A comprehensive in vivo and mathematic modeling-based kinetic characterization for aspirin-induced chemoprevention in colorectal cancer

Abstract

Accumulating evidence suggests that aspirin has anti-tumorigenic properties in colorectal cancer (CRC). Herein, we undertook a comprehensive and systematic series of in vivo animal experiments followed by 3D-mathematical modeling to determine the kinetics of aspirin’s anti-cancer effects on CRC growth. In this study, CRC xenografts were generated using four CRC cell lines with and without PIK3CA mutations and microsatellite instability, and the animals were administered with various aspirin doses (0, 15, 50, and 100 mg/kg) for 2 weeks. Cell proliferation, apoptosis and protein expression were evaluated, followed by 3D-mathematical modeling analysis to estimate cellular division and death rates and their impact on aspirin-mediated changes on tumor growth. We observed that aspirin resulted in a dose-dependent decrease in the cell division rate, and a concomitant increase in the cell death rates in xenografts from all cell lines. Aspirin significantly inhibited cell proliferation as measured by Ki67 staining (P < 0.05–0.01). The negative effect of aspirin on the rate of tumor cell proliferation was more significant in xenograft tumors derived from PIK3CA mutant versus wild-type cells. A computational model of 3D-tumor growth suggests that the growth inhibitory effect of aspirin on the tumor growth kinetics is due to a reduction of tumor colony formation, and that this effect is sufficiently strong to be an important contributor to the reduction of CRC incidence in aspirin-treated patients. In conclusion, we provide a detailed kinetics of aspirin-mediated inhibition of tumor cell proliferation, which support the epidemiological data for the observed protective effect of aspirin in CRC patients.

In this study, following in vivo colorectal cancer xenograft experiments and mathematical modeling, we have demonstrated the effects of aspirin on the rate of tumor cell proliferation and death, which might be associated with chemo preventive mechanism of aspirin in this disease.

Introduction

Colorectal cancer (CRC) is a well-recognized healthcare problem affecting approximately 5% of the population and remains one of the leading cause of cancer-associated deaths in the United States (1). Aspirin (acetylsalicylic acid) is a non-steroidal anti-inflammatory drug (NSAID), commonly used as an analgesic or antipyretic drug, and in prophylactic settings for cardiovascular or cerebrovascular diseases. Emerging epidemiological and clinical evidence indicates that long-term regular use of aspirin also results in reduction of CRC risk (2–4). Furthermore, several studies have shown that pre-diagnostic or post-diagnostic aspirin use improves the survival of CRC patients (5–12). Collectively, these data indicate that aspirin not only works as a chemo-preventive agent, but also appears to have an adjuvant role in the management of CRC.

Lynch syndrome (hereditary non-polyposis CRC) is an autosomal dominant hereditary disease of mismatch repair genes, such as MLH1, MSH2, MSH6, and PMS2, and patients with Lynch Syndrome have a 30–50% lifetime CRC risk (13,14). Previous studies have shown that aspirin administration reduces the risk of CRC development in patients with Lynch syndrome (15,16). Interestingly, the Cancer Prevention Programme 1 (CAPP1) study showed that aspirin intake did not significantly reduce polyp formation in patients with familial adenomatous polyposis (FAP) (17). However, the subsequent CAPP2 randomized clinical trial demonstrated that the daily intake of 600 mg of aspirin reduced cancer incidence after 4 years following the 25-month aspirin treatment (18–20). Despite the success of these studies, a high dose of aspirin (600 mg/day) is associated with an age-dependent increased risk of adverse effect such as gastrointestinal mucosal bleeding or brain hemorrhage (21). Therefore, the current CAPP3 study is interrogating multiple doses of aspirin (100 mg, 300 mg, and 600 mg/day) in Lynch Syndrome carriers to determine an optimal dose and treatment duration with aspirin (22).

Herein, we conducted a comprehensive in vivo study using three doses of aspirin that are comparable to the doses administered in the CAPP3 study (23) in four independent CRC cell line-derived xenografts with different mutational backgrounds including microsatellite instability (MSI) and mutations in the PIK3CA gene. Furthermore, we collected xenograft tumors of mice treated with various doses of aspirin treatments at multiple time points to examine the time-dependent efficacy of aspirin on cellular proliferation and apoptosis. Finally, we applied mathematical modeling to these data to determine how aspirin affects tumor growth kinetics in CRC—all of which can provide important insights into the molecular dynamics of this NSAID in cancer prevention.

