Cleaved Caspase-3 Transcriptionally Regulates Angiogenesis-Promoting Chemotherapy Resistance

Apoptosis is a well-known physiologic process that enables the maintenance of homeostasis in multicellular organisms. Apoptosis can be induced by different stimuli such as cytotoxic drugs, irradiation, and hypoxic stress. One of the best-known markers of apoptosis is the proteolytic cleavage of pro-caspase-3 into its active form, caspase-3 (Casp-3). Once cleaved, Casp-3 translocates into the nucleus, where it cleaves specific substrates (1). Depending on the context, activation of Casp-3 is considered a deleterious event in different pathophysiologic processes, such as neurodegenerative disorders or ischemia, or as a positive marker of effective therapeutic strategy such as in cancer. However, in recent years, new investigations have identified new nonapoptotic roles of caspases. Indeed, in cancer, activation of Casp-3 was described as a stimulator of tumor cell repopulation during radiotherapy treatment (2), or a driver of tumor cell proliferation in pancreatic cancer (3) and melanoma (4). Recently, Casp-3 was described as a mediator of neoangiogenesis in dying cells after X-ray therapy (5). Nevertheless, all nonapoptotic roles of Casp-3 rely on its proteolytic activity (6–9). In this article, we showed that Casp-3 activation induced transcriptional changes, leading to the upregulation of genes involved in angiogenesis and downregulation of genes involved in proapoptotic pathways. We have shown that Casp-3 was able to directly interact with the promoter of various proangiogenic genes, such as VEGFA gene, and induce transcription thanks to a specific DNA-binding domain. In vivo, the pharmacologic targeting of Casp-3 increased chemotherapy cytotoxicity in cancer cells.

All cell lines were cultured at 37°C under 5% CO_2. MCF7, MDA-MB-231, 4T1, and CT26 cells were obtained from ATCC. COLO-205 and HT29 cells sensitive and resistant to oxaliplatin were obtained from Pr. Olivier Adotevi and Pr Christophe Borg's lab (Inserm U1098, Besancon, France). MCF7, MCF7-Casp-3 (stably transfected), and MDA-MB-231 cells were cultured in DMEM (high glucose). 4T1, CT26, COLO-205, and HT-29 sensitive and resistant to oxaliplatin were cultured in RPMI1640. Mediums were supplemented with 10% (vol/vol) FCS, 1% penicillin, streptomycin, and amphotericin B, 4 mmol/L HEPES. Colo-205 and HT-29 resistant to oxaliplatin were cultured with oxaliplatin at 3 μg/mL. Using concerning culture conditions of our MCF7 cells, experiments were performed simultaneously with MCF7 Casp-3^−/− and Casp-3^+/+ within five cell subclonings. After five passages, new vials of both cells were thawed and new experiments were carried out. All cell lines were sequenced at the exome level and the transcriptomic level for authentication.

Total RNA was extracted with TRIzol reagent from MCF7 and Casp-3–transfected MCF7 cells treated or not with docetaxel. Libraries and analysis were performed as described previously (10). Chromatin immunoprecipitation (ChIP) sequencing was performed as described previously (11) for MCF7 or MCF7 Casp-3^+/+ cells treated or not with docetaxel. Next-generation sequencing data are available on the GEO website with the SuperSeries reference: GSE87227.

The study on patient samples was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the Centre Georges-François Leclerc (Dijon, France), the Comité Consultatif de Protection des Personnes en Recherche Biomédicale de Bourgogne. Written informed consent was obtained from all patients before enrollment.

All the mice were maintained in specific pathogen-free conditions and all experiments followed the guidelines of the Federation of European Animal Science Associations. All animal experiments were approved by the Ethics Committee of Université de Bourgogne (Dijon, France).

MDA-MB-231 cells were cultured for 24 hours with control inhibitor as control or with the Casp-3 inhibitor, Z-DEVD-FMK. Then, 5 × 10⁶ MDA-MB-231 cells in serum-free culture medium were inoculated subcutaneously into the right side of 8-week-old nude (nu/nu) BALB/c mice. As soon as tumors reached 400 mm², the mice were sacrificed and tumors were collected, included with optimal cutting temperature (OCT), and conserved at −80°C. Frozen sections were marked with CD31 antibody and DAPI.

