NRP1orNRP2gene invalidation resulted in inhibition of cell proliferation and migration
According to the papers of Cao Y et al [9, 10], neither NRP1 or NRP2 knock-down impacted cell proliferation and migration. These results were surprising since they described that NRP1 knock-down decreased the AKT activity, a major pathway involved in cell proliferation/survival. Moreover, the NRPs-mediated signaling pathways were associated with cell proliferation and migration in several cancers . We first tried to confirm the results of Cao Y et al particularly on cell proliferation and migration by using the same shRNA. The knock-down levels that we obtained were comparable to those described by Cao Y et al (Figure S1A-B). Interestingly, NRP1 knock-down decreased NRP2 expression, an observation that was not described in the Cao Y et al paper. This result suggests a crosstalk between the NRP1 and NRP2 signaling pathways (Figure S1A-B). Whereas Cao Y et al did not detect modifications of cell proliferation and migration at 24 hours, we observed a small but significant inhibition of cell proliferation for the shNRP1 cells at 72 hours and a surprising increased cell proliferation for the shNRP2 cells (Figure S1C). The migration velocity was also significantly inhibited for shNRP1 and shNRP2 cells (Figure S1D). The privileged NRP1 ligand VEGFA and the privileged NRP2 ligand VEGFC were not affected by the knockdowns (Figure S1E). These differences with the results of Cao Y et al incited us to decipher the role of NRP1 and NRP2 by knocking-out (KO) their genes by the CRISPR/Cas9 method in human (786-O) and mouse (RENCA) ccRCC cells. Two independent KO clones for NRP1 and NRP2 genes were obtained for 786-O (Fig. 1A-B), and one KO clone for NRP1 and NRP2 genes for RENCA (Figure S2A) cells. Specific NRP1 and NRP2 mRNA levels were very low and protein levels were almost undetectable in the KO clones (Fig. 1C-D and Figure S2B). However, NRP1 KO tends to increase NRP2 levels whereas NRP2 KO tends to decreaseNRP1 levels in 786-O cells at the mRNA and protein levels (Fig. 1D). Although the trend in decreased NRP1 levels were consistently observed in NRP2 KO RENCA cells, the NRP1 KO resulted in decreased expression of NRP2 in RENCA cells (Figure S2B). We obtained opposite results between KO and down-regulation of NRP1 by shRNA for NRP2 expression in 786-O cells (NRP1-directed shRNA decreased NRP2 levels whereas the KO induced NRP2 expression). All the other results, recapitulated in Tables S2, were consistent in 786-O and RENCA cells with down-regulation of KO. They suggest a fine-tuning crosstalk between the NRP1 and the NRP2 signaling, which mediates an equilibrated expression of each protein compatible with cell proliferation/survival. In 786-O cells, NRP1 KO moderately impacted cell proliferation while NRP2 KO decreased cell proliferation more importantly (Fig. 1E). These results were consistent for NRP1 down-regulation and KO whereas down-regulation of NRP2 stimulated and KO inhibited cell proliferation (Fig. 1E and Figure S1D). A moderate but still non-significant inhibition of cell proliferation was also observed in RENCA NRP1 KO cells. However, NRP2 KO resulted in inhibition of RENCA cell proliferation (Figure S2C). Except in one NRP1 KO clone, NRPs KO decreased the migration velocity of 786-O cells which is consistent with the results obtained by down-regulating NRP1 and NRP2 (Fig. 1F and Figure S1C). Since the NRPs signaling depends on stimulation by their ligands VEGFA and VEGFC, we tested their expressions in the KO cells. In the 786-O model, NRP1 KO resulted in increased expression of VEGFA without consistently affecting VEGFC expression and NRP2 KO induced VEGFC expression with inconsistent effects on VEGFA (Fig. 1G). Expression of VEGFA was consistently decreased in NRP1 and NRP2 RENCA KO cells and VEGFC was down-regulated only in NRP2 KO cells (Figure S2D). Table S3 recapitulates VEGFA and VEGFC in the different model cells. These results suggest the maintenance of a steady state levels of autocrine loops involving the respective NRP1/VEGFA and NRP2/VEGFC signaling pathways. This maintenance varies from a model to another and is compensated, at least in the human model, by increased expression of the ligand.
