CC BY-NC-ND
0NEOPLASIA
www.neoplasia.com
Volume 10 Number 7 July 2008 pp. 663-673 6 63
Abstract
Noninvasive methods are strongly needed to detect and quantify not only tumor growth in murine tumor models but also the development of vascularization and necrosis within tumors. This study investigates the use of a new imaging technique, flat-panel detector volume computed tomography (fpVCT), to monitor in vivo tumor progression and structural changes within tumors of two murine carcinoma models. After tumor cell inoculation, single fpVCT scans of the entire mice were performed at different time points. The acquired isotropic, high-resolution volume data sets enable an accurate real-time assessment and precise measurements of tumor volumes. Spreading of contrast agent-containing blood vessels around and within the tumors was clearly visible over time. Furthermore, fpVCT permits the identification of differences in the uptake of contrast media within tumors, thus delineating necrosis, tumor tissues, and blood vessels. Classification of tumor tissues based on the decomposition of the underlying mixture distribution of tissue-related Hounsfield units allowed the quantitative acquisition of necrotic tissues at each time point. Morphologic alterations of the tumor depicted by fpVCT were confirmed by histopathologic examination. Concluding, our data show that fpVCT may be highly suitable for the noninvasive evaluation of tumor responses to anticancer therapies during the course of the disease.
Neoplasia (2008) 10, 663-673
Morphologic Changes of Mammary Carcinomas in Mice over Time as Monitored by Flat-Panel Detector Volume Computed Tomography1
„ + + +
Jeannine Missbach-Guentner ' ' , Christian Dullin , Sarah Kimmina§, Marta Zientkowska*, Melanie Domeyer-Missbach*, Cordula Malz\ Eckhardt Grabbed Walter Stühmer* and Frauke Alves*
^Department of Hematology and Oncology, University Medicine, Robert-Koch-Str. 40, 37075 Göttingen, Germany; "Department of Diagnostic Radiology, University Medicine, Robert-Koch-Str. 40, 37075 Göttingen, Germany; *Max-Planck-Institute of Experimental Medicine, Hermann Rein Str. 3, 37075 Göttingen, Germany; §Department of Laboratory Animal Science, University Medicine, Robert-Koch-Str. 40, 37075 Göttingen, Germany; ^Department of Anatomy and Embryology, University Medicine, Robert-Koch-Str. 40, 37075 Göttingen, Germany
Introduction
To assess the increasing amounts of novel, targeted therapies to treat solid tumors in clinically relevant murine models, it is desirable to monitor tumor progression in detail during the course of the disease. At present, longitudinal studies of mouse cancer models require large cohorts because autopsy at different tumor stages has been the only reliable method to evaluate tumor progression and anticancer treatment efficacy. The assessment of tumor growth in small animal models has frequently relied on the rather imprecise calculations of tumor volumes by caliper measurements of maximum tumor diameter, tumor minor axis, and tumor length either during or toward the end of an experiment [1]. However, the assessment of therapy efficacy requires the accurate measurement of changes in tumor volume during the course
of the disease. Therefore, techniques have been developed, which enable imaging and exact monitoring of tumor growth and progression
Abbreviations: SCID, severe combined immunodeficient; fpVCT, flat-panel volume computed tomography; 3D, three-dimensional; H&E, hematoxylin and eosin; HU, Hounsfield units; 2D, two-dimensional
Address all correspondence to: Frauke Alves, MD, PhD, Department of Hematology and Oncology, University Medicine, Robert-Koch-Str. 40, 37075 Göttingen, Germany. E-mail: falves@gwdg.de
1This work was supported by a grant from the Deutsche Forschungsgemeinschaft (AL336/5-1) within the SPP1190 and by a tandem grant from the Max-Planck Society. Received 6 February 2008; Revised 10 April 2008; Accepted 11 April 2008
Copyright © 2008 Neoplasia Press, Inc. All rights reserved 1522-8002/08/$25.00 DOI 10.1593/neo.08270
in vivo, such as magnetic resonance imaging (MRI) [2], ultrasound [3-5], computed tomography (CT) [6,7], and nuclear imaging [8,9]. Each modality possesses a unique combination of strengths and weaknesses that affects their selection for use in a particular study.
Here, we introduce a novel approach, flat-panel detector volume computed tomography (fpVCT), which allows three-dimensional (3D) visualization of anatomic structures with isotropic imaging resolution [10,11]. Compared with other imaging modalities, fpVCT enables shorter scanning times of approximately 8 seconds. In addition, the mouse skeleton can be visualized in detail and with clear contours [12-14]. Flat-panel detector volume computed tomography therefore shows excellent sensitivity in detecting skeletal lesions. Using this method, we recently determined the accurate localization, size determination, and assessment of the progression of osteolytic bone lesions within the mouse skeleton in a metastatic breast tumor model [15]. Information about tumor volume changes is of great value in the assessment of the efficacy of tumor therapies. However, the development of blood vessels and necrotic tissues within these tumors are likewise important parameters in such studies. We evaluated, therefore, the use of the fpVCT method to monitor tumor growth and the morphologic changes in both orthotopic and subcutaneous severe combined immunodeficient (SCID) mouse models of MDA-MB-231 and R30C mammary carcinoma cells, respectively. Here, we demonstrate that fpVCT exhibits excellent sensitivity and accuracy in visualizing tumors and allows a precise real-time assessment oftumor growth by the detection of structural alterations within the tumors over time.
Materials and Methods
Cell Lines
The estrogen-independent human breast cancer cell line MDA-MB-231 was obtained from the American Type Culture Collection (Rockville, MD) and was maintained at 37°C in a humidified atmosphere containing 5% CO2 in DMEM supplemented with 10% fetal calf serum and 1% L-glutamine (all from PAN Systems, Aidenbach, Germany). The human mammary tumor cell line R30C was kindly provided by Dr. Ralf Bargou, Max-Delbrueck-Centre For Molecular Medicine, Berlin, Germany [6]. Cells were maintained in RPMI 1640 with GlutaMAX-I (Invitrogen, Karlsruhe, Germany), with 25 mM HEPES supplemented with 10% heat-inactivated fetal calf serum, 0.5 mM sodium pyruvate, 60 U/ml penicillin, and 60 pg/ml streptomycin. Cells were regularly certified free of mycoplasma contamination. Tumor cells were harvested near confluence by brief trypsinization in 0.25% trypsin-EDTA solution (Gibco, Carlsbad, CA), washed several times, and placed in sterile PBS shortly before implantation.
