In response to DNA damage, several signaling pathways that arrest the cell cycle in G1 and G2 are activated. The down-regulation of mitotic genes contributes to the stable maintenance of the G2 arrest. The human LINC or DREAM complex, together with the B-MYB transcription factor, plays an essential role in the expression of G2-M genes. Here, we show that DNA damage results in the p53-dependent binding of p130 and E2F4 to LINC and the dissociation of B-MYB from LINC. We find that B-MYB fails to dissociate from LINC in p53 mutant cells, that this contributes to increased G2-M gene expression in response to DNA damage in these cells, and, importantly, that B-MYB is required for recovery from the G2 DNA damage checkpoint in p53-negative cells. Reanalysis of microarray expression data sets revealed that high levels of B-MYB correlate with a p53 mutant status and an advanced tumor stage in primary human breast cancer. Taken together, these data suggest that B-MYB/LINC plays an important role in the DNA damage response downstream of p53. [Cancer Res 2009;69(9):4073–80]
In response to DNA damage, signaling pathways that stop the cell cycle and promote repair of damaged DNA are activated. These signaling pathways are triggered by protein kinases ATM and ATR, which activate checkpoint kinases Chk2 and Chk1, respectively (1–3). A major downstream effector of these signaling pathways is the tumor suppressor p53, which induces an arrest in G1 and G2 (4–6). Although G2 arrest is initiated independently on p53, it plays an important role in the maintenance of the G2 arrest (7, 8). The transcriptional down-regulation of mitotic genes by p53 contributes to the long-term arrest in G2. For example, it has been shown that p53 represses the transcription of cyclin B1, CDC2, and CDC25 (9–16). Although some mitotic genes are repressed through direct interaction of p53 with their promoters, p53 seems to mainly repress genes indirectly through activation of the cyclin kinase inhibitor p21 (17, 18). p21 inhibits the activity of cyclin D/cyclin-dependent kinase 4 (CDK4) complexes and, thus, results in hypophosphorylation of retinoblastoma proteins and formation of repressive pocket protein–E2F complexes on E2F-dependent promoters. Consistent with such an indirect mode of repression is the finding that DNA damage–induced inhibition of many G2-M genes depends on E2Fs and on pocket proteins, in particular on the p130 family member (19–22). In tumor cells that lack p53, the G1 checkpoint is selectively lost, making the cells dependent on its ability to arrest in G2.
We and others have recently identified a new multiprotein complex in mammalian cells that plays an important role in the activation of mitotic genes (23–25). This complex, called LINC or human DREAM, consists of a core module that dynamically associates with E2Fs, p130, and the B-MYB transcription factor in a cell cycle–dependent manner (23, 24, 26). In quiescent cells, LINC interacts selectively with p130 and E2F4. The binding of LINC with these proteins is disrupted in the S phase. At this time, B-MYB is incorporated into the complex. LINC plays an important role in the activation of mitotic genes, which is linked to the binding of B-MYB/LINC to the promoters of these genes (24, 25).
In this study, we found that, in response to DNA damage, binding of p130 to LINC is induced and that B-MYB is dissociated from LINC. We show that activation of the p53 > p21 pathway is necessary and sufficient to switch LINC from B-MYB to p130. We find that B-MYB/LINC is responsible, at least in part, for the increase in G2-M gene expression in response to DNA damage in p53 mutant cells and that B-MYB is required for these cells to escape from the G2 checkpoint. Finally, reanalysis of microarray expression data sets revealed that high levels of B-MYB correlate with a p53 mutant status in primary human breast cancer. Taken together, these studies identify a role for LINC and B-MYB in transcriptional regulation in response to DNA damage downstream of p53, and they reveal a dependency of p53-negative cells on B-MYB for checkpoint recovery.
Materials and Methods
Cell culture. HCT-116 wild type (wt), HCT-116 p53−/−, HCT-116 p21−/−, and MCF-7 cells were cultured in DMEM with 10% fetal bovine serum (FBS; Invitrogen), 100 units/mL penicillin, and 100 units/mL streptomycin and maintained at 37°C and 5% CO2.
Drug treatment. Cells were treated with 1 μmol/L doxorubicin (Sigma) for 2 h, washed twice with PBS, and returned to normal growth medium for the indicated time periods.
