Inhibition of Protein Synthesis Induced by CHK1 Inhibitors Discriminates Sensitive from Resistant Cancer Cells

Abstract

The DNA-damage-activated checkpoint protein CHK1 is required to prevent replication or mitosis in the presence of unrepaired DNA damage. Inhibitors of CHK1 (CHK1i) circumvent this checkpoint and enhance cell killing by DNA-damaging drugs. CHK1i also elicit single-agent cytotoxicity in a small subset of cell lines. Resolving the mechanisms underlying the single-agent activity may permit patient stratification and targeted therapy against sensitive tumors. Our recent comparison of three CHK1i demonstrated that they all inhibited protein synthesis only in sensitive cells. LY2606368, the most selective of these CHK1i, was used in the current study. Comparison across a panel of cell lines demonstrated that sensitive cells died upon incubation with LY2606368, whereas resistant cells underwent growth inhibition and/or cytostasis but failed to die. Sensitive cells exhibited inhibition of protein synthesis, elevated DNA damage, impaired DNA repair, and subsequently death. The consequence of CHK1 inhibition involved activation of cyclin A/CDK2 and MUS81, resulting in DNA damage. This damage led to activation of AMPK, dephosphorylation of 4E-BP1, and inhibition of protein synthesis. Inhibition of MUS81 prevented activation of AMPK, while inhibition of AMPK enhanced DNA repair and cell survival. The activation of AMPK may involve a combination of LKB1 and CaMKKβ. This study raises questions concerning the potential importance of the inhibition of protein synthesis in response to other drugs, alone or in combination with CHK1i. It also highlights the importance of clearly discriminating among growth inhibition, cytostasis, and cell death, as only the latter is likely to result in tumor regression.

Inhibitors of checkpoint kinase 1 (CHK1) are well-established to sensitize cells when used in combination with traditional chemotherapeutics. Many anticancer drugs induce DNA damage by causing DNA lesions resulting in single- or double-stranded breaks (DSBs) or by depleting deoxyribonucleotides (dNTPs), thereby stalling replication. The ATR–CHK1 axis is activated in each of these scenarios by the single-stranded DNA formed either by resection at the breaks or when replication forks stall as helicase continues to unwind DNA in the absence of ongoing synthesis. Activation of ATR–CHK1 triggers cell cycle checkpoints that arrest cells, allowing time to repair damaged DNA and thus thwarting the intended therapeutic action of the anticancer drug. Inhibitors of CHK1 (CHK1i) effectively abrogate this arrest by reinitiating replication on damaged DNA, even in the absence of dNTPS and before repair can be completed, resulting in replication or mitotic catastrophe.^1,2

Importantly, CHK1i are also effective when used alone in a subset of cell lines and xenografts. This represents a promising therapeutic option if patients can be appropriately stratified, as it may offer a targeted therapy against sensitive tumors. Approximately 15% of cell lines in a human-tumor derived cell line panel were shown to be acutely sensitive to single-agent MK-8776,³ and we recently obtained similar data for two additional CHK1i (SRA737 and LY2606368).⁴ The mechanism of action of single-agent CHK1i has been attributed to activation of cyclin-dependent kinase 2 (CDK2). Excessive activation of CDK2 in complex with cyclin A in S phase cells appears to be the primary driver of this sensitivity,^2,3,5 resulting in activity of the nucleases MRE11 and MUS81 followed by DSBs.^6,7 However, the precise molecular mechanisms by which aberrant CDK2/cyclin A activation leads to DNA damage and death have yet to be elucidated. Further characterization of monotherapy CHK1i activity is critical to inform the use of CHK1i as targeted therapies in patients.

In our recent comparison of MK-8776, SRA737, and LY2606368 (prexasertib), we made the novel discovery that all three of these CHK1i led to inhibition of protein synthesis (IPS).⁴ In the case of LY2606368, one explanation could be its reported inhibition of RSK1,⁸ but this is not a target for SRA737 or MK-8776 suggesting an alternate mechanism for inhibition of protein synthesis might be involved. We also observed an unexpected off-target effect for MK-8776 and SRA37 at slightly higher concentrations which limited analysis of these drugs in resistant cells.⁴ This off-target effect appeared to involve inhibition of CDK2. Consequently, we selected LY2606368 to further investigate the impact of CHK1i on protein translation.

The reprogramming of translational machinery is a tightly controlled response to various forms of cellular stress, including DNA damage, and can often enhance cell survival.⁹ Chemical-induced IPS can lead to cell death,¹⁰ suggesting that pathways to enhance survival might be circumvented. Here, we investigated the relationship between cell death, DNA damage, and IPS in response to CHK1i. Strikingly, we observe that while both sensitive and resistant cell lines inhibit CHK1 and sustain DNA damage, IPS and cell death only occur in sensitive cells. We find that CHK1i-induced IPS is driven by cyclin A/CDK2, DNA damage, and AMPK activation. The amount of DNA damage observed appears to be a reflection of ongoing DNA repair, which is prevented once cells activate AMPK and inhibit protein synthesis, and as a consequence cells die. Resistant cells fail to activate AMPK even in the presence of DNA damage and hence survive. The critical regulators of translation in this model remain to be determined. The impact of our current findings on the use of CHK1i as monotherapy as well as in potential targeted combination treatments is discussed.

