Amplification of TLK2 Induces Genomic Instability via Impairing the G₂–M Checkpoint

This article is featured in Highlights of This Issue, p. 893

Estrogen receptor–positive (ER⁺) breast cancers (also known as luminal breast cancers) account for a vast majority of all breast cancers and can be classified into A and B intrinsic subtypes. In contrast to the slow-growing and endocrine-sensitive luminal A tumors, the luminal B tumors are more aggressive form of ER⁺ breast cancers characterized by higher proliferation index and worse clinical outcome after endocrine therapy. Recent large-scale genomic profiling studies suggest that the markedly enhanced accumulation of chromosomal aberrations is characteristic of luminal B breast tumors (1). Chromosomal instability (CIN) is the major form of genomic instability in human cancers and is characterized by an increased rate of numerical and structural alterations in the chromosomes. CINs have been linked to disease progression, distant metastasis, and therapeutic resistance in breast cancer (1, 2), which pose a great challenge to clinical management. Because of mechanistic connection, it is increasingly accepted that numerical and structural CINs cannot be considered in isolation (3); thus, genome-wide copy number aberrations are often used as a surrogate marker to evaluate the level of CIN in cancer (4).

In the presence of DNA damage during S–G₂ phase, cell cycle is arrested at G₂–M checkpoint to ensure the cells to repair their DNA before enter into mitosis. The key regulatory step of mitotic entry is the activation of Cdk1 via dephosphorylation of Cdk1 at Thr14 and Tyr15, which is carried out by the Cdc25 phosphatase (5). G₂–M checkpoint signaling in response to DNA damage activates Chk1 and Chk2, which in turn repress Cdc25 phosphatases, resulting in the inactivation of CDK1 and cell-cycle arrest at G₂–M checkpoint (6). After DNA repair is completed, the mitotic kinases, such as AURKA and PLK1, have a key role in G₂–M checkpoint recovery. In multiple tumors, the amplification and overexpression of AURKA or PLK1 are known causes of G₂–M checkpoint defect and enhanced CIN, which is often related to increased tumor aggressiveness (7). Thus, targeted agents are being actively developed against these mitotic kinases, and the Aurora A kinase inhibitors are now in advanced clinical developments for treating solid tumors (8, 9). It is therefore critical to discover additional genetic aberrations of cell-cycle kinases independent of AURKA or PLK1 that promote G₂–M checkpoint defects and CIN in the luminal B breast cancers so as to develop new targeted therapies.

In our previous study, we have identified a cell-cycle kinase called “Tousled-like kinase 2” (TLK2) that is targeted for amplification in approximately 10.5% of ER⁺ breast tumors, and amplification of TLK2 appears to be enriched in the luminal B breast cancers. The resulting overexpression of TLK2 endows increased invasiveness of luminal breast cancer cells and correlates with a poorer outcome of ER⁺ breast cancer patients. In the current study, we discovered that TLK2 overexpression correlates with increased CIN of breast cancers measured by genome-wide copy number aberrations, which is independent of the known CIN causal factors, such as AURKA and PLK1. TLKs are nuclear-enriched cell-cycle kinases that have maximal activity during S phase and are rapidly inactivated in response to the DNA damage induced by ionizing radiation (IR; refs. 10, 11). As the role of TLK2 in DNA damage response (DDR) was largely based on the studies focusing on TLK1 (11, 12), the role of TLK2 in DDR is poorly understood, and there is no report about its function in CIN in breast cancer. Our further experimental studies revealed the crucial role of TLK2 overexpression in impairing G₂–M checkpoint signaling, delayed DNA repair, and increased CIN. These data further support the rationale to target the cell-cycle kinase TLK2 in the management of more aggressive luminal breast cancers.

Affymetrix SNP 6.0 array–based CNV data (level 3) were retrieved for 1,083 The Cancer Genome Atlas (TCGA) breast-invasive carcinoma samples. To extract a set of high-confidence copy number alternations (CNA) we used the segment mean threshold of 0.3 for copy number gain and −0.3 for copy number loss, as previously reported (13). For a given sample, we calculated the genomic instability index from these CNAs using the following equation:

The copy number break point index for each breast tumor was calculated as the sum of the copy number break points of each chromosome (= total number of copy number segments of each chromosome − 1). The cutoff of TLK2 overexpression was calculated on the basis of median + 1 × MAD (median absolute deviation). MAD is calculated using the R with default constant (=1.4826). PAM50-based clinical subtypes of breast cancer for TCGA samples were derived from the UCSC Cancer Genome Browser (https://genome-cancer.ucsc.edu/).

