The loss of Cdk5 expression results in increased sensitivity to PARP inhibition and the killing effects of IR
In order to investigate whether Cdk5 depletion has a direct impact on PARP-1 activity and sensitivity to a panel of DNA-damaging agents, we established a stably depleted Cdk5 HeLa cell line (Cdk5KD) using an RNAi system that targeted nucleotides 703–721. These Cdk5KD cells (clone 1,499) were cultivated for more than 80 days in culture with levels of the targeted proteins remaining below 30%, as assessed by the quantification of Western blots, of that seen in Control cells (Fig. 1a). A similar low level of Cdk5 protein expression was seen in two additional Cdk5-depleted clones, 1,500 and 1,501, in which nucleotides 455–473 and 41–59 were targeted, respectively (supplementary Fig. 4f).
The depletion of Cdk5 resulted in a significant increase in the sensitivity to cell killing following treatment with the PARP inhibitor ABT888 for 24 h compared to the response seen in Control cells under the same experimental conditions (p < 0.004) (Fig. 1b). In parallel experiments, comparing the survival of Control and HeLa cells depleted for PARP-1 cells after treatment with 10 μM of roscovitine for 24 h, an inhibitor of Cdk activity, a significantly lower survival was seen in the PARP-1KD cells (p < 0.002) (Fig. 1c). These results confirm those of Turner et al.  that synthetic lethality occurs when PARP and Cdk5 activity are both disrupted in stably depleted non-neuronal cell lines and lends support to a model that the DNA damage response is modified when both activities are compromised.
In order to further characterize the Cdk5KD cells, we compared the killing effects of a panel of three DNA-damaging agents, MMS, NCS, and IR, in these cells to that seen in unsynchronized cell populations of Control cells or cells invalidated for PARP-1 or DNA-PKcs. The different combinations were chosen to generate a spectrum of DNA lesions in backgrounds where different repair pathways are compromised. The base damage formed after exposure to MMS will either be eliminated from the DNA by spontaneous depurination or will be actively repaired through the action of glycosylases such as 3-methyladenine glycosylase and then through the later stages of the BER pathway . NCS is a complex consisting of a dodecadiyne antibiotic (NCSChrom) reversibly bound to a carrier protein. NCSChrom cleaves DNA through a suicide reaction, leaving no residual active drug after a few minutes of incubation. The major DNA lesions induced by NCSChrom in DNA result from radical attack  and consist of a blunt end break with a thymidine-5′-aldehyde residue on one strand, with an atypical abasic site at two-nucleotide interval on the complementary strand. This NCS-induced damage is rapidly converted into DSBs in living cells and thus NCS is considered to be a DSB-inducing agent. Finally, IR generates a variety of DNA base adducts, SSBs, and DSBs, as well as complex DNA lesions.
Cdk5KD and PARP-1KD cells were found to be approximately 1.6 times more sensitive to the killing effects of IR than Control cells, based on the surviving fraction at 2 Gy (SF2) (Fig. 2a). No increased sensitivity was seen in either the Cdk5KD cells or the PARP-1KD cells to either MMS or NCS. However, the DNA-PKcsKD cells were extremely sensitive to NCS (Fig. 2b, c). These similarities in cell survival between the Cdk5KD and the Control cells after treatment with MMS or NCS would suggest that the methylated DNA adducts and DSBs generated directly in the DNA after such treatments can be repaired using mechanisms that do not require Cdk5 but would suggest that Cdk5 is involved in the repair of SSBs, a process which involves PARP-1’s catalytic activity.
Cdk5 depletion modulates PARP-1 recruitment to sites of localized DNA damage and its enzymatic activity
In order to examine whether the depletion of Cdk5 had an impact on the recruitment of PARP-1, and its partner XRCC1, we made use of the technique of laser micro-irradiation using a 405-nm laser to generate localized DNA damage in the nucleus of the cells transiently expressing GFP or YFP tagged PARP-1 or XRCC1, respectively. Confocal microscopy was used to follow and quantify the recruitment and persistence of these proteins based on the increase in fluorescence at the site of DNA damage relative to the background fluorescence of the nucleus, reported as the relative spot intensity.