Results

LY2606368 Induces Cell Death in Sensitive Cells but Only Growth Inhibition in Resistant Cells

The majority of prior experiments with CHK1i, including ours, have addressed their ability to inhibit cell growth.^3−5,11 Many such assays are misleadingly called a “viability assay” despite showing no evidence of cell death (see critique in ref (12)). In our recent comparison of three CHK1i, we used a modified growth inhibition assay to specifically address the ability of each compound to induce cell death.⁴ This was achieved by increasing the initial cell density such that a decrease in total cell number could be reliably quantified. Total DNA was used as a measure of cell number, and a decrease below the starting inoculum was an indication of cell death. All three CHK1i killed sensitive AsPC-1 cells but not resistant SW620 cells. For example, 24 h of treatment with 25 nM LY2606368 killed AsPC-1 cells over the following several days, whereas 500 nM LY2606368 only induced cytostasis in SW620 cells even when incubated with CHK1i continuously for 8 days.⁴ This difference in sensitivity is far greater than observed when using a traditional growth inhibition assay, and is likely far more relevant to the differential that might be observed in a patient.

Here, we assessed growth and death in several additional cell lines. Each cell line was incubated with 0–1000 nM LY2606368 for 24 h, then the drug was washed out and cells allowed to recover in complete medium for the next 5 days (Figure 1, blue lines). Alternatively, the drug was left on cells for the entire 6 days (Figure 1, red lines). After harvest, wells were scored for DNA content as a surrogate for cell number. In the most sensitive cell lines, LY2606368 consistently induced cell death following the 24 h of treatment at concentrations ranging from 10 to 30 nM. Conversely, the majority of resistant cell lines showed no cell death regardless of incubation time or drug concentration. While some of these resistant lines initially showed inhibition of growth (Figure 2), others appeared to continue growing for 6 days despite the presence of high concentrations of drug. MDA-MB-231 is an example of a cell line that exhibited intermediate toxicity, cell death occurring during continuous 6 days of treatment with >60 nM LY2606368, or after 24 h of treatment with >500 nM LY2606368, both concentrations far in excess of the 10–30 nM required in sensitive cells. The results in Figure 1 also include a derivative of MDA-MB-231 deleted for ATM and studied in our prior paper on sensitivity to ATR inhibitors,⁵ although this deletion had little, if any, impact on sensitivity to LY2606368. The relative sensitivity of these cell lines is consistent with our prior papers using the more traditional growth inhibition assay.^3,4

Figure 1.

Differential sensitivity of cell lines to LY2606368-mediated growth inhibition and death. Cells were incubated with LY2606368 for 24 h, then without drug for the following 5 days (blue line). Alternately, cells were incubated continuously with LY2606368 for 6 days (red line). Wells were scored for total DNA at harvest. One plate was harvested at time zero (0 on y-axis) so that the increase or decrease in cell number could be determined. Values between 0 and −100 reflect percent death. Points above 0 reflect percent growth, the magnitude of which differed between cell lines.

Figure 2.

Kinetics of growth inhibition or cell death. Cells were incubated for 24 h with 0–100 nM LY2606368 (A) or 0–1000 nM LY2606368 (B), then harvested on days 0–8. Wells were scored for total DNA at harvest, and data were graphed so that the increase or decrease in cell number compared to day 0 is visualized. The lower panel for each cell line is a reanalysis of the data to emphasize the extent of inhibition or recovery observed each day.

We also investigated the time course for cell growth and death in several of the cell lines. In the sensitive U2OS and AsPC-1 cells, a decrease in cell numbers compared to those at time zero was first observed around 48 h (24 h after drug was removed) at 6 nM (Figure 2A, upper panel). In contrast, in the MDA-MB-231 cells, the decrease in cell number at high concentrations of drug was not observed until day 4 or longer (Figure 2B). SW620 cells again only demonstrated cytostasis at the highest concentrations.

The data in Figure 2 were reanalyzed to emphasize the ability of cells to recover after a 24 h incubation with LY2606368 (Figure 2, lower panel). For each time point, the data was recalculated such that untreated cells were expressed as 100% growth, and the drug-treated cells were expressed as a percent of growth or death. In the U2OS cells, about 50% of the cells had died by day 2 after 6 nM, while the remainder appeared to recover by 6–8 days, but after 25 nM, no recovery was observed. Very similar results were obtained in the AsPC-1 cells. In contrast, in MDA-MB-231 and SW620 cells, growth was still inhibited at fairly low concentrations of LY2606368 after 24 h, but with no evidence of cell death, cells were able to recover after most concentrations of the drug.

Figure 2 also shows an analysis of UACC62 cells, the most resistant in our prior analysis, as well as several derivatives of AsPC-1 with acquired resistance to LY2606368 (resistance on the order of 200- to 300-fold in the growth inhibition assay).⁴ These lines showed little growth inhibition at all concentrations. The rationale for developing and testing an LKB1 knockout derivative of AsPC-1 cells is discussed below.