T47D and MCF10A cells were obtained by Dr. Dean P. Edwards (Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, TX 77030, USA) from ATCC included in the NCI-ATCC ICBP 45 cell line kit. 293FT cells used for lentivirus packaging were purchased from Invitrogen. T47D cells were cultured in RPMI 1640 (Cellgro) with 10% FBS (Thermo Fisher Scientific). 293FT cells were cultured in DMEM (Thermo Fisher Scientific) with 10% FBS. MCF10A was cultured as described previously (14). All cell lines were authenticated by characterized cell line core facility of MD Anderson Cancer Center (Houston, TX) performing short tandem repeat analysis of DNA.

Cells were prepared on coverslips for immunostaining as described previously (15). For primary antibody, rabbit anti-γ-H2AX antibody (Bethyl Laboratories) or mouse anti-γ-H2AX (Millipore) with rabbit anti-TLK2 (Bethyl Laboratories) were used. A total of 500 cells were counted in each condition; >10 γ-H2AX foci–containing cells were considered as positive cells.

For cell-cycle analysis, cells were fixed in 70% ethanol (EtOH) and then stained with propidium iodide (Sigma-Aldrich). For mitotic analysis, cells were incubated with rabbit anti-phospho-H3 antibody (Cell Signaling Technology) for 2 hours at room temperature and then incubated with Alexa 488 goat anti-rabbit antibody (Invitrogen) for 1 hour at room temperature. Cells were analyzed using FACSCanto II Cell Analyzer (BD Biosciences).

Comet assay kit was purchased from TREVIGEN and assay was followed as per the manufacturer's protocols (https://www.trevigen.com/cat/1/3/0/CometAssay/). Briefly, 1 × 10⁵ cells were collected and suspended in 500 μL of cold PBS. A total of 20 μL of cell suspension was mixed with 200 μL of LM agarose, and then 50 μL of the cell mixture was placed to the sample area of slide. After incubation at 4°C in the dark for 30 minutes, slides were immersed in chilled lysis solution overnight at 4°C and incubated in the chilled neutral electrophoresis buffer for 30 minutes. Following the electrophoresis at 21 V for 75 minutes at 4°C, slides were immersed in DNA precipitation solution for 30 minutes at room temperature and then in 70% EtOH for 30 minutes at room temperature. After drying the samples, slides were incubated with 2.5 μg/mL of propidium iodide in the dark for 30 minutes, and then slides were dried at 37°C. Tail moment of each cell was analyzed by Comet assay IV software (Perceptive Instruments Ltd.).

T47D cells were treated with 10 mmol/L thymidine (Thy) for 18 hours, released for 9 hours after washing three times with PBS, and then blocked again with 10 mmol/L Thy for 22 hours.

Cells were extracted in RIPA lysis buffer (Sigma-Aldrich), supplemented with complete protease inhibitor cocktail tablet (Roche). Following primary antibodies were used for Western blot analysis: anti-TLK2 (Bethyl Laboratories), anti-GAPDH (Santa Cruz Biotechnology). Anti-pChk2 (T68), Chk2, pChk1 (S317), Chk1, pATM (S1981), ATM, pATR (S428), ATR, cyclin B1, pCdk1 (Y15), pH3 (S10), and pAurora kinases antibodies were purchased from Cell Signaling Technology.

From the full-length cDNA of TLK2 (Origene, SC115810), the open reading frame (ORF) was subcloned into an inducible lentiviral pTINDLE vector provided by Dr. Xuewen Pan. This vector contains an inducible promoter (pTRE-tight) and a transactivator (rtTA3) in a lentiviral backbone. We also engineered the ORF of yellow fluorescent protein into the pTINDLE vector as a control. After lentivirus packaging, containing doxycycline-inducible plasmid and infection, the stable lines expressing the TLK2 ORF were selected by treating with Geneticin (Invitrogen). Zero, 100, 200, or 2,000 ng/mL of doxycycline was used to express the TLK2 ORF.