RNA interference. The small interfering RNA (siRNA) targeting human LIN9 and the control siRNA have been described before (24). The following B-MYB–specific RNA was used: 5′-GCA GAG GAC AGT ATC AAC A-3′. Double-stranded RNA was purchased from MWG or Dharmacon. siRNAs were transfected in a final concentration of 65 to 130 nmol/L using Oligofectamine (Invitrogen) or Metafecten PRO (Biontex).
Propidium iodide staining and flow cytometry. Cells were trypsinized and fixed in 80% ethanol overnight at −20°C. Cells were treated with 100 μg/mL Rnase A in 38 mmol/L sodium citrate for 30 min at 37°C and stained for 10 with 69 μmol/L propidium iodide. Samples were analyzed on a Beckman Coulter FC500.
Identification of mitotic cells. Cells were harvested by trypsination, fixed in 2% paraformaldehyde in PBS for 10 min at 37°C, permeabilized with 90% methanol for 30 min at 4°C, washed in incubation buffer [bovine serum albumin (BSA) in PBS] and incubated with anti–phosphorylated histone H3–Alexa 488 (Cell Signaling) for 2 h at room temperature. Cells were then incubated with 75 μmol/L propidium iodide and 700 μg/mL Rnase A for 30 min at 37°C and analyzed by fluorescence-activated cell sorting (FACS) to determine the fraction of phosphorylated histone H3–positive cells.
Whole-cell lysates, immunoblotting, immunoprecipitation, and antibodies. Cells were lysed in TNN [50 mmol/L Tris (pH 7.5), 120 mmol/L NaCl, 5 mmol/L EDTA, 0.5% NP40, 10 mmol/L Na4P2O7, 2 mmol/L Na3VO4, 100 mmol/L NaF, 10 μg/mL phenylmethylsulfonyl fluoride, protease inhibitors (Sigma)]. Whole-cell lysates were immunoprecipitated with antibodies overnight at 4°C, bound to protein A or G–Sepharose for 2 h and washed five times with TNN lysis buffer. Proteins were detected by immunoblotting. α-LIN9, α-LIN37, α-LIN52, and α-LIN54 antisera were described before (24, 27). The B-MYB antibody used in Fig. 4 was a kind gift from Roger Watson. Other antibodies used were α-E2F4 (A-20 and C-20), α-p107 (C-18), α-p130 (C-20), and α-B-MYB (N-19 and H-115) from Santa Cruz and anti–β-tubulin (MAB3408) from Chemicon. Nonspecific IgG was from Sigma.
Chromatin immunoprecipitation. Cells were cross-linked with 1% formaldehyde for 10 min at room temperature. The reaction was stopped by adding 125 mmol/L glycine. Cells were lysed for 10 min in lysis buffer [5 mmol/L PIPES (pH 8.0), 85 mmol/L KCl, 0.5% NP40, protease inhibitors (Sigma)]. Nuclei were lysed in nuclei lysis buffer [50 mmol/L Tris (pH 8.1), 10 mmol/L EDTA, 1% SDS, protease inhibitors]. Chromatin was sonicated to an approximate length of 500 to 750 bp, diluted at 1:10 with dilution buffer [0.01% SDS, 1.1% Triton, 1.2 mmol/L EDTA, 16.7 mmol/L Tris (pH 8.2), 167 mmol/L NaCl, protease inhibitors], precleared with protein A–Sepharose for 1 h (blocked with 1 mg/mL BSA and 0.3 mg/mL ssDNA), and used for immunoprecipitation overnight. Beads were washed seven times with LiCl washing buffer [0.25 mmol/L LiCl, 0.5% NP40, 0.5% sodium deoxycholate, 1 mmol/L EDTA, 10 mmol/L Tris (pH 8.0), protease inhibitors] and eluted with elution buffer [50 mmol/L Tris (pH 8.0), 1% SDS, 10 mmol/L EDTA]. The cross-link was reversed overnight with 0.2 mol/L NaCl at 65°C. After proteinase K incubation for 3 h at 55°C, chromatin was purified using Qiagen DNA purification spin columns. Chromatin (1 μL) was used as template for quantitative real-time PCR, as described before (25).
