Posttransplant blockade of CXCR4 improves leukemia complete remission rates and donor stem cell engraftment without aggravating GVHD

Abstract

Allogeneic hematopoietic cell transplantation (allo-HCT) is a promising therapeutic option for hematological malignancies, but relapse resulting predominantly from residual disease in the bone marrow (BM) remains the major cause of treatment failure. Using immunodeficient mice grafted with laboratory-generated human B-ALL, our previous study suggested that leukemia cells within the BM are resistant to graft-versus-leukemia (GVL) effects and that mobilization with CXCR4 antagonists may dislodge leukemia cells from the BM, enabling them to be destroyed by GVL effects. In this study, we extended this approach to patient-derived xenograft (PDX) and murine T-ALL and AML models to determine its clinical relevance and effects on GVHD and donor hematopoietic engraftment. We found that posttransplant treatment with the CXCR4 antagonist AMD3100 significantly improved the eradication of leukemia cells in the BM in PDX mice grafted with B-ALL cells from multiple patients. AMD3100 also significantly improved GVL effects in murine T-ALL and AML models and promoted donor hematopoietic engraftment in mice following nonmyeloablative allo-HCT. Furthermore, posttransplant treatment with AMD3100 had no detectable deleterious effect related to acute or chronic GVHD. These findings provide important preclinical data supporting the initiation of clinical trials exploring combination therapy with CXCR4 antagonists and allo-HCT.

Introduction

The prognoses of patients with acute leukemia have been substantially improved over the last 2 decades with advances in subset recognition, risk stratification, newly emerging targeted therapies, and immunotherapies [1–5]. However, relapse, which occurs most commonly in the bone marrow (BM), remains the major obstacle for successful treatment of acute leukemia [6, 7]. The BM is a site of immune privilege that exists in steady state to allow for normal hematopoiesis [8]. The immunosuppressive microenvironment of the BM also protects leukemia cells from chemotherapeutic agents, targeted small molecule drugs, and immunotherapy [9–12]. Stromal cell-derived factor-1 (SDF-1, also known as CXCL12) and its receptor CXCR4 are integral to hematopoietic stem cell (HSC) and hematopoietic progenitor cell (HPC) retention in and homing to the BM and are involved in lymphocyte trafficking [13]. The CXCL12/CXCR4 axis is also crucial for the homing, residence, and survival of acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL) cells in the BM [14, 15]. In preclinical models, blocking CXCR4 by small molecule antagonists was found to enhance the efficacy of chemotherapeutic agents or kinase inhibitors [15, 16]. Several phase 1/2 clinical trials of chemotherapy combined with CXCR4 antagonists have also revealed encouraging remission rates in AML patients, although bridging to more effective treatments still needs to be considered [17, 18]. It is possible that the chemotherapeutic protocols used in these studies were not sufficiently effective to eradicate leukemia stem cells (LSCs) and leukemia-initiating cells (LICs) even after the cells were mobilized out of the BM. Thus, for an improved outcome, CXCR4 blockade needs to be combined with treatments that are more efficient in eradicating mobilized LSCs and LICs.

The potential of CXCR4 antagonists to improve the antileukemia or graft-vs.-leukemia (GVL) effects of allogeneic hematopoietic cell transplantation (allo-HCT) has been relatively poorly explored. Using immunocompromised mice grafted with human B-cell ALL (B-ALL) cells generated from human CD34⁺ cells with forced MLL-AF9 overexpression, we found that injection of the CXCR4 antagonist AMD3100 after allogeneic lymphocyte infusion (ALI) could enhance GVL effects, leading to more efficient eradication of leukemia cells within the immune-privileged BM [19]. Although the results are promising, several fundamental questions remain unaddressed. The first concern is the model, which used human B-ALL developed in mice, raising the question of whether the observed potentiating effect of CXCR4 antagonists on GVL activity is limited to this specific leukemia model. Another issue is the immunodeficient nature of the recipients, which rendered the study unable to evaluate the effect of CXCR4 antagonists on the development of graft-versus-host disease (GVHD) and rejection or engraftment of donor HSCs. Given that the CXCL12/CXCR4 axis is highly effective in mobilizing HSCs and regulating the migration/tissue retention of immune cells, these questions must be addressed to determine the possibility of translating this strategy into the clinic. In the present study, we sought to further confirm the efficacy of combination therapy with ALI and CXCR4 antagonists in clinically relevant patient-derived xenograft (PDX) models and to investigate how CXCR4 antagonists may affect GVL effects on murine primary T-ALL and AML, GVHD and donor hematopoietic engraftment in murine models of allo-HCT. We found that in both PDX and murine leukemia models, allo-HCT achieved significantly improved disease-free survival rates when combined with posttransplant administration of the CXCR4 antagonist AMD3100. Furthermore, treatment with AMD3100 improved donor hematopoietic engraftment without affecting the course of acute or chronic GVHD. This study offers a potentially translational strategy for preventing and treating relapse following allo-HCT and other types of cellular immunotherapy, such as infusion of T cells with leukemic antigen-specific T cell receptors or chimeric antigen receptors (CARs).

