Intermittent Dosing with Sulindac Provides Effective Colorectal Cancer Chemoprevention in the Azoxymethane-Treated Mouse Model

Colorectal cancer is the second leading cause of cancer-related deaths in the United States, and the third most common cancer in men and women (1). Although the 5-year survival rate for an early diagnosis of colorectal cancer in a localized area is relatively high (90.1%), the survival rate at a later diagnosis, after the cancer has metastasized, is low (13.5%; ref. 2). Because of this, early diagnosis and chemopreventive measures are necessary to increase the survival rate.

Sulindac is an NSAID that is proven to be an effective chemopreventive agent for colorectal cancer in mouse models (3, 4) and human clinical trials (4–6) Sulindac acts via cyclooxygenase-dependent mechanisms by inhibiting the activity of both cyclooxygenase-1 and -2 (COX-1/COX-2), enzymes implicated in tumor formation in colorectal carcinogenesis (4, 7), as well as COX-independent ways (8). NSAIDs and inhibitors of polyamine synthesis have been proven to suppress azoxymethane (AOM)-induced colon carcinogenesis (9). Sulindac activates polyamine catabolism and export by a COX-independent mechanism (10), and sulindac suppression of intestinal adenomas in the Apc^Min/+ mouse is polyamine dependent (11). It has been shown to act in both the initiation and promotion stages of colorectal carcinogenesis, and it has been demonstrated to reduce the increase in number of tumors if administered following the appearance of adenomas in the colon in the mouse (3).

Although regular use of sulindac and other NSAIDs can provide effective chemoprevention, these drugs have associated toxicities that have limited their use only to people with high risk of colorectal cancer (6). Sulindac is known to cause gastrointestinal and cardiovascular complications, including coronary artery disease, congestive heart failure, and diarrhea (12, 13). This is especially true of long-term use of sulindac, as daily dosing with approximately 158 mg for a mean period of 14 to 98 months led to 50% of participants developing gastrointestinal erosions. Subjects on a lower dose of sulindac generally experienced a decrease in adverse reactions (14). Another concern associated with the use of this type of chemopreventive agent is that withdrawal of the drug may lead to a rebound effect, causing incidence of adenomas to occur at an accelerated rate compared with those not treated with the drug. This may be of particular concern with larger doses of chemopreventives, as there is evidence that a higher dose of 400 mg twice daily of the COX-2 inhibitor celecoxib causes a trend toward more advanced adenomas following withdrawal of the drug (15). Therefore, there is a need to determine the minimal dosing of sulindac that can provide effective chemoprevention. One study demonstrated that short, discontinuous treatment with sulindac (300 mg for 2 months) was sufficient to provide chemopreventive effects for 12 months (16, 17). Another group of authors suggested that a future study be performed on short-term intermittent therapy of an NSAID such as sulindac, to determine whether this dosing scheme provides an effective means of chemoprevention with fewer risks of toxicity (18). For those on long-term chemoprevention, intermittent sulindac dosing may be a strategy for maintaining similar effectiveness to daily dosing with fewer toxicity concerns.

Investigation into the development of adenomas in mice has mainly been limited to terminal studies due to difficulties in tracking the progression of the disease (tumor count and burden) over time. Although white-light colonoscopy has been used to detect colorectal cancer in mice, it is technically challenging. One study demonstrated that it was able to identify 76% of colonic lesions, but it was unable to distinguish lymphoid infiltrates from small adenoma (19). Micro-CT has previously been used in mouse models, but it has limited sensitivity for detecting smaller adenoma and thus limited accuracy in estimating the total number of adenoma in a mouse (20, 21). Micro-CT also involves ionizing radiation, which is of concern for repeated imaging of the same animal due to possible effects on the disease model (22). An ideal imaging modality is simple to use, safe, accurate, and nondestructive. Optical coherence tomography (OCT) is a high-resolution, rapid imaging technology that uses near-infrared light to create cross-sectional images in scattering media like colon tissue (23). It is both minimally invasive and nondestructive, and it has previously been shown to rapidly and accurately detect adenomas and monitor disease progression in the mouse colon (3).

The current study tests whether intermittent dosing of sulindac (alternating 2 weeks on and 2 weeks off) can provide similarly effective chemoprevention as daily dosing, in a mouse model of colon cancer. This mouse model was treated with AOM injections prior to treatment with sulindac, as administration of drugs for cancer chemoprevention in a clinical setting is unlikely to start at a time prior to genetic mutations in the colon, and therefore, we believe our model is most clinically relevant. For instance, one current clinical trial (PACES) is investigating drug dosing during standard-of-care defined surveillance intervals to prevent colorectal cancers in patients that have already been treated for stages 0–III of colorectal cancer (24). In addition, the possible persistent chemopreventive effect of sulindac was examined in another dosing regimen (10 weeks on and 10 weeks off). This study used OCT to quantify the tumor count and tumor burden over time in all dosing groups. To verify the accuracy of OCT, tumor values obtained from histology and OCT images at the final time point of the study were compared.