Materials and methods

Cell culture and materials

CRC cell lines, including HCT116 (MSI and PIK3CA mutant), SW480 (MSS and PIK3CA wild type), HT29 (MSS and PIK3CA mutant), and LoVo (MSI and PIK3CA wild type) (24) were purchased from the American Type Culture Collection (Manassas, VA). These CRC cell-lines were authenticated and examined for mycoplasma contamination. The cells were grown in Dulbecco’s modified Eagle’s medium (Gibco, Carlsbad, CA), supplied with 10% fetal bovine serum (Gibco), 1% penicillin and streptomycin (Gibco), and maintained at 37°C in a humidified incubator at 5% CO₂. Aspirin was purchased from Sigma–Aldrich (product NO; A2093, St Louis, MO) and dissolved in 5% dimethyl sulfoxide (DMSO) for in vivo and in vitro experiments.

In vitro apoptosis assay

For these assays, 3 × 10⁵ cells/ml of cell culture medium were plated in 6-well plates overnight and subsequently treated with 1, 2.5, and 5 mM aspirin for 48 h (Figure 1A). All experiments were conducted in triplicate. Apoptosis assays were performed using the Muse Annexin V and Dead Cell Kit (MCH100105; Millipore, Chicago, IL) on a Muse Cell Analyzer (Millipore), according to the manufacturer’s instructions.

Figure 1.

The effect of aspirin on cellular apoptosis in multiple CRC cell-lines. (A) The schematic diagram of the in vitro experiment. (B) Percentage of apoptotic cells analyzed by apoptosis assays with each dose of aspirin. (C) Western immunoblotting illustrating the protein expression changes of key apoptosis-related markers (Bax, Bcl2, and Bcl-xl) in four CRC cell-lines with and without aspirin treatment. Abbreviations; Ctrl, control; Asp, aspirin. *P < 0.05, **P < 0.01, and ***P < 0.001 by one-way ANOVA.

Western immunoblotting

Western blotting experiments were performed according to the procedure described previously (25), and a schematic diagram of the overall in vitro experiments is shown in Figure 1A. Briefly, 3 × 10⁵ CRC cells/ml of cell culture medium were plated in 6-well plates overnight, and were treated with 5 mM of aspirin (Sigma–Aldrich) for HCT116, SW480, and HT29 cells, and with 2.5 mM of aspirin for LoVo cells for 48 h (Figure 1A). Proteins from CRC cells were extracted using a RIPA Lysis and Extraction Buffer (ThermoFisher Scientific, Waltham, MA) containing 0.1% protease inhibitor cocktail (ThermoFisher Scientific) and denatured with 2× Laemmli Sample Buffer (Bio-Rad Laboratories, Hercules, CA) containing 5% 2-mercaptoethanol (Sigma–Aldrich). Twenty micrograms of each denatured protein was loaded onto 12% sodium dodecyl sulfate polyacrylamide gels. The primary antibodies used were monoclonal anti-rabbit Bax antibody (#5023; 1:1000; Cell signaling technology, Danvers, MA), monoclonal anti-mouse Bcl-2 antibody (sc-7382; 1:200; Santa Cruz Biotechnology, Dallas, TX), and monoclonal anti-mouse Bcl-xl antibody (sc-8392; 1:200; Santa Cruz Biotechnology). Anti-mouse IgG-HRP (sc-2005; 1:5000; Santa Cruz Biotechnology) or anti-rabbit IgG-HRP (sc-2004; 1:5000; Santa Cruz Biotechnology) were used as secondary antibodies. A monoclonal mouse beta-actin antibody (A5441; 1:5000; Sigma–Aldrich) was used as the loading control. Chemiluminescence images were obtained using ChemiDoc-TMMP Imaging system (ver 5.2.1, BioRad Laboratories Inc, Hercules, CA). The band intensity was quantified using Image J ver. 1.52 (Bethesda, MD) (26), and shown as a ratio to beta-actin band intensity.

