| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Experimental Studies |
1 From the Departments of Radiology (J.B.K., I.N., O.C., R.A.K.) and Cancer Biology (P.T.), Beth Israel Deaconess Medical Center-West, One Deaconess Rd, Boston, MA 02215. From the 1998 RSNA scientific assembly. Received May 19, 1999; revision requested July 2; revision received August 12; accepted September 30. J.B.K. supported in part by the RSNA Research and Education Foundation as a General Electric Medical Systems/RSNA Scholar, a Biomedical Research Support Grant from Beth Israel Deaconess Medical Center, and a grant from the Beth Israel Deaconess Medical Center Radiology Foundation. Address correspondence to J.B.K. (e-mail: jkruskal@caregroup.harvard.edu).
 |
Abstract |
|---|
 TOP Abstract IntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: Seventy-six hepatic colorectal cancer metastases were studied in exteriorized livers of 18 mice by using intravital microscopy, US, and histologic examination of the same tumors.
RESULTS: Hypoechoic rims correlated with distended sinusoidal spaces in vivo. These spaces surrounded only locally invasive tumors (mean diameter, 0.85 mm) that had obstructed the supplying terminal portal venules. These spaces, containing adherent leukocytes and tumor cells, gave rise to new tumor vasculature. Results of histologic examination of rims (portal inflammation, congested or compressed sinusoids, cell atrophy) correlated with leukocyte endothelial adherence, occluded sinusoids, and new vessel formation in vivo.
CONCLUSION: Unlike results from previous studies, dynamic in vivo observations of peritumoral rims demonstrated distended sinusoidal spaces giving rise to new tumor-penetrating vessels. These sinusoids arose around locally invasive tumors and were associated with more advanced intrahepatic disease. These dynamic observations provide a pathophysiologic explanation for previous histologic correlates of peritumoral rims.
Index terms: Blood, flow dynamics, 95.92 • Colon, neoplasms, 75.32 • Liver neoplasms, blood supply, 95.92 • Liver neoplasms, secondary, 761.1267, 761.337 • Liver neoplasms, US, 761.12981, 761.12983 • Ultrasound (US), comparative studies
|
Introduction |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The purpose of our study, unlike that of previously reported studies, was to use in vivo microscopy to depict directly hepatic tumors at various phases of growth in live animals and to use this technique, modified for dynamic tumor imaging in our laboratory (10–14), to correlate directly with in vivo observations the peritumoral rims seen at corresponding US examinations performed in the same animals prior to microscopy. We thus avoided artifacts that may arise during microangiopathy studies or histologic specimen preparation.
|
MATERIALS AND METHODS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
Experimental Protocol
By using our established mouse model of hepatic colorectal cancer metastases (10–14), livers were imaged by using US, and in vivo videomicroscopy was then performed on the exteriorized tumor-bearing lobes of the same livers. During microscopy, particular attention was placed on imaging the same tumors that were identified on the US scans. Imaging results and in vivo microscopic findings were then compared with histologic findings in the same hepatic tumors.
Tumor Model
All experimental animals received intrasplenic inoculations of human CX-1 colorectal cancer cells according to protocols previously established in our laboratory (13,14). Cell lines were checked frequently for the presence of mycoplasma by means of staining (33258; Hoechst, Frankfurt am Main, Germany), and results were negative. Human CX-1 colorectal cancer cells (2 x 106 cells; 0.5 mL of phosphate-buffered saline solution, pH 7.4), with a moderately to well-differentiated human colorectal cancer line, were derived from the HT-29 cell line and provided by Dr. L. B. Chen (Dana-Faber Cancer Institute, Boston, Mass). These cells were maintained in a medium (RPMI 1640; Gibco Laboratories Life Technologies, Grand Island, NY) with 10% fetal bovine serum (Sigma Chemical Company, St Louis, Mo), 1% L-glutamine (Gibco), 100 units/mL penicillin G sodium (Gibco), and 100 mg/mL streptomycin sulfate (Gibco). Cells were injected intrasplenically into nude mice by means of a minilaparotomy incision by using a 27-gauge needle (Precisionglide; Becton Dickinson, Franklin Lakes, NJ). US imaging and in vivo microscopy of exteriorized mouse livers were performed 7–8 days later.
