Reactive Myelopoiesis and the Onset of Myeloid-Mediated Immune Suppression: Implications for Adoptive Cell Therapy

Abstract

Adoptive T cell therapy (ACT) in combination with lymphodepleting chemotherapy is an effective strategy to induce the eradication of cancer, providing long-term regressions in patients. However, only a minority of patients that receive ACT with tumor-specific T cells exhibit durable benefit. Thus, there is an urgent need to characterize mechanisms of resistance and define strategies to alleviate immunosuppression in the context of ACT in cancer. This article reviews the importance of lymphodepleting regimens in promoting the optimal engraftment and expansion of T cells in hosts after adoptive transfer. In addition, we discuss the role of concomitant immunosuppression and the accumulation of myeloid derived suppressor cells (MDSCs) during immune recovery after lymphodepleting regimens and mobilization regimens.

1. Introduction

In recent years, there has been an explosion of clinical trials launched to explore the safety and efficacy of ACT for treatment in various malignancies. According to clinicaltrials.gov, more than 300 clinical trials utilizing treatment with lymphodepleting chemotherapies that include cyclophosphamide in combination with ACT are actively recruiting patients. While, preparative lymphodepleting methods are essential to promote the engraftment of adoptively transferred T cells and augment their anti-tumor activity, the concomitant recovery of the endogenous immune system after lymphodepleting regimens can have a profound impact on the function of adoptively transferred T cells. The use of lymphodepleting regimens applies significant pressure on the bone marrow to reconstitute the immune system, which ultimately results in an increased abundance of immunosuppressive myeloid cells. This striking phenomenon is due in part to the mobilization of hematopoietic stem and progenitor cells (HSPCs) from the bone marrow and their differentiation into myeloid cells. The hematopoietic differentiation trajectory and the function of the immune system are hijacked by tumors to promote a growth advantage and evade immune clearance. A variety of factors contribute to the expansion of myeloid cells during stress-induced myelopoiesis which can influence the onset of myeloid-mediated immunosuppression. Tumors preferentially promote the expansion of myeloid cells with potent immunosuppressive functions, including myeloid derived suppressor cells (MDSCs). However, cancer-driven myelopoiesis may differ in the setting of ACT.

In this review, we provide an overview of the known literature pertaining to lymphodepleting regimens that precede ACT. We highlight the resulting accumulation of immunosuppressive myeloid cells which can have profoundly negative consequences on the anti-tumor activity elicited by adoptively transferred T cells. We draw parallels between the known mechanisms that drive myelopoiesis in the presence of pathogenic stimuli or mobilizing cytokines which may also promote the accumulation of myeloid cells in reaction to the stimuli provided by various cytotoxic, lymphodepleting agents. Finally, we provide rationale for strategies to target reactive myelopoiesis with a goal of enhancing therapeutic outcomes for the treatment of cancer with ACT.

2. Preparative lymphodepletion is essential to elicit durable therapeutic responses to ACT

For nearly 30 years, the infusion of T cells possessing the capability of recognizing and eliminating tumor cells has been explored in patients with cancer. Early studies pioneered by the Rosenberg group at the National Cancer Institute demonstrated that the administration of IL-2 could expand T cells in vivo, leading to durable regressions in patients with metastatic melanoma [1, 2]. This groundbreaking discovery was early evidence that therapies targeting the immune system, but not tumor cells, could lead to the eradication of cancer in human subjects. Subsequent studies demonstrated that the ex vivo expansion of tumor infiltrating lymphocytes (TILs) followed by the infusion of these cells to patients with metastatic melanoma led to complete regression of disease [3]. Although tumor cell death can be elicited after the infusion of tumor-specific T cells in patients, clinical response rates were not consistently observed until ACT was preceded by lymphodepletion, in the form of non-myeloablative (NMA) chemotherapy and/or total body radiation (TBI) [3, 4]. Similarly, durable remissions have been consistently observed in patients that receive ACT with T cells transduced with a chimeric antigen receptor (CAR) or transgenic T cell receptor (TCR) when combined with lymphodepleting chemotherapy [5–10]. The rationale to combine ACT with lymphodepleting regimens was drawn from the experience of creating optimal conditions in a host to facilitate engraftment after hematopoietic stem cell transplant (HSCT). In HSCT regimens, complete myeloablation is induced via chemotherapy to simultaneously eliminate malignant cells, provide a niche for the engraftment of donor cells, and prevent tissue rejection after allogeneic transplantation [11]. For ACT however, the goal of pre-conditioning regimens is to eliminate lymphocytes that compete for cytokines necessary for the proliferation and function of adoptively transferred T cells in vivo.

Preclinical mouse models have demonstrated that the infusion of gp100-specific pmel T cells in B16 tumor-bearing mice could significantly induce tumor regression, but only when ACT was combined with non-myeloablative TBI [12]. Several studies in mice have demonstrated the importance of IL-7 and IL-15 in a post-lymphodepletion setting [12–16]. Importantly, several clinical trials investigating the combination of lymphodepletion regimens with ACT demonstrated increases of IL-7 and/or IL-15 post-lymphodepletion and T cell infusion confirming the previous findings in murine models [4–6, 17–19]. Therefore, the increased availability of cytokines such as IL-7 and IL-15 are critical for efficacious responses in patients undergoing ACT in combination with lymphodepleting regimens.

