Formation of Perfusion Defects in Lymph Nodes During the Early Stage of Metastasis

A perfusion defect in metastatic lymph nodes (LNs) can be visualized as a localized area of low contrast on contrast-enhanced CT, MRI or ultrasound images. Hypotheses for perfusion defect include abnormal hemodynamics in neovascular vessels and decrease in blood ow in pre-existing blood vessels in the parenchyma due to compression of tumor growth in the LNs. However, the mechanism of perfusion defects in LNs during the early stage of LN metastasis has not yet been investigated. Here we show that the formation of a tumor mass with very few microvessels was associated with the development of a perfusion defect in the non-enlarged LN at the early stage of LN metastasis. We found in a mouse model of LN metastasis induced using non-keratinizing tumor cells that during formation of the perfusion defect in non-enlarged LN, the number of blood vessels £ 50 mm in diameter decreased, while the volume of existing blood vessels >50 mm in diameter increased using contrast-enhanced high-frequency ultrasound and contrast-enhanced micro-CT imaging systems with a maximum spatial resolution of > 30 mm. Furthermore, we found that tumor angiogenesis and pO 2 changes in the metastatic LN were not observed. Our results demonstrate that the perfusion defect appear to be a specic form of tumorigenesis in the LN as vascular-rich organ at the early stage of metastasis. We anticipate a perfusion defect on ultrasound, CT or MRI images to be used as an indicator in the non-enlarged LN at the early stage of LN metastasis. concept of LN-mediated hematogenous metastasis implies that tumor cells from vital via Our group developed a drug delivery system (LDDS) aims treat sentinel and their downstream anticancer directly into the sentinel LNs This therapeutic system can treat metastatic using smaller systemic imaging


Introduction
Sentinel lymph nodes (LNs) receive lymphatic drainage from the primary tumor and are the organs at highest risk of metastasis 1 . The presence or absence of LN metastasis is highly relevant to cancer staging, patient prognosis, treatment planning (including surgery, radiotherapy and adjuvant chemotherapy), and the risks of local tumor recurrence and cancer-speci c mortality 2 . Recent experiments in mice have indicated that tumor cells in afferent lymphatic vessels can invade the subcapsular sinus and in ltrate the veins running over the LN, thereby metastasizing to distant organs via hematogenous routes at the early stage of metastasis 3-5 . Therefore, early diagnosis and treatment of metastatic lymph node at the early stage of metastasis 6,7 are essential to prevent distant metastasis 8 .
Although the diagnosis of LN metastasis has been improved by the use of multiple biomarkers 9,10 , there are no perfect prognostic markers that can reliably predict cancer recurrence 11 . Diagnostic modalities such as CT, MRI and ultrasound cannot detect the pathological changes associated with LN metastasis, and their diagnostic accuracy is generally lower for smaller LNs 1 .
A perfusion defect in LNs can be visualized as a localized area of low contrast on contrast-enhanced CT, MRI or ultrasound images. Perfusion defects have been identi ed in many types of tumor cells 12,13 . The reported incidence of perfusion defects correlates with LN size, being 10-33% for metastatic LNs < 10 mm in diameter and 56-63% for metastatic LNs > 15 mm in diameter 14 . Measurement of the difference between the maximum and minimum signal intensity values on contrast-enhanced ultrasound images can be used to evaluate perfusion defects and to detect early LN metastasis in clinically node-negative breast cancer [15][16][17] . Central necrosis is an important feature used to diagnose small metastatic LNs, especially those that do not match the diagnostic criterion for size. The term 'central necrosis' is a synonym for a perfusion defect and is often used to refer to actual necrosis, the replacement of part of the LN parenchyma by tumor cells, or both. Since the mechanism of LN metastasis involves the proliferation of tumor cells that invade the subcapsular sinus from afferent lymphatic vessels, 'perfusion defect' is a more appropriate term than 'central necrosis.' Hypotheses for perfusion defect are an increase in the intranodal pressure due to tumor growth which compresses the pre-existing blood vessels in the parenchyma, as well as abnormal hemodynamics in neovascular vessels and hypoxia generated by angiogenesis during rapid tumor growth. Perfusion defect is observed even in metastatic LNs of welldifferentiated squamous epithelial cancer cells due to keratinization 18 . However, although tumor growth in the non-enlarged LN increases the internal pressure 19 , it does not induce neovascularization 20,21 because the LN already has a rich vascular network 3,13 ,22-24 . In addition, lymphangiography revealed decrease in the lymphatic sinus region due to the tumor formation at the early stage of metastasis, i.e., the parenchyma was replaced with tumor cells 25 . Therefore, a perfusion defect in a metastatic LN at an early stage is presumed to be an event that is distinct from those associated with hypoxia, angiogenesis, keratinization and tissue necrosis, which occur as tumor growth rapidly progresses 26,27 . The present study aims to elucidate the mechanisms of perfusion defect formation in the tumor mass in the nonenlarged LN at the early stage of metastasis.