US of Hepatic Tumors
After induction of anesthesia, animals were placed in a supine position, a 2.0-cm midline incision was made, and the left lobes of the liver were gently exposed and mobilized by incising the falciform ligament off the inferior diaphragmatic surface. A hepatic US examination was performed (J.B.K.) by using a 5- or 10-MHz intraoperative transducer (Entos CL; American Technology Laboratories, Seattle, Wash); both gray-scale and color Doppler images were obtained of all visible tumors in the left lateral lobes, the same lobes to be used for intravital microscopy immediately following the US study. The hepatic surface was kept moist with warm (37°C) lactated Ringer solution, which was also used to create a water path. US images were recorded (Ektascan Imagelink system; Eastman Kodak, Rochester, NY) and were compared with in vivo microscopic data and histologic findings in corresponding hepatic lobes and tumors. Specific findings evaluated included maximum tumor diameter, the presence or absence of a hypoechoic rim at the tumor periphery, the maximum and minimum thicknesses of this rim, and the presence or absence of flow in the tumor and the rim. A grading system was not used.
Hepatic in Vivo Videomicroscopy
Technique.—Videomicroscopy was performed on exteriorized livers 7–8 days after intrasplenic inoculation of human colon cancer cells, according to our previously described techniques (10–14). After induction of anesthesia, the abdomen was opened by means of a 2.0-cm vertical midline incision extended along the left subcostal margin, and the liver was mobilized by incising the falciform ligament off the diaphragm. Mice were then placed in the left lateral position, and the exteriorized peripheral lobules of the livers were placed on an angled quartz crystal on a modified microscopy stage and transilluminated with a cooled monochromatic (500–800-nm) halogen light source. The segment of liver to be studied was covered with moist saran wrap (Dow Chemical, Midland, Mich) to limit movement induced by respiration, cardiac motion, or peristalsis. At no point were livers handled manually; all hepatic manipulations were performed with cotton-tipped applicator sticks. Livers were kept warm and moist by means of continuous flushing with warm (37°C) lactated Ringer solution. The color of the mesenteric vessels was monitored regularly to evaluate and confirm adequate respiratory and circulatory status.
A standard compound binocular microscope was modified for in vivo microscopy and equipped for transillumination and epiillumination. Images were obtained at a magnification of x100–1,500 on a modified microscope system (Optiphot; Nikon, Tokyo, Japan) and were recorded (CCD-72 video camera; DAGE/MTI, Michigan City, Ind) continuously onto videotape by using a recorder (S-VHS; Panasonic Matsushita, Osaka, Japan) for subsequent data and image analysis and for comparison with US images and histologic findings. From these videotapes, images were captured onto a computer (Power Macintosh 93; Apple Computer, Cupertino, Calif) by using a real-time video frame grabbing and processing graphics accelerator card (Pixelpipe; Perceptics, Knoxville, Tenn) for use with image analysis software (NIH IMAGE 1.62; National Institutes of Health, Bethesda, Md; available at: ftp://rsbweb.nih.gov/pub/nih-image/.). When necessary, 35-mm color images were obtained by using a camera (N8008; Nikon) mounted on the microscope headpiece.
Data collection.—Videomicroscopic data were recorded over a wide range of standard magnifications (x100–1,500) and included (a) tumor sizes, acinar locations, and tumor blood supply; (b) direct observations of the tumor peripheries in tumors with and those without hypoechoic rims; (c) cell populations and vessels (hepatocytes, leukocytes, and Kupffer and endothelial cells) surrounding tumors; (d) leukocyte endothelial interactions; (e) portal vein, hepatic vein, hepatic arterial, sinusoidal, and neovascular flow rates adjacent to and within tumors; and ( f ) the sites and vessels of origin of new tumor vasculature. Since numerous tumors are seen at microscopy that are too small to be imaged at US, particular attention was placed on collecting data specifically from tumors seen at US prior to these microscopic studies.
Histologic Studies
Resected hepatic segments were fixed in 10% formalin; five sections, prepared from the same tumor-bearing regions seen during videomicroscopy, were examined histologically following hematoxylin-eosin staining. Slides from two tumors also were stained with trichrome stain to distinguish distended sinusoidal channels from portal and central veins. In all cases, the pathologist (I.N.) reviewing the slides was blinded to the in vivo observations and the US findings.
