Neutrophil extracellular traps induced by chemotherapy inhibit tumor growth in murine models of colorectal cancer

Abstract

Neutrophil extracellular traps (NETs), a web-like structure of cytosolic and granule proteins assembled on decondensed chromatin, kill pathogens and cause tissue damage in diseases. Whether NETs can kill cancer cells is unexplored. Here, we report that a combination of glutaminase inhibitor CB-839 and 5-FU inhibited the growth of PIK3CA-mutant colorectal cancers (CRCs) in xenograft, syngeneic, and genetically engineered mouse models in part through NETs. Disruption of NETs by either DNase I treatment or depletion of neutrophils in CRCs attenuated the efficacy of the drug combination. Moreover, NETs were present in tumor biopsies from patients treated with the drug combination in a phase II clinical trial. Increased NET levels in tumors were associated with longer progression-free survival. Mechanistically, the drug combination induced the expression of IL-8 preferentially in PIK3CA-mutant CRCs to attract neutrophils into the tumors. Further, the drug combination increased the levels of ROS in neutrophils, thereby inducing NETs. Cathepsin G (CTSG), a serine protease localized in NETs, entered CRC cells through the RAGE cell surface protein. The internalized CTSG cleaved 14-3-3 proteins, released BAX, and triggered apoptosis in CRC cells. Thus, our studies illuminate a previously unrecognized mechanism by which chemotherapy-induced NETs kill cancer cells.

Keywords: Colorectal cancer; Neutrophils; Oncogenes; Oncology.

Figure 1. The combination of CB-839 and 5-FU induces NETs in xenograft tumors in nude mice.

(A and B) Xenograft tumors of HCT116 were treated with the indicated drugs 5 days on and 2 days off in nude and NSG mice simultaneously, with the growth curves shown in (A) nude mice and (B) NSG mice (5 mice/group). (C and D) Mice were implanted with HCT116 cells, and after 6 days, mice were injected with either 100 μL liposome control or clodronate twice a week (C), IgG control or anti-GM1 twice a week (D), growth curves shown in C for macrophage depletion and D for NK cell depletion (5 mice/group). Mice were treated CB-839, 5-FU, and the drug combination daily continuously. (E–G) CRC xenograft tumors were treated with vehicle (veh) or drug combination (comb) with or without Ly6g antibody injection, tumor sizes were measured, and growth curves are shown in E for HCT116, F for DLD1, and G for RKO (5 mice/group). (H–L) Tumors shown in A and B were stained with antibodies against MPO, which marks neutrophils and NETs, and H3cit, which marks NETs. Representative images are shown in H. Scale bar: 50 μm, and quantifications are shown in I–L (n = 15/group). (M–P) Tumors treated with the indicated drugs were stained with antibodies against MPO or H3cit and quantified (n = 15/group). The growth curves of the drug treatment were published in Zhao et al. (10). Data in (A–G) are plotted as mean + SEM. 2-way ANOVA (A–G) or 1-way ANOVA (I–P) was used for statistical analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 2. Disruption of NETs by DNase I treatment or depletion of neutrophils attenuates the tumor-inhibitory effect of the combination of CB-839 and 5-FU.

(A–G) The indicated xenograft tumors in nude mice were treated with vehicle (veh) or the drug combination (comb) with or without DNase I (5 mice/group). Tumor growth curves are shown in A, D, and E. Tumors were stained with antibodies against MPO and H3cit. Representative images of HCT116 tumors are shown in B. Quantifications are shown in C, F, and G. (n = 15/group). Scale bar: 50 μm. 2-way ANOVA (A, D, and E) or 1-way ANOVA (C, F and G) was used for statistical analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 3. The combination of CB-839 and 5-FU induces IL-8 expression in CRCs to attract tumor-infiltrating neutrophils.

(A and B) The indicated cells were treated with the indicated drugs. IL-8 mRNA and protein levels were measured by qRT-PCR (A) and ELISA (B), respectively, (n = 3/group). (C–I) Parental HCT116, DLD1, RKO, and their IL-8 KO clones were grown as xenograft tumors in nude mice and treated with vehicle or the combination of CB-839 and 5-FU (5 mice/group). Tumor growth curves are shown in (C, F, and G). Tumors were stained with antibodies against MPO and H3cit. Representative images of HCT116 tumors are shown in (D). Quantifications are shown in (E, H and I), (n = 15/group). (J and K) The indicated CRC cells were transfected with scramble siRNA or 2 independent siRNA against IL-8. IL-8 mRNA levels are shown in (J), and secreted IL-8 levels are shown in (K), (n = 3/group). Scale bar: 50 μm. 2-way ANOVA (C, F, G, E, H, and I) or 1-way ANOVA (A, B, J, and K) was used for statistical analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 4. The combination of CB-839 and 5-FU acts on neutrophils to induce NETs and cause apoptosis in cancer cells.