LY2606368-Induced Cell Death Correlates with IPS Rather than CHK1 Inhibition

Previous work established that LY2606368 inhibits CHK1 at comparable concentrations in both sensitive and resistant cell lines and that DNA damage can also occur regardless of sensitivity, albeit a longer incubation may be required in the resistant cells.⁴ To expand these findings, we incubated two sensitive cell lines (AsPC-1 and U2OS) and two relatively resistant cell lines (MDA-MB-231 and SW620) with a concentration range of LY2606368 and assessed the impact on CHK1 inhibition, DNA damage, and IPS (Figure 3A). CHK1 inhibition was measured as a decrease in the autophosphorylation of CHK1 at ser-296, and DNA damage was assessed by phosphorylation of histone 2AX at ser-139 (γH2AX). As previously reported,⁴ CHK1 was inhibited at similar concentrations in all four cell lines, independent of sensitivity to LY2606368. In the sensitive cells, the inhibition of CHK1 correlated closely with the appearance of γH2AX, whereas in the resistant cells, the onset of γH2AX required about 10-fold higher concentrations.

Figure 3.

Inhibition of CHK1 and protein synthesis by LY2606368. (A) Four cell lines were incubated with 0–1000 nM LY2606368 for 24 h, with 1 μM puromycin added during the final hour. Cell lysates were analyzed by Western blotting. Multiplexed fluorescent images are shown for CHK1 and pCHK1. (B) The amount of puromycin incorporated in protein in panel A, together with 4 additional cell lines, was quantified from the fluorescent images. (C) AsPC-1 and SW620 cells were incubated with 30 or 1000 nM LY2606368, then harvested from 0–24 h, with puromycin added during the final hour. Lysates were analyzed as in A.

Protein synthesis was assessed using puromycin incorporation during the final hour before harvest as previously described.⁴ Concurrently, we assessed the phosphorylation status of the translational repressor 4E-BP1. Briefly, phosphorylation causes 4E-BP1 to dissociate from eukaryotic initiation factor eIF4E, allowing it to complex with other eIFs to form the initiation complex eIF4F. Conversely, dephosphorylated 4E-BP1 remains bound to eIF4E; thus, dephosphorylated 4E-BP1 should result in IPS. The switch between phosphorylated and dephosphorylated 4E-BP1 is detected as a mobility shift of total 4E-BP1 protein by Western blot. In the sensitive cells, almost complete loss of puromycin incorporation was observed at the same concentrations that inhibited CHK1 and induced γH2AX (Figure 3A). These same concentrations were associated with dephosphorylation of 4E-BP1, suggesting that these observations are an on-target consequence of LY2606368. This is also supported by our prior observations that three different CHK1i induced IPS.⁴ In contrast, little dephosphorylation of 4E-BP1 or IPS was observed in the resistant cells, even at concentrations that induced γH2AX. The amount of puromycin incorporation was quantified in these and four additional resistant cell lines, as summarized in Figure 3B, which emphasizes the difference between IPS in the sensitive versus resistant cells. As discussed further below, phosphorylation of 5′ AMP-activated protein kinase (AMPK) was also assessed as it is a repressor of mTOR and 4E-BP1, and this correlated with DNA damage and IPS in the sensitive and intermediate cells but not in the resistant SW620 cells (Figure 3A).

We next assessed the kinetics of DNA damage and IPS. In sensitive AsPC-1 cells, DNA damage was observed at 6 h, with more at 12 h, during incubation with both 30 and 1000 nM LY2606368 (Figure 3C). IPS was delayed until 18 h in both cases, as was the dephosphorylation of 4E-BP1. These data indicate that DNA damage occurs before IPS, but both occur before a decrease in cell number, indicative of cell death as observed above (Figure 2). The flow cytometry experiments discussed below also showed no evidence of cell death at 24 h (Figure 4). In the resistant SW620 cells, DNA damage was pronounced by 12 h (by 6 h at 1000 nM), but again no dephosphorylation of 4E-BP1 was observed, and only slight IPS at the highest concentration. The results in these resistant cells suggests that while DNA damage may be required it is not sufficient to inhibit protein synthesis.

Figure 4.

Induction of γH2AX by LY2606368. (A) AsPC-1 cells and a resistant derivative, AsPC-LYR(2020), were incubated with LY2606368 for 24 h, then analyzed by flow cytometry for cell cycle perturbation and γH2AX. The values reflect the percent cells gated for γH2AX and are graphed in panel B, along with values for additional cell lines (raw data for these lines in Figure S1). The values for MDA-MB-231 and SW620 are derived from our prior paper.⁴

Formation and Repair of DNA Damage

A more detailed analysis of DNA damage was performed by flow cytometry. Following a 24 h incubation with LY2606368, AsPC-1 and U2OS cells showed dramatic accumulation in S phase (up to 95%, and earlier in S phase at high concentrations), and all these cells were positive for γH2AX (Figures 4 and S1). In other cell lines, a higher concentration of LY2606368 was required, consistent with their degree of resistance, but the γH2AX still appeared predominantly in S phase.

