Bendamustine with total body irradiation limits murine graft-versus-host disease in part through effects on myeloid-derived suppressor cells

Abstract

Graft-versus-host disease (GvHD) remains a significant challenge in allogeneic hematopoietic cell transplantation (HCT). An under-investigated strategy to reduce GvHD is the modification of the preparative conditioning regimen. Our study aimed to evaluate GvHD associated with bendamustine (BEN) conditioning in conjunction with total body irradiation (TBI), as an alternative to the standard myeloablative regimen of cyclophosphamide (CY) and TBI. We demonstrate that BEN-TBI conditioning, while facilitating complete donor chimerism, results in significantly less GvHD compared to CY-TBI. In BEN-TBI conditioned mice, suppressive CD11b⁺Gr-1^high myeloid cells are increased in the blood, bone marrow (BM), spleen, and intestines. When Gr-1^high cells are depleted prior to transplant, the beneficial effects of BEN-TBI are partially lost. Alternatively, administration of G-CSF, which promotes CD11b⁺Gr-1⁺ myeloid cell expansion, trends towards an increase in survival in BEN-TBI mice. These findings indicate a potential role of myeloid-derived suppressor cells (MDSCs) in the mechanism by which BEN allows engraftment with reduced GvHD. BEN-TBI conditioning may present a safer alternative to CY-TBI conditioning for allogeneic HCT.

Graphical Abstract

INTRODUCTION

Allogeneic hematopoietic cell transplantation (HCT) can be curative for many patients with hematological disorders and malignancies, but graft-versus-host disease (GvHD) remains a significant barrier to its success. It is well-documented that the dose intensity and the specific agents used in pre-transplant conditioning have an impact on the incidence and severity of GvHD¹. Total body irradiation (TBI) based myeloablative conditioning (MAC) may be associated with a higher incidence of acute GvHD when compared to chemotherapy-based preparative regimens². However, TBI is widely used in HCT due to its anti-leukemic and immunosuppressive effects and its ability to treat extramedullary sanctuary sites of disease. While other agents such as etoposide³ or cytarabine⁴ have been evaluated in combination with TBI, cyclophosphamide (CY) and TBI has remained the most widely applied conditioning regimen for acute lymphoblastic leukemia (ALL) for almost half a century⁵ and is also a frequently applied regimen (albeit at lower doses) for nonmalignant conditions, such as severe aplastic anemia^6–10. There is a scarcity of knowledge on how GvHD may be altered by replacing CY with other agents in TBI-based conditioning regimens.

Bendamustine (BEN), an alkylating agent and purine analog, has been used as treatment for lymphomas^11–16 and chronic lymphocytic leukemia (CLL)^17–19 and has been shown to be safe and effective as a conditioning agent for autologous transplants^20,21. Additionally, when replacing CY and given in combination with fludarabine and rituximab (BFR) as allogeneic reduced intensity conditioning for CLL, BEN resulted in reduced myelosuppression and GvHD²², decreased treatment-related mortality, and superior survival²³.

We recently published that BEN can safely replace post-transplant cyclophosphamide (PT-CY) following murine haploidentical bone marrow transplantation (BMT), resulting in comparable protection from GvHD and superior graft-versus-leukemia (GvL) effects. We observed that post-transplant bendamustine (PT-BEN) preserved the myeloid compartment and resulted in an increased number of CD11b⁺Gr-1^high granulocytic myeloid-derived suppressor cells (MDSCs) compared to PT-CY. Additionally, BEN treatment in vitro enhanced MDSC function²⁴. It is well-documented that MDSCs play an important role in limiting GvHD and can be modulated by chemotherapy treatment²⁵. Adoptive transfer of MDSCs generated in vivo or in vitro can attenuate GvHD in an allogeneic murine BMT model^26–29. Furthermore, greater MDSC content in donor grafts is correlated with reduced incidence of acute GvHD in humans^30,31.

Based on these data, we hypothesized that BEN could effectively replace CY in the traditional CY-TBI myeloablative conditioning regimen, reducing GvHD through its effects on MDSCs. Here we demonstrate that BEN-TBI conditioning results in significantly less GvHD than the standard CY-TBI in part through effects on CD11b⁺Gr-1^high MDSCs.

MATERIALS AND METHODS

Mice

All strains of mice used were age-matched 6–10 week-old females purchased from The Jackson Laboratory (Bar Harbor, ME, USA). Mice were housed in specific pathogen-free conditions and cared for according to the guidelines of the University of Arizona Institutional Animal Care and Use Committee (IACUC).

