Abstract
To identify cell cycle regulators that enable cancer cells to replicate DNA and divide in an unrestricted manner, we performed a parallel genome-wide RNAi screen in normal and cancer cell lines. In addition to many shared regulators, we found that tumor and normal cells are differentially sensitive to loss of the histone genes transcriptional regulator CASP8AP2. In cancer cells, loss of CASP8AP2 leads to a failure to synthesize sufficient amount of histones in the S-phase of the cell cycle, resulting in slowing of individual replication forks. Despite this, DNA replication fails to arrest, and tumor cells progress in an elongated S-phase that lasts several days, finally resulting in death of most of the affected cells. In contrast, depletion of CASP8AP2 in normal cells triggers a response that arrests viable cells in S-phase. The arrest is dependent on p53, and preceded by accumulation of markers of DNA damage, indicating that nucleosome depletion is sensed in normal cells via a DNA-damage-like response that is defective in tumor cells.
Introduction
The cell cycle can be divided into two distinct periods, the interphase and the mitotic phase (M). During the interphase, cells are growing and duplicating their DNA, whereas in the mitotic phase, cells divide into two daughter cells. Interphase consists of three separate phases: gap1 (G1), DNA-synthesis (S) and gap2 (G2). Passing from one phase of cell cycle to the next is mainly regulated by cyclins and cyclin-dependent kinases. During the cell cycle, several checkpoints control that cells are ready to pass from one phase to the next. Interphase has two main checkpoints. One is in G1 phase, ensuring that everything is ready for DNA replication and the other is in G2 phase to verify that DNA replication is completed and that any damage to DNA is repaired. In addition, during the S phase, when the entire genome is replicated several checkpoint pathways can be activated as a response to DNA damage or stalled replication forks.
DNA replication is tightly coordinated with chromatin assembly, which depends on the recycling of parental histones and deposition of newly synthetized histones1. Drosophila and yeast S. cerevisiae cells can complete S phase without de novo histone synthesis2, 3. However, loss of histone expression or limiting assembly of nucleosomes to DNA by targeting chromatin assembly factors such as CAF-1, ASF1 and SLBP have been reported to induce S phase arrest in human tumor cells4–8. However, the mechanism of this arrest is still poorly understood.
Many regulators of the cell cycle have been identified by loss of function screens in yeast. Genome-wide RNAi screens have subsequently been used to identify both regulators that are conserved in and specific for higher organisms such as Drosophila9 and human10–14. In addition to differences between species, regulation of the cell cycle is often altered in normal and tumor cells from the same organism. This is thought to be at least in part due to mutations in cell cycle regulatory proteins such as RB115 and p5316. However, a systematic comparative study that would identify regulators that are differentially required in normal and tumor cells has not been performed.
Results
Genome-wide RNAi screen identifies differential regulation of S-phase progression in normal and cancer cells
To understand similarities and differences in cell cycle regulation in normal and cancer cells, we performed a genome-wide RNAi screen simultaneously in two distinct cell lines: the osteosarcoma cell line U2OS and the hTERT-immortalized normal retinal pigment epithelial cell line hTERT-RPE1.
U2OS and hTERT-RPE1 cells grown on 96-well plates were transfected in triplicate with the same transfection mixes, containing pooled siRNAs (Qiagen) targeting a single gene per well. A total of 23 348 genes were targeted, and the cell cycle phase of the cells was analyzed after 3 days using laser scanning cytometry (Fig. 1a). For 17 095 targeted genes a total of five phenotypes were analyzed: the total number of cells, the fraction of cells in G1, S and G2 phases, and the fraction of cells that had higher DNA content than G2 phase cells (overG2; Fig. 1b, Supplementary Table S1). These represent cells that have replicated DNA more than once without dividing.