Results
Metastasis in the PALN was detected from day 7 after KM-Luc/GFP cell inoculation into the SiLN and from day 14 after FM3A-Luc cell inoculation (Fig. 1a, b). Luciferase activity in both the SiLN and PALN increased with time ( Fig. 1a, b). Luciferase activity was higher in the SiLN (inoculation site) than in the PALN (metastatic site) after inoculation FM3A-Luc-inoculated solid tumors exhibited continuous growth ( Fig. 1S) that was similar to that reported for FM3A-Luc-inoculated solid tumors 19,20 . First, we compared the pO 2 between metastatic PALN, solid tumor, control SiLN, control PALN, tumor-inoculated SiLN and tail vein. The interior of the solid tumor became hypoxic and exhibited a signi cant decrease in pO 2 as it grew. By contrast, the pO 2 in the LNs remained constant irrespective of whether they contained tumor ( Fig. 1c; one-way ANOVA and Tukey's post-hoc test: P < 0.05, solid tumor vs metastatic PALN; P < 0.01, solid tumor vs tumor-inoculated SiLN; and P < 0.01, solid tumor vs tail vein).
Contrast-enhanced high-frequency ultrasound imaging was used to examine the volumetric and vascular changes in the PALN at different times after KM-Luc/GFP cell and FM3A-Luc cell inoculation (Fig. 2). Heat maps of the vessels in the PALN (Fig. 2a) did not show a perfusion defect at day 0 but did reveal a perfusion defect that lacked contrast agent (arrow) at day 10 after KM-Luc/GFP cell inoculation. Similarly, a perfusion defect (arrow) was observed on day 21 after FM3A-Luc cell inoculation, and the size of the defect expanded with time (days 28 and 35; Fig. 2a). Figure 2b shows data for the normalized vascular area ratio and normalized LN volume. There were no signi cant changes in LN volume up to day 10 following KM-Luc/GFP cell inoculation and up to day 35 following FM3A-Luc cell inoculation. Therefore, these LNs can be considered to be non-enlarged LNs based on the ultrasonography ndings.
There were also no signi cant changes in the vascular area ratio. However, this parameter showed a slight tendency to increase after KM-Luc/GFP cell inoculation and decrease following FM3A-Luc cell inoculation.
Next, contrast-enhanced micro-CT imaging was used to observe the vascular structure of the PALN (Fig.  3). Mice were perfused sequentially with PBS, 4% paraformaldehyde and gelatin-based barium contrast agent on days 4, 7 and 10 after KM-Luc/GFP cell inoculation and on days 14, 21, 28 and 35 after FM3A-Luc cell inoculation, and the PALNs were excised 2 h after cooling and then imaged. In the control group (day 0, no tumor cell inoculation), the LN was shaped like a at oval (Fig. 3a, b), and one side of its surface was covered with dense blood vessels, including the thoracoepigastric vein (TEV). In contrast, the reverse side had far fewer blood vessels. The arrows in Fig. 3 indicate the TEV anastomosed to many intranodal blood vessels (Video S1). The blood in the TEV ows from the PALN to the subclavian vein 4,5 . Perfusion defects (arrowheads) were observed following KM-Luc/GFP cell inoculation (Video S2) and FM3A-Luc cell inoculation (Video S3).
