In Vivo Optical Molecular Imaging of Matrix Metalloproteinase Activity Following Celecoxib Therapy for Colorectal Cancer

Abstract

We present an optical molecular imaging approach to measure the efficacy of the COX-2 inhibitor celecoxib on tumor growth rate through its effect on MMP activity. A xenograft model of colorectal cancer was generated in nude mice, which were then randomized to receive celecoxib vs vehicle. MMP activity was measured by an enzyme-activatable optical molecular probe. A novel genetically engineered mouse (GEM) model of colorectal cancer was also used to assess celecoxib’s effect on MMP activity, which was measured by quantitative fluorescence colonoscopy. Subcutaneously implanted xenograft tumors were 84% (SD 20.2%) smaller in volume in the treatment group versus control. Moreover, treated animals exhibited only a 7.6% (SEM 9%) increase in MMP activity, versus 106% (SEM 8%) for untreated animals. There was an apparent linear relationship (r = 0.91) between measured MMP activity and tumor growth rate. Finally, in the GEM model experiment, treated murine tumors remained relatively unchanged in volume and MMP activity; however, untreated tumors grew significantly and showed an increase in MMP activity. This method may provide for the improved identification of patients for whom COX-2 inhibition therapy is indicated, by allowing one to balance the patient’s cardiovascular risk with the cancer’s responsiveness to celecoxib.

Introduction

Colorectal cancer is the second most common cancer in the developed world and a major cause of morbidity and mortality in the United States ¹. It has been known since the early 1990’s that up to 85% of colorectal carcinomas exhibit significantly elevated levels of cyclooxygenase-2 (COX-2), an enzyme whose overexpression promotes tumorigenesis through a number of mechanisms ^{2, 3}. One such mechanism is the activation of matrix metalloproteinases (MMPs), a family of proteases that degrade the extracellular matrix, facilitating tumor cell migration and invasion of surrounding structures ^3–5. Production of active MMPs in the extra-cellular matrix is thought to be partially mediated by prostaglandins. Additionally, higher grade colorectal cancers demonstrate increased expression levels of MMPs⁶.

Given the prevalence of COX-2 overexpression in colorectal cancer, there has been much interest in the application of COX-2 inhibitors, such as celecoxib, to the chemoprevention and chemotherapy of this disease. Early pre-clinical studies in animals showed pronounced reduction in tumor growth rates and formation of adenomatous polyps with celecoxib and other COX-2 inhibitors ^7–12. Two large-scale human trials also confirmed the chemoprotective effects of celecoxib to prevent the occurrence of metachronous adenomas ^{1, 13, 14}. However, it was through these trials that the significant cardiovascular morbidity associated with chronic, high-dose celecoxib use became apparent, and the trials were ended early out of concerns for patient safety ^15–18.

Celecoxib is a potent pharmacologic agent to prevent and treat colorectal cancer, but due to its cardiovascular toxicity, the drug currently plays a very limited role in the prevention as well as treatment of this disease. Identifying a therapeutically beneficial yet safe dose of the drug is difficult, as robust metrics for the drug’s efficacy are lacking. Moreover, the ability to identify patients who have a propensity for COX-2 positive adenomas and who would thus benefit from COX-2 inhibition for chemoprevention remains elusive. Conventional methods to investigate celecoxib’s effectiveness, such as through immunohistochemical staining or Western blot analysis of protein expression levels, require tissue sampling and ex vivo processing.

Additionally, more so than expression levels, it is changes in the activity of enzymes such as MMPs caused by celecoxib that represent true markers of the drug’s tumor suppressive effects. As the activity of MMPs is contingent upon a complex network of activators and inhibitors, a single parameter, static assessment of enzyme expression fails to fully capture the temporal activity response. Studying dosing effects through clinical trials with standard endpoints such as reduction in adenoma recurrence rates are costly and time-consuming. The inability to monitor the biochemical effects of celecoxib in its role as an anti-cancer agent precludes the calculation of the drug’s dose response curve as well as its therapeutic window.