The abundance of homeostatic cytokines and enhanced tumor regression rates has been associated with increased intensity lymphodepleting regimens when in combination with ACT [4, 6]. Initial studies in mice indicated that an increased degree of lymphodepletion (i.e. myeloablation) with ACT in combination HSCT could enhance tumor regression [12, 20, 21]. Subsequent clinical studies initially validated these findings reporting that response rates were higher in melanoma patients that received ACT in combination with myeloablative therapy and HSCT compared to patients that received ACT with NMA chemotherapy alone (Overall response rate: 72% and 48% respectively) [4]. These findings indicated that an increased intensity of lymphodepletion could promote anti-tumor immunity elicited by ACT with TILs in human melanoma patients. However, the same group later conducted a randomized clinical trial and reported contradictory findings in that response rates were not significantly different between patients that received ACT with TILs in combination with myeloablative chemotherapy and HSCT in comparison to patients that received ACT plus NMA chemotherapy without HSCT [10]. The authors concluded that the contradiction of earlier findings could be attributed to patient selection and emphasized the importance of randomization for clinical trials. Overall, lymphodepletion is necessary to combine with ACT to achieve clinical efficacy, but the increased intensity of lymphodepleting regimens and subsequent HSCT may not have a clinical benefit to melanoma patients.

The rationale for increased intensity lymphodepleting regimens for ACT is not unfounded. Many clinical trials have investigated various lymphodepleting regimens in combination with ACT for the treatment of solid tumors and hematologic malignancies have demonstrated clinical benefits (Table 1). Despite the ample evidence supporting the combination of lymphodepleting regimens with ACT, increased dosing of chemotherapy and myeloablation is consistent with regimen-associated toxicity, including exacerbated neutropenia, infection, anemia, thrombotic microangiopathy, thrombocytopenia, pancreatitis, neurotoxicity, cardiotoxicity, and death [4, 8, 10, 22–24]. Thus, the utilization of cytotoxic agents to lymphodeplete cancer patients with a goal of augmenting the efficacy of ACT is not without significant risks.

Table 1.

Summary of clinical trials exploring various pre-conditioning regimens for ACT

Disease	ACT Strategy	Lymphodepleting Regimen	Additional Therapy	Key Findings	Reference
Melanoma	TIL	Cy/Flu	None	Enhanced survival with TBI combined with HSCT	[4]
		Cy/Flu + 200cGy TBI	HSCT
		Cy/Flu + 1200cGy TBI	HSCT
Melanoma	TIL	Cy/Flu	None	No survival difference observed with Cy/Flu+TBI+HSCT in randomized trial	[10]
		Cy/Flu + 1200cGy TBI	HSCT
Neuroblastoma	GD2-directed EBV-CTLs or ATCs	None	None	Similar persistence of both EBV-CTLs and ATCs Higher peak expansion of CAR EBV-CTLs CRs and NED observed	[27]
Neuroblastoma	GD2-directed ATCs	None	Anti-PD-1 (pembrolizumab)	Elevated IL-15 and CAR-T persistence in cohorts receiving Cy/Flu	[17]
		Cy/Flu
Pancreatic Cancer	Mesothelin CAR-T	None	None	Cy enhanced CAR-T expansion in vivo	[162]
		Cy
Multiple Myeloma	NY-ESO-1 TCR-transgenic T cells	Prior HSCT	Lenalidomide	Long-term persistence of T cells CRs and NED observed	[163]
Synovial Sarcoma	NY-ESO-1 TCR-transgenic T cells	Cyhigh/low/Flu	None	Elevated IL-7 & IL-15 and T cell engraftment with Cy/Flu vs. Cy alone	[6]
		Cy
NHL	CD19 CAR-T	Cylow/Flu	None	Elevated IL-7 and CCL-2 Cyhigh/Flu associated with increased CAR-T expansion & better clinical responses	[5]
		Cyhigh/Flu
NHL	CD19 CAR-T	Cy/Flu	HSCT 2° CAR-T	Enhanced CAR-T expansion & clinical responses with Flu	[8]
		Cy/Etoposide
		Cy
DLBCL	CD19 CAR-T	None	None	Poor ORR without LD ORR similar between Cy/Flu and bendamustine	[7]
		Cy/Flu
		Bendamustine
EBV+ HL/NHL	EBV-CTLs	None	2° EBV-CTLs	Long-term persistence of T cells CRs and NED observed	[28]
EBV+ HL	DN-TGFβRII EBV-CTLs	None	Various	Long-term persistence of T cells CRs and NED observed	[26]
EBV+ nasopharyngeal carcinoma	EBV CTLs	Anti-CD45 mAbs	None	1 CR Long-term persistence of T cells	[164]
B cell malignancies	Donor CD19 CAR-VSTs	Prior HSCT	1°−3° CAR-T	Long-term persistence of T cells CRs and NED observed	[28]

Abbreviations: TIL (tumor infiltrating lymphocytes); Cy/Flu (cyclophosphamide/fludarabine); cGy (centigrey); TBI (total body irradiation); HSCT (hematopoietic stem cell transplant); GD2 (disialoganglioside); EBV-CTLs (Epstein-Barr virus cytotoxic T lymphocytes); ATCs (activated T cells); CR (complete response); NED (no evidence of disease); NHL (non-Hodgkin’s lymphoma); HL (Hodgkin’s lymphoma); 1°, 2°, 3° (primary, secondary, or tertiary T cell infusions); DLBCL (Diffuse Large B cell Lymphoma).

In addition to cyclophosphamide and fludarabine, many other chemotherapeutic compounds also have lymphodepleting effects such as temozolomide, bendamustine, and etoposide and have been utilized in clinical trials and pre-clinical models investigating the efficacy of various ACT strategies [7, 8, 25], (NCT02664363). Thus, an attractive strategy would be to augment the success of ACT in human malignancies by exploring the utilization of chemotherapy agents that simultaneously lymphodeplete patients, providing a supportive niche for infused T cells, and also induce the cytotoxicity of tumor cells. Ultimately, continued efforts to eliminate lymphodepleting regimens prior to ACT may eventually be ideal to reduce treatment-associated morbidities and increase patient-treatment eligibility. In fact, some clinical trials have observed tumor shrinkage and complete regressions after ACT without any pre-conditioning lymphodepletion [26–28]. However, a universal solution that circumvents the utilization of cytotoxic lymphodepleting regimens to promote the activity of ACT, particularly in patients with solid tumors has yet to be implemented successfully in the clinic.

