Magnetic Resonance Imaging and Bioluminescence Imaging for Evaluating Tumor Burden in Orthotopic Colon Cancer

Quantifying tumor burden is important for following the natural history of orthotopic colon cancer and therapeutic efficacy. Bioluminescence imaging (BLI) is commonly used for such assessment and has both advantages and limitations. We compared BLI and magnetic resonance imaging (MRI) for quantifying orthotopic tumors in a mouse model of colon cancer. Among sequences tested, T2-based MRI imaging ranked best overall for colon cancer border delineation, contrast, and conspicuity. Longitudinal MRI detected tumor outside the colon, indistinguished by BLI. Colon tumor weights calculated from MRI in vivo correlated highly with tumor weights measured ex vivo whereas the BLI signal intensities correlated relatively poorly and this difference in correlations was highly significant. This suggests that MRI may more accurately assess tumor burden in longitudinal monitoring of orthotopic colon cancer in this model as well as in other models.

For MRI, various sequences may be applied, however, these have not been compared for which best depicts orthotopic colon cancer. Further, whether MRI or BLI better depict orthotopic colon cancer needs clarification. We compared two of the most widely used imaging modalities, MRI and BLI, for monitoring tumor progression in an orthotopic mouse model of colon cancer.

Results
Tumor take rate was 100% with HCT-116-Luc cells, with tumor formation in all 24 animals implanted. One week after cell inoculation, BLI and MRI were performed. Tumor growth at the primary site was evident by BLI and MRI as early as 1 week after cell inoculation (Figs 1A and 2A). Figure 1A shows BLI of animals longitudinally over a period of 6 weeks; and, Fig. 2A shows representative longitudinal MR images using different sequences of a tumor from the same mouse. Note that the T2-weighted images separate the tumor from the surrounding high-signal-intensity fat and low-signal-intensity air; this sequence ranked the highest among the sequences presented per mouse (n = 24, p ≤ 0.001) for colon tumor conspicuity, contrast, and margin delineation (Fig. 2B) and when sequences were presented randomly rated best for overall tumor conspicuity and margin delineation (Fig. 3A). Thus, the T2-based sequence was used to derive the tumor weight.
Quantification of the BLI signal and MRI-derived tumor weight over time demonstrated progressive increase in tumor growth (Figs 1B and 3B). The BLI signal intensities correlated with tumor weights measured ex vivo (r 2 = 0.57, n = 24, p ≤ 0.001; Fig. 1C), but the colon tumor weights calculated from MRI in vivo and the weights measured ex vivo were highly correlated with each other (r 2 = 0.998, n = 24, p ≤ 0.001; Fig. 3C3). Tumors outside the colon were observed in 3 of 24 animals that underwent MRI (Fig. 3D), suggesting metastatic disease versus seeding from the initial implantation. Tumors within versus outside the colon were not differentiated by BLI (Fig. 3D). The ex vivo tumor weight correlation with MRI-derived tumor weight was similar when mice with tumors outside the colon were excluded (r 2 = 0.996, n = 21). The correlation between BLI signal intensity and MRI-derived tumor weight was similar, whether including the animals with tumor outside the colon (r 2 = 0.57, n = 24) or excluding them (r 2 = 0.58, n = 21). Also, the correlation between ex vivo tumor weight with BLI signal intensity was similar, whether including the mice with tumor outside the colon (r 2 = 0.57, n = 24) or excluding them (r 2 = 0.58, n = 21). The Steiger method 18 was used to compare the data in Figs 1C and 3C, i.e. the correlation of the ex vivo tumor weight with the in vivo BLI-derived tumor signal intensity versus ex vivo tumor weight with the in vivo MRI-derived weight; these were statistically significantly different (p < 0.001, n = 24, Fig. 4), indicating that the MRI-derived tumor weight is superior to BLI-measured signal intensity values for estimating tumor burden.
Histopathological examination revealed clusters of large cells with a large oval or round acidophilic cytoplasm in well-differentiated tumors ( Fig. 5A-C). Histological examinations of tumor sections from mice euthanized at week 6 revealed small areas of hemorrhage or infarction within the tumor (Fig. 5C).

