Graphical Abstract

   graphic file with name fx1.jpg
   [29]Open in a new tab

Highlights

     * •
       Breast and lung cells show different sensitivities to
       chemotherapeutic drugs
     * •
       Lung cells activate Chk1 more strongly than breast cells with
       chemotherapeutic drugs
     * •
       Active Chk1 plays a regulatory role in p53 activation and apoptosis
       responses
     * •
       The balance of p53 and p21 dynamics drives the apoptosis response
       to DNA damage
     __________________________________________________________________

   Bioinformatics; Mathematical Biosciences; Systems Biology; Cancer
   Systems Biology

Introduction

   The DNA damage response (DDR) is an intricate signaling network that
   governs genomic integrity and protects against carcinogenesis. The core
   DDR network is formed by ATR and Chk1, ATM and Chk2, p53, and p21.
   ATR/Chk1 signaling is activated mostly by stresses involving
   single-strand DNA damage, whereas ATM and Chk2 are activated in
   response to double-strand breaks (DSBs) ([30]Reinhardt and Schumacher,
   2012, [31]Smith et al., 2010, [32]Weber and Ryan, 2015). Both ATR/Chk1
   and ATM/Chk2 can activate p53, a master regulator of the DDR, which
   regulates cellular responses such as cell cycle arrests, repair and
   survival, or cell death. An important determinant of outcome is the p53
   target protein p21, which can inhibit cell cycle progression and
   negatively regulate p53-mediated apoptosis ([33]Abbas and Dutta, 2009).

   Proper regulation of the DDR is important, because dysregulation is
   associated with aging and cancer ([34]Hoeijmakers, 2009). Tumors
   frequently contain mutations in DDR genes and are characterized by an
   aberrant response to DNA damage. Strikingly, cancers originating in
   distinct tissues have diverse mutational signatures ([35]Alexandrov
   et al., 2013, [36]Kandoth et al., 2013, [37]Zehir et al., 2017),
   suggesting that they have different DNA repair defects ([38]Chae
   et al., 2016, [39]Forestier et al., 2012, [40]Garner and Eastman,
   2011). Differences in DDR are also thought to result in specific
   sensitivities for DNA-damaging chemotherapy ([41]Chae et al., 2016,
   [42]Forestier et al., 2012, [43]Liu et al., 2017, [44]Sousa et al.,
   2015, [45]Stewart-Ornstein and Lahav, 2017), which differ per cancer
   type. For instance, non-small-cell lung cancer responds relatively well
   to the DNA adduct and strand cross-linking agent cisplatin
   ([46]Eastman, 1987, [47]Gettinger and Lynch, 2011; [48]Non-small Cell
   Lung Cancer Collaborative Group, 1995, [49]Laskin and Sandler, 2005,
   [50]Paoletti et al., 2011, [51]Rosenberg et al., 1969). In contrast,
   the DNA-DSB-inducing drug doxorubicin is more effective in treating
   breast cancer ([52]Arcamone et al., 1969, [53]Crown, 1998, [54]von
   Minckwitz, 2007).

   It is unclear whether the differences in DDR in cancer types arise
   during tumor evolution, or whether they reflect variations present in
   the original tissue. It is conceivable that healthy cells have
   differences in the DDR network. For instance, skin cells exposed to
   sunlight might express a higher amount of genes involved in the repair
   of UV-induced DNA lesions than other tissues. In addition, mutations in
   DDR genes affect cancer development across tissues differently. For
   example, germline mutations in BRCA1/2 and Chk2 enhance cancer risk in
   specific tissues such as breast, whereas the added risk for tumors
   occurring in other tissues like lung is small ([55]Levy-Lahad and
   Friedman, 2007, [56]Michailidou et al., 2017, [57]Nevanlinna and
   Bartek, 2006). Furthermore, the response to DNA damage is different
   across stem cell types. In particular intestinal stem cells are prone
   to induce apoptosis after DNA damage, whereas DNA damage in melanocytes
   results in terminal differentiation ([58]Blanpain et al., 2011,
   [59]Vitale et al., 2017). A systematic comparison of the DDR across
   tissues is lacking. Understanding what differences in DDR network arise
   during tumor evolution and which ones are already present will provide
   more insight into tissue-specific cancer risk. In addition, this
   knowledge could lead to more effective targeting of tumor cells for
   therapy.