A method to assess the efficacy of celecoxib by directly quantifying in vivo the drug’s effects on the molecular pathways that help drive tumor progression would greatly assist in the determination of appropriate, safe dosing regimens. We present herein an optical molecular imaging approach to measure the efficacy of celecoxib on tumor growth rate and aggressiveness through its effect on MMP activity. We do so with two different animal models of the disease. First, we use a nude mouse xenograft model to examine the relationship between celecoxib therapy and MMP activity in an implanted human colorectal cell line. We then apply our imaging technology to a genetically engineered mouse (GEM) model of colon cancer we developed to demonstrate that our method is readily clinically translatable.

We believe that the imaging method presented in this report may provide for the improved identification of patients for whom celecoxib chemoprevention is indicated, by allowing one to balance the patient’s cardiovascular risk with the cancer’s responsiveness to celecoxib. Additionally, this imaging method may improve our ability to select the appropriate dose of celecoxib chemotherapy in patients with metastatic disease by quantifying in vivo the efficacy of the drug on the tumor’s aggressiveness. Finally, this technique may assist in evaluating other pharmacologic agents that inhibit COX-2 by providing a cheaper and less time-consuming metric of effectiveness in pre-clinical trials.

Materials and Methods

MMP-activatable optical molecular probe

The commercially available MMP-activatable probe MMPSense ^19–21 (VisEn Medical, Woburn, MA) was purchased and stored at 4°C. This probe contains a gelatinase-cleavable peptide sequence and is cleaved by several members of the MMP family including MMP-2, MMP-9, and MMP-13. When administered, the probe is in an “optically silent” state, but when it is introduced into an environment containing MMPs, cleavage by these enzymes results in a many-fold increase in fluorescence signal. The probe fluoresces in the near-infrared (NIR), with excitation at 680nm and emission at 700nm.

Subcutaneous tumor experiment

The human metastatic colorectal cancer cell line HT-29 (ATCC, Manassas, VA) was grown in McCoy’s media supplemented with 10% fetal bovine serum (FBS) at 37°C with 5% CO₂. Before the cells reached confluence, they were harvested by trypsinization and injected subcutaneously into the flanks of (n = 10) nude athymic mice (nu/nu; Taconic, Germantown, NY). After the tumors were allowed to grow for 10 days, the animals were randomized into a treatment or control group. Animals in the former group received celecoxib 20mg/kg/day by intraperitoneal (i.p.) injection for 21 days; the injection was prepared by dissolving the drug in dimethylsulfoxide (DMSO). Animals in the control group received DMSO alone i.p. daily for 21 days. Tumor dimensions were measured daily with digital calipers to 0.1mm resolution, and tumor volumes were calculated by the following formula: V = L*W²*0.5, where V is tumor volume, L is length, and W is width.

MMP activity was measured at two time points during the 21 day treatment regimen, at days 0 and 15. Animals were administered 2nmol MMPSense via tail vein injection; twenty-four hours following injection, the animals were anesthetized with 2% isoflurane in 100% O₂ at 1L/min, and surface reflectance imaging of the MMPSense signal was performed using a commercial imaging system (bonSAI; Siemens, Erlangen, Germany). Excitation and emission filters optimized for NIR imaging were used. Regions of interest (ROIs) were drawn within the tumors to quantify the fluorescence intensity, which was normalized by the image exposure time so that data collected from animals imaged at differing exposure times could be directly compared.

Following the 21 day treatment regimen, the animals were sacrificed and their tumors harvested for subsequent ELISA analysis of COX-2, MMP-2, and MMP-9 protein concentrations.

Correlation of tumor growth with MMP activity

The relationship between tumor growth rate and MMP activity was evaluated in the following manner. MMP activity at days 0 and 15 in the subcutaneous animal experiment described above was measured using MMPSense, and the percent change in tumor fluorescence intensity across the 15 day time interval was calculated for each animal. To calculate tumor growth rates, the natural logarithm of the tumor volume data was computed, and the rate of change of these values per day was used as the tumor growth rate for each individual animal. Taking the logarithm of the tumor volume data allowed for the tumor growth curve, which resembles an exponential curve to a first order approximation, to be linearized. The growth rate was then plotted in a scatter plot against the change in MMP activity for each animal, and a best-fit line was calculated.