Animals
All animals were handled according to German regulations for animal experimentations, and all animal protocols were approved by the administration of Lower Saxony, Germany. For this study, SCID mice, strain C.B-17/Ztm-scid, were used and the latter were maintained in a sterile environment in special cages with filter huts and in a Scantainer (Scanbur, Koge, Denmark). Cages, bedding, and water were autoclaved, and the food was gamma-irradiated. Immunodeficiency ofSCID mice was verified by measuring serum immunoglobulin levels using an enzyme-linked immunosorbent assay.
Tumor Cell Implantation
For orthotopic tumor cell implantation, female SCID mice were anesthetized by peritoneal injection of 75 mg/kg ketamine hydro-
chloride with 15 mg/kg xylazine. A total of 1 x 106 MDA-MB-231 cells suspended in 25 pl of sterile PBS were implanted with an insulin syringe, 29-gauge x 0.5 in. (Becton Dickinson, Heidelberg, Germany) very slowly into the mammary fat pad of the fourth mammary complex. The cells were implanted to visibly infiltrate the breast tissue. The needle was slowly withdrawn after a 1-minute delay. The mammary gland was then returned, and the incision was closed using an interrupted Vicryl suture for the skin (5/0; Ethicon, Norderstedt, Germany). A total of 2 x were subcu-
taneously injected into both flanks of the SCID mice. All manipulations were conducted under aseptic conditions using a laminar flow hood. All animals tolerated the procedure well. After implantation, mice were inspected daily for body weight loss, general condition, and tumor formation.
Flat-Panel Detector Volume Computed Tomography Imaging
Mice were imaged with a fpVCT, a nonclinical volume CT prototype (GE Global Research, Niskayuna, NY). They were anesthetized with vaporized isoflurane at 0.8% to 1% concentration throughout the imaging session, centered on the fpVCT gantry axis of rotation, and placed perpendicularly to the z-axis of the system, so that it was possible to scan the whole mouse with one rotation. An iodine-containing contrast agent, Isovist 300 (at 150 pi per mouse; BayerSchering, Berlin, Germany), was applied intravenously approximately 30 seconds before scan.
Flat-panel detector volume computed tomography consists of a modified circular CT gantry and two amorphous silicon flat-panel X-ray detectors each of 20.5 x 20.5 cm2 with a matrix of 1024 x 1024, 200-pm detector elements. The fpVCT works in a step-and-shoot acquisition mode. Standard z-coverage of one step is 4.21 cm. All data sets were acquired with the same protocol: 1000 views per rotation, 8 seconds of rotation time, 360 used detector rows, and 80 kVp and 100 mA. A modified Feldkamp algorithm was used for image reconstruction resulting in isotropic high-resolution volume data sets (512 x 512 matrix, with an isotropic voxel size of approximately 100 pm).
Tumor Segmentation and Generation of a Tumor Density Histogram
To visualize morphologic changes within the tumor, tumors were segmented by a region-growing algorithm. Tumor capsules and peripheral blood vessels having a higher density than the tumor tissue were borders of segmentation. In the next step, a histogram was generated, displaying the relative occurrence of voxels over their Houns-field values within the tumor. Thus, the histogram represents the percentage of tumor tissue comprising the same density versus the measured density in Hounsfield units (HU). For tumor segmentation and computing, histogram data sets were analyzed with voxtools 3.0.64 Advantage Workstation 4.2 (GE Healthcare, Buckinghamshire, UK). For the evaluation of fpVCT-based volumetry, commercially available spherical phantoms (Spherotech, Fulda, Germany) of different materials and with known diameters were used.
Autopsies and Histologic Analysis
At the end of the experiment, autopsies of SCID mice were performed, tumors were excised and weighed, and the abdomen and thoracic cavity were examined systematically for the presence of metastases. At autopsy and after excision, visible tumors were externally measured with a caliper. Tumors were collected and placed in
phosphate-buffered 4% formalin for 16 hours at room temperature and were embedded in paraffin. Tissue sections (2.5 ^m) were obtained, stained with hematoxylin and eosin (H&E), and inspected by routine microscopic examinations.
Statistical Analysis
Five MDA-MB-231 tumors were analyzed for tumor growth rates and for morphologic alterations, respectively. Statistical analyses were performed using Microsoft Office Excel 2003 version 11.0 (Microsoft Corporation, Redmond, WA). All data were expressed as mean ± SD.
Results
Assessment of Mammary Carcinoma Growth by fpVCT
To investigate the use of fpVCT as a noninvasive imaging tool to monitor tumor growth and progression, whole bodies of breast tumor-bearing mice were scanned using fpVCT in combination with an iodine-containing contrast agent at certain time points. Orthotopic transplantation of MDA-MB-231 human breast tumor cells resulted in tumor development in the mammary gland of all five mice. As shown in Figure 1A, fpVCT enabled the visualization of the enlargement of the mammary carcinomas in 3D over time. By both the precise delineation of the tumor in vivo and the virtual isolation of the tumor from the fpVCT data sets, tumor volumes were calculated automatically by considering all three dimensions (Figure 1B). These calculations allowed the generation of precise growth kinetics for each tumor over time (Figure 1C). Volumes of the tumors ranged
between 63 and 486 mm3 at the end of the experiment and showed a change in weekly mean growth rates, 128 ± 80% and 68 ± 33% between days 21 and 28 and between 35 and 42 days, respectively (Figure 1C).
Using traditional caliper measurements, the length (L), width (W), and height (H) of each tumor were analyzed. Here, the presuming ellipsoid volume that provides the most precise measurement of tumor mass [16] was calculated from the following formula: 0.5 • L • W • H. Because most of the tumors developed either an irregular, nonellipsoid shape, or multiple lobes, tumor volumes calculated from the fpVCT data sets were generally 10.0 ± 5.85% larger compared with standard caliper measurements of volumes of postmortem-dissected tumors (Figure 1D). Small tumors showed the least discrepancy with a 1.8% deviation from tumor volumes determined using fpVCT. To assess the reliability of the automated process for the measurement of tumor volumes after tumor segmentation, two spherical phantoms with known diameters of 3.00 and 3.18 mm were scanned using fpVCT. Their volumes were determined in a manner similar to that performed for the determination of tumor volume using the region-growing tool; this method correlated very precisely with the mathematically determined volume. Relative measurement errors were between 0.99% and 1.01%.