Materials and methods

Animals and cells

C57BL/6 (B6; H-2D^b), BALB/c (H-2D^d), and CB6F1 (B6×BALB/c F1; H-2D^b×d) mice were purchased from Charles River Laboratory (Beijing, China). NOD-Prkdc^em26Cd52Il2rg^em26Cd22/Nju (hereafter abbreviated as NSG) mice were purchased from Nanjing Biomedical Research Institute of Nanjing University. Mice were housed in specific pathogen-free (SPF) conditions at the First Hospital Animal Center of Jilin University. Mice at 6–10 weeks of age were used in experiments. Murine primary AML cells of a B6 mouse background driven by an MLL-AF9 fusion gene (MLL-AF9-AML) were a gift from Tao Cheng (Institute of Hematology & Blood Disease Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College). Mouse primary T-ALL cells of a B6 mouse background driven by mutated Notch1 (Notch1-T-ALL) were prepared as previously described [20]. Both the leukemia cells expressed green fluorescent protein (GFP), which was used to identify leukemia cells via flow cytometry (FCM). Protocols involving the use of animals were approved by the Institutional Animal Care and Use Committee of the First Hospital of Jilin University, and all experiments were performed in accordance with protocols.

Human samples

BM or peripheral blood (PB) samples of patients with ALL were collected at diagnosis or relapse after informed consent was obtained from the Department of Hematology, the First Hospital of Jilin University. Peripheral blood mononuclear cells (PBMCs) from healthy volunteers were prepared by density gradient centrifugation. All procedures used in this study were approved by the ethics committee of the First Hospital of Jilin University.

Construction of patient-derived xenograft (PDX) mice and administration of allogeneic lymphocyte infusion

BM or PB mononuclear cells from individual patients with B-ALL were injected (i.v.) into NSG mice 4–8 h after conditioning with 1.2 Gy TBI. Leukemia cells in PBMCs were monitored by FCM with a combination of the following fluorochrome-conjugated mAbs: anti-mouse CD45 and anti-human CD45, CD19, and CD20. Spleens were harvested from these mice when high levels of leukemia cells were identified in PB and the mice became moribund and were used as the sources of human B-ALL cells for constructing PDX mice. Briefly, male NSG mice aged 8–10 weeks were conditioned with TBI (0.8 or 1.2 Gy as indicated) and injected (i.v.) with leukemia cells (2–250 × 10⁴/mouse as indicated). The leukemia cells in PBMCs and tissues were identified by FCM using anti-mouse CD45 and anti-human CD19 antibodies after donor-derived nonmalignant B cells disappeared. Allogeneic lymphocyte infusion (ALI) was performed by injection (i.v.) of allogeneic human PBMCs into leukemic NSG mice. Survival and subsets of human T cells in mice were evaluated by FCM using various combinations of mAbs specific for human CD45, CD19, CD20, CD3, CD4, and CD8.

Allo-HCT

Allo-HCT was performed in BALB/c-to-CB6F1 (parent to F1) and BALB/c-to-B6 combinations. Briefly, recipient mice were conditioned by total body irradiation (TBI; 2, 3, 7.5 or 9 Gy as indicated) and injected (i.v.) with BMCs (6–7.5 × 10⁶/mouse as indicated) and splenocytes (1.1–2.5 × 10⁷/mouse as indicated) from the indicated syngeneic or allogeneic donor mice 4–8 h after TBI. Leukemic recipients were additionally injected (i.v.) with T-ALL (1.5–2.0 × 10⁵/mouse as indicated) or AML (2.5 × 10⁵/mouse) cells along with HCT.

Treatment with the CXCR4 antagonist AMD3100

AMD3100 octahydrochloride hydrate (A5602, Sigma) was dissolved in phosphate-buffered saline (PBS), preserved in a dark environment at 4 °C, and used within 1 week. AMD3100 was given to ALI or allo-HCT recipients by subcutaneous injection (s.c; 5 mg/kg per injection) twice a day for 5 or 10 consecutive days as indicated. In the experiments evaluating mobilization effects, mice were given a single injection of AMD3100 at 5 mg/kg.

GVHD evaluation after allo-HCT

For the acute GVHD model, body weight and clinical GVHD scores of the recipient mice were monitored every two days. The GVHD clinical scores were evaluated as previously reported [21], including assessment of body-weight loss (0 points: <10%; 1 point: ≥10% to <25%; 2 points: ≥25%), posture (0 points: normal; 1 point: hunching noted only at rest; 2 points: severe hunching impaired movement), activity (0 points: normal; 1 point: mild to moderately decreased; 2 points: stationary unless stimulated), fur texture (0 points: normal; 1 point: mild to moderate ruffling; 2 points: severe ruffling/poor grooming), and skin integrity (0 points: normal; 1 point: scaling of paws/tail; 2 points: obvious areas of denuded skin). For the chronic GVHD model, body-weight and clinical GVHD scores of the recipient mice were determined every 2 to 7 days. The GVHD clinical scores were evaluated as previously reported [22], including assessment of body-weight loss (0 points: <2%; 1 point: ≥2% to <8%; 2 points: ≥8%), skin (0 points: healthy appearance; 1 point: skin lesions with alopecia <1 cm² in area; 2 points: skin lesions with alopecia ≥1 to <2 cm² in area; 3 points: skin lesions with alopecia ≥ 2 cm² in area; animals were assigned 0.3 points for each skin disease instance on the ears, tail, and paws), and hunch position (0 points: normal; 1 point: hunching noted only at rest; 2 points: hunching affected action).