A genetically engineered colorectal cancer murine model

A novel mouse model for sporadic colorectal cancer that we developed ²² was used to assess for celecoxib’s effect on MMP activity in an animal model that more accurately reflects sporadic human colon cancer than xenograft implantation. Compound mutant mice were generated containing a homozygous Apc allele flanked by LoxP and a heterozygous latent mutant K-ras allele preceded by a floxed stop codon. Tumors were induced as has been previously described ²³. In brief, animals were fasted overnight prior to infection. All procedures were conducted in a sterile laminar flow ventilated hood using isoflurane inhalation anesthesia. A segment of the mid-descending colon was isolated through a midline abdominal incision in the lower abdomen and flanking clips were placed. Adenovirus was introduced through the anus via small caliber catheter which terminated within the isolated bowel segment. The abdominal incision was closed in two layers with nylon suture. This method of adenoviral vector administration ensured that tumors developed only in the descending colon, which was the segment imaged by the fluorescence colonoscopy experiment described below.

GEM colorectal cancer imaging experiment

Expression levels of a number of MMPs in a novel GEM model were investigated by RNA microarray analysis. Tumor tissue or normal colon tissue (control) was excised from 6 mice each group, fixed overnight in RNAlater solution (Ambion, Austin, TX), and total RNA isolated using an RNeasy kit (Qiagen, Valencia, CA). Sample integrity was verified using a nanodrop spectrophotometer, and labeling, hybridization, and scanning was performed. Array values were normalized and compared using the dChip microarray analysis package and filtered to examine all available murine MMP family members covered by the Affymetrix 430 v2.0 microarray. Fold change was calculated by comparing average intensity across groups.

In vivo imaging of MMP activity was subsequently performed. Mice (n = 7) were randomized into either a treatment group (n = 4 mice) or a control group (n = 3). Screening colonoscopies were performed once weekly post-tumor induction to assess tumor development; an in-house designed endoscopic optical molecular imaging system that allows for real-time, quantitative fluorescence imaging was used for these colonoscopies ^{24, 25}. Once tumors were readily visible, animals in the treatment group received celecoxib 20mg/kg/day dissolved in DMSO and injected i.p., while animals in the control group received DMSO without drug i.p. MMP activity measurement with MMPSense was performed twice during the treatment regimen, on days 0 and 15. Animals were administered 2nmol of MMPSense via tail vein injection, and 24 hours later, underwent fluorescence colonoscopic imaging with the in-house built system described above.

ELISA assays

Tissue samples from the subcutaneous tumor model were lysed in 500μL of RIPA buffer and then mechanically homogenized. The samples were stored in a 4°C refrigerator for 1 hour to allow cell lysis to occur; after this, the samples were centrifuged and the supernatant was retained. The protein concentration of the supernatant was first quantified using a protein assay kit (BCA Protein Assay Kit; Thermo Fisher Scientific, Rockford, IL). Expression levels of human COX-2, MMP-2, and MMP-9 were then measured in the lysate by ELISA (COX-2, MMP-2, MMP-9 ELISA kits; Calbiochem, San Diego, CA). A 4-parameter non-linear least squares best-fit curve was calculated from the standards to estimate protein content from the tumor samples. COX-2, MMP-2, and MMP-9 protein concentration values were normalized by the samples’ total protein content to account for variations in the volume of tumor harvested for ELISA analysis.