BALB/c mice were subcutaneously injected with CT-26 or 4T1 cells and treated with 5-fluorouracil (50 mg/kg) + irinotecan (40 mg/kg) or docetaxel (10 mg/kg), respectively, 8 days after the tumor graft. Z-FA-FMK control inhibitor, Z-DEVD-FMK Casp-3 inhibitor, or anti-mouse VEGFA was injected intraperitoneally. siRNA transfections were performed twice a week as described previously (12) with anti-mouse pro-Casp-3 (GGATAGTGTTTCTAAGGAA), anti-mouse pro-caspase-6 (CCTGGTACATTCAGGATTT), and anti-mouse pro-caspase-7 (CCTGTTAGAGAAACCCAAA) siRNA.

Twelve-week-old TG3 transgenic mice (13) had been intraperitoneally treated with Z-DEVD-FMK or Z-FA-FMK as a negative control for 4 months. Mice were sacrificed and tail tumor nodes were counted.

GraphPad Prism software was used for the statistical analysis. Statistical significance is shown as *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.

It is known that an increase of standardized uptake value maximum (SUV_max corresponding to the ratio between tracer binding in the tumor on total injected radioactive activity normalized by the patient body weight) after one course of neoadjuvant docetaxel is a marker of nonresponse to treatment (14, 15). To understand the mechanism involved in the increase of SUV_max after docetaxel treatment, and consequently resistance to treatment, tumors generated with the aggressive MDA-MB-231 cells were treated or not with docetaxel. Two weeks later, FDG uptake was measured by PET scanner. First, we observed that docetaxel treatment induced FDG uptake compared with nontreated cells (Fig. 1A). As an increase of FDG uptake is correlated with tumor progression, we hypothesized that an inhibition of apoptosis by the pharmacologic Casp-3 inhibitor Z-DEVD-FMK would increase FDG uptake by tumors. Tumors treated by Casp-3 inhibitor alone presented a reduced size and, surprisingly, the inhibition of apoptosis completely abolished FDG uptake induced by docetaxel treatment (Fig. 1B and C). In agreement with literature, the decrease of FDG uptake, induced by Casp-3 inhibition, after one cycle of docetaxel was correlated with a decrease of tumor growth (Fig. 1D). As FDG uptake was described as being correlated with tumor angiogenesis (16), we wondered whether the result observed with FDG uptake was linked to tumor blood vessels. For that, we stained mouse blood vessels in the MDA-MB-231 tumors used for PET scanner experiments. We observed that docetaxel increased blood vessel density and that Casp-3 inhibitor treatment led to an inhibition of this phenomenon (Fig. 1E and F). In addition, quantification of tumor mouse vessels by CD31 mRNA expression analysis showed that docetaxel induced the expression of this blood vessel marker and that Casp-3 pharmacologic inhibition reduced its expression (Fig. 1G). To confirm that angiogenesis was impacted by docetaxel and Casp-3 pharmacologic inhibition, we performed RNA sequencing of human MDA-MB-231 injected into nude mice and treated or not with docetaxel ± the Casp-3 inhibitor Z-DEVD-FMK. The sequences obtained were first aligned with the mouse genome. This enabled us to identify mechanisms induced by tumor cells in the host. Unsupervised clustering revealed that untreated tumors' profile was closer to that of tumors treated with the Casp-3 inhibitor Z-DEVD-FMK, or docetaxel and Z-DEVD-FMK than to that of tumors treated with docetaxel alone (Fig. 1H). ClueGO software analysis of a cluster of genes specifically induced in tumors treated with docetaxel alone revealed enrichment in genes involved in mechanisms known to promote angiogenesis (blood vessel development, morphogenesis of branching structures, and response to wounding; Fig. 1H).

We secondly aligned sequences with the human genome. By comparing treated and nontreated tumors, we observed that the treatment of MDA-MB-231 tumors with docetaxel-induced pathways involved in angiogenesis according to ClueGO analysis (Supplementary Fig. S1A). By performing additional analysis using Enrichr software (17, 18), we observed that numerous pathways inducing angiogenesis were significantly enriched in docetaxel-treated cells such as alpha adrenergic (19), oxytocin (20–22), thyrotropin (23), muscarinic receptor (24), p38 MAPK (25), 5-hydroxytryptamine (26), and platelet-derived growth factor (27) pathways (Supplementary Fig. S1B).