BMT models

For the MHC-mismatched model used throughout, recipient BALB/c (H-2^d) mice received 40 mg/kg of BEN intravenously (iv) or 200 mg/kg of CY intraperitoneally (ip) on day −2 and 400 cGy TBI on day −1 using a Cesium 137 irradiator. Based on the literature, it is expected that drugs are cleared by 24 hours post-administration^32–34. On day 0, mice received 10⁷ C57BL/6 (H-2^b) bone marrow (BM) cells with 3×10⁶ spleen cells (SC) or 10⁷ T-cell depleted BM cells (TCD-BM) with 3×10⁶ isolated total T-cells (tT) iv. In some experiments, tT were isolated from congenic CD45.1⁺ BoyJ mice. Moribund mice were euthanized according to IACUC-approved criteria and procedures and survival was monitored daily. Mice were weighed every three to four days and percent of starting weight was calculated. Mice were also scored clinically on skin integrity, fur texture, posture and activity and cumulative GvHD scores were calculated³⁵. Mice given a cumulative score of 8 following day +8 were euthanized. A veterinary pathologist evaluated tissues for histological evidence of GvHD³⁶. For the haploidentical BMT model used, CB6F1 (H-2^b/d) mice received 50 mg/kg of BEN or 225 mg/kg of CY, 300 cGy TBI, and 10⁷ B6AF1 (H-2^b/k) BM cells with 3×10⁷ SC.

Preparation of total T-cells and T-cell-depleted BM

Total T-cells were isolated from naïve C57BL/6 spleens by negative selection using mouse Pan T-Cell Isolation Kit II (Miltenyi Biotec, Auburn, CA, USA) with a purity of >97%. T-cells were depleted from BM cells using the CD3ε MicroBead Kit (Miltenyi Biotec), with less than 0.3% CD3ε⁺ cells remaining.

Drug preparation and administration

CY and BEN were reconstituted and diluted as previously reported²⁴. Anti-Gr-1 depleting antibody (clone RB6–8C5; ThermoFisher Scientific, Waltham, Massachusetts, USA) and G-CSF (Amgen, Thousand Oaks, CA, USA) were diluted in sterile saline for injection. 200 μg of anti-Gr-1 was administered ip on days −3, −1, and +5 and 250 μg/kg of G-CSF was administered subcutaneously on days −2 through +11.

Flow cytometry

Prior to analysis by flow cytometry, blood was collected by cardiac puncture or tail tipping and red blood cells were lysed (BD Biosciences, San Jose, CA, USA). Spleens were processed to single cell suspension and red blood cells were lysed prior to flow cytometry. Intestines were digested as described below. Flow cytometry was performed as previously reported³⁷. Fluorescence data were collected with an LSRFortessa cell analyzer (BD Biosciences) and analyzed using FlowJo 2 (Tree Star, Ashland, OR, USA). Antibodies used were anti-mouse Gr-1 FITC (RB6–8C5), CD11b eFluor450 (M1/70), H2kb PerCP-eFluor710 (AF6–88.5.5.3), H2kd PE (SF1–1.1.1), CD45.1 APC (A20) (ThermoFisher Scientific), and CD45.1 PE-CF594 (A20) (BD Biosciences). Of note, anti-mouse Gr-1 clone RB6–8C5 has been shown to react strongly with Ly6G (a marker for granulocytic MDSCs) and more weakly with Ly6C (a marker for monocytic MDSCs), resulting in a delineated Gr-1^high population, representing the Ly6G⁺ granulocytic MDSCs, and a Gr-1^mid population, representing the Ly6C⁺ monocytic MDSCs³⁸. To determine absolute cell numbers in blood, white blood cell counts were determined using a HemaVet 950 (Drew Scientific, Miami Lakes, FL, USA)³⁹. Cells were analyzed for reactive oxygen species (ROS) using Abcam’s Cellular ROS Detection Assay Kit (Cambridge, United Kingdom).

Intestine digestion and immunofluorescence

To analyze the immune cells present in the intestines, flow cytometry and immunofluorescence were used. For flow cytometry, intestines were first digested to single cell suspension using a modification of a previously published protocol⁴⁰. Briefly, intestines were flushed and incubated at 37°C in a shaker with Hank’s Balanced Salt Solution (HBSS; ThermoFisher Scientific) with 5% FBS (Atlanta Biologicals, Flowery Branch, GA, USA), 10 mM HEPES (ThermoFisher Scientific), and 1 mM dithiothreitol (DTT; Bio-Rad Laboratories, Hercules, CA, USA) for 15 minutes. Intestines were then sequentially incubated in digestion solution (HBSS with 5% FBS and 10 mM HEPES, 100 U/mL Type I Collagenase, and 40 μg/mL DNase I, Grade II (MilliporeSigma, St. Louis, MO, USA)) for five and ten minutes. For immunofluorescence, intestines were fixed in formalin, embedded in paraffin, and slides were prepared by the UAC Pathology Services Lab. Slides were de-waxed using xylenes, antigen retrieval was performed by steaming in sodium citrate buffer (MilliporeSigma), and slides were incubated with anti-Gr-1 (1:125) and goat anti-rat–NL 637 (1:500; R&D Systems, Minneapolis, MN, USA) and mounted in Fluoroshield mounting medium with DAPI (Abcam). Staining was analyzed using a Keyence BZ-X700 digital fluorescence microscope (Itasca, IL, USA) and quantified using ImageJ (NIH, Bethesda, MD, USA).

