A synergistic two-drug therapy specifically targets a DNA repair dysregulation that occurs in p53-deficient colorectal and pancreatic cancers

Summary

The tumor-suppressor p53 is commonly inactivated in colorectal cancer and pancreatic ductal adenocarcinoma, but existing treatment options for p53-mutant (p53^Mut) cancer are largely ineffective. Here, we report a therapeutic strategy for p53^Mut tumors based on abnormalities in the DNA repair response. Investigation of DNA repair upon challenge with thymidine analogs reveals a dysregulation in DNA repair response in p53^Mut cells that leads to accumulation of DNA breaks. Thymidine analogs do not interrupt DNA synthesis but induce DNA repair that involves a p53-dependent checkpoint. Inhibitors of poly(ADP-ribose) polymerase (PARPis) markedly enhance DNA double-strand breaks and cell death induced by thymidine analogs in p53^Mut cells, whereas p53 wild-type cells respond with p53-dependent inhibition of the cell cycle. Combinations of trifluorothymidine and PARPi agents demonstrate superior anti-neoplastic activity in p53^Mut cancer models. These findings support a two-drug combination strategy to improve outcomes for patients with p53^Mut cancer.

Graphical abstract

Highlights

• •Trifluorothymidine induces DNA breaks in p53-mutant cells via post-replicative repair
• •PARP inhibitor enhances DNA breaks induced by trifluorothymidine, leading to cell death
• •Trifluorothymidine and PARP inhibitor act synergistically against p53-mutant cancers

Alruwaili et al. demonstrate that p53-mutant cancer cells accumulate DNA breaks due to post-replicative removal of thymidine analogs following DNA replication. Inhibition of PARP enhances DNA damage induced by thymidine analogs, leading to cell death. These findings support a two-drug combination strategy to improve outcomes for patients with p53-mutant cancer.

Introduction

Colorectal adenocarcinoma (CRC) and pancreatic ductal adenocarcinoma (PDAC) are among the leading causes of cancer-related death in the United States.¹ Although CRC is curable at an early stage, the majority of patients with CRC are either diagnosed with or develop metastatic disease with an expected 5-year overall survival (OS) rate of less than 15%.^2,3 Despite improvements in systemic therapy with regimens consisting of fluoropyrimidines, oxaliplatin, irinotecan, angiogenesis inhibitors, and molecularly targeted agents, most patients will succumb to this disease.^3,4,5 PDAC is also a notoriously aggressive and hard-to-treat malignancy with a 5-year OS rate of less than 8%.⁶ PDAC is routinely treated with chemotherapy such as gemcitabine plus nab-paclitaxel and the oxaliplatin, irinotecan, leucovorin, and 5-fluorouracil (5FU) combination.⁷ There is no US Food and Drug Administration (FDA)-approved drug that can extend OS more than 3 months in patients with chemotherapy-refractory, advanced CRC or PDAC.^3,8

Genetic alterations in the tumor-suppressor TP53 gene (p53) are present in most CRC and PDAC tumors.⁹ However, existing treatment options for p53-deficient cancers are ineffective and cause toxic side effects, emphasizing the need for better therapeutics.¹⁰ p53 is activated in response to DNA damage and regulates DNA repair machinery contributing to recognition of DNA damage, recruitment of repair factors, control of repair, or activation of cell death if DNA damage is irreparable.¹¹ While immense information is available on the functional consequences of p53 mutations (p53 mutant [p53^Mut]), therapeutic efforts targeting p53^Mut tumors have been largely unfruitful, and p53 status is mostly ignored in clinical management of patients.^10,12 Here, we describe a two-drug treatment strategy for p53-deficient CRC and PDAC tumors that takes advantage of their unique DNA repair abnormality. The study demonstrated that a thymidine analog, TAS102, selectively induces DNA breaks in p53-deficient cells through a mechanism involving DNA base excision repair (BER). Further, inhibitors of poly(ADP-ribose) polymerase (PARPis) enhanced double-strand breaks (DSBs) and cell death in p53-deficient cells, and the combination of TAS102 with PARPi agents showed greater anti-tumor activity in p53^Mut patient-derived xenograft (PDX) models than either drug alone, without toxic side effects.

Results

DNA replication and repair pathways in p53^Mut CRC and PDAC tumors

Aggressive clinical behavior of cancers is linked to high proliferation index and tumor mutational burden (TMB).^9,13 Stratification for p53 status of the genomic data in CRC and PDAC showed that p53 mutations are associated with elevated expression of replication-related genes (RRGs) and high TMB (Figures 1A, 1B, S1A, and S1B). DNA repair mechanisms such as mismatch repair (MMR) and BER normally remove mis-incorporated or modified nucleotides,^14,15 but their inadequate activity or genetic alterations may lead to high mutational burden. Genomic data analysis revealed that p53^Mut tumors express statistically significant higher levels of BER and MMR genes (Figures 1C and 1D). Unsupervised clustering showed that p53^Mut CRC and PDAC express elevated levels of MMR genes (i.e., MSH2 and MSH6) and BER genes, including UNG, TDG, and MUTYH (Figure S1C). Genomic data indicated a significant correlation in the expression of RRG and DNA repair genes (Figures 1F, 1G, 1I, and 1J) and BER-MMR genes (Figures 1E and 1H) in CRC and PDAC, whereas genomic alterations in the BER and MMR genes were rather infrequent (Figure S1C). These data indicated that p53^Mut CRC and PDAC exhibit high replicative activity and elevated TMB, despite high expression of DNA repair genes involved in the BER and MMR mechanisms. Thus, p53^Mut CRC/PDAC tumors may have an impediment in BER and MMR mechanisms even in the absence of genetic alterations in those repair genes.

Figure 1.

DNA replication and repair pathways in colorectal adenocarcinoma (CRC) and pancreatic ductal adenocarcinoma (PDAC)

(A) Tumor mutational burden (TMB) in CRC and PDAC stratified by p53 status.

(B–D) Expression of replication-related genes (RRG), base excision repair (BER), and mismatch repair (MMR) genes in CRC and PDAC; expression Z scores stratified by p53 status. Cell-cycle gene lists are derived from Database: Cyclebase 3.0, and BER and MMR gene lists are derived from KEGG. A comparison was made using the t test (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001).

(E and H) Correlation of BER and MMR pathways stratified by p53 status in CRC and PDAC.

(F, G, I, and J) Correlation of RRG and BER pathways (F and I) and RRG and MMR pathways (G and J) stratified by p53 status in CRC and PDAC. The correlation was calculated using Pearson’s correlation coefficient method.

BER-mediated DNA repair in p53-deficient cancer cells

The BER mechanism removes DNA base lesions arising from deamination (uracil), oxidation (e.g., 8-oxoguanine), or alkylation (e.g., 3-methyladenine).¹⁶ Mismatched nucleotides or insertions or deletions generally arise during DNA replication and are removed by the MMR mechanism, whereas p53 mutations are not typically associated with inactivation of MMR.¹⁷ Hence, we assessed the BER activity in MMR-deficient and -proficient cell lines by measuring removal of a thymidine analog (5-ethynyl-2′-deoxyuridine [EdUrd]) from genomic DNA following a 2-h pulse of EdUrd that incorporates into DNA.¹⁸ Genomic ethynyl-deoxyuracil (EdU) was fluorescently labeled using click-it chemistry and scored by flow cytometry. Isogenic HCT116 p53 wild-type (p53^WT) and p53 knockout (p53^KO) cell lines showed comparable genomic EdU levels after a pulse (Figure 2A). The EdU⁺ fraction in p53^WT cells was reduced within 24–72 h of incubation in EdUrd-free media, whereas p53^KO cells retained EdU for a prolonged amount of time (Figures 2A and 2C). Cell-cycle distribution was noticeably different in p53^WT and p53^KO cell lines. p53^WT cells after an initial increase of the G2/M fraction at 24–48 h regained, at 72 h, the cell-cycle distribution comparable to that of cells immediately after an EdUrd pulse. In contrast, EdU-pulsed p53^KO cells accumulated in G2/M at 48 and 72 h (Figure 2B). Only a small fraction of the EdU-pulsed cells were found in the sub-G1 fraction, ∼1% for both p53^WT and p53^KO at 72 h (Figure 2B).