   To address how the DDR is regulated across tissues, we used
   non-immortalized primary epithelial cells from multiple breast and lung
   donors. We focused on breast and lung cells for three reasons: (1)
   breast and lung tissues frequently give rise to tumors; (2) certain DDR
   gene mutations predispose for breast cancer, but not for lung cancer;
   and (3) breast and lung tumors have different sensitivities for
   genotoxic chemotherapy. By combining modeling and experimental
   approaches, we show that these cell types have different apoptotic
   sensitivities for genotoxic chemotherapy. We demonstrate that the
   difference in sensitivity originates at the earliest level of DNA
   damage detection: in the tissue-specific regulation of the ATR-Chk1 and
   ATM-Chk2 network. Our modeling and experimental analyses suggest that
   differential activation of Chk1 and Chk2 regulates the balance of p53
   and p21 dynamics and contributes to chemotherapy responses to DNA
   damage in breast and lung cells. This study sheds light on the
   tissue-specific DDR and provides clues on how to optimize chemotherapy
   response.

Results

An Equal Amount of Cisplatin-DNA Lesions Induces More Apoptosis in Primary
Breast Than Lung Cells

   The DDR pathway is a complex signaling network, which is activated upon
   many forms of damage and gives rise to multiple different outputs. We
   have focused on one DNA-damaging agent and one cellular outcome
   (apoptosis). To study the differences of the DDR in different cell
   types, we used primary breast and lung cells. We started with the
   platinum-containing drug cisplatin because it is a widely used
   chemotherapeutic that causes DNA adducts and inter- and intrastrand
   cross-links and is known to induce apoptosis ([60]Eastman, 1987, [61]Hu
   et al., 2016, [62]Rosenberg et al., 1969), which can be readily
   quantified using MTT
   (3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) cell
   viability assay. Briefly, cells were seeded at 50% density, followed by
   48 h of cisplatin exposure. This setup allowed us to measure cell death
   ([63]Figure 1A). Breast cells (IC[50]∼ 15–20 μM) were consistently more
   sensitive than lung cells (IC[50]∼40 μM; p = 1 × 10^−9; [64]Figure 1B).
   A different sensitivity for apoptosis was confirmed by the appearance
   of apoptotic morphology as visualized by DAPI staining ([65]Figures 1C
   and 1D).

Figure 1.

   [66]Figure 1
   [67]Open in a new tab

   Primary Breast Cells Are More Sensitive to Cisplatin-Induced DNA
   Lesions Than Primary Lung Cells

   (A) Overview of adapted MTT assay to measure cell death. Briefly, cells
   were seeded at 50% density and treated with high doses of cisplatin for
   24 or 48 h. Treatment under these conditions results in cell death
   (e.g., compare the wells treated with 60 μM cisplatin at 24 and 48 h).

   (B) Breast cells are more sensitive to cisplatin than lung cells. Human
   primary epithelial cells from six donors were treated with a range of
   cisplatin concentrations. After 48 h, the cell viability was determined
   by MTT assay. The concentration at which 50% loss of viability occurred
   (IC[50]) was calculated (n = 3–6). A t-test was used to compare IC[50]
   values between breast and lung cells.

   (C) Treatment with cisplatin results in apoptosis. Nuclear morphology
   of breast and lung cells was visualized by DAPI stain after 24-h
   exposure to 80 μM cisplatin (representative experiment, n = 4).

   (D) Breast cells more frequently display apoptotic morphology (DAPI
   stain) after 24 h of cisplatin treatment than lung cells
   (representative experiment, n = 4). A two-way ANOVA with Bonferroni
   post-hoc test was carried out to compare differences (*p < 0.05, ***p <
   0.001).

   (E) Breast cells are more sensitive to equal amounts of
   cisplatin-induced DNA lesions at 48 h. The amount of DNA-bound platinum
   was determined after 48 h of cisplatin treatment ([68]Figure S1A) and
   plotted against the cellular viability (MTT assay) at 48 h
   (representative experiment, n = 3). Regression analysis was used to
   test if the slope and intercept were different. Error bars represent
   the SD (n = 3).

   To control for differential cisplatin uptake, we looked directly at the
   relation between platinum bound to the DNA (Pt_DNA) and cellular
   outcome. Breast cells contained higher levels of Pt_DNA at similar
   cisplatin doses ([69]Figure S1A), and even when corrected for the
   amount of Pt_DNA, breast cells were consistently more sensitive for
   cisplatin ([70]Figure 1E). In summary, breast cells induce apoptosis at
   a lower amount of DNA damage than lung cells.