3. Reactive myelopoiesis promotes the expansion of MDSCs

Some of the earliest descriptions of MDSCs in mice and in human subjects were in the context of treatment regimens that included cytokine mobilization or the use of cytotoxic agents, including cyclophosphamide or TBI [29–32]. The immunosuppressive cells that were described in these studies were broadly characterized as suppressor cells from lymphoid organs after treatment with cyclophosphamide or resembled HSPCs in patients with head and neck cancer [30, 31]. Later studies characterizing immunosuppression after treatment with cyclophosphamide identified that cells of the myeloid lineage expanded and could potently suppress T cell activity [32, 33]. The reason for an increased accumulation of immunosuppressive myeloid cells, including MDSCs, after cyclophosphamide treatment is due to the mobilization of HSPCs from the bone marrow. Like granulocyte colony-stimulating factor (G-CSF), cyclophosphamide initiates the release of proteases, such as neutrophil elastase and cathepsins, within the bone marrow which disrupt HSPC adhesion to marrow-associated stromal cells allowing their egress into the periphery [34–36]. The majority of mobilized HSPCs then differentiate into CD11b⁺Gr-1⁺ cells, resulting in an increased number of these myeloid cells and their precursors in comparison to healthy control mice [35, 37]. Interestingly, the mobilized progenitors hold a significant myelopoietic bias and the recovery of other immune cell subsets can take considerably longer after treatment with lymphodepleting regimens, leaving the host lymphopenic (Figure 1) [38–40]. Collectively, this process of HSPCs mobilization and myeloid cell differentiation can be referred to as reactive myelopoiesis [41]. In contrast to emergency myelopoiesis which occurs in response to the presence of an infectious pathogen, reactive myelopoiesis is triggered by cytokines (G-CSF), tumor-mediated inflammation, chemical agents (cyclophosphamide, 5-fluorouracil, or alum), and physical insults (radiation) [41–49].

Figure 1. Immune recovery after non-myeloablative lymphodepletion.

Upon treatment with lymphodepleting chemotherapy or TBI, the absolute number of leukocytes decreases rapidly. Over time, each individual immune cell subset reconstitutes to its baseline frequency. However, lymphodepleting chemotherapy and TBI drives a reactive myelopoietic response characterized by a disproportional increase of immunosuppressive myeloid cells. Hematopoietic stem and progenitor cells are rapidly mobilized from the bone marrow which exhibit a myelopoietic bias. As a result, the total abundance of myeloid cells significantly exceeds baseline frequencies. The number of MDSCs, neutrophils, and monocytes are increased, but this expansion can contract after reaching a peak. Similarly, dendritic cells can be found at higher frequencies in the peripheral blood, lymph nodes, liver and BM. However, the recovery of macrophages after lymphodepletion is not well described. Lymphocytes, including NK cells, T cells, and B cells, remain depleted for several days to weeks and do not surpass the baseline threshold [35, 38–40, 90, 139].

Indeed, redundant mechanisms exist between emergency and reactive myelopoiesis characterized by overlapping cytokine signals and signal transduction pathways that activate myeloid cell differentiation pathways. For example, the production of G-CSF is increased in the presence of cancer or systemic bacterial infection (i.e. sepsis) [50–53]. As a result, these cytokine signals amplify JAK/STAT3 signaling which stimulate the expression of c-Myc and C/EBPβ leading to the rapid proliferation of myeloid progenitor cells and subsequent differentiation of monocytic and granulocytic cells [42, 54, 55]. Of note, STAT3 and C/EBPβ are established transcription factors that promote the expansion of MDSCs and regulate their immunosuppressive functions [56, 57]. Recent evidence has shown that emergency myelopoiesis in patients with severe COVID-19 exhibit an expansion of dysfunctional monocytes and neutrophils, which bear transcriptional similarities to M-MDSCs and PMN-MDSCs [58, 59]. Thus, convergent pathways that drive the expansion of MDSCs exist during both emergency and reactive myelopoiesis.

Despite the overlap of pathways that drive myelopoiesis, the context of the initial inflammatory stimulus likely influences the anti-microbial and immunosuppressive function of cells that arise during emergency or reactive myelopoiesis. For instance, many studies describing emergency myelopoiesis activated by pathogenic stimuli examine the anti-microbial roles of neutrophils [48, 60, 61]. The accumulation of MDSCs has been previously characterized as a result of an emergency myelopoiesis-like state in cancer patients [62]. Moreover, increased tumor burden in stage III/IV cancer patients correlates with an increased frequency of MDSCs and neutrophils [63–65]. Tumors produce an abundance of inflammatory cytokines and growth factors that constantly bombard the bone marrow, which applies pressure on HSPCs, resulting in the dysregulation of myelopoiesis and the accumulation of MDSCs. While this dysregulation of myeloid cell development in tumor-bearing hosts is characteristically similar to emergency and reactive myelopoiesis, there are key differences distinguishing these immunological phenomena. Emergency and reactive myelopoiesis are often acute processes that result in the mobilization of HSPCs and a dramatic elevation in the number of circulating granulocytic and monocytic cells within 1–2 weeks after administration of a mobilizing agent or onset of infection [35, 39, 66]. The apparent neutrophilia and monocytopenia in these settings typically resolves after reaching a peak and the frequency of myeloid cells eventually drops back to its baseline [35]. In contrast, the cancer-driven myelopoiesis is chronic and the accumulation of immunosuppressive myeloid cells increases as tumor-mediated inflammation is amplified during disease progression [63, 67]. Moreover, the pre-existing elevation of myeloid cells in cancer patients can be exacerbated by mobilizing agents that trigger acute reactive myelopoiesis (Figure 2) [66]. In addition, experimental MDSC depletion methods, including treatment with anti-Ly6G and anti-Gr-1 antibodies, effectively reduce the number of MDSCs; however, CD11b⁺Gr-1⁺ cells can resist antibody-mediated depletion and the cessation of antibody treatment can result in the increased abundance of MDSCs, as much as 5–10 fold higher than pre-depletion levels [68–71]. Thus, myelopoietic reactions to therapies that acutely deplete MDSCs can result in a rebound, thereby increasing the overall MDSC abundance.

