The SW480 and SW1116 cell lines were purchased from the American Type Culture Collection (ATCC) (Manassas, VA, United States), and all other cell lines were bought from the Cell Bank of the Chinese Academy of Sciences (Shanghai, China). Cell authenticity was confirmed by short tandem repeats (STR) profiling.

Alkannin (B50783) was bought from Shanghai Yuanye Biotechnology (Shanghai, China). Olaparib (S1060), niraparib (S2741), talazoparib (S7048) and veliparib (S-1004) were purchased from Selleck (Houston, TX, United States). The OGG1 inhibitor O8 (SML1697) and N-acetyl-L-cysteine (NAC) (BP907) were purchased from Sigma-Aldrich (St. Louis, MO, United States). Stock solutions of alkannin, olaparib or veliparib were made in 100% dimethyl sulfoxide (DMSO) (Sigma-Aldrich) and working solutions were prepared in complete cell culture medium. Vehicle controls were prepared similarly but without the test compound. Primary antibodies include γH2AX-pS139 polyclonal antibody (ab11174) (Abcam, Cambridge, United Kingdom); cleaved caspase-3 (9664S), Chk1-pS317 (12302S) and Chk2-pT68 (2661S) (Cell Signaling, Danvers, MA, United States); Chk1 (bs1681R), GAPDH (bs2188R), β-actin (bsm33036M), CDC25C (bs10579R), CDC25C-pS216 (bs3096R), p53-pS15 (bs3702R), p53 (bs2092R), and PARP1 (bs2138R) (Bioss, Beijing, China); H2AX (abs131731), Chk2 (abs131635) and BBC3 (PUMA) (abs131259) (Absin Bioscience, Shanghai, China); 53BP1 (A300-272A) (Bethyl, Montgomery, TX, USA); γH2AX-pS139 monoclonal antibody (14-9865-82) (ThermoFisher, Waltham, MA, USA). Secondary antibodies include goat anti-mouse-Alexa 488 (115-545-003) and goat anti-rabbit-Cy3 (111-165-003) (Jackson ImmunoResearch, West Grove, PA, USA); goat anti-mouse-HRP (bs-40296G-HRP) and goat anti-rabbit-HRP (bs-40295G-HRP) (Bioss).

Cells were seeded in 96-well plates at 4 × 10³ cells/well for 12 h and were then treated with drugs for the indicated times. When multiple drugs were used, they were added simultaneously. At the end of drug treatment, 20 μL of 5 mg/ml 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) (Sigma-Aldrich) was added to each well and incubated at 37 °C for 4 h. The media were carefully removed and 100 μL of DMSO was added to each well. The plate was left on a plate shaker for 30 min with gentle shaking at room temperature. Absorbance at 595 nm was measured. IC₅₀ values were determined by the GraphPad Prism software using nonlinear regression analysis. The combination index (CI) values were determined by the Chou-Talalay method according to the formula CI = (D)1/(Dx)1 + (D)2/(Dx)2. (Dx)1 and (Dx2) were concentrations of each drug alone to exert x% effect, while (D)1 and (D)2 were concentrations of drugs in combination to elicit the same effect. CI < 1 indicated synergism, and CI = 1 or >1 indicated additivity or antagonism, respectively (Chou, 2010, 2018).

Cells were seeded in 12-well plates at 500 cells/well for 12 h and were then treated with the indicated drugs for 7 days. Following washing with PBS, the cells were fixed in ice-cold methanol and briefly stained with crystal violet solution (0.5% crystal violet in 25% methanol) (Sigma-Aldrich). The plates were air-dried, images were taken, and the violet crystals were dissolved in 70% ethanol. Absorbance at 595 nm was measured.