Fifty-eight female A/J mice were used in this study. Thirty-four of these mice were treated with AOM subcutaneously and 24 were treated with saline as a control. See Table 1 for the number of mice included in each treatment group. Beginning at 6 weeks of age, the AOM-treated mice received 10 mg/kg of AOM (Sigma) once a week for 5 weeks, whereas the saline-treated mice received an equivalent volume of saline. Beginning one week after the last injection, mice were fed sulindac-laced chow, 100 ppm in AIN 93G food (Harlan Laboratories, Inc.), in four treatment groups shown in Fig. 1: (i) daily (sulindac-daily); (ii) for 2 weeks, then chow replaced with plain AIN 93G for 2 weeks, cycle repeated 5 times (sulindac-2); (iii) for 10 weeks, then chow replaced with plain AIN 93G for 10 weeks (sulindac-10); and (iv) plain AIN 93G with no sulindac (sulindac-none). Mice were included in analysis if they survived to at least 25 weeks of age (full study completion was 31 weeks of age). Three mice that died prior to the end of the study were excluded from OCT accuracy analysis due to lack of data regarding their final gross tumor counts.

The OCT system used in this study has been described in detail previously (22). Briefly, light from a superluminescent diode centered at 890 nm wavelength with a 150-nm full-width at half-maximum bandwidth was split into a reference arm and a sample arm using a 50:50 coupler. The sample arm consisted of a 2-mm diameter endoscope. The endoscope optics included a custom lens assembly coupled with a rod prism so that a side-viewing endoscope could be achieved. A glass envelope with a diameter of 2 mm enclosed the endoscope optics and was stationary during imaging. Linear and rotation actuators were used to control the lateral and angular position of the endoscope optics within the glass envelope. The reference arm contained a retroreflecting mirror at an equivalent optical path length to the endoscope envelope. Backreflected light from both the reference and sample arms recombined in the 50:50 coupler and was directed to a custom-built charge coupled device–based spectrometer, which measured the wavelength-dependent interference. A Fourier transform of this signal yielded the reflectivity of the sample as a function of depth. A total of 5,000 sampled spectra were transformed and combined to create the resulting 30 mm longitudinal × 2 mm deep images (5000 × 1024 pixels) with 4-μm axial resolution and 10-μm lateral resolution. These longitudinal cross-sectional images were obtained at 8 rotations around the circumference of the colon.

Imaging began 2 weeks following the final AOM (or saline) injection. The mice were imaged every 3 weeks (±1 week) until 20 weeks following the last injection, for seven imaging time points. Approximately 18 hours prior to the imaging session, the mice were placed in cages with wire cage bottoms and no bedding, and given Pedialyte in place of chow to minimize fecal matter in the colon during imaging. Immediately prior to imaging, the mice were weighed and anesthetized with a Ketamine/Xylazine mix injected intraperitoneally. Once the mice were completely anesthetized, the colon was flushed with 3 to 7 mL of warm saline to clear out any remaining feces or blood. The endoscope was coated with a biocompatible water-based lubricant and inserted approximately 32 mm into the colon. Eight longitudinal OCT images 45 degrees apart extending 30 mm into the colon were collected from each mouse at every time point. The mice were monitored following imaging until they were alert and ambulatory.

After the last imaging time point, the mice were euthanized and the distal 30 mm of the colon was explanted. The colon was then cut longitudinally and positioned flat with the lumen side facing up. The colon lumens were imaged along with a scale bar with a digital camera attached to a fixed stage. The colon lumens were analyzed using a dissecting microscope and tumors were circled on a printed image. The ex vivo analysis of the colons was used as the gold standard for tumor number, tumor diameter (used to compute tumor burden), and location.

The weight of each mouse was graphed over time, and the slope (weight change in grams per week) was derived from a linear mixed effects model with a random intercept to account for within-mice correlation and time, treatment, and the interaction between time and treatment as covariates. Slopes were compared between groups using one-way ANOVA to determine the significance of treatment on weight change.