Statistical Analysis and Image Presentation

All data presented herein represent means plus or minus the standard error. Fluorescent images, when shown in the same figure, were collected at identical exposure times and are presented with the same window and level settings. Fluorescence intensity was quantified on surface reflectance imaging data by measuring ROIs within the tumors and calculating the mean pixel intensity. Evaluation for a statistically significant difference in fluorescence intensity between the treated versus non-treated subcutaneous tumors in Figure 2 was conducted by a one-tail t-test. Likewise, comparison between MMP-2, MMP-9, and COX-2 protein concentrations in treated and non-treated tumors was performed by one-tail t-test.

Figure 2.

In vivo MMP imaging of HT-29 tumors following celecoxib therapy. A, nude mice with subcutaneously implanted HT-29 tumors were imaged using MMPSense for changes in MMP activity during treatment with celecoxib. Two representative mice from the treatment and non-treatment group are shown. Red arrows denote the location of the tumors. Surface reflectance optical molecular imaging was performed on day 0 (before the initiation of therapy) and on day 15 (after 2 weeks of therapy). Initially, tumor volumes and MMP activity were comparable in both experimental arms; however, by day 15, untreated tumors were significantly larger and demonstrated greater MMP activity than the treated tumors. B, MMP activity with celecoxib therapy. Summary of data collected from all (n = 10) animals shows a marked decrease of MMP activity following celecoxib therapy relative to untreated tumors.

Results

We examined the relationship between MMP activity and COX-2 expression first through a subcutaneous xenograft model of colorectal cancer. As Figure 1 illustrates, tumor growth rates were significantly reduced in animals that received the COX-2 inhibitor celecoxib 20mg/kg/day for 21 days. The percent difference in tumor size between non-treated and treated animals was 84% (SD 20.2%). This result is similar to findings from previously published reports ²⁶.

Figure 1.

Effect of celecoxib therapy on tumor growth rate. Volumes of subcutaneously implanted HT-29 tumors were measured over the course of 21 days in which one group of animals received celecoxib 20mg/kg/day while the non-treatment group received vehicle. There was an 84% (SD 20.2%) difference in tumor volume by the end of the 21 day experiment between the non-treated and treated animals.

One of the purported mechanisms for COX-2 mediated tumor growth is MMP-associated destruction of extracellular matrix proteins, which in turn allows tumor cells to migrate and invade into the surrounding parenchyma. We explored whether the suppression of tumor growth by celecoxib could be related to a reduction in MMP activity; this investigation was performed using an MMP-activatable optical molecular imaging probe. MMP activity was measured in vivo on day 0, before the initiation of celecoxib therapy, as well as on day 15 of the 21 day treatment course. These data are presented for two representative mice of the total (n = 10) mice in Figure 2A and are summarized in Figure 2B. On day 0, the tumors for the treatment and non-treatment mice are similar in size and express similar levels of imaged MMP activity. However, by day 15, the untreated animal’s tumor has grown significantly, with a dramatic increase in MMP activity, while the treated animal’s tumor has grown minimally and reveals a mild increase in MMP activity. These results were similar and statistically significant (p < 0.01) across all animals in the treatment and non-treatment group, as shown in Figure 2B.

We validated the imaging results with ELISA measurements of COX-2, MMP-2, and MMP-9 expression levels in tumors harvested from the animals in the above experiment (Table 1). The concentrations of these three enzymes were found to be significantly lower (p<0.01) in the tumors of animals that had received celecoxib versus the tumors from untreated animals. Although this assay detects concentration rather than activity, these data help corroborate the in vivo imaging results.

Table 1.

ELISA analysis of COX-2, MMP-2, and MMP-9 expression in HT-29 xenograft tumors following celecoxib therapy

	Treated (ng enzyme/mg protein)	Untreated (ng enzyme/mg protein)	Statistics
COX-2	0.018874866	0.06328125	p < 0.01
MMP-2	0.00024662	0.000869084	p < 0.01
MMP-9	0.002917449	0.005953757	p < 0.01

We next sought to investigate whether there was a direct correlation between tumor growth rate and MMP activity. The scatter plot in Figure 3 plots the rate of growth for each mouse’s tumor against the change in MMP activity between day 0 and day 15 measured from that tumor. There is an apparent linear relationship between the two variables, with increased MMP activity associated with greater rates of tumor growth. Additionally, this representation of the xenograft experiment data readily depicts the marked effects of celecoxib therapy: the treatment animals’ data points are closely clustered together around the origin, signifying that these animals’ tumors exhibited both low rates of growth and minimal change in MMP activity during the treatment period.