Xenograft experiments

Seven-week-old male athymic nude mice were purchased from Envigo (Houston, TX) and acclimatized under controlled conditions of light, fed ad libitum, with free access to water. The schematic diagram of the in vivo CRC xenograft experiment is shown in Figures 2A and 3A. Xenograft tumors from all four CRC cell-lines (HCT116, SW480, HT29, and LoVo) were generated by subcutaneous injection of 1 × 10⁶ tumor cells suspended in Matrigel matrix (BD Biosciences, Franklin Lakes, NJ) into both flanks of mice using a 25-gauge X 5/8 TB syringe (Becton Dickinson and Company, Franklin Lakes, NJ). A total of 432 mice, which comprised of 108 animals for each CRC cell-line divided into four groups of 27 animals per group: (i) a control group, (ii) a low dose aspirin treatment group (15 mg/kg per body weight, orally gavaged), (iii) a medium dose aspirin treatment group (50 mg/kg per body weight, orally gavaged), and (iv) a high dose aspirin treatment group (100 mg/kg per body weight, orally gavaged). Daily oral aspirin administration was performed using an 18-gauge animal feeding needle (catalog#14-825-251; Cadence science, Plainfield Pike Cranston, RI; Figure 2A). Treatment doses of aspirin were calculated using Human Equivalent dose, as described previously (23). In brief, 15 mg/kg aspirin for mice is approximately equivalent to 100 mg/day for human, 50–300 mg/day, and 100–600 mg/day, which corresponds to doses currently being interrogated in a CAPP3 human clinical trial (22). Tumor size was measured by calipers every day. Tumor volume was calculated according to the following formula: 1/2 (length × width × height), followed by normalization of all treatment groups by an average value of the control group as percentage. Three mice from each treatment group were euthanized for harvesting tumors at five separate time-points throughout the experiment (days 3, 5, 7, 9, and 11), and remaining animals were euthanized on day 15 (Figure 3A). Candidate mice for euthanasia at five harvesting time points were chosen randomly using a random number generating website (RANDOM.ORG). All tumor specimens were stored in formalin or RNA-later (Sigma–Aldrich) for subsequent analyses. The animal protocol was approved by the Institutional Animal Care and Use Committee, Baylor Scott & White Research Institute, Dallas, TX.

Figure 2.

The effect of aspirin on the regression of tumor xenografts in an animal model. (A) The schematic diagram for the in vivo experimental strategy for CRC xenografts. (B) Time-course tumor volume alterations with different dose(s) of aspirin treatment in xenografts derived from each cell-line. (C) Non-linear curves illustrating tumor volume with each aspirin dose after 15 days of treatment in PIK3CA mutated and wild-type cells (left) and MSI and non-MSI cells (right). *P < 0.05, **P < 0.01, and ***P < 0.001 by one-way ANOVA or two-tailed t-test as appropriate.

Figure 3.

The effect of aspirin on attenuation of tumor proliferation in xenografts derived from four different CRC cell-lines. (A) The schematic diagram and the concept of time-point for tumor collection in xenograft experiments. (B and C) Percentage of proliferative cells evaluated by Ki67 IHC staining in xenograft tumor tissues from each cell-line harvested on day-15. (D) Non-linear curves describing proliferative cell percentages with each aspirin dose after 15 days of treatment in PIK3CA mutated and wild-type cells (left) and MSI and non-MSI cells (right).*P < 0.05, **P < 0.01, and ***P < 0.001 by one-way ANOVA or two-tailed t-test as appropriate.

Ki67 immunohistochemistry staining

All xenograft tumor tissues were fixed in 10% formalin, and thereafter processed into formalin-fixed paraffin-embedded (FFPE) blocks. Subsequently, tissue sections (5 μm) were used for immunohistochemistry (IHC) analyses. Following antigen retrieval, blocking of nonspecific binding was performed by Dako Envision + Dual Link system-HRP (DAB+) kit (K4065; DAKO). Primary antibody for monoclonal anti-rabbit Ki67 was purchased from Cell Signaling Technology (#9027; 1:400; Danvers, MA) and incubated in a moist chamber overnight at 4°C. Loaded HRP Polymer (K4065; DAKO) was used as secondary antibody and incubated in a moist chamber for 60 min. DAB chromogen (K4065; DAKO) was used to identify target proteins. The sections were counterstained with hematoxylin and dehydrated in graded concentrations of ethanol and xylene. The percentage of proliferative cells evaluated in three microscopic fields per tumor section using ×100 magnification on an upright-microscope (Zeiss Axio Imager M2, Carl Zeiss Microscopy LLC, Thornwood, NY). Image J ver. 1.52 (Bethesda) were used to identify stained cells and total cells and the percentage of stained cells were measured against total cells (26). Further details are provided in Supplementary Materials and Methods, available at Carcinogenesis Online.

