CDK9 is a replication stress response protein

To validate that CDK9 is an RSR protein, we examined the ability of CDK9-silenced cells to recover from a transient replication fork arrest using cell cycle analysis by flow cytometry. U2OS cells treated with HU for 20 h were observed to be arrested in early S-phase (Fig 2A). At 10 h after removing HU, cells treated with nontargeting siRNA oligonucleotides progressed through S-phase and accumulated 4N DNA content, whereas U2OS cells treated with ATRIP, ATR or CDK9 siRNA oligonucleotides have a delayed progression through S-phase (Fig 2A,B). A similar impairment in recovery after CDK9 silencing was observed in human telomerase-immortalized epithelial cells, suggesting that the phenotype is not cell-type-specific (data not shown). Depletion of CDK9 caused a similar defect in recovery after a replication challenge of aphidicolin, a DNA polymerase inhibitor (Fig 2A,B). In the absence of exogenous damage, no changes in cell proliferation or apoptosis are seen after CDK9-silencing (supplementary Fig S2 online).

Open in a separate window

Figure 2

Cyclin-dependent kinase 9 is required for cells to complete DNA synthesis after replication stress. (A,B) U2OS cells were transfected with NT, ATR, ATRIP or CDK9 siRNA, treated with 3 mM HU or 15 μM APH for 20 h (arrested), and released into 1 μg/ml nocodazole for 10 h (released). DNA content was analysed by using flow cytometry. (B) The percentage (mean and s.d.) of cells that completed DNA synthesis in three replicate experiments is shown. (C) Depletion of CDK9 induces DNA damage signalling in replicating cells. U2OS cells were transfected with NT, CHK1 or CDK9 siRNA or treated with 5 Gy IR and processed 72 or 4 h later, respectively, for γH2AX staining by indirect immunofluorescence microscopy. The percentage of cells staining for γH2AX was scored. (D) U2OS cells were co-stained with γH2AX and CCNA or γH2AX and BrdU 72 h after transfection with siRNA. Quantitation of the percentage of cells co-staining with (E) γH2AX and CCNA or (F) γH2AX and BrdU 4 h after treatment with 5 Gy IR or 72 h after transfection with CHK1 or CDK9 siRNA. Mean and s.d. values from three replica experiments are shown. Asterisks in all panels indicate P<0.05. APH, aphidicolin; ATR, ataxia telangiectasia and Rad3-related protein; ATRIP, ATR-interacting protein; BrdU, 5-bromo-2-deoxyuridine; CCNA, cyclin A; CDK9, cyclin-dependent kinase 9; HU, hydroxyurea; IR, ionizing radiation; NT, nontargeting; siRNA, small interfering RNA.

Next, we examined the induction of DNA damage after CDK9 knockdown by immunofluorescence microscopy for phosphorylated histone γH2AX. Silencing many genes that function in RSR pathways causes an increase in spontaneous γH2AX staining during S-phase due to a failure to maintain replication fork stability, even in the absence of added genotoxic agents (Lovejoy et al, 2009; Paulsen et al, 2009). At 72 h after CDK9 silencing, phosphorylation of γH2AX was significantly increased compared with nontargeting silencing. To determine whether the induction of γH2AX occurs in replicating cells, we co-stained the cells for cyclin A—a cell marker in S- and G2-phase—and 5-bromo-2-deoxyuridine (BrdU)—a cell marker in S-phase. In contrast to cells treated with ionizing radiation—which causes damage in all phases of the cell cycle—cells in which CHK1 or CDK9 is silenced exhibit significantly increased co-staining for γH2AX and cyclin A (Fig 2D,E) and for γH2AX and BrdU (Fig 2D,F), suggesting that silencing of CDK9 induces DNA damage in replicating cells. Collectively, these findings demonstrate that CDK9 functions in an RSR pathway to maintain genome integrity during DNA replication.

CDK9 kinase activity is essential for its functions in the RSR

To assess whether the kinase activity of CDK9 is essential for mediating cell cycle recovery after replication stress, we generated U2OS cell lines stably expressing exogenous wild-type FLAG–HA (haemagglutinin)–CDK9 or FLAG–HA–CDK9 D167N—a kinase-dead mutant (Garriga et al, 1996)—and silenced endogenous CDK9 using siRNA targeting the 3′-untranslated region. Wild-type FLAG–HA–CDK9 but not FLAG–HA–CDK9 D167N complemented the HU and aphidicolin recovery deficits of CDK9-silenced cells (Fig 3A,B), suggesting that the kinase activity of CDK9 is essential for its functions in the RSR and confirming the siRNA phenotypes are not due to off-target effects. Western blot analysis demonstrated the expression of exogenous fusion proteins and knockdown of endogenous CDK9 in these experiments (Fig 3C).