Cisplatin Treatment Activates Chk1 Signaling More Strongly in Lung Cells,
Resulting in Increased Resistance

   To understand how an equal dose of DNA damage results in differential
   cell viability, we initially focused on p53 dynamics, as it is an
   important determinant of apoptosis. Intriguingly, p53 levels alone
   could not explain the differential apoptotic sensitivity. When breast
   and lung cells were treated with 20 and 30 μM cisplatin, DNA-bound
   cisplatin and total p53 levels were very similar ([71]Figures S1B and
   S1C). To understand the mechanism behind their differential cisplatin
   sensitivity, we used protein microarrays to assess temporal changes of
   the proteome and phosphoproteome (covering 366 unique UniProt IDs) in
   breast and lung cells. We analyzed the proteome at three time points:
   an early stage (6 h; p53 increase becomes apparent), an intermediate
   stage (12 h), and a late stage (24 h), in which signs of apoptosis can
   be observed. We thus generated an unbiased profile of the post-DNA
   damage proteome in healthy breast and lung cells. Among the proteins
   that were differentially expressed, 17 BioCarta pathways were
   significantly enriched (p < 0.1; [72]Figures 2A, 2B, and [73]S2;
   [74]Tables S1, [75]S2, and [76]S3). After antibody validation
   ([77]Table S2), we focused on five differentially regulated signaling
   modules: PTEN/Akt signaling, ERK, p38, JNK signaling, and DNA damage
   checkpoint activation. As shown in [78]Figures 2C and [79]S3,
   differential activation (as measured by phosphorylation on key
   residues) was observed for proteins in the mitogen-activated protein
   kinase (MAPK) signaling pathway. For example, p38 MAPK was activated to
   a higher degree in primary breast cells. Furthermore, differences in
   DNA damage checkpoint activity were observed ([80]Figures 2C and
   [81]S3E). Active ATR (p-S428) was present at higher levels in lung
   cells than breast cells. Accordingly, its downstream target Chk1 was
   activated (S345) to a greater extent in lung cells.

Figure 2.

   [82]Figure 2
   [83]Open in a new tab

   Proteomic and Functional Analysis Reveals an Increased Chk1 Activity in
   Lung Cells that Contributes to Cisplatin Resistance

   (A) Identification of genes involved in differential cisplatin
   response. Breast and lung cells were treated with a cisplatin dose that
   caused an equal amount of DNA damage (20 μM and 30 μM, respectively). A
   protein array (366 unique UniProt IDs covered by both pan-specific and
   phospho-specific antibodies) was carried out to assess proteomic
   changes at 6, 12, and 24 h. Gene set enrichment was performed on
   differentially expressed proteins to identify differentially regulated
   pathways. After validation of the array results using both antibodies
   present on the array and antibodies from different manufacturers,
   inhibitors were used to screen for genes that modified the cisplatin
   response. For details see [84]Transparent Methods and [85]Figure S2.

   (B) Overview of BioCarta pathways enriched for differentially expressed
   genes (DAVID pathway analysis). For details, see [86]Table S2.

   (C) Overview of the most differentially activated pathways after
   cisplatin treatment. After array antibody validation ([87]Table S2),
   activity of differentially regulated pathways was determined using
   quantitative western blot ([88]Figure S3). The colors indicate whether
   protein activity (e.g., highest phosphorylation levels) is the highest
   in lung (red) or breast (blue) cells.

   (D) Inhibition of Chk1 activity differentially affects cisplatin
   sensitivity. Cells were treated for 1 h with the Chk1 inhibitor
   PF477736 (1 μM) or an equivalent amount of DMSO, after which cisplatin
   was added. Viability was determined after 48 h. Error bars represent
   the SD (n = 7). A two-way ANOVA with Bonferroni post-hoc test was
   carried out to test if differences were significant (**p < 0.01).

   To investigate if the differentially regulated pathways affect
   cisplatin-induced apoptosis, we negatively regulated essential proteins
   in each pathway using specific inhibitors that produce fewer side
   effects than small interfering RNAs. Inhibitors for
   phosphatidylinositol 3-kinase, JNK, p38 kinase, and MEK did not
   significantly alter the cisplatin response ([89]Figure S4). We next
   focused on ATR, Chk1, and Chk2 using specific inhibitors (VE822 for
   ATR, PF477736 for Chk1, Chk2 inhibitor II for Chk2) ([90]Figure S5A).
   Strikingly, Chk1 inhibition sensitized primary lung cells to cisplatin
   (p < 0.001), but it had no effect on breast cells ([91]Figure 2D).
   Similar effects of Chk1 inhibition were observed in primary cells
   obtained from different donors ([92]Table S4), and ATR inhibition
   sensitized lung cells too ([93]Figures S5B and S5C). We also found that
   cell viability responses in lung cells were affected to a lesser extent
   by Chk2 inhibition than by Chk1 inhibition ([94]Figure S5D). Taken
   together, these results indicate that the higher activity of Chk1 in
   lung cells contributes to cisplatin resistance of lung cells.