The time-course of the recruitment of GFP-PARP-1 to damage sites seen in the Control cells was extremely rapid, as previously reported [12, 17–20]. However, in the Cdk5KD cells whilst GFP-PARP-1 was detected at damage sites immediately after their formation (Fig. 3a, first and second rows), the quantification of the relative spot intensity with time showed a statistically reduced level (p < 10−3) of GFP-PARP-1 at the damage sites in the Cdk5KD cells compared to Control cells from 50 s post-irradiation and which persisted up to 300 s post-irradiation (Fig. 3b). A statistically different (p < 10−3) profile of GFP-PARP-1 recruitment was also seen in the Cdk5 depleted clones 1,500 and 1,501 compared to the Control cells (supplementary Fig. 4a and b). As the three serine residues at amino acids 782, 785 and 786 in PARP-1 have been previously reported as in vitro target sites for phosphorylation by Cdk5 , we generated a mutant GFP-tagged PARP-1 plasmid in which all three were mutated to alanines (PARP-1 MUT-GFP). Under the same experimental conditions, we found that this mutated form of PARP-1 was also recruited to sites of DNA damage in Control cells immediately following micro-irradiation (0 min), however, its accumulation and persistence was less than that of the control GFP-PARP-1 and showed no statistical difference to the recruitment profile of GFP-PARP-1 in the Cdk5KD cells. (Fig. 3a rows 2 and 3 and 3b). These results would suggest that the phosphorylation of PARP-1 via Cdk5's kinase activity is necessary for its persistence at damage sites.
It has been well-documented that XRCC1 is recruited to local sites of DNA damage in a PAR and PARP-1-dependent fashion [12, 19, 21, 22]. Therefore, in order to verify the consequences of the reduced persistence of PARP-1 at sites of micro-irradiation-induced DNA damage sites in Cdk5KD cells, we next investigated the recruitment of XRCC1-YFP under the same experimental conditions (Fig. 3a). In agreement with other published studies using a similar experimental set-up, a rapid recruitment of XRCC1 was seen in the Control cells [12, 19, 21–23]. However, while we found that the relative intensity of XRCC1-YFP immediately after micro-irradiation (0 min) was not affected by the loss of Cdk5 expression (Fig. 3a rows 4, 5 and 3c), its accumulation was statistically lower (p < 10−4) in Cdk5KD cells at time points from 100 s after damage formation as compared to Control cells (Fig. 3c). This result was confirmed in the two additional Cdk5-depleted clones 1,500 and 1,501 (supplementary Fig. 4c and d). Supporting these findings, we observed a similar decreased level and slower accumulation of fluorescence from the YFP-tagged XRCC1 in Control cells exposed to the Cdk5 inhibitor roscovitine (supplementary Fig. 1).
The persistence of PARP-1 at sites of SSBs is regulated by the PAR chains formed during the automodification of the PARP-1 protein itself. These PAR chains impart a negative charge on PARP-1 and result in its eventual dissociation from the DNA as a consequence of charge repulsion between the negative charge of the DNA and the PAR chains [1, 22]. In order to investigate whether the reason for the reduced accumulation of PARP-1 observed in the Cdk5KD cells could be related to differences in the levels of PAR formed in the two cell types, we measured the basal levels and the levels formed in response to DNA damage induced by IR or H202 using immunofluorescence making use of a specific anti-PAR antibody.