To examine whether TLK2 overexpression correlates with CIN in breast tumors, we calibrated genome-wide genomic instability index (GII) for all TCGA breast tumor samples profiled by Affymetrix SNP 6.0 array. The GII score records the percentage of altered genome and is less affected by the noise of copy number datasets. As a result, we observed a significant positive correlation between TLK2 expression and GII (Spearman correlation: R = 0.393, P < 0.001), suggesting the role of TLK2 in the instability of the breast cancer genome (Fig. 1A, top). Interestingly, this correlation with GII was not observed for TLK1 expression (Spearman R = −0.09, P > 0.9). In addition to GII, we also enumerated the genome-wide copy number breakpoints (CNB, also known as copy number transitions) for TCGA breast tumors as a quantitative measure of the rearrangement events, leading to copy number aberrations (4, 16). The CNB approach will complement the GII approach that does not reflect the complex rearrangement events underlying these copy number alterations. As a result, this revealed a significant association of TLK2 expression with CNB index (Spearman R = 0.384, P < 0.001), but not TLK1 expression (Spearman = −0.067, P > 0.9; Fig. 1A, bottom). These data support the correlation of TLK2 upregulation with increased CIN in breast cancer and its distinct function from TLK1.

To assess the association of TLK2 overexpression with CIN in different breast cancer subtypes, we compared GII index of breast tumors classified by intrinsic breast cancer subtypes in the presence or absence of TLK2 overexpression. Here, the intrinsic subtyping is based on the 50-gene PAM50 predictor (see Materials and Methods; ref. 17). The GII indexes are significantly higher in TLK2-overexpressing luminal B, luminal A, and basal tumor samples as compared with TLK2 low samples (P < 0.001), and to a lesser degree in Her2 subtype samples (P < 0.01; Fig. 1B). This suggests that the increased GII associated with TLK2 overexpression may not be attributed to the enrichment of TLK2 overexpression in the luminal B tumors known to harbor increased CIN (1). To assess whether TLK2 overexpression correlates with other known drivers of CIN in breast cancer, such as AURKA and PLK1 (Supplementary Fig. S1; refs. 7, 18), we charted the status of TLK2, AURKA, and PLK1 overexpression together with chromosome instability scores (Supplementary Fig. S2). The resulting heatmap showed that TLK2 overexpression is independent of AURKA or PLK1 overexpression, suggesting TLK2 as an independent factor in promoting CIN in breast cancer (Supplementary Fig. S2A). In addition, the luminal B breast tumors overexpressing all three kinases (TLK2, AURKA, and PLK1) show a higher GII compared with the tumors overexpressing only one or two kinases, and the luminal B tumors that are negative for all three kinases showed the lowest average GII level (Supplementary Fig. S2B). These observations suggest the role of TLK2 overexpression in promoting CIN of breast cancers.

As TLK kinase activity is quickly inhibited following IR-induced DDR, we hypothesized that TLK2 amplification and overexpression may override the DDR signaling, leading to impaired double-strand break (DSB) repair. We thus assessed the DSB repair process induced by gamma irradiation (IR) in T47D and MCF10A cells inducibly expressing TLK2 using γ-H2AX foci formation assay (γ-H2AX is a biomarker for DSBs; ref. 19). Interestingly, induction of TLK2 overexpression in T47D or MCF10A cells treated with 2 Gy IR prominently delayed the DSB repair process, leading to an increase of DSBs compared with the controls without TLK2 induction, which is sustained for a prolonged period of time (Fig. 2A and Supplementary Fig. S3). This result was corroborated by the neutral comet assay directly visualizing the DSBs in the individual irradiated T47D cells and MCF10A-overexpressing TLK2 (Fig. 2B). In addition, immunofluorescence staining suggests that TLK2 forms nuclear foci in irradiated MCF10A cells, which partially colocalize with γ-H2AX foci (Supplementary Fig. S4). This implies the intimate association of TLK2 with DDR. Moreover, TLK2-driven DSB repair defect seems to be independent of p53 status as TLK2-overexpressing MCF10A (p53 wild type) and T47D (p53 mutant) presented similar DDR alternations (Fig. 2 and Supplementary Fig. S3).

