Immunopathogenic Mechanisms and Modulatory Approaches to Graft-versus-Host Disease Prevention in Acute Myeloid Leukaemia

Abstract

Allogeneic haematopoietic stem cell transplantation (HSCT) remains the only potential cure for intermediate to high-risk acute myeloid leukemia (AML). The therapeutic effect of HSCT is largely dependent on the powerful donor-derived immune response against recipient leukaemia cells, known as graft-versus-leukaemia effect (GvL). However, the donor-derived immune system can also cause acute or chronic damage to normal recipient organs and tissues, in a process known as graft-versus-host disease (GvHD). GvHD is a leading cause of non-relapse mortality in HSCT recipients. There are many similarities and cross talk between the immune pathways of GvL and GvHD. Studies have demonstrated that both processes require the presence of mismatched alloantigens between the donor and recipient, and activation of immune responses centered around donor T-cells, which can be further modulated by various recipient or donor factors. To dissect GvL from GvHD for more effective GvHD prevention and enhanced GvL has been the holy grail of HSCT research. In this review, we focused on the key factors that contribute to the immune responses of GvL and GvHD, the effect on GvL with different GvHD prophylactic strategies, and the potential impact of various AML relapse prevention therapy or treatments on GvHD.

Introduction

The application of allogeneic haematopoietic stem cell transplantation (HSCT) in human was first reported in 1957 based on the discovery in mice that allogeneic immune cells could exert powerful anti-leukemic effects, i.e., graft-versus-leukaemia (GvL) [1, 2]. Since then, major discoveries in human immune system and leukaemia biology have greatly improved transplant outcomes and molded the modern-day practice of HSCT [3, 4]. To date, HSCT remains the only potentially curative treatment for eligible patients with acute myeloid leukaemia (AML) that harbors adverse clinical, cytogenetic or molecular risk factors [5, 6]. AML is the most common indication for HSCT in the US and worldwide [7, 8]. Still, post-transplant relapse occurs in 15–60% of patients and carries a dismal prognosis [9]. To prevent relapse, it is crucial to maintain or enhance GvL early after HSCT.

Graft-versus-host disease (GvHD) is a unique side effect of HSCT. In GvHD, the donor immune system elicits an immune response towards the allogeneic donor tissue, causing a myriad of symptoms such as acute rash, diarrhea, and liver injury weeks to months post-HSCT (acute GvHD, aGvHD), and/or chronic malabsorption, sclerotic skin, sicca, and obstructive lung disease months to years later (chronic GvHD, cGvHD) [2, 10–12]. Even with declining incidence and better treatment, GvHD remains the leading cause of non-relapse mortality (NRM) post-HSCT [13–19]. Although both GvL and GvHD are dependent on sustained immune reconstitution of the donor immune system, the presence of GvHD does not always translate into stronger GvL and reduced relapse incidence (RI), suggesting differences in the immune responses which could be targeted to GvL without inducing GvHD, or reducing GvHD without dampening GvL [20, 21]. The immune responses that are implicated in GvL and GvHD are yet to be fully elucidated. Here we summarize the key components of GvL/GvHD immune responses and the mechanisms of action of selected GvHD prophylactic therapy and relapse-prevention/treatment regimens with a focus on AML.

Histocompatibility antigens

Human leukocyte antigens

The human leukocyte antigen (HLA) system comprises the main antigens involved in GvHD and GvL. HLAs are highly polymorphic and presented on all nucleated cells (class I) or APCs (class II), triggering CD8+ T-cell and CD4+ T-cell-dependent immune response, respectively [22]. Recognition of mismatched recipient alloantigens on leukaemia cells and the subsequent immune response centered around donor mature CD8+ cytotoxic T-cells is the key to GvL [23]. By losing the mismatched HLA, AML can evade GvL and relapse [24]. Alloantigens also participate in GvHD. Unlike pathogen-specific immunity, GvHD immune response is characterized by ubiquitously present alloantigen, does not require antigen presenting cells (APCs) maturation, and the immune response is constant as the alloantigens cannot be eliminated [25]. Higher disparity of HLA epitope between the donor and recipient theoretically enhances donor-derived immune reaction that leads to both GvL and GvHD.

Clinically, the effect of HLA-disparities on transplant outcome can be modulated by many other donor/recipient factors, conditioning and GvHD prophylaxis. As an example, haploidentical related donor (HRD) HSCT, once associated with prohibitive risks of severe GvHD due to high level HLA disparity, is widely used now owing to the development of modern GvHD prophylaxis strategies [26, 27]. Various studies examined the effect of HLA-matching on the outcome of HSCT in AML with some conflicting results. A multicenter study from China compared the outcome of HRD (n = 231) with HLA-identical matched sibling donor (MSD, n = 219) HSCT in patients with AML in 1^st complete remission (CR1), and found no significant difference in DFS, OS, incidence of relapse and non-relapse mortality (NRM) between the two, but the incidence of aGvHD grade II-IV/III-IV and cGvHD was much higher with HRD [28]. The EBMT conducted several large retrospective studies comparing donor type in different scenarios. In high-risk AML in first relapse, MUD HSCT was associated with better LFS and lower relapse incidence compared with MSD [29]. In AML in CR1 with previous induction failure, 10/10 MUD, HRD and MSD achieved similar RI, LFS and OS, but GvHD-free relapse-free survival (GRFS) was worse in 9/10 mismatched MUD HSCT [30]. In high-cytogenetic risk AML in CR1, HRD was associated with a higher incidence of GvHD, lower RI, and similar LFS, OS and GRFS compared with MSD, but in intermediate-risk AML MSD outperformed HRD owing to higher GvHD-NRM in the latter [31].