Figure 2. Stress-driven myelopoiesis promotes the accumulation of MDSCs.

Under steady state conditions, the frequency of lymphocytes and myeloid cells is at an equilibrium and MDSCs are exceedingly rare. An increase of MDSCs can arise during states of stress-driven myelopoiesis. Infectious pathogens, such as bacteria, can promote the mobilization of HSPCs which preferentially differentiate into myeloid cells, including MDSCs. Similarly, tumor-derived factors promote the accumulation of MDSCs which increases throughout disease progression. Reactive myelopoiesis can be triggered by chemotherapy, radiotherapy, cytokines (G-CSF, GM-CSF, TNFα, etc.), and possible antibody-mediated depletion. The addition of chemotherapy and/or radiotherapy can simultaneously deplete lymphocytes and cause the rapid increase of myeloid-biased HSPCs and MDSCs. The abundance of MDSCs after lymphodepleting therapy in tumor-bearing individuals is profoundly higher than the pre-lymphodepletion frequency of MDSCs.

4. The heterogeneity of myeloid cell expansion during reactive myelopoiesis

PMN-MDSCs and M-MDSCs have been shown to be genetically and functionally distinct from physiologic neutrophils and monocytes [72–74]. Generally, MDSCs have been described to be immunosuppressive, while neutrophils and monocytes from healthy donors or matched hosts do not have an immunosuppressive capacity [72, 74, 75]. However, neutrophils that expand after G-CSF-induced mobilization are phenotypically immature, have a defective chemotactic ability, and exhibit a reduced phagocytic capacity [54, 76]. These characteristics mirror that of PMN-MDSCs [73]. Moreover, high-density neutrophils in cancer patients and G-CSF mobilized healthy donors can suppress T cell proliferation and effector functions [77–80]. Indeed, PMN cells within the mononuclear layer after density gradient centrifugation are immunosuppressive PMN-MDSCs [72]. Yet, subsets of high-density neutrophils expressing PD-L1, arginase-1, and heightened levels of reactive oxygen species potently suppress T cell proliferation [77, 79, 80]. In contrast, other reports highlight that donor or autologous low high-density PMN cells and LOX-1⁻ PMN cells fail to suppress T cell proliferation in comparison to low-density, LOX-1⁺ PMN-MDSCs from the same patient [72, 81]. Thus, it is conceivable that the proportion of PMN-MDSCs and immunosuppressive neutrophils amongst other non-suppressive or even immunostimulatory PMN cells could vary in mobilized healthy donors and cancer patients [80].

In addition to immunosuppressive PMN cells from mobilized patients, monocytes and regulatory dendritic cells (DCs) expand in G-CSF mobilized donors. CD14⁺ monocytic cells from G-CSF-mobilized donors can suppress T cells and NK cells [82–86]. Moreover, CD34⁺ regulatory monocytes expand after G-CSF administration in healthy donors which inhibit graft-versus-host disease (GvHD) in patients receiving allogeneic hematopoietic stem cell transplantation [87]. IL-3Rα⁺ DCs expand in healthy donors mobilized with G-CSF and potentiate Th2 immune responses, which can prevent the onset of GvHD [88]. Conversely, immunostimulatory DCs can expand after treatment with mobilizing doses of cyclophosphamide and are necessary to promote the efficacy of ACT [89–91]. Neuroblastoma patients treated with various lymphodepleting regimens in combination with GD2-specific CAR-T cells exhibit increases in CD11b⁺CD33⁺CD163⁺ myeloid cells by one-week post-CAR-T cell infusion, which remained elevated for at least 6 weeks [17]. M-MDSCs were elevated in leukemia patients receiving CD19-directed CAR-T cells and the abundance of M-MDSCs negatively correlated with patient survival [92]. However, the immunosuppressive capacity of these M-MDSCs was not evaluated, so it is unclear if these MDSCs suppressed the CD19 CAR-T cells. Notably, the authors of this study examined myeloid cell frequencies in cryopreserved PBMC samples. PMN-MDSCs fail to survive freeze/thaw cycles thereby negatively impacting their detectability in cryopreserved PBMCs [93]. Nevertheless, in fresh PBMCs from melanoma and lung cancer patients, we demonstrated that highly suppressive PMN-MDSCs constitute the majority of myeloid cells that expand post-lymphodepletion and T cell infusion [66]. Thus, both PMN-MDSCs and M-MDSCs expand in cancer patients receiving ACT in combination with pre-conditioning lymphodepletion.

The inflammatory stimulus evoked by mobilization and lymphodepleting regimens appear to be sufficient to generate MDSCs, even in the absence of tumors. Indeed, doses of cyclophosphamide that are sufficient to mobilize HSPCs can promote the expansion of weakly suppressive MDSCs in non-tumor bearing mice [35, 94]. However, MDSCs from tumor-bearing mice that received cyclophosphamide were more suppressive than MDSCs taken from tumor-naïve mice given cyclophosphamide [94]. Thus, in healthy donor mobilized blood, MDSCs may be less suppressive than those cells taken from a cancer patient who is administered mobilizing agents, such as G-CSF or cyclophosphamide. It is likely that the added inflammatory state in tumor-bearing individuals increases the immunosuppressive capacity of MDSCs that accumulate post-mobilization. Notably, the clinical application of ACT in combination with lymphodepleting regimens utilize recombinant G-CSF to stimulate the expansion of neutrophils in patients that exhibit febrile neutropenia [95, 96]. Despite that the administration of recombinant G-CSF is necessary to resolve neutropenia in patients, it is possible that this treatment could also drive the expansion of MDSCs [97, 98].

