Vessel co-option mediates resistance to anti-angiogenic therapy in liver metastases

Abstract

The efficacy of angiogenesis inhibitors in cancer is limited by resistance mechanisms that are poorly understood. Notably, instead of through the induction of angiogenesis, tumor vascularization can occur through the nonangiogenic mechanism of vessel co-option. Here we show that vessel co-option is associated with a poor response to the anti-angiogenic agent bevacizumab in patients with colorectal cancer liver metastases. Moreover, we find that vessel co-option is also prevalent in human breast cancer liver metastases, a setting in which results with anti-angiogenic therapy have been disappointing. In preclinical mechanistic studies, we found that cancer cell motility mediated by the actin-related protein 2/3 complex (Arp2/3) is required for vessel co-option in liver metastases in vivo and that, in this setting, combined inhibition of angiogenesis and vessel co-option is more effective than the inhibition of angiogenesis alone. Vessel co-option is therefore a clinically relevant mechanism of resistance to anti-angiogenic therapy and combined inhibition of angiogenesis and vessel co-option might be a warranted therapeutic strategy.

Figure 1. Correlation between HGP and pathological response in patients treated preoperatively with bevacizumab

a. Diagrams illustrating the morphology of the normal liver or the morphology of the tumor-liver interface in liver metastases with a desmoplastic, pushing or replacement HGP. b. The HGPs and the pathological response to bev-chemo were scored in 59 CRCLMs from 33 patients treated preoperatively with bev-chemo at RM. Graph shows % HGP (replacement, desmoplastic, pushing) scored in each individual lesion and the data are grouped by pathological response score: >75%, 50–75%, 25–49% or <25% viable tumor. Median number of lesions examined per patient was 1 (range = 1 to 4 lesions per patient). c–e. Examples of H&E-stained lesions from the RM cohort are shown. Arrows point to examples of replacement HGP areas. Arrowheads point to examples of desmoplastic HGP areas. Asterisks indicate areas of infarct-like necrosis. f. The HGPs and the pathological response to bev-chemo were scored in 128 CRCLMs from 59 patients treated with bev-chemo at MUHC. Graph shows % HGP (replacement, desmoplastic, pushing) scored in each individual lesion and the data are grouped by pathological response score: >75%, 50–75%, 25–49% or <25% viable tumor. Median number of lesions examined per patient was 2 (range = 1 to 12 lesions per patient). The χ²-test was used to determine statistical significance (see 2 x 2 contingency tables in panels b and f). Scale bars, 1 mm.

Figure 2. Correlation between HGP and morphological response on CT in patients treated preoperatively with bevacizumab

a–f. CT scans of patients treated preoperatively with bev-chemo. Examples of optimal (OR), partial (PR) or absent (AR) morphological response are shown. a,b. OR; in the pre-treatment image a lesion in liver segment VII (arrowhead) is scored as group-3 (a); the same lesion imaged after 4 cycles of bevacizumab in combination with CAPOX is now scored as group-1 (b). c,d. PR; in the pre-treatment image a lesion in liver segment II (arrowhead) is scored as group-3 (c); the same lesion imaged after 4 cycles of bevacizumab in combination with CAPOX is now scored as group-2 (d). e,f. AR; in the pre-treatment image a lesion in liver segment VI (arrowhead) is scored as group-3 (e); the same lesion imaged after 6 cycles of bevacizumab in combination with FOLFIRI is still scored as group-3 (f). g. Morphological response and HGP were scored in 52 liver metastases from 31 patients treated preoperatively with bev-chemo at RM. Graph shows the % HGP scored in each individual lesion (replacement, desmoplastic, pushing). Lesions are grouped according to response: AR, PR or OR. Lesions scored as AR were classed as poor responders, whilst those scored as PR or OR were classed as good responders. Median number of lesions examined per patient was 1 (range = 1 to 4 lesions per patient). The χ² test was used to determine statistical significance (see 2 x 2 contingency table in panel g).

Figure 3. Cancer cells infiltrate the hepatic plates and co-opt sinusoidal blood vessels in the replacement HGP

a. An area of normal liver is shown. Staining is for hepatocyte specific antigen (HSA, green) to detect hepatocytes and collagen-3 (col-3, red) to detect liver sinusoidal blood vessels (SV). b–d. Staining for cancer cells (CK, red) and hepatocytes (HSA, green) at the tumor-liver interface (b,c) and within the tumor mass (d) in a replacement HGP liver metastasis of colorectal cancer. Examples of displaced hepatocytes are marked (arrowheads). e–g. Staining for cytokeratin 20 (CK20, brown) to identify cancer cells and CD31 to identify blood vessels (blue) in replacement HGP liver metastases of colorectal cancer. Arrows and arrowheads indicate examples of liver sinusoidal blood vessels where one end of the vessel is physically located in the liver parenchyma (arrows), whilst the other end is surrounded by cancer cells (arrowheads). Asterisk, tumor. Lv, normal liver. SV, sinusoidal blood vessel. Scale bars, 25 μM.