A few studies on optimal donor choice were controlled for GvHD prophylactic regimen. A CIBMTR study compared the outcome of 336 post-transplant cyclophosphamide (PtCy)-based HRD vs 869 calcineurin-inhibitor based MSD HSCT in AML in CR1, showing significantly lower incidence of cGvHD in the former (hazard ratio, HR 0.38), while OS, LFS, NRM, RI, incidence of grade II-IV aGvHD were similar [32]. When PtCy is used homogenously, another EBMT study showed that HRD HSCT (n = 789) carried a significantly higher risk of grade II-IV aGvHD (HR 1.6) and NRM (HR 2.6), lower RI (HR 0.7), similar LFS (HR 1.1) and GRFS (HR 1.0) compared with MSD (n = 215) and MUD (n = 235) HSCT for AML in CR1 [33]. A Spanish multicenter retrospective study where PtCy was uniformly used, however, demonstrated lower rates of aGvHD in MSD/MUD than HRD for patients with AML undergoing HSCT, while RI, NRM, OS, GRFS were similar [34]. Overall, the outcome of HRD is comparable to MSD and 10/10 MUD for high-risk AML patients.

Besides the degree of HLA mismatching, the effect of different HLA antigens on GvHD or GvL varies. For example, mismatch in class I antigens such as HLA-A or HLA-C is a strong risk factor for aGvHD, while class II antigen mismatch has a less prominent role in aGvHD [35]. Mismatches in HLA-A, -B, -C, and DPB1 were associated with grade III–IV aGvHD, mismatches in HLA-C with cGvHD, and both HLA-C and -DPB1 mismatches were associated with reduced RI [36]. The direction of HLA mismatch also matters. In HRD HSCT, HLA-DRB1 mismatches in the GvH direction are associated with lower NRM, and HLA-DPB1 non-permissive mismatches are associated with lower RI comparing to permissive mismatches. Further, mismatch in certain HLA single nucleotide polymorphisms (SNPs) may induce stronger T-cell alloimmunity and GvHD [37]. Deeper HLA sequencing to the SNP level may therefore reduce GvHD and improve disease-free survival (DFS) [38, 39]. Future endeavors in HLA research may direct towards finding more GvL-inducing and GvHD-reducing HLA-combinations to improve donor selection for optimal HSCT outcomes.

Non-HLA antigens

The most studied non-HLA antigens are the minor histocompatibility antigens (miHAs) [40]. The number of miHA in matched unrelated donor (MUD) is roughly twice as many as matched sibling donor (MSD) transplants, speaking to the greater genetic disparity of the former [41]. MiHA-mismatch is crucial to maintain CD8+ T-cell alloreactivity [42]. When pre-sensitized with donor antigen, CD4+ T-cells could develop immunotolerance across miHA barriers but CD8+ T-cells do not [43]. Infusion of miHA-specific CD8+ alloreactive T-cell clone could cause GvL, acute and chronic GvHD [44, 45]. Mismatch in non-HLA antigens, such as the SNP of Paired-immunoglobulin type 2-like receptor β, has been associated with increased transplant-related mortality (TRM) due to increased GvHD [46].

The ability of miHAs to induce GvHD and GvL depends not on the total number of miHA mismatches, but on the immunogenicity and the location of different miHAs [41, 47, 48]. As an example, miHA encoded by the Y-chromosome are associated with higher incidence of GvHD due to wider tissue distribution, while autosomal-encoded miHAs are associated with prolonged relapse-free survival (RFS) due to their more restricted distribution on haematopoietic cells [49, 50]. Immunodominance is an important phenomenon in which certain miHAs, such as HA-1, dominate alloreactivity, stimulate miHA-specific T-cell responses, and drive the presence of GvL or GvHD [50–52]. Variations in the presentation of miHAs by different MHC types also contribute to the clinical presentation of GvHD [53]. Identifying the haematopoietic tissue-specific miHA and the less immunogenic miHA or HLA-miHA couple could potentially alleviate GvHD and augment GvL. One such example is GRK4, a miHA highly expressed in AML and several other cancers, but has little to no expression in other normal tissues except the testis [41].

Proinflammatory T-cells

Sustained reconstitution of mature donor αβ T-cells is pivotal for both GvHD and GvL, demonstrated by a plethora of preclinical and clinical studies [43, 54–56]. Donor T-cell immune reconstitution is also essential for infection prevention and engraftment. Different T-cell subtypes exert distinctive effects in GvHD and GvL [54, 56–58]. The dose of CD3+ T-cells, especially the CD8+ subset, is predictive of the severity of aGvHD [59]. Among the CD4+ T-cells, naïve T-cells are more potent than memory T-cells for GvHD. In murine models, naïve T-cells induced a more severe acute GvHD than alloantigen-primed memory T-cells and had more vigorous homing activity to the colonic tissue and mesenteric lymph nodes [60]. Naïve T-cells induce the differentiation of miHA-specific alloreactive CD8+ T cells, whereas memory T-cells are more likely to generate pathogen-specific immune responses [61]. Unprimed memory T-cells lack the ability to induce GvHD due to an abortive immune response to alloantigen [62].

T-helper (Th)1 and Th17 cells are potent GvHD inducers and are derived from naïve T-cells upon stimulation by exposed alloantigens on tissues and altered intestinal microbiota signals [54, 63]. They produce inflammatory cytokines such as interferon(IFN)-γ, interleukin(IL)-17 and tumor necrosing factor (TNF)α and activate cytotoxic T-cells [64]. Murine studies showed that inhibiting IFN-γ could reduce gastrointestinal GvHD but increase Th17/IL-17 production and pulmonary damage, while IL-17 blockade would lead to Th1 proliferation, suggesting a reciprocal differentiation of Th1 and Th17 cells post-HSCT [65, 66]. Blocking the differentiation of both Th1 and Th17 cells could effectively prevent GvHD without affecting GvL [67]. Th22 cells are a recently defined Th subtype associated with inflammatory skin conditions such as scleroderma [68]. Th22 and its product IL-22 have been associated with the pathogenesis of cutaneous chronic GvHD [69]. Down regulation of IL-22 could attenuate skin GvHD while sparing the GvL effect [70]. Paradoxically, IL22 is protective in intestinal GvHD, highlighting the complexity of GvHD pathogenesis [71–73].

