autologous to leukemia) indeed lysed allogeneic leukemia and PHA blasts (G). (H-I) The liver of these mice contained mostly CD3⁺/CD4⁺ T cells. Fifty percent of these cells were FOXP3⁺ (light blue column), the remaining were FOXP3⁻ (blue column). (J) CD3⁺/CD4⁺ cells exhibited regulatory activity as they inhibited the MLR. (K) CD8⁺ T cells did not lyse allogeneic human leukemia and PHA blasts. Panel C also shows that mice that received leukemia and Tcons (and not Tregs) cleared leukemia but died with signs and histology that were typical of GVHD (top). (L-M,P-Q) Flow cytometry analyses at day +7 from Tcon infusion showed that the BM (L,M) and liver (P,Q) both contained human T cells that were predominantly CD8⁺ (green column). (N-O,R-S) As expected, these cells did not inhibit the MLR and lysed allogeneic leukemia and PHA blasts. Panel C shows that mice that had been infused with Tregs, but not Tcons, did not develop GVHD but died of leukemia (top). (T-U) The BM contained very low numbers of T cells (such that no functional analysis could be performed). (V-Y) In contrast, a few CD4⁺/FOXP3⁺ T cells (light blue column) were detected in the liver (V,W) and were indeed able to inhibit the MLR and did not lyse allogeneic human leukemia and PHA blasts (Y). Dot plots are merely representative of one flow cytometric analysis. Statistical significance was defined as ∗P < .05.

Mice that received leukemic cells and Tcons (without Tregs) cleared leukemia but died of GVHD (Figure 3B-C; supplemental Figure 3). Human T cells harvested from their BM, liver (Figure 3L-M,P-Q; supplemental Figure 6), gut, and lymph nodes (supplemental Figure 5C-D) were predominantly CD8⁺ and, as expected, were unable to inhibit the MLR and displayed alloreactivity against HLA-mismatched human leukemic cells, human PHA blasts (Figure 3N-O,R-S), and murine target cells (not shown).

Mice treated with Tregs alone died of leukemia (50%-70% of blasts detected in the BM; Figure 3B-C). Very low numbers of T cells were found in the BM (Figure 3T-U), thus showing that human peripheral blood Tregs do not home to the BM (supplemental Figure 7A-D). In contrast, human T cells were recovered in the liver (Figure 3V-W), gut, and lymph nodes (supplemental Figure 5E-F), 60% to 70% of which displayed a CD4⁺/FOXP3⁺ phenotype; 30% to 40% of the CD4⁺ T cells were FOXP3⁻ (supplemental Figure 7E-I). The whole CD4⁺ T-cell population inhibited MLR and had no ability to kill recipient targets (Figure 3X-Y).

Histological analyses of NSG BM, liver, gut, and skin were consistent with flow cytometry data. In fact, within the first 30 days after Treg/Tcon infusion, the BM was infiltrated by high numbers of human T cells that were FOXP3⁻. In contrast, human CD3⁺/FOXP3⁺ T cells were found in the liver (and gut, not shown; Figure 4A-D). In line with our hypothesis, mice infused with Tcons alone showed infiltration of CD3⁺/FOXP3⁻ T cells in both the BM and liver (supplemental Figure 8), whereas the mice cohort receiving Tregs alone displayed CD3⁺/FOXP3⁺ T cells in liver biopsies but not in the BM (supplemental Figure 9). To formally demonstrate that the CXCR4 molecule was responsible for the different BM homing features of the various T-cell subsets, NSG mice engrafted with AML were infused with human purified CD45RA⁺ Tregs (Figure 5A). These mice died of leukemia progression despite the coinfusion of large doses of Tcons (Figure 5B; supplemental Figure 10A). The purified CD4⁺ T cells recovered from the BM displayed regulatory function in the MLR. The purified BM CD8⁺ T cells displayed no ability to kill human leukemic cells, PHA blast cells (Figure 5C-F), or murine target cells. Thus, CD45RA⁺ Tregs homed to the BM and downregulated the GVL effect exerted by alloreactive Tcons. Similar to CD45RO⁺ Tregs, CD45RA⁺ Tregs homed to the liver and were able to inhibit the MLR between donor T cells and human leukemic cells (or human PHA blast or murine recipient cells) (Figure 5G-J). When CXCR4 on CD45RA⁺ Tregs was masked with an anti-CXCR4 blocking antibody in vitro before infusion of Tregs, Tregs could not home to the BM; Tcons exerted unopposed leukemic cell killing (Figure 5K-N; supplemental Figure 10A), and mice survived without leukemia (Figure 5B). As expected, the purified CD4⁺ T cells recovered from the liver (in the abovementioned experiments) inhibited the MLR and had no ability to kill recipient targets (Figure 5O-R). Conversely, mice infused with both human naïve CD45RA⁺ Tregs and Tcons treated with the anti-CXCR4 masking antibody died of leukemia (without developing GVHD), indicating that anti-CXCR4–treated Tcons were unable to exert antileukemic activity owing to impaired homing to the BM (Figure 5B; supplemental Figure 10A). Although low numbers of Tregs and Tcons were detected in the BM (supplemental Figure 10B), Treg and Tcon homing to the liver was preserved (supplemental Figure 10C).