Figure 2.

Removal of genomic thymidine analogs in HCT116 p53^WT and p53^KO cell lines after pulse (P) labeling

(A and E) Human colon cancer HCT116-p53^WT and HCT116-p53^KO cell lines were P labeled with EdUrd for 2 h or trifluorthymidine (TFT) for 4 h followed by wash and incubation for indicated time. Cells were fixed and stained for EdU using click-it chemistry, or TFT with antibody to BrdU, and analyzed by flow cytometry with Hoechst for DNA content.

(B and F) Cell-cycle data for cells treated as in (A) and (B).

(C and G) EdU⁺ or TFT⁺ populations were scored relative to initial levels.

(D) EdU⁺ cells with reduced Hoechst fluorescence (top left quartile). Comparison was made using two-way ANOVA (∗p < 0.05 and ∗∗p < 0.01; n = 3–4).

(H) Immunoblot data in HCT116-p53^WT and HCT116-p53^KO cells after 2 h EdUrd P and 24, 48, and 72 h after wash and incubation in EdUrd-free media. At least 3 biological repeats were done for EdUrd or TFT removal.

In addition, the EdU⁺ population of p53^KO cells was greater at 24 h compared to immediately after a pulse (Figure 2C; 0 h). This increase in genomic EdU in p53^KO cells may arise from cycles of futile BER-mediated DNA repair, a well-known phenomenon for 5-fluorouridine analogs.¹⁹ Metabolites of EdUrd and fluorinated analogs (EdUMP and FdUMP) inhibit thymidylate synthase (TS),^20,21 thereby stimulating incorporation of uridine residues into DNA and, consequently, activating BER.^19,22 Multiple BER cycles led to an accumulation of apyrimidinic sites, disruption of A-T pairs, and formation of DNA breaks.¹⁹ This hypothesis was validated with Hoechst dye, which selectively binds to AT-rich regions.²³ The Hoechst intensity in the S-phase population of p53^KO cells was decreased at 24–48 h, indicating a reduced binding of Hoechst dye to DNA due to generation of apyrimidinic sites and disruption of A-T pairs (Figure 2D). Comparable results were obtained in EdU-removal assays with MMR-proficient p53^Mut cell lines, i.e., a prolonged retention of genomic EdU and disruption of A-T base pairs (Figures S2A and S2C–S2E).

Next, we evaluated the removal of trifluorothymidine (TFT), a component of the TAS102 regimen approved by the FDA for treatment of advanced CRC. Flow cytometry assays showed that TFT removal was markedly delayed in p53^KO cells compared to p53^WT cells (Figures 2E and G). In p53^WT cells, the cell-cycle distribution was recovered at 72 h post-TFT incubation to control levels, whereas a large fraction of p53^KO cells was still in G2/M phase at 72 h (Figures 2F, S2F, and S2G). This result was consistent with our findings for EdU (Figure 2B).

Examination of p53 and DNA damage markers showed that exposure of p53^WT cells to EdUrd resulted in a transient activation of p53 (Figure 2H), an increase of TS, and phosphorylation of H2AX at Ser139 (γH2AX), a marker of DNA DSBs.²⁴ In contrast, HCT116 p53^KO cells responded to EdUrd with a prolonged induction of TS and the DNA damage marker for at least 72 h post-exposure (Figure 2H). Taken together, these data showed that p53^WT cells effectively removed genomic thymidine analogs through BER, while p53-deficient cells exhibited a prolonged delay in the analog removal, although the rates of analog removal were comparable in both cell lines. In p53^WT cells, the thymidine analog induced a transient activation of p53 and the DNA damage response, whereas p53-deficient cells responded with buildup of DNA damage and accumulation in G2/M phase.

DNA damage induced by clinical thymidine analogs in p53-deficient cells

We investigated further the response to clinically relevant thymidine analogs, floxuridine (FdUrd), and TFT, a component of TAS102. Treatment with TAS102 revealed a transient upregulation of DNA damage (γH2AX levels) and the p53-p21 signaling axis in p53^WT HCT116 and RKO cell lines (Figures 3A and 3B), whereas isogenic p53^KO cells accumulated DNA damage (γH2AX). A comparable response was observed with FdUrd (Figure 3C). Next, the response was assessed in MMR-proficient CRC and PDAC cell lines carrying p53^Mut. Treatment with TAS102 (Figures 3E and D) or FdUrd (Figures 3F and 3G) induced phosphorylation of p53 and accumulation of the DNA damage marker in all p53^Mut cell lines tested. As expected, p21/CDKN2A was not induced in p53^Mut cells. Thus, inactivation of p53 leads to accumulation of DNA damage upon exposure of p53-deficient CRC and PDAC cells to thymidine analogs.

Figure 3.

Accumulation of DNA damage in response to thymidine analogs in p53^Mut cells

(A, B, D, and E) Cells were treated with 0.5 μM TAS102 for indicated time.

(C, E, and F) Cells were treated with 5-fluoro-2′-deoxyuridine (FdUrd; 5 μM) for indicated time. Whole-cell lysates were probed with markers of p53 signaling and DNA damage marker. At least two biological repeats were done.

Inhibition of PARP promotes accumulation of G2-phase population in p53-deficient cells

We asked whether DNA damage induced by thymidine analogs in p53-deficient cancer cells could be further enhanced by interfering in BER-mediated DNA repair. Modified nucleotides (including thymidine analogs: 5-ethynyl-, 5-fluoro-, or 5-trifluoromethyl-uridine) are removed from genomic DNA by BER.^18,25 DNA glycosylases (e.g., UNG) recognize and excise modified bases, generating apurinic-pyrimidinic sites (AP-sites), followed by a single-strand DNA cleavage mediated by AP-endonuclease.²⁶ Subsequently, PARP1 binds to single-strand breaks (SSBs) and initiates recruitment of enzymes to carry out repair and restoration of the original DNA sequence.²⁶ Thus, inhibition of PARP1 may lead to accumulation of DNA breaks upon exposure to modified thymidine analogs in p53-deficient cells. To test this hypothesis, the cytotoxicity of two clinical PARPis, olaparib and talazoparib, was examined in CRC and PDAC cell lines. The cytotoxicity assays showed comparable IC50 values in p53^WT and p53-deficient CRC and PDAC cell lines, ranging from 2 to 4 μM for talazoparib and from 10 to 18 μM for olaparib (Figures 4A and 4B). The difference in IC50 values between two PARPi agents is likely associated with the greater DNA trapping ability of talazoparib compared to olaparib, while both compounds have comparable inhibition of enzymatic activity of PARP1.^27,28 Non-tumor human pancreatic nestin-expressing (HPNE) epithelial cells were the least sensitive to PARPi agents. The inhibitory activity of PARPi agents was also evaluated using antibodies that detect PARylation in proteins. Immunoblotting revealed that PARylation activity was inhibited by talazoparib at 50 nM and by olaparib at 100 nM in all tested cell lines, independent of the p53 status (Figures 4C–4E). Further, PARPi agents induced the p53-p21 axis in p53^WT cells, while p21 was not upregulated in p53-deficient CRC and PDAC cell lines, with a minimal impact on γH2AX levels (Figures 4C—4E). Cell-cycle assays showed that PARPi agents increased the G1 population in p53^WT cells (Figure 4F), whereas p53-deficient cells responded to PARPi agents with accumulation in G2 phase and this effect was greater in response to talazoparib (Figures 4G–4I). Together, these data demonstrated that PARPi agents induce p53-p21 signaling and increase the G1 population in p53^WT cells, whereas p53-deficient cells accumulated in G2 but showed limited DNA damage that is in line with a proficient status in homologous recombination (HR) in the tested cell lines.