The OCT images for each mouse at each time point were analyzed to determine tumor count and tumor burden. Adenoma requirements were standardized across all images using the diagnostic criteria determined and validated in a former study, through which 95% of adenoma were correctly diagnosed (23). Specifically, adenomas were distinguished from other structures due to a lateral dimension greater than 1 mL, moderate to marked protrusion of the colonic mucosa, moderate to severe image signal attenuation, and markedly faint to obscured tissue boundaries. Tumor burden was computed by measuring the maximal longitudinal dimension of each adenoma and assuming a spherical tumor using this dimension as the measured diameter. Using this method, the number of adenomas and total tumor burden were measured for each mouse at each time point.

The tumor count and burden from the OCT images at the final time point were compared with the tumor count and burden from the explanted colon using Pearson exact test. The slopes of tumor count and burden (change in number or mm³/week, respectively) were derived from a linear mixed effects model with squared root and logarithmic transformations, respectively, and random intercepts to account for within-mice correlation and time, treatment, and the interaction between time and treatment as covariates. Final time point tumor count and burden, as well as tumor count and burden slope, were compared across treatment groups using one-way ANOVA. Latency period (time until appearance of first tumor) was also compared between treatment groups using one-way ANOVA.

The OCT system used in this study produced high-resolution images that provided clear distinction between the different colonic tissues, including the mucosa, submucosa, and muscularis propria. It was also able to clearly distinguish adenoma from normal tissues and lymphoid aggregates, due to its ability to image about 1 mm deep into the colon tissue. An example OCT image series through four time points can be seen in Fig. 2.

Supplementary Fig. S1 shows the weight over time among the saline mice in the various sulindac treatment groups. There was no statistically significant difference in weight gain slopes between sulindac-2 mice and sulindac-none mice (P = 0.74). Sulindac-daily and sulindac-10 mice had statistically significantly lower weight gain slopes compared with mice that were not treated with sulindac (P < 0.01 and P < 0.0001, respectively). There was no appreciable difference in the visual observations of mouse health in the various groups. This study was fairly short and did not assess specific cardiovascular or gastrointestinal effects or other risk factors.

The Pearson correlation analysis demonstrated a strong correlation between the final time point OCT images and gross tumor counts (P < 0.00001), as well as gross tumor burdens (P = 0.008). OCT underestimated the number of tumors in mice with a large number of tumors, with an overall sensitivity of OCT to tumors identified in the gross image of 68.0%. There was only one false-positive tumor identified by OCT.

All sulindac groups were associated with an increase in latency period (time until appearance of first tumor). The average latency periods for sulindac-daily (15.7 weeks after conclusion of AOM treatment) and sulindac-10 (11.9 weeks) were not statistically different, as was expected (P = 0.11), and were both significantly longer than that of sulindac-none (5.0 weeks; P < 0.001 and P = 0.01, respectively). Sulindac-2 had a longer average latency period (10.4 weeks) than sulindac-none, with a difference that was just significantly (P = 0.04) shorter than either sulindac-daily or sulindac-10.

OCT imaging revealed that 92% of mice in all AOM-treated groups developed adenomas. More than half of all AOM-treated mice (64%) had tumors by 25 weeks of age (14 weeks after conclusion of AOM treatment). No mice in the saline groups developed adenoma. Figure 3 shows the progression of tumor count over the course of the study and the average slope of tumor count (change in number/week) for each treatment group. Sulindac-2 and sulindac-daily demonstrated a statistically significantly lower tumor count slope (P < 0.0001 for both) and final time point tumor count (P = 0.01 and P < 0.01, respectively) when compared with sulindac-none mice. The difference in tumor count slope between sulindac-daily and sulindac-2 was barely significant (P = 0.02), while the difference in final time point tumor count was not significant (P = 0.45).

Mice in the sulindac-10 treatment group had a statistically significantly lower overall tumor count slope compared with sulindac-none (P < 0.0001). There was no significant difference between the slopes of sulindac-10-off and sulindac-none (P = 0.8).

Figure 4 shows the progression of tumor burden for each treatment group over the course of the study, and the average slope of tumor burden [change in volume (mm³) per week] for each treatment group. All treatment groups showed a statistically significant reduction in tumor burden slope when compared with the mice that received no sulindac (P < 0.001). There were no statistically significant differences between sulindac-daily, sulindac-2, and sulindac-10 when these treatment groups were compared with one another.