Cell-Type-Specific DNA Damage Signaling Dynamics after Cisplatin Treatment

   The identification of Chk1 as a modifier of the apoptotic sensitivity
   for cisplatin was unexpected because its downstream target p53
   accumulates to similar levels in breast and lung cells
   ([95]Figure S1C). We therefore decided to compare the activity of the
   Chk1-Chk2-p53 signaling network in breast and lung cells more closely.
   We quantitatively measured the total levels and activities of Chk1,
   Chk2, p53, and p21 ([96]Figure 3A). After an equal amount of
   cisplatin-induced DNA damage, Chk1 activity, as measured by
   phosphorylation of sites S345, S317, and S296, increased to a much
   higher level in the lung cells ([97]Figure 3A, p < 0.01,
   [98]Figure S6). In contrast, Chk2 activation (T68, S516) was only
   moderately increased in both cell types. Although total p53 levels were
   similar in breast and lung cells, p53 was phosphorylated (S15/S20) to a
   larger extent in lung cells ([99]Figure 3A, p < 0.01, [100]Figure S6),
   correlating with higher Chk1 activation. In addition, p21 levels
   increased in lung cells but not in breast cells. Together, these
   results show that primary breast and lung cells display different
   dynamics of Chk1 activation, p53 phosphorylation, and p21 protein
   expression after cisplatin treatment.

Figure 3.

   [101]Figure 3
   [102]Open in a new tab

   Cisplatin Treatment Results in Different DNA Damage Signaling Dynamics
   in Breast and Lung Cells

   (A) Relative change of Chk1, Chk2, p53, and p21 dynamics after
   cisplatin treatment in breast and lung cells. Protein levels were
   analyzed using quantitative western blot and normalized by the number
   cells loaded in each sample. For every blot the highest expression
   value was set to 1, after which the average value was calculated for
   five to eight blots. Depicted is the average value normalized to the
   maximum value per graph. Error bars represent the SEM. Two-way ANOVAs
   were used to test if a protein was differentially expressed in breast
   versus lung cells. See also [103]Figure S6.

   (B) Comparison of model simulations to the experimental data of DNA
   damage signaling dynamics. To compare with the corresponding
   experimental data, the model simulation data were scaled as relative
   values (normalized to the maximal values in breast and lung cells) and
   depicted as blue lines. Data from individual blots are depicted as “+.”
   The average values are shown as red circles. Note that some data points
   with the same values appear overlapped in this scatterplot.

A Quantitative Model Predicts Differential Roles of Chk1 and Chk2 in the
Regulation of p53 Phosphorylation

   To better understand the relation between the dynamics of the DDR
   proteins and the chemotherapy response (apoptosis), we developed an
   integrated mathematical model based on our current data and previous
   studies ([104]Barr et al., 2017, [105]Batchelor et al., 2008,
   [106]Batchelor et al., 2011, [107]Ma et al., 2005,
   [108]Stewart-Ornstein and Lahav, 2017, [109]Zhang et al., 2011). This
   new model consists of coupled ordinary differential equations, and it
   focuses on the key interactions between Chk1, Chk2, p53, and p21
   ([110]Figure S7). In the model, we used the same model structure and
   reaction kinetics for breast and lung cells. Most parameters have the
   same values, except that some parameters are set with
   cell-type-specific values to distinguish different cisplatin-DNA
   binding dynamics, differential activation, and degradation of proteins
   observed in breast and lung cells (list of parameter values is given in
   [111]Tables S5–S7). A detailed description of the model development,
   model simulation, and parameter estimation is provided in the
   [112]Supplemental Information.

   The mathematical model was fitted to the experimental datasets that
   captured the temporal profiles of cisplatin-DNA binding dynamics; total
   levels of Chk1, Chk2, p53, and p21; the activation of Chk1, Chk2, and
   p53; as well as cell viability responses in breast and lung cells. In
   total, 98 average values (from 564 measured data points) were used to
   estimate the kinetic parameters of the model for cisplatin treatment.
   As shown in [113]Figures 3B and [114]S8, the model simulations
   quantitatively fit the signaling dynamics of Chk1, Chk2, p53, and p21
   proteins and DNA-bound cisplatin.