(A–C) Purified neutrophils were treated with the indicated compounds and stained with antibodies against MPO and H3cit. Representative images are shown in A, and quantifications are shown in B, (n = 15/group). The levels of ROS are shown in C (n = 4/group). Diphenyleneiodonium (DPI, 2 μM) is a ROS scavenger. Neutrophils were treated with phorbol myristate acetate (PMA, 1 μM) as a positive control. (D–I) (D) Schematics of the experiment setup. Cancer cells or neutrophils (NET) were treated with the combination of CB-839 and 5-FU, the conditioned media were collected, and protein concentrations of the conditioned media were measured. The indicated conditioned media were diluted to a final concentration of 10 μg/mL, or 30 μg/mL in fresh McCoy’s 5A medium to treat HCT116 cancer cells overnight. Cell viabilities are shown in E (n = 5/group), caspase 3 activities are shown in F (n = 3/group), annexin V positive cells are shown in G and H, (n = 3/group) and cleaved PARPs are shown in I. HCT116 cells were treated with camptothecin (3 μM) as a positive control. (J–O) Tunel staining was performed on the HCT116 xenograft tumors with indicated treatment. Representative images of HCT116 tumors are shown in J, L, and N, and quantifications are shown in K, M, and O (n = 15/group). Scale bar: 50 μm. 1-way ANOVA (B, C, E, F, H, K, M and O) was used for statistical analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 5. CTSG in NETs inhibits tumor growth.

(A–C) The HCT116 CRC cells were treated with the indicated recombinant proteins (10 μg/mL) overnight. Cell lysates were blotted with the indicated antibodies (A). The HCT116 cells were treated with the indicated recombinant proteins. Cells were treated with camptothecin (3 μM) in parallel as a positive control. Caspase 3 activities were measured (B) (n = 3/group). Cell numbers are shown in C (n = 3/group). (D–G) HCT116 cells were treated with indicated concentrations of recombinant CTSG. Caspase 3 activities are shown in D (n = 3/group), cell numbers were counted in E (n = 3/group). Annexin V staining and quantifications are shown in F and G (n = 3/group). (H–J) The indicated xenograft tumors were established in nude mice and treated with vehicle or the drug combination with or without CTSGi (intratumor injection, 5 mice/group). (K–N) Tunel staining of the tumors is shown in H–J. Representative images of HCT116 tumors are shown in K. Quantifications are shown in L–N (n = 15/group). (O–Q) Tumors from H stained with anti-MPO and anti-H3cit antibodies. Representative images of HCT116 tumors are shown in O, and quantifications are shown in P and Q (n = 15/group). Scale bar: 50 μm. 1-way ANOVA (B–E, G, L–N, P, and Q) or 2-way ANOVA (H–J) was used for statistical analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 6. CTSG enters cancer cells through RAGE, cleaves 14-3-3ε, results in BAX mitochondrial translocation, and triggers apoptosis.

(A and B) Parental HCT116, DLD1, RKO, and their RAGE KO cells were incubated with 5μg/mL recombinant CTSG for the indicated time. Representative images of immunofluorescent staining of an anti-CTSG antibody are shown in A, and Western blots of cleaved PARP are shown in B. (C–E) CRC RAGE WT and KO tumors treated with vehicle or drug combination, tumor growth curves are shown in C for HCT116, D for DLD1, and E for RKO (5 mice/group). (F) CRC cells were treated with recombinant CTSG or NET-conditioned medium for 16 hours, and cell lysates were blotted with indicated antibodies. Corrected loading control provided. (G–J) Tunel staining was performed on the CRC xenograft tumors from C–E. Representative images of HCT116 tumors are shown in G, and quantifications are shown in H for HCT116, I for DLD1, and J for RKO (n = 15/group). (K–M) Western blot of 14-3-3ε protein levels in the tumors with indicated treatment. (N–P) CRC cells were treated with 5μg/mL CTSG for the indicated time, and mitochondrial and cytosolic fractions were extracted by a cell fractionation kit and blotted with the indicated antibodies. 2-way ANOVA (C–E and H–J) was used for statistical analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 7. The combination of CB-839 and 5-FU induces NETs in syngeneic and GEM PIK3CA-mutant tumor models.

(A and B) CMT93 Pik3ca WT or E545 K mutant tumors were treated with indicated drugs, with growth curves shown in A for CMT93 Pik3ca WT and (B) for CMT93 Pik3ca E545K mutant. (5 mice/group). (C–E) The indicated tumors were stained with antibodies against MPO and H3cit. Representative images are shown in C. Quantifications shown in D and E (n = 15/group). (F and G) MC38 Pik3ca WT or mutant tumors with indicated treatment, growth curve shown in F for MC38 Pik3ca WT, G for MC38 PIK3CA mutant (5 mice/group). (H–J) The indicated tumors were stained with antibodies against MPO and H3cit. Representative images are shown in H. Quantifications shown in I and J (n = 15/group). (K) CDX2P-CreER^T2 Apc^fl/+ Kras^LSL–G12D/+ Pik3ca^{LSL–H1047R/+} mice were treated with tamoxifen and then treated with the indicated drug a week after tamoxifen treatment for 4 weeks. Kalplan-Meier curves of the mice are shown. A log-rank test was used to assess the statistical significance between the vehicle and the combination of CB-839 and 5-FU treatment groups. (L–N) CDX2P-CreER^T2 Apc^fl/+ Kras^LSL–G12D/+ Pik3ca^{LSL–H1047R/+} mice were treated with tamoxifen, and 4 weeks later, the mice were treated with the indicated drugs, and colon tumors were harvested and stained with antibodies against MPO and H3cit (3 mice/group). Representative images are shown in L. Quantifications of MPO are shown in M. Quantifications of H3cit are shown in N (n = 15/group). 2-way ANOVA (A, B, D–G, I and J) or 1-way ANOVA (N) was used for statistical analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Scale bar: 50 μm.