To address the role of Casp-3–dependent cell death in angiogenesis, the naturally Casp-3–deficient cells (MCF7) were stably transfected with an empty vector (Casp3^−/−) or with a plasmid encoding Casp-3 cDNA (Casp-3^+/+). Both cell lines were treated with docetaxel at 10 nmol/L. This dose of docetaxel induced Casp-3 (Supplementary Fig. S1C) and caspase-9 (Supplementary Fig. S1D) cleavage in Casp-3^+/+ cells with no differences observed neither in caspase-7 (Supplementary Fig. S1E) nor in caspase-8 (Supplementary Fig. S1F) cleavage nor in apoptosis (Supplementary Fig. S1G) nor in cell growth (Supplementary Fig. S1H) between both cell lines. After 24 hours of treatment, the medium containing docetaxel was removed; cells were washed and incubated for additional 24 hours with fresh docetaxel-free medium. Afterward, endothelial tube formation experiments were performed with the 24-hour conditioned drug-free cell medium. We observed that only docetaxel-treated Casp-3–proficient cells were able to induce tube formation and stabilize formed tubes (Fig. 1I). Altogether, these data suggest that Casp-3–dependent apoptosis could induce neovessel formation and its stabilization.

We previously observed that apoptosis and particularly Casp-3–dependent apoptosis correlated with increased neovessel formation and vasculature stabilization. To decipher the mechanism involved in this phenomenon, we performed massive mRNA sequencing on Casp-3–deficient and Casp-3–proficient MCF7 cells treated with 10 nmol/L docetaxel as mentioned previously. This analysis allowed the identification of a set of 522 transcripts specifically upregulated in Casp-3–deficient cells, a set of 283 transcripts overexpressed in both cell lines, and a set of 1,197 transcripts specifically upregulated in Casp-3–proficient cells (Fig. 2A; Supplementary Table S1). Each subset had an enrichment of particular pathways as detected by Enrichr software. In the subset of 522 transcripts, four pathways were significantly enriched (Supplementary Fig. S2A), whereas in the subset of 283 transcripts, eight pathways were enriched (Supplementary Fig. S2B) as for the subset of 1,197 transcripts (Supplementary Fig. S2C). Interestingly, in the third subset, appeared a pathway named “Signaling events mediated by VEGFR1 and VEGFR2.” Using ClueGO software, we confirmed that angiogenesis-related pathways were enriched in these 1,197 transcripts (Fig. 2B). As an example, the increased expression of the proangiogenic ANXA2, C1GALT1, and VEGFA genes was validated by qPCR exclusively in treated Casp-3–proficient cells (Fig. 2C). As VEGFA is a major proangiogenic gene involved in formation of neovessels and stabilization of vasculature, and as VEGFR1 and 2 signaling seemed activated only in Casp-3–proficient cells (Supplementary Fig. S2C), therefore, we decided to study its regulation by cleaved Casp-3 as an illustration of our observations. Although docetaxel cytotoxicity was identical in Casp-3–proficient and –deficient cells (Supplementary Fig. S1G and S1H), VEGFA mRNA expression increased continuously over time only in Casp-3–proficient cells (Fig. 2D), resulting in a higher VEGFA secretion in treated Casp-3–proficient cells (Fig. 2E) 72 hours after docetaxel treatment initiation (48 hours after docetaxel removal). Literature previously reported that VEGFA expression can be induced by NF-κB (28) or Casp-3–activated PKCδ (29). Using specific siRNAs, we showed that neither NF-κB (Fig. 2F) nor PKCδ (Fig. 2G) was involved in the VEGFA upregulation observed in our conditions. In contrast, extinction of Casp-3 by siRNA (Fig. 2H) or inhibition of Casp-3 activity by the pharmacologic inhibitor Z-DEVD-FMK (Supplementary Fig. S2D) prevented VEGFA mRNA increase. In 2015, it was hypothesized that AKT phosphorylation was responsible for the induction of VEGFA expression in response to Casp-3 cleavage (5). In our experimental conditions, the pharmacologic inhibition of AKT activity had no effect on VEGFA expression (Fig. 2I). Once cleaved, Casp-3 cleaves and activates the 2 downstream caspases-6 and -7. Still with siRNA strategy, it appeared that neither caspase-6 (Fig. 2J) nor caspase-7 extinction (Fig. 2K) had an impact on the induction of VEGFA mRNA expression following docetaxel treatment of Casp-3–proficient cells. Casp-3 can share substrates with its downstream caspases, especially with caspase-7. To confirm that our observation was linked only to cleaved Casp-3, we pharmacologically inhibited Casp-3 in combination with siRNA strategy against Casp-3, caspase-6, or caspase-7. These experiments indicated that the Z-DEVD-FMK Casp-3 inhibitor lost its activity only in presence of the siRNA targeting Casp-3 (Supplementary Fig. S2E). To be active, Casp-3 needs to be cleaved by using upstream caspases, such as caspases-8 and -9. siRNA strategy targeting both upstream caspases, we surprisingly observed that both caspase-8 (Supplementary Fig. S2F) and caspase-9 (Supplementary Fig. S2G) extinction induced a slight decrease of VEGFA mRNA expression compared with treated control. Nevertheless, when both upstream caspases were downregulated, we observed a similar impact to that obtained with casp-3 inhibition (Fig. 2L). Indeed, upon docetaxel treatment, Casp-3 activation is triggered by both caspase-8 and caspase-9 (30). Consequently, silencing of both caspases-8 and -9 was required to prevent Casp-3 activation. Finally, we also tested whether VEGFA mRNA expression could be influenced by chemotherapy-induced Casp-3 activation, without external targeting of Casp-3. For that, we have used the two colon cancer cell lines HT-29 and Colo-205. For each cell line, we disposed of a sensitive versus a resistant oxaliplatin clone. Treatment by oxaliplatin induced Casp-3 activation and cell death only in sensitive clones and not in resistant ones. Thus, VEGFA induction in both cell lines was exclusively observed in the sensitive clone (Fig. 2M). Taken together, these results demonstrated that only cleaved Casp-3 was involved in the induction of VEGFA and by extension in proangiogenic gene expression in response to chemotherapy.