Monitoring Tumor Blood Vessel Development by fpVCT
Furthermore, follow-up examination of mice was performed using fpVCT to assess fpVCT applicability to monitor the recruitment and formation of tumor vessels over time. As presented in Figure 2, contrast
Figure 1. Growth rate of an orthotopically implanted tumor depicted in vivo by 3D reconstructions of fpVCT data sets. (A) Representations of volume rendering of skin in the tumor area after virtual removal of fur. Visualization of the enlargement of a representative mammary carcinoma by repeated fpVCT scans in combination with contrast agent 21, 28, 35, and 42 days after MDA-MB-231 cell implantation. (B) Corresponding semiautomatically segmented tumors. The indicated volumes were automatically determined. (C) Growth rates of five MDA-MB-231 tumors for 3 weeks. (D) Comparison of automatically determined tumor volumes using fpVCT data sets (rhombuses) with caliper measurements postmortem (squares). Although tumor volumes were underestimated by caliper measurements, the data show the proportionality between both methods.
Figure 2. Tumor vessel development over time visualized in vivo using 3D fpVCT data sets. (A) Depiction of contrast agent-containing tumor vessels and their distribution and bifurcations both in the periphery and within a developing orthotopic mammary tumor by repeated fpVCT scans of a representative SCID mouse. Tumor vessels possessed small diameters on day 21 and their density increased by day 28. Tumor vessels appeared most enlarged after 35 days, followed by a decrease in the number of vessels with high density 1 week later. The tumor is framed. (B) In an MIP presentation of the final scan, blood vessels within the tumor were not visible, suggesting the loss of central blood vessels due to necrotic tissue after this period (left). One kidney is indicated (K). (C) Tumor vessel formation of a representative subcutaneous mammary carcinoma in combination with contrast agent (right). Tumor vessels were hardly visible 20 days after R30C tumor cell implantation and appeared after 25 days. The density of vessels around the tumor increased after 28 days and even more after 31 days. (B) An MIP presentation of vessel densities within the tumor indicates that a diffuse lattice of vessels infiltrated the tumor tissue (right). (D) Visualization of a single blood vessel supplying the orthotopic tumor (framed) and (E) the venous port of the subcutaneous tumor. The tumor is framed. (D) In the dorsal aspect of a tumor-bearing mouse after orthotopic implantation of MDA-MB-231 cells, a blood vessel (black arrow) coming from the tumor (white arrow) can be traced to the kidneys (K). In a higher magnification and after subtraction of the spine, a connection (arrowhead) between the tumor vessel (arrow) and the upper lumbar artery is visible. Aorta (A) and vena cava (V) are depicted as parallel running structures. (E) A large efferent vessel (white arrow) of the subcutaneous tumor (black arrow) discharges in the brachiocephalic vein (arrows) as shown in detail (right).
agent-containing blood vessels with sizes greater than 150 pm in diameter were monitored over time in arbitrary planes of the tumor periphery by applying different visualization protocols of the 3D fpVCT data sets. Only when the content of the contrast agent within the vas-culature is high the smaller blood vessels can be detected, although size cannot be assessed. Furthermore, tumor-infiltrating vessels could be clearly visualized, and their development was followed using fpVCT (Figure 2, A-C). The rate of vascularization in the orthotopic tumors increased between 3 and 5 weeks after tumor implantation (Figure 2A). After 3 to 4 weeks, the vessels possessed small diameters in the tumor periphery and in the tumor center. One week later, the vessels showed the greatest enlargement in diameter with highest density. However, after week 6, vessel density decreased, particularly in the tumor center, although increasing vessel diameter could be observed (Figure 2B, left).
In comparison, fpVCT scans revealed that subcutaneous R30C breast tumors recruited blood vessels later in the disease process and developed a regular distribution of peripheral blood vessels covering nearly the whole surface of the tumor (Figure 2C). Only after 25 days did blood vessels become visible; a dense network increased
continuously thereafter. In the 2D maximum intensity projection (MIP), the density of the tumor is indicative of a diffuse lattice of vessels, which crosses the tumor (Figure 2B, right).
Furthermore, the origin and distribution of blood vessels supplying both the orthotopic and subcutaneous tumors were defined using fpVCT (Figure 2, D and E). In both tumor models, the afferent blood vessels were primarily supplied by a connection to the upper lumbar artery (Figure 2D), whereas the efferent blood vessels were observed to discharge into the brachiocephalic vein (Figure 2E). Thus, the dynamic process of blood vessel constitution during tumor growth and the reorganization and loss of central blood vessels could be qualitatively imaged using fpVCT.
Qualitative and Quantitative Analysis of Tumor Density Ratios In Vivo
To quantify and visualize morphologic structures such as necrotic tissue, intact tumor tissues, and blood vessels within tumors, several steps of a semiautomatic procedure were performed. First, the tumor was segmented out of the entire 3D fpVCT data set as shown in Figure 3,
A-C. Next, a histogram of the relative frequencies of different densities in HU within the observed volume was created (Figure 3, D-F). Inflection points, defined as points on the curve at which the curvature changes its sign, were automatically determined in those cases in which the histogram did not fit a Gaussian distribution pattern (Figure 3E). Because the density values between two inflection points represent a particular portion of the entire volume, they were illustrated by various colors. Most of the examined tumors showed typical tripartite histograms that consist of a portion of low density, often in the negative HU range depicted in blue, a major portion of middle density, depicted in red, and a smaller portion of high density (up to 500 HU) depicted in white (Figure 3E). Hence, the major difference between these portions was the distinguishable X-ray absorption pattern due to the specific receptivity toward the contrast agent in specific tissues. Under the assumption that these different portions are the sum of at least three overlapping normal distributions, the histogram was predicted to be a mixed Gaussian distribution (Figure 3F).