NRPs KO in tumor cells inhibited experimental RCC growth in immunocompetent and immunodeficient mice
Considering the relevance of immune checkpoint inhibitors in the treatment of mccRCC (Calvo E. et al. Oncologist. 2019), we deciphered the specific role of NRPs expressed by tumor cells on the growth of experimental tumors in immunocompetent and immunodeficient mice. For that purpose, we compared the growth of experimental tumors generated with 786-O KO cells in nude mice and generated by RENCA KO cells grafted in nude and syngenic BalbC mice. The NRP1 and NRP2 786-O KO clones generated smaller tumors in nude mice as compared to the controls (Figure S3). This result is consistent with the NRPs-dependent proliferation in-vitro. Invalidation of NRP1 or NRP2 in RENCA cells delayed tumor development as compared to the control group in nude mice (Fig. 2A). However, injection of the same cells in immunocompetent mice did not generate tumors (Fig. 2B). These results strongly suggest that expression of NRPs on tumor cells inhibits the anti-tumor immune system.
The NRP inhibitor NRPa-308 inhibits ccRCC cell proliferation more efficiently than sunitinib and EG00229
The strong impact on tumor growth mediated by invalidation of NRPs encouraged us to test the relevance of the NRPs’ pharmacological inhibitor NRPa-308 on different parameters characterizing ccRCC cells (786-O and A498) aggressiveness in comparison with its effect on normal dermal fibroblasts (HDF). The NRPa-308 effects were compared to those of the reference treatment for ccRCC, sunitinib and to the commercially available NRPs inhibitor EG00229. EG00229 was not very efficient in inhibiting the proliferation of ccRCC cell lines (5% and 30% inhibition respectively for 786-O and A498 cells at the highest dose 2 µM, Fig. 3A-B). Sunitinib inhibited ccRCC cell proliferation more efficiently especially in 786-O cells as compared to A498 cells (40% versus 30% for the highest dose). IC50 values were lower for NRPa-308 as compared to sunitinib demonstrating its higher efficiency on cell proliferation. Moreover, the IC50 of NRPa-308 on normal cells (HDF) was superior to those on tumor cells (Fig. 3C). We then calculated the selectivity index (SI) to evaluate the selectivity to the targets. The normal cells (HDF) served as the reference values. The SI was inferior to 1 which indicates that NRPa-308 is more efficient on tumor cells and that its general toxicity is low. The SI of NRPa-308 was lower than the SI of sunitinib suggesting a more constant anti-proliferative and less toxic effects of NRPa-308 (Fig. 3D). Hence, NRPa-308 is more efficient and has a better anti-proliferative effect than the current reference treatment, sunitinib.
NRPa-308 exerts a wide range anti-proliferative effect on primary ccRCC cells
Nowadays, the major problem in the clinic are resistance to the current treatments especially with sunitinib . We previously generated sunitinib-resistant 786-O cells (786R) by chronic exposure to the drug . NRPa-308 was ineffective on these cells as compared to the parental cells with an IC50 > 2 µM. 786R cells presented a four-fold and a nine-fold reduction in the expression on NRP1 and NRP2 mRNA levels (Fig. 4A-B). These results are compatible with a strong dependence on cell proliferation/survival mediated by the VEGFA/NRP1 and VEGFC/NRP2 autocrine loops. We previously described primary RCC cells obtained from surgically operated tumor specimens and normal epithelial cells from the same donor . Tumor cells presented a wide range of sensitivity to NRPa-308 (from 40–0% inhibition) as compared to normal primary kidney cells and to 786-O cells (Fig. 4C). These results suggest that NRPs expression should be determined before the utilization of NRPa-308 in the clinic. Indeed, a difference in NRP1 and/or NRP2 expression seems to influence NRPa-308 efficacy (Fig. 4D).
NRPa-308-dependent inhibition of cell proliferation relied mainly on NRP2 in 786-O cells
NRP1 and NRP2 KO cells constitute ideal tools to test the specificity and the NRPs- dependent effect on cell proliferation of NRPa-308. NRPa-308 was designed to inhibit VEGFA binding on NRP1. VEGFA and VEGFC can interact with NRP1 and NRP2 and as described above VEGFA/NRP1 and VEGFC/NRP2 stimulates autocrine loops in ccRCC cells. Hence, we determined the specific anti-proliferative effects of NRPa-308 in NRPs KO clones. After 48 h of treatment, the IC50 of NRPa-308 were increased in each clone showing that it exerts its anti-proliferative effects via NRP1 and NRP2. However, the increase in IC50 was superior for the NRP2 KO clones. These results strongly suggest that NRPa-308 exerts its anti-proliferative effects mainly via NRP2 (Fig. 5A).
