Low Caspase-8 expression correlates with poor prognosis in cervical cancer patients

Cervical cancer possesses one of the highest Tumor Mutational Burden (TMB) of all cancers [60]. To determine the significance of Caspase-8 expression in the prognosis of cervical cancer patients, we splitted the patients from the CESC-TCGA database, based on their RNA-Seq data, into those with TMB lower or higher than the median. Surprisingly, patients with high non-synonymous somatic mutations and low CASP8 expression displayed significantly poorer OS (Overall Survival) and PFS (Progression-Free Survival) (Fig. 1A). In contrast, patients with TMB lower than the median, high CASP8 expression displayed significantly poorer OS and PFS (Fig. 1B).

Fig. 1

The effect of Caspase-8 expression on patient prognosis. A RNA-Seq data of cervical cancer patients with higher than median Tumor Mutational Burden (TMB) were obtained from the TCGA database (CESC-TCGA) and used to determine the correlation between high or low Caspase-8 expression with the Overall Survival (OS) and Progression-Free Survival (PFS) of 88 patients (upper panel). The tables show the mean OS and PFS of patients with low, intermediate, or high Caspase-8 expression, along with their 95% Confidence Interval (CI). B Correlation between high or low Caspase-8 expression with the OS and PFS of 88 patients with lower than median TMB based on their RNA-Seq data obtained from the CESC-TCGA database

In summary, these data highlighted that the correlation between low CASP8 expression and the prognosis of cervical cancer patients depends on their TMB status.

Knock-out of Caspase-8 alters cellular behaviors

To further explore the roles of Caspase-8 expression in the malignant behavior of cervical cancer cells that might contribute to poor patient prognosis, we used the CRISPR/Cas9 system to generate CASP8-/- HeLa and SiHa cells (Fig. 2A, Supplementary Fig. 1A). To avoid clonal variations [61], we mixed individual CASP8-/- knock-out clones—Clones K5, 7 and 8 for HeLa and Clones 7, 11 and 22 for SiHa, to form a mixed knock-out population (henceforth KO) and determined different biological traits of the resulting cells. We first determined the 2D cell-proliferation of the WT vs. KO cells over 120 h and observed that the KO HeLa cells proliferated at a significantly slower rate than the WT cells (Fig. 2B). Importantly, none of the individual knock-out clones displayed any significant differences in their cell-proliferations (Supplementary Fig. 2A), indicating that none of the individual clones significantly affected the differential proliferation of the KO population.

Fig. 2

Effects of the knock-out of Caspase-8 on the behavior of cervical cancer cell lines. A HeLa CASP8-/- knock-out clones were generated using the CRISPR/Cas9 genome editing system. Individual clones (K5, 7, 8, and 14) were lysed and subjected to a Western blot analysis using Caspase-8 and β-Actin antibodies. B The proliferation of HeLa WT and KO cells was measured using an CellTiter-Blue Cell Viability assay. Over a period of 120 h, the number of viable cells were quantified every 24 h and represented graphically [mean ± SD; n = 3 for each time point; p value (paired t test, two-tailed); * =  < 0.05]. C For cell-cycle analysis, HeLa WT and KO cells were synchronized by double-thymidine treatment for 16 h and then released for 0, 3, 6, and 9 h. Cells were harvested, permeabilized with ethanol, and treated with Propidium Iodide (PI). After RNase treatment, the DNA histograms were determined by FACS to reveal the cell-cycle distribution of the cells at each time point. Non-synchronized (NS) WT and KO cells were used as negative controls. Overlay of the histograms of PI-positive WT and KO cells at every time point has been displayed in the lower panel; D lysates of synchronized cells were analyzed by immunoblotting to check for cell-cycle markers PLK1, Cyclins B1, E1, and A1, as well as Caspase-8 and β-Actin. E For the BrdU assay, HeLa WT and KO cells were first synchronized with double-thymidine for 16 h and released in a fresh medium for 0, 3, 6, and 9 h. 10 µM of the thymidine analogue BrdU was added to the cells for 2 h (before each time point, except 0 h) to allow it to be incorporated into the newly synthesized DNA and was detected using an anti-BrdU antibody. The absorbance, measured at 370 nm, correlating with the amount of incorporated BrdU into the newly synthesized DNA, has been represented graphically. The Blank represents the fluorescence intensity of the anti-BrdU antibody, without the BrdU agent [mean ± SD; n = 3 for each time point; p value (paired t test, two-tailed); * =  < 0.05; ** =  < 0.005]. F The 2D migration of HeLa WT and KO cells was determined using ibidi migration chambers at 3 h intervals over 24 h. The reductions in the areas between the two cell populations at each time point, representing the migration of the cells, were measured, normalized to the area at 0 h, and represented graphically [mean ± SD; n = 3 for each time point; p value (paired t test, two-tailed); ** =  < 0.005]. G The 3D invasion of HeLa WT and KO cells was determined using Matrigel-coated invasion chambers was determined over a period of 24 h. The nuclei of the invaded cells were stained with DAPI (bottom panel), and their quantification was represented graphically. MCF-7 cells were used as a negative control [mean ± SD; n = 3 for each time point; p value (paired t test, two-tailed); ** =  < 0.005]