Flow cytometry (FCM)

FCM was used to analyze leukemia cells and the phenotypes of immune cells. Antibodies used in this study were purchased from BD, eBioscience or BioLegend. Analysis was performed on a Fortessa or Canto II (BD Biosciences). The following fluorochrome-conjugated mAbs were used in this study: anti-human CD45, CD19, CD20, CD3, CD4, and CD8 and anti-mouse CD45, CD3, CD4, CD8, B220, NK1.1, Gr-1, CD11b, CD25, Foxp3, CD44, CD62L, Lin, Sca-1, and c-kit (CD117). A LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Life) or 7-AAD staining was used to gate out dead cells.

Histology

Tissues were harvested from the indicated mice, fixed with 4% paraformaldehyde overnight, and embedded in paraffin. Serial sections (4 μm) were prepared and analyzed for H&E staining.

Statistical analysis

All analyses were performed using Prism 8.0 (GraphPad Software) or Statistics Package for Social Sciences (SPSS) software (version 20.0, SPSS Inc., Chicago, IL, USA). Unpaired t-tests (for normal distribution and homogeneity of variance) or Mann–Whitney tests (for nonnormal distribution or heterogeneity of variance) were used to compare two groups of measurement data. Two-way analysis of variance (ANOVA) was used to compare groups with Sidak’s multiple comparisons test used for post hoc analysis. Kaplan–Meier curves were generated for survival analysis, and the log-rank test was used to compare the differences between groups. A P-value of <0.05 was considered significant in all analyses. The P-value levels are indicated as follows: ^*P < 0.05; ^**P < 0.01; ^***P < 0.001; and ^****P < 0.0001.

Results

Human leukemia cells in the BM are resistant to ALI in PDX mice

Using a mouse model of human B-ALL in NSG mice derived from human CD34⁺ cells with forced overexpression of MLL-AF9, we recently reported that dislodgement of leukemia cells from the BM improves the potential of ALI to eradicate leukemia cells [19]. To further determine the clinical relevance of these findings, we repeated these experiments in PDX models created using B-ALL cells from patients (Table S1). Immunophenotypic analysis revealed that B-ALL cells from these patients were CXCR4⁺ (Table S1 and Fig. S1A). Furthermore, these patient-derived B-ALL cells were engraftable in NSG mice, and treatment of the engrafted mice with AMD3100 resulted in rapid but transient mobilization of human B-ALL cells, with the peak cell content in blood seen at 3 h after the treatment (Fig. S1B–C). PDX mice were created by injection (i.v.) of patient (Pt #1) B-ALL cells (2.5 × 10⁶/mouse) into 1.2 Gy-irradiated NSG mice; 4 weeks later, when leukemia cells became detectable in peripheral blood (PB; ~1–5% of cells), PDX mice were subjected to two rounds of combination treatment with ALI (given on day 0 and day 18) and AMD3100 (starting on day 10 and day 25, respectively) (Fig. 1A). Given its transient mobilization effect (Fig. S1), AMD3100 was given twice a day for 5 consecutive days. As ALI-derived human B cells were undetectable by day 10 following ALI (Fig. S2), CD19⁺ cells detected afterward were considered to represent B-ALL cells in the recipient mice. Although the levels of B-ALL cells in the PB were comparable between the AMD3100 and PBS control groups one day before and 10 days after ALI (immediately prior to AMD3100 treatment), recipient mice treated with AMD3100 showed a significant decrease in leukemia cells compared to mice treated with PBS control at day 15 after ALI (one day after the last injection of AMD3100; Fig. 1B, C). With the exception of one mouse in the PBS control group with detectable leukemia cells (~1% of cells), leukemia cells were not detected in the PB of mice from either group at day 30 (Fig. 1D, E). Furthermore, the leukemia cell levels were extremely low or undetectable in tissues, including spleen, liver, and lung, in both groups at day 30, indicating that the second ALI mediated strong GVL effects (Fig. 1D, E). However, ALI was able to effectively eradicate leukemia cells in the BM only when combined with AMD3100. In contrast to mice treated with AMD3100, in which leukemia cells were nearly undetectable in the BM, PBS control mice showed large numbers of surviving leukemia cells in the BM (Fig. 1D, E). The two groups of mice showed no difference in body-weight changes (Fig. 1F), indicating that AMD3100 did not significantly affect the course of GVHD. Similar results were observed in separate experiments, in which PDX mice were created using human B-ALL cells from the same patient (Pt #1; Fig. S3) or from a different patient (Pt #3; Fig. S4) and treated with one or two rounds of ALI.

Fig. 1.