TUNEL assay

After the tissue section was de-paraffinized with xylene and rehydrated with graded concentrations of ethanol, a TUNEL staining assay was performed according to manufacturer’s instructions (ApopTag Peroxidase In Situ Apoptosis Detection Kit S7100, EMD Millipore Corporation, Temecula, CA). The percentage of apoptotic cells were evaluated in three microscopic fields per tumor section using ×100 magnification on a Zeiss Axio Imager M2 upright-microscope (Carl Zeiss Microscopy LLC). Image J ver. 1.52 (Bethesda) were used to identify stained cells and total cells and the percentage of stained cells were measured against total cells (26).

Mathematical modeling

Detailed methods for mathematical modeling are provided in Supplementary Materials and Methods, available at Carcinogenesis Online.

Statistical analysis

All experiments were repeated three times. Comparison between groups were analyzed by a two-tailed t-test, and one-way ANOVA with Tukey’s post hoc test, as appropriate. A P-value of <0.05 was considered statistically significant. All error bars in the figures are represented as mean ± SEM. All statistical analyses were performed using Medcalc statistical software V.16.2.0 (Medcalc Software bvba, Ostend, Belgium), and GraphPad Prism V7.0 (GraphPad Software, San Diego, CA).

Results

Aspirin induces apoptosis in a dose-dependent manner in colorectal cancer cells

In our previous study, we have examined in vitro anti-proliferative effects of aspirin in multiple CRC cell lines and demonstrated that the efficacy of aspirin was dependent on PIK3CA mutation status (27). Therefore, given the importance of measuring apoptosis as a means to determine cell death and survival rates, in this study we analyzed cellular apoptosis to calculate growth kinetics of CRC cells (28,29), we initially conducted a series of in vitro experiments to confirm and strengthen this evidence in multiple CRC cell lines with diverse genetic backgrounds to solidify our rationale for their subsequent application in animal experiments. Accordingly, we administered 1, 2.5, and 5 mM of aspirin to HCT116, SW480, and HT29 cells (Figure 1A). Due to the shorter doubling time of HCT116 and LoVo cells, aspirin treatment appeared to be more effective against these cell lines. Nonetheless, as expected, the percentage of apoptotic cells were significantly increased in all cell lines, in a dose-dependent manner (Figure 1B); suggesting that aspirin induces apoptosis in all CRC cell lines regardless of their genetic background.

To further confirm that CRC cells undergo apoptosis but not necrosis following aspirin treatment, we harvested these cells and obtained protein lysates from all cell lines treated with varying doses of aspirin and examined the expression of various apoptosis-associated genes. Bax is a well-recognized apoptosis-activating protein that forms a heterodimer with Bcl2 (29). In support of our hypothesis, the protein expression of Bax was up-regulated in HCT116, SW480, and HT29 cells following aspirin treatment. In contrast, the expression of Bcl2 and Bcl-xl, the anti-apoptotic genes, was significantly down-regulated in all four cell-lines following aspirin treatment (Figure 1C). Collectively, these data confirm the previous notion that aspirin-mediated suppression of CRC growth is mediated through induction of apoptosis (27–29). Our findings highlight the fact that such effects occur in a dose-dependent manner, as well as in cells with or without MSI and wild-type or mutant PIK3CA, highlighting its broader role as a chemo-preventative for a broader population of patients with CRC.