Figure 4.

Inhibition of PARP induces G2 accumulation in p53^Mut cells and activates the p53-p21 axis in p53^WT cells

(A) Cytotoxicity curves for PARP inhibitors (PARPis) talazoparib and olaparib.

(B) Summary of IC50 cytotoxicity values for talazoparib and olaparib in the tested cell lines.

(C–E) Immunoblots of whole-cell extracts from cell lines treated with talazoparib (C–E) or olaparib (D and E) for 24 h at indicated concentrations.

(F–I) Cell-cycle data for cell lines treated with the PARPi olaparib (200–500 nM) or talazoparib (10–100 nM) or vehicle control for 24 h. Comparison was made using two-way ANOVA (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, ∗∗∗∗p < 0.0001). All experiments were repeated at least two times. Cytotoxicity experiments were done at least twice with six technical replicates.

DNA damage response to a two-drug combination

To explore the potential of drug interactions, we examined whether PARP inhibition enhances DNA damage induced by thymidine analogs in p53-deficient cells. First, we tested the response to FdUrd alone or in combination with PARPi at concentrations that block PARylation. In p53^WT HCT116 cells, FdUrd alone activated the p53-p21 axis and induced a transient increase in γH2AX levels, whereas PARPi alone increased p53 but did not induce appreciable levels of γH2AX (Figure S3A). These responses were not enhanced by the FdUrd-PARPi combination (Figure S3A). In contrast, FdUrd induced γH2AX levels, and this response was enhanced by PARPi in HCT116-p53^KO cells (Figure S3B). In p53^Mut CRC cell lines (HT29 and DLD1), PARPi strongly enhanced FdUrd-induced γH2AX levels (Figures S3C and S3D). As expected, neither treatment increased p21 in p53^KO and p53^Mut cells (Figures S3B–S3D).

Next, the response to TAS102 was examined in combination with PARPi. TAS102, alone or in combination with talazoparib, induced a transient induction of the p53-p21 axis and DNA damage response in p53^WT RKO cells (Figure 5A), whereas in isogenic p53-deficient RKO cells, the combination induced greater DNA damage (γH2AX levels) compared to either drug alone (Figure 5B). In p53^Mut CRC and PDAC cell lines, a TAS102-PARPi combination likewise showed enhanced DNA damage response (Figures 5C and 5D). To confirm DNA damage further, markers of double-strand DNA breaks (γH2AX and RAD51) were assessed by immunofluorescence (Figures 5E–5H). These assays revealed that talazoparib (PARPi) enhanced the intensity of both DNA break markers (γH2AX and RAD51) induced by TAS102 in p53-deficient HCT116 cells, whereas in p53^WT HCT116 cells, PARPi did not enhance these markers (Figures 5G and 5H). Microscopy showed co-localization of RAD51 and γH2AX, indicating the formation of DSBs (Figure 5F).

Figure 5.

Inhibition of PARP enhances DNA damage induced by TAS102 in p53^Mut cells

(A–D) Immunoblots of whole-cell extracts from cell lines treated with vehicle control (C), 100 nM talazoparib (T), 500 nM TAS102 (TAS), or the drug combination (TT) for indicated time.

(E) Immunofluorescence images of cells stained for γH2AX (green), RAD51 (red), and genomic DNA (Hoechst, blue). Cells were treated with vehicle control, 100 nM talazoparib (Tala), 500 nM TAS102, or their combination (TAS+Tala) for 48 h. Images were taken using a 100× lens. Scale bar: 20 μm.

(F) Enlarged images of highlighted areas in HCT116-p53^KO.

(G and H) The corrected total cellular fluorescence (CTCF) was determined for γH2AX (G) and RAD51 (H) fluorescence of at least 40 cells/group. Representative median scattered dot plots are shown. Comparison was made using the two-way ANOVA (∗p < 0.05, ∗∗∗∗p < 0.001).

(I and J) Immunoblots of whole-cell extracts from HT29 and RKO cell lines treated with vehicle control (C), 100 nM talazoparib (TAL), 500 nM TAS102 (TAS), their combination (TT), or 50 nM nocodazole (Nc) for 24 h.

(K) A model of DNA damage response (DDR) signaling induced by the TAS102-PARPi combination. Experiments were done at least two times.

Microscopic examination did not show accumulation of genomic instability marks in response to the two-drug combination (i.e., micronuclei and chromosome bridges; Hoechst stain) in p53^Mut cells (Figure 5F), while the flow cytometry data indicated accumulation of cells in G2 phase in response to thymidine analogs (Figures 2B and F) and PARPi (Figures 4G–4I). These observations suggested that the combination induces the G2 checkpoint. This hypothesis was examined in several cell lines by assessing markers of DNA damage response and the G2 checkpoint. In HT29 cells, the drug combination strongly induced markers of DNA damage response (phosphorylation of CHK1 and H2AX) and the G2 checkpoint (cyclin A2 and phosphorylation of CDC2/CDK1) (Figure 5I). This finding indicated activation of the G2 checkpoint by the TAS102-PARPi combination. In support of this, the combination did not increase phosphorylation of histone H3, indicating that cells did not move to mitosis, whereas in the control, nocodazole induced high levels of phospho-H3, indicating mitotic arrest (Figure 5I). These findings were validated in MiaPaca2 and RKO cell lines (Figures S4A and S4B). DNA damage response signaling was further investigated in RKO cells. Treatment with TAS102-talazoparib induced p53 levels, phospho-p53, DNA damage (γH2AX), and G2-checkpoint markers (phospho-CHK1 and phospho-CDC2/CDK1) in RKO cells (Figure 5J). A potent and selective ATMi (KU55933) reduced p53 total levels and phospho-p53 as well as phospho-H2AX induced by the TAS102-PARPi combination, indicating that ATM mediates activation of p53 and γH2AX signaling (Figure 5J). A selective ATRi (AZD6738) blocked induction of phospho-CHK1 and phospho-CDC2 but showed only a limited effect at higher doses on p53 (Figure 5J). Notably, ATRi increased γH2AX levels, indicating abrogation of the G2 checkpoint and increased DNA damage (Figure 5J). This latter response is likely associated with death of mitotic cells with incomplete DNA repair, also called mitotic catastrophe.²⁹ The experiments with ATMis and ATRis in p53-deficient cell lines further confirmed the role of ATM in γH2AX signaling and the role of ATR in induction of the G2 checkpoint in response to the TAS102-PARPi combination (Figures S4C and S4D). Together, these observations strongly indicate that the TAS102-PARPi combination activates double-strand DNA damage response signaling mediated by ATM kinase leading to activation of p53 (phospho-p53) and DNA repair (γH2AX), while ATR kinase is mainly responsible for the G2 checkpoint (phospho-CHK1, phospho-CDC2, cyclin A2). A working model is shown in Figure 5K.

Furthermore, the effect of the drugs on cell death was evaluated in caspase-3/7 (Casp-3/7) activity assays. TAS102 alone showed comparable induction in the Casp activity in p53^WT and p53^KO cell lines, while talazoparib alone had no effect (Figures S3E and S3F). The TAS102-talazoparib combination greatly induced Casp-3/7 activity in p53-deficient RKO and HCT116 cell lines compared to either drug alone (Figures S3E and S3F). Together, these data indicate that PARP inhibition cooperates with thymidine analogs in the induction of DNA damage and cell death in p53-deficient cells while activating the p53-p21 axis in p53^WT cells.

