Macrophages modulate tumor response to chemotherapy; in this issue, Lossos and colleagues show that high-dose alkylating agents instigate a synthetic lethal program in lymphoma cells that is independent of DNA damage and involves recruitment and priming of macrophages for antibody-mediated tumor phagocytosis. These findings implicate chemotherapy-elicited macrophages as critical effectors of lymphoma clearance during biological therapy.
See related article by Lossos et al., p. 944.
Macrophages infiltrate both solid and hematologic tumors in response to tissue growth and remodeling cues that emanate from the cancer cells and their subverted microenvironments. Once recruited, subsets of macrophages sustain tumor angiogenesis, promote immunosuppression, and facilitate cancer cell dissemination to distant organs (1). However, most preclinical studies indicate that macrophages are not indispensable for tumor growth and can acquire either tumor-promoting or antagonizing functions in response to distinct anticancer therapies (1, 2). For example, macrophages can engulf and kill cancer cells that have been opsonized with therapeutic antibodies, a process known as antibody-dependent cellular phagocytosis (ADCP). In this issue of Cancer Discovery, Lossos and colleagues use mice engrafted with patient-derived lymphomas to show that high doses of the alkylating agent cyclophosphamide instigate lymphoma cells to induce “superphagocytic” macrophages that clear the tumor in the presence of alemtuzumab, an anti-CD52 antibody (Fig. 1; ref. 3). Notably, doxorubicin, a nonalkylating chemotherapeutic agent, failed to elicit ADCP in the same lymphoma model. These results add weight to the notion that duly activated macrophages can synergize with anticancer agents to potentially eradicate tumors in mice.
Some aggressive B-cell lymphomas carry activating translocations of both MYC and BCL2 (or BCL6). Patients with these double-hit lymphomas (DHL) are orphans of standard-of-care therapy and are frequently treated with combinations of drugs that include cyclophosphamide, doxorubicin, prednisone (a glucocorticoid), and the anti-CD20 antibody rituximab, along with alternative agents. However, most of the patients succumb in spite of intense treatment (4). To establish a mouse model of treatment-refractory disease, Lossos and colleagues (3) isolated CD52+ lymphoma cells from patients who failed R-CHOP (rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone). After inoculation in immunodeficient mice, the lymphoma cells developed a lethal form of DHL that recapitulated some of the key features of recalcitrant human tumors, such as massive involvement of the bone marrow (BM). This preclinical DHL model allowed for investigating mechanisms of resistance to chemoimmunotherapy and exploring alternative treatment options.
The authors found that alemtuzumab in combination with a high but not a standard dose of cyclophosphamide debulked lymphoma cells in the BM and significantly extended the survival of the mice. Under the same conditions, the MTD of doxorubicin provided only limited therapeutic benefits. Because both cyclophosphamide and doxorubicin induced DNA damage and primed lymphoma cells for apoptosis, noncytotoxic mechanisms elicited by high-dose cyclophosphamide had likely contributed to DHL clearance. Indeed, the authors observed pervasive infiltration of macrophages in the BM of cyclophosphamide-treated mice, which occurred in both the presence and absence of alemtuzumab. Further studies indicated that lymphoma cells exposed to high-dose cyclophosphamide produced increased amounts of VEGFA, a cytokine that induces angiogenesis and vascular permeability through VEGFR2, but also monocyte and macrophage recruitment through VEGFR1.
A series of experiments demonstrated that the cyclophosphamide–VEGFA–macrophage axis contributed to the efficacy of alemtuzumab in the DHL model. First, the systemic elimination of macrophages with clodronate liposomes impeded lymphoma debulking in the BM of DHL mice treated with cyclophosphamide and alemtuzumab. Likewise, VEGFA blockade with the neutralizing antibody bevacizumab limited lymphoma clearance in response to cyclophosphamide and alemtuzumab. Third, both medium conditioned by cyclophosphamide-treated lymphoma cells and recombinant VEGFA increased phagocytosis of alemtuzumab-opsonized lymphoma cells by macrophages; this effect could be abrogated through inhibition of the SYK tyrosine kinase, which mediates ADCP downstream to immunoglobulin receptors (FCGR). Finally, high-dose cyclophosphamide decreased the expression of the “don't eat me signal” CD47 on lymphoma cells, which may have made them more amenable to macrophage-mediated phagocytosis. Together, these findings strongly argue that high-dose cyclophosphamide induces a non–cell autonomous synthetic lethal process that is mediated by lymphoma cell–derived VEGFA and culminates with enhanced macrophage ADCP (3).
RNA-sequencing analysis of lymphoma cells isolated from the BM of DHL mice revealed different gene signatures in response to cyclophosphamide and doxorubicin. High-dose cyclophosphamide induced an endoplasmic reticulum (ER) stress response and upregulated the expression of the ER stress–associated activating transcription factor 4 (ATF4), which was shown previously to induce VEGFA transcription (5). Accordingly, chromatin immunoprecipitation qPCR assays indicated that ATF4 was recruited to the VEGFA gene in cyclophosphamide-treated lymphoma cells, suggesting a potential mechanism for VEGFA induction by cyclophosphamide (3).