Figure 8. Disruption of NETs in syngeneic mouse tumors attenuates the tumor-inhibitory effect of the combination of CB-839 and 5-FU.

The indicated syngeneic tumors in C57/BL6 mice were treated with vehicle (veh), or the drug combination (comb) with or without DNase I. Tumor growth curves are shown in A for CMT93 Pik3ca E545K mutant tumors and B for MC38 Pik3ca E545K mutant tumors (5 mice/group), respectively. Tumors were stained with antibodies against MPO and H3cit. Representative images are shown in C for CMT93 Pik3ca E545K tumors and H for MC38 Pik3ca E545K tumors, respectively. Quantifications are shown in D–G for CMT93 Pik3ca E545K tumors and I–L for MC38 Pik3ca E545K tumors, respectively (n = 15/group). Scale bar: 50 μm. 2-way ANOVA (A and B) or 1-way ANOVA (D–G and I–L) was used for statistical analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 9. Pik3ca E545K mutation upregulates CXCL5 in mouse CRC cells.

(A and B) RT-PCR of CXCL1 and CXCL2 levels in CMT93 Pik3ca WT and mutant cells with the indicated treatment overnight (n = 3/group). (C) RT-PCR of CXCL1, CXCL2, and CXCL5 in CMT93 E545K cells (n = 3/group). (D and E) RT-PCR of CXCL5 in Pik3ca WT and E545K cells. D for CMT93, E for MC38 (n = 3/group). (F–I) CXCL5 protein level in culture medium (F and G) and tumors (H and I) was measured by ELISA (n = 3/group for F and H, n = 4/group for G and I). (J–O) p65 was knocked down by 2 independent siRNA. Cell lysates were blotted with the indicated antibodies (J and K). CXCL5 mRNA levels were measured by qRT-PCR (L and M). Secreted CXCL5 was measured by ELISA (N and O) (n = 3/group). (P) CMT93 PIK3CA mutant cells were treated with the combination of CB-839 and 5-FU. ChIP-PCRs were performed (n = 3/group). 2 tailed t test for P, 2-way ANOVA (B and D–I) or 1-way ANOVA (C and L–O) was used for statistical analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 10. CTSG in NETs induces apoptosis in the syngeneic model.

(A and B) The indicated syngeneic tumors were treated with vehicle or the drug combination with or without CTSGi (5 mice/group). Growth curves are shown in A and B. (C–E) Tunel staining was performed on the syngeneic tumors from A and B. Representative images of CMT93 tumors are shown in C, and quantifications are shown in D and E (n = 15/group), (F–J) Tumors from A and B were stained with antibodies against MPO and H3cit (F) and quantified (G–J). (K–M) MC38 Pik3ca E545K mutant cells were injected into C57/BL6 mice of the indicated genotypes and treated with vehicle or the combination of CB-839 and 5-FU (5 mice/group). The growth curves are shown in K. Tunel staining was performed on the syngeneic tumors; representative images are shown in L, and quantifications are shown in M. (N–P) The tumors shown in K were stained with antibodies against MPO and H3cit. Representative images are shown in N. Quantifications are shown in O and P (n = 15/group). (Q) mouse CRC cells treated with CTSG and NET medium for 16 hours, cell lysates were harvested and blotted with indicated antibodies. (R and S) tumors shown in A and B were blotted with indicated antibodies. 2-way ANOVA (A, B, K, M, and P) or 1-way ANOVA (D, E, H, and J) was used for statistical analysis. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001.

Figure 11. Higher tumor NET levels are associated with longer PFS.

(A) PFS of 32 metastatic PIK3CA mutant CRC patients in a phase II clinical trial of the combination of CB-839 and capecitabine, an oral prodrug of 5-FU. Red bars, stable disease; black bars, progressive disease; and gray bars, not evaluable. (B and C) Tumor biopsies were stained with anti-MPO and anti-H3cit antibodies. Representative images are shown in B, and quantifications are shown in C. ** P < 0.01; paired Student’s t test, 2-tailed. Scale bar: 100 μM. (D) Kaplan-Meier curves of PFS of patients are plotted with increased levels of H3cit in posttreatment biopsies compared with pretreatment biopsies versus decreased levels of H3cit levels in posttreatment biopsies compared with pretreatment biopsies. (E) Kaplan-Meier plot of PFS of patients whose posttreatment biopsies with high levels of H3cit versus those with low levels of H3cit. (F) A model of CB-839 plus 5-FU-induced NETs, which release CTSG, enter cancer cells, cleave 14-3-3ε, lead to BAX mitochondrial translocation, and trigger apoptosis.