We next decided to determine whether the reduced proliferative activity of KO cells could be attributed to any alterations in their cell-cycle profiles. For this, HeLa and SiHa WT and KO cells were synchronized with a double-thymidine block, released into a fresh medium, and analyzed by FACS. We observed that the cell-cycle profile of synchronized KO cells was retarded primarily at the 6 h time point, which corresponded with the S/G2-phase of the cell-cycle, compared to that of WT cells, albeit more prominently in HeLa than in SiHa cells (Fig. 2C, Supplementary Fig. 1B). The expression of cell-cycle markers for G1/S (Cyclin E1), S/G2 (Cyclin A1) and G2/M (PLK1, Cyclin B1) confirmed this cell-cycle delay in the KO cells of both cell lines (Fig. 2D, Supplementary Fig. 1C). Furthermore, the significant increase in the BrdU signals, which relies on the incorporation of the thymidine analogue 5-bromo-2′-deoxyuridine (BrdU) into the nascently synthesized DNA strands [44], after 3 and 6 h release from double-thymidine synchronization, confirmed the cell-cycle delay observed in the HeLa KO cells (Fig. 2E).

Moreover, comparing the 2D migration and 3D invasion between the KO and WT cells demonstrated that KO cells migrated (Fig. 2F, Supplementary Fig. 1D) and invaded (Fig. 2G, Supplementary Fig. 1E) at significantly faster rates than their WT counterparts. Importantly, none of the individual knock-out clones displayed any significant differences in their cell-migrations (Supplementary Fig. 2B), indicating that none of the individual clones significantly affected the differential migration of the KO population.

Knocking-down Caspase-8 expression using siCasp8 in HeLa and SiHa WT cells (Supplementary Fig. 3A) significantly enhanced their invasiveness as compared to Empty Flag-Vector (EV) transfected or non-transfected (NT) cells (Supplementary Fig. 3B). Reversibly, over-expressing a Flag-tagged-Casp8 (Flag-Casp8) in HeLa WT and KO cells significantly reduced invasion (except WT Flag-Casp8 vs. WT Empty Flag-Vector) and migration, as compared to their respective Empty Flag-Vector (EV) counterparts (Supplementary Figs. 3B, C).

In summary, these experiments demonstrated that the loss of Caspase-8 expression in HeLa and SiHa cells reduced cell-proliferation, which could be attributed to a delay in the cell-cycle progression at the S/G2 phase. In addition, the loss of Caspase-8 expression caused a significant enhancement of the 2D-migration and 3D-invasion of both cell lines. Importantly, we observed no significant increase in apoptosis in the KO cells, even after treating them with Trail (Supplementary Fig. 3D), further validating the veracity of our Caspase-8 knock-out.

CDK9 is an interacting partner of Caspase-8

The results of the previous experiments raised a fundamental question on how Caspase-8 alters the migration and invasion of cervical cancer cells. To elucidate this, we IPed Caspase-8 from non-synchronized (NS) exponentially growing and S/G2-phase synchronized (synch.) HeLa WT. The co-IPed proteins were subjected to an interactome analysis to identify Caspase-8 interacting proteins that might be involved in cell-migration and cell-invasion.

After performing the initial analysis of the non-synchronized (NS, ~ 1000 proteins) and S/G2-phase synchronized (synch., ~ 850 proteins) data sets, we compared the ratios of their Log2 FC ≥  ± 0.5 values (up- and down-regulated) to obtain 551 common proteins between the NS and synch. data sets (Supplementary Figs. 4A, B). Finally, we selected 291 proteins for further analysis, which were present in the non-synchronized (NS) and S/G2-phase synchronized (synch.) data sets, with a Log2 FC ≥  + 0.5 (up-regulated), representing the pool of potential Caspase-8 interacting proteins (Supplementary Fig. 4C). Expectedly, Caspase-8 had one of the highest Log2 FC values in the synchronized (NS) and S/G2-phase synchronized (synch.) data sets (Supplementary Fig. 4C). The step-by-step analysis of our interactome study, to obtain the 291 proteins, has been schematically represented here (Supplementary Fig. 4D).