Statistical Analyses
Data between groups were compared by using one-way analysis of variance. Results were considered significant at an overall difference of 5%.
|
RESULTS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
|
|
In Vivo Microscopic Appearance of Hepatic Tumors
Dynamic imaging of tumor periphery.—In vivo microscopic data were collected from all 76 colorectal cancer metastases that were imaged by means of US in 18 mice. The hypoechoic rims seen surrounding 55 tumor nodules at US corresponded to distended (43.6 µm ± .01; range, 35–55 µm) sinusoidal spaces tracking around the tumor periphery and connecting terminal portal venules to central veins (Figs 1, 2). These sinusoidal spaces contained increased numbers of adherent monocytes, which suggests endothelial activation. The intrinsic rolling flux, or mean numbers of temporarily and permanently adherent monocytes (per minute per high-power field), in the (a) terminal portal venules, (b) sinusoids that did not contain tumor, (c) pericentral venules, and (d) tumor-associated sinusoidal spaces were 18.4 ± 3.2, 22.6 ± 2.8, 40.5 ± 4.1 (P < .001 vs a and b), and 43.2 ± 3.7 (P < .001 vs a and b), respectively.
|
|
|
|
Tumor blood supply.—Tumors received a changing blood supply with increasing growth (Fig 2); the initial portal venous dominance led (Fig 2a, 2b) to an increasing neovascular supply as tumor growth restricted portal venous inflow (Fig 2b, 2c). Four general patterns of tumor growth were identified; these patterns correlated with tumor size and were distinguished by blood supply and zonal location. Single tumor cells or aggregates of less than five tumor cells were identified within terminal portal venules. With tumor growth, small nodules less than 0.5 mm in diameter surrounded the terminal portal venules. Tumor growth (nodules >0.5 mm in diameter) then continued along hepatic cords (Fig 2a) until venous occlusion occurred (Fig 2c), which resulted in the formation of distended sinusoidal spaces supplying the tumor periphery with a rich blood supply (Figs 1b, 2d). Where tumor growth started to produce narrowing of terminal portal venules, the mean tumor diameters were 0.72 mm. The distended sinusoidal spaces, which corresponded to rims on US scans, were thus associated with a more advanced stage of intrahepatic tumor growth. In each tumor nodule imaged, the terminal portal venules were obstructed partly or completely by tumor nodules growing into or around the vessels.
No tumors growing in (n = 8) or surrounding (n = 19) terminal portal venules had rims on the corresponding US scans, and distended spaces were not present in or near the tumor peripheries at microscopy. These tumor nodules were all less than 0.5 mm in diameter and were not associated with intratumoral neovascularity. These tumors clearly represented an earlier growth phase of hepatic metastases. This was in contradistinction to tumors greater than 0.5 mm in diameter, which grew as small centrilobular nodules (n = 38) surrounded by sinusoidal spaces, or locally invasive tumors (n = 11) growing along hepatic cords. Tumors in this latter group were always bordered by a distended sinusoidal space containing adherent leukocytes and also had hypoechoic rims.
Tumor neovascularity.—Videomicroscopy was used to characterize and to identify the precise origin of new tumor vessels. The small (mean diameter, 12 µm; range, 9–17 µm), irregular, nondistensible vessels arose directly from the distended sinusoidal spaces or from portal venules traversing tumors (Fig 2d). Tumor vessels were irregular in arrangement and size, and flow within the lumens was erratic in both direction and velocity. New vessels were identified in tumors located in all acinar zones and were identified only in tumors larger than 0.93 mm in diameter. New vessels were identified only in tumors that were surrounded by distended sinusoidal spaces in vivo and had hypoechoic rims at US. Even though all tumors containing microscopically visible new vessels had hypoechoic rims, no flow could be resolved in these vessels with our 10-MHz transducer.
Tumor growth patterns.—Initial tumor growth occurred in periportal and centrilobular zones and extended toward the pericentral regions with increased growth. At the time of microscopy, the proportions of tumors in the periportal, centrilobular, and pericentral zones were 36%, 54%, and 10%, respectively, of the 76 tumors. Each zone was compared for tumor size and the presence of rims: 34% of tumors located in the periportal zones contained rims, whereas 84% and 9% of tumors in the centrilobular and pericentral zones, respectively, contained rims. The mean tumor diameters in the periportal, centrilobular, and pericentral zones were 0.52, 0.91, and 0.43 mm, respectively.