PARPi synergizes with thymidine analogs in cytotoxicity against p53^Mut tumor cells

To define the outcome of enhanced DNA damage and cell death by the two-drug combination in p53-deficient cells, we examined the cytotoxicity responses to thymidine analogs and PARPi. TAS102 showed comparable IC50 values for p53^KO and p53^WT HCT116 and RKO cell lines (Figures 6A, 6B, and 6G). Importantly, PARPi at the concentrations inhibiting PARylation activity strongly enhanced cytotoxicity of TAS102 against p53^KO HCT116 and RKO cell lines, whereas the two drugs did not show interaction in isogenic p53^WT cell lines (Figures 6A, 6B, and 6G). Evaluation of the drug interaction using isobolograms revealed a combination index (CI) ranging from 0.16 to 0.40 in p53-null and p53^Mut CRC cell lines, indicating synergism, while p53^WT cell lines showed CI values close to 1.0, indicating the absence of a drug interaction (Figures 6D–6G). A synergistic interaction of TAS102 and PARPi drugs was also observed in p53^Mut PDAC cell lines (CI = 0.12–0.32), whereas no drug interaction was found in the non-malignant human pancreatic ductal hTERT-HPNE cell line (Figures 6G, S5G, and S5H). Further, PARPi enhanced cytotoxicity of thymidine analog FdUrd in p53^Mut CRC cell lines (Figures S5A–S5F). Together, these findings indicate that thymidine analogs synergistically interact with PARPi agents in cytotoxicity against p53-deficient CRC and PDAC tumor cells (Figure 6H), while no interaction was found in p53^WT tumor and non-tumor cells.

Figure 6.

Inhibition of PARP enhances cytotoxicity of TAS102 in p53-deficient CRC and PDAC cells

(A–C) Cytotoxicity curves in CRC cell lines for TAS102 alone and in combination with PARPi talazoparib (50, 100, and 500 nM).

(D–F) Isobolograms in CRC cell lines for TAS102 (y axis, fraction of IC50 TAS102) and for talazoparib (x axis, fraction of IC50 talazoparib) and mean combination index (CI).

(G) Summary of IC50 and CI values obtained from cytotoxicity assays repeated at least two times with six technical replicates.

(H) Model of the therapeutic strategy for p53^Mut cancers using the TAS102 and PARPi combination. Thymidine analog (i.e., TFT, a component of TAS102) is incorporated into DNA during replication, followed by BER-mediated removal resulting in a single-strand break (SSB) formation. PARP inhibition delays SSB repair, leading to double-strand breaks (DSBs) that activate DDR signaling mediated by ATM and ATR kinase pathways. ATM kinase mediates activation of the p53-p21 axis and DNA repair. ATR kinase controls the G2 checkpoint. The p53-p21 axis contributes to G1 and G2 checkpoints, while inactivation of p53 abrogates the G1 checkpoint, leading to accumulation of DNA damage induced by the analog. Accumulated DNA damage leads to cell death in p53^Mut cells.

Anti-tumor efficacy of the TAS102-PARPi treatment

We tested the anti-tumor efficacy of the TAS102-PARPi combination in CRC and PDAC models that do not carry genetic alterations in BRCA1/BRCA2. In the CRC HT29 model (p53^R273H, RAF^V600E), treatment with either TAS102 or talazoparib alone showed a limited effect on tumor growth, whereas the TAS102-talazoparib combination markedly reduced tumor growth (p < 0.001) and improved survival (Figures 7A and 7B). In the PDAC MiaPaca2 model (p53^R248W; KRAS^G12C), olaparib monotherapy did not affect tumor growth, and TAS102 alone showed a limited effect, whereas the TAS102-olaparib combination strongly reduced tumor growth and improved survival (Figures 7C and 7D). Next, the response to a two-drug therapy was examined in PDX models established from patients with PDAC or CRC. In PDAC PDX-14312-4p, carrying p53^R175H and KRAS^G12R mutations, olaparib monotherapy did not influence tumor growth, and TAS102 alone showed a limited effect (Figure 7E). The TAS102-olaparib combination markedly reduced tumor growth and extended survival by 27 days (97 vs. 70 days) compared to TAS102 alone (Figures 7E and 7F). Histological analysis of PDX PDAC tumors showed a reduction in proliferation Ki67 index and an increase in DNA damage (γH2AX) in the combination group compared to either drug alone (Figures 7G, 7H, and S6A). Assessment of active Casp3 showed a marked increase in this apoptotic marker (Figure 7I). TFT incorporation into DNA was comparable in TAS102 and TAS102-PARPi groups (Figure S6B). The tumor response to the TAS102-talazoparib combination was examined in the CRC PDX models PDX-01-03-RP-3p, carrying p53^WT, and PDX-02-07-RP-3p, carrying the p53^H179R mutant. In the p53^WT PDX model, TAS102 monotherapy and the combination showed similar effects on tumor growth and survival (Figures 7J and 7K). Histological analysis showed a comparable reduction in the proliferation Ki67 index and an increase in DNA damage by TAS102 monotherapy and the TAS102-PARPi combination, while Casp-3 activity was marginally elevated in the TAS102-PARPi regimen (Figures S8A–S8D). These findings were consistent with retrospective clinical data for CRC³⁰ and our cytotoxicity data (Figure 6). In the p53^H179R PDX model, TAS102 showed a limited effect, while the TAS102-talazoparib combination exhibited greater inhibition of the tumor growth and extended survival compared to TAS102 or talazoparib monotherapy (Figures 7L and 7M). The TAS102-talazoparib combination reduced the proliferation Ki67 index and increased DNA damage (γH2AX) and cell death (active Casp-3) (see Figures S7B and S7C). Evaluation of drug toxicities showed that neither monotherapies nor the combination affected mouse weight during the course of the treatments in the tumor xenograft studies, indicating low toxic side effects of the drug combinations (Figures S6C and S7A). Assessment of peripheral blood in the PDAC PDX study showed no significant changes in lymphoid and myeloid cell populations, confirming the low toxicity of the drug regimen (Figure S6D). Together, these findings demonstrated greater anti-tumor efficacy of the two-drug combination compared to either drug alone in the context of p53^Mut CRC and PDAC models.

Figure 7.

Anti-tumor activity of TAS102 in combination with PARPi agents in CRC and PDAC in vivo models

(A) Colon cancer HT29 cells were implanted subcutaneously (s.c.) into SCID mice (10 mice/group). When tumor size reached 100 mm³, mice were treated by daily oral gavage with vehicle control, talazoparib (0.15 mg/kg), TAS102 (50 mg/kg), or TAS102+talazoparib combination (50 mg/kg + 0.15 mg/kg) with the schedule 5 days on, 2 days off. Tumor size was measured two times per week. Comparisons were made using two-way ANOVA (∗p < 0.05).

(B) Survival was evaluated using Kaplan-Meier estimator based on time to arrive at 500 mm³ of tumor size. Median survival: 21 (vehicle), 23 (talazoparib), 25 (TAS102), and 35.5 days (TAS102+talazoparib). Comparison was made using the log rank test.

(C) Pancreatic cancer MIAPACA-2 cells were implanted s.c. into SCID mice (5 mice/group). When tumor size reached 100 mm³, mice were treated by daily oral gavage with vehicle control, olaparib (50 mg/kg), TAS102 (50 mg/kg), or TAS102+olaparib combination (50 mg/kg each drug) with the schedule 5 days on, 2 days off. Tumor size was measured two times per week. Comparison was made using two-way ANOVA (∗p < 0.05, ∗∗p < 0.01).

(D) Survival was evaluated using Kaplan-Meier estimator based on time to arrive at 600 mm³ of tumor size. Median survival: 21 (vehicle, olaparib), 29 (TAS102), and 40 days (TAS102+olaparib). Comparison was made using the log-rank test.

(E) Patient-derived tumor xenografts (PDXs) from a patient with PDAC (PDAC-14312-4p) were implanted s.c. into SCID mice (10 mice/group). Mice were treated and tumor sizes were measured and evaluated as described in (C).