Cells were seeded in 6-well plates at 1 × 10⁵ cells per well for 12 h and were then treated with drugs for the indicated times. Cellular ROS were detected using a cell-based ROS assay kit (Beyotime, Shanghai, China). Briefly, cells were washed with PBS, and incubated with 10 μM dichlorofluorescin diacetate (DCFH-DA) for 30 min at 37 °C in the dark. After washing three times in PBS, images were taken immediately, and the cells were then collected through trypsinization and analyzed on the BD FACS-Calibur flow cytometer (BD Biosciences, San Jose, CA, United States). Cellular ROS levels were expressed as the average DCF (di-chloro-fluorescin) intensity. Mitochondrial membrane potential (MMP) was measured using a commercial Mitochondrial Membrane Potential Assay kit with JC-1 following the manufacturer’s instructions (Beyotime). After drug treatment, cells were washed by PBS and incubated in JC-1 dye at 37 °C in the dark for 20 min, followed by washing twice with JC-1 buffer. Cells were covered by a thin layer of DMEM media for photographing.

Cells on round coverslips or cryosections of tumor tissues were fixed in 4% paraformaldehyde (PFA) for 30 min and washed three times with PBS. For 8-oxoG staining, fixed cells were incubated in Cy3-conjugated avidin (Struthers et al., 1998) (Rockland Immunochemicals, Limerick, PA, United States) (0.5 μg/ml) for 1 h at room temperature; for immunofluorescent staining, the samples were incubated sequentially in blocking buffer (3% fetal bovine serum, 0.1% Triton X-100 in PBS), primary and secondary antibodies, each for 1 h at room temperature; for TUNEL assay, tissue sections were incubated in working solutions from a One-Step TUNEL apoptosis assay kit (Beyotime) for 60 min at 37 °C. Samples were then washed three times in PBS and sealed on glass slides in the VECTASHIELD Mounting Medium with DAPI (Vector Laboratories, Burlingame, CA, United States).

Cells were seeded in 6-well plates at 1 × 10⁵ cells per well, treated with drugs for the lengths of time indicated in the figure legends, and collected through trypsinization. The cell pellets were washed twice in PBS before the following analyses. For cell cycle analysis, cell pellets were fixed in 70% ice-cold ethanol at −20 °C for 1 h. After washing in PBS twice again, cells were stained with working solutions from the Cell Cycle Detection kit (Bestbio, Shanghai, China). DNA content was revealed by propidium iodide. For the analysis of apoptosis, cells were resuspended in a binding buffer from the Annexin V-FITC Apoptosis Detection kit (Bestbio) according to the instructions provided by the manufacturer. The samples were analyzed by the MoFlo XDP Cell Sorter (Beckman Coulter, Indianapolis, IN, United States). 2 × 10⁴ cells were analyzed per sample.

Measurement of DNA strand breaks in individual cells by the alkaline comet assay (single cell gel electrophoresis assay) was performed according to the instructions included in the OxiSelect Comet Assay kit (Cell Biolabs, San Diego, CA, United States). After drug treatment, cells grown in 6-well plates were collected through trypsinization. Cell pellets were resuspended in 1.2% low-melting point agarose maintained at 37 °C at 10 × 10⁵ cells/ml, which were then layered on a frosted slide from the OxiSelect Comet Assay kit. The slides were stored at 4 °C overnight in pre-cooled lysis buffer (100 mM EDTA, 2.5 M NaCl, 10 mM Tris–HCl, 1% Triton X-100 and 10% DMSO, pH 10.0). After washing twice with an enzyme buffer (40 mM HEPES, 0.1 M KCl, 0.5 mM EDTA and 0.2 mg/ml BSA, pH 8.0), the slides were denatured in pre-chilled alkali buffer (300 mM NaOH, 1 mM EDTA) in a horizontal electrophoresis chamber for 30 min. Electrophoresis was then proceeded at 20 V and 300 mA in the same buffer for 30 min. After incubation in cold neutralizing buffer (250 mM Tris–HCl, pH 7.5) for 30 min, slides were immersed in cold 70% ethanol for 5 min and allowed to air dry. At the end, cells were stained with Vista Green DNA dye provided by the kit at room temperature for 15 min. Images were quantified using the free TriTek CometScore software (TriTek Corp., Sumerduck, VA, United States). At least 100 comets per sample were analyzed. Tail moment was determined as the percentage of DNA in the tail multiplied by the tail length.