Figure 4.

Histological analyses of human T cells in the BM and liver of NSG mice. (A-B) Human CD3⁺ (A,C) and FOXP3⁺ (B,D) T cells in representative BM and liver, respectively of NSG mice 3 weeks after the infusion of human Tregs and Tcons. Such histological analyses showed high numbers of human CD3⁺/FOXP3⁻ effector T cells and low numbers of human CD3⁺/FOXP3⁺ regulatory T cells in the BM of NSG mice. In contrast, human FOXP3⁺ regulatory T cells were present in the liver.

Figure 5.

In NSG mice engrafted with human leukemia, CXCR4 expression plays a key role in BM homing of human naïve CD45RA⁺ regulatory T cells. (A) NSG mice engrafted with human leukemia received adaptive immunotherapy with human naïve CD45RA⁺ Tregs and Tcons. (B) CD45RA⁺ Tregs expressed high levels of the CXCR4 BM homing receptor and indeed homed to the BM (refer to Figure 1C). Despite the coinfusion of large doses of Tcons, mice died of leukemia. (C-D) Flow cytometry analyses at day +7 from Tcon infusion showed that human CD4⁺ (blue and light blue columns) and CD8⁺ (green column) T cells were present in the BM of these mice. These cells were purified to assess their functionality separately. (E) Seventy percent of CD4⁺ T cells were FOXP3⁺ (light blue column) and inhibited the MLR. (F) CD8⁺ T cells had no ability to kill allogeneic leukemia and PHA blasts, indicating that CD45RA⁺ naïve Tregs had downregulated the Tcon effector function in the BM. (G-H) CD4⁺ (light blue and blue columns) and CD8⁺ (green column) T cells were also found in the liver. (I-J) Also in this case, CD4⁺ T cells (light blue column) inhibited MLR (I) and CD8⁺ T cells had no ability to kill allogeneic leukemia and PHA blasts (J). (K-N) When CXCR4 on naïve CD45RA⁺ Tregs was masked with an anti-CXCR4 blocking antibody in vitro, Tregs (light blue column) became unable to home to the BM (K,L) and, consequently, Tcons exerted unopposed leukemia killing (M,N) and mice cleared leukemia and did not develop GVHD and survived (refer to panel B). (O-Q) Flow cytometry analyses showed that CD4⁺ (light blue and blue columns) and CD8⁺ T cells were present in the liver (O,P). Sixty percent of CD4⁺ were FOXP3⁺ (light blue column) and inhibited the MLR compared with syngeneic controls (Q). (R) The CD8 T cells were not able to kill allogeneic leukemia and PHA blasts. Mice infused with human naïve CD45RA⁺ Tregs and Tcons, both treated with the anti-CXCR4 masking antibody, died of leukemia without developing GVHD, indicating that anti-CXCR4–treated Tcons were unable to home to the BM and exert antileukemic activity (refer to panel B). Dot plots are merely representative of one flow cytometric analysis. Statistical significance was defined as ∗P < .05.

Results in human Treg/Tcon transplants

In 20 patients with leukemia who received T-cell–depleted haplo-HSCTs in combination with adoptive transfer of Tregs/Tcons, we analyzed BM aspirates after transplant. One month after transplant, the BM contained 20% to 25% T cells, which were mostly (80%) CD8⁺ cells (Figure 6A), whereas Tregs were present in very low numbers (<1% of T cells). At this time point, in 5 of 20 patients, we performed functional analyses of BM T cells. T cells displayed cytotoxicity against cryopreserved recipient target cells (PHA blasts or leukemic cells) and, as expected, they did not inhibit the MLR (Figure 6B-C) in 3 of 5 patients. Immunohistochemistry analyses of BM biopsies showed a large predominance of T cells that exhibited a CD3⁺ and FOXP3⁻ phenotype (likely effector T cells; Figure 6D-E). One patient underwent a skin biopsy because of suspected but unconfirmed GVHD. The skin biopsy showed the presence of regulatory FOXP3⁺ T cells (Figure 6F-G).

Figure 6.

In patients who received haplo hematopoietic transplants and Treg/Tcon immunotherapy, BM T cells were able to kill leukemic cells. (A) Donor CD4⁺ (light blue and blue column) and CD8⁺ (green column) T cells in the BM of 20 patients who received haplo hematopoietic transplants and Treg/Tcon adoptive immunotherapy.14, 15, 16 CD8⁺ T cells were the predominant lymphocyte population. Within the first 30 days after transplant, CD3⁺/CD4⁺/CD25⁺/FOXP3⁺ Treg cells (light blue column) were virtually absent (<1% of T cells). (B-C) BM T cells from 5 patients were analyzed for cytotoxic activity and inhibitory function; T cells from 3 patients lysed recipient leukemia and PHA blasts (B) and, as expected, did not inhibit MLR between donor T cells and irradiated recipient leukemic cells (C). (D-G) Immunohistochemistry staining of human CD3⁺ (D,F) and FOXP3⁺ (E,G) T cells in the BM and skin biopsies, respectively from a patient who received T-cell–depleted haplo hematopoietic stem cells and Treg/Tcon adoptive immunotherapy (day +30 after transplant). Such histological analyses showed high numbers of CD3⁺/FOXP3⁻ effector T cells and low numbers of CD3⁺/FOXP3⁺ regulatory T cells in human BM. In contrast, in the human skin biopsy, FOXP3⁺ regulatory T cells were present.