These results raise the question as to whether less DNA damage is induced in the resistant cells, or whether they are better able to tolerate it through repair of the lesions. The inhibition of protein synthesis in the sensitive cells would logically impede DNA repair resulting in higher levels of damage. To begin to dissect this possibility, we incubated cells for 1–3 days in complete growth medium after removing LY2606368 (Figures 5 and S2). In the sensitive cells, there was only a small decrease in the number of cells positive for γH2AX over time, and the cells with decreased γH2AX still remained arrested in the S phase. There was an increasing number of cells with sub-G1 DNA content at later time points, and fewer cells were available for analysis. This is consistent with increased cell death as recorded in Figure 2, and most of the dead cells were likely gated out of the flow profiles.

Figure 5.

Recovery of cells from DNA damage induced by LY2606368. (A) AsPC-1 cells were incubated with 10–100 nM LY2606368 for 24 h, while the resistant derivative, AsPC-LYR(2020), were incubated with 100–1000 nM LY2606368. Some cells were harvested at 24 h (day 1). For the remainder, the media was replaced, and the cells were harvested on days 2–4. Cells were evaluated for cell cycle perturbation and γH2AX. The values reflect the percent cells gated for γH2AX and are graphed in panel B, along with comparable values for additional cell lines (flow data for these lines in Figure S2).

Similar experiments in the more resistant cells showed a rapid reduction in the number of cells positive for γH2AX 24 h after removal of drug, and in the most resistant cells (UACC62), this repair occurred even after incubation at the highest concentration (1 μM) (Figures 5 and S2). The rate of repair was not solely dependent on the percent positive at 24 h, as some of the cell lines exhibited 40–65% positivity at 24 h yet exhibited extensive repair by 48 h. In addition, any S or G2 arrest observed at 24 h rapidly disappeared. This apparent rate of repair cannot be explained simply on the basis of overgrowth by nondamaged cells, as the decrease in positive cells occurs much faster than the rate of cell doubling.

CDK2/Cyclin A Activity Drives Both DNA Damage and Cell Death

Given the emerging correlation between IPS and cell death, we next sought to define the pathway connecting CHK1 inhibition to IPS. Inhibition of CHK1 activates the CDC25A phosphatase and in turn CDK2, and aberrant CDK2 activity is responsible for the monotherapy activity of CHK1i.³ In AsPC-1 cells, DNA damage and IPS were induced with 30 nM LY2606368. Concurrent incubation with the CDK2 inhibitor CVT-313 prevented DNA damage and dephosphorylation of 4E-BP1 at 5–10 μM (Figure 6). IPS was also inhibited, albeit not completely, and higher concentrations of CVT-313 appeared less effective. CVT-313 also inhibits CDK1 at higher concentrations, but we previously demonstrated that 5 μM CVT-313 prevented DNA damage induced by CHK1i, without inducing cell cycle arrest, consistent with selective inhibition of CDK2.³

Figure 6.

Inhibition of protein synthesis is rescued by suppressing CDK2 and cyclin A. (A) AsPC-1 cells were incubated with 30 nM LY2606368 for 24 h concurrent with 0–80 μM CVT-313; puromycin was added for the final hour. Cell lysates were analyzed by Western blotting. (B) AsPC-1 cells were transfected with siRNA targeting cyclin A or a control siRNA, then incubated with or without 30 nM LY2606368 for 24 h, and analyzed by Western blotting. The amount of puromycin incorporated is presented in the histograms. A repeat of this experiment is shown in Figure S3 with additional data for pAMPK. (C) AsPC-1 cells were incubated with 0–100 nM LY2606368 for 24 h, concurrent with 0–10 μM CVT-313, then incubated for an additional 5 days after removing drugs (left panel). CVT-313 was also added from 24–48 h after removal of LY2606368 then incubated for an additional 4 days (right panel).

In parallel, we assessed the impact of CVT-313 on cell growth and death. A 24 h concurrent incubation with low concentrations of CVT-313 partially rescued cells from LY2606368-induced cell death, while extending the duration of CVT-313 treatment to 48 h (24 h after removing LY2606368) prevented cell death at 25 nM LY2606368 (Figure 6C). Higher concentrations of CVT-313 alone inhibited growth which likely results from inhibition of CDK1.

CDK2 forms active complexes with either cyclin A or cyclin E at different stages of the cell cycle,¹³ and the role of CDK2 in sensitivity to MK-8776 was specifically attributed to CDK2 complexed with cyclin A.³ To test whether this is also the case for LY2606368 and to evaluate the role of cyclin A in IPS, we used siRNA to knock down cyclin A in AsPC-1 cells (Figure 6B). Knockdown of cyclin A effectively prevented LY2606368-induced γH2AX, as well as IPS and dephosphorylation of 4E-BP1. A repeat of this experiment also showed that suppression of cyclin A prevented activation of AMPK (Figure S3). Taken together, these data support the hypothesis that cyclin A/CDK2 is responsible for driving LY2606368-induced IPS, consistent with this phenomenon being an on-target effect of the drug.

DNA Damage Inhibits Protein Synthesis

We previously demonstrated that γH2AX induced by CHK1i was dependent on the MUS81 nuclease.⁶ MUS81 is used to resolve regressed forks that result from stalled replication, so presumably γH2AX results from transient or unrepaired breaks resulting from this cleavage. Using an siRNA targeting MUS81, we reconfirmed the role of MUS81 in response to LY2606368 (Figure 7A). Importantly, we observed that siMUS81 also prevented pAMPK and IPS. Hence, DNA damage is a critical mediator in signaling from CHK1 to IPS. The fact that resistant SW620 cells do not exhibit pAMPK or IPS (Figure 3) suggests that some step in this pathway between damage and activation of AMPK may determine the difference between sensitive and resistant cells.