Targeting proinflammatory T-cells to prevent GvHD in AML

GvHD prophylactic regimens evolve around how to eliminate or reduce the function of proinflammatory T-cells. A T-cell replete peripheral blood (PB) graft carries a higher volume of mature donor T-cells which confers stronger GvL and GvHD effects. Studies across donor types and GvHD prophylaxis regimens invariably demonstrated faster engraftment, less infection, reduced RI, improved DFS, but higher risk of aGvHD and cGvHD using T-cell replete PB than bone marrow (BM) graft [74–77]. Non-selective ex vivo T-cell depletion (TCD) can significantly reduce the risk of GvHD but at the cost of significantly higher RI, risk of infection, graft failure, and post-transplant lymphoproliferative disease [78–80]. Delayed regeneration of T-cell dependent functions and reduced diversity of T-cell receptor (TCR) repertoire were observed in TCD BM HSCT [81, 82]. However, high intensity conditioning (myeloablative chemotherapy plus myeloablative total body irradiation) seemed to ameliorate the detrimental effect of TCD on relapse while preserving its anti-GvHD effect [83]. Selective TCD by ex vivo αβ TCD to less than 1×10⁵/kg αβ T-cells, frequently in combination with CD19+ B-cell depletion, has been studied mainly in pediatric patients [84]. The risk of aGvHD II-IV was seen in 3–39% and cGvHD in 0–30%. Relapse occurred in 22–58% of patients and was the main cause of death [84]. Naïve TCD has also been studied. In a total of 138 patients with acute leukaemia, naïve TCD lead to low incidence of grade III–IV aGvHD (4%) and moderate to severe cGvHD (1%), without apparent increase of RI or NRM [85].

Pharmacologic immunosuppressive therapy (IST) such as methotrexate (MTX), calcineurin inhibitors and mycophenolate mofetil (MMF) reduces T cell activation, proliferation and differentiation. ISTs have been widely used for GvHD prophylaxis since the inception of HSCT, with some studies suggesting higher RI with IST [86, 87]. Therefore, timely discontinuation of IST is important after HSCT for AML, especially those transplanted with measurable residual disease (MRD) or for those who represent MRD features after prior clearance. There is no consensus regarding the optimal timing of IST discontinuation or whether IST should be tapered, though most practices stop or taper IST around 60–100 days post-HSCT in the absence of GvHD.

Anti-thymocyte globulins (ATG) are polyclonal antibodies extracted from horse or rabbit sera, immunized with human thymic cells or T lymphoblastoid cell line [88]. ATG has been used for ex vivo and in vivo TCD. When used in vivo, both host T-cells and incoming allograft T-cells can be effectively reduced due to ATG’s long half-life [89]. ATG reduces GvHD by reducing proinflammatory T-cells, increasing regulatory T-cells, selective enhancement of NK-cells, inducing B-cell apoptosis and reducing APC function [90–93]. Among different ATG formulations, rabbit ATG is more potent than horse ATG in preventing GvHD, likely due to stronger T-cell depleting ability [94, 95]. ATG has been widely studied in MUD HSCT, and in HRD HSCT in China. Randomized and non-randomized studies have consistently showed less incidence and severity of acute or chronic GvHD, and no major change in RI in patients with haematologic malignancies receiving ATG for GvHD prophylaxis [89, 96–101].

PtCy has demonstrated a powerful ability to reduce GvHD in murine studies and was soon applied to human HRD HSCT [102–104]. PtCy not only reduces the function, proliferation and differentiation of alloreactive effector T-cells like IST, but also promotes the preferential recovery of CD4+ regulatory T-cells [105, 106]. In addition, PtCy restores early post-transplant B-cell homeostasis favoring naïve B-cells and suppresses activation of B-cells at peripheral lymph nodes, adding additional protection against cGvHD [107]. With PtCy, HRD HSCT now has comparable outcomes to non-PtCy based MSD HSCT, greatly enlarging the donor pool [32, 108]. Besides the HRD setting, PtCy-based GvHD prophylaxis has been increasingly applied to HLA-matched HSCT [109–112] ref.

Whether the use of PtCy affects GvL remains inconclusive. The occurrence of aGvHD or cGvHD after PtCy has not been associated with reduced RI, suggesting a dissociation between GvL and GvHD in this setting [113, 114]. The higher RI after PtCy observed in some studies may well be explained by the difference in conditioning intensity and not the PtCy [115]. When controlled for conditioning intensity, PtCy-based GvHD prophylaxis significantly lowered the incidence of both aGvHD and cGvHD compared with non-PtCy, without change in RI [110]. Patients who relapse after PtCy-based HSCT showed different T- and NK-cell activation signatures than those who did not relapse [116]. Several large, randomized studies compared PtCy-based GvHD prophylaxis with other GvHD prophylaxis methods. The BMTCTN 1301 study compared CD34-selected PB (n = 114) vs single agent PtCy BM (n = 109) vs tacrolimus/MTX BM HLA-matched HCT (n = 114), with around 60% of patients in each group transplanted for AML [111]. The PtCy group had the lowest RI compared to the other two groups (HR 0.52, p = 0.037), comparable OS to tacrolimus/methotrexate and better OS than CD34-selection [111]. The BMTCTN 1703 compared PtCy/tacrolimus/MMF vs tacrolimus/MTX for MSD, MUD, or 7/8 mismatched unrelated donor reduced intensity conditioning HSCT, in which about 50% of patients in each group had AML [112]. The RI and OS at 1-year were similar between the groups, but the PtCy group had better GRFS (HR 0.64, p = 0.001) due to lower incidence of aGvHD and cGvHD [112].

Post-transplant Bendamustine (PtB) has been studied as an alternative to PtCy as murine studies provided evidence that PtB was as effective as PtCy in mitigating GvHD, less myelosuppressive, and preserved GvL likely by increasing myeloid-derived suppressor cells [117]. Katsanis et al. reported a phase 1 study substituting day +4 PtCy with PtB. The PtCy/PtB combination was well-tolerated, had faster engraftment and lower incidence of CMV reactivation compared with contemporaneous controls using PtCy [118]. Moiseev et al. reported a dose-finding prospective clinical trial studying PtB in patients with refractory leukaemia (22 AML and 5 ALL) receiving myeloablative conditioning [119]. Compared with historical PtCy data, PtB was associated with a faster and more successful engraftment even in refractory leukaemia. As a single agent, PtB prevented aGvHD well but not cGvHD. Unexpectedly, PtB was associated with cytokine release syndrome that mimicked macrophage activation syndrome, which was not seen in pre-clinical models [119]. Ongoing studies on immunomodulatory GvHD prevention are summarized in Table 1.