There are a variety of sources that contribute to the expansion of MDSCs including the presence of advanced metastatic cancer, the type of tumor, and T cell-derived cytokines which can make it challenging to identify which specific factors dictate the expansion of MDSCs in the setting of ACT. The combination of these factors and the source of stimulation provides more complexity in the prediction of myelopoietic responses. This is further complicated by the heterogeneous nature of MDSCs and that particular factors could drive the expansion of M-MDSCs or PMN-MDSCs separately. For instance, G-CSF can potently induce the differentiation of PMN-MDSCs from CD34⁺ stem cells, while stimulation with GM-CSF or M-CSF preferentially leads to the expansion of M-MDSCs [99]. It appears that the inclusion of IL-3 in cytokine cocktails can prevent the differentiation of CD15⁺ PMN-MDSCs from hematopoietic stem cells [100]. The entire myelopoietic milieu including the tumor microenvironment, bone marrow, secondary lymphoid organs, and blood must be considered when determining which factors drive the expansion of MDSC subsets.

5. Factors that drive reactive myelopoiesis in ACT: Opportunity for therapeutic targets?

Indeed, lymphodepleting regimens can deplete immunosuppressive regulatory T cells (Tregs) and MDSCs, but have disparate effects on immune cell reconstitution [101–103]. Tregs appear to be more sensitive to TBI in comparison to Cy/Flu-based lymphodepletion regimens [103]. Notably, the kinetics of MDSC expansion after lymphodepleting TBI is different than mice treated with chemotherapy. The depletion of MDSCs in TBI-based lymphodepleting regimens is transient and their frequency increases within 2–3 weeks after the initial radiation dose [40]. In contrast, Cy/Flu lymphodepleting regimens promote the rapid reconstitution and over-production of CD11b⁺Gr-1⁺ MDSCs within 7–10 days [13]. Interestingly, the addition of TBI to cyclophosphamide or Cy/Flu lymphodepleting regimens can attenuate the expansion of myeloid cells [13]. This suggests that the repression of bone marrow reconstitution by radiation could abrogate cyclophosphamide-mediated reactive myelopoiesis. Similarly, docetaxel or gemcitabine can reduce MDSC accumulation in mice treated with TBI- or cyclophosphamide-based lymphodepleting regimens, respectively [40, 47, 91]. However, the added chemotherapy can apply more stress on the bone marrow and could increase the risk of fatal toxicity. Specifically, we observed that the addition of docetaxel to Cy/Flu lymphodepletion resulted in a 100% mortality rate in mice within one week after the start of therapy (data unpublished). In contrast, gemcitabine was well tolerated in mice that received lymphodepletion with cyclophosphamide alone [47, 91]. Thus, additional chemotherapy agents can be added to lymphodepleting regimens to deplete MDSCs, but the dosing schedule and type of chemotherapy agent must be carefully considered to prevent therapy-induced toxicities.

While chemotherapy agents can reduce the number of MDSCs that expand after lymphodepleting regimens, their selectivity is broad and can result in undesired toxicities. Hence, targeting mechanistic processes more specific to MDSCs are preferable to ameliorate immunosuppression during ACT regimens (Figure 3). For instance, the blockade of MDSC chemotaxis can reduce the number of MDSCs in the spleens and tumors of mice that receive ACT. CCR2 and CXCR2 are key chemotactic receptors for M-MDSCs and PMN-MDSCs, respectively [47, 104, 105]. The blockade of CCR2 has been shown to restrict the accumulation of M-MDSCs in mice treated with lymphodepleting cyclophosphamide in combination with CD4⁺ ACT, while CXCR2 blockade can reduce the number of tumor-associated PMN-MDSCs, leading to an increased number of adoptively transferred T cells within tumor beds [47, 106]. The blockade of CCR2 results in a compensatory accumulation of PMN cells in the periphery and in tumors. The inverse occurs in mice treated with anti-CXCR2 antibodies resulting in an increase of monocytic cells and macrophages [107]. Additionally, monocytic cells accumulate in the bone marrow during CCR2 blocking therapy and the cessation of this chemotactic blockade results in the increased accumulation of monocytes in tissues, leading to the priming of metastatic niches [108, 109]. Thus, the compensatory mechanisms in response to MDSC chemotaxis blockade can likely lead to therapeutic resistance.

Figure 3. MDSC accumulation in the setting of adoptive T cell therapy.

Pre-conditioning lymphodepleting regimens initiate the release of G-CSF, NE (neutrophil elastase), CG (cathepsin G), and HGF. These factors can disrupt the adhesion of HSPCs to bone marrow stroma including interactions with CXCR4, which promote the mobilization of HSPCs and MDSCs from the bone marrow after lymphodepletion. Numerous cytokines and transcription factors stimulate the differentiation of HSPCs to MDSCs which potently suppress adoptively transferred T cells through various mechanisms (Arg1, ROS, PD-L1, PGE2, etc). The abundance of MDSCs can be attenuated by chemotherapy (gemcitabine, docetaxel), radiotherapy, and the exploitation of known apoptosis-inducing pathways (TRAIL-R2, Fas, NKG2D-ligands). Similarly, the number of MDSCs can be reduced through the neutralization of chemotactic factors, limiting their ability to traffic to tumors. Mobilized HSPCs exhibit a MDSC-differentiation bias that can be skewed by DC-promoting factors to enhance to accumulation of DCs, thereby augmenting the activity of adoptively transferred T cells. T cells can be modified to express factors to promote the differentiation (Flt3L) and activation (IL-12, CD40L) of immunostimulatory DCs which can overcome MDSC-mediated immunosuppression after adoptive transfer.