Suppression assays

Suppression assays were conducted and analyzed as previously reported²⁴. Briefly, T-cells were isolated from spleens of naïve C57BL/6 mice, stained, and stimulated. Gr-1^high MDSCs were isolated from spleens using a mouse Myeloid-Derived Suppressor Cell Isolation kit (Miltenyi Biotec), with >95% purity. MDSCs were co-incubated with T-cells at various ratios for 3 days. Flow cytometry was followed by Modfit (Verity Software House, Topsham, ME, USA) analysis to determine the proliferation index (PI) of the T-cells, in order to calculate % proliferation.

qRT-PCR

Cells were frozen in dry pellets prior to mRNA isolation using an RNeasy Kit (Qiagen, Hilden, Germany). cDNA was generated and quantitative real-time polymerase chain reaction (qRTPCR) was performed and analyzed as previously reported^39,41.

Statistics

Kaplan–Meier survival curves were generated and the log-rank statistic was used to evaluate differences between groups^42,43. Mann–Whitney tests were used to determine other differences between groups.

RESULTS

BEN-TBI conditioning results in improved survival compared to CY-TBI

Using a fully MHC-mismatched murine BMT model (C57BL/6, H2^b→BALB/c, H2^d), we compared BEN-TBI to the traditional CY-TBI conditioning. Equivalent doses, ~50% of the maximum tolerated dose, of BEN and CY were used (Figure S1). With no post-transplant GvHD prophylaxis, BEN-TBI resulted in significantly increased survival over CY-TBI in both a severe GvHD model (3×10⁶ T-cells) (Figure 1A) and a milder model (3×10⁶ SC) (Figure 1B). To further confirm these findings, SC (3×10⁶) were given alongside a comparable number of purified T-cells (10⁶), with BEN-TBI conditioned mice demonstrating similarly increased survival over CY-TBI conditioned mice, regardless of the inoculum (Figure 1C). This indicates that the presence of other donor immune cells in the graft is not required for the protective effects of BEN-TBI and that this difference in survival holds true over a range of T-cell doses. Moreover, BEN-TBI resulted in increased survival compared to CY-TBI over a range of drug doses (Figure 1D). To verify that this difference in GvHD was not due to graft rejection, donor cell engraftment was evaluated in the blood at various intervals. BEN-TBI and CY-TBI conditioning both resulted in full donor chimerism with no apparent difference in engraftment kinetics or myeloid/lymphoid subset chimerism (Figure 1E, S2). Greater than 90% donor chimerism was observed in the bone marrow on days +7 and +14, confirming lack of graft failure (data not shown). Donor cell engraftment was confirmed in all experiments and for all donor T-cell inoculums and conditioning drug doses utilized. Additionally, complete donor cell engraftment in the blood and spleen was confirmed at the conclusion of experiments (data not shown). Importantly, by utilizing an F1→F1 haploidentical murine BMT model, we demonstrated that the effects of BEN-TBI on GvHD are not model-specific. In CB6F1 (H2^b/d) mice transplanted with B6AF1 (H2^b/a) BM and SC, BEN-TBI resulted in significantly increased survival (Figure S3) and equivalent full donor chimerism when compared to CY-TBI (data not shown).

Figure 1. BEN-TBI conditioning results in significantly increased survival compared to CY-TBI.

BALB/c recipient mice received 40 mg/kg BEN iv or 200 mg/kg CY ip on day −2, 400 cGy TBI on day −1, and 10⁷ TCD-BM cells with 3×10⁶ purified T-cells (tT) (A), 10⁷ BM with 3×10⁶ SC (B, D), or 10⁷ TCD-BM with 3×10⁶ SC or 10⁶ purified T-cells (C) from naïve C57BL/6 mice on day 0. (A-C) Survival is shown. (A) Pooled data from 4 experiments are shown, n=16 mice/group, p=0.0072. (B) Pooled data from 6 experiments are shown, n=31 mice/group, p<0.0001. (C) Representative data from 2 experiments are shown, n=4 mice/group, BEN vs. CY: p=0.0067, SC vs. T-cells p=n.s.. (D) BALB/c mice were given BEN or CY at various doses, followed by 400 cGy TBI and 10⁷ BM with 3×10⁶ SC. Pooled data from 2 experiments are shown, n=6–8 mice/group. For all doses, BEN vs. CY: p<0.01. (E) Peripheral blood was collected on days +7, +14, and +21 and analyzed by flow cytometry. H2kb⁺ cells were considered to be of donor origin. Pooled data from 3 experiments are shown, n=7–13 mice/group/time point.