AMD3100 facilitates eradication of leukemia in the BM by ALI in PDX mice. PDX mice were created using patient (Pt #1) B-ALL cells (i.v.; 2.5 × 10⁶/mouse), and leukemia development in these mice was confirmed by FCM analysis of PBMCs (1–5% human B-ALL cells) ~4 weeks after leukemia cell injection. On the following day (designated day 0), PDX mice were given two rounds of combination treatment with ALI (4.5 × 10⁶ and 2.5 × 10⁶ allogeneic PBMCs/mouse given on day 0 and day 18, respectively) plus AMD3100 (5 mg/kg twice daily for 5 days starting on day 10 and day 25, respectively) or PBS (with the same volume and schedule). A Schematic outline of the experimental design. B Ratios (%; mean ± SEM) of human B-ALL cells in PBMCs at the indicated time points. C Numbers (mean ± SEM) of human leukemia cells in PBMCs at day 15 post-ALI. D, E Representative staining profiles (D) and percentages (mean ± SEM) of human leukemia cells in the indicated tissues at day 30 post-ALI (E). F Body-weight changes (mean ± SEM). ^*P < 0.05 (B: two-way ANOVA followed by Sidak’s multiple comparisons test; C: unpaired t-test)

Given that posttransplant treatment with AMD3100 appeared to have a limited effect in PDX mice generated with leukemia cells from another patient (Pt #2; in which leukemia growth in the BM was highly aggressive, with an average of over 90% leukemia cells at day 20 post-ALI (Fig. S5)), we investigated whether bridge chemotherapy could improve the outcome of these mice. Briefly, PDX mice were created by injection (i.v.) of B-ALL cells from patient (Pt #2) into NSG mice (2 × 10⁶/mouse) and were treated with cyclophosphamide (CTX) and dexamethasone (Dem) when large numbers of leukemia cells were detected in the PB ( > 50%); 1 week later, ALI (4 × 10⁶/mouse) plus AMD3100 or PBS was administered (5 consecutive days staring on day 10 with respect to ALI; Fig. 2A). Although leukemia cells in the PB, spleen, liver, and lung were effectively eradicated in both groups (Fig. 2B–D), complete eradication of leukemia cells in the BM was only detected in mice treated with ALI plus AMD3100, and mice treated with ALI alone (with PBS) had large numbers of surviving leukemia cells in the BM (Fig. 2C, D). Collectively, these results confirmed that the BM, as an immune-privileged site, protects leukemia cells from alloimmunity-related killing and that treatment with AMD3100 is effective in improving GVL effects by mobilizing leukemia cells out of the BM in clinically relevant PDX models. Furthermore, bridge chemotherapy may significantly improve the efficacy of post-allo-HCT treatment with AMD3100 for patients with a high leukemia burden or aggressive leukemia.

Fig. 2.

Combination treatment with ALI and AMD3100 enhances GVL effects in PDX mice pretreated with chemotherapy. PDX mice were created using patient (Pt #2) B-ALL cells (i.v.; 2 × 10⁶/mouse), and leukemia development in these mice was confirmed by FCM analysis of PBMCs (50–90% human B-ALL cells) ~4.5 weeks after leukemia cell injection. The following day, PDX mice were treated with CTX and Dem. Two days after chemotherapy (designated day 0), the mice were treated with ALI (4 × 10⁶ allogeneic PBMCs/mouse given on day 0) plus AMD3100 or PBS (starting on day 10). AMD3100 was given at 5 mg/kg twice a day for 5 consecutive days. A Schematic outline of the experimental design. B Ratios (%; mean ± SEM) of human B-ALL cells in PBMCs at the indicated time points. C, D Percentages (mean ± SEM) (C) and representative staining profiles (D) of human leukemia cells in the indicated tissues at day 15 post-ALI. ^*P < 0.05 (unpaired t-test)

AMD3100 treatment dislodges leukemia cells from the BM and improves GVL effects in murine acute lymphoblastic and myeloid leukemia models

Although PDX models closely resemble patients in the clinic, these models lack an intact host immune system that is autologous or syngeneic to the leukemia and therefore are unsuitable for mechanistic studies of ALI-associated alloimmune responses, including GVHD and donor hematopoietic engraftment. Thus, we addressed these questions in murine models of T-ALL and AML, which were derived from B6 mouse HSCs with forced overexpression of an activating Notch-1 mutation [20] and MLL-AF9 [23], respectively. FCM analysis revealed that both the Notch1-T-ALL and MLL-AF9-AML cells were positive for CXCR4 (Fig. 3A, E). By using leukemia-bearing syngeneic mice, we further confirmed that, like the PDX models (Fig. S1B-C), both the Notch1-T-ALL (Fig. 3B–D) and MLL-AF9 AML cells (Fig. 3F–H) showed a rapid and transient mobilization response to AMD3100, with the numbers of leukemia cells peaking within 3 h and returning to normal by 6 h after the treatment.

Fig. 3.

Mobilization of murine leukemia cells by AMD3100 in mice bearing syngeneic T-ALL or AML. B6 mice were injected (i.v.) with murine T-ALL (8 × 10⁵/mouse; A–D) or AML cells (5 × 10⁵/mouse; E–H). Two weeks later, the mice were treated with PBS or AMD3100 (5 mg/kg) the day after confirmation of leukemia development by FCM analysis of PBMCs. PBMCs were prepared 3, 6, and 12 h posttreatment, and leukemia cells were analyzed by FCM. CXCR4 expression on T-ALL (A) and AML (E) cells, representative FCM staining profiles (B, F) and levels (mean ± SEM) of GFP⁺ leukemia cells in PBMCs at the day prior to (%; C, G) and the indicated time points after (D, H; left, %; right, cell counts) AMD3100 injection in mice bearing T-ALL (B–D; n = 3/group) or AML (F–H; n = 5/group) cells are shown. ^**P < 0.01, ^****P < 0.0001 (two-way ANOVA followed by Sidak’s multiple comparisons test)