The results of this study indicate that a variety of sulindac dosing regimens are effective in suppressing tumor formation in the AOM-treated mouse model of colon carcinogenesis. Although preclinical, translational and clinical evidence supports the efficacy of NSAIDs, including sulindac, to suppress colon cancer development, an important unanswered question is whether NSAIDs need to be administered continuously to interrupt this process. Clinical trial results suggest that colorectal adenoma growth rebounds to rates exceeding placebo groups in patients previously treated with celecoxib (a selective COX-2 inhibitor) when celecoxib is terminated (15). A major finding of the study described here is that the rate of colon adenoma growth does resume after cessation of sulindac treatment in the AOM-treated mouse model (compare sulindac-none, sulindac-daily, and sulindac-10 groups in Fig. 3). However, that growth does not exceed the sulindac-none group. As seen in Fig. 4, the rate of tumor burden increase in the sulindac-10 group actually remains similar to that of sulindac-daily, indicating a persistent sulindac effect on tumor burden, at least in this context.

Mice that were not treated with sulindac developed high tumor counts and tumor burdens by the end of the study. The tumor count growth rate for this group was statistically significantly higher when compared with sulindac-daily and sulindac-2, while the tumor burden growth rate showed a statistically significantly greater value when compared with all treatment groups. The high tumor burden for this group evident by 31 weeks of age limited the length of the study to the 7 time points that were observed. However, mice in the sulindac-daily group showed minimal increase in tumor count, with an average of 0.83 tumors by 31 weeks of age. Similarly, tumor burden grew at a slow rate. These findings compare favorably with previous studies that have demonstrated that sulindac acts as a chemopreventive for colorectal cancer (5, 7, 23).

Although daily doses are most effective, the sulindac-2 regimen provides similar effects. The sulindac-2 treatment group had a minimally statistically significant difference in tumor count slope when compared with sulindac-daily, but it had a very statistically significant difference when compared with sulindac-none. With regards to tumor burden slope, there was no statistical difference between sulindac-2 and sulindac-daily. These findings suggest that 2-week cycles of sulindac separated by 2 weeks of no drug may be a viable alternative to daily intake of sulindac in colorectal cancer chemoprevention.

The sulindac-10 dosing regimen effectively delayed tumor appearance and growth. Sulindac-10 mice had biphasic tumor number and burden slopes, linked to the varying sulindac dosage received. The latency periods for both the sulindac-10 mice and the sulindac-daily mice were statistically indistinguishable, which is expected as the treatments during the first half of the study were identical. During the “off” portion of treatment, tumor number slope was similar to the sulindac-none group. The tumor burden slope for sulindac-10 was significantly lower than sulindac-none; however, as tumor burden appears to follow an exponential growth pattern as seen in Fig. 4, it is possible that the tumor burden slope would approach that of the sulindac-none group had the sulindac-10 mice been taken to later time points. These findings suggest that sulindac only provides effective chemoprevention while it is being administered, effectively delaying carcinogenesis. Once treatment is stopped for more than a few weeks, tumor appearance and growth may follow trends similar to mice who received no sulindac. Despite this, there was no apparent rebound effect during the time period that was studied, as tumor appearance and growth for the sulindac-10 treatment group did not surpass that of sulindac-none at any point. These results are consistent with those found by Takayama and colleagues, in which sulindac was shown to provide chemopreventive effects for 12 months after treatment (16, 17). The latency periods of sulindac-daily, sulindac-2, and sulindac-10 all demonstrated a statistically significant difference when compared with that of sulindac-none, but sulindac-daily and sulindac-10 had a significantly longer latency period than sulindac-2, suggesting that continuous dosing of sulindac best prolongs latency period.

OCT has previously been shown to identify distinct layers of tissue in the colon (mucosa, submucosa, and muscularis propria) and distinguish between normal tissue, lymphoid aggregates, and adenoma in mouse models of colorectal cancer (23, 25). This type of imaging technique has also previously been utilized to observe disease progression over time by detecting changes in the colonic tissue of these mouse models (23). In this study, OCT performed as expected. Imaging took only 5 minutes, required minimal training, and as the endoscope envelope was stationary and the optics were always in a known position and orientation, location and tumor diameter were easily measured.

This study used endoscopic OCT time-serial imaging to examine the effects of varying sulindac dosing schedules on chemoprevention efficacy. Although there has been some research into the long-term effects of continuous sulindac (16, 17), no previous studies to our knowledge have tested the efficacy of intermittent doses of sulindac. Endoscopic OCT allowed the observation of progression of tumor number and tumor burden at multiple intermediate time points. This nondestructive imaging system gave us the means to visualize trends in disease progression in each treatment group, especially enabling the effect of changes in dosing pattern (going from daily to no sulindac in the sulindac-10 group) to be quantified.

No mouse deaths were attributable to the OCT imaging procedure. Thirty-six of the 58 mice lived through the entirety of the study. Forty-four mice survived long enough to be utilized in data analysis. Eight of the mice were euthanized due to exhibiting morbidity criteria, including excessive bleeding. One mouse died from anesthesia. The remaining mice were found dead in their cages from unknown causes. The highest percent attrition occurred in AOM-treated mice that received no sulindac, the group with the highest tumor number and burden.