Discussion
T2-weighted MR imaging better depicted orthotopic colon tumors compared to BLI in the mouse model, enabling longitudinal tumor monitoring in vivo. Rating and ranking studies found the T2-weighted sequence to be superior to other MR sequences tested. The findings were supported by the high correlation between in vivo MRI-derived tumor weights and ex vivo tumor weights 19 . In comparison, BLI signal intensities correlated less well with ex vivo colon tumor weights. Moreover, unlike BLI, MRI distinguished colonic vs extracolonic tumors. Longitudinal BLI and MRI confirmed the progression of orthotopic tumor growth with time. As far as we are aware, multiple MRI sequences for imaging colon tumors have not been compared in orthotopic mouse models. In the current study, we found that among the MRI sequences, the T2-weighted sequence fared best for lesion detection and correlated highly with the excised tumor weight. These findings were supported by reader studies in which T2-based sequences ranked higher than other sequences for colon tumor conspicuity, contrast, and margin delineation. Intraperitoneal tumor can be seen with cancer metastasis in humans. Two-dimensional BLI, used as a semiquantitative tool to measure tumor burden 9,20 , did not distinguish between tumors inside and outside the colon. In addition, BLI tumor signal intensity is dependent on several factors such as the lesion depth and is limited by its need for tumors to express a luciferase reporter gene 9,13-16,20 ; thus, BLI is possible in cultured cell tumor models but can be significantly more challenging in spontaneous, transgenic, or xenograft models. In contrast to optical imaging techniques, it should be possible to perform MRI in most colon tumor-bearing mice, including spontaneous, transgenic and xenograft models, such as patient-derived xenografts.
Studies comparing BLI and MRI to gold standard excised tumor weight are wanting. Jost et al. 21 compared BLI and MRI derived tumor volumes in a mouse model of glioblastoma and found a moderate correlation (r 2 = 0.78) www.nature.com/scientificreports www.nature.com/scientificreports/ among smaller tumors (less than 40 mm 3 volume) and a comparative loss of this correlation (r 2 = 0.13) with larger tumors that grew beyond 40 mm 3 ; they then utilized both BLI and MRI together to measure tumor volumes for their preclinical evaluations; the study compared BLI to MRI measured tumor volumes but did not compare to gold standard excised tumor weights/volumes. In a hepatocellular carcinoma model, Wu et al. 22 found r 2 = 0.43 comparing BLI to tumor size by MRI. Thalhemier et al. 23    www.nature.com/scientificreports www.nature.com/scientificreports/ volume/liver volume) × 100) in a SW1222 colon liver metastasis model 24 whereas Fleten et al. 25 found r 2 = 0.55 comparing BLI of HCT116 colon liver metastases to excised tumor weight.
The ability to visualize and quantify tumor burden before therapy aids evaluation of treatment response. The present study demonstrated that to measure tumor growth longitudinally, T2-weighted MRI better depicted the orthotopic colon tumors than did the 2D BLI. Currently, nearly all BLI as practiced is performed in 2D. In the current study, 2D analysis did not distinguish colonic from extracolonic tumor but instead merged them into one global signal. Recently, bioluminescence imaging system have been equipped with software specifically designed to run 3D Diffuse Luminescent Imaging Tomography (DLIT) algorithms 26,27 . Whether these will aid distinguishing colonic vs extracolonic tumors or demonstrate improved correlation with tumor burden needs to be tested.
Relative merits of BLI and MR imaging are presented in Table 1. For 48 mice, the cost of MR may be approximately ten fold greater than BLI depending on institutional cost structure and number of mice imaged at a time. The balance of experiment cost vs time and image sensitivity need to be closely examined for every study. Given cost and time, BLI may be appropriate for several applications, particularly when alterations in tumor burden are expected to be large. It may also be appropriate for detecting smaller, live tumors expressing luciferase that may not be detectable by MR. We found that MR has the advantage of higher correlation with tumor burden than BLI, which may be needed in several applications such as assessing response to therapy. Approximately 0.75 correlation between BLI and MR have been found in other models such as glioma 28 . In the current study, compared to ex vivo tumor burden assessment, BLI correlation coefficient was approximately 0.75, whereas, MR was 0.99. MR also may be useful in transplantable models and in established genetically modified animal models where gene transfer of luciferase is difficult. In some situations, BLI and MR may be complimentary, such as using the former for earlier detection of tumors expressing luciferase and the latter for assessing the primary and metastases. Other techniques may also have advantages such as high-frequency US for tumor burden assessment 29 and functional assessment, and may be further compared in the future.
In the present study, orthotopic tumor growth was readily detected and observed over time by serial BLI and MRI measurements. BLI is a sensitive semi-quantitative tool for longitudinally monitoring orthotopic colon cancer. MRI can longitudinally detect orthotopic colon cancer and more accurately assess tumor burden. MRI should provide a safe, non-invasive method for monitoring the effects of new therapeutics, determining the optimal time for the collection of samples for molecular analysis, and correlating growth response with other functional response indicators. Colonoscopy is the primary screening test for colon cancer in patients. Radiologically, colon tumor detection has relied on CT colonography, which is commonly used in circumstances such as failed or incomplete colonoscopy or patient preference. MR is already used clinically for staging of rectal tumors. The current study suggests potential for clinical translation of MRI in the setting of colon tumors.

Conclusion
Among MR sequences tested, the T2-weighted sequence best depicted orthotopic colon tumors. BLI may have sufficient accuracy for assessing tumor burden in several scenarios evaluating orthotopic colon cancer and has advantages of low cost and ease of use. MRI better depicted the orthotopic colon tumor location and volumes/weights than did BLI; thus, MRI may more accurately assess tumor burden in longitudinal monitoring of orthotopic colon cancer in this model as well as in other models, which may be advantageous in a variety of applications.