Next, we performed cell-cycle analysis to determine whether T47D and MCF10A cells overexpressing TLK2 have a defect in cell-cycle arrest after DNA damage (Fig. 3A and B). The asynchronized T47D cells inducibly expressing TLK2 were irradiated with 2 Gy IR and collected at different time points until 21 hours. The asynchronized MCF10A cells inducibly expressing TLK2 were treated with different doses of IR and collected after 18 hours. Cell-cycle changes were assessed by flow cytometry measuring DNA content. In both models, a prominently delayed accumulation of G₂–M phase cells was observed after IR when TLK2 was overexpressed, suggesting a possible defect of G₂–M arrest. To verify the G₂–M checkpoint defect, we synchronized the T47D cells at G₁–S border by double thymidine (DT) block. At 5 hours after release from DT block, but before entering G₂–M phase (Fig. 3C and Supplementary Fig. S5), cells were irradiated, and mitotic entry was analyzed by phospho-H3 staining (a mitotic biomarker). This revealed a marked increase in mitotic cells (phospho-H3 percentage) in T47D cells overexpressing TLK2 after irradiation, supporting a G₂–M checkpoint defect (Fig. 3C). As G₂–M checkpoint defect has been linked to increased sensitivity to irradiation and genotoxic agents, as seen in the cancer cells treated with Chk1 inhibitors (20), we postulate that TLK2 overexpression may have a similar effect in breast cancer cells. We thus treated the T47D cells overexpressing TLK2 with irradiation or doxorubicin. Indeed, a modest but significant sensitizing effect was observed for both treatments with TLK2 overexpression, which supports our reasoning (Fig. 4A and B).

Next, we went on to investigate how TLK2 overexpression leads to a G₂–M checkpoint defect. The key step to initiate the mitotic processes is activation of the Cdk1/cyclin B complex by dephosphorylation of Cdk1 on the inhibitory tyrosine pY15 residue (5), which is directly controlled by the G₂–M checkpoint signaling (21). We thus performed Western blot analysis to detect alterations in IR-induced G₂–M checkpoint signaling in T47D cells overexpressing TLK2 (Fig. 4C). Impressively, TLK2 overexpression markedly repressed the phosphorylation of Chk1 and Chk2 in response to IR. In addition, a decrease in pY15 Cdk1 (inactive form) and cyclin B1 level is observed with TLK2 overexpression, followed by IR. Cyclin B1 starts to accumulate in G₂ phase, whereas dephosphorylation of pY15 Cdk1 and activation of Cdk1 is specific to mitotic cells (5). As a vast majority of cells with 4n DNA content (G₂–M population) attributes to G₂ phase, the total cyclin B1 level will be primarily affected by the G₂ cell population (22). Thus, the decrease of pY15 Cdk1 and cyclin B1 implies an increase in the mitotic cell population and a decrease in the G₂ population following TLK2 overexpression. Furthermore, consistent with the G₂–M checkpoint defect, TLK2 overexpression leads to increased phosphorylation and activation of mitotic kinases, such as Aurora A and increased phospho-H3 (Fig. 4C). Together, these data suggest that overexpression of TLK2 may override the DNA damage checkpoint signaling via repressing Chk1/2, leading to G₂–M checkpoint defect and delayed DSB repair. A schematic illustrating the mechanisms of G₂–M checkpoint signaling impaired by TLK2 overexpression is shown in Supplementary Fig. S6.

Our genome-wide GII and CNB analysis of TCGA breast tumor samples revealed that TLK2 amplification might be one of the genetic factors contributing to the outbreak of CIN in the luminal B breast tumors. Interestingly, a most recent phosphoproteomics study of TCGA breast cancers by The Clinical Proteomic Tumor Analysis Consortium (CPTAC) independently identified TLK2 as a highly phosphorylated kinase associated with genomic amplifications at these loci that preferentially present in luminal breast cancer (23). This further supports the importance of TLK2 amplification in luminal tumors. The current study will timely complement the CPTAC study, revealing the role of TLK2 amplification in G₂–M checkpoint defect and CIN. Our experimental data suggest that ectopic expression of TLK2 in the T47D luminal breast cancer cells or MCF10A benign breast epithelial leads to delayed DNA repair, as evidenced by the γ-H2AX foci formation and neutral comet assays. Further cell-cycle analysis after irradiation of T47D or MCF10A cells suggested a G₂–M checkpoint defect associated with TLK2 overexpression, which is further verified via phospho-H3 staining (a mitotic biomarker). These data support the role of TLK2 in G₂–M checkpoint defect and CIN. Presumably, such a G₂–M checkpoint defect will allow unrepaired DSBs to enter into mitosis, leading to accumulation of DNA CNAs as described previously (24). In addition, our mechanistic studies show that TLK2 overexpression in T47D cells potently represses the phosphorylation of Chk1 (S317) and Chk2 (T68), which may explain the impaired G₂–M checkpoint signaling (Fig. 4C and Supplementary Fig. S6). Finally, our result revealed that TLK2 overexpression in T47D cells modestly increases the sensitivity of breast cancer cells to DNA-damaging agents, such as gamma irradiation and doxorubicin (Fig. 4A and B), consistent with the G₂–M checkpoint defect in TLK2-overexpressing cells. Future studies will be required to further examine the effect of TLK2 overexpression on breast cancer cell sensitivity to DNA-damaging agents and the dependence of such an effect on p53 status. In addition, it will be interesting to further examine whether TLK2 overexpression induces CIN during tumor initiation process.