Correlation of Hepatic Histologic Findings with in Vivo Microscopic Findings
Histologic and US findings were compared with in vivo microscopic observations from the same tumors. The Table summarizes comparative data from this study. Correlation was poor between the in vivo observations and the histologic findings (Table). In vivo evaluation of the peritumoral regions revealed no compressed hepatic parenchyma, fibrous pseudocapsules, necrotic debris, or edema, as was expected from results of previously published pathology studies (1–7). However, these previously recognized findings were observed histologically in the same hepatic tumors. The compressed parenchyma adjacent to tumors corresponded to collapsed, occluded, nonperfused sinusoids. No peritumoral edema was observed in vivo, specifically in the Disse spaces. No pseudocapsules were observed in vivo. Necrotic debris filling distended sinusoids corresponded to sinusoids containing adherent leukocytes and erythrocytes.
|
|
DISCUSSION |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
However, the histologic findings differed from in vivo observations in the same tumors. Our in vitro observations of hemorrhage, inflammation, and congested sinusoids correlated with trapped erythrocytes, adherent leukocytes, and occluded sinusoids in vivo. Portal inflammatory changes corresponded to leukocytes adhering to sinusoidal lining cells in vivo; congested sinusoids corresponded to adherent tumor cells, leukocytes, and erythrocytes. Unlike histologic observations of sinusoidal compression by tumor cells, in vivo observations demonstrated collapsed sinusoids adjacent to tumors; these resulted from nonperfusion of sinusoids owing to occlusion by adherent leukocytes or tumors. Dynamic videomicroscopy provided a pathophysiologic explanation for histologic findings and an in vivo explanation for an important imaging sign.
Outwater et al (15) identified peripheral hypointense rims on T2-weighted MR images in approximately 25% of patients with colorectal metastases. Histologic examination of rims in resected specimens revealed compressed hepatic parenchyma and sinusoids, hepatocyte atrophy, and fibrosis. Authors of other studies (5,16) have ascribed rims to peritumoral edema. The results of both the in vitro study by Outwater et al (15) and our in vivo study failed to show peritumoral edema, and we agree with Outwater et al (15) that concentric hepatic edema does not contribute substantially to ring patterns of low or high signal intensity.
Mergo et al (17) showed blood-pool-phase ring enhancement around human hepatic tumors after the administration of an intravascular paramagnetic contrast agent. In microangiopathy studies of barium-perfused human livers, Marchal et al (1) showed peritumoral compression of hepatocytes corresponding to US halos, and they identified peritumoral vascularization in two cases corresponding to halos. These authors identified pooling of contrast agent in large intercellular spaces packed with viable cells surrounding some tumors. The distended sinusoidal spaces present in our in vivo studies may correspond to those identified by Marchal et al (1). Our study results thus have provided in vivo confirmation for previous postulates concerning the nature of halos surrounding certain hepatic tumors.
At in vivo microscopic study of subcutaneous human adenocarcinoma xenografts in mice with severe combined immunodeficiency, Patan et al (18) identified widened vessels that looked like blood lakes on tumor surfaces and that were similar to those we identified. Sprouting and pillar formation occurred from lateral vessel walls into adjacent nonvascularized tissue and represented the earliest stages of angiogenesis (19). As we have shown in this study of hepatic metastases, distended sinusoids were a necessary precursor for angiogenesis and microvascular growth to occur and indicated the earliest conversion of a dormant tumor into a vascularized locally advanced tumor. New tumor vessels arose from distended sinusoidal spaces only after growth was sufficient to occlude small terminal portal venules.
These vessels arising from distended sinusoids may represent new tumor vasculature that is seen angiographically and is associated with more advanced tumor growth (20). For a tumor to grow beyond a limited volume, tumor cells must recruit and induce growth of new vessels (19). Thus, identification of neovascularity surrounding a tumor suggests a more advanced stage of growth. In MR and microangiopathy studies of implanted hepatomas in rats, Ni et al (2) showed that peritumoral rims were related exclusively to the presence of highly malignant tumors. The rims corresponded in part to zones of malignant infiltration that surrounded parenchymal compression and bile duct proliferation (2).
Fibrous pseudocapsules surround certain hepatic tumors on MR images (21,22). Histologic findings in our specimens showed fibrous capsules, but we did not identify capsules or compression of adjacent parenchyma in vivo. Lee et al (4) demonstrated that a halo of high signal intensity on T2-weighted images corresponded to sinusoidal dilatation and edema in the Disse spaces. We observed Disse spaces at histologic examination, but we did not observe these spaces or perisinusoidal edema in vivo. Using microvascular casts of hepatic metastases, Lin et al (23) showed compression of sinusoids adjacent to hepatic metastases. We demonstrated enhanced leukocyte-endothelial adherence narrowing sinusoidal lumens and reducing sinusoidal inflow and perfusion. Histologic findings in these regions revealed congested sinusoids at tumor edges, which corresponded to these adherent leukocytes.