Figure 3.

Correlation of rate of tumor growth versus change in MMP activity. Scatter plot charts the rate of tumor growth, presented as the rate of change of the natural log of the tumor volume per day, against the percent change in MMP activity between day 0 and day 15. The dotted line represents a best-fit line (r = 0.91) that illustrates a possible linear correlation between the two variables. These data suggest that optical molecular imaging of MMP activity can be used to predict tumor growth rates as well as responsiveness to therapy.

We then examined the effects of celecoxib therapy on MMPs in a novel mouse model for sporadic colon cancer ²². Mice that are homozygous for a floxed Apc allele and heterozygous for a latent activated Kras allele were treated with adenovirus expressing Cre recombinase, resulting in isolated tumors localized to the distal colon. Microarray analysis revealed that multiple members of the MMP family are over-expressed in this mouse model (Table 2). These mice were randomized to receive either celecoxib 20mg/kg/day or no treatment for 15 days; MMPSense imaging was performed on treatment days 0 and 15. Figure 4A shows quantitative fluorescence colonoscopic data from representative mice in both experimental groups. Before the initiation of celecoxib therapy, tumors visualized in the distal colons of both mice were of an appreciable size and had comparable MMP activity. After two weeks of celecoxib therapy, the treatment mouse’s tumor remained relatively unchanged in volume and MMPSense signal; however, the non-treatment mouse’s tumor showed a marked increase in size and MMP activity. These findings were consistent with the other animals in both treatment and non-treatment groups, as shown in Figure 4B, which shows a statistically significant decrease in MMP activity in the former versus latter group (p<0.01).

Table 2.

Microarray analysis of MMP RNA expression levels in the GEM model. Several MMPs that are particularly over-expressed are highlighted.

MMP	GEM mean	Control mean	Fold increase (GEM/control)
2	4477.61	1578.71	2.836246049
3	4516.2	56.11	80.4883265
7	11110.41	226.71	49.00714569
8	818.5	21.72	37.68416206
9	1340.36	26.79	50.03210153
10	9841.88	113.25	86.90401766
12	1130.41	33.29	33.95644338
13	7454.54	119.63	62.31329934
14	7249.01	301.79	24.02004705
19	338.31	74.15	4.562508429

Figure 4.

In vivo imaging of MMP activity following celecoxib therapy in a GEM model. A, quantitative fluorescence colonoscopy was performed in a GEM model prior to and two weeks into treatment with celecoxib 20mg/kg/day. Before initiation of therapy, tumor volumes were similar in these two animals representative of the treatment and non-treatment groups. However, by day 15, tumors in the non-treatment group had markedly grown, with an associated elevation in MMP activity, in comparison to the tumors in the treatment group. B, in vivo MMP activity with COX-2 inhibition. A statistically significant (p<0.01) decrease in MMP activity measured from tumors by quantitative fluorescence colonoscopy was found following celecoxib therapy in this GEM model. C, ex vivo surface reflectance imaging demonstrates high fluorescence intensity within a colonic wall tumor.

Discussion

Colorectal cancer is a significant cause of cancer-related morbidity and mortality in the United States ¹. The majority of colorectal carcinomas are known to over-express the enzyme COX-2, and much attention has been paid to examine the use of COX-2 inhibitors such as celecoxib for the chemoprevention and chemotherapy of this disease. Pre-clinical studies with the Apc^Min mouse model, which spontaneously generates adenomas in the small intestine ²⁷ due to a mutation in the murine homolog of the Apc gene ²⁸, have demonstrated that non-selective COX inhibitors are effective at decreasing overall polyp burden ^{29, 30}. A chemical carcinogen-induced rodent model ³¹ was used to illustrate that celecoxib is a potent inhibitor of carcinogenesis ³². Moreover, tumor growth rate was found to be suppressed in a dose-dependent fashion with celecoxib therapy in a nude mouse xenograft model ²⁶.