Aspirin suppresses tumor growth dose-dependently in a xenograft animal model

Considering that it is difficult to examine the dose–response for the anti-tumorigenic effects of aspirin in clinical scenarios, we utilized a xenograft animal model derived from four CRC cell lines with different genetic backgrounds. We analyzed the impact of three different animal aspirin doses, which are comparable to the doses being evaluated in an ongoing human clinical trial, to gain insights into the effects of aspirin on CRC growth dynamics. These animal equivalent doses were categorized as low (15 mg/kg), medium (50 mg/kg), and high (100 mg/kg), which represent 100, 300, and 600 mg human doses respectively in the CAPP3 trial (22,23). From a genomic diversity viewpoint, we selected the four cell lines with different PIK3CA mutation and MSI status (Figure 2A), as these have been shown to associate with aspirin’s chemo-preventative efficacy. In all cell lines, tumor volume at day 15 was significantly reduced in animals treated with all aspirin doses, while the high dose aspirin yielded the most pronounced reduction in tumor volume versus the control group (HCT116; P < 0.001, SW480; P < 0.001, HT29; P < 0.001, LoVo; P < 0.001, Figure 2B). In the medium dose treatment, HCT116, HT29, and LOVO cell-derived xenografts exhibited the most significant attenuation of tumor volume compared to that of control animals at day 15 (HCT116; P < 0.001, HT29; P < 0.001, LoVo; P < 0.001, Figure 2B). In contrast, tumor growth was attenuated by low dose aspirin treatment primarily in HCT116 and HT29 cell-derived xenografts (P < 0.01 and P < 0.001, respectively, Figure 2B). Collectively, these data demonstrate that aspirin consistently suppresses tumor growth in all cell lines, but its efficiency appears to vary somewhat in cell lines with diverse genetic backgrounds.

Therefore, we next examined the effectiveness of aspirin on xenograft tumor growth based on the mutational backgrounds (MSI and PIK3CA status) of these CRC cell lines. Although aspirin treatment was not influenced by the MSI status, it was significantly more effective in xenografts derived from PIK3CA-mutated cell lines (HCT116 and HT29) versus the PIK3CA-wild type cell lines (SW480 and LoVo), particularly at low and medium doses (Figure 2C). In addition, xenograft tumor weight was consistent with the estimated tumor volume (Supplementary Figure 1, available at Carcinogenesis Online). Taken together, these data demonstrate that aspirin suppresses tumor growth in this animal model in a dose-dependent manner, and its efficacy, especially at lower doses, was significantly higher in a PIK3CA mutant background.

Aspirin suppresses tumor proliferation rates in CRC xenografts

In order to further analyze the anti-tumorigenic effects of aspirin in vivo, we collected a total of six xenograft tumors from each experimental group on days 3, 5, 7, 9, and 11, respectively, and the remaining tumors on day 15 (Figure 3A). First, we investigated the percentage of proliferative cells assessed by Ki67 staining over the course of the entire experiment. Although at individual time points, the differences for cellular proliferation with various aspirin doses were not statistically significant (Supplementary Figure 2A, available at Carcinogenesis Online), when we combined data for all cell lines, we observed a significant suppression of cellular proliferation with 50 mg/kg aspirin dose (P < 0.05) and 100 mg/kg regimen (P < 0.001; Figure 3B) in all cell lines. However, in HCT116 and HT29-derived xenografts, even the low-dose aspirin resulted in a significant reduction of cellular proliferation rates compared to controls (P < 0.05, Figure 3B). When all four cell lines were analyzed together, aspirin treatment suppressed proliferative cell percentages at all doses (all P < 0.01, Figure 3C). Next, we investigated whether aspirin inhibits cellular proliferation more effectively based on the PIK3CA mutational status and MSI status. Consistent with our previous findings for apoptotic rates, while the effectiveness of aspirin did not differ based on MSI status, it was significantly more effective in inhibiting proliferation of PIK3CA-mutant versus wild-type tumors at all doses (15 mg/kg; P < 0.01, 50 mg/kg; P < 0.05 Figure 3D), highlighting that aspirin’s CRC suppressive activity in part is mediated by inhibiting cellular proliferation.

Aspirin treatment induces apoptosis in CRC xenografts

Next, we examined the effect of aspirin treatment on the rate of apoptosis in xenograft tissues by TUNEL assay. Similar to cellular proliferation, at individual time points, we did not observe statistical differences between aspirin doses due to limited sample size (Supplementary Figure 2B, available at Carcinogenesis Online); however, when all the datasets were combined, the percentage of apoptotic cells significantly increased in a dose-dependent manner in xenografts derived from all four CRC cell-lines treated with aspirin (Figure 4A and B). In particular, all xenograft tissues exhibited a statistically significant increase in the number of apoptotic cells at 50 and 100 mg/kg doses, compared to controls. Subsequently we evaluated the effect of aspirin in the context of MSI and PIK3CA mutation status, and the subgroup analysis revealed that the effectiveness of aspirin-induced apoptosis was independent of these features (Figure 4C).

Figure 4.