In addition to blocking MDSC trafficking, targeting suppressive mechanisms is an effective strategy to relieve immunosuppression and enhance the efficacy of ACT. The inhibition of c-MET can restrict the accumulation of immunosuppressive CD11b⁺Ly6G⁺ cells (presumably PMN-MDSCs) in lymph nodes and tumors in mice treated with a cyclophosphamide pre-conditioning regimen in combination with an adenoviral gp100 vaccine and ACT [110]. Intriguingly, the depletion of PMN-MDSCs in peripheral lymphoid organs and tumors was associated with a concomitant increase of these cells within the bone marrow, suggesting that the inhibition of c-MET restricted the mobilization of PMN-MDSCs. Importantly, the reduction of PMN-MDSCs led to the expansion of adoptively transferred pmel T cells and enhanced the survival in B16 tumor-bearing mice [110]. Thus, the HGF-c-MET axis could directly promote the accumulation of MDSCs during reactive myelopoiesis. In addition, the blockade of PD-1/PD-L1 axis can enhance the efficacy of CD4⁺ ACT combined with lymphodepleting cyclophosphamide [47]. The authors of this study show that M-MDSCs, but not CD11b⁺Ly6G⁺ cells, are immunosuppressive after cyclophosphamide treatment in BALB/c mice with lymphoma and PD-1/PD-L1 axis abrogation enhanced CD4⁺ ACT. The suppressive capacity of MDSCs is highly dependent on fatty acid oxidation (FAO) and the administration of a carnitine palmitotransferase-1 inhibitor, etomoxir, enhanced the efficacy of cyclophosphamide or ACT [111]. Notably, the authors used a lymphodepleting dose of cyclophosphamide, but cyclophosphamide was not included with ACT, etomoxir combination therapy. Presumably, the inhibition of FAO would abrogate the suppressive capacity of MDSCs that expand as a result of lymphodepleting cyclophosphamide resulting in an enhanced ACT efficacy.

Indeed, targeting known pathways that promote the suppressive capacity of MDSCs can be effective at enhancing the efficacy of ACT. However, the modification of T cells to resist myeloid-mediated immunosuppression or eliminate suppressor cells represent alternative strategies to boost anti-tumor activity of adoptively transferred T cells. T cells engineered to express IL-12 and an anti-VEGFR-2 CAR can significantly reduce the number of CD11b⁺Gr-1⁺ cells within murine B16 tumors (99). The adoptive transfer of IL-12-expressing T cells can increase the expression of Fas in MDSCs, leading to an enhanced infiltration of the infused T cells within tumors and significantly improve the rate of tumor regression [112, 113]. In contrast to MDSCs from untreated tumor-bearing mice, we demonstrated that lymphodepletion-generated MDSCs expressed lower levels of Fas and were more resistant to Fas-induced apoptosis when in contact with T cells when compared to MDSCs taken from untreated tumor-bearing mice [66]. In addition to the ability of T cell-mediated Fas-induced apoptosis of MDSCs, the expression of NKG2D ligands enable NK cells to eliminate MDSCs [114, 115]. Specifically, NK cells engineered to express a CAR comprised of NKG2D and the ζ-chain of the T cell receptor can kill human M-MDSCs in vitro and within xenograft tumors upon adoptive transfer [115].

In addition to modifying ACT products to enhance the elimination of MDSCs, the modification of T cells to resist immunosuppression can also be advantageous. EBV-specific T cells engineered to express a dominant negative TGF-βRII (dnTGF-βRII) effectively expand and induce complete regressions in Hodgkin’s lymphoma patients [26]. TGF-β is well known to promote MDSC activity and the production of TGF-β by MDSCs can induce the differentiation of CD4⁺ T cells to Tregs [116–118]. In addition to TGF-β, IL-4 is a well-established regulator of MDSC activity. While IL-4 is dispensable for the promotion of MDSC accumulation, IL-4Rα⁺ MDSCs accumulate in cancer patients [65, 119]. CAR-T cells specific for prostate stem cell (PSCA) engineered to express a hybrid cytokine receptor comprised of an IL-4R exodomain fused to the IL-7R endodomain converted the inhibitory effects of IL-4 into the stimulatory effects of IL-7R signaling [120]. Thus, the engineering of T cells to resist immunosuppressive factors produced by tumors and/or MDSCs is an effective strategy to boost the effectiveness of ACT.

In the setting of ACT, the increased number of circulating HSPCs could perhaps provide a window of opportunity to skew myeloid differentiation. Retinoic acid related orphan receptor (RORC1) deficiency was associated with the failure myeloid precursors to differentiate into MDSCs and macrophages within spleens and tumors [62]. Moreover, RORC1 ablation resulted in the increased abundance of BM-hematopoietic stem cells, common myeloid progenitors (CMPs), and a concomitant decrease in granulocyte-macrophage progenitors (GMPs) [62]. Thus, targeting RORC1 could dramatically alter the landscape in the setting of reactive myelopoiesis. While RORC1 is a positive regulator of cancer-driven myelopoiesis, Speckle-type BTB-POZ protein (Spop) has recently emerged as a master repressor of emergency myelopoiesis. Spop-deficient mice are lymphopenic and die of lethal neutrophilia after a single dose of poly(I:C) or LPS [48]. Mechanistic studies revealed that Spop is a critical mediator of MyD88 ubiqutination and degradation in hematopoietic stem cells, leading to the downstream neutrophilia in Spop-deficent mice which could be reversed upon IL-1R blockade [48]. However, it is unclear if Spop promotes reactive myelopoiesis prompted by other stimuli including cytokines (i.e. G-CSF), or lymphodepleting chemotherapy regimens.