Table 1.

Ongoing investigational immunomodulatory approaches in AML HSCT.

Immune target	Clinicaltrials.gov identifier #	Clinical Setting, if applicable	Country
miHA
MiHA-loaded PD-L-silenced DC Vaccination	NCT02528682		Netherlands
Ex vivo T-cell depletion
alpha-beta T-cell depletion, with CD19+ B cell depletion	NCT04337515, NCT02323867, NCT05794880 (with targeted ATG dosing), NCT04099966		United States
TCRαβ-depleted Progenitor Cell Graft with Additional Memory T-cell DLI	NCT03849651		United States
Naïve T cell depletion	NCT03779854	Pediatric	United States
Allogeneic CD34+-enriched and CD45RA-depleted PBSCs	NCT03970096		United States
Ex Vivo T-Cell Depletion of Mobilized PBSCs via CD34-Selection	NCT01189786		United States
Ex Vivo Expanded/Activated Gamma Delta T-cell Infusion	NCT03533816	After HRD with PtCy	United States
Orca-T: T-cell-Depleted Graft with Additional Infusion of Conventional T Cells and Regulatory T Cells	NCT05316701		United States
PT-Cy
PTCy + ATG	NCT04202835, NCT03357159, NCT03608059		Canada, Australia, Israel, China
PT-Cy vs. ATG	NCT04888741, NCT05153226		United Kingdom, Germany
PT-Cy + methotrexate	NCT04622956		Brazil
PT-Cy + PT-Bendamustine or PT-Bendamustine alone	NCT02996773, NCT04022239		United States
PT-Cy + Bortezomib	NCT03850366		United States
PT-Cy, Bortezomib + Abatacept	NCT05289167	MSD or ≥7/8 MUD	United States
PT-Cy, Abatacept + Tacrolimus	NCT05621759	HRD	United States
PT-Cy + uhCG/EGF	NCT04886726	MMUD	United States
PT-Cy + Itacitinib	NCT05364762, NCT05823571	8/8 MRD or MUD HRD	United States
PT-Cy + Ruxolitinib	NCT05622318		United States
PT-Cy + total marrow and lymphoid irradiation (TMLI)	NCT03467386	10/10 MRD or MUD	United States
PT-Cy, Abatacept, Vedolizumab + CNI	NCT05515029	Pediatric/AYA	Russia
PT-Cy dose de-escalation	NCT05436418, NCT04959175, NCT03983850, NCT05158608		United States Russia
DLI
T-cell Receptor α/β Depleted Donor Lymphocyte Infusions	NCT05350163		United States
Siremadlin Alone and in Combination with DLI	NCT05447663		United States, Israel, Italy, Spain, Germany
Regulatory T-cells
Treg-enriched donor cell infusion	NCT05095649 NCT04678401	SR-CGVHD Peri-HRD transplant	Spain United States
Purified regulatory T-cells (Treg) plus CD34+ HSPC	NCT05088356		United States
T-allo10 cells: T regulatory type 1 (Tr1) cells	NCT03198234	MMRD, MMUD	United States
NK Cells
Ex Vivo Expanded Allogeneic NK Cells Infusion	NCT05250362		China
Cytokine Induced Memorylike NK Cell Adoptive Therapy	NCT02782546	After HRD	United States
Expanded Natural Killer Cells	NCT03300492		Switzerland
Other Adoptive Cell Therapies
TSC-100, TSC-101 (donorderived T cells targeting HA-1 and HA-2, respectively)	NCT05473910	HLA A*02:01 positive patients, HRD	United States
Interferon γ-Primed Mesenchymal Stromal Cells	NCT04328714		United States
Immunomodulators for GVHD prophylaxis
Vorinostat	NCT03842696	Pediatric, AYA	United States
Tocilizumab	NCT03434730, NCT04688021		United States China
Itacitinib	NCT03755414		United States
Ponatinib	NCT03690115		France
Baricitinib	NCT04131738		United States
Sitagliptin	NCT05149365		China
Belimumab	NCT03207958		United States
RGI-2001	NCT04014790		United States
Alpha-1 Antitrypsin (AAT)	NCT03805789		United States, Australia, Germany, Spain, United Kingdom
High Dose Vitamin A	NCT03719092		United States
Tildrakizumab	NCT04112810		United States
Ustekinumab	NCT04572815		United States
Anti-CD40L Domain Antibody (BMS-986004)	NCT03605927		United States
Microbiota
Prebiotic Galactooligosaccharide	NCT04373057		United States
Oral Pooled Fecal Microbiotherapy (MaaT033)	NCT05762211		France
Sodium butyrate	NCT05808985		Korea
Fecal Microbiota Transplantation (FMT)	NCT04935684		France
Post-HSCT Maintenance/ Relapse Prevention
Low Dose Azacitidine	NCT01995578		United States
Glasdegib	NCT04168502		Italy
Tagraxofusp	NCT05233618	CD123+ AML	United States
Magrolimab + Azacitidine	NCT05823480		United States
Venetoclax + Azacitidine vs decitabine/cedazuridine	NCT03613532	After ven + conditioning	United States
Decitabine	NCT05528354	After ven + decitabine + conditioning	China
Azacitidine + Chidamide	NCT05270200		China

Regulatory T-cells

Regulatory T-cells (Treg) are important negative modulators of GvHD. They are crucial for immune tolerance and can alleviate GvHD by cytolysis and metabolic suppression of effector T-cells, modulating APC maturation and function, and producing immunosuppressive cytokines [120]. The number of Treg in the graft has been inversely associated with aGvHD preclinically and clinically [121]. Thymic injury that occurs due to peri-HSCT events including aGvHD, will impair negative selection, corrupt APC function, leading to Treg failure and ultimately cGvHD [53, 121]. Tregs have higher affinity to IL-2 than conventional CD4+ T-cells. In vivo expansion of Treg using IL-2 can restore Treg and alleviate cGvHD symptoms in small phase 1/2 clinical trials for patients with GvHD [122–124]. However, higher leukaemia relapse incidence was seen in small clinical trials using low dose IL-2 as GvHD prophylaxis [125, 126]. Treg infusion trials have not shown convincing efficacy in preventing GvHD [127, 128]. The disappointing results of Treg-enhancing therapy could be explained by the plasticity of late-stage Treg such as Treg to Th-17 conversion triggered by APCs [129]. Belumosudil, a ROCK2 inhibitor FDA-approved for refractory cGvHD, exerts its therapeutic effect partly through shifting the Th17/Treg balance towards Treg [130].