Affinity tests have been carried out to determine the efficacy of NRPa-308 to inhibit the VEGFA binding on NRP1 and VEGFC binding on NRP2. NRPa-308 inhibited VEGFA/NRP1 binding in a dose-dependent manner but inhibited VEGFC/NRP2 binding in a reverse dose-dependent manner. Hence, low doses of NRPa-308 are sufficient to prevent VEGFC binding to NRP2 which also suggests a stronger affinity for NRP2 as compared to NRP1 (Fig. 5B).
NRPa-308 binding mode is different between NRP1 and NRP2
To understand the selectivity mechanisms of NRPa-308 towards NRP1 and NRP2, we conducted a docking study. The NRPa-308 predicted binding mode completely differs between NRP1 and NRP2 (Fig. 6A). The orientation of NRPa-308 into NRP1 binding site is flipped relatively to those obtained into the NRP2 binding site. In both cases, NRPa-308 is stabilized in the binding site through hydrogen bonds, π-stacking and hydrophobic interactions (Fig. 6B), but most of the interacting residues are distinct. Few residues involved in these interactions are conserved in NRP1 and NRP2 (W301/304, S346/349, E348/351, Y353/356 according to the NRP1/NRP2 numeration) but they establish interactions with different part of NRPa-308. Comparison of NRP1 and NRP2 structures revealed that the residues forming each binding site differ and consequently the NRP2 binding site is larger and more open than the NRP1 binding site. This result explains the docking study obtained with NRPa-308 but also the difference of affinity experimentally obtained (Fig. 5B).
NRPa-308 inhibited 786-O cell migration velocity more efficiently than sunitinib
As above-described, NRPs down -regulation and KO resulted in the inhibition of cell migration. Therefore, the ability of NRPa-308 to inhibit this parameter of tumor cell aggressiveness was tested. NRPa-308 reduced 786-O cell migration velocity more efficiently than sunitinib at a very low concentration (0.02 µM compared to 2 µM for sunitinib, Fig. 7A-B). This result suggests an anti-metastatic effect of NRPa-308.
High NRPa-308 concentration stimulated the production of NRPs ligands and of pro-angiogenic/pro-inflammatory cytokines
We observed that NRPs KO resulted in increased production of their ligands (Fig. 1). Although these ligands cannot influence tumor cells KO for NRPs, their paracrine effects can be highly detrimental by stimulating angio/lymphangiogenesis and by inducing immunotolerance. Hence, we evaluated the minimal NRPa-308 concentration that inhibits cell proliferation without influencing their secretome. As described in Fig. 3, the maximal antiproliferation effect of NRPa-308 was obtained at 0.2 µM. Increasing further the concentration did not result in a better efficacy of the drug. Hence, we analyzed the expression of VEGFA and VEGFC following NRPa-308 treatment. We also determined the expression of the ELR + CXCL cytokines CXCL5 and CXCL8 since they are involved in resistance to bevacizumab and sunitinib in ccRCC as we previously described [16, 25]. Sunitinib, at these low concentrations (below the IC50 ), had no influence on VEGFA and VEGFC expression but increased CXCL5 and CXCL8 expression as previously shown (Fig. 8A-D) . NRPa-308 increased the expression of these different factors at the highest concentration of 2 µM (Fig. 8A-D). The lowest concentration (0.2 µM), only stimulated the expression of VEGFC. These results strongly suggest that the best ratio (beneficial/detrimental effects) can be obtained at low doses of NRPa-308 in the context of ccRCC.
NRPa-308 decreases experimental ccRCC growth in a reverse dose-dependent manner
In previous studies , NRPa-308 inhibited the growth of experimental breast cancers at an optimal dose of 50 mg/kg. A pilot experiment using this concentration, on the growth of experimental ccRCC in nude mice using 786-O cells was unsuccessful (no inhibition of tumor growth). Considering the possible detrimental paracrine effects induced by high concentrations of NRPa-308, increasing concentrations (5 µg/kg, 500 µg/kg and 50 mg/kg) were used to inhibit the growth of experimental ccRCC in immunodeficient and immunocompetent mice. Considering a full distribution in the blood and a 1.5 ml of blood in a mouse of 25 grams, the respective blood concentrations of the drug administered at 5 µg/kg, 500 µg/kg and 50 mg/kg should be around 0.2, 20 and 2000 µmol/L. Of course, these blood concentrations are a rough estimation that does not consider the biological distribution in. The lowest concentration is in the range of concentration inhibiting cell proliferation and migration without affecting the production of pro-angio/lymphangiogenic and pro-inflammatory cytokines. Again, no effect of the highest NRPa-308 dose was observed. However, tumor growth was inhibited significantly by the lower amounts, especially the lowest concentration of 5 µg/kg in both mouse models (Fig. 9A-B). Immunostaining were carried out on tumors generated in immunodeficient mice. The number of blood vessels (CD31 labelling) and of pericytes (αSMA labelling) per cm2 was high in the control group and increased in a dose-dependent manner (Fig. 9C-D). However, the number of functional blood vessels (CD31/αSMA co-labelling) decreased with NRPa-308 concentrations (Fig. 9E-F). These experiments showed that NRPa-308 represents an interesting therapeutic strategy for ccRCC at a low concentration which is a good compromise associating efficacy and low toxicity (no modification of mouse weight at low doses) (Figure S4).