   Intriguingly, the model suggested that the contribution of Chk1 and
   Chk2 to the regulation of p53 and p21 dynamics is different. As shown
   in [115]Figure S9, the modeling analyses indicate that Chk2 activities
   in breast and lung cells are relatively high and give rise to a
   saturated p53 phosphorylation rate. On the other hand, in breast cells
   Chk1 activity is too low to induce p53 phosphorylation, whereas its
   level in lung cells is relatively high and induces additional p53
   phosphorylation on top of Chk2's activity. We next studied what happens
   to downstream p53 and p21 responses after in silico inhibition of Chk1
   and Chk2 activities (details are in [116]Supplemental Information). The
   model predicted that Chk1 inhibition has almost no effect on p53 and
   p21 dynamics in breast cells, where Chk1 phosphorylation is low
   ([117]Figure 4A). In contrast, in lung cells a strong effect of Chk1
   inhibition on p53 and p21 dynamics is expected ([118]Figure 4B).
   Interestingly, it is also predicted that Chk2 inhibition will strongly
   decrease the levels of total p53, p53 phosphorylation, and p21 in both
   breast and lung cells ([119]Figures 4A and 4B). These results thus
   indicate that the contributions of Chk1 and Chk2 in regulating p53 and
   p21 dynamics are different.

Figure 4.

   [120]Figure 4
   [121]Open in a new tab

   Inhibitor Studies Reveal Different Roles for Chk1 and Chk2 in Breast
   and Lung Cells

   (A and B) Model predictions for the inhibition of Chk1 and Chk2 in
   breast (A) and lung cells (B).

   (C and D) Inhibition of Chk1 and Chk2 alter p53 and p21 dynamics.
   Primary breast (C) and lung cells (D) were pretreated for 1 h with
   inhibitors against Chk1 (PF477736; 1 μM) and Chk2 (Inhibitor II, 10 μM)
   or an equivalent amount of DMSO, followed by cisplatin treatment.
   Depicted is the average relative expression (n = 3–7) for each
   condition normalized to the maximum value. Error bars represent the
   SEM. Two-way ANOVAs were used to test if a protein was differentially
   expressed between treatments.

   See also [122]Figures S10 and [123]S11.

   To experimentally validate the model predictions, we treated breast and
   lung cells with inhibitors for Chk1 and Chk2, or an equivalent amount
   of DMSO, followed by cisplatin treatment. As predicted by the model,
   Chk1 inhibition did not have a significant effect on downstream targets
   in breast cells ([124]Figures 4C and [125]S10). In lung cells, however,
   Chk1 inhibition significantly altered p53 and p21 dynamics. Total p53
   levels were unaffected, whereas p53 phosphorylation was diminished and
   p21 was not upregulated after cisplatin treatment ([126]Figures 4D and
   [127]S11). In line with model predictions, Chk2 inhibition resulted in
   the loss of p21 expression in both breast and lung cells. In addition,
   total and phospho-p53 levels remained low when Chk2 activity was
   inhibited ([128]Figures 4C and 4D). Together, our model analyses and
   experimental data suggest that Chk1 activity plays a regulatory role in
   p53 activation, whereas Chk2 is essential for p53 activation and
   expression.

Breast Cells Are More Resistant to DNA Double-Strand Breaks Than Lung Cells

   Chk1 is strongly activated by single-strand DNA damage, whereas Chk2 is
   mostly activated by DNA DSBs ([129]Bartek and Lukas, 2003, [130]Smith
   et al., 2010). We therefore wondered what happens if we treat primary
   breast and lung cells with the DSB-inducing drug doxorubicin.
   Doxorubicin induces DNA adducts and also causes DNA DSBs by inhibiting
   topoisomerase II activity ([131]Arcamone et al., 1969, [132]Yang
   et al., 2014). When cell viability response data were plotted against
   the level of intracellular doxorubicin at 48 h ([133]Figure S12),
   breast cells were more resistant than lung cells ([134]Figure 5A).
   Surprisingly, this is opposite to the effect of cisplatin, which
   induces more apoptosis in breast cells, and suggests that the
   sensitivities of breast and lung cells to genotoxic stress depend on
   the nature of DNA damage.

Figure 5.

   [135]Figure 5
   [136]Open in a new tab

   Breast Cells Are More Resistant to Doxorubicin Than Lung Cells

   (A) Cell viability response to equal intracellular amounts of
   doxorubicin. After 48 h of doxorubicin treatment, intracellular
   doxorubicin fluorescence was determined by fluorescence-activated cell
   sorting ([137]Figure S12). The intracellular doxorubicin levels
   (arbitrary units) were plotted against the cellular viability (as
   determined by MTT assay) (n = 3). Error bars represent the SD.
   Regression analysis was used to test if the slope and intercept were
   different.

   (B) Relative change of Chk1, Chk2, p53, and p21 dynamics after
   doxorubicin treatment in breast and lung cells. Depicted is the average
   relative expression per treatment (n = 4–6) normalized to the maximum
   value per graph. Error bars represent the SEM. Two-way ANOVAs were used
   to test if a protein was differentially expressed in breast versus lung
   cells. See also [138]Figure S13.