In addition to its effects in promoting the accumulation of granulocytes, tumor-derived G-CSF disrupts the development of DCs. The dearth of conventional DCs (cDCs) and DC-precursors in pancreatic and breast tumors is due in part to the antagonization of IRF8 activation in DC-precursors by G-CSF, which favors the accumulation of PMN-MDSCs and neutrophils [121]. This imbalance to myelopoiesis within the bone marrow can be partially restored when mice are given recombinant Flt3L to drive the expansion of conventional DCs at the expense of the development of PMN-MDSCs [121]. Mice that lack Flt3L exhibit an impaired accumulation of DCs after treatment with lymphodepleting cyclophosphamide and infusion of pmel T cells [91]. Similar to Flt3L, treatment with the TLR3 agonist, poly(I:C), can effectively increase the number of DCs within lymph nodes after cyclophosphamide treatment, which led to the expansion of adoptively transferred T cells [90]. Moreover, the power of Flt3L to drive the differentiation of immunostimulatory DCs can be leveraged through the modification of T cells to express Flt3L. In mice with breast tumors expressing ovalbumin (OVA) and Her2, the infusion of anti-Her2 CAR-T cells engineered to express Flt3L effectively expanded DCs and promoted epitope spreading against the OVA protein [122]. Provided that target antigen escape is a common resistance mechanism associated with CAR-T cell therapy, promoting the development of cDCs to generate polyclonal T cell responses may hold substantial relevance to the clinical application of ACT [123]. While Flt3L can promote the development of cDCs, engineering T cells to express CD40L or IL-12 can promote the licensing of APCs and negate the requirement of lymphodepleting cyclophosphamide [124, 125]. All-trans retinoic acid (ATRA) can promote the development of DCs at the expense of the development of monocytes and M-MDSCs [126, 127]. ATRA attenuates the expression of key mediators of immunosuppression in cyclophosphamide-induced MDSCs, including Arg1, iNOS, TGF-β, ROS, and VEGF [94]. Moreover, ATRA has been shown to be successful in limiting the ability of murine MDSCs to suppress human CAR-T cells in mice with xenograft tumors [128]. TNFα can have disparate effects on myelopoiesis by simultaneously inducing the apoptosis of pre-existing myeloid progenitor cells, but promoting the survival of hematopoietic stem cells and skewing their fate towards the development of myeloid cells [129]. In tumor bearing mice, the neutralization of TNFα can reduce the number of CMPs, GMPs, and MDSCs within tumor-bearing mice [130]. In the setting of ACT with CAR-T cells, TNFα amongst other cytokines, such as IL-1β and IL-6, are associated with cytokine release syndrome [131, 132]. Given the well-established roles of IL-1β and IL-6 in promoting the differentiation and suppressive functions of MDSCs, it is feasible that these factors play an important role in driving the accumulation of MDSCs after lymphodepletion and CAR-T cell infusion [56, 133, 134].

It is important to note that MDSCs that expand after mobilization and lymphodepleting regimens are transcriptionally and functionally distinct from MDSCs taken from untreated tumor-bearing mice [66]. We recently showed that MDSCs from tumor-bearing mice lymphodepleted with Cy/Flu have an enhanced survival capacity and reduced Fas expression in comparison to MDSCs from non-lymphodepleted mice [66]. We demonstrated that the increased abundance of MDSCs in mice and human cancer patients was associated with increases in HSPCs, which exhibited a myeloid differentiation bias in adoptive transfer experiments. In contrast, mobilized HSPCs transferred to non-lymphodepleted tumor-bearing recipient favored the development of lymphocytes at the expense of myelopoiesis. This striking result indicated that mobilized HSPCs have a myeloid lineage bias, but host factors are responsible for enforcing this bias and pushing the trajectory of myeloid cell development. We go on to show that IL-6 was critical during the differentiation of mobilized HSPCs to provide a protective effect to MDSCs against Fas-induced apoptosis conducted by T cells. The abrogation of IL-6 signals increased the expression of Fas in MDSCs, sensitized them to apoptosis, and enhanced the efficacy of ACT in mice.

6. The role of danger signals in reactive myelopoiesis

Mounting evidence has highlighted the importance of the microbiota and their products in driving immune responses in cancer [135, 136]. In mice, lymphodepleting doses of cyclophosphamide (i.e. 200mg/kg body weight in a mouse) can disrupt the gut-epithelial barrier, promoting dysbiosis and the translocation of bacteria from the gut [137]. The efficacy of ACT and effector T cell responses in mice pre-conditioned with cyclophosphamide and antibiotics are reduced in comparison to mice without antibiotic treatment [136, 138]. Moreover, treatment with antibiotics enhanced the persistence of murine CD19-CAR-T cells and prolonged B cell aplasia in mice with lymphoma, but did not enhance the survival of mice [136]. This suggests that the involvement of the microbiota in driving therapeutic responses may differ depending on the type of malignancy being treated and nature of the adoptive T cell therapy. Additionally, TBI-based lymphodepletion regimens can also promote the translocation of gut bacteria leading to DC activation and expansion of adoptively transferred T cells [20]. This suggests that microbial stimuli can augment the anti-tumor activity of adoptively transferred T cells through the activation of myeloid cells and antigen presenting cells. Similar strategies have identified that the TLR3 agonist, poly(I:C), could boost the activation of DCs in mice lymphodepleted with cyclophosphamide, resulting in an expansion of adoptively transferred T cells within the peripheral blood, lymph nodes, and tumors [90, 139]. Conversely, the activation of TLRs and the composition of gut microbiomes can concomitantly promote the accumulation of MDSCs [140, 141]. Thus, triggering certain TLR signaling pathways can yield different effects on myelopoiesis and the development of MDSCs. The mobilization of GMPs and monocytedendritic cell progenitors are triggered by LPS and CpG, respectively [142]. CpG-mediated activation of TLR9 can negatively impact the suppressive capacity of MDSCs [143]. In contrast, the activation of TLR4 by LPS can promote IL-10 production in MDSCs, leading to an impaired cross talk with macrophages leading to the down-regulation of IL-12 [144].