RNA isolation, reverse transcription, and quantitative real-time PCR. Total RNA was isolated with Total RNA Isolation Reagent (Thermo Scientific). RNA (1–2 μg) was transcribed using 100 units of M-MLV-RT (Promega) or 125 units of MMuLv (Thermo Scientific). Quantitative real-time PCR reagents were from Thermo Scientific, and real-time PCR was performed using the Mx3000 (Stratagene) detection system. Expression differences were calculated as described before (25).
To analyze the composition of LINC after induction of DNA damage, we performed immunoprecipitation with antibodies directed at LIN9, a core component of LINC, and analyzed the binding of B-MYB and p130 by immunoblotting. DNA damage was induced by treatment of MCF-7 cells with the chemotherapeutic drug doxorubicin. Doxorubicin treatment induced binding of p130 to LIN9 and resulted in dissociation of B-MYB from LIN9 (Fig. 1A) Binding of LIN9 to LIN37 and LIN54, two components of the LINC core module, did not change by doxorubicin treatment, suggesting that the LINC core complex does not dissociate in response to DNA damage (Fig. 1B). Binding of p130 to LIN9 was already detected at 6 hours after addition of doxorubicin and gradually increased over the course of the experiment. In contrast, significant changes in cell cycle distribution were first observed at 18 hours after addition of doxorubicin. This suggests that binding of p130 to LINC following DNA damage is not simply an indirect effect of cell cycle inhibition (Fig. 1C). Because at 6 and 9 hours after induction of DNA damage, both B-MYB and p130 coprecipitated with LIN9, it is possible that B-MYB and p130 simultaneously bind to the LINC core module at these time points. To address this, we asked whether p130 coprecipitates with B-MYB (Fig. 1D). p130 was not detected in anti–B-MYB immunoprecipitations at any time point, indicating that binding of B-MYB and p130 to the LINC core complex is mutually exclusive and that a LINC complex that contains both p130 and B-MYB does not exist.
We next investigated the possibility that dephosphorylation of p130 induces the rearrangement of LINC upon DNA damage. Immunoprecipitation experiments showed that LIN9 selectively binds to the faster migrating hypophosphorylated form of p130 present after DNA damage, but not to hyperphosphosphorlyated p130 present in untreated cells (Figs. 1 and 2A). Because DNA damage activates p53, which in turn induces the cdk inhibitor p21, the p53 > p21 pathway could regulate the composition of LINC. Consistent with this notion, p21 induction at 6 hours after doxorubicin treatment correlated with p130 dephosphorylation and binding of p130 to LINC (Fig. 1A).
To directly test whether p21 is necessary for the DNA damage–induced association of p130 with LINC, we used HCT-116 p21−/− cells, in which both copies of the p21 gene have been disrupted by homologous recombination (32). As a control, isogenic HCT-116 p21+/+ cells were used. As was seen before in MCF-7 cells, p130 associated with LIN9 in HCT-116 p21+/+ cells after treatment with doxorubicin (Fig. 2B). In contrast, p130 failed to associate with LIN9 in HCT-116 p21−/− cells, indicating that p21 is necessary for the DNA damage–induced binding of p130 to LINC.
Because p21 is induced by p53 in response to DNA damage, we next asked whether activation of p53 in the absence of DNA damage switches the composition of LINC. To directly activate p53, we treated cells with the HDM2 antagonist Nutlin-3. Nutlin-3 induced binding of p130 to LIN9 and disrupted binding between LIN9 and B-MYB (Fig. 2C). This indicates that activation of p53 is sufficient to switch LINC from B-MYB to p130.
Next, we used p53-deficient HCT-116 cells to test whether reorganization of LINC after DNA damage is also dependent on p53 (7). Indeed, in HCT-116 p53−/− cells, binding of p130 to LIN9 was not induced by doxorubicin treatment (Fig. 2D). Interestingly, binding of B-MYB to LIN9 was not decreased but even further increased upon doxorubicin treatment compared with untreated cells. Taken together, these data indicate that p53 is necessary and sufficient to switch LINC from B-MYB to p130.
Although DNA damage triggers a rapid p53-independent pathway, which leads to G2 arrest, p53 is required to maintain the G2 arrest until the DNA repair has been completed. Previous studies have shown that p53-dependent down-regulation of mitotic genes contributes to a stable G2 arrest. Because LINC plays an important role in the expression of mitotic genes in the cell cycle, we therefore asked whether LINC is involved in the transcriptional repression of G2-M genes in response to p53 activation.