BEN-TBI results in decreased GvHD morbidity

To further evaluate GvHD differences between these two conditioning regimens, mouse weights and clinical GvHD scores were monitored over time³⁵. Although BEN-TBI conditioned mice initially showed weight loss and GvHD scores comparable to CY-TBI mice, they began to improve about two weeks post-BMT, while CY-TBI treated mice continued to deteriorate (Figure 2A and 2B). In agreement with the clinical GvHD scores seen early after BMT, pathologic examination of liver, skin, and intestines demonstrated comparable histological evidence of GvHD (representative images shown; Figure S4). As mice receiving CY conditioning die from GvHD by 3–4 weeks after BMT, we were not able to compare differences in histological GvHD between the two groups at later time points. To confirm that the clinical signs used to evaluate morbidity were actually GvHD rather than conditioning regimen related toxicity, additional mice received BEN-TBI or CY-TBI and syngeneic BM and SC. These mice did not display clinical or histologic evidence of GvHD (Figure S5).

Figure 2. BEN-TBI results in decreased GvHD morbidity.

BALB/c mice received 40 mg/kg BEN or 200 mg/kg CY on day −2, 400 cGy TBI on day −1, and 10⁷ BM with 3×10⁶ SC from C57BL/6 mice on day 0. Mice were weighed and clinically scored twice a week. (A) The mean % weight change from the starting weight with SEM is shown. (B) The weekly average of the mean clinical GvHD score per group is shown with SEM. Representative data from 6 experiments are shown, n=5 mice/group.

BEN-TBI conditioning results in fewer donor T-cells and more Gr-1⁺ cells in the intestines early post-BMT

Increased intestinal T-cell infiltration has been correlated with more severe GvHD^44,45. Additionally, increased MDSC frequency in the intestines has been associated with reduced GvHD⁴⁶. BEN-TBI resulted in a trend towards more Gr-1⁺ cells in the small and large intestines compared to CY-TBI early after conditioning (Figure 3A). We also demonstrated a significantly higher frequency of donor T-cells in the large and small intestines of CY-TBI conditioned mice early post-BMT (Figure 3B,C). A negative control for CD45.1⁺ intestine infiltration is shown in Figure S6. Conversely, early post-BMT, significantly more Gr-1⁺ cells were present in the intestines of BEN-TBI conditioned mice. Compared to intestines from naïve mice, both BEN-TBI and CY-TBI treated mice showed increased Gr-1⁺ cells, which includes both Ly6G⁺ and Ly6C⁺ cells, consistent with reports that allogeneic BMT leads to increased numbers of Ly6G⁺ cells in the gut⁴⁷ (Figure 3D). In summary, early post-BMT, BEN-TBI conditioned mice have a lower T-cell to MDSC ratio, which is consistent with less GvHD. Additionally, on day +14, more Gr-1⁺ cells were detected in the intestines of BEN-TBI conditioned mice, indicating an enduring effect from the conditioning regimen (Figure 3E).

Figure 3. BEN-TBI conditioning results in fewer donor T-cells and more Gr-1⁺ cells in the intestines early post-BMT.

(A) BALB/c mice received 40 mg/kg BEN or 200 mg/kg CY on day −2 and 400 cGy TBI on day −1. On day +4, intestines were collected for analysis by immunofluorescence. Intestines were stained for Gr-1 and the number of Gr-1⁺ cells per area of intestines was quantified. Representative images taken at 20X are shown (DAPI blue, Gr-1 red). Pooled data from 2 experiments are shown, n=7–8 mice/group. (B-D) BALB/c mice received 40 mg/kg BEN or 200 mg/kg CY on day −2, 400 cGy TBI on day −1, and 10⁷ TCD-BM from C57BL/6 mice with 3×10⁶ purified T-cells stained with CellTrace Violet from BoyJ (CD45.1⁺) mice on day 0. Large and small intestines were analyzed by flow cytometry. (B) Percentages of donor T-cells on day +3 in the leukocytes recovered from intestine digestion are shown. Pooled data from 2 experiments are shown, n=9–10 mice/group. (C) Percentages of donor T-cells on day +5 in the leukocytes recovered from intestine digestion are shown. Representative data from 2 experiments are shown, n=4 mice/group. Representative flow plots are shown as SSC versus CD45.1 (used to identify donor T-cells). (D) Percentages of Gr-1⁺ cells on day +5 in the leukocytes recovered from intestine digestion are shown. Naïve mice were used as a control. Pooled data from 2 experiments are shown, n=5–8 mice/group. (E) BALB/c mice received 40 mg/kg BEN iv or 200 mg/kg CY ip on day −2 and 400 cGy TBI on day −1, and 10⁷ BM and 3×10⁶ SC from C57BL/6 mice on day 0. Intestines were harvested on day +14 and stained for Gr-1. Pooled data from 2 experiments are shown, n=7–8 mice/group. * p<0.05, ** p<0.01

