Pancreatic adenocarcinoma is the fourth leading cause of death from cancer in humans,1 with a dramatic increase in incidence over the past 50 years.2, 3, 4 Each year about 10/100,000 people die of pancreatic cancer. The 5-year survival rate is only about 3%, and the median survival is less than 6 months.5, 6 Treatment strategies involving combined radiochemotherapy or radioimmunotherapy7 are only palliative and do not offer a significant benefit for long-term survival. The only curative treatment is surgical removal of the tumorous pancreas. Although the protocols to diagnose the disease8, 9, 10 and its metastatic spread11, 12 are rapidly evolving, only 5–15% of the tumors are resectable.13 This is either due to local invasion of adjacent structures such as stomach or major vessels or due to metastasis, mainly to the liver and lymph nodes. Advances in pathologic classification and genetic evaluations have improved the understanding of pancreatic cancer, but important aspects of the biology of the disease remain unknown.5
Faced with such a bad prognosis, new treatment strategies for pancreatic cancer have been studied extensively. As an experimental model, tumors have been implanted subcutaneously primarily in mice. However, disease in these animal models does not resemble the human disease adequately, because these tumors are often surrounded by a pseudocapsule and rarely metastasize. Furthermore, no adjacent major anatomic structures (such as major vessels or gastrointestinal organs) are invaded by the tumor.14, 15 In 1985 the first orthotopic pancreas tumor model with the tumor growing in the pancreas itself was presented.16 Orthotopic pancreatic tumors more closely resemble the biologic characteristics of the human tumor, and thus represent the clinical situation in patients better. Several orthotopic models have been recently described with either transplantation of histologic intact tumor specimen or injection of cell suspensions into the pancreas.14, 15, 17, 18 Due to the location of the tumors within the abdomen and not superficially under the skin, tumor growth is difficult to monitor. Palpating the tumor can lead to an inaccurate estimation of its actual size; direct observation requires invasive surgery. Therefore, the assessment can only be performed reliably at final autopsy. A noninvasive technique to monitor the tumor size and to assess the therapeutic efficacy of intervention would thus be advantageous over existing techniques and highly desirable. This also applies to other pancreatic diseases such as diabetes for which several animal models do exist.19
In order to monitor tumor growth within the pancreas, diagnostic radiologic methods such as magnetic resonance imaging (MRI) can be used. MRI is advantageous over computer-assisted tomography (CT), due to its better soft tissue contrast, and is superior to positron emission tomography (PET), due to its better spatial resolution.20 Clinical MRI scanners already offer good spatial resolution, especially with dedicated animal coils that are optimized to receive a signal from a small volume such as a mouse.21, 22, 23 However, pancreas imaging in mice is particularly challenging since the organ is not a defined solid retroperitoneal organ as it is in humans, but rather a thin membrane, spread throughout the upper abdomen and lying immediately adjacent to the gut. If no precautions are taken, artifacts due to peristaltic and breathing movements as well as from the gas within the intestines can easily affect the image quality.
We here report on the MRI of the pancreas and of pancreatic adenocarcinoma in an orthotopic mouse tumor model and propose that MRI offers an optimized imaging technique and a possibility to noninvasively monitor orthotopic pancreatic tumors.
MATERIAL AND METHODS
The animal experiments were performed in accordance with the guidelines of the Massachusetts General Hospital's animal handling protocol. Five SCID mice (Jackson Laboratories, Bar Harbor, MA) were used for our study. For tumor implantation and imaging, the mice were anesthetized with an injection of a mixture of ketamine/xylazine intraperitoneally (80 mg/kg per 12 mg/kg BW, Parke-Davis, Morris Plains, NJ, and Miles Inc., Shawnee Mission, KS).
Orthotopic tumor model
CAPAN-2 human pancreatic adenocarcinoma cells24, 25 (HTB-80, American Type Culture Collection (ATCC), Manassas, VA) were implanted into 4 SCID mice. The cells were grown in MCCoy's 5A medium supplemented with 10% FBS (Cellgro Mediatech, Washington, DC) in humidified 5% CO2 atmosphere (ThermoForma incubator, Thermo, Marietta, OH) according to the ATCC guidelines. For the implantation, a median incision was made in the abdomen and the peritoneum was separated from the abdominal wall and opened.14 The splenic portion of the pancreas was carefully exposed, and 2 × 106 tumor cells in PBS (total volume of 15 μl) were slowly injected into the pancreatic tissue. The incision was closed in 2 layers (peritoneum and abdominal wall) with interrupted sutures using 4.0 Vicryl (1.5 metric; Ethicon, Somerville, NJ). The mice tolerated the procedure well without any complications.