We next investigated the molecular networks/pathways in which these 291 proteins participate and the potential involvement of Caspase-8 in them using Ingenuity Pathway Analysis (IPA). Intriguingly, IPA analysis revealed a wide variety of molecular pathways critical for the growth and survival of tumor cells, including those involved in tumor cell-migration (Fig. 3A, highlighted in black). Furthermore, GOTERM_BP (Biological Process) analysis using the DAVID bioinformatics tool revealed several cell-migration-associated biological functions (Fig. 3B, highlighted in black).

Fig. 3

Caspase-8 interactome. 291 differentially expressed proteins with a Log2 FC ≥  + 0.5 (up-regulated), representing the pool of potentially Caspase-8 interacting proteins, were identified through the comparative analysis of HeLa WT non-synchronized (NS) and S/G2-phase synchronized (synch.) data sets. They were analyzed for their cellular functions using A the Ingenuity Pathway Analysis (IPA) bioinformatics tool. The top 20 most significantly regulated pathways are shown here, predicted by IPA. The black bars represent cancer cell-migration associated functions; and B the DAVID bioinformatics tool (www.david.ncifcrf.gov, v.6.8). Shown here are the top 15 most significant processes, predicted using the Gene Ontology Term (GOTERM) “Biological Process (BP)” function of DAVID. The black bars represent cancer cell-migration-associated processes. C The 291 proteins predicted by IPA to be associated with cell-migration were further analyzed by IPA to determine whether these proteins, including Caspase-8, were directly involved with cell-migration or indirectly, via other proteins, which may or may not be present in our data-set of 291 proteins. The black arrow highlights Caspase-8. D The “Diseases and Functions” feature of the IPA tool also predicted that the cell-migration-associated proteins were also involved in several other biological processes. Shown here are the top 10 processes regulated by these proteins and their associations with Caspase-8. E The proteins involved in both cell-migration and transcription were further analyzed using IPA to determine whether Caspase-8 could directly or indirectly interact with them. The black and red arrows highlight Caspase-8 and CDK9, respectively. The interactome analysis was performed in triplicate

We next used IPA to perform their network analysis to elucidate the correlation of Caspase-8 with proteins involved in these pathways. Interestingly, IPA predicted that Caspase-8 is not only directly involved with the cell-migration pathway (Fig. 3C, black arrow), whose constituent proteins, including Caspase-8, appear to form a tight cluster (Supplementary Fig. 4E, black arrow), but also with several other biological processes, including transcription (Fig. 3D). This data encouraged us to investigate potential correlations between cell-migration and transcription-associated proteins. Notably, IPA identified the kinase CDK9, which was involved in both the pathways, to interact with Caspase-8, albeit indirectly via the transcription factor STAT3 and the CDC37/HSP90 chaperone complex (Fig. 3E, red arrow).

In summary, the interactome analysis of Caspase-8 revealed that it could potentially interact with a wide variety of proteins, including CDK9, regulating different molecular pathways, including those involved in cell-migration. Surprisingly, in both analyses, IPA predicted Caspase-8 as a nuclear protein (Fig. 3C, E,black arrow), even though it is generally known to be a cytoplasmic protein [10, 17]. Based on these predictions and the fact that CDK9 is a predominantly nuclear protein [38], we next aimed to study the potential interaction between Caspase-8 and CDK9.

To achieve this goal, we first performed a ‘pull-down’ assay from the lysates of HeLa KO cells, using GST-Casp8-WT, -NT (N-terminal region/prodomain), -p18, and -p12 domains (making up its C-terminal region/catalytic domain) [10] (Fig. 4A) and observed that endogenous CDK9 interacted with the GST-Casp8-WT, specifically with the p12 domain (Fig. 4B). Since most cellular CDK9 is found either as part of the large negative regulatory complex—7SK snRNP, within the nucleus or with the CDC37/HSP90 chaperone complex, within the cytosol [38, 62], we next decided to identify whether the interaction between CDK9 and Caspase-8 occurred directly, or via any intermediate proteins, which could be a part of these complexes. For this purpose, we incubated a commercially available, recombinant, active P-TEFb (GST-CDK9/Cyclin K) of Baculovirus origin with recombinant GST-Casp8-WT of bacterial origin in a cell-free system and performed immunoprecipitations using specific antibodies. IP with an anti-CDK9 antibody immunoprecipitated GST-CDK9 and co-immunoprecipitated the GST-Casp8-WT, whereas the IP with an anti-Caspase-8 antibody immunoprecipitated GST-Casp8-WT and co-immunoprecipitated the GST-CDK9. This confirmed that the interaction between the two proteins occurs directly, without the involvement of any intermediate bridging proteins (Fig. 4C).