   (C) Comparison of model simulations to the experimental data in
   response to doxorubicin treatment. To compare with the corresponding
   experimental data, the model simulation data were scaled as relative
   values (normalized to the maximal values in breast and lung cells) and
   depicted as blue lines. Data from individual blots are depicted as “+.”
   The average values are shown as red circles. Note that some data points
   with the same values appear overlapped in this scatterplot.

   To compare how DNA damage signaling is differently regulated after
   doxorubicin treatment, we quantitatively measured the dynamics of
   DNA-damaging signaling proteins. Interestingly, doxorubicin treatment
   led to strong activation of both ATM (S1981) and Chk2 (T68 and S516) in
   both cell types ([139]Figures 5B and [140]S13), although breast cells
   showed higher levels of active Chk2. In addition, lung cells showed a
   significant higher level of ATR and Chk1 activation (S345, S317, S296)
   than breast cells. The induction of total and phospho-p53 was higher in
   lung cells, whereas p21 levels showed a weak increase in both breast
   and lung cells.

   To understand the DDR to doxorubicin treatment, we slightly modified
   the values of the model parameters that are specific for cisplatin
   treatment. These parameters were estimated with 94 average values from
   412 measured data points for doxorubicin treatment (parameters values
   are listed in [141]Table S7). As shown in [142]Figures 5C and [143]S14,
   the models for doxorubicin treatment fitted well to the temporal
   profiles of different DNA-damage-signaling proteins in both breast and
   lung cells. Similar to the cisplatin model, the doxorubicin model
   predicted that Chk2 activity is required for p53 phosphorylation and
   accumulation, as well as p21 induction in both breast and lung cells
   ([144]Figures S15A and S15B). Chk1 activity contributed to p53
   phosphorylation, p53 accumulation, and p21 induction in lung cells.
   However, it has only a minor contribution in breast cells. The model
   predictions were confirmed by experiments with Chk1 and Chk2 inhibitors
   ([145]Figures S15C, S15D, [146]S16, and [147]S17).

A “Friction Model” Suggests that the Balance of p53 and p21 Dynamics
Contributes to Chemotherapy Responses

   The different apoptotic sensitivities to cisplatin and doxorubicin are
   counterintuitive. Therefore, it is interesting to investigate how the
   same DNA-damaging core network can lead to a different apoptotic
   sensitivity depending on the cell type and the nature of
   chemotherapeutic agents. Previous studies have indicated that a
   threshold mechanism mediates p53 cell fate decision to induce apoptosis
   ([148]Kracikova et al., 2013, [149]Paek et al., 2016). In this work, we
   proposed a novel “friction model” to describe the cell apoptosis
   response using a coarse-grained approach, in which a hypothetical
   threshold mechanism is used to trigger apoptosis. The model assumes
   that cells undergo apoptosis when the total p53 level is higher than a
   certain threshold, θ. Instead of using a fixed threshold, we assume
   that the threshold θ is proportional to the p21 expression level as p21
   can act as a negative regulator for apoptosis ([150]Abbas and Dutta,
   2009, [151]Roninson, 2002). The dynamic threshold is analogous to the
   friction force exerted by a surface that resists the motion of an
   object, which is proportional to its mass. To move cells to apoptosis,
   the driving force (caused by p53) should be larger than the friction
   force (induced by p21). As shown in [152]Figure S18, this simplified
   friction model was able to reproduce the different cell viability
   responses in breast and lung cells after cisplatin or doxorubicin
   treatment.

   Additional modeling analyses indicated that the balance between p53 and
   p21 dynamics could explain the differential sensitivities to diverse
   chemotherapeutic agents. In case of cisplatin treatment, both breast
   and lung cells have a similar accumulation of p53 protein (similar
   driving forces), whereas p21 expression is induced more in lung cells.
   Therefore the overall friction force is larger in lung cells, which
   makes lung cells more resistant to apoptosis after cisplatin treatment
   ([153]Figures 6A and 6B). Indeed, an inhibitor of p21 transcription
   could sensitize lung cells to cisplatin treatment (p < 0.001;
   [154]Figure S19). On the other hand, after doxorubicin treatment the
   p53 level in lung cells is higher, whereas p21 dynamics is similar.
   Hence, the driving force is lower in breast cells, explaining why
   breast cells are more resistant ([155]Figures 6C and 6D). In summary,
   the model suggests that the balance of p53 and p21 dynamics contributes
   to the decision to undergo apoptosis after DNA damage.

Figure 6.