Efficient NRPa-308 dose decreased the expression of pro-tumoral factors
To understand the better efficacy of the lowest dose of NRPa-308, we evaluated the expression of genes involved in tumor aggressiveness especially those regulating proliferation, angio/lymphangiogenesis, epithelial/mesenchymal transition (EMT) and immune tolerance. The modifications to the mRNA analyzed by qPCR were compiled in Fig. 10. Genes associated with lymphangiogenesis were the most downregulated by the lowest dose of the drug including human NRP2, Prox1 and VEGFC and murine Prox1 and VEGFC in the immunodeficient model (Fig. 10A) and NRP2, Prox1 and VEGFC in the immunocompetent model (Fig. 10B). Only murine Prox1 and VEGFC were downregulated by the intermediate dose in the immunodeficient model (Fig. 10A). Human NRP2, Prox1, and VEGFC and murine NRP2 were upregulated in the presence of the highest dose in the immunodeficient model (Fig. 10A). In the immunodeficient model, proangiogenic genes including human NRP1, VEGFA and murine NRP1, VEGFA, VEGFR1 and VEGFR2 were downregulated by the highest dose (Fig. 10A). Some of them were also downregulated by the lowest or intermediate dose including human VEGFA and VEGFR1 and murine NRP1, VEGFR1 and VEGFR2 (Fig. 10A). Human NRP1 and VEGFR1, and murine VEGFA and VEGFR2 were upregulated by using the lowest or the highest dose (Fig. 10A). In the immunocompetent model, the proangiogenic genes NRP1 and VEGFR1 were downregulated for the two doses (Fig. 10B). The murine gene involved in immunotolerance, PDL1 was downregulated by the lowest and intermediate dose and was unchanged for the highest dose in immunodeficient mice (Fig. 10A). It was downregulated by the two doses in immunocompetent mice. In the immunodeficient model, genes involved in EMT including human MET and HGF and murine MET and HGF were downregulated by the lowest or intermediate dose (Fig. 10A). Only murine MET was downregulated by the highest dose and human MET and HGF and murine HGF were upregulated by the intermediate or highest dose (Fig. 10A). In the immunocompetent model MET and HGF were downregulated by the two doses (Fig. 10B). In the immunodeficient model, mCD69, a marker of the lymphocytes’ activation, is upregulated, which is synonymous of an activation of the immune response (Fig. 10A). The M2 macrophages marker mARG1 was decreased, which reflects a beneficial polarization of macrophages (Fig. 10A). According to these differences, we attempted to establish a score of good or bad prognosis depending on the up or downregulation of genes involved in tumor aggressiveness. We gave a score of 2 when a gene of poor prognosis decreased and a score of -2 when it increased and vice versa for a gene of good prognosis. The global score for the lowest concentration was respectively 18 and 12 in the immunodeficient (Fig. 10A) and immunocompetent models (Fig. 10B). It was of 20 for the intermediate dose in the immunodeficient model (Fig. 10A) and of 2 and 6 respectively for the highest dose in the immunodeficient (Fig. 10A) and the immunocompetent models (Fig. 10B). This evaluation, in addition to the reduction of tumor growth consistently favored the notion that a low dose of NRPa-308 had the best therapeutic efficacy.