Fig. 4

Interaction between Caspase-8 and CDK9. A Domain structure of the Caspase-8 protein. NT N-Terminus, CT C-Terminus and DED Death Effector Domain. B GST-tagged-Casp8 (full-length)-WT, -NT, -p18, and -p12 fragments were used to pull-down endogenous CDK9 from the lysates of HeLa WT cells. Pull-down analyses using an uncoupled GST protein (GST) were used as the negative control. An anti-CDK9 antibody was used to detect the presence of CDK9 in an immunoblot of the pull-down assay, whereas anti-Caspase-8 and -GST antibodies were used to identify/confirm the presence of the GST-Caspase-8 (WT and fragments)/GST bait proteins in the same immunoblot. Coomassie staining was performed to determine the expressions of the different GST-tagged proteins (lowest panel). C Commercially available, recombinant GST-CDK9 of Baculovirus origin was incubated with recombinant GST-Casp8-WT of bacterial origin. CDK9 and Caspase-8 specific antibodies were used to immunoprecipitate GST-CDK9 and GST-Casp8-WT, respectively. Input and IgG controls have also been included. D GST-tagged-CDK9 (GST-CDK9) was used to pull-down endogenous Caspase-8 from lysates of HeLa and SiHa (WT, KO) cells. An immunoblot of the lysate input was probed for Caspase-8 and β-Actin (left panel), whereas the pull-down was probed for Caspase-8 and GST-CDK9 bait protein (right panel). Pull-downs from HeLa and SiHa WT lysates, using GST, were included as negative controls. E A specific antibody was used to precipitate Caspase-8 from the lysates of WT and two knock-out clones (K5, K7) of HeLa cells and probed for CDK9 and Caspase-8 in an immunoblot (IB). F A GFP-tagged-Caspase-8 (GFP-Casp8) was first expressed in HeLa WT cells and then used to perform a CDK9 IP. Immunoblot of the lysate input (left lane), CDK9 IP (center lane), and IgG control (right lane) were probed for Caspase-8 and CDK9. G A Proximity Ligation Assay (PLA) was performed by staining HeLa WT cells using anti-CDK9 or -Casp8 antibodies individually or together. The cells were then stained with fluorescent-tagged PLA probes. PLA signals (red spots) were observed under a fluorescence microscope. DAPI and Tubulin staining were used to highlight the nucleus and cytoplasm of the cells, respectively. H The cervical cancer cell lines (HeLa) and the high-grade serous ovarian cancer cell lines (OVCAR-3, OVCAR-8) were fractionated into their respective cytosol and nuclear fractions. These fractions were immunoblotted to check for Caspase-8, pCDK9, CDK9, β-Tubulin (cytosolic marker), and NUP98 (nuclear marker). I HeLa KO cells expressing either an Empty Flag-Vector (EV, negative control) or Flag-tagged-Caspase-8 (Flag-Casp8) were similarly fractionated into their cytosol and nuclear fractions. Identically fractionated non-transfected (NT) KO cells were also included. All cytosol and nuclear fractions were immunoblotted to check for Flag-Caspase-8, CDK9, CDC37 (cytosolic marker), and β-Actin (whole-cell marker). J Caspase-8 was IPed, using an anti-Caspase-8 antibody from the cytosol (C) and nuclear (N) fractions of HeLa WT siCtrl and CDK9-depleted cells (siCDK9). Lysate input, shown on the left side, and the IP on the right, were checked for Caspase-8 and CDK9, whereas the inputs were additionally checked for GAPDH (cytosolic marker) and NUP98. IgG (R) control has also been demonstrated

Reversibly, GST-CDK9 also pulled down endogenous Caspase-8 in both HeLa and SiHa WT cells, but, expectedly, not from their respective KO cells (Fig. 4D). Furthermore, endogenous CDK9 was also co-IPed in a Caspase-8 IP of HeLa WT cells, but expectedly, not using two KO clones (Fig. 4E). Reversibly, exogenous GFP-Casp8, overexpressed in HeLa WT cells, was also co-IPed with CDK9 (Fig. 4F). A Proximity Ligation Assay (PLA) further showed a tremendous increase in red fluorescent spots in HeLa WT cells stained together with anti-CDK9 and -Caspase-8 antibodies, as compared to when they were used individually (Fig. 4G), indicating an interaction between CDK9 and Caspase-8 in living cells. In summary, these experiments confirmed that CDK9 directly interacts with Caspase-8 both in vitro and in cells.