Figure 7.

Genetic manipulation of regulators of protein synthesis. (A) AsPC-1 cells were transfected with an siRNA targeting MUS81 or control siRNA, then incubated with or without 30 nM LY2606368 for 24 h. Puromycin was added for the final hour, and cell lysates were analyzed by Western blotting. (B) AsPC-1 LKB1-KO cells were incubated with LY2606368 for 24 h, with puromycin added for the final hour. Lysates were then analyzed by Western blotting. The first two lanes show cells transfected with a control guide RNA to confirm the knockout of LKB1 and to contrast the level of pAMPK that occurs. Further analysis of the sgControl derivative and a second LKB1-KO clone is shown in Figure S5. (C) AsPC-1 control (top) and LKB1-KO cells (bottom) were incubated with either 10 or 50 μM STO-609 or takinib, or the combination, concurrently with 30 nM LY2606368 for 24 h, with puromycin added for the final hour. Cell lysates were then probed by Western blotting.

LY2606368-Induced Cell Death Is Driven by AMPK

The dephosphorylation of 4E-BP1 in LY2606368-sensitive cell lines (Figure 3) suggests upstream inhibition of mechanistic target of rapamycin complex 1 (mTORC1), the primary regulator of 4E-BP1. Under growth-permissive conditions, activation of the PI3K-AKT pathway stimulates mTORC1 by inhibiting tuberous sclerosis complex 2 (TSC2), a negative regulator of mTORC1.¹⁴ Active mTORC1 then promotes translation initiation by phosphorylating 4E-BP1. Under conditions of cellular stress, AMPK phosphorylates TSC2 to negatively regulate mTORC1, preventing 4E-BP1 phosphorylation and thus inhibiting translation.¹⁵ We therefore hypothesized that LY2606368-induced IPS might be regulated by AMPK activation.

The primary mode of AMPK activation is through phosphorylation at thr-172.¹⁶ LY2606368 robustly induced pAMPK in AsPC-1 and U2OS cells at concentrations that correspond closely to those at which 4E-BP1 dephosphorylation and IPS occurred (Figure 3). pAMPK was also observed in MDA-MB-231 cells at the slightly higher concentrations that induced DNA damage. However, no pAMPK or dephosphorylation of 4E-BP1 was observed in SW620 despite the appearance of γH2AX.

We expanded this analysis to AsPC-1 cells with acquired resistance to LY2606368, and these also failed to phosphorylate AMPK or dephosphorylate 4E-BP1 (Figure S4). Additionally, these cells required considerably higher concentrations of drug to induce γH2AX.

We next determined whether inhibition of AMPK could prevent IPS and rescue sensitive cells from LY2606368. AsPC-1 cells were incubated with 0–100 nM LY2606368 concurrent with 0–10 μM AMPK inhibitor dorsomorphin (Figure 8A). Incubation for 24 h with 5–10 μM dorsomorphin resulted in a decrease in pAMPK, partial rescue of phosphorylated 4E-BP1, and rescue from IPS. In a parallel toxicity assay, partial protection from growth inhibition was observed during concurrent incubation of LY2606368 and dorsomorphin (Figure 8B). Continuing dorsomorphin treatment for an additional 24 h after removing LY2606368 increased protection, such that 5–10 μM dorsomorphin almost completely prevented cell death induced by 50–100 μM LY2606368. It should be noted that dorsomorphin is a poorly selective drug that inhibits several other kinases more potently than AMPK,¹⁷ but the close correlation between the concentration of dorsomorphin that reduces pAMPK and dephopsphoylation of 4E-BP1, prevents IPS, and rescues cells from cell death is consistent with activation of AMPK being required. Furthermore, the activation of this pathway can only occur in viable cells, supporting the premise that IPS is a cause rather than a consequence of cell death.

Figure 8.

Inhibition of AMPK prevents IPS and rescues cells from LY2606368. (A) AsPC-1 cells were incubated with 30 nM LY2606368 for 24 h concurrent with 0–10 μM dorsomorphin. Puromycin was added for the final hour, and cell lysates were analyzed by Western blotting. (B) AsPC-1 cells were incubated with 0–100 nM LY2606368 for 24 h concurrently with 0–10 μM dorsomorphin (left panel) or for an additional 24 h with just dorsomorphin (right panel). Cells were harvested on day 6 and analyzed for cell growth or death; 0 on the y-axis reflects the cell number of day 0. (C) AsPC-1 cells were incubated with 10 or 30 nM LY2606368 for 24 h concurrent with 0, 5, or 10 μM dorsomorphin, then harvested for assessment of γH2AX by flow cytometry (left panel). LY2606368 was also removed after 24 h, and cells were incubated for an additional 24 h in the presence of dorsomorphin before harvest (right panel). The values are derived from flow cytometry shown in Figure S6.