Cells grown in 6-well plates were scraped off the plates in 100 μL of radioimmunoprecipitation buffer (150 mM NaCl, 1.0% IGEPAL CA-630, 0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate, and 50 mM Tris, pH 8.0) (Sigma) with 1 mM phenylmethane sulfonylfluoride (Sigma). Samples were centrifuged at 4 °C for 20 min at 12,000 g and protein concentrations were determined by a BCA Protein Assay kit (Dingguo, Changchun, China). Proteins were denatured at 95 °C for 10 min, separated on a 12% SDS-PAGE gel, and transferred to PVDF membranes. The membranes were blocked in 5% (w/v) non-fat milk in TBST (10 mM Tris, 100 mM NaCl, 0.1% Tween 20, pH 7.5) for 1 h at room temperature, and then probed in specific first and second antibodies. Signals were developed using a Tanon-5200 chemiluminescence image analysis system (Tanon, shanghai, China).

All animal studies were conducted in compliance with animal protocols approved by the Institutional Animal Care and Use Committee of Jilin University. 2 × 10⁶ SW480 cells were inoculated subcutaneously into the flank of 8-week old male athymic BALB/c nude mice (Charles River, Boston, MA, United States), and the mice were randomly placed in control and treatment groups (5 animals per group). Fifteen days after inoculation, tumors reached about 150 mm³ and the mice were treated once daily with 20 mg/kg alkannin oral gavage (p.o.) or 50 mg/kg olaparib intraperitoneal injection (i.p.), or a combination of alkannin and olaparib, for 15 days. Tumor volume was measured every 3 days, and tumor weight was measured at the end of treatment.

All in vitro experimental data were expressed as mean ± standard deviation (SD) of three independent experiments. All statistical comparisons were completed by the two-tailed Student’s t-test using the GraphPad Prism software (GraphPad Software, Inc., La Jolla, CA, United States). p < 0.05 was considered as statistically significant.

Antitumor activity of alkannin was initially demonstrated in a mass screening of the US National Cancer Institute’s natural and synthetic compound repository in 1974 (Driscoll et al., 1974). Since then, a large body of studies has confirmed the cytotoxicity of alkannin against diverse types of cancer cell lines with IC₅₀ values in the low micromolar range (1–30 μM) (Huang et al., 2004; Wu et al., 2005; Bogurcu et al., 2011). In agreement with the literature, we found that alkannin dose-dependently inhibited the growth of a panel of human solid tumor cell lines (Supplementary Figure S1A). After 24 h of treatment, the IC₅₀ values of alkannin against the SW480 and SW1116 human colorectal cancer cell lines were 3.98 and 4.23 μM respectively (Figure 1A). Accordingly, 3 μM alkannin significantly suppressed the clonogenic growth of these two cell lines, but 1.5 and 0.75 μM alkannin had no impact on their clonogenic survival (Figure 1B). Nevertheless, all three doses of alkannin induced a marked increase in ROS levels in both cancer cell lines (Figure 1C; Supplementary Figure S1B).

Flow cytometry analyses revealed that, treatment with 0.75 μM alkannin for 3 h caused a significant increase in ROS levels in both SW480 and SW1116 cancer cells (Figures 1D–E, Supplementary Figure S1C). In contrast, similar treatment caused no change in ROS levels in the NCM460 normal human colon epithelial cell line (Figure 1D–E). The ROS inhibitor N-acetyl-L-cysteine (NAC) effectively reduced the alkannin-induced ROS increase in the cancer cells (Figures 1D–E; Supplementary Figure S1C). Together, these results showed that significant increases in ROS levels were induced in the colorectal cancer but not in the noncancerous colon epithelial cells by nontoxic doses of alkannin.