In a parallel set of experiments, single-cell RNA-sequencing analysis of macrophages isolated from the BM of DHL mice identified two macrophage clusters that expanded in response to cyclophosphamide. Both cluster 2 (C2) and cluster 4 (C4) macrophages were enriched in ADCP-associated genes, including several FCGRs. In particular, C4 macrophages accounted for about 20% to 25% of BM macrophages after cyclophosphamide, had a gene signature consistent with upstream VEGFA regulation (according to ingenuity pathway analysis), expressed the activatory FCGRIV, and had superior ADCP activity toward opsonized lymphoma cells compared with macrophages belonging to other clusters. Collectively, these data implicate cyclophosphamide-induced “superphagocytic” macrophages as key effectors of DHL clearance by alemtuzumab.
The preclinical findings of Lossos and colleagues (3) have several potential implications for the treatment of human DHL. The authors observed that resistance to alemtuzumab monotherapy correlated with disease burden in the mouse BM, which was associated with a proportional demise of macrophage infiltrates. Thus, a low lymphoma:macrophage ratio predicts resistance to alemtuzumab, at least in the DHL model employed by the authors. In one clinical study, high pretreatment macrophage infiltration in B-cell follicular lymphomas was associated with a more favorable outcome after R-CHOP therapy (6). In another study, a high macrophage count in diffuse large B-cell lymphomas predicted poor response to chemotherapy, but this effect was reversed with a rituximab-containing regimen (7). Together, these clinical results make a compelling case for macrophages being antitumoral in the context of chemoimmunotherapy. However, the dramatically different sensitivity of distinct B-cell lymphoma subtypes to R-CHOP complicates generalizations of the prognostic value of macrophage infiltrates in lymphoid neoplasms.
Alemtuzumab is not yet indicated for DHL, but is approved for a number of other indications, such as chronic lymphocytic leukemia. A clinical trial investigating high-dose cyclophosphamide plus alemtuzumab (NCT03132584) was prematurely terminated, so clinical validation of the authors' findings may not become available in the near future. Conventional treatments for DHL involve lower cyclophosphamide doses than those employed by Lossos and colleagues; notably, standard cyclophosphamide doses failed to induce ER stress, ATF4 expression, and secretion of VEGFA in lymphoma cells, emphasizing the importance of high-dose cyclophosphamide for unleashing synthetic lethal mechanisms mediated by macrophages (3). Also, it remains to be seen whether alemtuzumab and rituximab would be interchangeable and equally effective after induction therapy with high-dose cyclophosphamide. Although the authors showed that high-dose cyclophosphamide could sensitize CD20+ Raji cells to macrophage ADCP mediated by rituximab, the DHL xenografts were CD52-positive but CD20-negative. So, it was impossible to test whether high-dose cyclophosphamide could reverse resistance to rituximab in the mouse model.
Most of the findings reported by Lossos and colleagues (3) reproduce and extend previous work in a genetically engineered model of human DHL (8). Nonetheless, Lossos and colleagues offer new interesting insight into the effects of high-dose cyclophosphamide on the phenotypic and functional heterogeneity of lymphoma-associated macrophages. Importantly, molecular profiling of FCGRIV+ (C4) “superphagocytic” macrophages might offer clues for pharmacologically inducing or expanding this macrophage subset in patients with DHL without concurrent high-dose cyclophosphamide, which has limiting toxicity in the generally older patients with DHL. Further work should also address some limitations of this study. For example, it is unclear whether C4-like macrophages would also expand in the BM of patients with lymphoma treated with high-dose cyclophosphamide. The human:mouse chimeric model employed by the authors has several operational advantages, but it cannot be ruled out that cyclophosphamide-induced expansion of C4 mouse macrophages represents an idiosyncratic consequence of allospecific cellular interactions. Another potential limitation involves the use of clodronate liposomes to deplete macrophages. Clodronate is a bisphosphonate that preferentially and efficiently kills various tissue- and tumor-resident macrophage populations. However, in doing so it also induces preapoptotic macrophages to acutely release cytokines and metabolites that might indirectly affect the biology of lymphoma cells and their microenvironments.
Whereas macrophages phagocytose malignant B cells in response to high-dose alkylating agents (3, 8), they support tumor growth in other experimental settings. For example, the selective elimination of macrophages with an anti-CSF1R antibody delayed progression of chronic lymphocytic leukemia by reprogramming the tumor microenvironment and sensitizing leukemic cells to rituximab (9). Adding to the complexity, FCGRs expressed on macrophages can also avidly sequester therapeutic antibodies, such as immune checkpoint inhibitors, preventing their binding to cognate cellular targets and blunting therapeutic responses in mouse models of cancer (10). These observations underscore the multifaceted and dynamic nature of tumor–macrophage interactions and mandate that interventions based on targeting macrophages be carefully pondered against the effects of concomitant therapy on these resourceful yet elusive cells.
Disclosure of Potential Conflicts of Interest
M. De Palma reports receiving commercial research grants from Deciphera Pharmaceuticals and Hoffmann-La Roche, and is a consultant/advisory board member for Deciphera Pharmaceuticals. No potential conflicts of interest were disclosed by the other author.
M. De Palma acknowledges support from the European Research Council (ERC-EVOLVE-725051), the Swiss National Science Foundation (SNF 31003A-165963), and the Swiss Cancer League (KFS-3759-08-2015). Scientific images displayed in Fig. 1 were retrieved from Smart (Servier Medical Art).