Overcoming adaptive resistance to anti-VEGF therapy by targeting CD5L

Abstract

Antiangiogenic treatment targeting the vascular endothelial growth factor (VEGF) pathway is a powerful tool to combat tumor growth and progression; however, drug resistance frequently emerges. We identify CD5L (CD5 antigen-like precursor) as an important gene upregulated in response to antiangiogenic therapy leading to the emergence of adaptive resistance. By using both an RNA-aptamer and a monoclonal antibody targeting CD5L, we are able to abate the pro-angiogenic effects of CD5L overexpression in both in vitro and in vivo settings. In addition, we find that increased expression of vascular CD5L in cancer patients is associated with bevacizumab resistance and worse overall survival. These findings implicate CD5L as an important factor in adaptive resistance to antiangiogenic therapy and suggest that modalities to target CD5L have potentially important clinical utility.

Fig. 1. Upregulation of CD5L in anti-VEGF therapy–resistant endothelial cells promotes angiogenesis properties.

A Time point at which SKOV3ip1 ovarian cancer tumors were isolated during the course of B20 treatment. Tumor progression was identified by an increase in bioluminescence (Data represented as mean ± SD; n = 5 mice for control IgG and n = 10 mice for B20 antibody treatment). B Heat map from gene expression profiling of endothelial cells isolated from B20-resistant tumors compared with endothelial cells isolated from B20-sensitive tumors. The microarray data were deposited in GEO (Accession number GSE180687). C CD5L staining in endothelial cells from mouse tumors sensitive or resistant to B20 antibody (n = 4 mice; scale bar = 100 µm). D CD5L protein expression in RF24 endothelial cells containing CD5L-overexpressing plasmid versus empty vector (EV). Western blotting was performed two times as technical replicates; in each repeat, the blotting, including loading control, was performed using the same sample processing controls. E Cell proliferation in RF24 endothelial cells containing CD5L-overexpressing plasmid versus EV. F, G Tube formation (F) and cell migration (G) in RF24 endothelial cells containing CD5L-overexpressing plasmid versus EV (scale bar = 200 µm). H Concentration of CD5L in media collected from RF24 endothelial cells containing CD5L-overexpressing plasmid versus EV (n = 2 biologically independent experiments). I Levels of CD5L protein in RF24 endothelial cells treated with siCD5L versus siControl. Western blotting was performed two times as technical replicates; in each repeat, the blotting, including loading control, was performed using the same sample processing controls. J–L Cell proliferation (J), tube formation (K), and cell migration (L) in RF24 cells treated with siCD5L versus siControl; (scale bar = 200 µm for K and L). Data represented as mean ± SD, determined by two-tailed Student’s t-test; n = 3 for E, F, and J and n = 4 for G, K, and L biologically independent experiments.

Fig. 2. CD5L is upregulated through hypoxia-induced PPARG overexpression.

A, B CD5L mRNA (A) and protein expression (B) in RF24 endothelial cells containing PPARG overexpressing plasmid versus empty vector (EV). C CD5L promoter constructs activation using RF24 endothelial cells containing PPARG overexpressing plasmid versus EV. D, E PPARG and CD5L mRNA (D) and protein (E) expression in RF24 endothelial cells treated with siPPARG versus siControl. F CD5L promoter construct activation using RF24 endothelial cells treated with siPPARG versus siControl. G Luciferase expression in RF24 endothelial cells after co-transfection of PPARG overexpressing plasmid and CD5L promoter construct harboring mutated PPARG binding site. H, I PPARG and CD5L mRNA (H) and protein (I) expression in RF24 endothelial cells cultured in hypoxic or normoxic conditions. J, K PPARG and CD5L mRNA (J) and protein (K) expression in RF24 endothelial cells treated for 6 and 30 h with cobalt chloride (HIF1α stabilizer). Western blots were performed from two independent technical replicates; in each repeat, the blotting, including loading control, was performed using the same sample processing controls (B, E, I, K); L PPARG and CD5L mRNA expression in RF24 endothelial cells treated with YC-1 or topotecan under hypoxic conditions. M CD5L WT promoter construct activation in RF24 endothelial cells cultured in hypoxic and normoxic conditions. N Chromatin immunoprecipitation (ChIP) analysis of the CD5L promoter using an anti-PPARG antibody under hypoxic and normoxic conditions. Data represented as mean values ± SD, determined by two-tailed Student’s t-test for A, C, F, G, H, M; one-way ANOVA Tukey’s multiple comparisons for J and L; two-way Anova Tukey’s multiple comparisons for D and N; n = 3 for D, G, H, J left panel, M and N; n = 4 for A, C, and L left panel; n = 5 for F; n = 2 for J right panel and L right panel.

Fig. 3. Exogenous CD5L treatment of RF24 endothelial cells results in upregulation of PI3K/AKT signaling.

A AKT pathway activation was measured by pAKT/AKT in RF24 cells after an exogenous CD5L protein treatment. B–D AKT pathway activation, tube formation (C), and cell migration (D) in RF24 cells treated with CD5L protein and either LY294002 (PI3K inhibitor) or DMSO. Scale bar = 200 µm for C and D. E CD36 mRNA expression in RF24 cells treated with CD5L protein. F AKT pathway activation in RF24 cells treated with siCD36. Western blots were performed from two independent technical replicates; in each repeat, the blotting, including loading control, was performed using the same sample processing controls (A, B, F). G, H Cell viability of RF24 cells at increasing concentrations of bevacizumab with the addition of either CD5L protein (G) or siCD5L (H). Data represented as mean values ± SD, determined by two-tailed Student’s t-test except for two-way Anova Tukey’s multiple comparisons for G and H. (n = 3 for C, D, and E; n = 4 for G and H biologically independent experiments).