(F) Survival was evaluated using Kaplan-Meier estimator based on time to arrive at 750 mm³ of tumor size. Median survival: 58 (vehicle), 68 (olaparib), 70 (TAS102), and 97 days (combination). Comparison was made using the log-rank test.

(G–I) Immunohistochemistry (IHC) evaluation of proliferation Ki67 index, DNA damage (γH2AX), and active caspase-3. Three tumors per group, at least 9 fields per stain. Comparison was made using two-way ANOVA (∗∗∗p < 0.001).

(J and K) Colorectal cancer PDX specimens (PDX-01-03-RP-3p) were implanted s.c. into NSG mice (10 mice/group). Mice were treated by oral gavage with vehicle, talazoparib (0.15 mg/kg), TAS102 (50 mg/kg), or TAS102+talazoparib (50 mg/kg + 0.15 mg/kg) with the schedule 5 days on, 2 days off. (K) Survival was evaluated using Kaplan-Meier estimator based on time to arrive at 550 mm³ of tumor size, and comparisons were made as described above (B, D, and F).

(L) Colorectal cancer PDX specimens (PDX-02-07-RP-B-3p) were implanted s.c. into NSG mice (10 mice/group). Mice were treated as described in (J) and (K). Comparison was made using two-way ANOVA (∗p < 0.05, ∗∗p < 0.01).

(M) Survival was evaluated using Kaplan-Meier estimator based on time to arrive at 750 mm³ of tumor size. Median survival: 16 (vehicle), 20 (talazoparib), 28 (TAS102), and >40 days (combination). Comparison was made using the log-rank test.

Discussion

Treatment of CRC and PDAC is a significant clinical problem, in part due to lack of effective targeted therapy for the majority of patients. Although genetic alterations in p53 are found in most cancers, including CRC and PDAC, only a limited number of experimental therapeutic strategies exploit this genetic abnormality.^10,12 Here, we present evidence for a two-drug strategy that specifically targets p53^Mut tumors and improves survival in preclinical CRC and PDAC models (Figure 6H). The study revealed that p53^Mut cancer cells are compromised in the repair of DNA lesions upon incorporation of thymidine analogs. These analogs did not interfere with DNA replication, but rather their removal by the BER-mediated mechanism was considerably delayed in p53-deficient cancer cells, leading to accumulation of DNA breaks. Inhibition of PARP further enhanced DNA damage induced by thymidine analogs selectively in p53-deficient cancer cells, whereas in p53^WT cells, PARPi agents activated the p53-p21 axis, leading to accumulation of p53^WT cells in the G1 phase. In preclinical in vivo experiments, the combination of the thymidine analog TAS102 and PARPi showed greater anti-tumor activity in p53^Mut CRC and PDAC PDX mouse models, without toxic side effects.

The two-drug treatment (Figure 6H) is initiated by incorporation into DNA of a thymidine analog (i.e., TFT, the main component of TAS102), followed by BER-mediated removal of a modified base and formation of SSBs). TFT belongs to a class of fluorinated uridine analogs that includes FdUrd and 5FU, which have been utilized in clinic for several decades.^31,32 TFT and FdUrd are incorporated into DNA, whereas 5FU is predominantly incorporated into RNA.²⁶ Similar to the other fluorinated analogs, TFT inhibits TS via a common metabolite, fluorinated dUMP, thereby reducing dTTP levels and promoting incorporation of dUTP and the analog into the genome,^26,32 although TFT has a lower inhibitory capacity for TS compared to the other analogs.^20,21 The thymidine analogs and uracil are removed by DNA glycosylases (i.e., UNG, TDG, SMUG1), initiating the BER process,^26,33 which involves multiple rounds of BER in the presence of the fluorinated analogs,¹⁹ as evidenced by increased disruption of A-T pairs (Figure 2). In non-tumor cells, BER-produced DNA lesions activate the p53-p21 axis, resulting in G1 arrest, whereas in p53-deficient cells, this response is compromised, therefore leading to the buildup of DNA lesions, demonstrated by elevated γH2AX levels and accumulation of the G2 population compared to untreated cells. The BER mechanism is evidently effective in p53^WT and p53-deficient cells, as comparable removal rates for thymidine analogs were observed in both cell lines (Figure 2). The key reason for accumulation of DNA damage in p53-deficient cells is likely to be loss of the replication control via the p53-p21 axis. In support of this notion, inactivation of the p53-p21 axis by depletion of either p53 or p21 in non-tumor p53^WT cells was sufficient to confer the response to the thymidine analogs comparably to p53^Mut cancer cells.¹⁸ These findings are consistent with the role of p53 in DNA repair^11,12 and the ability of p21 to negatively regulate PCNA-dependent DNA replication.³⁴ Thus, the treatment with TFT selectively targets p53-deficient cancer cells, while non-tumor tissues effectively repair DNA lesions. In comparison, radiation and conventional chemotherapeutic agents affect rapidly dividing cells and, therefore, are unable to discriminate between malignant and non-malignant cells.

PARPi agents further enhanced DNA breaks and cytotoxicity of the thymidine analogs in p53-deficient cancer cells at doses that inhibited PARP activity and over 20 times less than IC50 values (Figures 5 and 6). Mechanistically, the collision of the replication fork with SSBs induced by BER-mediated removal of a thymidine analog likely to result in DSBs. PARP activity is required for SSB repair induced by BER as well as for restarting the stalled replication fork.²⁷ PARPi agents can also disrupt processing of Okazaki fragments, leading to replication-associated single-strand DNA gaps.³⁵ In addition, PARP trapping at single-strand DNA breaks can cause replication fork stalling or collapse and lead to formation of one-ended DSBs.^27,36 DSB repair in S phase is typically mediated by HR, and PARPi agents have been used for the treatment of cancers with HR deficiency that include patients with PDAC carrying germline pathogenic BRCA1/2 mutations.³⁷ The same concept can theoretically be applied to CRC with HR deficiency.^38,39 However, only a small proportion of patients have pathogenic BRCA1/2 mutations: 4%–8% of patients with PDAC and 5% of patients with CRC.^37,40,41 In addition, PARPi resistance mechanisms are frequently associated with restoration of HR activity via reverse mutations in BRCA1/2, PALB2, and RAD51C.²⁸ However, all models tested in this study carry normal BRCA1/2 genes and are HR proficient. Thus, the two-drug treatment strategy (Figure 6H) expands the utility of PARPi agents to the majority of CRC and PDAC that are p53 deficient but have normal BRCA1/2 status.

The thymidine analog-PARPi combination (TAS102-PARPi) induces DNA damage response signaling and activation of p53 via a mechanism requiring ATM kinase (Figures 5I–5K and S4). This observation is consistent with a known role of the ATM-CHK2 axis in the cellular response to DNA DSBs to mediate initiation of DNA repair by the HR mechanism.⁴² The ATM-CHK2 axis may stabilize and activate p53 by various mechanisms including disruption of the p53-MDM2 interaction, inactivation of MDMX, and phosphorylation of HNRNPK.⁴³ In addition, the TAS102-PARPi combination activated the G2 checkpoint through a mechanism involving the ATR-CHK1 axis (Figures 5I, 5J, and S4). Blockade of ATR in cells treated with TAS102-PARPi strongly increased γH2AX and phospho-HH3 levels, indicating that the cells are released from the G2 checkpoint, leading to mitotic cell death. These findings suggest a therapeutic option for p53^Mut cancers by combinatorial application of TAS102-PARPi and ATR blockade given the availability of clinical drugs for ATR.