We previously observed that activation of Casp-3 induced the expression of proangiogenic genes. As this induction seemed to be exclusively due to cleaved Casp-3 and as cleaved Casp-3 translocated into the nucleus under docetaxel treatment (Fig. 3A and B), we wondered whether cleaved Casp-3 could be able to interact with DNA. To elucidate this point, we performed ChIP sequencing (ChIP-seq) experiments by immunoprecipitating cleaved Casp-3 after docetaxel treatment in Casp-3–deficient MCF7 and Casp-3–proficient MCF7 cells. We first observed interactions between cleaved Casp-3 and DNA. These interactions were at 2% with gene promoters, 40% in intergenic regions, 30% in introns, and 1% in exons (Fig. 3C). From aligned data, we obtained three putative consensus DNA sequences of interaction between DNA and cleaved Casp-3 (Fig. 3D). Next, bioinformatics analyses revealed that cleaved Casp-3 interacted with 1,406 genes whose 117 genes had an expression induced only in the presence of Casp-3 (Fig. 3E; Supplementary Table S2). Among these 117 genes, the most enriched pathway was the “Signaling events mediated by VEGFR1 and VEGFR2” (Supplementary Fig. S3A), suggesting that cleaved Casp-3 was able to bind and activate genes involved in this pathway. Notably, ChIP-seq revealed an interaction between cleaved Casp-3 and the promoter of VEGFA (Fig. 3F), confirmed by ChIP PCR experiments (Supplementary Fig. S3B). We also observed that cleaved Casp-3 interacted with promoters of ANXA2 (Fig. 3G) and C1GALT1 (Fig. 3H), both induced in treated Casp-3–proficient cells (Fig. 2C). On the contrary, no interaction was observed with VEGFB and VEGFC promoters (Supplementary Fig. S3C) that were not induced by Casp-3 activation. As cleaved Casp-3 seemed to interact with DNA, we wondered whether this interaction was functional. Consequently, we performed luciferase assays. The transfection of the VEGFA promoter upstream from the luciferase gene was associated with luciferase signal detection only in presence of Casp-3 (Fig. 3I), whereas absence of pro-caspases-6 and -7 did not impact luciferase signal detection (Supplementary Fig. S3D). Pro-caspase-3 is structured in three main domains: a prodomain, a large subunit, and a small subunit (Supplementary Fig. S3E). To decipher the involvement of each pro-caspase-3 domain in the interaction between cleaved Casp-3 and DNA, we generated a Casp-3 protein without prodomain. This truncation induces a specific nuclear location of noncleaved Casp-3 (31). This prodomain-truncated Casp-3 did not induce any VEGFA mRNA expression after treatment into Casp-3–deficient cells, whereas the full-length Casp-3 did (Fig. 3J). This observation suggests that activation of Casp-3 by cleavage is necessary for its transcriptional activity. Similarly, we generated a Casp-3 protein lacking the small subunit. The expression of this second truncated form of Casp-3 in Casp-3–deficient cells was not associated with an increase of VEGFA mRNA (Fig. 3K). This result suggested that the interaction between cleaved Casp-3 and DNA needed the small subunit of Casp-3. Furthermore, we analyzed the catalytically inactive C163A isoform of Casp-3 that can adopt a conformation equivalent to wild-type Casp-3 but with no protease activity (32). Overexpression of this proteolytic-inactive mutant of Casp-3 induced VEGFA promoter activity as the wild-type Casp-3 (Supplementary Fig. S3F), suggesting that the catalytic activity of Casp-3 is not necessary for its transcriptional activity, contrarily to its conformation. Next, we wondered whether we could detect an amino acid sequence enabling an interaction with DNA. By using BindN tool (33), we observed the presence of a 15-amino acid sequence “AYSTAPGYYSWRNSK” with putative DNA interaction properties located in the small subunit of Casp-3 (Supplementary Fig. S3E). To test the DNA-binding ability of this 15 amino acid sequence, we generated a Casp-3 protein lacking the 15 amino acids and performed luciferase assays with the VEGFA promoter. The overexpression of the mutated Casp-3 was not able to induce luciferase activity in Casp-3–deficient cells (Fig. 3L) and completely abolished the luciferase activity induced by docetaxel (Supplementary Fig. S3G), even in the absence of caspases-3, -6, or -7 (Supplementary Fig. S3H). This observation suggested that this 15 amino acid sequence might be involved in DNA binding and consequently gene transcription induction by Casp-3 as the putative DNA-binding domain truncated Casp-3 was not able to interact with VEGFA promoter as demonstrated by ChIP experiment (Fig. 3H). Finally, as this putative DNA-binding domain is located in the small subunit of Casp-3 and as this small subunit was necessary for transcription induction, we wondered whether the small subunit alone was sufficient for transcription induction. We therefore generated a plasmid encoding only the small subunit of Casp-3 and performed a luciferase assay with the VEGFA promoter. As obtained with the DNA-binding domain truncated Casp-3, overexpression of the small subunit alone did not induce luciferase activity in Casp-3–deficient cells (Fig. 3L), and no interaction between the small subunit of Casp-3 and the VEGFA promoter was observed (Fig. 3M). In Casp-3–proficient cells, the overexpression induced the same activation of the VEGFA promoter as the control (Supplementary Fig. S3I). Nevertheless, when we associated this overexpression with the extinction of caspases-3, -6, or -7, the activation of VEGFA promoter was completely abolished in absence of Casp-3 (Supplementary Fig. S3J), suggesting that the induction observed in Supplementary Fig. S3G was due to endogenous Casp-3. These results suggested that both subunits of cleaved Casp-3, and consequently a specific 3D structure of Casp-3, were required for the transcriptional activity of cleaved Casp-3.