We first assessed the effect of AMD3100 on GVL effects against murine T-ALL. CB6F1 mice were conditioned with sublethal TBI (3 Gy) and injected with BMCs (7.5 × 10⁶/mouse) and splenocytes (2 × 10⁷/mouse) from syngeneic CB6F1 or allogeneic BALB/c donor mice. Leukemic recipients were additionally injected with Notch1-T-ALL cells (2 × 10⁵ per mouse) on the day of HCT and treated with PBS or AMD3100 (5 mg/kg, twice per day) for 5 consecutive days starting from day 8 post-HCT (Fig. 4A). Allo-HCT recipients without leukemia did not develop significant GVHD, as shown by a lack of detectable body-weight loss compared to nonleukemic recipients of syngeneic HCT (syn-HCT) or a lack of mortality (Fig. 4B and Fig. S6). Thus, leukemia was the major cause of death in mice receiving T-ALL cells (Fig. 4C), and in support of this conclusion, these mice showed an increase in GFP⁺ leukemia cells (Fig. 4D) and enlarged spleen and liver at autopsy (Fig. 4E). Although prolonged survival was observed in allo-HCT recipients compared to syn-HCT controls, mice treated with allo-HCT plus AMD3100 survived significantly longer than those receiving allo-HCT alone (Fig. 4C). In line with the survival results, FCM analysis of PB cells at different time points after HCT revealed that the frequencies of leukemia cells in mice receiving allo-HCT plus AMD3100 were markedly lower than those in mice receiving allo-HCT alone, although the latter also had fewer leukemia cells than the syn-HCT recipients (Fig. 4D). However, there was no detectable difference in survival or leukemia cell burden between syn-HCT recipients with and without AMD3100 treatment (Fig. 4C, D).

Fig. 4.

AMD3100 enhances GVL effects against murine T-ALL and AML cells. A–G CB6F1 mice were conditioned with 3 Gy TBI and injected (i.v.) with BMCs (7.5 × 10⁶/mouse) and splenocytes (2 × 10⁷/mouse) from syngeneic CB6F1 or allogeneic BALB/c mice injected with or without 2 × 10⁵ (A–E; n = 6 per group) or 1.5 × 10⁵ (F, G; n = 13 per group; data from two independent experiments were combined) T-ALL cells. Some recipients were treated with PBS or AMD3100 (5 mg/kg, twice per day) for 5 consecutive days starting on day 8 post-HCT. Shown are schematic outlines of the experimental design (A), body-weight changes (mean ± SEM) (B), survival (C, F), levels (%; mean ± SEM) of GFP⁺ T-ALL cells in PBMCs at the indicated time points (D, G), and representative images of spleens (top) and livers (bottom) from a normal mouse and recipients of allo-HCT plus 2 × 10⁵/mouse T-ALL cells that were treated with AMD3100 or PBS (both died at day 44) (E). As over half of the leukemic recipient mice in the syngeneic group died before day 34, day 34 data are not shown for these mice in Fig. 4D. H, I TBI (3 Gy)-conditioned CB6F1 mice were injected (i.v.) with BMCs (7.5 × 10⁶/mouse) plus splenocytes (2 × 10⁷/mouse) from BALB/c mice and AML cells (2.5 × 10⁵/mouse) and treated with AMD3100 (5 mg/kg, twice per day) or PBS for 5 consecutive days starting on day 8 post-HCT (n = 6 per group). Shown are the survival rates (H) and percentages (mean ± SEM) of leukemia cells in PBMCs at the indicated time points (I). ^*P < 0.05; ^**P < 0.01; ^***P < 0.001, ^****P < 0.0001 (C, F, H: log-rank test; D, G, I: two-way ANOVA followed by Sidak’s multiple comparisons test)

In a separate experiment in which allo-HCT recipients were injected with a lower number of T-ALL cells (1.5 × 10⁵ per mouse), most AMD3100-treated allo-HCT recipients (~85%) survived to the end of the experiment (8 months), while most PBS-injected controls (~85%) died of leukemia (Fig. 4F). Furthermore, all PBS controls showed a continuous and dramatic increase in T-ALL cells in PB before death, while the T-ALL cell levels in most of the AMD3100-treated mice were profoundly low and became undetectable by 68 days (Fig. 4G). All long-term survivors who received allo-HCT (11 out of 13 and 2 out 13 in the AMD3100 and PBS groups, respectively) were leukemia-free, as shown by a lack of evidence for tumors at autopsy (Fig. S7A) or a lack of detectable recipient class I (H-2D^b)⁺ or GFP⁺ leukemia cells by FCM analysis (Fig. S7B, C). We then evaluated the potential of AMD3100 to improve GVL effects against AML cells in 3 Gy-conditioned CB6F1 mice that were injected with MLL-AF9-AML cells and treated with allo-HCT from BALB/c mice with or without AMD3100 treatment. Allo-HCT recipients treated with AMD3100 showed significantly prolonged survival (Fig. 4H) and fewer leukemia cells in the PB (Fig. 4I) than those injected with PBS. Taken together, these results provide direct evidence that the posttransplant treatment with AMD3100 significantly enhanced the efficacy of allo-HCT in eradicating leukemia cells in murine models of T-ALL and AML.