Elevated ROS can cause oxidative DNA damage including nucleobase oxidization and single-strand DNA breaks (SSB); double-strand DNA breaks (DSB) may arise when a DNA replication fork collides with a SSB during DNA replication. Consistent with the increase in ROS levels, staining of 8-oxoguanine (8-oxoG), the major type of oxidized nucleobases (Markkanen, 2017), showed that within 3 h of treatment, both 1.5 and 0.75 μM alkannin induced a significant increase in 8-oxoG levels in the SW480 and SW1116 cancer but not in the NCM460 noncancerous cells (Figure 2A; Supplementary Figure S2A). NAC suppressed 8-oxoG increase in the cancer cells (Figure 2A; Supplementary Figure S2A), correlating 8-oxoG elevation with sublethal alkannin-induced ROS overload.

Similarly, alkaline comet assay revealed that 3 h of treatment by 0.75 μM alkannin caused a significant increase in DNA strand breaks in the SW480 and SW1116 cancer cells (Figure 2B; Supplementary Figure S2B). Moreover, 53BP1 foci, which represent sites of DSB (Schultz et al., 2000), were induced in the two colorectal cancer cell lines by 0.75 μM alkannin in a time-dependent manner (Figure 2C, Supplementary Figure S2C), suggesting that some of the DNA strand breaks detected by the alkaline comet assay were DSBs. No increase in DNA strand breaks was induced in the noncancerous NCM460 colon epithelial cells (Supplementary Figures S2B,C), and both the comet tails and 53BP1 foci in the cancer cells were effectively reversed by NAC (Figures 2B,C; Supplementary Figure S2C), indicating that the DNA strand breaks resulted from sublethal alkannin-induced ROS elevation.

The induction of oxidative DNA damage led us to ask if nontoxic doses of alkannin could sensitize colorectal cancer cells to inhibition of PARP1/2. Homologous recombination (HR)-deficient cancer cells are exquisitely sensitive to PARP inhibition even in the absence of exogenous DNA damage. However, most types of cancer cells are HR-proficient and are thus resistant to PARP inhibition. Indeed, the SW480 and SW1116 cells readily formed RAD51 foci in response to treatment by 5 μM SN-38 (Figure 3A, Supplementary Figure S3A), indicating normal HR function (Graeser et al., 2010). Consistently, both cell lines were highly resistant to the PARP inhibitor olaparib (Figure 3B, Supplementary Figure S3B).

Remarkably, however, in the presence of 0.75 μM alkannin, olaparib dose-dependently inhibited the viability of the SW480 and SW1116 cancer cells (Figure 3B, Supplementary Figure S3B). Similarly, colony formation assays showed that neither 0.75 μM alkannin nor 10 μM olaparib had any impact on the clonogenic growth of the SW480 and SW1116 cancer cells but strikingly, the combination of 0.75 μM alkannin and 10 μM olaparib almost completely inhibited their clonogenic survival (Figure 3C, Supplementary Figure S3C). In contrast, the clonogenic growth of the NCM460 normal colon epithelial cells was not affected by the combination of 0.75 μM alkannin and 10 μM olaparib (Figure 3C). Thus, sublethal alkannin selectively sensitized the SW480 and SW1116 cancer cells to olaparib to yield synergistic cancer cell toxicity at nontoxic doses of both drugs.

To evaluate the synergism between alkannin and olaparib, the combination index (CI) was determined by the Chou-Talalay method using the CompuSyn software. The CI values for combinations between either 1.5 or 0.75 μM alkannin and a wide range of olaparib concentrations were far below 0.5, indicating strong synergy between alkannin and olaparib (Figure 3D, Supplementary Figure S3D).

The synergistic cytotoxicity between alkannin and olaparib was greatly attenuated by NAC (Figure 3C, Supplementary Figure S3C), suggesting that it was dependent on alkannin-induced oxidative DNA damage. The 8-oxoguanine glycosylase (OGG1) initiates base excision repair (BER) of 8-oxoG by excising oxidized guanin bases and further cleaving the DNA backbone, leading to generation of SSB. PARP1/2 bind SSBs produced directly or as intermediates of BER, and recruit XRCC1 to assemble the repair machinery (Ray Chaudhuri and Nussenzweig, 2017b). Thus, inhibition of PARP would lead to accumulation of SSBs, while inhibition of OGG1 would reduce the generation of SSB and mitigate the impact of PARP inhibition. Consistently, the OGG1 inhibitor O8 (Donley et al., 2015) greatly reduced the cytotoxicity of the alkannin and olaparib combination (Figure 3B, Supplementary Figure S3B), suggesting that OGG1-initiated base excision repair of alkannin-induced 8-oxoG was a major cellular activity that synergized with olaparib to generate synergistic cytotoxicity.