Apoptosis inducing effect with different doses of aspirin in xenograft tumors derived from four CRC cell-lines. (A and B) Percentage of proliferative cells evaluated by TUNEL staining using each cell-line xenograft tumor tissues harvested on day 15. (C) Non-linear curve showing apoptotic cell percentages at each aspirin dose in PIK3CA mutated versus wild-type cells (left) and MSI versus non-MSI cells (right). **P < 0.01, and ***P < 0.001 by one-way ANOVA or two-tailed t-test as appropriate.

Mathematical modeling of data reveals that aspirin treatment results in increased cell death rates and reduced cellular division in xenograft tumors

While the experimental measurements provided insights into the number of proliferating and apoptotic cells, we additionally applied mathematical models to the experimental data in order to measure the rates of cell division and cell death. Having estimates of these kinetic parameters in turn allows us to perform calculations about how aspirin influences the probability of tumor cell colonies to survive and become established, which can give insights into the correlates of aspirin-mediated protection against CRC.

Mathematical models were fit to the temporal data of xenograft growth in order to obtain the best fitting parameter combinations. We have previously performed such an analysis in the context of in vitro experiments (27). In contrast to cultured cells however, xenografts exhibit three-dimensional spatial structure, which makes this effort more complex. In order to do this, we have designed and validated a model of three-dimensional tumor growth, as described in Materials and methods. We then used this model to extract parameter values from the data by fitting the model to the tumor growth data. Because measures of apoptosis markers were only available for a subset of the experimental runs and time points for each cell line and aspirin treatment regime, we calculated the death rates based on the average number of cells undergoing apoptosis for each experimental condition. Using these average death rates, we then fit the model to all individual tumor growth curves to estimate the division rate for each growing tumor. For each cell line and treatment condition, we then calculated the average division rate. We further compared the average division rate among all PIK3CA mutated and non-mutated tumors. Additional details are described in Materials and methods.

The Figure 5A illustrates the dependence of the estimated division and death rates on aspirin dose for the four cell lines. We can observe that as the aspirin dose increases, two separate trends emerge: (i) the death rates of cells increase and (ii) the division rates of cells decrease. To determine whether this result depends on our choice of parameters, we performed a sensitivity study (see Supplementary Figure 3, available at Carcinogenesis Online). While we fixed the cell volume, v = 2.57 × 10⁻⁶ mm³, we explored the uncertainties in parameters α and n, by varying α between 0.1 and 1 day⁻¹ (from left to right), and parameter n from 1 to 100 (from top to bottom). While the numerical values obtained for the rates L (live) and D (dead) change with these parameters, the trends remain the same: the death rates increase, and the division rates decrease with aspirin dose. Comparing the parameter estimates for PIK3CA mutated and non-mutated tumors, we find that the estimated division rates are significantly lower during aspirin treatment for PIK3CA mutated versus wild-type xenograft tumors (Figure 5B). This applies to all aspirin doses, and no significant differences in the estimated death rates were observed.

Figure 5.

Results from mathematical modeling analysis. (A) Cell death and division rates were estimated by fitting a mathematical model of spatial tumor growth to the xenograft data. Death rates (left panel), normalized with respect to the zero dose mean death rates. Death rates were calculated from the experimentally measured numbers of apoptotic cells using formula. Standard errors are displayed. Division rates (right panel), normalized: For each aspirin dose, and each cell line, and each animal, we fit the model to calculate the division rates. There are on average 20 repeats for each cell line/dose; the division rates are normalized with respect to the zero dose mean division rate. The results are plotted with standard error. (B) Histograms of calculated division rates for the repeat experiments, grouped as PIK3CA wild-type (SW480 and LOVO, orange) and PIK3CA mutants (HCT116 and HT29, blue). Aspirin dose 15, 50, and 100 mg/kg. The average normalized division rates for PIK3CA mutant cell lines were significantly different compared to those of PIK3CA wt cell lines for all aspirin doses. For dose = 20, P = 0.0008; for dose = 50, P = 0.014; for dose = 100, P = 0.008 (t-test with 2 sets of about 40 points each). (C) Prediction from the agent-based model regarding the fraction of tumors that fail to successfully expand due to 100 mg aspirin administration. The predictions are made for a range of virtual tumors, characterized by different turnover rates (baseline values of D/L). The fold-change of division and death rates was given by the average fold-change of the parameter estimates over all cell lines.