As described previously, we surprisingly observed that Casp-3 inhibition completely abolished the FDG uptake in tumors treated with docetaxel through angiogenesis inhibition. Moreover it also dramatically decreased tumor growth. These observations suggested that Casp-3 activation promoted aggressiveness of cancer cells by promoting neovessel formation and stabilization. To validate this observation in humans, we first studied the extent of the tumor vascular network, through the analysis of CD31 mRNA expression, before and after neoadjuvant chemotherapy in 61 locally advanced breast tumors and correlated data with relapse-free survival (RFS). We observed a significantly worse RFS in patients with increased CD31 mRNA expression after cytotoxic treatment than in patients with a decreased CD31 mRNA expression (Fig. 4A). On the same patients, we studied Casp-3 activation by IHC at the same time point. It appeared that the number of tumors with increased nuclear staining (Casp-3 cleavage) was about twice more than in tumors with decreased nuclear staining, indicating that neoadjuvant treatment induced Casp-3 activation. Moreover, 73% of tumors presenting increased CD31 expression also showed an increase of Casp-3 cleavage, and 67% of tumors with decreased CD31 expression showed a decrease of Casp-3 activation (Fig. 4B). This difference between both groups was statistically significant (P = 0.041). These findings indicated that neoadjuvant chemotherapy induced Casp-3 activation and that angiogenesis was increased in the majority of human breast tumors harboring Casp-3 activation. As the observations done in mouse were validated in humans, we wondered whether cleaved Casp-3 could be a therapeutic target. For that, we have used the mouse colon CT26 and mammary 4T1 cancer cell lines for in vivo experiments. Both mouse cell lines harbor the same behavior as human cell lines in response to Casp-3 activation by chemotherapy. Both increased VEGFA expression in response to cytotoxic treatment in the presence of cleaved Casp-3 (Supplementary Fig. S4A). In vivo, treatment of CT26 tumors with 5- fluorouracil + irinotecan did not affect tumor growth, whereas intratumor injection of anti-mouse Casp-3 siRNA alone slightly decreased tumor growth. Interestingly, the combination of siRNA with cytotoxic treatment dramatically increased the antitumor effects of 5-fluorouracil + irinotecan (Fig. 4C). In parallel, we also tested the impact of anti-mouse caspase-6 and anti-mouse caspase-7 siRNAs on tumor growth in absence and presence of cytotoxic treatment. Contrary to Casp-3 targeting, the targeting of caspase-6 and caspase-7 did not affect tumor growth (Supplementary Fig. S4B). The same observation was obtained with systemic inhibition of Casp-3 with the pharmacologic inhibitor Z-DEVD-FMK, administered intraperitoneally (Fig. 4D). Moreover, treatment of tumors with an anti-mouse VEGFA therapeutic antibody in association with chemotherapy showed the same benefit (Fig. 4E) as that observed with inhibition of Casp-3 by siRNA or the Z-DEVD-FMK inhibitor (Fig. 4C and D). Similar observations were obtained with the mammary 4T1 cancer cell model, where both inhibition of Casp-3 (Fig. 4F) and VEGFA targeting (Fig. 4G) further increased docetaxel growth inhibition.