Open in a separate window

Figure 3

Cyclin-dependent kinase 9 activity is essential for its activities in the replication stress response. (A,B) U2OS cells stably expressing an empty vector, wild-type FLAG–HA–CDK9 or FLAG–HA–CDK9 D167N were transfected with NT or CDK9 siRNA directed against the 3′-UTR of CDK9, treated with 3 mM HU or 15 μM APH for 20 h and released into nocodazole for 10 h. DNA content was analysed by flow cytometry. (B) The percentage (mean and s.d.) of cells that completed DNA synthesis in three replicate experiments is shown. (C) Western blot analysis demonstrating expression of fusion proteins and knockdown of endogenous CDK9. APH, aphidicolin; CDK9, cyclin-dependent kinase 9; HA, haemagglutinin; HU, hydroxyurea; NT, nontargeting; siRNA, small interfering RNA; UTR, untranslated region.

Cyclin K is a replication stress response protein

To determine which regulatory subunit works with CDK9 in the RSR, we examined cell cycle recovery after a replication challenge of HU or aphidicolin in cells silenced for cyclins T1, T2 and K. Four siRNAs targeting cyclin K strongly impaired cell cycle recovery (Fig 4A,B). By contrast, silencing cyclin T1 and cyclin T2 did not cause a deficit in cell cycle recovery (Fig 4A,B; supplementary Fig S3 online), suggesting that cyclin K, but not cyclin T1 or cyclin T2, is the regulatory subunit of CDK9, which mediates its activities in the RSR. Western blot analysis of cells treated with siRNA against cyclin K, T1 and T2 confirmed the knockdown of protein expression (Fig 4C–E). Silencing of cyclin K in the absence of exogenous damage also induced a significant increase in phosphorylation of γH2AX compared with nontargeting silencing (Fig 4F), consistent with a role for cyclin K in maintaining genome integrity.

Open in a separate window

Figure 4

The cyclin-dependent kinase 9 regulatory subunit cyclin K, but not cyclin T1 or cyclin T2, is required for cell cycle recovery after replication arrest. (A,B) U2OS cells were transfected with NT, cyclin K, cyclin T1 or cyclin T2 siRNA, treated with 3 mM HU or 15 μM APH for 20 h and released into nocodazole for 10 h. DNA content was analysed by using flow cytometry. (B) The percentage (mean and s.d.) of cells that completed DNA synthesis in three replicate experiments is shown. (C–E) Western blot analysis demonstrating efficiency of knockdown with four unique siRNAs. (F) Depletion of cyclin K induces DNA damage signalling. U2OS cells were transfected with NT, CHK1 or cyclin K siRNA or treated with 5 Gy IR and processed 72 or 4 h later, respectively, for γH2AX by indirect immunofluorescence microscopy. Mean and s.d. values from three replicate experiments is shown. Asterisks indicate P<0.05. APH, aphidicolin; CDK9, cyclin-dependent kinase 9; HU, hydroxyurea; IB, immunoblotted; IR, ionizing radiation; NT, nontargeting; siRNA, small interfering RNA.

The absence of a deficit in cell cycle recovery after silencing of cyclin T1 and cyclin T2—well-characterized regulatory subunits of the P-TEFb complex—suggests that the RSR phenotypes observed after CDK9 knockdown might be independent of P-TEFb-mediated transcription regulation. To further examine this issue, we performed a microarray genome-wide expression analysis in U2OS cells treated with nontargeting or CDK9 siRNA under the same conditions in which we obtained our loss-of-function phenotypes. The microarray results are summarized in supplementary Table S2 online. The expression levels of 60 genes without treatment with HU and 127 genes treated with HU were significantly downregulated by at least 1.5-fold, and the expression levels of 75 genes without treatment with HU and 138 genes treated with HU were significantly upregulated after CDK9 silencing, compared with nontargeting silencing. However, few of the regulated genes have any functional link to the cell cycle or DNA damage, suggesting that transcriptional changes might not account for RSR defects in CDK9-silenced cells.