The dynamics of xenograft tumor colony formation are more significant in cell lines with higher cellular turnover rates

The division and death rates of the tumor cells are crucial determinants of the probability for a cell clone to successfully grow from low numbers, rather than to go extinct. To investigate the role of aspirin in tumor survival, we considered a stochastic agent-based model that tracks individual cells in a three-dimensional spatial setting. The rules of this simulation are described in Supplementary Materials and Methods, available at Carcinogenesis Online, and the corresponding results illustrated in Supplementary Figures 4–6, available at Carcinogenesis Online. Computer simulations were started with one cell in the middle of the space. Simulations were run repeatedly, and the fraction of runs during which the cell population went extinct or persisted were recorded.

The above-described parameter estimates were used for these computer simulations. The absolute division and death rates, however, are of limited value because they were obtained from robustly growing tumor cell lines, and newly transformed tumor cells are likely to be characterized by different rates in vivo. The relative effect of aspirin on the cell division and death rates, however, might be more broadly applicable. Therefore, we considered a collection of “virtual cell lines” with different baseline division and death rates, expressed by different ratios of D/L (ratio of death over division rates). Different tumors should lie at different points of this D/L range. Those with low D/L ratios are low turnover tumors with significantly more division than death events. Those with larger D/L ratios are higher turnover tumors where division and death events are more balanced. We assumed that the magnitude of the aspirin-induced change in these parameters for each dose is given by the fold increase or decrease, estimated from the growth data (Figure 5A), averaged over all cell lines.

Figure 5C demonstrates the probability of cells with different D/L ratios to successfully establish a growing cell colony in the absence of aspirin (black lines), and in the presence of 100 mg/kg aspirin dose (red line). The blue line shows the fraction of tumors that failed to grow due to aspirin. This is an indication of the extent that aspirin can prevent tumor occurrence simply based on the experimentally documented modulation of division and death rates. For low turnover tumors (low D/L), the effect of aspirin on colony formation is small. The effect, however, clearly increases at higher turnover rates (higher D/L). A 30% reduction in colony formation in the presence of aspirin is observed for D/L = 0.1. For values of D/L approaching 0.3, aspirin treatment is predicted to prevent all colony formation attempts, corresponding to maximum possible protection. Our data highlight that the variation in basic birth and death kinetics can potentially account for the different levels of protection conferred by aspirin against different kinds of tumors.

Discussion

In the present study, we comprehensively examined the anti-tumorigenic effects of aspirin on CRC kinetics in a series of xenograft animal studies. The present study was modeled based upon aspirin concentrations that are currently being interrogated in ongoing human clinical trials. We undertook a careful and exhaustive analysis for tumor growth data at multiple aspirin doses at various time intervals, and the percent of cells expressing the cellular proliferation marker Ki67 and the number of cells undergoing apoptosis were quantified in various subsets of tumors. Mathematical models were fit to these experimental data to estimate the rates of cell division and cell death for each aspirin dose. Overall, we observed that aspirin induced a dose-dependent increase in the cell death rates, as well as reduction in the rate of cell division.

Interestingly, we observed that the reduction in the cell division rate was significantly more pronounced for xenograft tumors derived from cell lines with PIK3CA mutations versus wild-type cells. The PIK3CA mutant cells harbor a mutation within the p110a subunit of PI3K, which results in constitutive activation of the PI3K-Akt pathway (30) and affects 20% of CRC patients (31–33). Furthermore, in the adjuvant setting, aspirin use has been shown to be substantially beneficial for overall prognosis in patients with PIK3CA-mutated tumors than in those without (34). Our results underscore the notion that the PIK3CA mutational status contributes to determining the extent of protection from CRC achieved by aspirin administration.

In contrast, MSI status did not appear to influence anti-tumorigenic efficacy of aspirin. Although the role of MSI status on the anti-tumorigenic efficacy of aspirin remains unclear, aspirin appears to be just as effective in both MSS and MSI cell derived tumors in the current study. The results of the present study were consistent with our previous study where we reported that aspirin-induced tumor growth attenuation was not affected by MSI status (27). Nevertheless, considering that only four cell lines were used in the present study, further studies may be needed to clarify whether MSI status has any influence on the effectiveness of aspirin in colorectal cancer.