In Fig. 1C, we observed that the pharmacologic inhibition of Casp-3 with Z-DEVD-FMK inhibitor strongly decreased tumorigenicity of highly tumorigenic MDA-MB-231 cancer cells. On the basis of the data previously described, we could hypothesize that cleaved Casp-3 promoted tumor development by inducing tumor angiogenesis through the induction of transcription of proangiogenic genes as VEGFA. To test this hypothesis, we wondered whether the pharmacologic inhibition of Casp-3, in a spontaneous melanoma mouse model, could decrease the development of tumors. The Casp-3 inhibitor Z-DEVD-FMK, or the control inhibitor Z-FA-FMK, was intraperitoneally injected once a week in 12-week-old TG3 mice. Four months later, mice were sacrificed and tail tumor nodes were counted. It appeared that mice treated with control Z-FA-FMK harbored a mean of 10 (7–12) nodes, whereas mice treated with the Casp-3 inhibitor Z-DEVD-FMK had a mean of only 3.4 (0–7) nodes (Fig. 4H), confirming that activation of Casp-3 could be tumorigenic. Finally, we wondered whether Casp-3 activation, detected by IHC in 31 locally advanced breast tumors, could be correlated with an increased VEGFA mRNA expression. We observed that tumors with a low Casp-3 activation (less than 20% of stained nuclei) harbored a significantly lower expression of VEGFA mRNA than did tumors with high Casp-3 activation (more than 20% of stained nuclei; Fig. 4I).

Using data obtained from massive RNA sequencing comparing Casp-3–deficient and Casp-3–proficient cells treated or not with docetaxel, we observed that treated Casp-3–proficient cells specifically downregulated 486 genes. The analysis of these downregulated genes indicated a significant enrichment of eight pathways (Fig. 5A). Surprisingly, among these eight pathways specifically downregulated in treated Casp-3–proficient cells, seven were involved in apoptosis induction: FAS, TRAIL, IFNγ, ceramide, TNF receptor, RAC1, and Tap63 pathways. This observation suggested that the activation of Casp-3 decreases the expression of genes inducing apoptosis, especially extrinsic apoptosis. As we previously showed that cleaved Casp-3 was able to bind DNA and promote gene transcription, we wondered whether it was also able to inhibit transcription of proapoptotic genes by direct interaction with DNA. For that, we compared the list of the 486 downregulated genes with the list of the 1,407 Casp-3–interacted genes. Among the 36 genes common between both lists (Fig. 5B), no pathways were significantly enriched (Fig. 5C), suggesting that the mechanism underlying downregulation was not due to a direct interaction of Casp-3 with DNA. We then focused on genes present in enriched pathways of Fig. 5A. Sixteen genes were represented in these downregulated pathways: ARRB1, BIRC2, DHRS3, GPX2, IQGAP1, IRF1, MAP3K1, MAP3K11, NQO1, PIK3CB, PKN1, PRKCD, RACGAP1, RIPK1, SMPD1, and ZYX (Fig. 5D). Among these 16 genes, except RACGAP1 for which Casp-3 interacted with an enhancer region of this gene, none was a target of Casp-3 DNA-binding activity (Fig. 5E), suggesting that these downregulations were probably due to an indirect mechanism. However, the analysis of these 15 genes without interaction with Casp-3 showed that their expression was mainly under the control of two transcription factors: NFE2L2 and FOXA2, suggesting that a putative indirect mechanism explaining downregulation of antiapoptotic genes in Casp-3–proficient cells might be an inhibition of NFE2L2 and FOXA2 (Fig. 5F).