To address this question, we first compared G2-M gene expression in HCT-116 p53+/+ and p53−/− cells in response to DNA damage. In HCT-116 p53+/+ cells, addition of doxorubicin inhibited the expression of cyclin B1, CDC2, BIRC5, and UBCH10, as expected (Fig. 3A and Supplementary Fig. S2). In p53−/− cells, basal levels of these G2-M genes were already elevated compared with wt cells and their expression further increased upon addition of doxorubicin (Fig. 3A). As a control, we analyzed proliferating cell nuclear antigen (PCNA), which is an E2F-dependent gene activated at G1-S, and found that it was not inhibited in p53+/+ cells and that its expression did not significantly differ between HCT-116 wt and p53−/− cells.
Next, we analyzed binding of B-MYB, E2F4, p130, and LIN9 to the promoters of G2-M genes by chromatin immunoprecipitation. Chromatin was isolated from control or doxorubicin-treated cells and immunoprecipitated with antibodies specific for E2F4, B-MYB, p130, and LIN9. As a control, we used nonspecific IgG. Promoters were detected by quantitative real-time PCR. In HCT-116 wt cells, doxorubicin induced binding of E2F4 and p130 to G2-M promoters (Fig. 3B and Supplementary Fig. S2). LIN9 binding was not affected by doxorubicin treatment, consistent with the presence of LIN9 in repressive and activating LINC complexes. Binding of B-MYB to G2-M promoters was reduced 48 hours after treatment. A strikingly different pattern was observed in HCT-116 p53−/− cells, wherein binding of E2F4 to G2-M promoters was already reduced compared with p53+/+ cells and further decreased upon DNA damage (Fig. 3B and Supplementary Fig. S2). p130 also failed to associate with G2-M promoters in p53−/− cells in response to DNA damage. Furthermore, B-MYB did not dissociate from G2-M promoters in p53−/− cells upon doxorubicin treatment (Fig. 3B). Importantly, B-MYB, E2F4, p130, and LIN9 were not recruited to the glyceraldehyde-3-phosphate dehydrogenase promoter, showing the specificity of the chromatin immunoprecipitation assays (Supplementary Fig. S2).
Similar changes in binding of E2F4, p130, and B-MYB to G2-M promoters were observed upon induction of DNA damage by doxorubicin in primary human BJ and MCF-7 cells and p53 activation in MCF-7 cells (Supplementary Figs. S3–S5). This indicates that the changes in promoter binding of LINC are not restricted to one cell type and that they are not limited to DNA damage–induced cell cycle inhibition at G2. Together, these observations are consistent with a role for LINC in the down-regulation of G2-M genes in response to p53 activation.
Decreased binding of E2F4 and the failure of B-MYB to leave G2-M promoters after the induction of DNA damage in p53−/− cells correlate with increased expression of G2-M genes in these cells. To directly test whether B-MYB is required for hyperactivation of G2-M genes in p53-deficient cells, we depleted B-MYB by RNA interference (RNAi) and then analyzed gene expression before and after treatment with doxorubicin. Doxorubicin inhibited G2-M gene expression in control-depleted HCT-116 wt cells, as expected (Fig. 4A and Supplementary Fig. S6). Expression of G2-M genes was reduced in B-MYB depleted cells and further decreased by doxorubicin treatment. Depletion of B-MYB in p53−/− cells reduced basal levels of G2-M gene expression and significantly attenuated the induction of G2-M genes by doxorubicin. Together, these observations suggest that B-MYB contributes to increased expression of G2-M genes in p53−/− cells in response to DNA damage.