MRI was performed using an FDA-approved 90 cm bore 7T human head MR scanner (NMR Magnex Scientific Inc., Concord, CA, and Siemens, Erlangen, Germany). After native T1- and T2-weighted images were acquired, all mice received gadolinium (Gd)-DTPA (0.1 mmol/kg BW, Magnevist™; Schering, Berlin, Germany) intraperitoneally 15 min prior to further imaging to enhance the contrast of the pancreas to the surrounding tissue and to better delineate the tumors. Due to the large surface area of the peritoneum, the intraperitoneally injected Gd-DTPA was absorbed very quickly into the circulation without showing any peritoneal enhancement. To test for gut relaxation, 2 of the 5 mice (one without and one with tumor) also received butylscopolamine (0.5 mg/kg BW; Buscopan™, Böhringer Ingelheim Pharma KG, Ingelheim, Germany), a preparation widely used to relax the gut and to decrease peristaltic movements of the guts. To test another method to reduce peristaltic movements, the remaining 3 tumor-bearing mice were starved for 6 hr prior to imaging to empty the stomach and guts. All mice were positioned in a supine position and fixed with tape on all extremities and the tail. The abdomen was centered between 2 coils (5 and 10 cm in diameter; NMR Imaging Center, Massachusetts General Hospital, Charlestown, MA) consisting of 2 copper loops. The 10 cm coil functions as a transmitting coil that sends out the pulse, and the other as receiver of the MR signal. Both coils were tuned manually for each animal for optimal performance. The imaging protocol was the same for all mice and consisted of a T1-weighted spin echo and a T2-weighted turbo spin echo sequence. The sequence parameters are shown in Table I. The imaging planes were coronal and transverse.
Table I. MR Pulse Sequence Parameters Used in the Study
The field of view (FOV) was slightly adapted to the positioning of the mouse abdomen.
In order to confirm the correct identification of the pancreas on the MR images, the mouse without tumor was sacrificed after imaging. The abdomen of the mouse was opened and compared to the images, identifying the entire length of the pancreas. The organs were partly removed and the mouse was imaged again to assure correct identification.
The tumor-bearing mice were imaged consecutively over several weeks (prior to implantation, 10, 15 and 30 days after implantation), using the imaging protocol described above.
The tumor-bearing animals were sacrificed on day 30; their pancreases were excised, snap-frozen and cut into 5 μm sections utilizing a cryostat (CM1900, Leica, Bannockburn, IL). Histology was performed with H&E staining. The sections were examined using a Nikon Eclipse E400 light microscope (Nikon, Melville, NY).
First we examined the delineation of the pancreas itself in MR images in a mouse without tumor. Figure 1a shows the white light image of the abdomen of the mouse that was imaged immediately prior to sacrifice; the corresponding MR images are presented in Figure 1b–e. The tail of the pancreas is closely adjacent to 2 other organs: spleen and stomach. The spleen presents with a triangular shape in coronal slices; the stomach is an oval structure with a low (or mixed) signal lumen, continuing to the right into the duodenum and lying adjacent to the right lobe of the liver. The pancreas's tail is usually situated distal of the stomach, medial of and partly covering the spleen and also the upper pole of the left kidney, filling a triangular-shaped space between stomach, kidney and spleen. Following the tail of the pancreas to the right allows easy delineation of the body of the pancreas, which lies immediately adjacent to the duodenum. In order to identify the pancreatic structures it is crucial to obtain thin adjacent slices of 1 mm or less slice thickness without any distance factor, which allows differentiating the pancreatic tissue from adjacent gut structures. Although deployed previously, the use of contrast agent proved not to be necessary for the identification of the pancreas since the adjacent structures (spleen, stomach and duodenum) can be used as landmarks.
We used N-butylscopolamine in 2 mice (without and with tumor) to reduce artifacts from the peristaltic movements of the intestine, but this approach can only be partially recommended. Though N-butylscopolamine relaxes the gut, it has a very short half-life, taking effect for only about 5–10 min. This makes repeated injections necessary. Our second approach to reduce peristaltic derived artifacts was to deprive the mice of food 6 hr prior to imaging. This is a more feasible method to relax the gut since there is no digestion and therefore decreased peristaltic movement. A reduction of moving artifacts from breathing could be achieved by placing the mouse in a supine position since the excursion of the abdomen during breathing is smaller than in the prone position.
All tumor-bearing animals were imaged with and without application of Gd-DTPA. On T2-weighted images (Fig. 2c and 2e) the tumors appeared as high signal intensity structures. Areas of heterogeneous signal within the tumor can be delineated with even higher signal intensity on T2-weighted images (Fig. 3b and 3d), corresponding to lower signal intensity without much enhancement in T1-weighted images (Fig. 3a and 3c). After application of contrast agent the tumors showed only a peripheral enhancement on T1-weighted images on days 10 and 15 (Fig. 2b and 2d), and on day 30 an additional low enhancement within the tumor as well (Fig. 3a and 3c). The growth of the tumors could be monitored and the size measured by MRI, showing a doubling in size from 2–4 mm (mean) in 2 weeks.