CDK9–cyclin K complex interacts with other RSR proteins

To test whether CDK9 functions directly in the ATR-dependent RSR pathway, we tested whether CDK9 could form complexes with ATR or other RSR proteins. Immunoprecipitation of CDK9–HA expressed in HEK293T cells pulled down endogenous ATR, Claspin and ATRIP, but not PCNA or RPA32 (Fig 5A). Endogenous CDK9 also co-immunoprecipitated with endogenous ATR, Claspin and ATRIP (Fig 5B,C), suggesting that the proteins interact in a complex or complexes. These interactions are not DNA-dependent or regulated by replication stress as they are preserved after treatment of the immunoprecipitation reactions with benzonase nuclease and are unchanged after treatment of the cells with HU (data not shown). We further investigated whether any of the regulatory cyclins interacted with components of the RSR. Immunoprecipitation of ATR pulled down both CDK9 and cyclin K but not cyclin T1 or cyclin T2 (Fig 5C,D). These data further confirm that cyclin K, not cyclin T, mediates the RSR functions of CDK9.

Open in a separate window

Figure 5

Cyclin-dependent kinase 9 interacts with components of the replication stress response, localizes to chromatin and reduces chromatin-bound RPA after replication stress. (A) Cells were transfected with CDK9–HA, harvested to prepare nuclear extracts and immunoprecipitated with HA or control IgG antibodies. Bound proteins were washed, separated by SDS–PAGE and immunoblotted with antibodies against ATR, Claspin, ATRIP, HA, RPA32, and PCNA. (B–D) Endogenous CDK9, ATR, Claspin or ATRIP were immunoprecipitated from cell lysates as indicated. Immunocomplexes were washed, separated by SDS–PAGE and immunoblotted with antibodies against the indicated proteins. (E) U2OS cells were transfected with the indicated siRNAs and tested for sensitivity to HU. Mean and s.d. values from three replicate experiments is shown. (F) Soluble and chromatin fraction of U2OS cells transfected with NT or CDK9-3 siRNA for 72 h and treated with or without 3 mM HU for 20 h. Chromatin-bound RPA70/ORC2 and CDK9/ORC2 ratio of representative blot from three independent experiments is shown. APH, aphidicolin; ATR, ataxia telangiectasia and Rad3-related protein; ATRIP, ATR-interacting protein; CDK9, cyclin-dependent kinase 9; HA, haemagglutinin; HU, hydroxyurea; IB, immunoblotted; IR, ionizing radiation; ORC2, origin recognition complex 2; PCNA, proliferating cell nuclear antigen; RPA, replication protein A; SDS–PAGE, sodium dodecyl sulphate–polyacrylamide gel electrophoresis; siRNA, small interfering RNA.

To determine whether CDK9 and ATR function in a common pathway, we performed epistasis experiments for HU sensitivity after silencing of ATR, ATRIP and CDK9. The silencing of ATR and ATRIP showed a similar HU sensitivity compared with ATR silencing alone, consistent with ATR and ATRIP functioning in the same pathway (Fig 5E). Similarly, ATR silencing and CDK9 silencing also caused a similar HU sensitivity compared with ATR silencing alone, suggesting that ATR and CDK9 act in a common pathway that sensitizes to HU (Fig 5E).

To investigate the function of CDK9 in the ATR signalling pathway, we examined the requirement of CDK9 for ATR-dependent signalling in response to replication stress. Silencing of CDK9 did not impair CHK1 or minichromosome maintenance 2 phosphorylation in response to HU treatment, suggesting that CDK9 does not regulate ATR activation (supplementary Fig S4A online). Furthermore, CHK1 autophosphorylation is not impaired in CDK9-silenced cells (supplementary Fig S4B online). We did observe an increase in CDC25A protein levels (with no change in mRNA levels, P=0.515) in CDK9-silenced cells both before and after treatment with HU, which was rescued by the siRNA-resistant CDK9 complementary DNA (supplementary Fig S4B online). These data suggest that CDK9 might function either downstream from CHK1 or perhaps in a parallel pathway to regulate CDC25A stability.

We thank D. Hallahan for useful discussions. This study was supported by National Cancer Institute grant R01CA136933 to D.C.; the Vanderbilt Center in Molecular Toxicology (P30ES000267), Vanderbilt-Ingram Cancer Center and Vanderbilt Institute for Clinical and Translational Research (UL1 024975); and a Department of Defense Breast Cancer Research Programme Era of Hope Postdoctoral Award BC085283 to D.S.Y., who is also a recipient of the American Board of Radiology Holman Research Pathway fellowship.