Posttransplant treatment with AMD3100 does not augment GVHD

CXCR4 is also expressed on immune cells and contributes to their retention in the BM [24, 25]. FCM analysis revealed that among the immune cells within mouse BM, almost all CD3⁺CD8⁺ T cells expressed a high level of CXCR4, and a proportion of CD3⁺CD4⁺CD25⁺Foxp3⁺ Tregs, B220⁺ B cells, and NK1.1⁺ cells were also clearly positive for CXCR4 (Fig. S8). Accordingly, AMD3100 treatment induced rapid mobilization of these immune cell populations from the BM into circulation (Fig. S9). These findings suggest that AMD3100 administration may affect the course of GVHD by inducing immune cell tissue redistribution. We first assessed the effect of AMD3100 on acute GVHD in nonleukemic recipients: lethally irradiated B6 mice that were injected with BMCs (7.5 × 10⁶/mouse) and splenocytes (1.5 × 10⁷/mouse) from BALB/c mice or B6 mice, some of which were further treated with PBS or AMD3100 (5 mg/kg, twice per day) for 5 consecutive days starting from day 6 (Fig. 5A). Although allo-HCT induced significant body-weight loss compared to syn-HCT or allo-BMC injection only, there was no significant difference between mice in the AMD3100 and PBS groups (Fig. 5B). The two groups of allo-HCT recipients also showed similar mortality rates (Fig. 5C), GVHD clinical scores (Fig. 5D), and pathological changes in major GVHD target organs, including the intestine, liver, lung, and skin (Fig. 5E). Similar results were observed in two additional independent experiments, in which allo-HCT recipients received reduced numbers of donor splenocytes and developed less severe GVHD (Fig. S10).

Fig. 5.

AMD3100 does not exacerbate acute GVHD. B6 mice were conditioned with 9 Gy TBI and injected (i.v.) with BMCs (7.5 × 10⁶/mouse) and splenocytes (1.5 × 10⁷/mouse) from syngeneic B6 mice (n = 11) or from allogeneic BALB/c mice (allo-HCT) followed by treatment with PBS (n = 13) or AMD3100 (n = 14). B6 mice treated with TBI only (n = 6) or treated with TBI and injected with allogeneic BMCs only without splenocytes (allo-BMCs; n = 7) were used as controls. PBS or AMD3100 (5 mg/kg/injection) was given twice per day for 5 consecutive days starting on day 6 post-BMT. Shown are schematic outlines of the experimental design (A), body-weight changes (mean ± SEM; B), survival (C), clinical GVHD scores (mean ± SEM; D; the symbols of syn-HCT are completely embedded in those of allo-BMCs), and pathological presentations of the indicated tissues for representative syn-HCT controls (sacrificed at day 30) and allo-HCT recipients treated with PBS (died at day 14) or AMD (died at day 11) (E)

We also evaluated the effect of AMD3100 on chronic GVHD in leukemic allo-HCT recipients: CB6F1 mice were conditioned with 7.5 Gy TBI and injected (i.v.) with BMCs (7.5 × 10⁶/mouse) and splenocytes (2.5 × 10⁷/mouse) from syngeneic CB6F1 or allogeneic BALB/c mice along with T-ALL cells (1.5 × 10⁵/mouse). In this model, allo-HCT recipients developed chronic GVHD characterized by body-weight loss, skin lesions (hair loss, redness, flaking, and scabbing), and hunched postures; this outcome was not affected by AMD3100 treatment, as the two groups of mice showed no difference in mortality rates (Fig. 6A), body-weight changes (Fig. 6B), overall GVHD scores (Fig. 6C), skin GVHD scores (Fig. 6D, E), or pathological presentations of GVHD target tissues (Fig. 6F). All syngeneic controls died of leukemia, while leukemia cells were completely eradicated in allo-HCT recipients (in both the AMD3100 and PBS groups; Fig. S11), indicating that GVHD was the cause of death following allo-HCT (Fig. 6A). Together, these results indicate that the post-HCT treatment with AMD3100 had no deleterious effect related to acute or chronic GVHD.

Fig. 6.

AMD3100 does not exacerbate chronic GVHD. CB6F1 mice were conditioned with 7.5 Gy TBI and injected (i.v.) with BMCs (7.5 × 10⁶/mouse) and splenocytes (2.5 × 10⁷/mouse) from syngeneic CB6F1 or allogeneic BALB/c mice along with T-ALL cells (1.5 × 10⁵/mouse). The allo-HCT recipients were treated with PBS or AMD3100 (5 mg/kg, twice per day) for 5 consecutive days starting on day 8 post-HCT. Shown are survival (A), body-weight changes (mean ± SEM; B), clinical GVHD scores (mean ± SEM; C), skin GVHD scores (mean ± SEM; D), photos of representative allo-HCT recipients at day 64 post-HCT (E), and pathological presentations of the indicated tissues harvested from a normal CB6F1 mouse and allo-HCT recipients treated with PBS (died at day 62) or AMD (died at day 58) (F)

Posttransplant treatment with AMD3100 promotes engraftment of donor hematopoietic stem/progenitor cells