Discussion

In the present study, in murine transplantation models and in patients undergoing haplo-HSCT, we showed, for the first time to our knowledge, the mechanism whereby immunotherapy with human regulatory and effector T cells protect against GVHD without loss of beneficial T-cell–dependent GVL effect. The migratory properties of Tregs have been investigated by several groups who demonstrated that Tregs migrate to peripheral tissues (ie, liver, gut, etc) where they help control Tcon alloreactivity and prevent GVHD. In mouse models of mismatched HSCT, Nguyen et al¹⁷ showed that, upon Treg infusion, those Tregs that are alloantigen specific become activated and expand in the lymph nodes. Afterward, they migrate to peripheral tissues depending on the expression of homing molecules for various sites.¹⁸ Thus, alloantigen-specific Tregs monitor the reactivity of Tcons not only in the lymph nodes but also in peripheral tissues that are targets of GVHD.

Although we confirmed the Treg immune regulatory properties in preventing GVHD in murine Treg/Tcon immunotherapy models,¹¹ we demonstrated that both murine spleen and human peripheral blood CD4⁺/CD25⁺/FOXP3⁺ Tregs expressed the CXCR4 BM homing receptor at lower intensity than Tcons. In fact, in our murine and xenogeneic transplant models and in patients, donor peripheral blood Tregs did not home to the BM. In contrast, donor effector T cells did express CXCR4 and, indeed, homed to the BM. When harvested from either murine or human BM, donor Tcons exerted leukemic cell killing. These results are consistent with the findings of Fujisaki et al²⁰ who showed that Tregs played a key role in making the stem cell niche an immunologically privileged site where they protected normal stem cells from T-cell alloreactivity. One may speculate that the HSCT conditioning regimen destroyed the recipient BM Tregs and, as a result, leukemic cells had no shield against alloreactive effector T lymphocytes that killed them. Therefore, peripheral donor Tregs protect against GVHD in the periphery, whereas adoptive immunotherapy with immediately available donor Tcons allows unleashed killing of leukemic cells in the BM. Obviously, these results could be limited to protocols that use peripheral blood Tregs. In fact, Tregs derived from different graft sources may exhibit distinct phenotypic and functional characteristics, particularly in terms of their memory vs naïve phenotype and expression of homing receptors such as CXCR4, thereby influencing posttransplant immune regulation. Cord blood grafts are typically enriched for CD45RA⁺ naïve Tregs,²¹ which predominantly express high levels of CXCR4. Similarly, BM grafts contain a higher percentage of CD45RA⁺ Tregs associated with increased CXCR4 expression.¹⁸ As a result, their homing to the BM may be enhanced, potentially increasing their ability to suppress the GVL activity mediated by Tcons. Again, in the case of plerixafor-mobilized peripheral blood grafts, plerixafor (a CXCR4 receptor antagonist) could mobilize naïve Tregs into the peripheral blood. It is plausible that Tregs mobilized under plerixafor treatment exhibit higher CXCR4 expression than steady-state peripheral blood Tregs. However, data on the Treg phenotype after plerixafor treatment remain limited.

The goal of allogeneic HSCT is to elicit an alloimmune response that eradicates resistant leukemias. In this context, leukemia immunotherapy induced by conventional T-cell–replete transplant may be impaired by immunosuppressive posttransplant prophylaxis, which could attenuate the strength of the T-cell–mediated GVL effect. However, further studies are required to directly compare the efficacy of the GVL response between T-cell–depleted and T-cell–replete transplant settings.

In pioneering HLA-haplo-HSCT, our group took an entirely different approach and invested in transplant of megadoses of T-cell–depleted hematopoietic stem cells in the absence of any posttransplant immunosuppression.22, 23, 24 This platform offered the opportunity to separate GVL from GVHD given that it allowed for the discovery of the effectiveness of donor natural killer cell alloreactivity.²⁵ More recently, we added the option of exploiting T-cell alloreactivity for leukemic cell killing through the use of Treg/Tcon adoptive immunotherapy.14, 15, 16 In this setting, Tregs made it possible to cotransplant a sufficient number of Tcons to eradicate leukemia without damaging normal tissues. Indeed, the recent clinical trial in patients with high-risk AML was associated with very low leukemia relapse and cGVHD rates that resulted in an unprecedented 75% cGVHD/relapse-free survival.¹⁶

Conflict-of-interest disclosure: The authors declare no competing financial interests.