As dorsomorphin prevented IPS, we next assessed its impact on DNA damage and repair. AsPC-1 cells were incubated with 10 or 30 nM LY2606368 for 24 h concurrent with 0, 5, or 10 μM dorsomorphin. There was a dramatic decrease in the amount of γH2AX, with 10 μM dorsomorphin almost completely preventing DNA damage (Figures 8C and S6). In addition, cells that were positive for γH2AX appeared later in S phase. Cells were further incubated with dorsomorphin for an additional 24 h after removal of LY2606368, and most of the residual γH2AX disappeared.

These results could be explained if dorsomorphin prevented cells from entering S phase which would protect them from incurring DNA damage. However, incubation with 5 μM dorsomorphin caused little cell cycle perturbation. While 10 μM dorsomorphin did reduce the growth rate which was associated with arrest in the G2 phase (not shown; presumably through some other target), the cells would not have been able to reach G2 unless they had first passed through S phase. Hence, it appears that the ability of dorsomorphin to prevent IPS is due to the rescue of DNA repair resulting in far less DNA damage, and thereby enhanced survival. Again, this data gives strong support for the hypothesis that IPS impedes repair and thereby enhances cell killing.

Upstream Regulation of AMPK

The major kinase that phosphorylates AMPK is liver kinase B1 (LKB1). The A549 cells in Figure 1 are known to be defective for LKB1 and are resistant to LY2606368. Two other cell lines in our original cell panel are also defective for LKB1: Of these, H23 are sensitive to other CHK1i, while H460 are resistant.³ To further investigate the potential role of LKB1 in regulating AMPK, we generated AsPC-1 LKB1-KO cells. Upon addition of LY2606368, 4E-BP1 was still dephosphorylated, IPS still occurred, and the cells retained sensitivity to LY2606368 (Figures 2, 7B, and S5). Interestingly, while pAMPK was still observed, it was greatly reduced compared to AsPC-1 wildtype cells, suggesting that another pathway may also contribute to pAMPK and result in IPS.

There are two other kinases that phosphorylate AMPK: transforming growth factor-β-activated kinase (TAK1), and Ca²⁺/calmodulin-dependent protein kinase β (CaMKKβ).^16,18 We hypothesized that one of these kinases is likely responsible for the residual AMPK activation in response to LY2606368 in LKB1-KO cells. Incubation of AsPC-1 cells or LKB1-KO cells with the TAK1 inhibitor takinib had no impact on LY260638-induced pAMPK or IPS (Figure 7C). However, incubation with 50 μM CaMKKβ inhibitor STO-609 significantly reduced pAMPK, independent of the LKB1 status, although it only slightly rescued from IPS (Figure 7C). The combination of takinib and STO-609 had no greater inhibition of pAMPK or rescue from IPS, suggesting that TAK1 does not contribute to the response of cells to LY2606368. The potential involvement of CaMKKβ is consistent with a recent report that this pathway contributes to AMPK activation in response to replication stress (the authors used 25 μM STO-609).¹⁹ Further experimentation is needed to elucidate the potential roles of CAMKKβ and LKB1 in the observed AMPK phosphorylation and IPS.

Discussion

Cells are continuously exposed to DNA damage, both from endogenous and environmental sources, making CHK1 important for protection even in the absence of drug-induced damage. Much of our recent research has dissected the sensitivity to CHK1i as a single-agent, and we have demonstrated that the underlying mechanism is different than that involved when used in combination with DNA damaging agents.^2,20,21 Sensitivity to CHK1i depends on activation of CDK2 which surpasses some threshold level resulting in differential phosphorylation of, as yet, unknown substrates in S phase. This high threshold level is also dependent on cyclin A. In contrast, CHK1i-induced sensitivity to DNA-damaging agents occurs in all cell lines tested, and it only requires a low level of CDK2 activity sufficient to reinitiate replication on damaged DNA. We proposed that the differential sensitivity to CHK1i depends on the differential activation of CDK2/cyclin A,² and the critical role of this complex was confirmed in the current study. However, the results presented here provide an additional (or alternate) factor that appears critical for sensitivity, namely IPS.

This current research was driven by toxicity experiments showing that sensitive cells decreased their cell number below the starting inoculum over time indicating cell death, whereas resistant cells, at best, underwent reversible cytostasis (Figure 1). Assessment of cell death appears much more relevant to tumor regression in xenografts and patients as previously discussed.¹² The cause of cell death in the sensitive cells appears dependent on IPS, as cells could be rescued when various steps in the pathway were inhibited.

In addition to cyclin A/CDK2, the pathway for IPS and sensitivity depends on DNA damage, activation of AMPK, and dephosphorylation of 4E-BP1. DNA damage was assessed by staining for γH2AX. While γH2AX is usually considered to be associated with DSB, it may also be induced by single-strand DNA. However, the latter possibility may be questioned. The primary rationale for suspecting that γH2AX responds to single-strand DNA is that γH2AX can be observed in the absence of DSB or can be induced by agents not thought to cause DSB.^22,23 However, subsequent research has demonstrated that stalled replication forks regress resulting in a “one-sided” DSB²⁴ which might result in γH2AX. Regressed forks are resolved with the aid of the MUS81 nuclease,²⁵ and suppression of MUS81 has been shown to inhibit CHK1i-induced γH2AX.^6,7 We reiterated this observation here using siMUS81. This finding is intriguing as it suggests that γH2AX may not occur on the regressed forks which occur before MUS81 cleavage, but rather after they have been cleaved by MUS81. Such cleavage events should be repaired through homologous recombination, yet we observe persistent γH2AX in sensitive cells after removal of CHK1i. The sensitive cells do not appear defective in homologous recombination because when IPS is suppressed downstream of damage (e.g., dorsomorphin) repair is rapid and the cells survive.