The rim surrounding hepatic tumors on MR images has been evaluated as a sign for distinguishing benign from malignant tumors (4,15,16,23). In one study (22), a halo was seen in 29% of malignant primary tumors and 13% of hepatic metastases. Mergo et al (17) found no difference in the presence of rims surrounding hepatomas and hepatic metastases, but they identified rims surrounding hepatocellular adenomas. Wernecke et al (5) found that 88% of malignant hepatic tumors were surrounded by halos, as opposed to 14% of benign tumors. They also observed halos around focal nodular hyperplasia, hepatic abscesses, and hemangiomas. The pathogenesis of halos is probably different in these latter entities. Animal models for these tumors do not exist currently. Therefore, only colorectal cancer metastases were evaluated in this study, and the diagnostic value of the US halo could not be assessed. Thus, even in patients with known extrahepatic malignancies, the presence of rims around hepatic tumors cannot be used to infer malignancy. A rim surrounding a known malignant hepatic tumor may indicate more locally advanced disease.
With the establishment of the in vivo nature of peritumoral rims, it is important to identify possible prognostic and clinical implications of these rims. We identified endoluminal tumor cells coexisting with adherent leukocytes. Small tumor-leukocyte microemboli migrated into draining pericentral venules to enter the pulmonary circulation. Transhepatic spread of micrometastases was identified from tumor nodules greater than 0.5 mm in diameter. Clusters of tumor cells also were observed passing through the liver without adhering to sinusoidal lining cells. These clusters bypassed hepatic sinusoids via peripheral channels, which are normally present and connect terminal portal venules with central veins. In patients with colorectal cancer, pulmonary metastases are thought to arise from hepatic metastases. The observation of tumor cells in the splenic vein effectively bypassing the liver to enter the pulmonary circulation suggests that pulmonary metastases may occur in patients with gastrointestinal malignancies without hepatic metastases.
Despite the lack of direct tumor penetration beyond the sinusoidal spaces, viable tumor cells were identified in these spaces, which suggests that local image-guided ablative therapies should be extended at least to include the entire rim. This would also target the site where new tumor-penetrating vessels arise, vessels that are necessary to maintain and enhance tumor growth. Also, it is possible that targeted diagnostic and therapeutic agents could be developed for better localization to the tumor site. For instance, hepatic endothelial cell and macrophage activation occurs in patients with malignancies (24). We demonstrated enhanced leukocyte endothelial adherence, which suggests endothelial cell activation; according to results from our previous study (25) in isolated livers, endothelial cell activation occurs owing to endothelial rather than leukocyte receptor up-regulation.
One cannot make direct comparisons between results from animal studies and observations in human clinical studies. It is possible that the mouse liver responds in a different manner to the presence of metastases. The use of animal models of hepatic metastases that mimic the sequence of events in human metastasis development is essential for evaluating hepatic tumors (26). Our model, which relies on spontaneous passage of tumor cells from the spleen to the liver (12), is therefore well suited for answering the specific questions posed by this study.
In summary, using an in vivo technique to depict peritumoral regions shown to represent rims or halos on images, we identified an alternate pathophysiologic explanation for previous histologic findings ascribed to rims or halos surrounding these same tumors.
Practical application: Since a US halo surrounding colorectal hepatic metastases indicates a more locally invasive angiogenic tumor, the rim should be incorporated in the ablation field in interstitial therapies. Also, since new vessels arise from these spaces, antiangiogenic therapies should be investigated as adjuncts to traditional treatment regimens.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. B. Kruskal, A. Azouz, H. Korideck, M. El-Hallak, S. C. Robson, P. Thomas, and S. N. Goldberg Hepatic Colorectal Cancer Metastases: Imaging Initial Steps of Formation in Mice Radiology, June 1, 2007; 243(3): 703 - 711. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Liu and O. Matsui Changes of Intratumoral Microvessels and Blood Perfusion during Establishment of Hepatic Metastases in Mice Radiology, May 1, 2007; 243(2): 386 - 395. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. B. Kruskal, P. Thomas, R. A. Kane, and S. N. Goldberg Hepatic Perfusion Changes in Mice Livers with Developing Colorectal Cancer Metastases Radiology, May 1, 2004; 231(2): 482 - 490. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