Antigen-presenting cells

Host APCs participate in T-cell-dependent aGvHD [131]. Host APCs are estimated to be 100–1000 times more potent than donor APCs in inducing aGvHD via HLA class II. Among host APCs, the tissue-resident non-haematopoietic APCs such as intestinal epithelial cells are more potent than professional APCs in the lymphoid organs [132, 133]. Conditioning-induced tissue damage and microbiota disruption are important activators of APCs [134, 135]. In mice, depletion of host macrophages and dendritic cells (DC) from the liver and spleen could reduce recruitment of donor CD8+ T-cells to the liver and inhibit liver aGvHD [136]. In the gut, the propagation of aGvHD is dependent on a profound feed-forward loop between host DC and naïve donor DC [137]. The antigen presentation by donor DC subsequently triggers the migration of proinflammatory Th1/Th17 to the gut to cause GvHD [138]. Despite the prominent role of host APC in GvHD, depletion of host APC cannot prevent or ameliorate GvHD, given the fact that all host tissues harbor alloantigens and the existing redundancy of antigen-presentation in GvHD immune response [139].

Donor APCs intensify GvHD but are not essential for GvL [140]. Donor DC creates a feed-forward cascade by antigen presentation and cytokine release in GvHD [137]. The level CD123+ plasmacytoid DC in the graft is a strong predictor of aGvHD [59]. Depletion of donor CD11b- DC from the BM graft led to significant expansion of donor CD4+ memory T-cells, enhanced GvL without producing GvHD in a murine BMT model [141]. Another way to exploit donor DC in the setting of AML is to selectively enrich the leukaemia antigen-specific DC, which was shown to enhance GvL without causing GvHD in a postallograft relapse animal model [142].

Humoral immunity

B-cells are involved in both acute and chronic GvHD [107, 143]. Higher B-cell content in the graft is associated with higher risk of aGvHD [59]. In cGvHD, B-cells often express exhaustion markers such as CD22 [144]. Patients with cGvHD often have elevated immunoglobulin and auto/alloantibody levels, but the high immunoglobin levels do not translate into effective antimicrobial response [145]. Several B-cell markers have been proposed as biomarkers for cGvHD, such as plasma B-cell activating factor (BAFF) level, BAFF/CD19+ ratio, and level of CD19⁺CD21^low transitional B-cells [146–148]. Various B-cell targeting agents have been applied to the treatment of cGvHD, highlighting the importance of B-cells in the pathogenesis [149–151]. Although B-cells may participate in GvL by producing anti-miHA antibodies, the role is less prominent than in GvHD [152]. Treating cGvHD with B-cell targeting agents such as ibrutinib and fostamatinib did not increase the incidence leukaemia relapse [150, 151].

Innate immune cells

Innate immune cells are promising therapeutics for AML due to their unique properties of inducing GvL without exacerbating GvHD, or even reducing GvHD. Natural killer (NK)-cells are the first donor-derived lymphocytes to recover after HSCT [153]. Timely NK-cell recovery is associated with decreased incidence of leukaemia relapse, viral reactivation and GvHD [154]. In a landmark study by Ruggeri et al., HSCT with NK-cell alloreactivity, compared with those without, had significantly lower incidence of AML relapse (0 vs 75%, p < 0.0008), graft rejection (0 vs 15.5%, p ≤ 0.01) and aGvHD grade ≥ 2 (0 vs 13.7%, p ≤ 0.01) [155]. In patients receiving PtCy, timely NK-cell recovery is predictive of improved OS, lower OS and RI [116]. Donor memory-like NK cells have potent anti-AML effects when given to patients with post-transplant relapse and did not induce GvHD [156, 157]. Both donor and recipient NK-cells can down-regulate GvHD. Donor NK-cells can eliminate recipient APCs and prevent donor T-cell activation upon recognition of non-self HLA ligands by surface killer immunoglobulin-like receptors (KIRs) [155, 158]. Donor NK-cells suppress the number and activity of alloreactive T-cells via perforin- and Fas ligand-dependent pathways [159]. Delayed NK-cell recovery, especially the CD56-high CD16-dim naïve NK-cells, is associated with increased incidence and severity of aGvHD [160, 161]. Recipient NK-cells, on the other hand, can reduce alloreactive T-cell expansion in target organs. Pre-HSCT conditioning may promote GvHD by eliminating recipient NK-cells and disrupting the intestinal barrier [162].

Γδ T-cells are an integral part of both the innate and adaptive immune system [163]. They comprise the majority of residential T-cells in the skin and mucosa [164]. Unlike αβ T-cells, γδ T-cells usually do not express the CD4 or CD8 coreceptors and undergo activation in an HLA-unrestricted manner [163]. As a result, γδ T-cells kill leukaemia cells via direct recognition of surface antigens independent of HLA [165]. Many studies have demonstrated that γδ T-cells are crucial for GvL and anti-microbial immune responses [164, 166, 167]. In one study of 153 patients with acute leukaemia, patients with high vs low/normal post-HSCT γδ T-cells had better 5-year LFS (54.4 vs 19.1%, p<0.0003) and OS (70.8 vs 19.6, p<0.0001), there were no difference in GvHD risk [168, 169]. In AML patients with pre-transplant MRD, early γδ T-cell recovery reduced the risk of relapse by 60% and significantly improved LFS and OS [170]. The beneficial effect of high γδ T-cell levels on survival may also be attributed to reduced incidence of infection as a result of more robust immune reconstitution [171]. Several studies have suggested a limited role of γδ T-cells in GvHD pathogenesis. A meta-analysis of 11 studies (919 patients) showed that high γδ T-cell levels post-HSCT was associated with less relapse, fewer viral infections, higher overall and disease-free survivals, and no association with aGvHD [172]. Several clinical trials that sought to increase γδ T-cell levels post-HSCT by in vivo expansion or infusion of activated γδ T-cells are currently underway, with promising preliminary results showing prolonged survival and low incidence of GvHD [165, 173].