Although the anti-tumorigenic properties of aspirin have been recognized for decades, the exact mechanisms that account for this protection are incompletely understood. Because aspirin is an anti-inflammatory agent and inflammation promotes carcinogenesis (35), it has been suggested that the observed protective effect of aspirin might be due to reduced levels of inflammation in the tissue microenvironment (36). The work presented here, however, shows that pharmacologically relevant aspirin concentrations can have direct effects on the division and death rates of tumor cells in vitro, which in turn can itself contribute to a reduced tumor incidence. Based on the estimated fold-change in cell division and death rates brought about by aspirin administration, we predicted the percentage of tumors that fail to grow due to aspirin administration at maximal dose. This was done in the context of a 3D computational model of tumor growth, using virtual tumors that are characterized by different baseline cell division and death rates. This yielded two insights. First, the magnitude of the aspirin-induced changes of cell kinetics shown in our experiments can bring about a reduction in tumor incidence that is consistent with the degree of protection observed in epidemiological data (2–4). Second, the extent to which tumor growth is prevented by aspirin is predicted to depend strongly on the turnover rate of the tumor cells, defined by the ratio of the death and division rates of cells. The incidence of low turnover tumors, where the cells divide and die less frequently, is predicted to be least affected by aspirin administration. Higher levels of protection are predicted for higher-turnover tumors, characterized by higher rates of cell death relative to the rate of cell division. These results provide new insights into how aspirin might contribute to protection against CRC, and further suggest that in addition to the PIK3CA mutational status, variation in the extent of protection might be explained by variation in the baseline turnover rate of the tumor cells.

Although the anti-tumorigenic effect of aspirin has been documented in several in vitro systems, the dose of aspirin utilized in the majority of such studies was often deemed too high (28,37). Therefore, for years, claims of the chemo-preventive effects of aspirin were met with skepticism. While it is true that aspirin doses studied in our in vitro assays might not reflect the true physiologically intact tumor microenvironment. However, since we were able to demonstrate the efficacy of these doses in an in vivo system, as well as the fact that the doses used in our study corresponded to salicylate concentrations found in the plasma of human patients who use aspirin to control arthritis (38), we feel that the aspirin doses in our study might be reasonable. In addition, the CAPP clinical studies have shown in Lynch Syndrome patients that daily intake of aspirin reduced the cancer incidence (17,19,20). In an attempt to identify the most optimal dose of aspirin, the CAPP3 study is currently testing three doses of aspirin on Lynch Syndrome patients (22)

Our present study illustrates that the doses used in CAPP3 can bring about changes in the in vivo kinetics of tumor cell division and death, which can result in a reduction in tumor growth that is consistent with the epidemiologically observed protective effect of aspirin against CRC.

In summary, aspirin is a cost-effective and well-tolerated gent that appears to have anti-tumorigenic effects through attenuation of tumor growth and inhibition of CRC recurrence. In this study, we for the first time used comprehensive in vivo experiments to systematically evaluate the anti-tumorigenic kinetics of aspirin’s activity, which in part are modulated through suppression of cellular proliferation and induction of apoptosis. We used mathematical modeling approaches to demonstrate that this can lead to a significant reduction of successful tumor growth, which might contribute to the epidemiologically documented protective effects of aspirin in patients at risk of CRC.

Funding

This work was supported by CA72851, CA184792, CA187956, and CA202797 grants from the National Cancer Institute, National Institutes of Health (NCI/NIH); RP140784 from the Cancer Prevention Research Institute of Texas (CPRIT); grants from the Baylor Foundation and Baylor Scott & White Research Institute, Dallas, TX, USA.

Conflict of Interest Statement: The authors have no conflicts of interest to disclose.

Authors’ contribution

Concept and design: T.S., S.T., D.W., and A.G.; Acquisition, analysis, or interpretation of data: T.S., S.T., D.W., C.R.B., and N.K.; Drafting of the manuscript: T.S., S.T., D.W., and A.G.; Statistical analysis: T.S., S.T., D.W., and N.K.; Administrative, technical, or material support: T.S., S.T., and A.G.; Supervision: A.G.

Glossary

Abbreviations

CRC

colorectal cancer;

CAPP

Cancer Prevention Programme

DMSO

dimethyl sulfoxide;

FAP

familial adenomatous polyposis;

FFPE

formalin-fixed paraffin-embedded;

IHC

immunohistochemistry;

MSI

microsatellite instability;

NSAID

non-steroidal anti-inflammatory drug