To characterize the different underlying normal distributions within the tumor and their percentages per total volume, we implemented
a specific expectation maximization (EM) algorithm as described by Dullin [17] in which the portion percentage, the mean, and the SD of the now-separated single distributions within the entire volume of the segmented tumor are estimated. Next, we assumed that the different distributions within the histogram were a consequence of the different behaviors of the tumor tissue due to its biologic receptivity toward the contrast agent. A tumor with necrotic portions consists ideally of three different types of tissues: 1) normal tumor tissue that takes up small amounts of the applied contrast agent because of sufficient vascularization, 2) necrotic tumor tissue with a more waterlike density that assimilates no contrast agent, and 3) contrast media-containing blood vessels. Therefore, we correlated single distributions of the histogram with the approximate amount of ne-crotic tissue, depicted in blue as a low-density region and intact tumor tissues within the segmented tumor, depicted in red (Figure 3, E-1). The most precise determination of the amount of single portions was observed in the minimal overlapping distributions with well-separated peaks. Initially, this pattern was generated from the distinct absorption of the contrast agent. Otherwise, a clear separation would be disturbed by the partial volume effect, leading to
Figure 3. Quantitative and qualitative analyses of the density ratios within an orthotopic tumor in vivo using fpVCT. (A) A 3D volume-rendering image of the tumor in situ (arrow) 42 days after orthotopic implantation of MDA-MB-231 cells. Superficial layers were subtracted. (B) A 2D fpVCT image of the same mouse in a coronal view; note that the tumor is framed. (C) Semiautomatic segmentation of the tumor after subtraction of all additional data sets of the mouse. (D) Histogram of the percentage of tumor tissue within the tumor volume versus its density in HU measured by fpVCT. (E) Inflection points of the curve were determined automatically. Density values between two inflection points (divided by black marks) belong to a specific density portion, marked with different colors. The blue distribution indicates a fraction of low density, red indicates a portion of an intermediate-density range, and white fraction indicates the portion of highest density. (F) These density portions within the tumor volume were assumed to be three overlapping Gaussian distributions that possessed their specific peaks in different density regions of the histogram. (G-I) Volume-rendering images of the tumor in three serial median slices. Distribution of the colors corresponds to the colors of the histogram and suggests to represent the central necrotic area in blue as a low-density region, the surrounding tumor tissue in red, and the contrast agent-containing blood vessels in white with the highest density on the surface of the tumor. Scale bar in (I), 5 mm.
merging at the boundaries of both low- and the high-density distributions within the central portion.
Due to the high spatial resolution of the 3D fpVCT data sets, virtual serial sections of the tumor with a distance of 200 p m were generated. Corresponding to the colors used for the histogram (Figure 3, E and F ), the sections were dyed in blue and red to show the exact localization of both the low- and high-density areas within the tumor having central necrosis and surrounding intact tumor tissue, respectively (Figure 3, G -I ).
Depiction of Tumor Density Ratios by fpVCT Compared with Histologic Examination
To validate the finding that different tissues within the tumor can be classified using fpVCT in vivo, we compared the density profiles of tumors generated by the 3D depiction of the fpVCT data sets with both macroscopic and histologic examinations. As shown in Figure 4, A-D, after 6 weeks, fpVCT data sets revealed that low-density areas corresponding to necrosis were primarily located in the center of orthotopic tumors and were surrounded by tissue of higher density (Figure 4C). Macroscopic examination of the same tumor shortly dissected after fpVCT scan confirmed the presence of necrotic areas within the center of the tumor (Figure 4D). Therefore, fpVCT images reliably depict density ratios of tumors with high spatial resolution and thereby allow classification of tumor tissues in vivo.
By comparing histologic cross sections of the excised tumors with corresponding serial sections of the fpVCT images (Figure 4E ), we could show that the morphologic structures closely correlated to the corresponding in vivo data acquired using fpVCT (Figure 4E, detailed view). The central area of necrosis as shown by histologic examination is clearly visible using fpVCT (Figure 4E, left) and is thereby predictable in vivo. Even discrete features, such as a small blood vessel appearing within the necrotic area and the surrounding intact tumor tissue, were depicted authentically (Figure 4E, detailed view).
Determination ofLongitudinal, Morphologic Changes of Both Necrotic and Nonnecrotic Tumors Using fpVCT In Vivo
By an exact delineation of density and density ratios within the tumors and the quantification of the main components, such as tumor tissue, necrotic tissue, and blood vessels, the amount of necrosis was quantified based on the histograms for each scan time (Figure 5).
A representative, orthotopically implanted tumor was chosen to illustrate the morphologic alterations that are characteristic of this tumor model (Figure 5, A and B). Twenty-one days after the orthotopic implantation of MDA-MB-231 cells, the major tumor volume displayed a low density. These findings can be explained by the missing receptivity toward the contrast agent as a consequence of a depauperate tumor blood vessel system. After 28 days, the histogram showed the characteristic shift to a positive HU range (Figure 5A) that correlates with the occurrence of more compact structures within the tumor (Figure 5B). In the histograms for days 21 and 28 (Figure 5A), only single components of the entire tumor volumes were depicted. After these first time points, the curves shift gradually to the negative HU range, suggestive of the development of necrosis. During this observation period, the curves of the histograms divide explicitly into at least two portions, suggesting the separation of a larger necrotic area and a smaller area of intact tumor tissue (Figure 5A). The corresponding 3D visualizations of the tumor were in agreement with an increase in a central area of less density (Figure 5B). After the imple-
mentation of the self-developed EM algorithm, the portions of necrosis after both 35 and 42 days constituted 49% and 51% of the entire tumor volume, respectively. The increase in necrosis between weeks 5 and 6 was accompanied by a decrease in intact tumor tissue from 27% to 20%. Furthermore, an increase of the high-density tumor portion was observed in accordance with an increase of vessel diameter, as depicted in Figure 2A.
