Perivascular M2 Macrophages Stimulate Tumor Relapse after Chemotherapy

Abstract

Tumor relapse after chemotherapy-induced regression is a major clinical problem, because it often involves inoperable metastatic disease. Tumor-associated macrophages (TAM) are known to limit the cytotoxic effects of chemotherapy in preclinical models of cancer. Here, we report that an alternatively activated (M2) subpopulation of TAMs (MRC1(+)TIE2(Hi)CXCR4(Hi)) accumulate around blood vessels in tumors after chemotherapy, where they promote tumor revascularization and relapse, in part, via VEGF-A release. A similar perivascular, M2-related TAM subset was present in human breast carcinomas and bone metastases after chemotherapy. Although a small proportion of M2 TAMs were also present in hypoxic tumor areas, when we genetically ablated their ability to respond to hypoxia via hypoxia-inducible factors 1 and 2, tumor relapse was unaffected. TAMs were the predominant cells expressing immunoreactive CXCR4 in chemotherapy-treated mouse tumors, with the highest levels expressed by MRC1(+) TAMs clustering around the tumor vasculature. Furthermore, the primary CXCR4 ligand, CXCL12, was upregulated in these perivascular sites after chemotherapy, where it was selectively chemotactic for MRC1(+) TAMs. Interestingly, HMOX-1, a marker of oxidative stress, was also upregulated in perivascular areas after chemotherapy. This enzyme generates carbon monoxide from the breakdown of heme, a gas known to upregulate CXCL12. Finally, pharmacologic blockade of CXCR4 selectively reduced M2-related TAMs after chemotherapy, especially those in direct contact with blood vessels, thereby reducing tumor revascularization and regrowth. Our studies rationalize a strategy to leverage chemotherapeutic efficacy by selectively targeting this perivascular, relapse-promoting M2-related TAM cell population.

Figure 1

Effects of cyclophosphamide (CTX) on tumor growth and accumulation of M2 TAMs in LLC1 tumors. A and B, growth kinetics of LLC1s after three i.p. injections with either 150 mg/kg cyclophosphamide or PBS (left; A), the number of MRC1⁺ or MRC1⁻ F4/80⁺ TAMs (right; n = 7–8/group), and their cell-surface expression of MRC1 (B), as detected by flow cytometry (n = 5–6/group), 2 days after the last dose of cyclophosphamide (day 6, i.e., before the regrowth phase). C, flow-cytometric analysis of TIE2 and MRC1 on TAMs from cyclophosphamide-treated tumors. D, colocalization of MRC1 and TIE2 on TAMs in cyclophosphamide-treated tumors (see yellow arrows and inset for dual-stained cells). V, vessel. Bars, 50 μm. ^*, P < 0.05.

Figure 2

Effects of adoptive transfer of MRC1^Hi vs MRC1^Lo TAMs on LLC1 relapse after cyclophosphamide (CTX). A, design schematic of the TAM transfer experiment. B, quantification of MRC1⁺ TAMs in tumors receiving adoptive transfer of MRC1^−/Lo or MRC1^+/Hi TAMs. C, LLC1 regrowth after the last cyclophosphamide injection (day 0). D and E, vascularization (D) and proliferation (E) in LLC1 tumors receiving MRC1^Hi or MRC1^Lo (n = 5/group). Costaining of cyclophosphamide-treated LLC1 tumor sections with antibodies against the pan-leukocyte marker (CD45) and BrdUrd. Fewer than 1% of proliferating cells were leukocytes in either group. ^*, P < 0.05. Bars, 50 μm.

Figure 3

Effect of cyclophosphamide (CTX) on the number of MRC1⁺ VEGFA⁺ TAMs in LLC1 tumors and VEGFA release by TAMs in vitro. A, representative immunostaining for F4/80⁺, MRC1⁺, and VEGFA⁺. VEGF-expressing F4/80⁻ cells (yellow arrows) and F4/80⁺ TAMs (orange arrows). B, VEGFA colocalization with MRC1 in TAMs. Bar, 50 μm. C, number of MRC1⁺ VEGFA⁺ TAMs. D, VEGFA release by CD45⁺CD11b⁺Ly6G⁻F4/80⁺ TAMs (VEGFA release in medium over 16 hours, standardized by live TAM numbers). ^*, P < 0.05.