Cancer cell lines. Human colon cancer HCT-116 cells were purchased from the American Type Culture
Collection. Cells were maintained in McCoy 5A medium containing 10% fetal bovine serum in a 5% CO 2 atmosphere.

Establishment of HCT-116-Luc cells. 293T cells were infected with a lentiviral vector carrying a firefly
luciferase gene, along with package vectors psPAX2 and pMD2.G, in 60-mm dishes using Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA), according to the manufacturer's protocol. Culture medium containing virus particles was collected 48 hours later and added to the HCT-116 cells, which were then selected with 1 μg/ml puromycin to eliminate uninfected cells. orthotopic animal model. For the orthotopic xenograft mouse model, 1 × 10 5 luciferase-expressing HCT-116 cells (HCT-116-Luc) were suspended in 30 μl of HBSS medium and injected into the cecal wall of 24 8-9-week-old male athymic nude mice, as previously described 3 . To minimize the retrograde leakage of cells into the peritoneum, a 1.5-mm diameter fat pad was harvested from the abdominal fat of a donor male nude mouse and was adhered using surgical glue (Dermabond; Ethicon, Inc., Somerville, NJ) on top of the injection site. If any fluid seepage occurred from the injection site, the mouse was excluded from the study to limit extracolonic tumor related peritoneal tumor formation. For tumor monitoring, groups of 3 mice were imaged then sacrificed each week to analyze the distribution and weight of tumors for a period of 6 weeks; another 6 mice were imaged longitudinally for 6 weeks and then euthanized to analyze tumor distribution and weight.

BLI.
For BLI, mice were injected intraperitoneally with 150 mg/kg of D-Luciferin (Caliper Life Sciences). After 10 minutes, mice were anesthetized with 2% isoflurane; they were then placed supine, with their abdomen toward the camera, and imaged, following the manufacturer's recommendations, for 10 seconds using the Xenogen IVIS 200 BLI system (Caliper Life Sciences). Using Living Image 2.5 software, regions of interest (ROIs) were drawn around the tumor and total radiant efficiency (in photons/sec/cm 2 /Sr) of the tumor was measured.

MRI.
For MRI, animals were anesthetized with 2% isoflurane. Body temperature was maintained at 37 °C by laying the mice on a bed heated with circulating warm water. The animals were imaged in a prone position. MRI data were collected on a Bruker MR (Bruker Biospec, Bruker Biospin, Billerica, MA) with a 7T/200-mm horizontal bore magnet and a 150 milli-Tesla/meter (mT/m) gradient insert. After scout images were obtained, the following imaging parameters were used to acquire the images: T2-weighted rapid acquisition with relaxation acquisition  www.nature.com/scientificreports www.nature.com/scientificreports/

Interpretation of MRI.
To rate the quality of colon tumor border delineation, contrast, and conspicuity, all MRI sequence images were presented randomly; blinded interpretation was performed at different times by a board-certified radiologist and a research scientist focused on small animal imaging. The images were rated on a scale of 1 to 5, with 1 = not seen, 2 = poor, 3 = fair, 4 = good, and 5 = excellent. In addition, the different sequence-derived images for each mouse were presented and ranked in order from 1 to 7, with 1 being the best.
Colon tumor measurements on MRI. Multiple 0.75-mm slices that spanned the entire colon were used to generate the ROIs. These were manually traced around tumors in each slice using the National Institutes of Health's ImageJ software 27 . The sum of the volumes of contiguous tumor ROIs was tabulated; to control for volume averaging, ½ the volume of the most cranial or caudal image was added. Assuming a tumor density of 1 g/ml, tumor volumes (mm 3 ) were converted to weight (g) for analysis.
Histological examination. Colon tumors were fixed with 4% formaldehyde, dehydrated in ethanol, and embedded in paraffin. Five-micrometer-thick sections were cut, dried, mounted on glass slides, deparaffinized in xylene, and rehydrated in a series of ethanol and phosphate-buffered saline 28 . Sections were stained using hematoxylin and eosin.
statistics. Pearson correlation coefficients between BLI tumor signal intensity, MRI-derived tumor weight, and gold standard ex vivo tumor weight as well as t-tests were performed using spreadsheet software (Microsoft Office Excel 2003, Microsoft, Seattle, WA). For comparing rating and ranking scores for MR sequences, linear mixed model was performed using SAS version 9 (SAS Institute, Cary, NC). Reader, time, and mice were treated as random effects to account for correlations between readings from the same mice. Pairwise comparisons were made between all pairs of sequences. Tukey-Kramer adjustment was used to control overall type I error rate at 5%. Differences in the correlation between ex vivo tumor weight and BLI signal intensity versus ex vivo tumor weight and MRI-derived tumor weight were evaluated using the Steiger method 18 and for this, the signal intensity was square root transformed while the MRI-derived tumor weight and true weight were log-transformed. All tests were 2-sided, and p-values of 0.05 or less were considered statistically significant.