   [156]Figure 6
   [157]Open in a new tab

   Mathematical Modeling Shows that the Balance of p53 and p21 Levels
   Determines Cell Viability after DNA Damage

   (A–D) Model simulations for the dynamics of cell viability, apoptosis
   rate, p53, p21, and threshold for apoptosis in response to cisplatin (A
   and B) or doxorubicin (C and D) in breast and lung cells.

Discussion

   We report here that primary non-transformed breast and lung cells have
   different sensitivities for DNA-damaging agents. Breast cells are more
   sensitive to the DNA adduct and cross-linking agent cisplatin, whereas
   lung cells are more sensitive to the DNA adduct and DSB-inducing agent
   doxorubicin. These differences in sensitivity can be explained by the
   differential activation of Chk1, which is phosphorylated to a greater
   extent in lung cells. In contrast, in breast cells, Chk2 is activated
   to a higher extent after doxorubicin treatment. The differential
   activities of Chk1 and Chk2 affect p53 and p21 dynamics and hence
   contribute to drug sensitivity. Although differences in p53 and p21
   dynamics and downstream pathways have been described in tissue-specific
   stem cells before ([158]Blanpain et al., 2011, [159]Insinga et al.,
   2013, [160]Vitale et al., 2017), they were considered to be stem cell
   properties that change after differentiation ([161]Insinga et al.,
   2013). Our study shows for the first time that healthy epithelial cell
   types have a different DDR. Moreover, our observation that differences
   in DDR can already be observed at the level of DNA damage detection is
   novel.

   An intriguing question is why breast and lung cells show different Chk1
   and Chk2 activities. The ATR-Chk1 pathway is most strongly activated by
   single-strand DNA damage, whereas the ATM-Chk2 pathway plays a role in
   DSB repair. As cell types experience different kinds of DNA damage, it
   may be advantageous to tailor DDR protein dynamics to the form of DNA
   damage that occurs most frequently. For instance, the high oxygen
   levels in lung likely result in increased numbers of free radicals,
   which can damage the DNA by causing DNA adducts and other forms of
   single-strand damage ([162]Cadet and Wagner, 2013, [163]Pham-Huy
   et al., 2008). This may explain why Chk1 is activated to a higher
   extent in lung cells than breast cells. In contrast, in breast,
   metabolic by-products of estrogen signaling ([164]Yasuda et al., 2017)
   are known to give rise to DSBs ([165]Savage et al., 2014), indicating
   that breast cells may need enhanced ATM-Chk2 signaling. The causes of
   the differences in Chk1 and Chk2 activities in breast and lung cells
   are unknown. This will be the subject of future study.

   We observed that primary breast cells are more sensitive to cisplatin,
   and lung cells, to doxorubicin treatment. Paradoxically, first-line
   chemotherapeutic regimens for breast and lung cancer are mostly based
   on doxorubicin or cisplatin, respectively ([166]Crown, 1998,
   [167]Laskin and Sandler, 2005, [168]Paoletti et al., 2011, [169]von
   Minckwitz, 2007). It is likely that during tumorigenesis breast and
   lung cancer cells become more sensitive to doxorubicin and cisplatin,
   respectively, as breast cancer cell lines respond well to doxorubicin
   and non-small-cell lung cancer cell lines are generally sensitive to
   cisplatin ([170]Alley et al., 1988, [171]Shoemaker, 2006). The goal of
   cancer therapy is to destroy cancer cells while sparing healthy cells.
   The difference in sensitivity between normal and transformed cells
   would provide a window to selectively kill cancer cells. It would be
   interesting to investigate if and how the sensitivity for DNA damage
   changes during tumorigenesis.

   Chk1 inhibitors have recently attracted attention as sensitizers for
   chemotherapy as Chk1 is frequently overexpressed in tumors ([172]Zhang
   and Hunter, 2014). Our model confirms that Chk1 overexpression could
   lead to increased cisplatin resistance and hence provides a rationale
   for co-treatment of Chk1 inhibitors with cisplatin. Indeed, Chk1
   inhibition has been found to decrease cisplatin resistance in cancer
   cell lines ([173]Gadhikar et al., 2013, [174]Li et al., 2016,
   [175]Thompson et al., 2012). However, the observation that Chk1
   inhibition did not sensitize breast cells to cisplatin may indicate
   that only tumors with overactive Chk1 may respond to Chk1 inhibition,
   and hence calls for careful patient stratification in clinical trials.