Genomic TCGA data showed high expression levels of RRGs in p53^Mut CRC and PDAC tumors, and these levels correlated with expression of DNA repair BER and MMR genes (Figure 1). High replication activity consumes significant nucleotide resources, provoking incorporation of mismatched, ribo-, or modified nucleotides into DNA,^14,15 and may lead to replication stress and high mutational burden.^{44,45,46,47,48} Apparently, high expression of BER and MMR genes counteracts this effect, in agreement with their role in repair of those DNA lesions.^16,49

Our two-drug therapeutic strategy provides a strategic opportunity for treatment of p53-deficient CRC and PDAC. Clinical data demonstrated that TAS102 exhibits efficacy in the treatment of metastatic colorectal cancer that progressed on standard therapy, including 5FU.⁵⁰ However, retrospective clinical data indicate a limited clinical benefit of TAS102 in refractory metastatic p53^Mut CRC compared to p53^WT CRC.³⁰ Consistent with this clinical observation, our preclinical studies showed that TAS102 monotherapy is effective against p53^WT tumor xenografts, whereas it is ineffective against p53^Mut cancer models (Figure 7). Furthermore, in all tested p53^Mut tumor xenograft mouse models, the TAS102-PARPi combination greatly reduced tumor growth and enhanced survival (Figure 7). Together, these observations strongly support the TAS102-PARPi combination as a therapeutic option for p53-deficient CRC and PDAC cancers. There are five clinical trials testing TAS102, alone or in combination with other anti-neoplastic agents, in patients with PDAC. While outcomes are not yet available for three of these studies, one TAS102 monotherapy study was terminated early without apparent anti-tumor efficacy (ClinicalTrials.gov: NCT02921737). The addition of bevacizumab (Avastin) improves the efficacy of TAS102 but also increases hematologic toxicity. However, responses are infrequent, and there is no biomarker-defined population that may derive benefit from this approach.^51,52 In addition, the combination of TAS102 with gemcitabine and nab-paclitaxel does not appear to add any appreciable clinical benefit.⁵³ In contrast, our work demonstrates that the TAS102-PARPi combination has strong activity against p53^Mut CRC and PDAC compared to TAS102 and PARPi monotherapies (Figure 7). Presently, this two-drug combinatorial strategy is being tested in a phase 1 clinical study for refractory CRC using TAS102 and talazoparib (ClinicalTrials.gov: NCT04511039). Tumor-suppressor p53 is mutated in most solid tumors.⁹ While existing treatment options for p53^Mut cancers are largely ineffective and cause toxic side effects,¹⁰ the results of this work and our therapeutic strategy (Figure 6H) can be applicable to different types of cancers beyond CRC and PDAC.

Limitations of the study

The study demonstrates an increased efficacy of the TAS102-PARPi combination against p53^Mut CRC and PDAC metastatic cell lines and PDX models. Although our prior work showed that this combination blocks spontaneous breast cancer metastases,¹⁸ the current study did not directly assess the effect of the combinatorial drug treatment on metastatic burden (e.g., liver) in PDAC and CRC in vivo models. This question will be the subject of future investigations.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
anti-p53 (Mouse Monoclonal antibody, Clone do-1)	Santa Cruz Biotechnology	Cat# sc-126; RRID:AB_628082
anti-p21 (CDKN1A), F-5 mouse monoclonal	Santa Cruz Biotechnology	Cat# sc-6246; RRID:AB_628073
anti-phospho-Ser15 p53 (rabbit, polyclonal)	Cell Signaling Technology	Cat# 9284; RRID:AB_331464
anti-PAR (mouse monoclonal anti-PADPr, Clone 10h)	Santa Cruz Biotechnology	Cat# sc-56198, RRID:AB_785249
anti-PARP1 (mouse monoclonal antibody, Clone 5a5)	Santa Cruz Biotechnology	Cat# sc-56197; RRID:AB_630080
anti-phospho-Ser139 H2AX (γH2AX; rabbit polyclonal)	Abcam	Cat# ab-11174; RRID:AB_297813
anti-Thymidylate Synthase (TS106)	Millipore	Cat# MAB4130, RRID:AB_2210729
anti-phospho-Tyr15-CdC2	Santa Cruz	Cat# sc-136014, RRID:AB_2016914
anti-phospho-Chk1 (Ser325)	Cell signaling	Cat# 2348, RRID:AB_331212
anti-phospho HH3 (Ser10)	Cell signaling	Cat# 31261, RRID:AB_3076725
anti-Cyclin A2	Invitrogen	Cat# MA1-180, RRID:AB_2633305
anti-bromodeoxyuridine, anti-BrdU	Dako	Cat# M0744, RRID:AB_10013660
anti-cleaved Caspase-3 (Asp175)	Cell signaling	Cat# 9661, RRID:AB_2341188
Rabbit anti-gamma-H2AX (phospho-Ser139) Antibody	Bethyl	Cat# A300-081A, RRID:AB_203288
anti-Human Ki67/MKI67 Antibody	R&D Systems	Cat#: IC7617S, RRID:AB_2889352
anti-Rad51 antibody	Abcam	Cat# ab133534, RRID:AB_2722613
anti-phospho-Ser139 H2AX (yH2AX; mouse mAb, JBW301)	Millipore	Cat# 05–636; RRID:AB_309864
anti-BrdU antibody (3D4), FITC conjugate	Biolegend	Cat#364104; RRID:AB_2564481
Goat Anti-Rabbit IgG (H L)-HRP Conjugate antibody	Bio-Rad	Cat# 170–6515; RRID:AB_11125142
Goat Anti-Mouse IgG (H L)-HRP Conjugate antibody	Bio-Rad	Cat# 170–6516; RRID:AB_11125547
Biological samples: Patient derived xenografts (PDXs)
PDAC PDX-14312-4p	Dr. Robert Straubinger (University at Buffalo)	SUNY at Buffalo
CRC PDX-01-03-RP-3p	Dr. Andrei Bakin (Roswell Park)	PDX-01-03-RP
CRC PDX-02-07-RP-B-3p	Dr. Andrei Bakin (Roswell Park)	PDX-02-07-RP-B
Chemicals, peptides, and recombinant proteins
Olaparib (Synonyms: AZD2281; KU0059436)	MedChemExpress	Cat# HY-10162
Talazoparib (Synonyms: BMN-673)	Adooq	Cat# A15533
TAS102 (trifluorthymidine (trifluridine)/tipiracil HCl mix)	Adooq	Cat# A11243
Floxuridine (5-fluoro-2′-deoxyuridine; FdUrd)	MedChemExpress	Cat# HY-B0097
5-ethynyl-2′-deoxyuridine (EdUrd)	Lumiprobe	Cat# 20540
Cy3-azide	Lumiprobe	Cat#B1030
CellTrace Violet Dye	Invitrogen	Cat#C34557
AFDye 488 Azide	Click Chemistry Tools, Scottsdale, AZ	Cat#1275-1
THPTA (tris-hydroxypropyltriazolylmethylamine)	Click Chemistry Tools, Scottsdale, AZ	Cat#1010-500
RNase A (10 mg/mL)	Thermo Fisher Scientific	Cat#EN0531
D-luciferin	Gold Biotechnology, St Louis	LUCK-1G
Copper sulfate, Cu(II)SO4	Sigma-Aldrich	Cat# 61230
Hydroxyurea	Sigma-Aldrich	Cat#55291
(2-Hydroxypropyl)-β-cyclodextrin (HPCD; Cavasol)	Sigma-Aldrich	Cat#778907
Bovine serum albumin (BSA)	Sigma-Aldrich	Cat#A2153
Dimethyl sulfoxide (DMSO)	Sigma-Aldrich	Cat#D8418
Boric acid	Sigma-Aldrich	Cat#B6768
Propidium iodide	Sigma-Aldrich	Cat#P4170
Hoechst 33342	Sigma-Aldrich	Cat#B2261
Methylene Blue hydrate	Sigma-Aldrich	Cat#M9140-100G
Paraformaldehyde	Sigma-Aldrich	Cat#158127
Dialyzed FBS (fetal bovine serum)	Gibco	Cat#A33820
ECL Western Blotting Substrate, Pierce	Thermo Fisher Scientific	Cat# 32106
SuperSignal™ West Atto Ultimate Sensitivity Substrate	Thermo Fisher Scientific	Cat#A38555
SDS (Sodium Dodecyl Sulfate)	Bio-Rad	Cat#1610301
DC™ Protein Assay Kit I	Bio-Rad	Cat#5000111
Tween 20, 100% Nonionic Detergent	Bio-Rad	Cat#1706531
Matrigel Matrix GFR	VWR International	Cat# 47743-720
Critical commercial assays
Caspase-Glo® 3/7 Assay System	Promega	Cat#G8090
Deposited data
Gene Expression and Mutation Counts	Pan-CANCER-TCGA	Cbioportal.org
Experimental models: Cell lines
HCT-116 (p53-WT)	Dr. Bert Vogelstein (Johns Hopkins)	RRID:CVCL_0291
HCT-116-p53-KO; HCT-116 TP53(−/−); HCT116-p53-null	Dr. Bert Vogelstein (Johns Hopkins)	RRID:CVCL_HD97
RKO	Dr. Alessandro Carugo (MD Anderson Cancer Center)	RRID:CVCL_0504
RKO-p53(−/−); RKO-p53KO; RKO-p53-null	Dr. Alessandro Carugo (MD Anderson Cancer Center)
HT29; HT-29	ATCC	RRID:CVCL_0320
MIAPACA2; MiaPaCa-2	ATCC	RRID:CVCL_0428
SUIT2; Suit-2	ATCC	RRID:CVCL_3172
HPNE; hTERT-Human Pancreatic Nestin-Expressing cells	ATCC	RRID:CVCL_C466
DLD1	ATCC	RRID:CVCL_0248
PANC-1	ATCC	RRID:CVCL_0480
Experimental models: Organisms/strains
Mouse: SCID/CB17, Balb/c background	Roswell Park Colony	N/A
Mouse: NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ (NSG) mice, Balb/c background	Jackson Laboratory	N/A
Software and algorithms
GraphPad PRISM 10.1.2	GraphPad Software	https://www.graphpad.com/
R/RStudio v4.0.4	The R Foundation	https://www.r-project.org/
FCS Express 7 (Version 7.04.0016)	De Novo software	FCS Express Flow Cytometry Software - De Novo Software
ImageJ	NIH	https://imagej.net/software/imagej/
HeskaView Integrated Software (version 2.5.2)	Heska	Element HT5 - Heska
Aperio ImageScope (version 12.4.0.5043)	Leica Biosystems	N/A
MetaVue imaging software (Version 7.7.3)	Molecular Devices	N/A
FACSDiva (Version 6.1.3)	Beckman Coulter	N/A
Other
Nikon TE2000-E inverted microscope	Nikon, USA	N/A
CoolSNAP HQ (CCD camera)	Photometrics, USA	N/A
LSRFortessa Cytometer	BD Biosciences, USA	N/A
VERITAS microplate luminometer	Promega, USA	N/A
Nanodrop	Thermo Fisher, USA	N/A