More importantly, our result revealed the distinct functions of TLK2 from TLK1 in DDR. TLK1 has been considered as a guardian of genome integrity (12). As opposed to the CIN driven by TLK2 overexpression, upregulation of TLK1 led to enhanced DNA repair and increased genomic stability (25). In response to DNA damage, TLK1 localizes to DSBs and functions as a molecular chaperone to recruit the Rad9–Hus1–Rad1 (9-1-1) complex (25), which initiates an ATR-Chk1–mediated cell-cycle checkpoint (26). In our result, overexpressed TLK2 in MCF10A was also recruited to DNA damage site after IR (Supplementary Fig. S4), and TLK2 overexpression led to delayed DNA repair, showing 50% to 100% higher number of cells containing more than 10 γ-H2AX foci compared with control after IR (Fig. 2A). As TLK2 shares less homology with TLK1 in the non-STK domains (10), it may be possible that TLK2 may lack the DDR chaperone function possessed by TLK1 and thus competitively inhibit the TLK1 function as a DDR chaperone after DNA damage. Furthermore, a new study just published during our submission has reported TLK2 as a key regulator of checkpoint recovery from DNA damage–induced G₂ arrest (27). Thus, besides repressing IR-induced Chk1/2 activation, TLK2 overexpression may also contribute to the premature mitotic entry via its function in G₂ checkpoint recovery. Future studies are needed to investigate the interaction of the two tousled-like kinases in DDR, and pinpoint the precise mechanisms engaged by TLK2 to repress Chk1/2, and impair G₂–M checkpoint.

Together, our data suggest that TLK2 amplification may contribute to the increased CIN of luminal breast cancer via impairing the G₂–M DNA damage checkpoint. In addition to this observation, our previous study showed that TLK2 overexpression promotes more aggressive phenotypes in luminal breast cancers and correlates with poor prognosis regardless of endocrine therapy. Thus, targeting TLK2 may present an attractive therapeutic strategy for the luminal breast tumors harboring TLK2 amplifications with enhanced aggressiveness and increased CIN.

No potential conflicts of interest were disclosed.

Conception and design: J.-A. Kim, R. Schiff, K. Li, X. Wang

Development of methodology: J.-A. Kim, X. Wang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.-A. Kim, J. Veeraraghavan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.-A. Kim, M. Anurag, K. Li, X. Wang

Writing, review, and/or revision of the manuscript: J.-A. Kim, M. Anurag, R. Schiff, K. Li, X. Wang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.-A. Kim

Study supervision: R. Schiff, X. Wang

The results published here are in part based upon the data generated by TCGA (dbGaP accession: phs000178.v6.p6). We thank Dr. Dean P. Edwards (Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, TX) for providing cell lines from the ATCC ICBP45 cell line panel.

This study was supported by Susan G. Komen foundation PDF12231561 (to J.-A. Kim), CDMRP grants W81XWH-12-1-0166 (to X-S. Wang), W81XWH-12-1-0167 (to R. Schiff), W81XWH-13-1-0431 (to J. Veeraraghavan), and NIH grants CA181368 (to X-S. Wang), CA183976 (to X-S. Wang), and P30-125123-06. The computational infrastructure was supported by Bayer College of Medicine Dan L. Duncan Cancer Center Biostatistics and Informatics Shared Resource (supported by NCIP30 CA125123). This project was also supported by the Cytometry and Cell Sorting Core at the Bayer College of Medicine with funding from the NIH (P30 AI036211, P30 CA125123, and S10 RR024574).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.