As shown in Figure 6A, an overlay of the histograms from all scans of the orthotopic tumor depicts the sum of morphologic changes at different time points. The development of necrosis, which constitutes the main portion of the tumor after 35 days, is indicated by the incremental shifts of the peaks of the histogram to the negative HU range. In addition, the different peaks and breadths of the curves for each time point represent the amount of voxels within the histograms and, hence, show tumor growth by an increase in volume. To determine the extent of necrosis in the orthotopic mammary carcinoma model, five mice were analyzed on the day of dissection by the equal application of contrast agent 28 seconds before the scan. All assessed tumors showed distinct central necrosis with a mean portion of 49 ± 6%. Note that the smallest tumor showed the smallest portion of necrosis (data not shown).
In contrast, the subcutaneous tumor showed no central necrosis during the observation period as clearly depicted in Figure 5, C and D. Only the broadening of the curve in the histogram after 31 days indicates a slow separation of density portions, suggesting the development of necrosis (Figure 5C). Although only a single component was recognized after the application of the EM algorithm, small, scattered necrotic foci could be visualized by 3D depiction (Figure 5D).
In accordance with these results, the overlay of the histograms from all scans of the nonnecrotic tumor displayed no development of a primary necrotic portion. Figure 6B, the nonnecrotic tumor showed a shift of peaks toward the positive HU sector during the observation period.
Comparison of these fpVCT images of the final scans with the corresponding histologic sections confirmed the presence of a large central area of necrosis in the orthotopic tumor (Figure 6, C-E) and the predominate existence of intact tumor tissue in the subcutaneous tumor (Figure 6, F -H ). Furthermore, these results underscore the reliability and the potential of this technique to both detect and quantify morphologic alterations within a defined tumor.
Discussion
In this study, we demonstrated that the use of volume CT with flat-panel detectors enabled the noninvasive determination of morphologic changes within two estrogen-independent mammary carcinoma tumor models. Invasive and highly dedifferentiated human MDA-MB-231 mammary carcinoma cells were implanted into the orthotopic environment of the mouse mammary fat pad, whereas human mammary carcinoma R30C cells were injected subcutaneously into mice, a rather nonphysiological environment [6]. Weekly scanning using fpVCT revealed fast-growing MDA-MB-231 mammary carcinomas in all SCID mice and precise determination of their growth rates. Generally, tumor growth in murine tumor models has been determined by intravital measurement of the tumor using a cal-iper [18,19]. This technique is rather limited not only because it demands the expertise of the researcher but also because it is only applicable for subcutaneously or superficially growing tumors. Only approximate tumor volumes can be determined using this method
Figure 4. Comparison of images generated by fpVCT with both macroscopic and histopathologic examinations. (A) A 3D fpVCT image of an orthotopic mammary tumor (arrow) 6 weeks after tumor cell implantation in situ. (B) Macroscopic appearance of the tumor (arrow) in situ at the time of dissection. (C) An fpVCT-generated representative 3D central section depicts the halves of the tumor with their density profile: the central necrotic area (blue, black arrow) and intact tumor tissue, including densely packed mammary gland tissue (red, white arrow). (D) Macroscopic appearance of a central slice of the dissected tumor confirmed the existence of necrotic tissue (black arrow) between the intact tumor and mammary gland tissues (white arrow). Note that the analog distribution of these features in the fpVCT image is indicated with arrows in (C). (E) An fpVCT volume-rendering presentation of a different mouse bearing an orthotopic mammary tumor (blue). (E, detailed view, left) Serial sections of the tumor virtually cut off the fpVCT data sets that were generated with contrast agent on the day of dissection. Visualization of densities within tumor sections demonstrates a distinct central necrotic region (blue), surrounding tumor tissue (red), and a cross section of a blood vessel (top left, arrow). (E, detailed view, top right) Median paraffin section of the same tumor stained with H&E closely corresponds to the predicted distribution of the morphologic features necrosis (N), intact tumor tissue (T), and a vessel (V, arrow) depicted by the fpVCT data sets as shown in (E, top left). At a 2.5-fold higher magnification (E, bottom right), the display window according to the frame in (E, bottom left) shows the occurrence of intact tumor tissue (T) between an area of central necrosis (N). Scale bars: bottom and top left, 4 mm; top right, 3 mm; bottom right, 800 fjm.
because only two dimensions are measured and the theoretical thickness of the skin must be subtracted [20]. Image reconstruction from fpVCT measurements, however, results in isotropic high-resolution 3D volume data sets, enabling the determination of tumor volumes in vivo with high precision and, thus, allowing the objective acquisition of tumor growth kinetics in longitudinal studies. To evaluate preclinical treatment responses to anticancer therapies, further information regarding tumor microvasculature and the occurrence of necrosis are of great interest. Here, we demonstrate that short, single fpVCT scans at certain time intervals allow the precise observation of the development of structural alterations within tumors of both cancer models. Our results were confirmed by histologic examination and revealed that orthotopic mammary carcinomas developed central necrotic areas over time, whereas only small, scattered necrotic areas were observed in the subcutaneous tumor.
Similar to fpVCT, 3D ultrasound microimaging generates 3D data sets and has been shown to determine exact tumor volumes in living mice over time. A previous study showed that by correlating ultrasound image texture with cell density, both the foci of necrotic areas and blood vessel distribution in mouse prostate cancer could be depicted [5]. This technique has a high spatial resolution comparable to fpVCT; however, it is primarily used for the analysis of soft tissues that do not require the
use of contrast agent. Compared with fpVCT, ultrasound uses a smaller penetration depth and has a smaller field of view. Furthermore, it requires the depilation of fur before image acquisition.
Micro CT (pCT), with a very high spatial resolution of 5 to 10 pm [21], generates 3D data sets comparable to fpVCT and ultrasound; however, it is accompanied by the need for high doses of radiation and long scanning times [22,23]. A precise calculation of tumor volumes in combination with a good soft tissue contrast agent is feasible using MRI [24-26]. Furthermore, Overhauser-enhanced MRI or [19F]-perfluorocarbon MRI allow the direct measurement of oxygen concentrations within the tumor, which reflects the hypoxic status of the tissue [27,28]. Moreover, due to the vascular state within the tumor, a depiction of necrotic areas is also possible [29].