In general, RNAi treatments targeting genes known to be required for mitosis (e.g. CDK1, PLK1, KIF11) had similar phenotypes in both cell lines (Fig. 1b, Supplementary Fig. 1a). However, several known S-phase genes (e.g. PCNA, POLA1, RBX1, RRM1) had much stronger S phase arrested phenotypes in hTERT-RPE1 cells compared to U2OS (Fig. 1c). One of the strongest S-phase phenotypes in both cell lines was caused by siRNAs targeting caspase 8 associated protein 2 (CASP8AP2, also known as FLASH; Fig. 1c). CASP8AP2was also the strongest S-phase regulator in a secondary screen with a Dharmacon siRNA library targeting 55 of the identified cell cycle genes in nine different cell lines (Supplementary Table S2; Supplementary Fig. 1b). siRNA targeting of two other known regulators of histone gene transcription, NPAT and HINFP also resulted in an increase in the fraction of cells in the S-phase in most of the nine cell lines studied.
Loss of histone gene transcription regulators differentially affects S-phase progression
To validate disruption of S-phase progression by loss of the regulators of histone genes we transfected U2OS and hTERT-RPE1 cells with CASP8AP2, NPAT, HINFP and control siRNA pools and then measured the DNA synthesis rate by incorporation of the thymidine analogue 5-Ethynyl-2′-deoxyuridine (EdU). In both U2OS and hTERT-RPE1 cells, knockdown of CASP8AP2 dramatically decreased EdU incorporation in S-phase. Knockdown of NPAT and HINFP had a similar effect in U2OS cells with accumulation of cells with poor EdU incorporation. However, in hTERT-RPE1 cells depletion of NPAT and HINFP failed to appreciably affect S-phase progression (Fig. 2a).
CASP8AP2, NPAT, HINFP and E2F1 have different impact on histone gene expression
To determine the effect of loss of CASP8AP2, NPAT and HINFP on histone gene expression, we profiled gene-expression in siRNA treated U2OS and hTERT-RPE1 cells using Affymetrix WT1.1 arrays (Supplementary Table S3). We found that CASP8AP2, NPAT and HINFP do not regulate expression of each other, but mainly affect the expression of histone genes. Most histone genes were downregulated in U2OS cells following loss of CASP8AP2, NPAT or HINFP. In normal cells, some highly expressed histone genes were downregulated, albeit less than in tumor cells. In addition, many histone genes that are normally expressed at lower levels were upregulated (Fig. 2b; Supplementary Table S3).
To identify whether CASP8AP2, NPAT and HINFP directly bind to the histone gene promoter regions we carried out ChIP-Seq in U2OS and hTERT-RPE1 cells. Consistent with previous findings, HINFP was found enriched near transcription start sites (TSSs) of replication-dependent histones H4 and H2B17–20 (Supplementary Table S4 and S5). We also found that HINFP regulated two replication-independent histone H1 genes, H1F0 and H1FX (Supplementary Table S4 and S5). In contrast, CASP8AP2 and NPAT ChIP-Seq peaks were only found colocalized at replication-dependent histone genes on chromosomes 1, 6 and 12 in both cell lines (Fig. 2c, Supplementary Table S4 and S5). These results indicate that CASP8AP2 and NPAT regulate only replication-dependent histones, whereas HINFP regulates a subset of replication dependent histones (H4 and H2B), and two replication indipendent H1 variants (H1F0 and H1FX).
Another histone gene regulator, E2F121, 22, also bound to TSSs of many histone genes, including both replication dependent and independent histones (Supplementary Table S4 and S5). In addition, E2F1 bound to the promoter of CASP8AP2, suggesting that E2F proteins control CASP8AP2 and histone expression directly and via a feed-forward loop, respectively.