In addition to inhibiting the activity of PARP1/2, many PARP inhibitors, such as olaparib, rucaparib, niraparib and talazoparib also trap the PARP protein at DNA damage sites to generate additional cytotoxicity (Murai et al., 2012; Lord and Ashworth, 2017; Murai and Pommier, 2019). In contrast, veliparib is much weaker at trapping PARP1/2 on damaged DNA and is therefore considered as a PARP inhibitor with no PARP-trapping potency. Interestingly, the combination of 0.75 μM alkannin and 10 μM veliparib produced no impact on the clonogenic growth of the SW480 and SW1116 cancer cells (Figure 3C, Supplementary Figure S3C), suggesting that veliparib did not synergize with the sublethal alkannin to generate synergistic cytotoxicity. On the contrary, both niraparib and talazoparib, two PARP inhibitors with higher PARP-trapping potency than olaparib, dose-dependently inhibited the viability of the SW480 and SW1116 cancer cells in the presence of 0.75 μM alkannin (Figures 3E,F, Supplementary Figures S3E,F). The OGG1 inhibitor O8 also greatly reduced the cytotoxicity of the alkannin and niraparib or alkannin and talazoparib combination (Figures 3E,F, Supplementary Figures S3E,F). Similarly, the combination of 0.75 μM alkannin and 5 μM niraparib, or 0.75 μM alkannin and 0.25 μM talazoparib, completely inhibited the clonogenic survival of the SW480 and SW1116 cancer cells (Figure 3C, Supplementary Figure S3C). Thus, PARP-trapping was likely the major cause of the synergistic cytotoxicity between alkannin and PARP inhibitors with PARP-trapping potency.

Inhibition of PARP1/2 leads to accumulation of SSBs. In addition, PARP inhibitors with PARP-trapping property produce trapped PARP-DNA complexes which, together with unrepaired SSBs, collide with DNA replication forks, leading to fork stalling and replication stress (Pommier et al., 2016; Ray Chaudhuri and Nussenzweig, 2017b). Staining of γH2AX, a molecular marker of DNA replication stress and/or DSBs (Ward and Chen, 2001), revealed that 3 h of treatment by the combination of 0.75 μM alkannin and 10 μM olaparib, but not by each agent alone, caused a marked increase in the number of γH2AX positive SW480 and SW1116 cells, most of which displayed strong pan-nuclear γH2AX staining (Figure 4A, Supplementary Figure S4A), indicating induction of intense replication stress (Moeglin et al., 2019) specifically by the combination of alkannin and olaparib. The staining of γH2AX was reversed by NAC (Figure 4A, Supplementary Figure S4A), correlating the intense replication stress with sublethal alkannin-induced oxidative DNA damage.

Because PARP1 and PARP2 also play critical roles in the stabilization and restart of stalled DNA replication forks, PARP inhibition can impair replication fork stabilization and restart, which, together with the replication stress simultaneously induced by PARP inhibition, will promote replication fork collapse and generation of one-ended DSBs (Hanzlikova and Caldecott, 2019). Consistent with the specific induction of strong replication stress, alkaline comet assays revealed that the combination of alkannin and olaparib induced significantly higher levels of DNA strand breaks than the individual drugs (Figure 4B, Supplementary Figure S4B). Similarly, 53BP1 staining showed that significantly more 53BP1 positive SW480 and SW1116 cells were induced by the combination than by each agent alone (Figure 4C, Supplementary Figure S4C), suggesting that the intense replication stress induced by the combination of alkannin and olaparib was converted to DSBs.