In contrast to the brief imaging examinations by fpVCT, data acquisition by MRI, pCT, or ultrasound demands much longer scanning times and, therefore, requires the maintenance of body temperature and administration of anesthesia through inlet tubes for the mice. Positron emission tomography (PET) imaging assesses functional tumor characteristics, such as glucose metabolism, as indirect markers for the vitality of tumor tissues and provides information about blood flow and blood volume [30]. However, only in combination with CT can certain biologic processes be determined as PET can be correlated to
0.140 cm3
.Necrosis 4D % J5d
Necrosis ÍÍU&. 42|J
0.440 em3
Figure 5. Comparison of growth kinetics and morphologic changes in both necrotic and nonnecrotic tumors analyzed by fpVCT imaging in vivo. (A) Graphic representations of density distributions within a tumor with developing necrosis over time. The state of vascularization of this tumor is depicted in Figure 2A. At 21 days after implantation, a wide Gaussian curve in a negative HU range indicates a lack of blood vessels in the mammary fat pad. After 4 weeks, a small bell-shaped curve with a homogeneous distribution of only one density portion in a positive HU range is detectable. During days 35 and 42, the Gaussian distribution divides into two components, one of low density (developing necrosis) and one of higher density (tumor tissue and contrast agent-containing blood vessels). (B) Corresponding serial sections of the 3D visualization of the tumor were generated using fpVCT. Similar to the corresponding histograms (A), necrotic tissue developed regularly (blue) and displaced intact tumor tissue (red). Surrounding blood vessels (white) detected by contrast agent only occur in the tumor periphery (B, bottom). (C) The histograms for the nonnecrotic tumor showed no movement into the negative HU range between days 20, 25, 28, and 31, although the curve spreads at the latest time point because of an increase of different density portions. (D) Corresponding serial sections of the 3D visualization of the tumor are shown. As depicted in the corresponding histograms, no central area of necrosis was observed in this tumor over time.
Figure 6. Morphologic structures of both tumor types depicted by fpVCT compared with histopathologic analysis. (A and B) Overlays of the histograms over time allowed the comparison of the shift of the curves during tumor development. (A) Necrotic tumor: 21 days after orthotopic implantation of MDA-MB-231 cells, the tumor volume primarily consisted of low density, which is explained by its lack of receptivity toward the contrast agent. After 28 days, the histogram depicts the characteristic shift to the positive HU sector. At this time point, the tumor consisted primarily of intact tumor tissue and had a threefold greater volume than the previous week. After days 35 and 42, the curve shifts progressively into the negative HU sector, suggestive of the development of necrosis. (B) Nonnecrotic tumor: 20 days after subcutaneous implantation of R30C cells, the tumor volume primarily consisted of an area of relatively low density similar to the necrotic tumor in (A) at the first time point. After 25 days, the tumor volume increased nearly twofold. Until day 28, a clear shift of the graph to the positive HU sector remained throughout the peak of the curve; after 31 days, it shifts broadly in a negative direction. (C-H) fpVCT-generated images of the necrotic tumor (C and E) and the homogeneous nonnecrotic tumor in situ (F and H). H&E-stained paraffin sections of the tumors depict the distinct central area of necrosis of the orthotopically implanted tumor (D) and the generally intact tumor tissue of the subcutaneous tumor (G). Note that the latter tumor grew in multiple lobes (F, arrow) and that only one lobe was analyzed in detail (G and H). The predicted distributions of necrosis as illustrated in blue and the intact tissue as depicted in red within the tumors are determined using fpVCT (E and H). Scale bars: D, 3 mm; E, 6 mm; G, 800 jm; H, 1600 jm.
anatomic structures. The use of PET is also rather expensive because it depends on short-lived isotopes and requires a cyclotron and nearby radiochemistry facilities [31]. Imaging by fpVCT reliably mirrors the structural alterations within the different mammary carcinomas, including the spreading and loss of contrast agent-containing blood vessels during the course of tumor progression.
After 5 weeks of monitoring orthotopically implanted tumors using fpVCT, the vessel density within the center of the tumor appears to be reduced, clearly the first sign of the development of central necrosis. Simultaneously, the broadening of vessel diameters around the tumor suggested the enhanced permeability of the vessels, resulting in a rapid discharge of iodine-containing contrast agent into the interstitial space as shown previously [11]. The enhanced permeability of vessels together with abnormal endothelial stratification and altered basement membranes, arteriovenous shunts, and blind ends are characteristic features of tumor-induced vascularization [32]. In contrast, the subcutaneous tumors continuously recruit increasing peripheral and inner tumor vessels during the observation period. Furthermore, they develop a close lattice of vessels through the tu-
mor, allowing sufficient blood supply to prevent the development of central necrosis.
It has previously been shown that the application of iodine-containing contrast agents during p,CT and fpVCT scans allow the observation of specific blood-flow distribution and blood supply [33]. Functional characteristics of tumor vessels, such as permeability that results from defective endothelial barrier function, in combination with high interstitial pressures [34] or perfusion measurements of blood flow to predict blood volume within a single vessel, can also be obtained using contrast-enhanced p.CT [35,36].
One of the most challenging problems using fpVCT is the spatial resolution, which, in our study, was limited to 150 to 200 p,m. Therefore, sizes and number of microvessels cannot be depicted by fpVCT imaging. Due to the partial volume effect caused by the limited resolution of 150 p,m, changes in microvessel density will only result in an alteration in the normal tumor tissue compartment of the histogram. Therefore, we propose that the quantitatively observed increase in the high-density portion within the tumor volume is caused by a dilatation of preexisting vessels. Here, we repeatedly administered an
iodine-containing contrast agent at a concentration that was well tolerated in all mice. Smaller vessel diameters of 40 to 50 p,m have been visualized in tumors by Kiessling et al. [11] by applying higher doses of iodine- or barium-containing contrast agents. However, this method does not allow follow-up studies because of the high, and therefore toxic, concentrations of these contrast agents. To optimize the spatial resolution to visualize blood vessels, there is a strong need for blood-pool contrast agents, such as the macromolecular media that remain intravascular [37,38].
Such a development will bridge the gap between the high spatial resolution of either light or scanning electron microscopy for the analyses of microvessel architecture ex vivo [39-41] and the moderate resolution of emerging imaging techniques for the dynamic, 3D, in vivo morphogenic studies over time.