Donor clonal haematopoiesis

Several retrospective studies demonstrated the possible connection between donor clonal hematopoiesis (CH), relapse, and the development of acute or chronic GvHD [174–177]. Specifically, mutations in DNMT3A, TET2, and ASXL1 are associated with a 2-fold and 2 to 8-fold increase in the incidence of grade II–IV and grade III–IV aGvHD, respectively, likely due to enhanced production of proinflammatory cytokines and Th17 in the presence of these mutations [174]. Donor DNMT3A-CH is also associated with reduced RI and increased cGvHD likely by augmenting both GvL and GvHD through proinflammatory T-cell cytokines [176]. Donor CH carries the risk of donor-derived myeloid malignancies [176, 178]. These data suggest a potential benefit to screen for donor CH prior to transplant especially in older donors where CH is more prevalent.

Microbiome

The fact that organs/tissues with microbiota are mostly affected by GvHD is not a mere coincidence but suggestive of the importance of dysbiosis in the pathogenesis of GvHD. There is also growing evidence that microbiota are associated with anti-leukaemia immunity. Disrupted gut microbiota may reduce anti-leukaemia immunity by altered secretion of bacterial metabolites [179]. A phase 3 clinical trial demonstrated that gut-decontamination with oral metronidazole could lower the risk of grade II–IV aGvHD [180]. A small phase 2 study showed that gut decontamination with oral vancomycin-polymycin B did not significantly change the gut microbiome diversity, T-cell or B-cell subset reconstitution, and clinical outcomes, but reduced the risk of blood stream infection [181]. The diversity of gut microbiome is inversely correlated with TRM and the risk of GvHD [182]. Specifically, elimination of certain Clostrida species is associated with the development of GvHD [183]. The abundance of certain species, such as Blautia, has been associated with lower GvHD-related mortality [182]. Small case series demonstrated that restoration of gut microbiome diversity and symbiosis by fecal microbiota transplantation may be useful in severe intestinal SR-GvHD [184]. Similarly, decreased skin microbiome diversity has been observed in cutaneous GvHD [185]. Various peri-HSCT events can affect the health of human microbiome, such as pre-transplant chemotherapy, conditioning regimens, and antibiotics. Abnormal microbiome can cause reduced production of molecules such as short chain fatty acids, decrease bile acid clearance, thereby impairing endothelial repair [134]. Bacterial antigens could trigger DC activation and subsequent recruitment of proinflammatory T-cells, eventually causing GvHD [186]. Besides bacteria, fungal elements may also recruit proinflammatory T-cells and play a role in GvHD [187]. It is plausible that better preservation of microbiome homeostasis during AML treatment would improve immune reconstitution, leading to lower GvHD and relapse risk.

In summary, the mechanism of GvHD is a complicated process and has frequent overlap with GvL (Figure 1). The mismatched donor alloantigens and other antigens such as altered microbial elements, activate GvHD immune response centered around pro-inflammatory donor T-cells. The immune response attacks recipient tissues and organs, which are already vulnerable after chemotherapy, radiation, infection and other insults, leading to GvHD. Therefore, the prevention of GvHD should aim at reducing recipient tissue damage, antigen mismatching, and immune reactivation especially proinflammatory donor T-cells, while preserving the strong anti-leukaemia effect exerted by the graft.

Figure 1.

Key factors that intersect with the immunopathogenesis of GvHD and GvL in alloHSCT for AML

Post-transplant therapy and GvHD

Post-HSCT relapse of AML remains one of the main causes of treatment failure and is associated with an extremely poor prognosis, with 1-year OS after relapse of 20–25% [188]. Multiple evasion mechanisms have been identified in AML that help the leukaemia cells escape immune surveillance by donor-derived immune system [189]. To prevent relapse, various post-transplant therapies are being studied retrospectively or in a clinical trial setting (Table 1). The goal of post-HSCT therapy is to suppress residual AML, prevent immune evasion, while avoiding the induction of GvHD.

Preemptive and prophylactic DLI

In patients with mixed chimerism, positive MRD, or at high risk of leukaemia relapse, donor lymphocyte infusions (DLI) given preemptively (preDLI) or prophylactically (proDLI) may be employed to prevent recurrent disease. The former refers to DLI given to patients with MRD, molecular relapse, or mixed chimerism (MC); the latter refers to DLI given to patients with high-risk disease or T-cell depleted grafts [190]. A series of murine studies demonstrated that after non-myeloablative transplant, preDLI could eliminate mixed chimerism without causing clinically significant GvHD [191, 192].

A German two-center retrospective study showed that proDLI doubled the 7-year OS of high-risk AML patients (DLI vs no DLI, 67% vs 31%, p < 0.001) [193]. Similarly, a multicenter study from China demonstrated that proDLI significantly improving the 3-year OS and LFS in patients with high-risk acute leukaemia [194]. According to a retrospective EBMT study of 105 patients with AML, pre/proDLI was associated with improved 5-year OS in AML patients with unfavorable cytogenetics or transplant beyond CR1 (compared with controls, 69.8% vs 40.2%, p = 0.027) [195]. In another EBMT study of 318 patients (AML, n = 249), the 5-year RI of preDLI for MRD, pre-DLI for MC and proDLI were 44%, 28% and 28%, and 5-year OS were 51%, 63%, and 68%, suggesting the detrimental effect of MRD on relapse and survival. Patients with MRD/mixed chimerism who responded to preDLI had better 5-year OS than non-responders (63%/76% vs 30%) [190].