Next, to address whether activation of G2-M genes by B-MYB in p53−/− cells is mediated by the LINC complex or whether it is an isolated function of B-MYB, we depleted LIN9 by RNAi. Similar to B-MYB, depletion of LIN9 also significantly reduced the activation of G2-M genes by doxorubicin (Fig. 4B and Supplementary Fig. S6). As a control we analyzed the expression of PCNA, which was unchanged by the depletion of LIN9. Codepletion of LIN9 and B-MYB resulted in a similar reduction in G2-M gene expression compared with the depletion of LIN9 alone, suggesting that B-MYB and LIN9 function in the same pathway (Fig. 4B). B-MYB or LIN9 depletion also inhibited the DNA damage–mediated induction of CDC2 and cyclin B1 in p53−/− cells on the protein level (Fig. 4C). Taken together, these observations suggest that increased expression of G2-M genes in p53−/− cells is mediated by the B-MYB/LINC complex. The finding that mitotic genes are still induced, albeit at lower levels in B-MYB and LIN9 depleted cells, suggests that there are other factors that also contribute to the expression of mitotic genes in p53−/− cells.
To substantiate the relevance of these observations, we analyzed the role of B-MYB in the restart of the cell cycle after DNA damage. It has been shown that p53 is required to prevent entry into mitosis after the induction of DNA damage by γ radiation (7). To test whether p53 is also necessary to prevent entry into mitosis after treatment with doxorubicin, we treated HCT-116 p53+/+ and HCT-116 p53−/− cells with doxorubicin for 2 hours and then analyzed the fraction of cells that entered into mitosis by staining with an antibody specific for phosphorylated histone H3. Indeed, a fraction of HCT-116 p53−/− cells, but not wt cells, entered into mitosis between 48 and 72 hours after DNA damage (Fig. 5A). We then asked whether the cell cycle reentry of p53−/− cells after DNA damage requires B-MYB and/or LIN9. To test this, we depleted B-MYB or LIN9 in HCT-116 p53−/− cells by RNAi and treated them with doxorubicin for 2 hours. As a control, HCT-116 p53−/− cells transfected with the control siRNA were used. As expected, a fraction of HCT-116 p53−/− cells transfected with the control siRNA entered into mitosis after doxorubicin treatment (Fig. 5B). In striking contrast, B-MYB and LIN9 depleted p53−/− cells failed to enter into mitosis. Taken together, in p53−/− cells, B-MYB and LIN9 are required for entry into mitosis after DNA damage, which is consistent with a role for B-MYB/LINC in the activation of mitotic genes in these cells.
To substantiate the relevance of these observations, we investigated the correlation between the p53 status and B-MYB expression in published cancer microarray data sets.
B-MYB is one of the small number of genes that has consistently reported to be up-regulated in breast cancer (30, 33–35). High levels of B-MYB can predict the probability of breast cancer recurrence (33, 35). B-MYB is also included in a set of 374 genes that were extracted from 44 published gene lists and are significantly associated with breast cancer recurrence (36). Furthermore, B-MYB is one of the marker genes of a commercially marketed 21-gene assay that can predict recurrence in estrogen receptor–positive, node-negative patients (37). We therefore examined three published breast cancer microarray data sets, in which the p53 status and B-MYB have been reported (28–30). First, the expression of B-MYB was plotted against the p53 status in these data sets. As shown in Fig. 6A, high levels of B-MYB are consistently associated with mutated p53. Furthermore, high levels of B-MYB correlate with an advanced disease stage, consistent with previous reports (Fig. 6B). Importantly, we found a strong correlation between the expression of B-MYB, mutation of p53, and the expression of mitotic B-MYB target genes cyclin B1 and CDC2 (Fig. 6C). In contrast, expression of dihydrofolate deductase (DHFR), a G1-S regulated gene, was not strongly correlated with B-MYB expression, and importantly, DHFR expression was not correlated with the p53 status in this data set.
LINC or human DREAM is a recently described multiprotein complex that plays an important role in the transcriptional regulation of G2-M genes (23, 25, 26). In this study, we show that binding of the retinoblastoma protein p130 to LINC is induced by DNA damage. An opposite binding pattern was observed for B-MYB whose association to LINC is disrupted after DNA damage. This switch in LINC composition plays a role in transcriptional repression of G2-M genes in response to DNA damage (see model in Supplementary Fig. S7). A causal role for LINC in the DNA damage response is supported by previous studies, which showed a role for pocket proteins in the DNA damage–induced cell cycle arrest (19, 21, 22).