The NRP2 associated pathways is more determinant for the aggressiveness of mccRCC
We were puzzled by confronting our previous results on breast cancers with those enclosed in this manuscript; high dose of NRPa-308 was the most efficient on models of breast cancers whereas it has no effect in models of ccRCC . Considering the striking therapeutic value of targeting NRPs for both models of cancers, we first analyzed the expression of NRPs and their ligands VEGFA and VEGFC on a panel of cell lines available in the TCGA data base. In most of the cell lines representative of aggressive ccRCC and breast cancers, VEGFA and NRP1 are expressed at high levels especially in the cell lines used in our respective experimental tumor growth (786-O and MDAMB231) (Fig. 11A-C). VEGFC and NRP2 are expressed by all the ccRCC cell lines but VEGFC levels are very low in three out of five breast cancer cell lines and NRP2 levels are very low in all the breast cell lines including MDA-MB-231 (Fig. 11B-D). These very low levels in MDAMB231 and the more specific effects of NRPa-308 on NRP2, partly explained the results obtained on experimental tumor growth. Our next step was to deep insight into the prognostic role of the NRP1 and NRP2 and their known partners VEGFA, VEGFR1, VEGFR2, Semaphorin 3A (Sema3A) and plexin A1 (PLXNA1) (all NRP1 partners) and VEGFC, VEGFR3, Semaphorin 3F (Sema3F), plexin A2 (PLXNA1) and Prospero homeobox protein 1 (Prox1), a master transcription factor of lymphangiogenesis (all NRP2 partners). For that purpose, we correlated the expression of these different partners to disease free survival (DFS, non-metastatic patients M0), progression free survival (PFS, metastatic patients M1) and overall survival (OS) in patients with ccRCC and in patients with the most severe triple negative breast cancers (TNBC). For each gene, we defined the best cut off that determines a survival difference. 425 samples were from M0 and 103 from M1 ccRCC patients. 115 samples were from TNBC patients. For M0 ccRCC patients, expression of VEGFR2, NRP2, VEGFC, VEGFR3, PLXA2 above their respective best cut off was of good prognosis for DFS (trend (T, p between 0.08 and 0.06) for NRP2, VEGFC) and significant (S) for VEGFR2, VEGFR3, PLXA2). Expression above the best cut off for Sema3A and Prox1 was of poor prognosis for DFS. For M0 patients OS, NRP1, VEGFR1, VEGFR2, NRP2, VEGFC, VEGFR3, Sema3F and PLXA2 (T for NRP1, VEGFC, VEGFR3 S for VEGFR1, VEGFR2, NRP2 and Sema3F). We performed the same analysis for the PFS and OS of M1 patients PFS. Only Sema3F (T) was associated with a longer DFS whereas eight parameters were correlated to a worse prognosis (NRP1 (S), VEGFR2 (S), Sema3A (S), NRP2 (S), VEGFC (S), VEGFR (S)3, PLXA2 (T) and Prox1 (S)). NRP1 (S), VEGFR1 (S), VEGFR2 (S), Sema3F (S) and PLXA2 (S) were correlated with a longer OS while Sema3A (S), PLXA1 (S), NRP2 (S), VEGFC (S) and Prox1 (S) were correlated to a shorter one.
For TNBC, VEGFA was curiously associated with a longer PFS (T) but NRP1 (S), VEGFR1 (S), PLXNA1 (S), NRP2 (T), VEGFC (T), Sema3F (S) and PLXNA2 (T) were correlated with a shorter one. NRP1 (S), VEGFR1 (S), VEGFR2 (T), PLXNA1 (S), VEGFC (T), Sema3F (S) and PLXNA2 (T) were correlated with a shorter OS. NRP2 was not correlated to survival in that case. We defined a score by attributing a relative weight of -2 for a gene associated with a significant poor prognosis and a relative weight of -1 for a trend. Inversely, a relative weight of 2 was given for a gene associated with a significant good prognosis and 1 for a trend. NRP1 and NRP2 pathways were considered separately. For the NRP1 pathway, a -1 score was obtained for the DFS and OS of M0 ccRCC patients, -6 and 2 scores for PFS and OS of M1 ccRCC patients, and − 5 and − 7 scores for the PFS and OS of TNBC patients. For the NRP2 pathway, positive score of 4 and 6 were obtained for the DFS and OS of ccRCC patients, -8 and − 2 scores for M1 ccRCC patients and − 5 and − 7 scores for the PFS and OS of TNBC patients.
These results showed that NRP1 and NRP2 signaling pathways, in general, strongly correlate with shorter survival for the most aggressive cancers M1 ccRCC and TNBC. However, NRP2 is correlated with shorter DFS and OS in ccRCC while NRP1 is more involved in TNBC patients’ survival. Hence, NRP1 targeting is more adapted for TNBC while NRP2 targeting is more adapted for ccRCC.