GvHD is a major contributor to post-DLI NRM, but it remains unclear whether pre/proDLI truly increases the risk of GvHD. One EBMT study suggested that the increase in GvHD risk associated with DLI was not significant [195]. In contrast, the aforementioned Chinese study showed that prophylactic DLI increased the incidence of cGvHD (DLI vs control, 2-year incidence of cGvHD 38% vs 17%, p=0.021), but not aGvHD (incidence of grade II–IV aGvHD, 17% vs 23%, p=0.35) [194]. In the pre/proDLI setting, DLI-induced GvHD is not associated with reduced RI but clearly associated with higher NRM and poorer OS [190]. Risk factors for post-DLI GvHD included age > 60 years (p=0.046), transplantation beyond CR1 (p=0.003), interval from HSCT to DLI < 6 months (p=0.018), and a history of grade II–IV aGvHD before DLI (p=0.036) [190]. Careful selection of patients for pre/proDLI may be necessary to avoid the risk of GvHD and preserve the benefit of GvL.

Maintenance therapy

Post-HSCT pharmacologic maintenance therapy should optimally reduce the incidence of relapse but should do so without exerting hematological side effects or detrimental drug interactions with other post-transplant medications. Hypomethylating agents (HMAs) are effective therapy for AML and have been studied as post-HSCT maintenance. In the post-transplant setting, preclinical studies suggested that HMAs could epigenetically modulate donor-derived immunity, leading to Treg expansion, induce cytotoxic T-cell response to upregulated tumor antigens, and enhance NK-cell function by inducing KIR expression [196–198]. Therefore, HMAs could theoretically enhance GvL and reduce GvHD [199]. Moreover, HMAs are less myelosuppressive than conventional chemotherapy, making them desirable agents to test in the post-HSCT maintenance setting. In clinical trials, however, there are conflicting results whether post-HSCT HMA maintenance could reduce RI or improvement in RFS [200–204]. In a phase 3 study randomizing 187 patients with high-risk myeloid malignancy (around 60% high-risk AML) to azacitidine vs no azacitidine post-HSCT, the two groups had no significant difference in OS (HR 0.84, p = 0.43) or RFS (HR 0.73, p = 0.14) [204]. However, less than 30% patients were able to complete the planned total 12 cycles of therapy, raising concern about the tolerability and feasibility of maintenance azacitidine [204]. The addition of granulocyte colony-stimulating factor (G-CSF) may reduce hematologic toxicity of HMA and enable delivery of therapy. A randomized phase II study from China studied the G-CSF/decitabine maintenance vs no maintenance in 204 high-risk AML patients post-transplant and found significantly lower 2-year cumulative RI in the maintenance group (15.0% vs 38.3%, p < 0.01) and no significant increase in the rate of cGvHD [205].

Venetoclax (VEN) has been studied off-label in the post-HSCT maintenance setting. In single center studies, the main adverse events of venetoclax included cytopenia and diarrhea, but it did not increase the risk of GvHD beyond historic controls [206, 207]. A small study tested the combination of low dose decitabine with venetoclax as post-HSCT maintenance in 20 patients (AML, 17), showing a 2-year OS of 85.2%, aGvHD (any grade) in 55% and cGvHD in 20% of patients [207]. A randomized, phase 3 study of azacitidine/venetoclax as post-HSCT maintenance (VIALE-T) is currently being conducted [208].

Targeted agents against leukaemia specific mutations have been applied to post-HSCT maintenance. Sorafenib is a FLT3 inhibitor that has on-target effect against FLT3-mutated AML, furthermore, it may enhance GvL through autocrine cytokine production [209]. Two randomized multicenter studies in Europe and China, respectively, showed that sorafenib was associated with reduced RI, improved RFS and OS [210, 211]. However, difficulty in recruiting and premature discontinuation of treatment have been observed in these trials, questioning the feasibility and tolerability of sorafenib. Moreover, sorafenib was associated with increased risk of aGvHD and cGvHD [211]. Another FLT3 inhibitor, midostaurin, was assessed in the randomized RADIUS trial as post-HSCT maintenance. The addition of midostaurin achieved a 40% relative reduction in the risk of relapse but did not significantly improve RFS or OS [212]. Although midostaurin was not associated with higher risk of GvHD, it significantly increased gastrointestinal adverse events [212]. Gilteritinib, a newer generation FLT3 inhibitor, demonstrated promising effect in preventing AML relapse in small retrospective studies [213]. However, preliminary results from the randomized phase 3 MORPHO trial (BMTCTN 1506) showed no improvement in RFS with gilteritinib maintenance [214]. IDH1 and IDH2 inhibitors have also been tested in small phase 1 studies as post-HSCT maintenance for patients with respective mutations. Enasidenib appeared safe post-HSCT [215]. Results from the ivosidenib maintenance trial is still pending (NCT03564821).

Adoptive NK-cell therapy

The efficacy and risk of adoptive NK-cell therapy to prevent AML relapse has been studied in small, early phase clinical trials [216]. The NK-cell donor may be different from the stem cell donor (third party) and selected based on KIR ligand mismatch and KIR B motif number [217, 218]. Infusing NK-cells from a third party may have lower risk of GvHD [218]. Prophylactic NK-cell therapy may be performed before or after HSCT. A phase 1 clinical trial delivered expanded NK-cell transfer from a third party haploidentical donor prior to HSCT for high-risk myeloid malignancy, and demonstrated a positive correlation between survival and the NK-cell dose [218]. In another phase 1 clinical trial, infusion of ex vivo expanded, purified NK-cells before and after (day −2, +7 and +28) HRD HSCT for high-risk myeloid malignancies was safe, with lower incidence of relapse, and no increase in GvHD observed in the 13 patients enrolled [219]. A phase 1/2 study studied the efficacy and safety of IL-15 and IL-21 expanded HSCT donor-derived NK-cell delivered at 2-week and 3-week post-HSCT in patients with high-risk acute leukaemia (AML, 78%, n = 32), showing significant reduction in relapse with NK-cell infusion, and no increase in GvHD [220]. However, another phase 2 study from the same group showed that additional NK-cell infusion at day +6 and +9 did not further reduce relapse but was associated with higher toxicity from cytokine release syndrome [221]. Although adoptive NK-cell therapy showed promising results in prevention post-HSCT relapse, ex vivo expansion and in vivo persistence of the transferred NK-cells remain the major limitations. The optimal timing and dosage of NK-cell infusion also need to be defined [222].