The excised pancreas revealed the presence of well-defined tumor masses at the site of the implantation in the tail of the pancreas (Fig. 4). The tumors show heterogeneity within the mass, correlating with the MR images. On T2-weighted images (Fig. 3b and 3d) tumors showed high signal intensity in the center of the tumor; on T1-weighted images (Fig. 3a and 3c) we observed a lower signal intensity, which corresponded to an area of less densely packed and less intensely stained cells and more pancreatic ducts on histologic sections.
With the method described in this report, the pancreas and orthotopic implanted pancreatic tumors could be reliably imaged in mice utilizing MRI. We could show that the size of orthotopic implanted tumors can be evaluated with MRI, which has evident advantages over CT, PET or ultrasound in terms of soft tissue contrast, radiation, resolution or user independence. Major problems in imaging of the pancreas in small rodents (motion artifacts from breathing and peristaltic) could be reduced to produce acceptable images with our protocol. We could even distinguish inhomogeneities within the tumor tissue that corresponded to the histologic appearance.
To our knowledge there are only 2 reports on rodent pancreatic MRI.18, 26 In comparison, our approach used a higher spatial resolution (0.1 × 0.1 × 0.5 mm) in both T1- and T2-weighted sequences. He et al.18 achieved an in-plane resolution of 0.1 mm2 for T2, but only 0.3 mm2 for T1. Seki et al.26 reported an in-plane resolution of 0.4 mm2 for both T1 and T2. The work by He et al. described a method to monitor the viability of the mouse during the imaging session by imaging the blood flow in major blood vessels. Although an interesting and advanced method, this complicates the procedure and does not seem to be necessary in our view since the viability of the mouse can be easily monitored by observing the breathing motion of the chest and abdomen that can be seen on each image using an adequate window. Seki et al.26 found that the splenic segment showed a poor contrast to surrounding tissue in rats, which are considerably larger. Our results in mice showed a nice delineation of the splenic segments of the pancreas from the spleen and the stomach.
The behavior of a tumor implanted in the pancreas is not necessarily the same as that of a tumor originating from the cells in the pancreas. We did not observe any desmoplastic reaction that is commonly observed in human pancreatic adenocarcinoma. A possible explanation for this different appearance may lay in the fact that human cells were xenografted into a mouse. Ideally, a mouse model with endogenous pancreatic tumors should be used; these are, however, not easily available. On the other hand, orthotopically implanted tumors are certainly closer to the clinical situation in a patient than subcutaneously implanted pancreatic tumors.
This is the first report to utilize a paramagnetic contrast agent to image orthotopic implanted tumors in the pancreas of mice. He et al.18 did not use a contrast agent due to difficulties in its administration via the tail vein. In our experiments, paramagnetic contrast agent was administrated intraperitoneally and produced high-quality MR images. Due to the large surface area of the peritoneum, the contrast agent is quickly absorbed into the circulation, and there is no contrast enhancement of the peritoneum itself after 15 min.
The signal-to-noise ratio (SNR) of the T2-weighted sequence was less than that of the T1-weighted sequence, and some ghosting and blurring as well as some motion artifacts remained, but this did not compromise the image quality of the T2-weighted sequence. We also found that motion artifacts occurred less in the supine than in the prone position of the animal. Artifacts resulting from peristaltic movements can be reduced either by injecting a relaxation agent or, more simply and effectively, by fasting the animals 6 hr prior to imaging. In clinical human imaging, motion artifacts are usually overcome by using fast turbo spin echo or gradient echo sequences in breath-hold technique. This cannot be accomplished in small animal imaging except with elaborate respiration techniques that usually require endotracheal intubation. Here we report on the imaging of mouse pancreas in vivo in a high-field MR scanner, using imaging sequences that utilized high resolution and a small field of view without the necessity of special equipment for anesthesia.
In conclusion, our results show that MRI permits the imaging of the pancreas in mice and the monitoring of tumors within the pancreatic tissue, using sequences implemented on clinical scanners and simple methods to reduce motion artifacts. The method described here can be utilized to image the pancreas, for example, in nonmalignant diseases such as diabetes.27 Tumors within the pancreas were readily identified and their development could be followed easily, demonstrating the advantage of monitoring orthotopically implanted tumor growth with MRI, and therefore treatment response in experimental models.
We thank Ralph Weissleder for his valuable discussion and suggestions. Jan Grimm is supported by a fellowship grant from the German Research Society (Deutsche Forschungsgemeinschft, DFG). The authors would like to acknowledge the Office of National Drug Control Policy (Counterdrug Technology Assessment Center, Director Dr. Albert E. Brandenstein) for their financial support of the Martinos Center at MGH.