Fig. 4. PPARG silencing inhibits tumor growth and angiogenesis in the ID8 xenograft model.

A Photographs of representative mice from wild-type (WT) and Tie2-cre;PPARG KO mice. B, C Tumor weight (g) (B) and the number of tumor nodules (C). D, E Ki67 IHC (D) and CD31 IF (E) staining of tumors from WT versus PPARG KO mice. For statistical analysis, five randomly selected tumors per group were stained, and five random fields per tumor were scored. Scale bar = 200 µm for D and E. F Survival plot for B20, anti-VEGF antibody treatment. B20 was injected into the peritoneal cavity twice weekly at a dose of 5 mg/kg. G Expression of pAKT relative to AKT in tumor samples from WT versus Tie2-cre;PPARG KO mice (pAKT to AKT ratio determined after normalization of pAKT to AKT). Data represented as mean values ± SD, determined by two-tailed Student’s t-test; n = 3–4 mice for D, E, and G; n = 5 mice for B, C, and F.

Fig. 5. Antibodies targeting CD5L exhibit antitumor and antiangiogenic effects.

A Photographs of representative mice of control antibody and anti-CD5L antibody (H-447 and R-35) treated groups. Mice were treated intraperitoneally with either PBS or anti-CD5L antibody (10 mg/kg) starting on Day 8 after tumor injection until Day 35. B, C Tumor weight (B) and the number of tumor nodules (C). D CD31 immunofluorescence staining of tumors from control versus anti-CD5L antibody-treated groups. For statistical analysis, five randomly selected tumors per group were stained, and five random fields per tumor were scored. Scale bar = 100 µm. E, F Tube formation; scale bar = 500 µm (E) and cell migration; scale bar = 200 µm (F) of RF24 cells treated with either control antibody alone, control antibody + CD5L protein, or R-35 antibody + CD5L protein. Data represented as mean values ±  SEM determined by the Mann–Whitney test for B, C (n = 13 for control Ab; n = 7 for H-447 and R-35 antibodies respectively), ordinary one-way ANOVA Tukey’s multiple comparisons test for E and F and the two-sided Student’s t-test used for D; (n = 3 for D and E and 9 for F).

Fig. 6. S76.T (CD5L)-aptamer blocks resistance to anti-VEGF therapy.

A Mouse adaptive resistance tumor model. S76.T-aptamer was injected intravenously every 3 days starting on day 38, after 21 days of B20 treatment. B Photographs of representative mice treated with scramble aptamer, scramble aptamer + B20, CD5L aptamer (S76.T) + IgG, and S76.T + B20. C, D Tumor weight (g) and the number of tumor nodules. E, F CD31 immunohistochemical (E) and Ki67 immunohistochemical (F) staining of tumors from scramble aptamer, B20, S76.T, and combination of S76.T + B20 treated mice (scale bar = 100 µm). For statistical analysis, five randomly selected tumors per group were stained, and five random fields per tumor were scored. Data represented as mean values ± SD determined by the ordinary one-way ANOVA Tukey’s multiple comparisons test; n = 5 for both C & D; for D; n = 4 for E and F.

Fig. 7. CD5L overexpression is associated with bevacizumab resistance and worse overall survival in ovarian cancer patients.

A Representative images of CD5L expression measured by immunohistochemical (IHC) analysis in patients considered as bevacizumab responder (n = 25) or non-responder (n = 11) scale bar = 100 µm. Quantification of CD5L expression measured by IHC analysis scaled from 0 (absent) to 10 (high) in bevacizumab responders (n = 25) versus non-responders (n = 11). B CD5L serum protein levels in ovarian cancer patients that were classified as either responsive (n = 8) or non-responsive (n = 7) to bevacizumab. C Representative images of low (upper) and high (lower) CD5L protein expression in tumor endothelial cells from a human ovarian cancer patient cohort; scale bar = 100 µm. D Kaplan–Meier curve of overall survival in patients with high-grade serous ovarian cancer (HGSC) stratified according to CD5L protein expression level (n = 30 for Low and n = 24 for High) as measured by IHC analysis, from a patient cohort in panel (C). The inset on the right shows the curves from 0–80 months only. Data represented as mean values ± SD determined by a two-sided Student t-test was used for statistical calculations, aside from panel (D), which was generated with the use of the log-rank test.

Fig. 8. Mechanism of CD5L-induced AVA resistance.

Anti-VEGF treatment may initially cause tumor regression via decreased angiogenesis (tumor with low vessel density); however, adaptive resistance frequently emerges over time, leading to tumor growth and increased angiogenesis (larger tumor with high vessel density). Inset demonstrates tumor endothelial cells showing that local tumor hypoxia leads to increased CD5L secretion by overexpression of transcription factor PPARG. Secreted CD5L binds to the CD36 receptor, causing activation of the AKT pathway and ultimately leading to increased cell proliferation and angiogenesis.