AMD3100 treatment mobilizes not only leukemia cells but also normal HSCs and HPCs from the BM [26]; as such, donor hematopoietic engraftment is potentially affected in allo-HCT recipients. Previous studies using murine congenic or syngeneic transplant models showed that CXCR4 antagonists selectively improve donor hematopoietic engraftment in myeloablative recipients but not in nonmyeloablative recipients [27, 28]. However, it remains unclear whether AMD3100 affects donor hematopoietic engraftment in allo-HCT settings. CB6F1 mice conditioned with 3 Gy TBI were injected (i.v.) with BMCs and splenocytes from BALB/c mice and treated with AMD3100 or PBS for 5 consecutive days starting on day 8 and analyzed for donor chimerism 1 day after the last injection of AMD3100 (13 days after allo-HCT; Fig. 7A). The recipient mice had mixed chimerism at day 7 post-allo-HCT (one day before AMD3100 treatment; with cell percentages of 52.32 ± 4.84% and 17.62 ± 5.12% in the PB and BM, respectively; Fig. 7B), but all mice achieved nearly full donor chimerism at day 13 post-HCT regardless of whether they were treated with AMD3100 or PBS (Fig. 7C), and there were no significant differences in the frequencies of T, B, or myeloid cells in PB and BM between the two groups of mice (Fig. 7D, E). However, AMD3100 treatment induced a significant increase in both the frequency and number of LSK (Lin^-Sca-1⁺c-kit⁺) cells in the BM (Fig. 7F), supporting a beneficial effect of AMD3100 treatment on the engraftment and/or survival of donor HSCs and HPCs. Similar results were observed in another experiment, in which AMD3100 treatment resulted in a significant increase in donor hematopoietic chimerism in the BM in allo-HCT recipients that were conditioned with a reduced dose (2 Gy) of TBI (Fig. S12).

Fig. 7.

The effect of AMD3100 on donor hematopoietic engraftment. CB6F1 mice (n = 12) were conditioned with 3 Gy TBI and injected (i.v.) with BMCs (6.8 × 10⁶/mouse) and splenocytes (1.2 × 10⁷/mouse) from allogeneic BALB/c mice. The recipient mice were randomly divided into two groups (n = 6 per group) and treated with PBS or AMD3100 (5 mg/kg, twice per day) for 5 consecutive days starting on day 8 post-HCT. BM aspiration was performed 1 day before the first (day 7; five mice were randomly selected) and 1 day after the last (day 13; n = 6 per group) injection of PBS or AMD3100. A Schematic outline of the experimental design. B Representative FCM staining profiles (left) and percentages (mean ± SEM) (right) of donor (H-2D^b-H-2D^d+) cells in the PB and BM at day 7 post-HCT. C Representative FCM staining profiles (top) and percentages (mean ± SEM) (bottom) of donor (H-2D^b-H-2D^d+) cells in the PB and BM at day 13 post-HCT. D, E Percentages (mean ± SEM) (left) and numbers (mean ± SEM) (right) of donor CD3⁺, B220⁺, and CD11b⁺ cells in the PB (D) and BM (E). F Representative FCM staining profiles (left), percentages (middle; mean ± SEM) and numbers (right; mean ± SEM) of donor Lin^-Sca-1⁺c-kit⁺ (LSK) cells in the BM. ^*P < 0.05 (unpaired t-test)

Discussion

Although allo-HCT is highly appreciated to mediate powerful GVL effects with the potential to eliminate residual leukemia cells or measurable residual disease (MRD), relapse is inevitable for many patients [29]. LSCs in the BM are responsible for leukemogenesis and MRD [30, 31]. CXCR4 is widely expressed on leukemia cells, and higher CXCR4 expression is associated with a poorer prognosis [32, 33]. The CXCL12/CXCR4 axis is crucial for the homing of AML and ALL cells to the BM and their residence and survival in this location [14, 15]. We previously reported that dislodgement of leukemia cells from the BM by the CXCR4 antagonist AMD3100 improves MRD-negative remission rates in a mouse model of human B-ALL, in which NSG mice were grafted with human B-ALL cells derived from CD34⁺ cells with forced expression of MLL-AF9 [19]. In this study, to further determine the clinical relevance and understand the mechanism, we extended our study to include human B-ALL PDX models made with patient-derived leukemia cells and murine T-ALL and AML models that permit assessment of GVL effects and GVHD in allogeneic settings with immunocompetent recipients. We showed that in all these models, leukemia cells in the BM were more resistant to GVL effects than those in other tissues and that leukemia cell mobilization with the CXCR4 antagonist AMD3100 markedly improved complete remission rates. Furthermore, treatment with AMD3100 significantly improved donor hematopoietic engraftment without increasing GVHD. Although dislodgement of LSCs out of their niches may compromise their survival and disease-initiating potential, this treatment appeared to have limited benefits even in patients receiving CXCR4 antagonists in combination with chemotherapy [17, 18]. As the BM microenvironment is highly immunosuppressive [8], the CXCR4/CXCL12 axis not only improves LSC survival and function but also protects these cells from alloreactive T cells by keeping them in the BM niches. Thus, forcing LSCs out of the niche using CXCR4 antagonists may expose them to and enable them to be killed by alloreactive T cells, which is considered the main mechanism by which post-HCT treatment with AMD3100 improves complete remission rates. The immunoprivileged nature of the BM is maintained by multiple factors and complex mechanisms involving Tregs, mesenchymal stromal cells (MSCs), immature myeloid cells, and hypoxia [8, 34–36]. Thus, dislodging leukemia cells from the BM is considered a simpler and more effective strategy for improving MRD-free survival than directly targeting the BM immunosuppressive environment.