We find that p53 is both necessary and sufficient to switch LINC from B-MYB to p130. p53 induces p21, which inhibits the activity of CDKs and thus leads to accumulation of dephosphorylated p130. Together with the observation that only hypophosphorylated p130 binds to LINC, this suggests that association of p130 with LINC is regulated by the phosphorylation status of p130. It is possible that hypophosphorylated p130 displaces B-MYB from the LINC core complex. Consistent with this notion, a complex that simultaneously contains B-MYB and p130 cannot be detected.
Like DNA damage, p53 activation by Nutlin-3, which primarily arrests cells in G1, also switches the LINC complex and results in dissociation of B-MYB from G2-M promoters and in recruitment of p130/E2F4. This suggests that LINC is not exclusively involved in the G2-M checkpoint response and that it also contributes to the repression of G2-M genes in G1 arrested cells. LINC-mediated p53-dependent repression of G2-M genes could contribute to a stable cell cycle arrest (Supplementary Fig. S7).
The pathways leading to p21 induction and p130 dephosphorylation are not activated in p53 mutant cells. Consequently, p130 fails to associate with LINC in these cells upon DNA damage. Instead, we find that high levels of B-MYB/LIN9 accumulate in p53 mutant cells in response to DNA damage and that this promotes mitotic entry after activation of the G2 checkpoint (Supplementary Fig. S7). p53-negative cells cannot arrest in the G1 phase of the cell cycle and thus show a stronger dependence on G2 DNA damage checkpoint. However, the G2 checkpoint of p53 mutant cells is weakened, and a fraction of these cells is able to enter into mitosis after a genotoxic insult (Fig. 5; ref. 7). We find that this checkpoint recovery of p53−/− cells is mediated, at least in part, by B-MYB. The finding that higher levels of B-MYB bind to mitotic promoters in p53−/− cells than in p53+/+ cells suggests that the ability of B-MYB to promote checkpoint recovery of p53-negative cells is directly linked to its ability to up-regulate the transcription of mitotic genes. This is supported by the finding that activation of G2-M genes in p53−/− cells is partially inhibited by the depletion of B-MYB. LIN9 depletion also prevents mitotic entry of p53 mutant cells in response to DNA damage. Thus, the ability of B-MYB to promote checkpoint recovery seems to be linked to the activation of mitotic genes as opposed to a transcription-independent function of B-MYB, such as the recently reported mitotic function of the B-MYB-Clafi complex (38). The concept that B-MYB promotes checkpoint recovery through increased mitotic transcription is also supported by the finding that high levels of cyclin B1, a B-MYB target gene, shortens the G2 checkpoint in HeLa cells (39, 40). Two other B-MYB target genes, Polo-like kinase 1 and Aurora A, have also recently been implicated in checkpoint adaptation (41, 42).
Interestingly, it has recently been shown that chicken DT40 cells lacking B-MYB are more sensitive to DNA damage (43). Because DT40 cells do not express p53 (44), it is possible that, similar to what we observed in human cells, these cells are especially dependent on B-MYB to recover from G2 checkpoint and to survive DNA damage.
Our finding that B-MYB promotes checkpoint recovery in p53 mutant cells suggests that therapeutic agents that inhibit the expression or function of B-MYB may be clinically relevant in a p53 mutant background. Increased expression of B-MYB has been found in different types of human cancers. For example, B-MYB overexpression is associated with a poor prognosis in neuroblastoma (45). B-MYB is also one of the small number of genes that has consistently reported to be up-regulated in breast cancer (30, 33–35). Our reanalysis of human breast cancer microarray data sets shows that high levels of B-MYB strongly correlate with a p53 mutant status (Fig. 6). Based on our findings, it is possible that overexpression of B-MYB in these tumors allows cell cycle restart and proliferation of cells with damaged DNA. This could contribute to their genomic instability and tumor progression. Inhibition of B-MYB in p53-negative tumors might block checkpoint recovery and thus lead to a more efficient G2 arrest upon treatment with DNA-damaging agents. Thus, treatment of patients with agents that inhibit B-MYB may be effective as adjuvant therapy.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
Grant support: Sander Foundation and DFG (TR17) grant (S. Gaubatz).
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.
We thank Roger Watson, Matthias Dobbelstein, and Thorsten Stiewe for reagents and discussions; Susi Spahr for technical help; Leona Probst for her assistance with HCT-116 p21−/− cells; and all the members of the laboratory for their suggestions and critical reading of the manuscript.