The translocation of bacteria and activation of microbial-sensing pathways can promote both anti-tumor immune responses and exacerbate immunosuppression. However, it appears that the development of MDSCs during reactive myelopoiesis could be skewed by targeting specific pattern recognition receptors (PRRs). Similarly, the activation of various PRRs by damage associated molecular patterns (DAMPs) can yield disparate effects on pro- and anti-tumor immune responses [145]. The release of DAMPs is widely known to promote immunogenic cell death in tumors through the activation of DCs and enhanced recruitment of T cells to tumors [146–148]. In contrast, high mobility group box 1 (HMGB1), S100A8/9, Hsp72, and calprotectin are DAMPs that can promote the accumulation and suppressive functions of MDSCs [59, 149–152]. In the context of ACT, lymphodepletion with TBI or chemotherapy induces systemic cell death and could promote the release of DAMPs [153]. It is conceivable that microbial sensing pathways and the release of DAMPs in response to lymphodepletion may play critical roles in regulating the expansion of MDSCs during myelopoiesis in patients receiving ACT. The distinction between reactive myelopoiesis and emergency myelopoiesis becomes less clear when considering that lymphodepleting TBI or chemotherapy perturb the gut microbiome. It is likely that both microbial-associated stimuli and the effects of lymphodepleting chemotherapy simultaneously impact myelopoieisis after adoptive T cell transfer.

7. Conclusions and future perspectives

Many studies in preclinical models have not included lymphodepleting regimens when studying the anti-tumor effect of adoptively transferred T cells. Indeed, providing an excess of tumor-specific T cells to a tumor-bearing host without any pre-conditioning holds significant value in the investigation of native interactions between cancer and the adaptive immune system. However, the true test of the ability of adoptively transferred T cells to control tumor growth must combine ACT with lymphodepletion to fully recapitulate the dynamic changes that occur during the recovery of the endogenous immune system. In preclinical models, numerous studies utilize TBI, rather than lymphodepleting regimens with cyclophosphamide alone or in combination with fludarabine. Indeed, both TBI and chemotherapy-based lymphodepleting regimens enhance the efficacy of ACT in mice [12, 13]. However, we discussed that the kinetics of immune recovery differ between TBI- and Cy/Flu-based lymphodepleting regimens. Notably, TBI is less translationally relevant in comparison to Cy/Flu pre-conditioning regimens in the clinical application ACT. Hence, we emphasize that the recovery of the immune system after lymphodepleting regimens is a critical mediator that can impact the efficacy of ACT. Furthermore, attempts to link the significance of ACT in human subjects must consider the use of translationally relevant lymphodepleting regimens.

Lymphodepleting regimens foster the engraftment and persistence of adoptively transferred T cells, but the concomitant onset of myeloid-mediated immunosuppression is key mechanism that restricts anti-tumor immune responses. Notably, irradiation and chemotherapy can profoundly disrupt the integrity of the bone marrow niche which consequently impacts hematopoiesis and disrupts lymphopoiesis long-term [154–157]. The effects of lymphodepleting regimens result in the acute accumulation of MDSCs and may potentiate indelible effects that could impact hematopoiesis and modulate the interactions between tumors and the immune system. Despite these consequences, lymphodepleting regimens provide a unique window of opportunity to reduce the accumulation of MDSCs by targeting myelopoietic differentiation. The first few weeks after adoptive T cell transfer are critical for potentiating the maximal expansion and long-term persistence of infused T cells, which is paramount for durable clinical responses [4, 23, 158, 159]. The rapid accumulation of MDSCs is clearly undesirable as the abundance of these immunosuppressive cells negatively impact the long-term persistence of adoptively transferred T cells [66]. However, the initiation of HSPC mobilization by lymphodepleting chemotherapy does not immediately result in the accumulation of MDSCs. Thus, we propose that this short window of time during HSPC mobilization and differentiation can be exploited to skew myelopoiesis away from the development of MDSCs, which ultimately could lead to longer persistence and improved functions of adoptively transferred T cell in vivo.

Collectively, a better understanding of reactive myelopoiesis in the setting of ACT can lead to improvements in patient outcomes. Additional clarity is needed to better understand the vast heterogeneity of immunostimulatory and immunosuppressive myeloid cells that expand after lymphodepleting and mobilization regimens. Age, obesity, the presence of cancer, and the type of mobilizing agent likely influence the immunosuppressive capacity of MDSCs and the accumulation of other monocytic and granulocytic cell populations that expand during mobilization regimens [77, 160, 161]. Only a few studies have identified negative associations with the presence of MDSCs and the expansion and/or persistence of adoptively transferred TILs or CAR-T cells [66, 92]. The contribution of the type of malignancy (hematological vs. solid tumors) and the modality of ACT (TIL, CAR-T, Transgenic TCR, etc.) are also factors that could influence the expansion of MDSCs after T cell infusion and lymphodepletion. Furthermore, it is unknown if TILs or engineered T cells are distinct from one another in terms of their sensitivity to MDSC-mediated suppression. It remains unclear if any of these factors predispose individuals to the onset of myeloid-mediated immunosuppression during the course of HSPC and myeloid cell mobilization.

The majority of studies that have investigated immune recovery after lymphodepletion have largely focused on the accumulation of MDSCs in the periphery. This gap in the field prompts intriguing questions: How does the tumor-immune milieu change after lymphodepleting chemotherapy? Does the recovery of tumor-associated immune cells antagonize or promote the eradication of tumors conducted by adoptively transferred T cells? Hopefully, an enhanced understanding of reactive myelopoiesis in the context of ACT can lead to the generation of new therapies that neutralize the immunosuppression that arises in response to lymphodepleting regimens.

Highlights.

• The impact of MDSCs on adoptive T cell therapy is reviewed.

• Reactive myelopoiesis is driven by lymphodepleting regimens.

• Strategies to inhibit lymphodepletion-generated MDSCs are discussed.