BEN-TBI conditioning leads to more granulocytic MDSCs in the BM, blood, and spleen

MDSCs are also very important when present in blood and spleen post-BMT and have been shown to suppress T-cell activation²⁹, reduce T-cell infiltration of the intestines⁴⁸, and attenuate the overall impact of GvHD²⁹. Five days following conditioning, BEN-TBI resulted in a higher number of host CD11b⁺Gr-1^high cells (granulocytic MDSCs; gating shown in Figure 4A) in the BM, blood, and spleen when compared to CY-TBI (Figure 4B–D). More CD11b⁺Gr-1^high cells were detected in the spleen on day 0, prior to transplant (Figure 4E). While BEN-TBI conditioned mice had more granulocytic MDSCs than those receiving CY-TBI, both groups had reduced numbers in the bone marrow, blood, and spleen compared to the number of CD11b⁺Gr-1^high cells found in naïve mice (Figure 4B,C,E). This indicates that BEN-TBI preserves the myeloid compartment, specifically the granulocytic MDSCs, more so than CY-TBI. The same difference was not seen in the monocytic subset (data not shown). Additionally, on day +7 post-BMT, host MDSCs in the blood were more proliferative by Ki-67 expression in BEN-TBI conditioned mice and donor MDSCs were more abundant in the spleens of BEN-TBI conditioned mice (Figure 4F,G), indicating enduring differential effects of BEN on MDSCs. Of note, CY-TBI mice had more neutrophils in the blood on day +7, coinciding with a higher white blood cell count, with no differences at later time points, as determined by complete blood count analysis (Figure S7).

Figure 4. BEN-TBI conditioning leads to more granulocytic MDSCs in the BM, blood, and spleen.

(A) Representative flow cytometry gating of the CD11b⁺Gr-1^high cell population is shown. (B-D) BALB/c mice received 40 mg/kg BEN or 200 mg/kg CY on day −2 and 400 cGy TBI on day −1. Bone marrow, blood, and spleens were collected on day +4. Naïve mice were used as a control. Bone marrow (B), blood (C), and spleen (D) were analyzed by flow cytometry and CBCs and percentages of CD11b⁺Gr-1^high cells were used to calculate absolute numbers. Pooled data from 2 experiments are shown, n=7–8 mice/group. (E) Spleens were also collected on day 0 and CD11b⁺Gr-1^high cells were isolated by Miltenyi kit. Number of cells isolated are shown. Pooled data from 3 experiments are shown, n=11–12 mice/group. (F-G) BALB/c mice received 40 mg/kg BEN iv or 200 mg/kg CY ip on day −2 and 400 cGy TBI on day −1, and 10⁷ BM and 3×10⁶ SC from C57BL/6 mice on day 0. On day +7, blood and spleen cells were analyzed by flow cytometry. (F) The average Ki-67 expression by mean fluorescence intensity (MFI) is shown with SEM for host CD11b⁺Gr-1^high cells in the blood. Pooled data from two experiments are shown, n=7 mice/group. (G) The numbers of splenic CD11b⁺Gr-1^high cells are shown. Pooled data from 2 experiments are shown, n=9 mice/group. * p<0.05, ** p<0.01.

BEN-TBI conditioning does not result in altered granulocytic MDSC function compared to CY-TBI

Given differences in MDSC number in BEN-TBI compared to CY-TBI conditioned mice, we next sought to evaluate potential differences in MDSC function. Splenic CD11b⁺Gr-1^high cells from each group suppressed T-cell proliferation, confirming their identity as MDSCs (Figure S8). No difference in suppressive function was observed on day 0 before transplant (host MDSCs) or days +7 (mixed chimerism MDSCs) and +14 (donor MDSCs) post-BMT (Figure 5A). We next investigated the mechanisms underlying MDSC suppressive function. MDSCs are able to suppress T-cell proliferation using a variety of mechanisms, including the expression of arginase-1 (arg-1)⁴⁹, inducible nitric oxide synthase (iNOS)^50,51, and indoleamine 2, 3-dioxygenase (IDO)⁵² and ROS production⁵³. No difference was detected in ROS production by MDSCs isolated between BEN-TBI and CY-TBI conditioned mice on day 0 (Figure 5B). On the day of transplant and days +7 and +14 post-BMT, no significant differences in mRNA levels of iNOS were observed between BEN-TBI and CY-TBI conditioned mice (Figure 5C). Arg-1 was not detectable in naïve CD11b⁺Gr-1^high cells or in most day 0 MDSC samples (data not shown). On days +7 and +14, differences in arg-1 levels, though trending, were not significant (Figure 5D). IDO was only consistently detectable in MDSC samples on day +7 post-BMT, where BEN-TBI and CY-TBI conditioned mice both demonstrated similarly elevated levels of IDO mRNA compared to naïve mice (Figure 5E). In summary, we were unable to provide clear evidence of differences in the cell-intrinsic suppressive function of MDSCs between BEN-TBI and CY-TBI conditioned mice.