GvHD and the Treatment for post-HSCT Relapse

There is a lack of consensus on the best treatment for post-HSCT relapse. A combination strategy of tapering immunosuppressants, chemotherapy, DLI, and/or second HSCT is commonly deployed. The complexity of treatment often makes it difficult to distinguish the effect on GvHD from each individual approach.

In medically-fit patients with post-SCT relapse, conventional intensive chemotherapy may be associated with better overall response rates and OS compared with hypomethylating agents (HMA) [223]. In different studies, CR rates after conventional chemotherapy were 40–70%, and around 10% to HMA [223–227]. HMA in combination with VEN has been used in post-HSCT relapse. There is no clear evidence that HMA/VEN triggers or exacerbates GvHD. In a small case series of four patients, one developed limited skin GvHD after HMA/VEN while also receiving four sessions of DLI [228]. In one multicenter retrospective study, 32 patients received HMA/VEN for post-transplant relapse, none developed new aGvHD, one developed cGvHD in the setting of DLI[229].

DLI is a commonly used treatment for post-HSCT relapse. In a German single center study of 51 patients with post-HSCT AML relapse who were given DLI, median 2-year and 5-year OS were 25.5 and 9.8%, respectively [230]. In an EBMT multicenter retrospective study of adult patients with first relapse post-HSCT, the 2-year OS was 21% vs 9% in DLI-recipients (n = 171) vs no DLI (n = 228, p < 0.0001) [231]. DLI improves the CR rates and survival in both chemotherapy- and HMA-treated patients with relapsed AML [223]. Like prophylactic DLI, therapeutic DLI is also associated with significantly increased, and even higher, risks of acute and chronic GvHD. Schmid et al, showed that the incidence of post-DLI grade II–IV aGvHD approached 40% and was associated with inferior survival (relative risk, RR, 0.54, p = 0.003), whereas the risk of cGvHD was 46% and associated with a better outcome among survivors more than 100 days after DLI (RR, 2.29, p = 0.004) [231].

Post-DLI GvHD and GvL are both dose-dependent, but some patients could enjoy a potent GvL effect without clinical GvHD [232, 233]. The risk of GvHD after DLI for relapsed AML is similar to pre/proDLI [230]. Therefore, a DLI dose escalation schedule was widely accepted [234, 235]. Initial CD3+ cell dose higher than 100 million/kg was associated with higher incidence of GvHD, and did not improve the RI or survival [236]. Pre-DLI lymphodepletion with fludarabine and cyclophosphamide was also associated with significantly higher incidence of aGvHD due to increased T-cell proliferation. Risks of grade II-IV aGvHD and grade III-IV aGvHD were 60% and 47% in patients receiving lymphodepletion and DLI, vs 24% and 14% in DLI alone (both p = 0.01) [237]. Immunomodulating agents can potentially ameliorate the pro-GvHD effect of DLI. In murine studies, post-mismatched DLI cyclophosphamide could eradicate alloreactive T-cells, prevent lethal GvHD while preserving a potent GvL effect [238].

Second HSCT can be considered in medically fit patients especially those who relapsed ≥ 6 months post-first HSCT. The choice of donor may be the same donor, a different matched donor, or a different HRD. The EBMT compared the outcomes of 2^nd HSCT in AML using different donors, and found similar LFS, incidence of GvHD and relapse among the groups, but the risk of NRM doubled in HRD [239].

Relapsed AML has increased expression of immune evasion associated proteins (e.g., PD-L1) and/or HLA class II silencing [240]. Targeting the PD-1/PD-L1 pathway has yielded conflicting results in small phase 1/2 clinical trials for relapsed/refractory AML [241]. Pre-transplant PD-1/PD-L1 blockade is associated with exacerbated GvHD, which can be ameliorated by PtCy [242–244]. PD-1/PD-L1 blockade has not been studied in post-HSCT settings. Other novel checkpoint inhibitors, such as magrolimab, sabatolimab, have not been studied for post-transplant relapse either [245, 246]. Adoptive cellular therapies are in early development for relapsed AML. A phase 1 study showed poor feasibility of anti-HA-1 T-cells in post-HSCT relapse [247]. Off-the-shelf CD33 CAR-NK cells are safe in relapsed/refractory AML, but the feasibility and efficacy of CAR-NK warrants further testing in larger clinical trials [248].

Summary

Despite frequent cross-talks between the immune mechanism of GvL and GvHD, several key differences have been suggested by pre-clinical or clinical studies, such as hematologic tissue-restricted vs more ubiquitously expressed alloantigens, different T-cell differentiation profiles and the higher dependence on humoral immunity by GvHD. Innate immune cells such as NK-cells and γδ T-cells may have potent antileukemic effects while remaining suppressive or neutral on GvHD. Deeper understanding of the differences in GvL and GvHD immunopathogenesis will likely inspire better GvHD prevention, boost the GvL effect, and lead to more-effective anti-relapse therapy, ultimately to fully harness the power of HSCT in AML.

Practice Points.

HSCT remains the only potential curative therapeutic approach for patients with intermediate or high-risk AML.

The main causes of failure of HSCT are AML relapse and GvHD.

Detailed understanding of the immunopathogenesis of GvHD has enabled the employment of more effective GvHD prevention strategies.

Research Agenda.

GvHD remains the leading cause of NRM for patients with AML undergoing HSCT.

Deeper understanding of the differences between the mechanisms of GvHD and GvL immune responses could facilitate the development of better HSCT platforms to suppress the former and enhance the latter.

Treatment of post-HSCT AML relapse should avoid triggering or exacerbating GvHD.