Macrophage migration inhibitory factor downregulation: a novel mechanism of resistance to anti-angiogenic therapy

Abstract

Anti-angiogenic therapies for cancer such as VEGF neutralizing antibody bevacizumab have limited durability. While mechanisms of resistance remain undefined, it is likely that acquired resistance to anti-angiogenic therapy will involve alterations of the tumor microenvironment. We confirmed increased tumor-associated macrophages in bevacizumab-resistant glioblastoma patient specimens and two novel glioblastoma xenograft models of bevacizumab resistance. Microarray analysis suggested downregulated macrophage migration inhibitory factor (MIF) to be the most pertinent mediator of increased macrophages. Bevacizumab-resistant patient glioblastomas and both novel xenograft models of resistance had less MIF than bevacizumab-naive tumors, and harbored more M2/protumoral macrophages that specifically localized to the tumor edge. Xenografts expressing MIF-shRNA grew more rapidly with greater angiogenesis and had macrophages localizing to the tumor edge which were more prevalent and proliferative, and displayed M2 polarization, whereas bevacizumab-resistant xenografts transduced to upregulate MIF exhibited the opposite changes. Bone marrow-derived macrophage were polarized to an M2 phenotype in the presence of condition-media derived from bevacizumab-resistant xenograft-derived cells, while recombinant MIF drove M1 polarization. Media from macrophages exposed to bevacizumab-resistant tumor cell conditioned media increased glioma cell proliferation compared with media from macrophages exposed to bevacizumab-responsive tumor cell media, suggesting that macrophage polarization in bevacizumab-resistant xenografts is the source of their aggressive biology and results from a secreted factor. Two mechanisms of bevacizumab-induced MIF reduction were identified: (1) bevacizumab bound MIF and blocked MIF-induced M1 polarization of macrophages; and (2) VEGF increased glioma MIF production in a VEGFR2-dependent manner, suggesting that bevacizumab-induced VEGF depletion would downregulate MIF. Site-directed biopsies revealed enriched MIF and VEGF at the enhancing edge in bevacizumab-naive patients. This MIF enrichment was lost in bevacizumab-resistant glioblastomas, driving a tumor edge M1-to-M2 transition. Thus, bevacizumab resistance is driven by reduced MIF at the tumor edge causing proliferative expansion of M2 macrophages, which in turn promotes tumor growth.

Figure 1

Increased M2 macrophages at the tumor edge associated with bevacizumab-resistant glioblastoma. (a) Human glioblastomas (n=8/group) exhibited increased IBA1⁺ macrophages upon becoming bevacizumab-resistant compared to paired pre-treatment bevacizumab-naive specimens (P=0.02), while no such increase occurred in bevacizumab-naive glioblastomas after recurrence compared with before (P=0.8) Y-axis represents percent of the high-powered field that was immunopositive as derived by ImageJ software. (b) Absolute qPCR for microglia marker CX3CR1 and myeloid macrophage marker CCR2 on CD11b⁺ cells isolated from site-directed biopsies of the central core, enhancing edge, and infiltrated white matter of patient glioblastomas revealed that the ratio of CCR2:CX3CR1 expression increased robustly in all sites of bevacizumab-resistant (n=3) versus bevacizumab-naive (n=5) glioblastomas (P=0.006 central core; P=0.01 enhancing edge; P=0.02 infiltrated white matter) (c) U87-Bev^S and U87-Bev^R intracranial xenografts exhibited increased CD11b⁺ macrophages (red) upon treatment with bevacizumab compared with treatment with IgG (P=0.01). These macrophages were recruited to the tumor (green) edge (n=5/group). (d) THP-1 human monocytes exhibited decreased chemotaxis in response to U87-Bev^R conditioned media (CM) compared with U87-Bev^S CM (P<0.0001) (n=4/group), whereas (e) murine bone marrow derived macrophages had greater cell numbers when cultured in U87-Bev^R conditioned media for 48 h compared with U87-Bev^S conditioned media (P=0.04), with both U87-Bev^R (P<0.001) and U87-Bev^S (P=0.04) conditioned media increasing macrophage numbers (n=8/group). (f) qPCR of FACS-sorted CD11b⁺ TAMs isolated from intracranial U87-Bev^R and U87-Bev^S xenograft (n=5/group) revealed increased M2 macrophages in U87-Bev^R intracranial xenografts relative to U87-Bev^S intracranial xenografts, independent of treatment (P<0.0001), as revealed by M2/M1 polarization ratios determined by multiplying the qPCR fold increases in three different M2 primers divided by the qPCR fold increases in three different M1 primers normalized to results for U87-Bev^S xenografts treated with IgG.

Figure 2

Decreased MIF expression associated with bevacizumab-resistant glioblastoma. (a) Microarray analysis of U87-Bev^R relative to U87-Bev^S (n=3/group) was performed to assess tumor-secreted factors known to influence macrophages, with only two changes being statistically significant (*raw P<0.001): upregulated CCL2 and downgregulated MIF. (b) qPCR revealed reduced CCL2 and MIF transcript (P<0.05 each), while (c) Immunoblotting revealed reduced CCL2 and MIF protein in U87-Bev^R xenografts (n=5) compared to U87-Bev^S intracranial xenografts (n=5). (d) Immuoblotting revealed reduced MIF in bevacizumab-resistant patient glioblastomas (n=4) compared to bevacizumab-naive patient glioblastomas (n=4).