A major difference between sensitive and resistant cells is the level of γH2AX observed after a 24 h incubation (Figure 4). This value reflects the initial amount of DNA damage minus that fraction repaired during this time frame. Hence, when two cell lines exhibit different levels of γH2AX, this could be a consequence of a different initial level of damage or that one is more competent at repair (or both). That repair may play an important role is suggested by the rapid decrease in γH2AX following removal of CHK1i in resistant but not sensitive cells (Figure 5). Our results support the hypothesis that the limited repair in sensitive cells is a consequence of IPS, and this may be critical for their subsequent death.

Our results suggest that AMPK is a critical intermediate between DNA damage and IPS. We initially assumed that phosphorylation of AMPK was mediated by LKB1, yet while knockout of LKB1 did decrease pAMPK, it was incomplete and did not rescue cells from IPS. Inhibition of CaMKKb, but not TAK1, also reduced pAMPK, with partial reduction in IPS. A recent paper also identified a pathway from DNA damage through CaMKKβ to AMPK, albeit the damage was caused by hydroxyurea-induced stalled replication.¹⁹ Activation of AMPK resulted in inhibition of EXO1 thereby protecting forks from excess resection. This would appear contrary to our proposal that activation of AMPK is detrimental to a cell through IPS. However, as CHK1i induces DSB, any suppression of EXO1 may be too late to protect cells from aberrant resection. The authors also did not consider IPS which is an expected consequence of the activation of AMPK, perhaps because only short-term effects were assessed (4 h). However, that paper does support our observation that DNA damage may lead to AMPK activation through CaMKKβ.

Limitations and Outstanding Questions

We recognize that this study is incomplete due to the COVID-19 pandemic, followed by closure of the laboratory due to retirement. For example, greater confidence in the conclusions would require replicate experiments with additional siRNA or knockout derivatives, as well as reiterating our conclusions in additional cell lines. Such cell lines could be used to further dissect the apparent association between IPS and loss of DNA repair. A number of the cytotoxicity assays would benefit from repeating, albeit the observations across multiple cell lines, as well as consistency with our prior experiments,^3,4 does provide confidence for the differential sensitivities observed. There are many other questions that could be addressed, some of which could significantly impact cancer drug development.

Dissection of the signaling from DNA damage to translation is clearly incomplete. What is the relevant importance of LKB1 or CaMKKβ to activation of AMPK; are both pathways necessary or are they redundant? How are they activated by DNA damage? Are there alternate pathways contributing to IPS?

Is the high initial DNA damage observed in sensitive cells simply a reflection of IPS, or do resistant cells have an additional factor that limits DNA damage? In prior experiments, we demonstrated that sensitive cells exhibit much more DNA damage after only 6 h, which is before IPS is observed.⁴ This suggests that the extent of DNA damage cannot be explained solely by IPS. This is consistent with our previous proposal that differential overactivation of CDK2/cyclin A is critical for the excess damage in sensitive cells.^2,3 We recently reported that high activity of CDK2/cyclin A elicits feedback inhibition on CDC25A, thereby limiting inappropriate activation of CDK2, and perhaps this feedback loop contributes to differential DNA damage and sensitivity of cells to CHK1i.²⁶

Is 24 h of target inhibition by CHK1i sufficient to kill other sensitive cell lines? The lethal DNA damage occurs only in S phase cells, so efficacy will depend on cell cycle length. While the AsPC-1 and U2OS cells all accumulate in S phase during a 24 h incubation and exhibit IPS within 12–18 h, if other cell lines take longer to accumulate in S and exhibit IPS, then this could suggest that extended exposure times in a patient might improve therapy. Further investigation aimed at defining the required exposure time in vivo and developing administration schedules that can achieve this is critical.

While the current investigation has focused on CHKi as a monotherapy, our prior studies have included extensive analysis of the combination of CHK1i with other DNA damaging agents.²⁷⁻²⁹ We now question what role IPS may have in sensitivity to these combinations and whether differential IPS may result in differential response to such treatments.

A final question is how these observations relate to the response to other drugs. How many drugs fail because they induce cytostasis rather than cytotoxicity thereby permitting tumors to recover? The corollary is whether observed tumor regression is dependent on IPS? Preliminary experiments have observed that dephosphorylation of 4E-BP1 is a much more common occurrence with ATR inhibitors, consistent with may more cell lines being sensitive.⁵ Hence, our observations could relate to the efficacy, or lack thereof, for numerous drugs.

One of the major driving forces underlying our investigations has been the potential for CHK1i to provide a selective, targeted therapy against sensitive tumors. This has not yet been realized because of the difficulty in stratifying patients due to a failure to understand the underlying mechanism of sensitivity and to inadequate assessment of the length of target inhibition in the tumor. We hope our studies on the mechanism of action of CHK1i will help realize this goal.