CASP8AP2 knockdown results in low histone H3 protein levels and slows progression of replication forks in osteosarcoma cells
To analyze the long-term effect of CASP8AP2 loss on S-phase progression and histone protein levels, we treated U2OS and hTERT-RPE1 cells with CASP8AP2 siRNAs, and analyzed DNA content, histone H3 protein level, and EdU incorporation by flow cytometry in the same population of the cells. We found that CASP8AP2 siRNA treatment did not completely arrest U2OS cells in S-phase, but instead dramatically slowed down S-phase progression, resulting in an S-phase that lasted more than 3 days (Fig. 3a). The slowdown in S-phase was accompanied by increased cell death, and a marked decrease in histone H3 protein levels in the surviving cells that continued to replicate (Fig. 3b). In contrast, viable hTERT-RPE1 cells were arrested in S-phase by CASP8AP2 siRNA, and very little if any effect on histone H3 protein levels was detected (Supplementary Fig 2a-d).
To identify the mechanism of the slowdown of the cell cycle in the U2OS cells, we analyzed the speed of replication forks in CASP8AP2 siRNA treated U2OS cells using a DNA fiber assay, where DNA is labelled by two different labels consecutively, and then analyzed visually to detect the length of the labelled regions. This analysis confirmed earlier observations23 that CASP8AP2 siRNA generally decreases the speed by which replication forks progress (Fig. 3c–e), suggesting that in human cancer cells, individual replication forks are affected by nucleosomes loading behind them.
The ability of normal cells to activate H2AX phosphorylation in response to deregulation of histone expression is p53 dependent
To determine why normal and tumor cells respond differentially to CASP8AP2 loss, we examined the expression profiles of non-histone genes after CASP8AP2, NPAT and HINFP siRNA treatment. This analysis revealed that p53 target genes were upregulated in hTERT-RPE1, but not in p53-proficient U2OS cells (Fig. 4a). The p53 target genes were upregulated relatively late, clearly, after the changes in histone gene expression were observed (Supplementary Fig. 3 a,b; Supplementary Table S6).
One possibility that would explain the activation of p53 is that the slow replication causes DNA damage or a DNA damage-like state that is sensed by p53. Consistently with this hypothesis, an increase in a marker commonly associated with DNA damage, gamma-H2AX, was observed more in hTERT-RPE1 cell line (Fig. 4b; Supplementary Fig. 3c,d) and less in U2OS cell line (Supplementary Fig. 3c,d; 4a-d). To address whether the p53-dependent pathway is responsible for the S-phase arrest observed in normal but not in tumor cells, we analyzed the expression of the p53 target gene p21. We found that after CASP8AP2 RNAi treatment, p21 protein was induced more in hTERT-RPE1 cells (Supplementary Fig. 5a), and only in S-phase cells that express elevated gamma-H2AX (Fig. 4c). Similar results were found in two other normal human cell lines, HFL1 and CCD-1112Sk (Supplementary Fig. 6a-d).
To test if the effect is dependent on p53, we generated a derivative of hTERT-RPE1 cell line that lacks p53 using CRISPR/Cas9 genome engineering. This cell line failed to completely arrest in S-phase or induce p21 in response to CASP8AP2 RNAi (Supplementary Fig. 5a,b). In addition, significantly lower levels of gamma-H2AX accumulation were observed in the p53 deficient cell line (Supplementary Figs. 3c,d and 6a-d). The response of the p53 deficient cell line was similar to that of tumor cells, including U2OS and HCT116 (Supplementary Fig. 6a-d).
Discussion
Genome-wide RNAi screens have become very powerful and informative approaches for the analysis of cell cycle regulation and for identification of new regulators of cell cycle progression and checkpoint control. Unlikely previously published genome-wide RNAi screens performed to identify genes affecting cell cycle and cell size in Drosophila cells9 and in human cancer cells10–14 the present study compares two distinct cell lines: the osteosarcoma cell line U2OS and the hTERT-immortalized normal retinal pigment epithelial cell line hTERT-RPE1. As expected, several known mitotic regulators, and the known transcriptional regulator of histones, CASP8AP2 displayed strong phenotypes in both types of cells. However, some DNA-replication regulators, including two other transcriptional regulators of histones, HINFP and NPAT, displayed differential effects in the two cell types.