One of the hallmarks of cancer is resistance to apoptosis (34), suggesting that apoptosis induction in cancer cells is an important anticancer mechanism. Since a few years, some works tend to demonstrate that caspases' activation (6–8, 35), and especially Casp-3 (2, 4, 5, 9, 36–38), could have impact independently of apoptosis that could even have inverse effects on cell death. In this work, we showed that apoptosis-triggering Casp-3 activation is responsible for transcriptional changes such as upregulation and downregulation of some genes compared with apoptosis in absence of Casp-3. Among upregulated genes, we observed that some genes involved in angiogenesis, another hallmark of cancer, were significantly represented. This overexpression of genes, such as VEGFA, the main driver of angiogenesis, was correlated with an increased angiogenesis in vivo. This observation is in agreement with two previous studies, which showed that Casp-3 activation was linked to angiogenesis induction in a hepatocyte model (39) and in a cancer model after treatment with ionizing radiations (5). Consequently, we decided to focus on VEGFA expression as an illustration of Casp-3 transcriptional activity. We showed that the induction of VEGFA was exclusively due to cleaved Casp-3 and not NF-κB (28) in our experimental conditions. Our observation could be explained by the proteolytic cleavage and therefore inhibition of NF-κB by cleaved Casp-3 (40), contrary to the angiogenesis stimulated by Casp-8 in glioblastoma that is due to NF-κB activation (41).

Our results also contradict the hypotheses that the induction of angiogenesis by Casp-3 requires the activating proteolytic cleavage of PKCδ by Casp-3 (29), or AKT activation (5, 29). Our results could be in accordance with different studies. Indeed, PKCδ was shown to induce stabilization of VEGFA translation without impacting its transcription (42). Moreover, AKT was described as a target of cleaved Casp-3, inducing an abolishment of its activity (43, 44). As once cleaved, Casp-3 translocates into the nucleus, we wondered whether the induction of proangiogenic genes could be due to an unexpected transcriptional function of cleaved Casp-3. After its proteolytic activation, Casp-3 migrates to the nucleus and directly interacts with DNA, especially in regions containing proangiogenic genes, to induce transcription. This interaction needs the 15 amino acid motif “AYSTAPGYYSWRNSK” and the formation of the tertiary structure between small and large subunits of Casp-3. Nevertheless, the proteolytic activity of Casp-3 is not necessary for its transcriptional activity, as the overexpression of the inactive C163A Casp-3 isoform was able to induce VEGFA expression to the same extent as wild-type Casp-3. We demonstrate that cleaved Casp-3 translocates into the nucleus for a function other than its canonical proteolytic activity. Moreover, we confirmed a link between Casp-3 activation and angiogenesis in breast tumors. Almost 75% of breast tumors presenting an increase in CD31 mRNA after chemotherapy also showed an increase in Casp-3 activation. Besides, increased angiogenesis could be related to a worse RFS in breast tumors. In addition to its role in the induction of angiogenesis, VEGFA is also described as an inducer of resistance in cancer cells (45). Recently, VEGFA (46) and Casp-3 (47) were associated with the development of fibrosis in pancreatic cancer, a marker of resistance to chemotherapeutic treatment (48). VEGFA has also been associated with an increased infiltration of immunosuppressive cells, myeloid-derived suppressor cell, in ovarian cancer, resulting in local immunity inhibition and a poor prognosis (49). In parallel with the direct upregulation of genes, Casp-3 was also able to induce downregulation of other genes, some involved in proapoptotic pathways. This downregulation was not due to a direct interaction with promoters but possibly through the inhibition of NFE2L2 and FOXA2. At the transcriptional level, FOXA2 was not detected in our RNA-seq analysis, whereas NFE2LE was. The inhibition of NFE2L2 could be due to its cleavage by cleaved Casp-3 as described in HeLa cells during TNFα-mediated apoptosis (50). This double effect of Casp-3 activation (upregulation of proangiogenic genes and downregulation of proapoptotic genes) could explain the observation showing that Casp-3 inhibition improved chemotherapy efficacy. Consequently Casp-3 can be a new therapeutic target. Moreover, this double effect could explain that Casp-3 targeting could prevent tumor growth in mice but also tumor development in genetically predisposed mice, suggesting that Casp-3 inhibition could be a prophylactic treatment for patients with a familial risk of cancer. The therapeutic use of Casp-3 inhibitors could be rapidly tested in clinics as a feasibility study was performed for the use of Casp-3 inhibitors in Pemphigus vulgaris (51), and a randomized clinical trial of emricasan, a pan-caspase inhibitor, was performed in subjects with nonalcoholic fatty liver disease (52).