Our data show that fpVCT enables the quantitative determination of various density portions that correlate to distinct tissue types within a defined tumor volume. At present, data such as these have only been obtained by both a time- and cost-intensive approximation due to the analyses of serial histologic sections postmortem, which does not allow the determination of the kinetics of necrotic progression. Because the depiction of necrosis is dependent on the altered distribution of contrast agent, only strict compliance to standardized imaging protocols will allow the obtainment of comparable and reproducible data and, thereby, the exact delineation and quantification of different tumor components. First, both the amount and time point of the application of contrast agent must be equalized for each scan. Second, the tumor must possess sufficient blood vessels, even in the periphery, to avoid the dispersion of contrast agent within the area of necrosis during the scan. A delay in scanning after the application of contrast agent, as well as an increased permeation of contrast agent, will result in the overestimation of viable tumor tissue. So far, the influence of high intratumoral pressure on the uptake of contrast agent into the tumor is unknown and will be the focus of further investigations. In this study, however, histologic analysis confirmed that the reduced receptivity for contrast medium in the center of the tumor is due to central necrosis.
During irradiation or antiangiogenic cancer treatment, effects on the permeability of vessel walls, which might lead to an over- or underestimation of single density portions such as necrosis, have to be considered. To prevent misinterpretation of fpVCT data sets regarding morphologic changes within the tumor, comparative analyses of fpVCT images taken over time to early fpVCT examinations, to controls, and to histologic results are mandatory.
In summary, our study illustrates the ability of fpVCT under standard conditions to allow the precise and noninvasive quantification of tumor growth and morphologic changes during the course of the disease. This imaging technique will thus limit the statistical variability of preclinical studies, reduce the number of mice required for each study, and increase the amount of data obtained from each animal. The novel approach using fpVCT presented here is therefore both a useful and cost-effective tool for the preclinical evaluation of anticancer therapies. Furthermore, within the fast-growing field of oncological experiments with genetically engineered mice, fpVCT imaging will provide unique, valuable information on the role of single genes in tumor growth and progression and in altered tumor vessel recruitment.
Acknowledgments
The authors acknowledge the excellent technical assistance of Roswitha Streich, Johanna Widera, and Sarah Greco, the technical support of
Kristin Hammer, Tomasz Karykowski, and Karin Stapp-Kurz for running the fpVCT, and the critical comments on the manuscript from Irmi Sures and Carlos E. Guentner.
References
[1] Somville J, De Beuckeleer L, De Schepper A, Verstreken J, and Taminiau A (2001). Reliability of measuring volume by different methods for tumors of the musculoskeletal system. Acta Orthop Belg 67, 338—343.
[2] Schmidt KF, Ziu M, Schmidt NO, Vaghasia P, Cargioli TG, Doshi S, Albert MS, Black PM, Carroll RS, and Sun Y (2004). Volume reconstruction techniques improve the correlation between histological and in vivo tumor volume measurements in mouse models of human gliomas. J Neurooncol 68, 207—215.
[3] McCarville MB, Streck CJ, Dickson PV, Li CS, Nathwani AC, and Davidoff AM (2006). Angiogenesis inhibitors in a murine neuroblastoma model: quantitative assessment of intratumoral blood flow with contrast-enhanced grey-scale US. Radiology 240, 73-81.
[4] Wu G, Wang L, Yu L, Wang H, andXuan JW (2005). The use of three-dimensional ultrasound micro-imaging to monitor prostate tumor development in a transgenic prostate cancer mouse model. Tohoku J Exp Med 207, 181-189.
[5] Wirtzfeld LA, Wu G, Bygrave M, Yamasaki Y, Sakai H, Moussa M, Izawa JI, Downey DB, Greenberg NM, Fenster A, et al. (2005). A new three-dimensional ultrasound microimaging technology for preclinical studies using a transgenic prostate cancer mouse model. Cancer Res 65, 5231-5237.
[6] Krneta J, Kroll J, Alves F, Prahst C, Sananbenesi F, Dullin C, Kimmina S, Phillips DJ, and Augustin HG (2006). Dissociation of angiogenesis and tumorigenesis in follistatin- and activin-expressing tumors. Cancer Res 66, 5686-5695.
[7] Cavanough D, Johnson E, Price RE, Kurie J, Travis EL, and Cody DD (2004). In vivo respiratory-gated micro-CT imaging in small animal oncology models. Mol Imaging 3, 55-62.
[8] Yang Y, Tai YC, Siegel S, Newport DF, Bai B, Li Q, Leahy RM, and Cherry SR (2004). Optimization and performance evaluation of the microPET II scanner for in vivo small-animal imaging. Phys Med Biol 49, 2527-2545.
[9] Del Guerra A and Belcari N (2002). Advances in animal PET scanners. QJNucl Med 46, 35-47.
[10] Greschus S, Kiessling F, Lichy MP, Moll J, Mueller MM, Savai R, Rose F, Ruppert C, Günther A, Luecke M, et al. (2005). Potential applications of flat-panel volumetric CT in morphological and functional small animal imaging. Neoplasia 7, 730-740.
[11] Kiessling F, Greschus S, Lichy MP, Bock M, Fink C, Vosseler S, Moll J, Mueller MM, Fusenig NE, Traupe H, et al. (2005). Volumetric computed tomography (VCT): a new technology for non-invasive, high-resolution monitoring of tumor angiogenesis. Nat Med 10, 1133-1138.
[12] Dullin C, Missbach-Guentner J, Vogel WF, Grabbe E, and Alves F (2007). Semi-automatic classification of skeletal morphology in genetically altered mice using flat-panel volume computed tomography. PloS Genet 3, e118.
[13] Obenauer S, Dullin C, Alves F, Missbach-Guentner J, Grabbe E, and Heuser M (2007). Flat-panel-detector-based volumetric CT: performance evaluation of imaging for skeletal structures of small animals in comparison to multislice CT. Clin Imaging 31, 18-22.
[14] Obert M, Ahlemeyer B, Baumgart-Vogt E, and Traupe H (2005). Flat-panel volumetric computed tomography: a new method for visualizing fine bone detail in living mice. J Comput Assist Tomogr 29, 560-565.
[15] Missbach-Guentner J, Dullin C, Zientkowska M, Domeyer-Missbach M, Kimmina S, Obenauer S, Kauer F, Stühmer W, Grabbe E, Vogel WF, et al. (2007). Flat-panel detector based fpVCT: a novel 3D imaging technique to monitor osteolytic bone lesions in a mouse tumor metastasis model. Neoplasia 9, 755-765.