Figure 4. The replacement HGP occurs in progressive disease and is associated with a poor outcome in patients treated with bevacizumab

a. Left: HGPs in untreated CRCLMs (n = 32 lesions from 19 MUHC patients). Middle: HGPs in pre-existing CRCLMs (n = 128 lesions from 59 MUHC patients). Right: HGPs in new CRCLMs (n = 35 lesions from 13 MUHC patients). Graphs show % replacement (R), % desmoplastic (D) and % pushing (P) HGP per lesion ± SEM. b. Kaplan-Meier estimates of OS for 62 MUHC patients treated preoperatively with bev-chemo stratified into two groups: predominant replacement HGP (26 patients) or predominant desmoplastic HGP (35 patients). c. Kaplan-Meier estimates of OS for 29 MUHC patients treated preoperatively with chemotherapy alone stratified into two groups: predominant replacement HGP (12 patients) or predominant desmoplastic HGP (16 patients). d. Kaplan-Meier estimates of OS for 51 MUHC patients with a predominant desmoplastic HGP stratified into two groups: desmoplastic HGP treated with bev-chemo (35 patients) or desmoplastic HGP treated with chemotherapy alone (16 patients). e. Kaplan-Meier estimates of OS for 38 MUHC patients with a predominant replacement HGP stratified into two groups: replacement HGP treated with bev-chemo (26 patients) or replacement HGP treated with chemotherapy alone (12 patients). Kruskall-Walls test (a) or the Log-Rank test (b–e) were used to determine statistical significance. Hazard ratios were calculated using Cox-regression. * P<0.001.

Figure 5. The replacement HGP predominates in breast cancer liver metastases

a. The HGPs were examined in breast cancer liver metastases (BCLMs) from 17 patients. Graph shows the % HGP (replacement, desmoplastic, pushing) scored in each case. The cases are grouped by intrinsic subtype of breast cancer. Lum A, luminal A. Lum B (HER2^-), luminal B HER2 negative. Lum B (HER2⁺), luminal B HER2 positive. TN, triple negative. b–g. Morphology of the replacement growth pattern of BCLMs.Diagram of the tumor-liver interface in the replacement HGP (b). H&E-stained human BCLM sample illustrating the tumor-liver interface (c). Co-staining for hepatocyte specific antigen (HSA) to label hepatocytes and pan-cytokeratin (CK) to label cancer cells confirms that breast cancer cells infiltrate the liver parenchyma and replace hepatocytes in BCLM (d). Co-staining for alpha smooth muscle actin (αSMA) to label fibroblasts and CK to label cancer cells confirms the absence of a desmoplastic stroma at the tumor-liver interface in BCLM (e). Co-staining for collagen-3 (col-3) to label sinusoidal vessels and CK to label cancer cells shows that the vascular architecture of the adjacent liver is preserved at the tumor-liver interface in BCLM (f). Co-staining for CD31 to label blood vessels and cytokeratin 19 (CK19) to label cancer cells confirms the infiltrative pattern of tumor growth that facilitates vessel co-option in BCLM (g). Asterisk, cancer cells; Lv, normal liver. Scale bars, 50 μM.

Figure 6. Inhibition of vessel co-option and angiogenesis is more effective than targeting angiogenesis alone

a,b. Areas of replacement (a) and desmoplastic (b) HGP are shown in a preclinical (HT29 cell line) orthotopic model of advanced liver metastasis. Staining shown is for: H&E, CK and HSA, CK and col-3, CK and αSMA or cytokeratin 20 (CK20) and CD31, as indicated. c,d. Characterization of parental HT29 cells (parent) and HT29 cells transduced with control non-targeting shRNA (control shRNA) or ARPC3-targeting shRNAs (shARPC3-1, shARPC3-2 or shARPC3-3). In c, ARPC3 expression was determined by western blotting (see also Supplementary Data Set 1). Graph shows ARPC3 expression relative to parental HT29 cells ± SEM (n = 3 independent western blots). In d, cell motility was measured by time-lapse microscopy. Graph shows cell velocity (μm per minute) relative to parental HT29 cells ± SEM (n = 30 tracked cells per group pooled from 2 independent experiments). e. Quantification of the HGPs in control- and ARPC3-knockdown tumors. Graph shows the % replacement (R), % desmoplastic (D) and % pushing (P) HGP per group ± SEM (n = 6 mice per group). f–h. Tumors with normal ARPC3 levels (control shRNA) or ARPC3 knockdown (shARPC3-3) were established in the livers of mice, followed by treatment with B20-4.1.1 plus capecitabine (BC) or vehicle alone (Vh) for two weeks followed by histopathological analysis. Graph in f shows the % HGP per group ± SEM (n = 8 mice per group). Graph in g shows liver tumor burden expressed in terms of lesion area ± SEM (n = 8 mice per group). Graph in h shows tumor vessel density in terms of vessels per mm² ± SEM (n = 8 mice per group). For statistical analysis, Student’s t-test (panels c,g,h) or Mann Whitney U-test (panels d,e,f) were used. *P<0.05, **P<0.01 ***P<0.001, ****P<0.0001. n.s., no significant difference. Asterisk, cancer cells; DS, desmoplastic stroma; Lv, normal liver. Scale bars, 50 μM.