Screening of a kinase library reveals novel pro-senescence kinases and their common NF-κB-dependent transcriptional program

A kinase screen identifies new pro-senescence kinases and pathways

Little is known about the ability of activated kinases other than those acting upstream or downstream from RAS signaling to induce premature senescence. To identify novel cell-senescence-inducing kinases, we stably transduced each kinase gene of a previously described and validated library [7] into normal human fibroblasts (Supplementary Table 1). A first selection on the basis of decreased cell proliferation yielded 53 kinases (Supplementary Table 2). We next examined the effects of these 53 kinases on the levels of transcripts of four well-described SASP components: IL1A (Fig. 1A), IL1B (Fig. 1B), IL6 (Fig. 1C), and IL8 (Fig. 1D) [8,9]. Most of the tested kinases proved able to induce expression of the SASP-component genes, sometimes very strongly (in some cases, more than 100-fold induction was observed) (Fig. 1A-D). Generally, variations in the level of one transcript correlated with changes in the levels of the other three. In others words, the more strongly a kinase induced expression of a SASP-component gene, the more strongly it induced expression of the others (Fig. 1E-F). We found 44 kinases to both decrease cell proliferation and induce expression of SASP-component genes (Supplementary Table 3).

We next focused on p16 transcripts. The p16 protein is a known cyclin-dependent kinase inhibitor (CDKI) whose expression generally increases in senescent cells [10]. The level of p16 transcripts was found to increase in response to numerous kinases, but the fold induction did not exceed about 2.5, in contrast to what was observed with SASP-component transcripts (Fig. 2A). We found 37 kinases both to decrease cell proliferation and to induce p16 gene expression (Supplementary Table 4). As shown in a Venn diagram compiling all the data, we found 33 kinases to decrease cell proliferation and to increase expression of both the p16 gene and the SASP-component genes (Fig. 2B and Table 1). Although the same 33 kinases induced expression of p16 and the SASP, p16 and SASP components appeared in different main branches after hierarchical clustering (Fig. 2C). This suggests that they might be regulated by different transcriptional programs. Still, there was a significant correlation between p16 and the different SASP components (Fig. 2D). A KEGG pathway analysis of these 33 kinases reveals over-representation of some signaling pathways (Table 2) and some of them represent new senescence regulating pathways. In conclusion, we have identified 33 kinases and some associated signaling pathways capable of inducing senescence.

NF-κB pathway activation is a shared characteristic of several pro-senescence kinases

To see what the identified kinases might have in common, we first interrogated the STRING database to see if at least those displaying the greatest pro-senescence effect, as judged from the level of SASP induction, might share a common protein association network. We built a network based on the seven kinases showing the strongest SASP-inducing effect (Fig. 3). In a KEGG pathway analysis of this network performed to identify overrepresented pathways, we detected strong enrichment in actors of the NF-κB signaling pathway (p=5.8-16) (green circles, Fig. 3). Interestingly, SASP induction during replicative senescence or in response to senescence triggering by the RAS signaling pathway, through RAS or MEK activation, is reported to be mediated by the NF-κB transcription factors [11,12]. These factors are thus good candidate mediators of SASP-component induction by the pro-senescence kinases identified in our screen.

To confirm experimentally the results of this bioinformatic analysis, we first investigated the ability of the three strongest SASP and p16 inducers identified here (Fig. 2C): the MAP3K7, PRKCD, and MATK kinases to activate the NF-κB pathway using a 3κB-Luc reporter vector [13]. The luciferase activity was strongly induced by the 3 kinases indicating that these pro-senescent kinases activate the NF-κB pathway. We next investigated the role of the NF-κB pathway in mediating SASP-component induction by these kinases. We transduced each pro-senescent kinase into human fibroblasts, in combination or not with the gene encoding the IκBα super-repressor (IKBAm), a stabilized NF-κB inhibitor. As expected, the levels of IL1A, IL1B, IL6, and IL8 transcripts strongly increased upon constitutive expression of the kinases (Fig. 4B-E). Strikingly, this induction was almost completely abolished when NF-κB activity was inhibited by IKBAm (Fig. 4B-E). These data demonstrate that SASP component induction by the tested pro-senescence kinases depends on NF-κB transcription factor activity, as anticipated from the STRING bioinformatic analysis.

We next wondered if these pro-senescence kinases might activate a broader NF-κB-dependent program affecting more than just SASP components. To answer this question, we examined the effect of our 33 kinases on transcript-level expression of the genes encoding three known direct intracellular NF-κB targets: IκBα, SOD2, and COX2. The last two genes are known to be involved in NF-κB-induced senescence [14-16]. The kinases were found to induce expression of all three genes (Fig. 5A-C). Upon hierarchical clustering of the SASP components, the intracellular NF-κB targets, and p16 on the basis of their induction profiles, the SASP components and intracellular NF-κB targets were found to cluster on the same main branch, whereas p16 was still on an independent branch (Fig. 5D). This again suggests that expression of the p16 gene is not controlled directly by an NF-κB-dependent transcriptio-nal program. Nevertheless, correlations were observed between p16, individual SASP-component, and individual intracellular NF-κB-target mRNA levels (Fig. 5E).

SASP components can propagate and amplify the senescence signal [8,9,11,17]. COX2 and SOD2 can also promote cellular senescence induced by Rel/NF-κB transcription factors [14-16]. As the responses of all these factors to our pro-senescence kinases appear to be under the transcriptional control of NF-κB, we examined whether constitutive NF-κB inhibition by IKBAm might prevent senescence induction by pro-senescence kinases. As expected, constitutive expression of PRKCD, MATK, or MAP3K7 blocked cell proliferation (Fig. 6A, upper panel) and induced p16 expression (Fig. 6B). Inhibiting NF-κB with IKBAm did not prevent the kinase-promoted proliferation arrest (Fig. 6A) and only partially inhibited p16 induction (Fig. 6B). This suggests that the tested pro-senescence kinases, although strongly inducing transcription of SASP-component and other pro-senescence regulator genes via NF-κB transcription factors, do not induce senescence solely through the NF-κB pathway.