Methods

Cell Culture

Cells were obtained from the Developmental Therapeutics Program of the National Cancer Institute or the American Type Culture Collection. Cells were replaced approximately every 3 months. Cells were maintained in RPMI1640 (Corning/Mediatech) plus 10% fetal bovine serum (Hyclone) and 1% antibiotic/antimycotic (Gibco). AsPC-1 cells with acquired resistance were previously generated by incubation with increasing concentrations of LY2606368 over 9 months.⁴ The “2019” derivative was approximately 200-fold resistant, while the “2020” derivative was about 300-fold resistant. Cell lines were tested for mycoplasma using the MycoAlert Mycoplasma Detection Kit (Lonza).

Chemicals

LY2606368, dorsomorphin, STO-609, and takinib were purchased from Selleckchem (stored in DMSO); CVT-313 was obtained from Sigma (all stored in DMSO). Puromycin was obtained from Gibco (10 mg/mL in 20 mM HEPES).

Growth Inhibition and Cell Death

Cells were plated at 10 000 cells per 100 μL in each well of a 96-well plate. The following day, one plate was harvested to provide a starting value, and drugs were added to the other plates at 2-fold dilutions (8 wells per concentration). After 24 h, the drug was removed, and wells washed with phosphate-buffered saline (PBS). Fresh media added for an additional 5 days. Alternately, drugs were left on the cells until harvest on day 6, or cells were harvested throughout an 8 day time course. Plates were washed in PBS and stored at −80 °C until analysis. Cells were lysed, and DNA was stained with Hoechst 33258 as previously described.^27,30 Fluorescence was read on a microplate spectrofluorometer. Results were expressed relative to the starting cell number.

Cell Cycle

Cell cycle analysis was conducted by flow cytometry.²⁸ Both attached and any floating cells were pooled, fixed in 70% ethanol, and labeled with Alexa 488-conjugated γH2AX (Cell Signaling Technology) and propidium iodide. Cells were analyzed on a Becton Dickinson Gallios flow cytometer.

Western Blotting

Cells were rinsed in PBS, lysed in Laemmli lysis buffer, and boiled for 5 min. Proteins were separated by SDS-PAGE and transferred to polyvinylidene difluoride membranes. Western blotting was performed with the following primary antibodies: Cell Signaling Technology: pS296-CHK1 (2349S), γH2AX (9718S), pT172-AMPK (50081S), 4E-BP1 (9452S), LKB1 (3050S), and cyclin E (20808S). Santa Cruz Biotechnology: CHK1 (sc8408), AMPK (sc-74461), MUS81 (sc-53382), and vinculin (sc073541). EMD Millipore: puromycin (MABE343). GeneTex: cyclin A (GTX103042). Fluorescent secondary antibodies were obtained from Cell Signaling Technology: mouse IgG-DyLight 800 (5257), rabbit IgG-DyLight 800 (5151), and mouse IgG-DyLight 680 (5470). Images were generated using a Licor Odyssey fluorescent imager and processed using Image Studio Lite.

Gene Suppression and Knockout Derivatives

Cyclin A siRNA was Dharmacon On-Target Plus SMARTpool CCNA2 (L-003205–00–0005); and the control siRNA was Ambion Silencer Select Negative Control no. 1 (4390843). MUS81 siRNA was obtained from Invitrogen (4392429), together with negative controls [no. 1 siRNA (4390843) and no. 2 siRNA (4390846)]. Cells were transfected with 25 pmol of siRNA using Invitrogen Lipofectamine RNAiMax transfection reagent. After 24 or 48 h, the medium was replaced, and cells were treated ±30 nM LY2606368 for 24 h.

CRISPR guide sequences targeting LKB1 (STK11) were designed using Benchling Biology Software (https://benchling.com). Oligonucleotides (Integrated DNA technologies) 5′-CACCGCCACCGCATCGACTCCACCG-3′ and 5′-AAACCGGTGGAGTCGATGCGGTGGC-3′ were cloned into LentiCRISPRv2-puro (LCv2, Addgene no. 52961) following the Zhang Lab General Cloning Protocol (https://www.addgene.org/crispr/zhang). LCv2 carrying an EGFP-targeting guide was used as a control. Lentivirus was prepared in HEK293 cells, and AsPC-1 cells were transduced overnight with concentrated virus. Transduced cells were selected with 2 μg/mL puromycin (Gibco). Clonal LKB1^–/– cell lines were derived by limiting dilution from the surviving pool following selection and screened by Western blotting for loss of LKB1 protein.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsptsci.1c00150.

• Original flow cytometry experiments that were quantified and summarized and additional Western blots (PDF)

Author Contributions

All authors designed the experiments. J.W.H. generated and analyzed all genetically modified cell lines. J.P.D. performed protein analyses. J.P.D. and A.E. performed cytotoxicity assays and flow cytometry. All authors wrote and reviewed the manuscript.

This work was supported by grant CA117874 from the National Cancer Institute, and a Cancer Center Support Grant to the Norris Cotton Cancer Center (CA23108).

The authors declare no competing financial interest.