A comparison of the final OCT tumor count and gross tumor count, as well as the final OCT tumor burden and gross tumor burden, showed a significant correlation between OCT and gross tumor images. However, in this study, OCT frequently underestimated the number and burden of adenoma, leading to a sensitivity of only 68%. This finding is lower than our previous studies, one of which identified adenoma with a sensitivity of 89% (3) and another that identified adenoma with a sensitivity of 95% (23). OCT was highly specific with only one false positive obtained in the study. The low sensitivity and underestimation of tumor burden could be due to sampling error, as images were only obtained every 45°, and it is possible that some tumors were overlooked. However, this is unlikely as the circumferential distance between images was less than 1 mm. Adenoma may be compressed when the endoscope is inserted, which is likely the cause of most underestimation in tumor count and burden. In highly diseased colons, individual adenomas in close proximity to each other often appear as a single, larger adenoma. Image quality was also lower in highly diseased colons (in the sulindac-none group), as it was sometimes extremely difficult to remove residual feces or blood. This study had more highly diseased mice than previous studies using OCT to measure tumor number, likely explaining the lower sensitivity. Despite some inaccuracy, a favorable comparison of between-group statistical analysis using OCT at the last time point or gross image tumor counts strongly indicates that OCT is an appropriate measurement tool, with the advantages of being nondestructive, time serial, and rapid. This supports our previous findings that show that OCT is successful in tracking colonic disease progression (23).

Although our endoscopic OCT system was successful in measuring tumor count and tumor burden in a useful fashion, future studies may be made more sensitive by using a spiral-scanning OCT system recently developed by the authors that can obtain three-dimensional images of the colon (26). Full rendering of the colon would allow more sensitive detection of adenoma, as well as a more accurate measurement of their size and shape.

Our results show that, although daily doses of sulindac are most effective in reducing tumor count and tumor burden, intermittent doses of sulindac in a 50% duty cycle with an overall 4-week period, as shown by the sulindac-2 treatment model, may also provide highly effective chemoprevention of colorectal cancer while receiving only half the overall dose of sulindac compared with the sulindac-daily group. Regular treatment with sulindac appears to be necessary to show prolonged chemopreventive effects. Intermittent dosing for chemopreventive agents has previously been tested in animal models with drugs such as oltipraz and combinations of arzoxifene and the rexinoid LG100268. These studies demonstrated high rates of efficacy in chemoprevention with minimal side effects, which showed promise for clinical use of the drugs (27, 28). This type of short-term intermittent therapy has also been previously tested in synthetic lethality-based cancer therapy in mice to target genes linked to ovarian, colorectal, and lung cancers (18). These previous studies have shown potential for intermittent dosing to lower toxicity levels while maintaining the efficacy of these drugs. In addition, combinations of chemopreventive agents have been shown to provide only minimal benefits with regards to tumor count, while providing significant improvement to tumor burden, suggesting the benefits of paying careful attention to tumor burden as a clinical endpoint in this type of study (29). We only assessed the effect of intermittent sulindac dosing on adenoma number and burden, and not on adenoma grade, in this model. However, it has previously been shown in the Apc^Min/+ mouse that sulindac dramatically reduces the number of both high-grade and low-grade intestinal tumors, but that of the remaining adenoma, there is a higher percentage of advanced adenomas compared with control (no drug) mice (30). Future studies should assess tumor grade outcomes in relation to intermittent dosing of sulindac. If intermittent doses of sulindac are successful in chemoprevention of colorectal cancer, toxicity concerns associated with this NSAID may also be mitigated.

E.W. Gerner has ownership interest (including patents) in Cancer Prevention Pharmaceuticals and Gerner Pharmaceutical Investors. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Conception and design: E.W. Gerner, J.K. Barton

Development of methodology: P.F. Rice, J.K. Barton

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.C. Nymeyer, P.F. Rice, J.K. Barton

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Chandra, P.F. Rice, E.W. Gerner, J.K. Barton

Writing, review, and/or revision of the manuscript: S. Chandra, A.C. Nymeyer, P.F. Rice, E.W. Gerner, J.K. Barton

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.C. Nymeyer, P.F. Rice

Study supervision: P.F. Rice, J.K. Barton

Research reported in this publication was supported by the NCI of the NIH under award number R01CA109385 (S. Chandra, A.C. Nymeyer, P.S. Rice, and J.K. Barton) and was supported in part by the Biostatistics Shared Resource, University of Arizona Cancer Center (P30 CA023074).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.