A statistically higher basal level of PAR was detected in the Cdk5KD cells and also in the clones 1,500 and 1,501 compared to the Control cells (Fig. 4b, d and supplementary Fig. 4e). After exposure to IR, the PAR levels were increased in both cell types with significantly higher absolute levels being found in the Cdk5KD compared to the Control cells (p < 10−4, n = 300–500 cells) (Fig. 4a, b). A similar response was seen after exposure to 1 mM H2O2 (p < 10−5, n = 400–800 cells) (Fig. 4c, d). Confirmatory data obtained from clones 1,500 and 1,501 showing significantly higher absolute PAR levels compared to Control cells after treatment with IR is shown in supplementary Fig. 4e (p < 10−5, n = 200–300 cells).
These observations of higher levels of PAR formation were validated by measuring the changes in NAD(P)H levels using a colorimetric assay . However, due to the technical limitations associated with the assay, this could only be done after the H2O2 treatment, where the exposure could be carried out over a longer period of time. While the PARP-1KD cells showed a negligible decrease in NAD(P)H levels under these experimental conditions, the Cdk5KD cells showed a significantly greater consumption of NAD(P)H compared to the Control cells (p < 10−4), indicative of a higher PARP activity after DNA damage induced by H2O2 (supplementary Fig. 2).
A possible explanation for the finding of higher polymer formation in Cdk5KD cells is that the activity of poly(ADP-ribosyl) glycohydrolase (PARG), responsible for removing the PAR polymers from PARP-1 , is reduced in the absence of Cdk5. Post-translational modifications of PARG have been reported at a number of positions, several of which are located in the N-terminal putative regulatory region raising the possibility that as for PARP-1, phosphorylation events may modulate PARG activity [5, 25, 26]. This N-terminal region contains a sequence around Ser 137 that shows a high degree of homology with the consensus sequence for Cdk5 phosphorylation (KS/TPXK) . However, no differences were seen between the Cdk5KD and Control cells in either their PARG mRNA or protein levels (data not shown), nor in either the basal PARG activity or after exposure of the cells to IR (supplementary Fig. 3). Based on these results and the recruitment data, we hypothesize that the phosphorylation of the PARP-1 protein by Cdk5 on one or more of the serines 782, 785, and 786 results in an attenuation of its ribosylating activity facilitating its persistence at the sites of DNA damage.
Cells lacking Cdk5 are able to effectively repair DNA single-strand breaks (SSB)
In order to investigate whether the impaired recruitment of PARP-1 and XRCC1 in the Cdk5KD cells had an impact on SSB rejoining, the kinetics was compared in unsynchronized Control and Cdk5KD cells after exposure to gamma irradiation (5 Gy) using alkaline elution. Immediately following irradiation, as little as 5% of the total [2-14C] thymidine-labeled DNA from lysed Control and Cdk5KD cells was retained on the filters and in a time-dependent manner over 1 h, the SSB repair proceeds to completion with nearly 100% of the DNA retained on the filter, indicating complete SSB rejoining, by this time point in both cell lines (Fig. 5).
DNA SSB repair can also proceed through the alternative long patch (LP) BER sub-pathway involving the binding of PCNA to sites of DNA damage. In order to investigate whether this mechanism could contribute to the repair of SSBs in the Cdk5KD cells, GFP-tagged PCNA was transiently transfected into these and Control cells, and its recruitment to DNA-damage sites produced by laser micro-irradiation were assessed as previously described [12, 18, 20, 21]. In the Control cells, the fluorescence intensity of PCNA at the irradiated site was significantly lower than for XRCC1, which is in good agreement with earlier studies [12, 18, 23], and increased slowly over the observation period. However, in the Cdk5KD cells, the amount and rate of PCNA-GFP accumulation was significantly higher (Fig. 6).
These results suggest that the SSBs generated might also be repaired via a PCNA-dependent pathway compensating for the deficiencies in the recruitment of the PARP-1 and XRCC1 in the Cdk5KD cells. Thus, although the loss of Cdk5 does not prevent cells from resolving SSBs, it does have an impact on PARP-1 activity and radiation sensitivity.