Human clinical trials have also been performed to investigate the chemoprotective effects of celecoxib. The drug was found to significantly reduce the number of intestinal polyps in patients with familial adenomatous polyposis ¹⁴. Two large-scale trials ^{1, 13} designed to study the reduction in recurrent adenomas with celecoxib showed a profound effect but were ended early when it became clear that there was an increased incidence of cardiovascular events, primarily stroke and myocardial infarction, in the treatment group. Thus, despite promising early results in animals and humans, larger trials revealed a very significant cardiovascular risk with celecoxib use ^{16, 17}, a limitation that has marginalized the role of COX-2 inhibitors for the prevention and treatment of colorectal cancer.

COX-2 upregulation promotes carcinogenesis and tumor proliferation by many purported mechanisms ^{2, 3}, one of which is the increased activation of MMPs ⁵. MMPs themselves have myriad pro-neoplastic effects, including promoting cellular proliferation, activating growth factors, and inhibiting cell death ⁴. MMPs have been shown to contribute to all stages of colorectal cancer development ⁶.

We hypothesized that the inhibition of MMP activity is a significant contributor to the anti-tumor effect seen with celecoxib and that we could quantify this reduction in protease activity through the use of an optical molecular imaging approach. We studied the effect of celecoxib in two animal models, a nude mouse xenograft model and a novel GEM model. The latter results in isolated distal colonic tumors that are homozygous for Apc deletion and heterozygous for activating Kras mutations, two of the early genetic modifications shown to be important in colon carcinogenesis³³. These genetic changes, along with the low tumor multiplicity and anatomic restriction to the distal bowel, makes this a more robust surrogate for human disease than other mouse models. In this study, we have demonstrated that not only does celecoxib dramatically reduce the rate of tumor growth in both xenograft and sporadic animal models, but it also avidly diminishes MMP activity within tumors. We additionally illustrate the ability of optical molecular imaging of MMP activity to predict tumor growth rate and responsiveness to therapy.

From a basic science perspective, the data reveal a direct correlation between tumor growth and MMP activity. This finding highlights the central role MMPs play in the expansion and invasiveness of colorectal cancers. Moreover, it supports the theory that MMP activity suppression is a key component of prevention of colorectal cancer growth with COX-2 inhibitors: the selective inhibition of COX-2 activity led directly to reduced MMP activity and the consequent retardation of tumor growth. One caveat is that COX-2 independent mechanisms for anti-tumor effects have been described for celecoxib ^{34, 35}, including a COX-independent pathway that involves inhibition of Notch-1, resulting in cellular apoptosis³⁶; however, very high concentrations of the drug are required for these effects, and this phenomenon is unlikely to be relevant in this experiment.

From a clinical perspective, we believe these results have important translational implications. First, the ability to characterize the effects of COX-2 on a molecular level has potential utility as an end-point for pre-clinical and clinical trials of new therapies that affect this molecular pathway to determine safe, effective dosing regimens. For example, new, non-NSAID drugs that suppress COX-2 activity and reduce adenoma formation, presumably without the harmful cardiovascular side effects, may be important chemopreventative options in the future ³⁷, and the imaging approach we have presented would be ideally suited for investigating further such pharmacologic approaches in subsequent studies. For example, in cases of familial adenomatous polyposis, patients present with 100’s to 1000’s of polyps, necessitating total colectomy. Prophylaxis with COX-2 inhibitors, such as sulindac or celecoxib, has been shown to be efficacious ¹⁴. Because of the known complications from chronic COX-2 inhibitor therapy, it would be most advantageous to minimize drug dosages by identifying in real time and with high spatial resolution tumor responsiveness to COX-2 inhibition.