Figure 5. BEN-TBI conditioning does not result in altered granulocytic MDSC function compared to CY-TBI.

BALB/c mice received 40 mg/kg BEN or 200 mg/kg CY on day −2, 400 cGy TBI on day −1, and 10⁷ BM with 3×10⁶ SC from C57BL/6 mice on day 0. (A) Splenic MDSCs were isolated and plated at various ratios in a suppression assay with CellTrace Violet stained CD3/CD28 bead-activated C57BL/6 T-cells. Proliferation was assessed by flow after 3 days of co-culture. Average percent proliferation with SEM is shown, compared to the control of no MDSCs. Representative data from 2 experiments are shown, n=4 mice/group. (B) ROS production by day 0 MDSCs was measured by flow cytometry is shown. Pooled data from 2 experiments are shown, n=8 mice/group. (C,D,E) Expression of iNOS, arg-1, and IDO in cDNA generated from MDSCs isolated on day 0 were assessed by qRT-PCR. Average expression fold compared to CY-TBI CD11b⁺Gr-1^high cells (C,D) or CD11b⁺Gr-1^high cells from naïve mice (E) is shown with SEM. Representative data from 2 experiments are shown, n=4 mice/group.

The beneficial effects of BEN-TBI rely on Gr-1^high cells

To further evaluate the role of MDSCs in GvHD protection following BEN-TBI, we depleted Gr-1^high cells pre- and early post-BMT, eliminating residual host MDSCs, as well as early donor MDSCs. Depletion was confirmed in the blood and spleen on the day of BEN or CY administration and transplant (data not shown). As shown in Figure 6A, on day +7 in the blood, >99% of CD11b⁺Gr-1^high cells (granulocytic MDSCs) were depleted, while monocytic MDSCs (CD11b⁺Gr-1^mid cells) were spared. Gr-1^high cell depletion significantly decreased survival from GvHD in BEN-TBI conditioned mice. No significant change in survival was seen in CY-TBI treated mice (Figure 6B). This indicates that pre- and early post-BMT Gr-1^high cells are required for BEN-TBI conditioning to suppress GvHD.

Figure 6. The beneficial effects of BEN-TBI rely on Gr-1^high cells.

BALB/c mice received 40 mg/kg BEN or 200 mg/kg CY on day −2, 400 cGy TBI on day −1, and 10⁷ BM and 3×10⁶ SC from C57BL/6 mice on day 0. On days −3, −1, and +5, the appropriate groups received 200 μg of anti-Gr-1 monoclonal antibody ip. (A) Successful depletion of CD11b⁺Gr-1^high cells was confirmed in the blood by flow on day +7. Representative flow plots are shown. (B) Survival data are shown. Pooled data from 2 experiments, n=8 mice/group. BEN vs. CY: p<0.0001, BEN vs. Gr-1 dep + BEN: p=0.045, Gr-1 dep + BEN vs. Gr-1 dep + CY: p=0.09, CY vs. Gr-1 dep + CY: p=0.106, Gr-1 dep + BEN vs. CY: p=0.1234.

G-CSF administration accentuates the difference in GvHD between BEN-TBI and CY-TBI