However, the question remained about the cellular compartment where this interaction occurs. This is because CDK9 is predominantly nuclear, where it undergoes activation by phosphorylation at its Thr186 residue, promoting its binding to its cyclin partners, predominantly Cyclin T1, to form P-TEFb [38]. To verify IPA’s prediction of the nuclear presence of Caspase-8, we separated the lysates of HeLa, SiHa, and two High-Grade Serous Ovarian Cancer (HGSOC) cell lines—OVCAR-3 [63] and OVCAR-8 [64], into their cytosol and nuclear fractions. We observed that while endogenous Caspase-8 was predominantly cytoplasmic, it was present, at lower levels, within the nucleus of every cell line (Fig. 4H). Furthermore, similar fractionation of HeLa KO cells, over-expressing Flag-Casp8, also exhibited low expression of the exogenous Caspase-8 within the nucleus (Fig. 4I). To investigate whether Caspase-8 and CDK9 would interact within the nucleus, we IPed Caspase-8 from the cytosol and nucleus of HeLa WT cells, in which CDK9 expression was downregulated with siCDK9. Indeed, Caspase-8 co-IPed CDK9 only from the nuclear fraction of the siCtrl (-) cells (Fig. 4J).

In summary, we have confirmed that CDK9 interacts with the p12 domain of Caspase-8 within the nucleus of cervical cancer cells. We cannot rule out any potential interaction between Caspase-8 and CDK9 in the cytosol. However, at least under our experimental conditions, this interaction in the cytosol was below the limit of detection of the Caspase-8 IP (Fig. 4J).

Caspase-8 inhibits the activity of CDK9

Having demonstrated the interaction between Caspase-8 and CDK9, the next question was about the effect of this interaction on each other. As CDK9 is a kinase, we first asked whether Caspase-8 may be a target for phosphorylation. Therefore, we performed an in vitro kinase assay with γ-32P ATP involving the active P-TEFb (GST-CDK9/Cyclin K) incubated with increasing amounts of GST-Casp8 or GST (Fig. 5A). Surprisingly, instead of P-TEFb phosphorylating Caspase-8, we observed a GST-Casp8 amount-dependent reduction of P-TEFb phosphorylation. Increasing amounts of GST, used as a control, did not influence the phosphorylation of P-TEFb (Fig. 5A, Supplementary Fig. 5A). Importantly, both GST-Caspase-8 and GST did not affect the levels of non-phosphorylated GST-CDK9, as determined by immunoblot (Fig. 5A, lower panel).

Fig. 5

Effect of Caspase-8 on the autophosphorylation and enzymatic activity of CDK9. A For in vitro radioactive kinase assays increasing amounts of GST-Casp8 or GST were incubated with a commercially available, active GST-CDK9/Cyclin K and radioactive γ-32P-labeled ATP. The autoradiogram of phosphorylated GST-CDK9 (upper panel) and the Coomassie staining of GST-Casp8/GST (middle panel) are shown. Immunoblot of GST-CDK9 probed with a CDK9-specific antibody has been shown in the lower panel. B Increasing amounts of GST-Casp8 or GST were incubated with active GST-CDK9/Cyclin K along with His-tagged C-terminal domain (His-CTD) of RNAPII in a radioactive kinase assay (upper panel). Coomassie staining of GST-Casp8/GST (middle panel) and His-CTD (lower panel) is shown. C SPT5 was IPed from lysates of HeLa WT cells using a specific SPT5 antibody. The immunoblot of the IP and input control, which were probed for SPT, is shown (upper panel). The precipitated SPT5 protein was incubated with increasing amounts of GST-Casp8 or GST, active GST-CDK9/Cyclin K, and radioactive γ-32P labeled ATP (middle panel). The Coomassie staining of GST-Casp8/GST is depicted (lower panel). Immunoblot showing the levels of the precipitated, non-phosphorylated SPT5, used as the substrate in the kinase assay, probed with the SPT5 antibody, has been shown in the lowest panel. D HeLa KO cells were transfected with either an Empty Flag-Vector (EV, negative control) or increasing amounts of Flag-Casp8, immunoblotted, and checked for Flag-Casp8, phospho-RNAPII, RNAPII, pCDK9, CDK9, and β-Actin (upper panel). The levels of pCDK9 and phospho-RNAPII, in the presence of increasing amounts of Flag-Casp8, normalized to their respective CDK9 and RNAPII levels, are shown in a bar graph (lower panel). E HeLa WT cells were either transfected with siCtrl or with increasing concentrations of siCasp8, immunoblotted, and checked for Caspase-8, phospho-RNAPII, RNAPII, pCDK9, CDK9, and β-Actin (upper panel). The levels of pCDK9 and phospho-RNAPII, in the presence of increasing concentrations of siCasp8, normalized to their respective CDK9 and RNAPII levels, have been shown graphically (lower panel). F The cytosol (C) and nucleus (N) of HeLa WT and KO cells were fractionated, immunoblotted, and checked for pCDK9, CDK9, Cyclin T1, phospho-RNAPII, RNAPII, transcription regulators BRD4 and NELF-A, Caspase-8, β-Tubulin (cytosolic marker) and Histone H3 (nuclear marker). G HeLa WT and KO cells were treated with 5-ethynyl uridine (EU) or DAPI for staining nuclei (left panel). The ratio of fluorescence intensity of EU to DAPI stain was quantified, normalized, and graphically represented (right panel) [mean ± SD; n = 3; p value (paired t test, two-tailed); **** =  < 0.0001]