[16] Tomayko MM and Reynolds CP (1989). Determination of subcutaneous tumor size in athymic nude mice. Cancer Chemother Pharmacol 24, 148-154.
[17] Dullin C (2001). Entwicklung einer Methode zur Segmentierung verschiedener Strukturen im Datenmaterial bildgebender Messverfahren in der Medizin [diploma thesis]. Jena, Germany: Friedrich-Schiller Universität.
[18] Rygaard K and Spang-Thomsen M (1997). Quantitation and Gompertzian analysis of tumor growth. Breast Cancer Res Treat 46, 303-312.
[19] Reynolds CP, Sun BC, DeClerck YA, and Moats RA (2005). In Methods in Molecular Medicine, Vol. 111: Chemosensitivity: Vol. 2:In vivo Models, Imaging, and Molecular Regulators. RD Blumenthal (Ed.). Totowa, NJ: Humana Press, pp. 335-350.
[20] Spang-Thomsen M, Nielsen A, and Visfeldt J (1980). Growth curves of three human malignant tumors transplanted to nude mice. Exp Cell Biol 48, 138-154.
[21] Engelke K, Karolczak M, Lutz A, Seibert U, Schaller S, and Kalender W (1999). Micro-CT. Technology and application for assessing bone structure [in German]. Radiologe 39, 203—212.
[22] Ford NL, Thornton MM, and Holdsworth DW (2003). Fundamental image quality limits for microcomputed tomography in small animals. Med Phys 30, 2869—2877.
[23] Taschereau R, Chow PL, and Chatziionnaou AF (2006). Monte Carlo simulations of dose from microCT imaging procedures in a realistic mouse phantom. Med Phys 33, 216-224.
[24] Cline HE, Dumoulin CL, Hart HR, Lorensen WE, and Ludke S (1987). 3D reconstruction of the brain from magnetic resonance images using a connectivity algorithm. Magn Reson Imaging 5, 345-352.
[25] Vaidyanathan M, Clarke LP, Hall LO, Heidtman C, Velthuizen R, Gosche K, Phuphanich S, Wagner H, Greenberg H, and Silbiger ML (1997). Monitoring brain tumor response to therapy using MRI segmentation. Magn Reson Imaging 15, 323-334.
[26] Bouvet M, Spernyak J, Katz MH, Mazurchuk RV, Takimoto S, Bernacki R, Rustum YM, Moossa AR, and Hoffman RM (2005). High correlation of whole body red fluorescent protein imaging and magnetic resonance imaging on an orthotopic model of pancreatic cancer. Cancer Res 65, 9829-9833.
[27] Krishna MC, Subramanian S, Kuppusamy P, and Mitchell JB (2001). Magnetic resonance imaging for in vivo assessment of tissue oxygen concentration. Semin Radiat Oncol 11, 58-69.
[28] Williams BB, al Hallaq H, Chandramouli GV, Barth ED, Rivers JN, Lewis M, Galtsev VE, Karczmar GS, and Halpern HJ (2002). Imaging spin probe distribution in the tumor of a living mouse with 250 MHz EPR correlation with BOLD MRI. Magn Reson Med 47, 634-638.
[29] Bhujwalla ZM, Artemov D, Natarajan K, Ackerstaff E, and Solaiyappan M (2001). Vascular differences detected by MRI for metastatic versus non-metastatic breast and prostate cancer xenografts. Neoplasia 3, 143-153.
[30] Blankenberg FG, Eckelman WC, Strauss HW, Welch MJ, Alavi A, Anderson C, Bacharach S, Blasberg RG, Graham MM, and Weber W (2000). Role of radionuclide imaging in trials of antiangiogenic therapy. Acad Radiol 7, 851-867.
[31] Padhani AR and Neeman M (2001). Challenges for imaging angiogenesis. Br J Radiol 74, 886-890.
[32] Vaupel P, Kallinowski F, and Okunieff P (1989). Blood flow, oxygen and nutrient supply and metabolic microenvironment of human tumors: a review. Cancer Res 49, 6449-6465.
[33] Savai R, Wolf JC, Greschus S, Eul BG, Schermuly RT, Hänze J, Voswinckel R, Langheinrich AC, Grimminger F, Traupe H, et al. (2005). Analysis of tumor vessel supply in Lewis lung carcinoma in mice by fluorescent microsphere distribution and imaging with micro- and flat-panel computed tomography. Am J Pathol 167, 937-946.
[34] Jain RK (1990). Vascular and interstitial barriers to delivery of therapeutic agents in tumors. Cancer Metastasis Rev 9, 253-266.
[35] Purdie TG, Lee TY, Iizuka M, and Sherar MD (2000). Dynamic contrast enhanced CT measurement of blood flow during interstitial laser photocoagulation: comparison with an Arrhenius damage model. Phys Med Biol 45, 1115-1126.
[36] Eastwood JD, Lev MH, Azhari T, Lee TY, Barboriak DP, Delong DM, Fitzek C, Herzau M, Wintermark M, Meuli R, et al. (2002). CT perfusion scanning with deconvolution analysis: pilot study in patients with acute middle cerebral artery stroke. Radiology 222, 227-236.
[37] Wiart M, Corot C, Berthezene Y, Violas X, and Canet E (2001). CT pulmonary angiography with a macromolecular contrast medium: a comparative study versus iobitridol in rabbits. Invest Radiol 36, 547-553.
[38] Idee JM, Port M, Robert P, Raynal I, Prigent P, Dencausse A, Le Greneur S, Tichkowsky I, Le Lem G, Bourrinet P, et al. (2001). Preclinical profile of the monodisperse iodinated macromolecular blood pool agent P743. Invest Radiol 36, 41^9.
[39] McDonald DM and Choyke PL (2003). Imaging of angiogenesis: from microscope to clinic. Nat Med 9, 713-725.
[40] Fox SB (1997). Tumour angiogenesis and prognosis. Histopathology 30, 294-301.
[41] Augustin HG (2001). Tubes, branches and pillars. The many ways of forming new vasculature. Circ Res 89, 645-647.