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, prof. Andrei Bakin (Andrei.Bakin@roswellpark.org).

Materials availability

Patient-derived xenograft colorectal cancer models are available from the lead contact (A.B.) under a Material Transfer Agreement.

Data and code availability

• •This paper does not report original code.
• •This article includes all data associated with the study in the accompanying tables, figures, and supplementary materials. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and study participant details

Cell culture models

All human cell lines were authenticated using short tandem repeat profiling by ATCC and/or the Roswell Park Genomic Shared Resource within the last three years. The cells were routinely screened for mycoplasma, and all studies made use of mycoplasma-free cells. Cell cultures were maintained in media supplemented with 10% heat-inactivated fetal bovine serum (FBS) and penicillin/streptomycin at 37°C with 5–10% CO2 in a humidified incubator. The genetics information details for the cell line and PDX models is presented in Table S1.

Xenograft animal models

Female SCID/CB17 mice (6-7-week-old) were obtained from a colony of SCID/CB17 mice bred and maintained at the Animal Facility of the Roswell Park Comprehensive Cancer Center (RP).

Pancreatic patient derived xenografts (Panc-PDX-14312-4p) were implanted sc into SCID mice, while colorectal patient derived xenografts (CRC PDX-01-03-3p and PDX-02-07-RP-B-3p) were grown in NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ mice (Jackson Laboratory, Bar Harbor, ME), also known as NOD SCID gamma (NSG). Animals were kept in microinsulator units and provided with food and water ad libitum according to a protocol and guidelines approved by the Institute Animal Care and Use Committee (IACUC). The facility is certified by the American Association for Accreditation of Laboratory Animal Care (AAALAC) and in accordance with current regulation and standards of the US Department of Agriculture and the US Department of Health and Human Services.

Method details

Cell-derived xenografts in mice

For tumor cell-derived xenografts (CDX), female SCID/CB17 mice (6-7-week-old) were inoculated subcutaneously in the flank with exponentially growing HT29 or MIAPACA-2 tumor cells (1x10⁶/mouse). Tumor growth was monitored by measuring tumor diameters with electronic calipers twice/week. Volumes were calculated using the formula (length) × (width)²/2. Once tumor volume reached 100 mm³, mice were randomly divided into four groups: vehicle-control, PARPi, TAS-102, and TAS-102+PARPi. PARP inhibitors (Olaparib or Talazoparib) and TAS102 were dissolved in 12% HPCD, (2-Hydroxypropyl)-β-cyclodextrin, in Dulbecco’s Phosphate Buffered Saline (DPBS). TAS102 and Olaparib were given at 50 mg/kg and Talazoparib at 0.15 mg/kg by oral gavage on schedule 5-day-on and 2-day-off. At the endpoint (tumor diameter 2 cm), mice were euthanized and subjected to necropsy and organ collection. For the tumor growth assessment, when tumor size reaches a tumor volume threshold in about 20% of mice and they are removed from study, tumor volume progression is no longer plotted. However, the survival data were still collected on the study. Tumor tissues were collected for RNA and protein analyses by snap-freezing in liquid nitrogen. Blood was collected for CBC by cardiac puncture.

Patient derived xenograft (PDX)

Donor and experimental pancreatic patient derived xenografts (Panc-PDX-14312-4p) were implanted sc into SCID mice, while colorectal patient derived xenografts (CRC PDX-01-03-3p and PDX-02-07-RP-B-3p) were grown in NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ mice, also known as NOD SCID gamma (NSG). Tumors were grown in donor mice to volume of ∼1500mm,³ harvested rapidly after euthanasia, cut into 2 × 2 × 2 mm³ pieces under sterile conditions, and surgically implanted into the left flank. Surgical staples were removed about ten days post implantation, and tumor growth and mouse weights were measured two times a week. Once the average tumor volume reached approximately 100mm³, mice were randomized and treated as described above.

Complete blood Count (CBC)

At the endpoint, blood was collected by cardiac puncture into EDTA solution to prevent coagulation. Analysis was performed using the HemaTrue Analyzer and HeskaView Integrated Software (Version 2.5.2).

Cytotoxicity assays

Cells were plated at a density of 5,000 cells/well in a 96 well-plate and treated with the appropriate drugs at varying concentrations for 24 h. Media was replenished with media with or without PARP inhibitors followed by incubation for 96 h. Cells were stained with 1% Methylene Blue for 30 min, rinsed with water, dried, solubilized in 5% SDS in PBS, and read at 650 nm IC₅₀ values were generated using GraphPad Prism (Version 10.0.3).

Caspase-3/7 activity

Cells were seeded in black-wall 96-well plates at 5,000 cells/well and incubated at 37°C, 5% CO₂. Next day, treatments were given in triplicates for 48hrs, and Caspase-3/7 activity was determined using Promega Caspase-Glo 3/7 Assay following the manufacturer’s protocol. Luminescence was measured in the VERITAS microplate luminometer, presented as relative luminescence units (RLU) normalized to vehicle-control, and quantified with Microsoft Excel. Experiments were performed at least in two biological repeats.