Our data indicate that Casp-3 could be a therapeutic target to enhance the efficacy of chemotherapy. In contrast, in other medical domains such as neurodegenerative diseases or ischemic disorders of heart or brain, our data could encourage the development of Casp-3–protective treatments. Indeed, in cases of myocardial infarction, Casp-3 is rapidly cleaved (53), resulting in a putative induction of proangiogenic genes and decrease of proapoptotic genes, which could protect cardiac muscle (54, 55). In a previous study, we showed that the inhibition of caspase-6 blocked apoptosis without blocking Casp-3 activation (12). In these conditions, we could speculate that the inhibition of caspase-6 could block the deleterious impact of apoptosis on heart cells without affecting the benefit of Casp-3 activation on the production of other proangiogenic factors. A similar rationale could be applied to ischemic stroke, in which cleaved Casp-3 is essential for neuroprotection (56) and an increase in VEGFA is neuroprotective (57). Finally, VEGF stimulates neural stem cells' proliferation after cerebral ischemia and could enable the treatment of neurodegenerative disorders (58–60).

Our study highlights unexpected functions of Casp-3 in cells. Indeed, we showed that cleaved Casp-3 is able to interact with DNA and more specifically with promoters of proangiogenic genes. This interaction induces transcription of these genes and leads to neoangiogenesis (Fig. 5G). In parallel, activation of Casp-3 induced, by an indirect mechanism, downregulation of genes involved in proapoptotic pathways. These observations could explain why the activation of Casp-3 is required for tumorigenesis and is a factor of resistance to cytotoxic drugs (Fig. 5G). Finally, we identified cleaved Casp-3 as a new therapeutic target in cancer as its inhibition sensitizes cells to chemotherapy drugs and decreases the development of tumors in genetically predisposed mice.

No potential conflicts of interest were disclosed.

Conception and design: L. Apetoh, F. Végran, R. Boidot

Development of methodology: S. Chevrier, V. Dérangère, F. Végran, R. Boidot

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Bernard, S. Chevrier, F. Beltjens, M. Dosset, E. Viltard, A. Lagrange, A. Oudot, B. Collin, L. Apetoh, O. Feron, S. Chen, L. Arnould, F. Végran, R. Boidot

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Bernard, F. Beltjens, A. Lagrange, A. Oudot, B. Collin, L. Arnould, F. Végran, R. Boidot

Writing, review, and/or revision of the manuscript: L. Apetoh, O. Feron, S. Chen, L. Arnould, F. Végran, R. Boidot

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Ghiringhelli, R. Boidot

Study supervision: R. Boidot

We thank Pr. Christophe Borg and Pr. Olivier Adotevi for supplying oxaliplatin-sensitive and -resistant clones of Colo-205 and HT-29 cell lines. We thank Dr. Alexia Cotte for her critical read of the manuscript and Isabel Gregoire for manuscript editing. We also thank “Ligue contre le Cancer de Côte d'Or” and “Ligue contre le Cancer de Saône-et-Loire” for the financial support of this work. We thank “Ligue contre le Cancer” for financially supporting A. Bernard. Imaging experiments were supported by a French Government Grant managed by the French National Research Agency (ANR) under the program “Investissements d'Avenir” (reference ANR-10-EQPX-05-01/IMAPPI Equipex).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.