G-CSF and Bv8 are increased in refractory tumors. (A) Cytokine levels in the CM of tumor cells lines. All cell lines were incubated in DMEM plus 10% FBS for 72 h. CM were then collected and concentrated by Amicon Spin Columns (10-kDa MW cutoff; Millipore). Levels of GM-CSF, SDF1α, PlGF, Bv8, and G-CSF were measured by ELISA. (B and C) Balb-c nude mice were implanted with sensitive or refractory tumors as described, and were treated with control or anti-VEGF mAbs. By using ELISA, levels of GM-CSF, SDF1α, PlGF, Bv8, and G-CSF were quantified in the tumors (B) and plasma (C). *, significant difference (P < 0.05) when comparing levels of each cytokine in refractory tumors in the control/anti-VEGF treated mice with the corresponding ones in sensitive tumors.

Anti-G-CSF treatment targets CD11b⁺Gr1⁺ cells. Single cells were isolated from the tumors (A), PB (B), or BMMNCs (C) of mice bearing sensitive or refractory tumors, as indicated in the main text. Tumors were harvested when they reached ≈2,000 mm³ (late tumor growth), and were stained with rat anti-mouse CD11b-APC and rat anti-mouse Gr1-PE antibodies as described in SI Methods. Graphs represent mean frequency of CD11b⁺Gr1⁺ cells in each tissue, and bars represent SEM. *, significant difference (P < 0.05) in the frequency of myeloid cells when comparing each treatment vs. corresponding control-treated tumors. ▼, significant difference (P < 0.05) in combination treatments (anti-Bv8 plus anti-VEGF or anti-G-CSF plus anti-VEGF) vs. anti-VEGF monotherapy. Insets indicate frequency of myeloid cells in nontumor bearing mice.

Anti-G-CSF or anti-Bv8 treatments inhibit growth of refractory tumors. Balb-c nude mice [n = 10; 3 × 10⁶ cells per mouse] were s.c. implanted with B16F1 (A), Tib6 (B), EL4 (C), or LLC (D) cells. Mice were treated with control IgG, anti-VEGF, anti-Bv8, anti-G-CSF, anti-Bv8 plus anti-VEGF, or anti-G-CSF plus anti-VEGF, as described in Methods. Tumor volume was measured at several time points as indicated, starting at day 1 posttumor cell inoculation. Graphs represent mean tumor volume ± SEM. *, significant difference (P < 0.05) in tumor volume when comparing each treatment vs. corresponding control-treated tumors. ▼, significant difference (P < 0.05) when comparing combination treatments (anti-Bv8 plus anti-VEGF or anti-G-CSF plus anti-VEGF) with anti-VEGF monotherapy.

Anti-Bv8 or anti-G-CSF reduces tumor angiogenesis. (A) Sections from tumors treated with mono or combination treatments were stained with anti-CD31 (shown in red), anti-CD11b (shown in green), and anti-Gr1 (shown in pink) antibodies as described in SI Methods. Images of tumors (20×) were taken in a Zeiss confocal microscope. Due to space limitation, images from the control, anti-VEGF, anti-G-CSF, and combination of anti-VEGF and anti-G-CSF treatments have been included here. Images from other treatment groups are shown in Fig. S4. Consistent with FACS data (Fig. 2A), anti-G-CSF treatment significantly reduced CD11b⁺Gr1⁺ cells in the tumors. (B) Quantification of tumor vasculature indicated that anti-G-CSF or anti-Bv8 inhibits angiogenesis as monotherapy or in combination with anti-VEGF in refractory tumors. By using ImageJ, areas of CD31⁺ in each section were measured in 20–25 images from each treatment (n = 3). Bars represent the mean VSA ± SEM in each treatment. *, significant difference (P < 0.05) when comparing VSA in mono or combination therapy vs. controls. ▼, difference in combination treatment vs. anti-VEGF alone is significant (P < 0.05).

G-CSF reduces responsiveness to anti-VEGF. (A) Balb-c nude mice (n = 10) were implanted with B16F1 cells [3 × 10⁶ cells per mouse]. Mice received recombinant G-CSF or PBS i.p. for the first 4 days after tumor implantation and then at alternative days. Treatment with anti-VEGF or control mAbs was started at day 5 after tumor cell inoculation. Data represent mean tumor volumes ± SEM, and asterisks indicate significant difference when comparing G-CSF treated tumors in anti-VEGF treated mice vs. the corresponding control group. (B–E) Frequency of CD11b⁺Gr1⁺ cells in the tumors (B), PB (C), BMMNCs (D), and tumor-associated endothelial cells (CD31⁺CD45^neg; E) were measured by using the same FACS technique as described. Asterisks indicate a significant difference (P < 0.05) when comparing myeloid cells in G-CSF treated mice with those in the PBS treated group. Treatment with G-CSF confers reduced responsiveness to anti-VEGF through induction of angiogenesis and infiltration of myeloid cells. (F) B16F1 cells were cotransfected with a mG-CSF expression plasmid and a vector conferring Zeocin resistance. The transfected cells were selected, expanded in culture, and characterized as described in SI Methods. Nude mice (n = 10) were implanted with 5 × 10⁶ G-CSF- or control- transfected cells, and were treated with anti-VEGF, anti-Bv8, or control mAbs, starting at day 1 postinoculation. Data shown represent mean tumor volumes ± SEM, and asterisks indicate significant difference in B16F1-G-CSF tumors treated with anti-Bv8 or anti-VEGF vs. corresponding groups in the Vector tumors. (G–I) By using ELISA, levels of G-CSF (G) and Bv8 (H) in the tumors, and Bv8 in BMMNCs (I) were measured in all of the treatment groups. G-CSF is highly expressed in tumors derived from the G-CSF transfected lines, resulting in higher expression of Bv8 in tumors and BM. Note that the presence of anti-Bv8 antibodies may interfere with Bv8 measurements by ELISA.