We also believe that our method of measuring COX-2 activity in vivo opens the door for the personalized identification of patients who would most benefit from chronic COX-2 inhibitor therapy for the chemoprevention of colorectal cancer. Individual variations such as polymorphisms in the cytochrome P450 enzymes ³⁸ and the ornithine decarboxylase gene ³⁹ have clinically significant impacts on responsiveness to COX-2 inhibitors, as well as possibly the risk of cardiovascular morbidity with this therapy. Moreover, through histologic analyses of resected adenomas, we know that COX inhibition by regular aspirin use has a significantly increased reduction in colorectal cancer-related as well as overall mortality following the diagnosis of colorectal cancer for patients whose tumors express COX-2 ⁴⁰. Therefore, it is clear that the identification of patients, on an individual basis, whose tumors are quantifiably responsive to COX-2 inhibition therapy is essential for the risk stratification of initiating COX-2 inhibitor for colorectal cancer chemoprevention. However, our current methods to assess response, such as immunohistochemistry and agarose-based techniques, are limited by the fact that they are ex vivo methods, and that they measure enzyme expression levels, rather than activity levels. While it is ultimately the activity of an enzyme that is of critical importance, this parameter is often difficult to quantify in an ex vivo setting, as it is highly dependent for many enzymes upon the activity of other complementary as well as inhibitory enzymes in the intra- and extra-cellular milieu; as such, singular, ex vivo measurements of the enzyme’s concentration offer a limited snapshot of an enzyme’s contribution to carcinogenesis. In contrast, we have presented a molecular imaging approach that directly reports upon enzyme activity in a pathway upregulated by COX-2 overexpression in vivo and with high spatial and temporal resolution. We believe this technique has the potential to usher in a personalized medicine approach to the identification of patients most likely to benefit from colorectal cancer chemoprevention by COX-2 inhibition ⁴¹.

In addition to the identification of patients for COX-2 inhibition, we also believe the method’s ability to quantify reductions in MMP activity in vivo has important implications in improving the dosing regimens of COX-2 inhibitors. While the aggregate patient data in the APC trial showed a statistically significant reduction in adenoma formation with both low and high dose celecoxib administration, as well as no clear dose relationship for cardiovascular toxicity, subsequent analyses have revealed that genetic polymorphisms render some patients’ tumors responsive only to high-dose celecoxib ³⁸. The approach presented here offers the possibility of directly quantifying the changes in the molecular pathways affected by celecoxib therapy, thereby allowing one to adjust the dosing of the medication on a patient-by-patient, individualized basis.

We acknowledge several important limitations in our study. We did not directly assess whether the reduction in MMP activity following celecoxib therapy could be attributed to COX-2 independent pathways, by evaluating, for example, changes in MMP activity following celecoxib administration in a COX-2 negative mouse model of colorectal cancer. We also did not investigate our imaging technique in a metastatic model of colorectal cancer, and so we were not able to conclude upon our method’s ability to quantify response of metastatic tumors to COX-2 inhibitor therapy used in a chemotherapeutic role, which remains an intriguing application that has been evaluated in human studies as well ⁴².

In conclusion, we demonstrate an optical molecular imaging approach to quantify the effect of celecoxib therapy on MMP activity in a xenograft as well as a GEM model of colorectal cancer. We demonstrate that our approach is able to measure the impact of COX-2 overexpression by downstream activation of MMPs in real time, in vivo, and with high resolution, in both a surface fluorescence imaging and fluorescence colonoscopy setting. Our data illustrate a significant reduction in MMP activity following celecoxib therapy. We also demonstrate a direct relationship between MMP activity and tumor rate of growth, suggesting that the proteolytic activity of these enzymes is integral for the growth and progression of colorectal cancer. We believe that these results may potentially allow for a personalized medicine approach to the selection and treatment of patients with COX-2 inhibitors for colorectal cancer.

Abbreviations

COX-2

cyclooxygenase-2

GEM

genetically engineered mouse

MMP

matrix metalloproteinase

NIR

near infrared