Immunohistochemistry (IHC)

Tumors and organ tissues were fixed in 10% (v/v) formalin, before embedding in paraffin by the Roswell Park Pathology Core. H&E and other stains were carried out by the Pathology Core as described in.¹⁸ Details of antibodies and reagents, and expanded methodology for immunohistochemistry and statistical analysis can be found in the Supplementary Information. Three tumors from each treatment group were used for quantification. Slides were scanned using Aperio ImageScope version 12.4.0.5043. For each tumor section, at least five images at 20× magnification (fields of view (FOV)) were captured and analyzed. Data were analyzed in Microsoft Excel and presented as scattered dot plots using GraphPad Prism 10 (Version 10.0.3).

Immunoblotting

Whole-cell lysates were prepared using NP40 lysis buffer containing PMSF, Na-Orthovanadate, and protease inhibitor cocktail. Protein concentrations in lysates were quantitated and 15 μg of protein/lane were resolved on SDS-PAGE gels. Proteins were transferred onto a nitrocellulose membrane in 10% Methanol-SDS buffer and probed with appropriate antibodies. ECL reagent was used to visualize immune complexes on radiographic films.

Immunofluorescence microscopy

For evaluation of γH2AX and RAD51, cells were grown on glass coverslips and treated with 500nM TAS102, 100nM Talazoparib or their combination for indicated time. Cells were fixed with 4% PFA and permeabilized with 0.05% Triton X-100. Samples were blocked with 3% milk in PBS and probed with appropriate antibodies in 1% milk, followed by wash in PBS and incubation with fluorescent secondary antibodies in 0.5% milk. Samples were washed with PBS three times and DNA was labeled with Hoechst 33342 before mounting on glass slides. Fluorescence images were taken with a Plan Apochromat 100×/1.40 NA oil objective at ambient temperature using a Nikon TE2000-E inverted microscope equipped with a charge-coupled device camera (CoolSNAP HQ; Photometrics). The levels of γH2AX and RAD51 were assessed by measuring cellular fluorescence intensity as the Corrected Total Cellular Fluorescence (CTCF). Briefly, a minimum of six random fields per treatment group were evaluated using MetaVue imaging software (Version 7.7.3, Molecular Devices). Three background readings were measured for each field of view and individual cells that were positively stained with γH2AX and RAD51 were selected with a region freehand tool. For each treatment group >40 cells were evaluated and the CTCF values were calculated using the following formula: integrated density – (area of cell × average background fluorescence). Data were analyzed in Microsoft Excel and presented as scattered dot plots using GraphPad Prism (Version 10.0.3).

Flow cytometry

For the EdUrd pulse experiments, 300,000 cells per well were seeded in a 6-well plate, and the following day media was replaced with base media containing 5% dialyzed FBS. Cells were then incubated with 10μM EdUrd for 2 h, while untreated cells served as the negative control. Following the 2-h pulse, cells were washed twice with DPBS and the media was replenished. Collection of cells began at t = 0h up to t = 72hrs post EdUrd-pulse. Cells were collected using standard trypsinization, washed in 1% BSA in DPBS, and fixed in 4% paraformaldehyde for 15 min. Cells underwent two more washes in 1% BSA/DPBS before being permeabilized in 1X saponin buffer. To label the incorporated EdU, cells were subjected to ‘click-it’ reaction with Cu(II)SO₄, Tris-pH 8.5, THTPA, ascorbic acid, and either Cy3 azide or AFDye 488 azide for 30 min. DNA content was labeled with either Hoechst 33342 or Propidium Iodide containing 1 μg/mL of RNAse A. Samples were subsequently washed in 1% BSA/DPBS, resuspended in 1X saponin buffer, transferred to polystyrene tubes, and analyzed on an LSRFortessa Cytometer (BD Biosciences) running FACSDiva (Version 6.1.3), and the data were processed using FCS Express 7 (Version 7.04.0016). Experiments were repeated three times and representative histograms and dot-plots shown.

For trifluorthymidine removal assay, cells were seeded as above and then incubated with 5μM TAS102 for 4 h. After the TAS102 pulse, cells were washed with DPBS twice and the media was replenished with media containing 5% dialyzed FBS. Samples were collected at time = 0 h up to 72 h post the TAS102 pulse. Cells were collected by trypsinization, washed twice with 1% BSA in DPSA, fixed with 75% ethanol for 30 min at 4°C. Cells were incubated with 0.5 mL of 2N HCl, 0.5% Triton X-100 in 1X DPBS for 30 min at room temperature (RT), and then neutralized with 1.5 mL 0.1M sodium tetraborate, pH 8.5, for 2 min. Cells were washed twice with 1% BSA in 1X DPBS and resuspended in 50μL of 0.5% Tween 20, 1% BSA in 1X DPBS. Cells were then stained with 1 μg/10⁶cells FITC-α-BrdU antibody for 1 h at RT, washed twice by 1% BSA in 1X DPBS. DNA content was labeled with Hoechst 33342 containing RNase A. Samples were washed in 1% BSA in DPBS, and resuspended with 1X saponin buffer, then transferred to polystyrene tubes. Samples were analyzed on an LSRFortessa Cytometer (BD Biosciences) running FACSDiva (Version 6.1.3), and the data were processed using FCS Express 7 (Version 7.04.0016). Experiments were repeated three times and representative histograms and dot-plots shown.

For cell cycle analysis, cells were seeded at 250,000 cells/well in 6-well plates and then treated with various amounts of Olaparib and Talazoparib for 24 h. Collected cells were fixed for 2 h in ice-cold 70% ethanol and stained for 2 h at 4°C in Krishan DNA Buffer (propidium iodide, sodium citrate, RNase A, NP40, and 0.1 mM HCl). Samples were sorted using a BD LSRFortessa cytometer running FACSDiva (Version 6.1.3), and the data were analyzed using FCS Express 7 (Version 7.04.0016). Experiments were repeated twice with representative histograms shown.

Metadata analysis

Clinical, mutation and expression Z-scores data for Colorectal cancer (CRC) and Pancreatic adenocarcinoma (PDAC/PAAD) were downloaded from the TCGA Pan-Cancer (PANCAN) via the cBioPortal tool https://www.cbioportal.org/. Outlier data was removed using Interquartile Range (IQR) calculation using outlier R package. Gene lists for cell-cycle related genes are generated using Cyclebase_3.0 database http://www.cyclebase.org.⁵⁶ DNA repair gene lists were derived from KEGG database http://www.genome.jp/kegg/.⁵⁷ Heatmap, boxplots and correlation plots were generated using ComplexHeatmap and ggplot packages in R/RStudio 4.0.4 version (University at Buffalo, Center for Computational Research (UBCCR)). The represented boxplot is the average expression of DNA repair and replication-related genes and reported as mean and standard deviation.

Quantification and statistical analysis

Statistical analysis was performed using GraphPad Prism software (Version 10.0.3) unless otherwise indicated. The data represent biological replicates (n) as mean value with standard error (mean ± SE). Statistical significance of data comparisons was determined using the Student’s unpaired t-test with a two-tailed distribution, one way ANOVA, or two-way ANOVA with a Tukey multiple comparisons test or with Sidak’s multiple comparisons test as indicated in figure legends. Statistical significance was achieved when p < 0.05. The correlation analysis was performed using Pearson correlation approach in RStudio. Survival was evaluated using the Kaplan-Meier estimator with the log rank test, based on time-to-arrive at a defined tumor volume using GraphPad Prism. Cytotoxicity data were normalized to control and IC50 was calculated using three-parameter nonlinear regression.

Published: February 21, 2024