   Both ATR-Chk1 and ATM-Chk2 signaling are known to regulate p53
   stabilization and phosphorylation. Our model predicts that Chk1 and
   Chk2 affect p53 dynamics differently. Chk2 activity is required for p53
   phosphorylation, whereas Chk1 activity is not essential, but
   contributes to p53 phosphorylation ([176]Figure 7A). As a consequence
   of its higher activity in lung cells, Chk1 contributes to p53
   phosphorylation in lung cells but not in breast cells. As expected, a
   Chk1 inhibitor reduced p53 phosphorylation levels only in lung cells.
   Although Chk1 has been demonstrated to be able to phosphorylate p53 on
   S15 and S20 ([177]Shieh et al., 2000), some studies reported that Chk1
   inhibition did not diminish p53 phosphorylation ([178]Tian et al.,
   2002, [179]Wu et al., 2013). Our data indicate that this may be a
   cell-type or DNA-damage-specific effect.

Figure 7.

   [180]Figure 7
   [181]Open in a new tab

   A Friction Model of Cell Apoptosis Response to DNA Damage Signaling

   (A) A simple scheme for Chk1, Chk2, p53, and p21 protein interactions.

   (B) A cartoon of the “friction model.” A cyan-to-orange gradient
   indicates the balance between survival (cyan region) and apoptosis
   (orange region). Cellular outcome to DNA damage is determined by the
   balance of the driving force (caused by p53) and the friction force
   (induced by p21). To move cells to the apoptotic state, the driving
   force should be larger than the friction force. The size of arrows
   depicts the relative strength of the driving and friction forces in
   breast and lung cells in response to cisplatin or doxorubicin
   treatment. p-Chk1 and p-Chk2 contribute to p53 phosphorylation, and
   both p-p53 and Chk2 can induce p21 transcription. However, whether p21
   expression increases also depends on the nature of the stress.

   We propose a “friction model” to explain the differences in apoptotic
   sensitivity for cisplatin and doxorubicin in breast and lung cells. The
   induction of apoptosis is a complex process that may involve
   caspase-dependent or caspase-independent pathways and is regulated by
   pro- and anti-survival proteins. We did not incorporate these effects
   in our model because we found that a simplified version of apoptosis
   induction, based on the balance between p53 and p21 dynamics, was
   sufficient to explain the apoptotic response. Moreover, additional
   processes will increase the complexity of the model, and studying these
   detailed regulations is not the focus of this work. Our new friction
   model incorporates the idea of a dynamic p53 threshold for apoptosis
   induction ([182]Paek et al., 2016), but defines it as a function of p21
   levels. After cisplatin treatment, p53 dynamics are similar in breast
   and lung cells causing a similar apoptotic stimulus ([183]Figure 7B).
   The difference in cisplatin resistance can be explained because in lung
   cells increased amounts of p53 phosphorylation (as result of increased
   Chk1 activity) give rise to p21 induction, resulting in an increased
   apoptotic threshold. In contrast, after doxorubicin treatment p21
   dynamics are similar in breast and lung cells. In this case, increased
   p53 phosphorylation in lung cells leads to stabilization of p53, which
   translates into a higher apoptotic stimulus and enhanced sensitivity to
   doxorubicin ([184]Figure 7B). A possible role for p21 as a friction
   force is supported by data for the p21 inhibitor UC2288. Although the
   mechanism of action of this inhibitor is incompletely understood, we
   observed down-regulation of p21 after treatment. We cannot exclude,
   however, that the role of p21 is cellular context dependent. In
   addition, it is likely that other proteins could contribute to the
   friction force as well, such as the inhibitors of apoptosis (IAP)
   family proteins ([185]Deveraux and Reed, 1999, [186]Paek et al., 2016).

   In summary, our study provides new insights into the apoptotic
   sensitivity for DNA-damaging agents. It will be interesting to
   investigate how modulating Chk1 levels and the balance between p53 and
   p21 may improve the effectiveness of chemotherapy.

Limitations of the Study

   In this study we made use of primary epithelial cells to investigate
   the DDR in healthy cells. Although this study provided novel insights
   into how Chk1, Chk2, p53, and p21 collectively affect the apoptotic
   response to DNA-damaging chemotherapy in different tissues, subsequent
   studies are necessary to investigate the implication for cancer cells
   and cancer treatment. In addition, in this study, we described a
   friction model to explain the difference in apoptotic sensitivity
   between breast and lung cells for two different anti-cancer drugs. This
   friction model is based on the balance of p53 and p21. It is likely,
   however, that other anti-apoptotic proteins (such as the IAP family of
   proteins) could contribute to the friction force as well. Furthermore,
   in our mathematical model, different p53 negative regulators are lumped
   as one abstract variable “negative feedback regulator”, which makes it
   difficult to study the contributions of specific negative regulators.
   As different p53 negative regulators may affect p53 expression and
   activity in different ways, their exact roles and contributions need to
   be further studied.

Methods

   All methods can be found in the accompanying [187]Transparent Methods
   supplemental file.

Acknowledgments