The principal cellular function of P-TEFb is to mediate the productive transcriptional elongation by phosphorylating the Ser2 residue of RNAPII [38]. As we have demonstrated that Caspase-8 impedes the phosphorylation of P-TEFb in vitro (Fig. 5A, Supplementary Fig. 5A), we next sought to investigate whether Caspase-8’s interaction with P-TEFb would affect its kinase activity toward RNAPII. With this aim, we performed a similar kinase assay but also included the carboxy-terminal domain (CTD) of RNAPII as a His-tagged recombinant protein (His-CTD). Intriguingly, the presence of GST-Caspase-8, but not GST, triggered an amount-dependent reduction in the phosphorylation of His-CTD (Fig. 5B, Supplementary Fig. 5B). These data suggested that Caspase-8’s interaction with CDK9 blocked P-TEFb’s kinase activity, which in turn affected the P-TEFb dependent phosphorylation of RNAPII.

Next, we were curious whether this inhibition of P-TEFb’s kinase activity also extended to other substrates. P-TEFb has been reported to phosphorylate the SPT5 subunit of DSIF [(DRB) Sensitivity-Inducing Factor], a negative regulator of transcription elongation [38]. As no commercial phsospho-specific antibodies against SPT5 exist, we IPed SPT5 from HeLa WT cells and performed a similar kinase assay with P-TEFb. In line with the previous results, we observed a GST-Casp8, but not GST, amount-dependent reduction in the phosphorylation of the IPed SPT5 by P-TEFb (Fig. 5C, Supplementary Fig. 5C). Once again, both GST-Caspase-8 and GST did not affect the endogenous levels of SPT5, as determined by immunoblot of the IPd SPT5 (Fig. 5C, lowest panel). In summary, we could confirm that in vitro, Caspase-8 inhibited the activity of P-TEFb toward at least two major substrates.

Our next objective was to specify whether this inhibition occurred at Thr186, the activation site of CDK9, and Ser2, CDK9’s phosphorylation site on RNAPII [38]. With this aim, we performed a non-radioactive kinase assay by first precipitating active CDK9, using a Cyclin T1 antibody from WT and two KO clones of HeLa cells, and incubating the precipitates with His-CTD and ATP. We observed higher levels of phospho-Ser2-His-CTD, using a specific phospho-RNAPII (Ser2) antibody, in both the KO clones as compared to the WT cells (Supplementary Fig. 6A). To verify whether Caspase-8 manifests these inhibitory effects within cells as well, we next over-expressed increasing amounts of Flag-Casp8 in HeLa KO cells and observed a Flag-Casp8 amount-dependent decrease in Ser2 phosphorylation of RNAPII (henceforth phospho-RNAPII) and Thr186 phosphorylation of CDK9 (henceforth pCDK9) (Fig. 5D, bar graph). On the other hand, down-regulating CASP8 in HeLa WT cells showed a siCasp8 concentration-dependent increase in phospho-RNAPII and pCDK9 (Fig. 5E, bar graph). Additionally, the cytoplasm and nuclear fractions of WT and KO HeLa (Fig. 5F) and SiHa (Supplementary Fig. 6B) cells also showed increased levels of pCDK9 and phospho-RNAPII in the nuclei of the KO cells as compared to the WT cells.

In summary, these experiments have confirmed Caspase-8’s role in negatively regulating the Thr186 phosphorylation of CDK9 and its kinase activity on the Ser2 phosphorylation of RNAPII within cells.

Due to phospho-RNAPII’s role in transcriptional regulation, we next investigated whether Caspase-8 expression altered overall transcription in the cells. For this, we performed an EU assay, which relies upon incorporating the uridine analogue—5-ethynyl uridine (EU) into the nascently synthesized RNA [45]. Interestingly, we observed a significant increase in the EU uptake in the KO vs. WT cells, suggesting an alteration in overall transcription in the absence of Caspase-8 expression (Fig. 5G). To further confirm Caspase-8’s role in altering overall transcription, we repeated the EU assay by expressing increasing amounts of Flag-Casp8 or siCasp8 in the KO and WT cells, respectively. We observed a significant Flag-Casp8 amount-dependent reduction of EU uptake, as indicated by the decrease in fluorescent signal, as compared to Empty Flag-Vector (EV, negative control) (Supplementary Fig. 6C). In contrast, we observed a significant siCasp8 concentration-dependent increase in fluorescent signal, as compared to siCtrl transfected WT cells (Supplementary Fig. 6D).