G-CSF administration has been shown to expand the Gr-1⁺ myeloid cell compartment^28,54,55. Given that Gr-1 depletion exacerbated GvHD, we sought to determine if G-CSF administration would further alleviate GvHD. G-CSF administration resulted in an overall increase in CD11b⁺Gr-1⁺ cells in the blood on day +7. Both monocytic MDSCs (Gr-1^mid) and granulocytic MDSCs (Gr-1^high) appeared to increase but only the difference in monocytic was significant (Figure 7A). G-CSF administration resulted in a trend towards increased survival in BEN-TBI conditioned mice, no observed effect in CY-TBI treated mice, and a widening of the survival gap between BEN-TBI and CY-TBI (Figure 7B). This points to a synergistic effect of BEN and G-CSF and supports the role of MDSCs in improved GvHD survival in BEN-TBI conditioned mice. GM-CSF and G-CSF levels, both of which can contribute to expansion or survival of these cells^28,54,56, were evaluated in plasma. GM-CSF was not detectable by our methods and higher levels of G-CSF were not seen in BEN-TBI conditioned mice (data not shown). We also investigated the transcriptional expression of G-CSF and GM-CSF receptors in the bone marrow and splenic MDSCs, to determine if increased receptor expression in BEN-TBI mice could explain the increased numbers of MDSCs. No difference was seen in receptor expression in bone marrow (Figure S9A) or in splenic MDSCs isolated on day +7 or +14 (data not shown). On day 0, BEN-TBI conditioned MDSCs showed increased mRNA levels of Csf2rb, indicating a potential increased sensitivity to GM-CSF (Figure S9B). This could contribute to the survival or expansion of MDSCs. Bone marrow cells were also evaluated for levels of IFR8 (a negative regulator of MDSC development)^57–59 and PU.1 (a transcription factor upregulated during myeloid lineage commitment)^60,61 mRNA, revealing no difference between the groups (data not shown).

Figure 7. G-CSF administration accentuates the difference in GvHD between BEN-TBI and CY-TBI.

BALB/c mice received 40 mg/kg BEN or 200 mg/kg CY on day −2, 400 cGy TBI on day −1, and 10⁷ TCD-BM with 3×10⁶ purified T-cells (tT) from C57BL/6 mice on day 0. Appropriate groups received 250 μg/kg G-CSF subcutaneously on days −2 through +11. (A) Expansion of the CD11b⁺Gr-1⁺ population was confirmed by flow. Representative flow plots and gating of high and mid populations are shown, as well as average absolute numbers of these cells in the blood on day +7. (B) Survival is shown. Pooled data from 2 experiments, n=9 mice/group. BEN vs. CY: p=0.0084, BEN vs. G-CSF + BEN: p=0.068, G-CSF + BEN vs. G-CSF + CY: p=0.0003, CY vs. G-CSF + CY: p=0.26.

DISCUSSION

GvHD remains a serious and often fatal complication of HCT, despite the application of numerous prophylactic and therapeutic immunosuppressive approaches post-BMT. Pre-transplant conditioning damages host tissues, resulting in release of inflammatory stimuli that promote activation of antigen-presenting cells, setting the stage for development of GvHD⁶². CY-TBI is a commonly used myeloablative regimen associated with significant tissue damage and GvHD^1,63. Despite the importance of conditioning in the pathophysiology of GvHD, research on the effects of specific preparative chemotherapeutic agents on GvHD induction has been limited. We, therefore, sought to examine whether replacing CY with BEN in TBI-based conditioning would alter GvHD. We chose to focus on BEN due to our previous report revealing its immunomodulatory effects on MDSCs when given post-transplant²⁴. Although BEN has successfully been incorporated into chemotherapy-based preparative regimens previously, we provide the first experimental evidence that substituting BEN for CY in conjunction with TBI may have significant advantages in reducing GvHD morbidity and mortality. This approach may provide a safer alternative for patients requiring TBI as part of their conditioning.

MDSCs, including the Gr-1^high granulocytic and the Gr-1^mid monocytic subsets, are recognized as important immunosuppressive cell populations in the control of GvHD^29,48. In allogeneic murine BMT models, adoptive transfer of MDSCs generated in vivo or in vitro can reduce GvHD mortality and morbidity^26–29. Furthermore, donor grafts with a higher proportion of MDSCs are correlated with reduced acute GvHD in human allogeneic HCT^30,31. We show that BEN-TBI results in increased numbers of MDSCs, particularly the granulocytic subset, in the spleen, BM, blood, and intestines compared to CY-TBI (Figure 4). We additionally show decreased donor T-cells in the intestines of BEN-TBI mice compared to CY-TBI (Figure 3). This increased number of suppressive CD11b⁺Gr-1⁺ cells, particularly in target GvHD organs, may play at least a partial role in the improved survival of BEN-TBI mice compared to CY-TBI conditioned mice. Although we were not able to show any difference in the suppressive activity of MDSCs between the two groups on a per cell basis, the increase in number of MDSCs in multiple tissues following BEN-TBI conditioning can account for the decrease in GvHD. It has been shown that the ratio of MDSCs to T-cells at the time of engraftment predicts the development of GvHD, particularly in the gut⁴⁸, which is where we found a clear increase in the MDSC to T-cell ratio in BEN conditioned mice (~20 Gr-1⁺ cells to 1 donor T-cell compared to 4 to 1 with CY-TBI) (Figure 3). The importance of MDSCs in BEN-TBI mice was further confirmed using Gr-1 depleting antibodies. When Gr-1^high cells were depleted, the survival of BEN-TBI conditioned mice significantly decreased (Figure 6). This suggests that granulocytic MDSCs are required for the beneficial effects of BEN-TBI conditioning, and importantly, that the monocytic MDSCs alone, which were not depleted, are not sufficient for the BEN-induced suppression of GvHD. These results should be interpreted with some caution due to the lack of an isotype control in the in vivo experiments.