Figure 4

Effect of cytotoxic drugs on the distribution of MRC1⁺ TAMs in mouse and human tumors: colocalization with VEGFA. A, immunostaining of MRC1⁺ TAMs in PIMO⁻VA and hypoxic (PIMO⁺) areas of LLC1 tumors 48 hours after last injection of cyclophosphamide (CTX; left and middle), and their abundance in PIMO⁻ VA areas normalized by CD31⁺ area (right; n = 4/group). B, MRC1⁺ TAM accumulation in vessel-associated (VA) areas in 4T1 (n = 8–9/group) and MMTV-PyMT tumor implants after treatment with paclitaxel (PTX) or doxorubicin (DOX) respectively (n = 7/group). C, overall number of MRC1⁺VEGFA⁺ TAMs in PIMO⁻ VA (top) and the number in direct contact with CD31⁺ vessels (abluminal; bottom) in control and cyclophosphamide-treated LLCs 48 hours after the final dose of cyclophosphamide (both normalized by total CD31⁺ area). D, MRC1⁺ TAMs in vascular or avascular areas of human primary carcinomas 3 weeks after three cycles of paclitaxel treatment (black arrows; n = 4 biopsies). MRC1⁺ macrophages near vessels in bone metastases from patients with advanced breast cancer after treatment (red arrows; n = 4 biopsies). Bars, 50 μm. ^*, P < 0.05.

Figure 5

Effect of TAM-derived VEGFA on relapse of orthotopic MMTV-PyMT tumors after doxorubicin (DOX) treatment. A, representative immunostaining for VEGFA in MRC1⁺ TAMs in vascularized (CD31⁺) areas of doxorubicin-treated, Cre⁻, MMTV-PyMT tumors. This was not seen for MRC1⁻ TAMs anywhere in the same tumors (B). C, female Tg(Csf1r-Mer-iCre-Mer)_1jwp; Vegfa^fl/fl) mice bearing MMTV-PyMT tumors were administered tamoxifen for 24 hours to delete VEGFA expression in TAMs (right panels) after a single injection of either vehicle alone or doxorubicin. D and E, growth of tumors treated with vehicle alone in Cre⁺ and Cre⁻ mice (n = 3–6/group; D) and regrowth of tumors in Cre⁺ and Cre⁻ mice after treatment with doxorubicin (n = 3–4/group; E). F, CD31 staining of vessels in tumors in Cre⁻ or Cre⁺ mice given doxorubicin. Bars, 50 μm. ^*, P < 0.05 with respect to tumors at the same time point in the respective Cre⁻ group.

Figure 6

Expression of CXCR4 by MRC1⁺ TAMs and upregulation of CXCLl2 in LLC1 tumors after cyclophosphamide (CTX) treatment. A and B, immunostaining of F4/80 and CXCR4 (A), and the percentage of CXCR4⁺ cells coexpressing F4/80 in control and cyclophosphamide-treated LLCs (n = 4/group; B). C, immunostaining of CXCR4 and CD31 in cyclophosphamide-treated LLCs. D, flow-cytometric analysis of TAM expression of CXCR4 and MRC1 in dispersed cyclophosphamide-treated LLCs (left), CXCR4 MFI on MRC1⁺ vs. MRC1⁻ TAMs (middle; n = 4–5/group), and tumor levels of immunodetectable CXCL12 protein in control and cyclophosphamide-treated tumors (n = 4/group; right). E, CXCL12⁺ cells were perivascular in cyclophosphamide-treated LLC1s (left) and exogenous recombinant human CXCL12 was chemotactic for MRC1^+/Hi (but not MRC^−/Lo) TAMs isolated from LLC1 tumors. Bars, 50 μm. ^*, P < 0.05.

Figure 7

Effect of the CXCR4 inhibitor AMD3100 on LLC1 relapse after cyclophosphamide (CTX): role of perivascular MRC1⁺CXCR4^Hi TAMs. A, regrowth of tumors in mice treated with PBS, PBS + AMD3100, cyclophosphamide + PBS, or cyclophosphamide + AMD3100 (n = 5–9/group). B and C, F4/80⁺MRC1⁺ TAMs in the PIMO⁻ VA and hypoxic (PIMO⁺) areas of cyclophosphamide + PBS or cyclophosphamide + AMD3100 treated tumors (n = 4–6/group) at day 6 (B) and day 10 (n = 7–9/group; C). D, total CD31⁺ vessel area in tumors treated with cyclophosphamide + PBS and cyclophosphamide + AMD3100 (at day 6 in A). E, MRC1⁺ TAMs in direct contact (abluminal) or not in contact with (non-abluminal) with CD31⁺ endothelial cells in PIMO⁻ VA areas of either cyclophosphamide + PBS or cyclophosphamide + AMD3100-treated LLCs (normalized to total CD31⁺ area in each field). ^*, P < 0.05.