In summary, the EU experiments suggested that the loss of Caspase-8 expression altered the overall transcription in HeLa and SiHa cells, possibly by increasing the phosphorylation of CDK9 at Thr186 and CDK9’s targeted phosphorylation of SPT5 and RNAPII.

Caspase-8 inhibits the phosphorylation of CDK9 in primary cervical cancer samples

Our work so far has demonstrated a direct correlation between Caspase-8 expression and pCDK9 levels in cervical cancer cell lines. However, we were curious whether this correlation also extended to primary cervical cancer tissues. To investigate this, we performed IHC staining of pre-treatment biopsies from 69 cervical cancer patients (FIGO IB–IVA) for Caspase-8 and pCDK9 (Fig. 6A). The median pCDK9 index for all patients was 2.0% (range of 0%—50%). The Weighted Score (WS) of Caspase-8 expression revealed a significant correlation with pCDK9 levels in corresponding tumor tissues (p = 0.05; Fig. 6B). Due to the limited number of patients, the WS for Caspase-8 expression was arbitrarily dichotomized with a score of ≤ 6 being classified as “low” and > 6 as “high” expression. The median pCDK9 signal significantly increased in tumors with a low Caspase-8 expression (p = 0.004; Fig. 6C, Table 1).

Fig. 6

Correlation between Caspase-8 expression and pCDK9 in cervical cancer patients. A Examples of cervix cancer biopsies with high and low immunohistochemical detection of Caspase-8 and pCDK9 (Thr186). Original magnification × 100, scale bar: 100 μm. B The association of the immunohistochemical detection of pCDK9 (percentage of positive cells) and the Caspase-8 weighted score (intensity of staining × percent of positive tumor cells) in pretreatment biopsies of 69 patients. C The association of the immunohistochemical detection of pCDK9 with high (score 8–12) and low (score 0–6) Caspase-8 expression. The tick line is the median value, the solid box is the interquartile range, and the whiskers are the 5th and 95th percentiles. D Immunoblot of the lysate of a cervical cancer primary patient cells, in which Caspase-8 expression was either knocked down using siCasp8 or transfected with siCtrl. The immunoblot was checked for the levels of Caspase-8, phospho-RNAPII, RNAPII, pCDK9, CDK9, and GAPDH. The levels of pCDK9 and phospho-RNAPII in siCtrl and siCasp8 transfected cells, normalized to their respective CDK9 and RNAPII levels, have been shown graphically. E 3D cell-invasion assay of patient-derived cervical cancer primary cells transfected with either siCasp8 or siCtrl or was non-transfected (NT). Invaded cells were stained with DAPI, and their quantification was graphically represented. The bottom panel shows the fluorescent images of DAPI stained invaded cells of each type [n = 3; mean ± SD; p value (paired t test, two-tailed); **** =  < 0.0001]

Table 1 Correlation between CASP8 expression and pCDK9 level in cervical cancer patients

To investigate whether the correlation between Caspase-8 expression and pCDK9 levels in cervical cancer patients also extended to phospho-RNAPII levels, we next knocked-down Caspase-8 expression in primary cells derived from a cervical tumor. Immunoblotting revealed enhanced pCDK9 and phospho-RNAPII in the siCasp8-treated cells compared to siCtrl transfected cells (Fig. 6D, bar graph). In addition, the knock-down of Caspase-8 expression in cervical cancer patient-derived cells also led to a significant increase in their invasiveness compared to non-transfected (NT) or siCtrl transfected cells (Fig. 6E).

In summary, these data validated that, in cervical cancer patients, low or high Caspase-8 expression significantly correlated with increased or decreased pCDK9 levels and, by extension, phospho-RNAPII levels, respectively. These results are concomitant with our cellular observations and strongly suggest that low Caspase-8 expression in cervical cancer patients might stimulate enhanced cell-migration and cell-invasion.