Substantiating these observations, MDSC expansion via G-CSF administration resulted in a trend towards increased survival in BEN-TBI mice and widened the survival gap between BEN and CY (Figure 7). Importantly, G-CSF expanded both the monocytic and granulocytic subsets of MDSCs, with a greater effect on the monocytic subset, while our data indicate that the MDSC-related effects of BEN are primarily attributed to the granulocytic subset. G-CSF is commonly given to patients following HCT to increase neutrophil counts⁶⁴. Though retrospective clinical studies have not shown a reduction in GvHD with G-CSF administration^65,66, the synergistic beneficial effect of G-CSF and BEN-TBI conditioning in our murine model warrants further investigation as another potential advantage BEN-TBI conditioning may offer clinically.

We observed a moderate increase in the GM-CSF receptor in BEN-TBI treated MDSCs, which may contribute to the preservation of these cells (Figure S9B). As it has been shown that certain cytokines, including IL-6, VEGF, Flt3L, and M-CSF^67–69, can also promote MDSC expansion or survival, additional studies are needed to further elucidate how BEN-TBI is affecting these factors and promoting increased MDSC numbers.

We saw no difference in the suppressive functions of granulocytic MDSCs by T-cell suppression assays, ROS production, and mRNA levels of arg-1, iNOS, and IDO. However, our investigation of MDSC function was not exhaustive and other functional aspects need further investigation in the context of BEN-TBI. In addition to overall suppression of T-cell proliferation, MDSCs can confer antigen-specific tolerance^70,71. MDSCs have been shown to modify tyrosine residues in the T-cell receptor of CD8 T-cells, resulting in their inability to bind peptide-MHC, but allowing them to retain their ability to respond to nonspecific stimulation⁷⁰. Additionally, MDSCs have been shown to engage in bidirectional crosstalk with dendritic cells and macrophages, resulting in increased immunosuppression^72,73, partially through increased IL-10 production. BEN has been shown to increase IL-10 production specifically in B-cells⁷⁴, supporting the notion that BEN-TBI creates an anti-inflammatory cytokine milieu. We have not yet explored how BEN-TBI may be impacting the induction of antigen-specific tolerance by MDSCs or the crosstalk of MDSCs with other cell types. As MDSCs have complex, multifaceted functions and means of expansion and survival, much remains to be elucidated on how BEN-TBI is affecting CD11b⁺Gr-1^high granulocytic MDSCs and their interactions with other immune cells. While our data indicate that MDSCs are playing an important role in the mechanism by which BEN-TBI limits GvHD, it is likely that other immune cells are playing salient roles as well, warranting further investigation.

Interestingly, recent publications by Zeiser et al. have highlighted the important role of Gr-1⁺ granulocytes in GvHD, particularly in the gut. These studies demonstrated that Ly6G⁺ cell depletion can mitigate GvHD in murine models^47,75. Granulocytes (defined by Gr-1 or Ly6G expression) are a functionally heterogeneous population comprised of both pro-inflammatory (as seen in these reports) and anti-inflammatory (as demonstrated by our data) cells. Our research findings add to the literature on this dynamic, complex subset of cells. Our data indicate that Gr-1^high cell depletion, comparable to Ly6G⁺ cell depletion, exacerbates GvHD when BEN-TBI is used as conditioning and has no effect on survival when CY-TBI is used (Figure 6). It is important to note that Zeiser et al. utilized conditioning regimens of TBI, CY and busulfan, and CY and fludarabine. They did not combine chemotherapy with TBI, offering a potential explanation for these different results, as the combination of CY and TBI has been shown to have synergistic tissue damaging effects⁷⁶. They also did not utilize BEN, which we believe differentially affects myeloid cells. Additionally, our two groups utilized different timing of depletion antibody administration. Moreover, Zeiser et al. observed that the effect of Ly6G⁺ granulocytes on GvHD was dependent on intestinal microbiota^47,75. How BEN-TBI may affect intestinal microbiota differently than CY-TBI and other conditioning regimens, thereby modifying the function and role of Ly6G⁺ cells, remains to be determined.

In summary, BEN-TBI is associated with reduced GvHD, in part due to differential effects of BEN compared to CY on CD11b⁺Gr-1^high myeloid cells. This conditioning regimen warrants further investigation as an alternative for patients requiring TBI-based conditioning for HCT and our laboratory has ongoing studies to evaluate the effect of BEN-TBI on GvL.