In response to replication stress and/or DSB, DDR is activated to regulate cell cycle checkpoints and DNA repair that maintain cell viability; or to induce programmed cell death that eliminates the cells with irreparable problems. In the SW480 and SW1116 cancer cells, 3 h of treatment by the combination of 0.75 μM alkannin and 10 μM olaparib induced a marked increase in the levels of phosphorylated Chk1, Chk2 and p53 (Figure 5A, Supplementary Figure S5A), demonstrating activation of both the ATM-Chk2 and the ATR-Chk1 DDR pathways (Ciccia and Elledge, 2010). The levels of phosphorylated Chk1, Chk2 and p53 were notably reduced by NAC (Figure 5A, Supplementary Figure S5A), correlating DDR activation with the specific induction of intense replication stress and extensive DNA strand breaks by the combination of alkannin and olaparib. A weaker increase in the levels of phosphorylated Chk1, Chk2 and p53, as well as a significant increase in total p53, was induced by olaparib alone (Figure 5A), indicating that blocking PARP1/2 activity alone was able to activate DDR but could not produce cytotoxicity.

In accordance with DDR activation, phosphorylation of CDC25C, one of the important downstream targets of DDR, was significantly increased, which was accompanied with a corresponding significant decrease in the levels of the total CDC25C protein (Figure 5A, Supplementary Figure S5A). Consistently, flow cytometry analyses showed that, 48 h of treatment by the combination of 0.75 μM alkannin and 10 μM olaparib, but not by each alone, induced a dramatic decrease in the number of cells in G₁ and a marked increase in the number of cells in G₂/M (Figure 5B, Supplementary Figure S5B), indicating thorough activation of the G₂/M cell cycle checkpoint specifically by the combination of alkannin and olaparib.

Importantly, the proapoptotic BH3-only protein PUMA, a downstream target of p53 and direct activator of apoptosis, was significantly upregulated specifically by the combination of alkannin and olaparib (Figure 5A, Supplementary Figure S5A), consistent with that the SW480 and SW1116 colorectal cancer cells have normal p53 function (Rochette et al., 2005; Leroy et al., 2014). Correspondingly, staining with the fluorescent probe JC-1 showed a profound dissipation of mitochondrial membrane potential (Figure 5C, Supplementary Figure S5C), and Western blot analyses revealed a significant increase in the levels of activated caspase 3 and cleaved PARP (Figure 5D, Supplementary Figure S5D), in the SW480 and SW1116 cancer cells treated by the combination of 1.5 or 0.75 μM alkannin and 10 μM olaparib for 12 h. Furthermore, flow cytometry analysis showed that 48 h of treatment by the combination of 0.75 μM alkannin and 10 μM olaparib, but not by each of them alone, induced a marked increase in the percentage of Annexin V-positive cells (Figure 5E, Supplementary Figure S5E), demonstrating full activation of the mitochondrial apoptotic pathway.

To assess the clinical potential of the alkannin and olaparib combination, we evaluated the effects of treatment with alkannin and olaparib alone and combined on tumor xenografts in vivo. Mice bearing SW480 tumor xenografts were treated with either alkannin (dosed once daily by oral gavage, 20 mg/kg) or olaparib (dosed once daily by intraperitoneal injection, 50 mg/kg), or the combination of alkannin and olaparib, for 15 days. The results showed that alkannin or olaparib alone had no impact on the growth of the tumor xenografts, however, coadministration of alkannin and olaparib resulted in significant regression of the tumor xenografts (Figures 6A–C), demonstrating in vivo efficacy for the alkannin and olaparib combination. On the other hand, no significant difference in body weight was observed between control and the three drug-treated groups (Figure 6D), suggesting that the two drugs were well tolerated either alone or combined. Immunohistochemical staining of the tumor tissues revealed that only the group treated by the combination of alkannin and olaparib showed significantly increased γH2AX signals and intense apoptosis (Figure 6E), suggesting that DNA damage-induced apoptosis caused tumor regression.