Caspase-8 negatively regulates the transcription of cell-migration- or cell-invasion-associated genes

Having demonstrated the altered regulation of Caspase-8 on RNAPII-mediated global transcription, our next question was whether this could lead to alterations in the transcription of genes associated with cell-migration and cell-invasion. This might explain the enhancements in these cellular behaviors in the Caspase-8 KO cells (Fig. 2F, G, Supplementary Figs. 1D, E). With this intent, we first performed a transcriptomics analysis of non-synchronized (NS) and S/G2-phase synchronized (synch.) WT and KO HeLa cells (Fig. 7A). After normalization, ~ 300 transcripts were obtained using the cut-offs of Log10 (p value) of ≤ 1.30 and Log2 FC of ≥  ± 0.5, whose expressions were altered either in the WT [non-synchronized (NS) and S/G2-phase synchronized (synch.)] or KO [non-synchronized (NS) and S/G2-phase synchronized (synch.)] cells. The DAVID GOTERM_BP analysis of the ~ 300 transcripts revealed several cell-migration associated functions (Fig. 7B, highlighted in black). To determine whether the differential mRNA expressions of these genes, due to the presence/absence of Caspase-8 expression, also translated to their protein expression, we also performed proteomics analysis of non-synchronized (NS) and S/G2-phase synchronized (synch.) WT and KO HeLa cells (Fig. 7C). Once again, using the cut-offs of Log10 (p-value) of ≤ 1.30 and Log2 FC [non-synchronized (NS) and S/G2-phase synchronized (synch.)] of ≥  ± 0.5, we identified 60 proteins whose DAVID GOTERM_BP analysis revealed several cell-migration associated functions (Fig. 7D, highlighted in black). Comparison of the 60 proteins and ~ 300 genes registered 14 common genes between the two data sets, of which 10, like STOM, SLC9A3R1 (NHERF1), were similarly over-expressed in the WT and 4, like TGM2 and EPHB2, in the KO cells (Fig. 7E), whose DAVID GOTERM_BP analysis again revealed several cell-migration associated functions (Fig. 7F, highlighted in black).

Fig. 7

Identification of differentially expressed genes in the presence/absence of Caspase-8. A Comparison of differentially expressed transcripts in HeLa WT and KO cells, as detected through HumanHT-12 v3 microarrays. > 48,000 transcripts were detected when not adjusted for FDR q value. The red dashed lines represent a cut-off of − Log10 (p value) of ≤ 1.30 (y-axis) and Log2 FC [non-synchronized (NS) and S/G2-phase synchronized (synch.)] of ≥ 0.5 (over-expressed in the KO cells) and ≥ − 0.5 (over-expressed in the WT cells) (x-axis). Normalization of the raw data by adjusting for FDR q value of ≤ 0.05 resulted in 302 transcripts, with an over-expression of 78 transcripts in the KO and 224 in the WT cells. (B) The top 15 functions regulated by these 302 transcripts were determined using the DAVID GOTERM_BP analysis. Highlighted in black are the processes associated with cell-migration. C Comparison of proteins differentially expressed in Tandem Mass Tag (TMT) labeled HeLa WT and KO cells detected through MS analysis. > 5000 proteins were initially detected, of which 60 proteins (33 in WT and 27 in KO cells) passed the cut-offs of − Log10 (p value) of ≤ 1.30 and Log2 FC [non-synchronized (NS) and S/G2-phase synchronized (synch.)] of ≥  ± 0.5, as represented by the red dashed lines. D The top 15 functions regulated by these 60 proteins, as determined using the DAVID GOTERM_BP analysis. Highlighted in black are the processes associated with cell-migration. E Of the 302 genes from transcriptome analysis and 60 protein-encoding genes from the proteome analysis, 14 were shared between the two data sets. Representation of the expression profiles of the 14 common genes in the transcriptome and proteome data sets. F The top 10 functions are regulated by the 14 common genes, as determined using the DAVID GOTERM_BP analysis. Highlighted in black are the processes associated with cell-migration. Both the transcriptome and proteome analyses were performed in triplicate. G Immunoblot of HeLa WT and KO cells representing the protein levels of TGM2, Stomatin, Caspase-8, and GAPDH

To confirm our findings, we checked the protein expressions of two of these genes—STOM and TGM2 in HeLa WT and KO cells due to their reported roles in promoting metastasis in different cancer entities [65, 66]. We observed that TGM2 expression was up-regulated, while Stomatin was down-regulated in the KO cells (Fig. 7G), precisely matching our analysis (Fig. 7E). Finally, the knock-down of TGM2 (siTGM2) in HeLa and SiHa WT and KO cells significantly reduced their 2D migration, as compared to their respective siCtrl transfected counterparts. This suggested that the enhanced expression of TGM2 is, at least in part, responsible for the enhanced migration of cervical cancer cells, lacking Caspase-8 expression (Supplementary Figs. 7A, B).

This finding is crucial as we could demonstrate that when the HeLa WT and KO cells were treated with increasing concentrations of the classic Epithelial to Mesenchymal Transition (EMT) inducing agent—TGF-β1 [67], the HeLa KO cells displayed significantly enhanced protein and mRNA levels of the mesenchymal marker—Vimentin [68] (Supplementary Figs. 8A, B) and significantly reduced mRNA levels of the epithelial marker—E-Cadherin [