Figure 3

In cultured cells, MIF promotes M1 polarization, causing macrophages to be less phagocytic and exert a less pro-tumoral effect. (a) Murine bone marrow-derived macrophage precursors were matured for 5 days and rested for 3 days prior to an 18 hour polarization in media containing 200 ng/ml and 800 ng/ml MIF, 20 ng/ml IFN-γ (for M1 polarization), 20 ng/ml IL-4 (for M2 polarization), or conditioned media from U87-Bev^R or U87-Bev^S cells (n=3/group). M1/M2 polarization ratio was assessed by multiplication of the qPCR fold increases in three different M1 primers divided by the qPCR fold increases in three different M2 primers normalized to results when incubating with media alone. Recombinant MIF drove M1 polarization in a dose-dependent manner (P<0.0001), while, unlike media from U87-Bev^S cells, media from U87-Bev^R cells drove an M2 macrophage polarization (P<0.0001). (b) The phagocytic activity of THP-1 monocyte-derived macrophages without stimulation, cultured with cytokines (recombinant MIF, IFN-γ, and IL-4), or cultured with conditioned media (CM) was assessed via uptake of fluorescent heat-killed E. coli with subsequent measurement of the proportion of fluorescent cells (n=6/group). M1 polarized macrophages were less phagocytic whereas M2 polarized macrophages were more phagocytic relative to unstimulated control macrophages. Macrophages treated with 800 ng/ml MIF were less phagocytic relative to control (P=0.005). Macrophages treated with CM from U87-Bev^R cells were more phagocytic than macrophages treated with CM from U87-Bev^S (P=0.01) (example images shown to the right). (c) Sequential conditioned media (SCM) experiments were performed as illustrated to the left. Briefly, media from U87-Bev^R or U87-Bev^S cells was applied to THP-1-derived macrophages, and the media was then taken from those macrophages and applied to U87 cells, with numbers of U87 cells counted 48 h later (n=24/group). U87-Bev^R-derived macrophage conditioned media (U87-Bev^R MCM) stimulated a significantly greater expansion of U87 cells than U87-Bev^S-derived macrophage conditioned media (U87-Bev^S MCM) (P<0.0001), with U87-Bev^R MCM and U87-Bev^S MCM stimulating more (P=0.005) and less (P=0.003) U87 expansion than control media, respectively.

Figure 4

MIF reduction increases tumor-associated macrophages and drives M2 polarization in vivo. Intracranial xenografts established from U87 glioma cells transduced with two different MIF-targeted shRNA sequences versus control-targeted shRNA sequences exhibited (n=3/group) (a) larger tumor weights (P=0.04 U87/shMIF1 and P=0.03 U87/shMIF2); (b) more TAMs (red) (P=0.01), with the TAMs preferentially localizing to the tumor (green) periphery; (c) increased M2 TAM polarization based on elevated M2/M1 ratios (P<0.0001) in CD11b⁺ FACS-sorted TAMs analyzed by qPCR; (d) more proliferative TAMs based on FACS revealing a larger percentage of CD11b⁺ TAMs to be positive for the proliferation marker Ki67 (P<0.05); and (e) more vessel density based on CD31 vessel staining (P=0.03). Scale bars, 50 μm.

Figure 5

MIF overexpression in bevacizumab-resistant cells decreases tumor-associated macrophages and drives M1 polarization in vivo. Intracranial xenografts established from U87-Bev^R glioma cells transduced to overexpress MIF in two different clones relative to U87-Bev^R cells transduced with empty vector (EV) exhibited (n=3/group) (a) decreased tumor weight (P=0.03 U87-Bev^R/MIF1 and P=0.009 U87-Bev^R/MIF2); (b) increased CD11b+ TAMs (P=0.006 for both U87-Bev^R/MIF1 and U87-Bev^R/MIF2 compared to U87-Bev^R/EV) (CD11b⁺ cells red, GFP⁺ tumor cells green, DAPI rendering nuclei blue); (c) increased M1/M2 ratio (P=0.002 U87/Bev^R-MIF1, P=0.01 U87/Bev^R-MIF2); and (d) decreased vascularity (P=0.046 U87-Bev^R/MIF1 and P=0.04 U87-Bev^R/MIF2). Scale bars, 50 μm.

Figure 6

Dual mechanisms of bevacizumab-induced MIF depletion. (a) Immunoprecipitation revealed that bevacizumab bound MIF, with bevacizumab binding of VEGF serving as a positive control (upper portion of figure). Bevacizumab blocked the ability of recombinant MIF to drive M1 polarization in cultured bone marrow cells (lower portion of figure). (b) Treatment of cultured U87 and U251 cells with 100 ng/ml recombinant VEGF for varying time points up to 48 hours revealed that VEGF increased MIF transcription by qPCR, intracellular protein by western blot of whole cell lysate, and protein secretion by ELISA (P<0.01), offering a potential mechanism by which bevacizumab could reduce tumoral MIF. (c) Cultured U87 cells were treated with VEGF in the presence of blocking antibodies targeting VEGF receptors-1 and 2 (VEGR-1 and VEGFR-2). Blocking VEGFR-2 eliminated VEGF-induced MIF expression, while blocking VEGFR-1 had no effect on VEGF-induced MIF expression. (d) Site-directed biopsies taken from four bevacizumab-naive recurrent glioblastomas revealed increased MIF and VEGF RNA copies at the enhancing edge relative to the central core and FLAIR bright non-enhancing periphery, as assessed by absolute quantification qPCR (P=0.03). In contrast, site-directed biopsies taken from a bevacizumab-resistant recurrent glioblastoma revealed the same regional pattern in VEGF RNA levels, but loss of the spike in MIF RNA that typically occurs at the enhancing edge (P=0.02). Absolute quantification qPCR of column purified CD11b⁺ cells from bevacizumab-naive versus recurrent glioblastoma revealed elevated M2/M1 ratio at the central core and infiltrated white matter in both tumor types, but at the enhancing edge the bevacizumab-naive tumor had predominantly M1 macrophages, while the resistant tumor had predominantly M2 macrophages (P=0.008-0.01).