Inhibition of the p53 or Rb pathway does not prevent senescence or induction of SASP-component gene expression

As the p53 and p16/Rb pathways are important mediators of cell senescence [18,19], we wondered whether they might mediate the cell proliferation arrest and upregulation of SASP-component expression observed in response to pro-senescence kinases. We thus used protein E6 to inhibit the P53 pathway and protein E7 to inhibit the p16/Rb pathway. The ability of E6 or E7 to strongly inhibit its target pathway was demonstrated by acceleration of cell proliferation in normal human fibroblasts infected with an E6- or E7-encoding vector, as compared to cells infected with a control vector (Fig. 7A). Yet even in the presence of E6 or E7, constitutive expression of the MAP3K7, PRKCD, or MATK kinases was still found to promote cell proliferation arrest (Fig. 7B) and increased IL1A, IL1B, IL6, and IL8 transcript levels (Fig. 7C-F). Combination of p53 and NF-kB inhibition has been showed to prevent senescence induced by an oncogenic RAS [12]. We then tested whether MAP3K7-, PRKCD-, or MATK-induced senescence might be overcome by inhibition of p53 by E6 and NF-κB by IKBAm. Inhibition of p53 and NF-κB largely reverted senescence induced by MAP3K7 or PRKCD but had no effect on senescence induced by MATK (Supplemental Figure).

Together, these results demonstrate that inhibition of the p16/Rb or p53 pathway is not sufficient to prevent the proliferation arrest and induction of SASP-component expression triggered by these pro-senescence kinases. Combined inhibition of NF-kB and p53 can revert senescence induced by only some of these pro-senescence kinases.

Cell culture

IMR-90 (ATCC), MRC-5 (ATCC), and GP293 cells (Clontech) were cultured in DMEM. All media were supplemented with 10% FBS (Sigma) and 1% penicillin/streptomycin (Invitrogen). Upon receipt, cells were thawed and amplified and aliquots frozen. Experiments were performed on the aliquots within a month. The cells were maintained at 37°C under a 5% CO2 atmosphere.

Vectors, transfection, and infection

The library used was the Myristoylated Kinase Library described in [7] (Kit # 1000000012, Addgene). The retroviral vectors employed were pBabe-Puro-IκBα-mut (super repressor) [7], pLXN-E6, pLXN-E7, and pLXN-E6E7 [31] (Addgene). The 3κB-Luc reporter vector and the protocol for the transactivation assays were previously described [13]. The protocols used to transfect virus-producing GP293 cells and infect target cells have been described previously [32]. The infection protocols were designed so that practically all cells were infected, as judged from the fluorescence observed after infection with a GFP-expressing retroviral vector.

Colony formation assays

Colony formation assays were carried out in 6-well plates. Depending on the experiment, 50,000 to 100,000 cells were seeded per well. One day later they were infected. Five to ten days after seeding, the cells were washed with PBS, fixed with 4% paraformaldehyde, and stained with 0.05% crystal violet (Sigma-Aldrich).

RNA extraction, reverse transcription, and PCR

To isolate RNA, we used a phenol-chloroform extraction method involving cell lysis in RNA-ISOL lysis reagent (Dutscher). PhaseLockGel tubes (Prime) were used to separate the phases. Thereafter, the Dynamo cDNA Synthesis Kit (Fisher Scientific) was used for cDNA synthesis from 1 μg total RNA. The RT reaction mixture was diluted 1/20 and the cDNA template used for qPCR analysis. TaqMan quantitative PCR analysis was carried out in the CFX96 Connect Real-Time PCR Detection System (Bio-Rad). The FastStart Essential Probes Master (Roche) was used as PCR mix. The ACTB housekeeping gene was used for normalization. Real-time intron-spanning PCR assays were designed with the ProbeFinder software (Roche Applied Science). The following primers and UPL probes were used: ACTB-Forward ATTGGCAATGAGCGGTTC and ACTB-Reverse GGATGCCACAGGACTCCAT (UPL probe: 11), IL8-F agacagcagagcacacaagc and IL8-R atggttccttccggtggt (UPL probe: 72), IL6-F caggagcccagctatgaac and IL6-R gaaggcagcaggcaacac (UPL probe: 7), IL1A-F ggttgagtttaagccaatcca and IL1A-R tgctgacctaggcttgatga (UPL probe: 6), IL1B-F tacctgtcctgcgtgttgaa and IL1B-R tctttgggtaatttttgggatct (UPL probe: 78), p16-F gtggacctggctgaggag and p16-R ctttcaatcggggatgtctg (UPL probe: 34), IKBA-F gtcaaggagctgcaggagat and IKBA-R atggccaagtgcaggaac (UPL probe: 38), COX2-F gctttatgctgaagccctatga and COX2-R tccaactctgcagacatttcc (UPL probe: 2), SOD2-F aatcaggatccactgcaagg and SOD2-R taagcgtgctcccacacat (UPL probe: 3).

Heatmaps and correlations

Heatmaps and correlation analyses were done with R (gplots, corrgram, and Hmisc package). Briefly, normalized RT-qPCR tables were loaded in R and relative fold induction was calculated for each cytokine. One in FC induction is set to Zero, maximum FC is set to One (100% induction). Details of calculation are described as supplementary Methods.