Treatment with CXCR4 antagonists mobilizes not only HSCs and LSCs but also immune cells that express CXCR4 [37, 38]. We found that most immune cell populations in the mouse BM expressed CXCR4 and hence were responsive to AMD3100, consistent with a previous report [37]. However, it should be noted that the effect of AMD3100 on immune cell trafficking and tissue relocation in recipients of allo-HCT is more complex than that in naïve mice, as both CXCR4 expression on immune cells and the production of its ligand CXCL12 may differ depending on the immune activation status and cytokine milieu [39, 40], and recipient immune cells are expected to be substantially eliminated and replaced by donor-derived immune cells. It is conceivable that treatment with CXCR4 antagonists will also affect donor-derived immune cells and hence GVHD. However, it should be noted that immune cells may express different types and/or levels of chemokine receptors depending on their activation status [40] and hence are simultaneously regulated by multiple chemokines. Nonetheless, our results did not identify any difference in the development of acute or chronic GVHD between allo-HCT recipients with and without AMD3100 treatment, indicating that transiently blocking the CXCL12/CXCR4 axis had no or minimal effects on GVHD. In support of our results, the addition of the CXCR4 antagonist plerixafor to chemotherapy [17, 18] or allo-HCT [41, 42] was also reported to be safe in patients with hematologic malignancies. Together, these studies support the safety of using CXCR4 antagonists in patients following allo-HCT.

CXCR4 antagonists dislodge not only leukemia cells but also HSCs/HPCs from the BM [26]. In congenic and syngeneic mouse models of myeloablative transplant, posttransplant administration of AMD3100 improved donor LSK cell engraftment in the BM and facilitated the recovery of donor blood cells [27]. However, posttransplant mobilization by CXCR4 antagonists failed to achieve a significant improvement in donor LSK cell engraftment or multilineage hematopoietic cell recovery, although a transient increase in neutrophils was observed in the PB following nonmyeloablative conditioning [28]. These studies suggest that blockade of CXCR4 may have a similar impact on recipient and donor cells following congenic or syngeneic transplantation and therefore could only significantly improve donor engraftment in myeloablative but not nonmyeloablative recipients. However, AMD3100 is expected to improve donor hematopoietic engraftment following nonmyeloablative allo-HCT because although AMD3100 mobilizes both recipient and donor HSCs/HPCs, the mobilized recipient cells will be selectively destroyed by alloreactive donor T cells. In support of this possibility, we found that treatment with AMD3100 significantly improved donor hematopoietic engraftment, as shown by the increased frequency and number of donor hematopoietic cells, including LSK cells, in the BM in mice following nonmyeloablative allo-HCT. In agreement with our results, AMD3100 treatment was also reported to facilitate neutrophil and platelet recovery in patients following allo-HCT [41].

We have shown that posttransplant treatment with AMD3100 is highly effective in certain patients in whom leukemia cells can be efficiently eliminated in the periphery after being mobilized from the BM, i.e., in recipients with a low leukemia burden in the periphery and ongoing GVL responses. Therefore, AMD3100 in combination with donor lymphocyte infusion (DLI) offers a potential therapy for leukemia patients with early-stage relapse. AMD3100 may also be used as a preemptive treatment in patients with high-risk diseases or MRD positivity after allo-HCT with or without DLI when immunosuppressants are withdrawn to prevent disease relapse. However, for patients with a high leukemia burden, preconditioning or bridging therapy to reduce the leukemia burden is likely to be necessary prior to combination therapy with allo-HCT and AMD3100. In addition, this study did not test other mobilization drugs that are also capable of mobilizing HSCs and LSCs, affecting donor engraftment, and altering immune responses [43]. Granulocyte colony-stimulating factor (G-CSF) is the most widely used agent for HSC mobilization in the clinic. G-CSF mobilizes not only HSCs from the BM but also leukemia cells and has been used to treat patients with leukemia [44, 45]. Posttransplant treatment with G-CSF facilitates the engraftment and recovery of donor cells [46, 47]. G-CSF also induces immune tolerance when used in HSC mobilization of allo-HCT [24, 48]. However, further studies are needed to determine whether posttransplant treatment with different mobilization drugs may lead to different outcomes or mediate different effects depending on the stage of disease or type of leukemia.

In summary, this study provides direct evidence that posttransplant treatment with AMD3100 significantly improved complete remission rates in both clinically relevant PDX and multiple murine allo-HCT models. Furthermore, AMD3100 treatment facilitated donor hematopoietic engraftment without increasing GVHD, offering a potentially safe approach for improving the efficacy of allo-HCT or other types of cellular immunotherapy, such as infusion of tumor antigen-specific TCR-expressing T cells or CAR-T cells. This study provides an impetus for further clinical trials exploring the posttransplant use of CXCR4 antagonists in allo-HCT.

Author contributions

Conceptualization: LS, ZH, Y-GY; Methodology: LS, M-HF, JZ, S-JG, X-YG, ZH, Y-GY; Investigation: LS, M-HF, JZ, S-JG, X-YG, X-DM, XW; Visualization: LS, X-YG, XW; Funding acquisition: LS, ZH, Y-GY; Project administration: LS, ZH, Y-GY; Supervision: ZH, Y-GY; Writing—original draft: LS, ZH, Y-GY; Writing—review & editing: all authors.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41423-021-00775-9.