The canonical histones are exclusively expressed in S phase of cell cycle22, 24, 25. This process is regulated by E2F and cyclin E-Cdk2 kinase through phosphorylation of NPAT19, 21, 26,28. Based on microarray data and flow cytometry, the depletion of CASP8AP2 caused strong downregulation of replication-dependent histones in U2OS cells. In normal cells, loss of CASP8AP2 also resulted in deregulation of histone expression, resulting in both positive and negative effects at the level of individual histone genes. The transcriptional regulation of histone genes by CASP8AP2 and NPAT is likely direct, based on their known interaction29, and our chromatin immunoprecipitation followed by sequencing (ChIP-seq) results showing localization of NPAT and CASP8AP2 almost exclusively to the replication-dependent histone loci.
In order to understand the mechanism behind the differential response of normal and tumor cells, we analyzed the S-phase phenotypes in more detail, using EdU incorporation assays in multiple tumor and normal cells depleted of CASP8AP2. This analysis revealed that the increase in cells in the S-phase observed in tumor cells after CASP8AP2 loss was not caused by an S-phase arrest. Instead, it resulted from a dramatic slowdown of DNA replication. In contrast, the S-phase progression of normal cells was arrested. Further analysis revealed also that the deregulation of nucleosome gene expression triggers a p53-dependent pathway in normal cells, but not in U2OS tumor cells. Thus, p53 activation is a consequence, not the cause, of the observed DNA replication phenotype.
One possibility that would explain the activation of p53 is that imbalance of histone proteins and the disruption of replication process causes DNA damage or a DNA damage-like state that is sensed by p53. Consistently with this hypothesis, an increase in a marker commonly associated with DNA damage, gamma-H2AX, was observed in hTERT-RPE1 cells following CASP8AP2 knockdown. Much lower levels of the marker were seen in similarly treated U2OS cells. Thus, downregulation of histones in tested cancer cells could cause slow progression through S-phase without sufficient accumulation of DNA damage marker to activate p53-dependent cell cycle arrest. It consistent with previous result where short-term depletion of histone genes by CASP8AP2 knockdown did not induce DNA damage markers23.
In summary, we have through a genome-wide analysis of normal and tumor cells identified a significant difference in regulation of cell cycle arrest in response to abnormal regulation of histone genes transcription. In both normal and tumor cells, the depletion of CASP8AP2 alters mainly the expression of replication-dependent histones. Whereas hTERT-RPE1 cells respond by triggering a p53-dependent checkpoint, which leads to cell cycle arrest and recovery, U2OS cells continue to progress in S-phase, leading to cell death. The difference was observed in all tested cell types, including eight tumor cell lines and three different normal cell types. Further studies are necessary for determining the precise mechanisms leading to cell death in cells that continue to progress in S-phase despite lack of sufficient levels of histones. The identified defect in sensing histone expression levels could also in part explain the abnormal nuclear morphology commonly associated with cancer cells. Furthermore, the identified novel vulnerability of cancer cells can potentially be used for therapeutic purposes in the future.
Author contributions
M.S. performed most of the experiments with help from M.Tur. O.M. performed DNA fiber assay, P.H. performed immunofluorescence experiments. T.K. performed siRNA sequences mapping to human transcripts. M.S. carried out data analyses and interpreted the results with input from M.B., M.T., T.K. and A.V. T. H. and J.T. supervised experiments and data analysis. M.S and J.T. wrote the manuscript. M.S., M.Tur., T.K., A.V., M.B., M.T., T.H. and J.T. discussed the results and commented on the manuscript.
Competing financial interests
The authors declare no competing financial interests.
Acknowledgments
We thank Sini Miettinen for technical assistance, Drs. M. Laiho and G. Wei for for supplying several cell lines and Dr. Bernhard Schmierer for critical review of the manuscript. This work was supported by the Academy of Finland Center of Excellence in Cancer Genetics and Finnish Cancer Organizations.