The volume of the PALN and the volume of its internal vessels were calculated from 3D contrastenhanced micro-CT images (Fig. 4). There were no signi cant changes in PALN volume over time following the inoculation of KM-Luc/GFP cells or FM3A-Luc cells (i.e., the PALN can be considered a nonenlarged LN), in agreement with the data shown in Fig. 2. Vessel volume increased after KM-Luc/GFP cell inoculation (one-way ANOVA: P < 0.05, day 0 vs day 7; P < 0.01, day 0 vs day 10) and FM3A-Luc cell inoculation (one-way ANOVA: P < 0.05, day 0 vs day 21; P < 0.01, day 0 vs day 28) (Fig. 4a). Total blood vessel length in the PALN, calculated from the CT images, showed no changes in both tumor cell groups (Fig. 4b).
Figure 4c, d shows the distribution of blood vessel diameter in the PALN after tumor cell inoculation into the SiLN. Blood vessels with a diameter < 50 µm accounted for 80% of all blood vessels before the inoculation of both tumor cell types into the SiLN (day 0; Fig. 4c, d). However, following the inoculation of KM-Luc/GFP cells (Fig. 4c), the number of blood vessels with a diameter < 50 µm decreased (one-way ANOVA: P < 0.000, day 0 vs day 4; P < 0.000, day 0 vs day 7; P < 0.000, day 0 vs day 10; P < 0.01, day 4 vs day 7; P < 0.01 day 4 vs day 10; P < 0.05, day 7 vs day 10) and the number of blood vessels with a diameter > 50 µm increased as tumor growth progressed. The above ndings are consistent with the dilation of existing blood vessels. After the inoculation of FM3A-Luc cells (Fig. 4d), the number of blood vessels with a diameter < 50 µm decreased (one-way ANOVA: P < 0.000, day 0 vs day 10; P < 0.01, day 4 vs day 7; P < 0.001, day 4 vs day 10; P < 0.01, day 7 vs day 10), and those with a diameter > 50 µm showed increases over time as the tumor grew. Figure 5 shows PALN tissue sections obtained after CT imaging. KM-Luc/GFP cells formed a tumor mass in the LN and proliferated expansively, but the boundary with surrounding normal tissue was well de ned.
There were no mature blood vessels containing blood in the tumor mass. However, blood vessels around the tumor mass appeared compressed, and some dilated blood vessels were also seen (Fig. 5a). FM3A-Luc cells proliferated invasively in the PALN, resulting in a poorly de ned boundary with surrounding normal tissue. No mature blood vessels were evident in the tumor tissue. Some tumor cells had in ltrated the surrounding tissue's blood vessels to destroy the vasculature at the invasion site, and no dilated blood vessels were found in the tissue surrounding the tumor (Fig. 5a). Figure 5b shows the CD31-positive vascular density calculated from histological images (one-way ANOVA, KM-Luc/GFP group: P < 0.05, day 0 vs day 10; one-way ANOVA, FM3A-Luc group: P < 0.01, day 0 vs day 28) and the vascular density calculated from CT images (one-way ANOVA, KM-Luc/GFP group: P < 0.01, day 0 vs day 4; P < 0.01, day 0 vs day 7; P < 0.01, day 0 vs day 10; one-way ANOVA, FM3A-Luc group: P < 0.01, day 0 vs day 28). Vascular density increased with time, indicating that the perfusion defect was formed in a microenvironment in which no tumor angiogenesis was induced and the existing blood vessels were dilated.

Discussion
In the present study, we used contrast-enhanced high-frequency ultrasound and contrast-enhanced ex vivo micro-CT imaging to examine the processes by which perfusion defects are formed in non-enlarged LNs after inoculation with two different types of non-keratinizing tumor cells. The blood vessel network is well developed in normal LNs and provides an adequate supply of oxygen and nutrients (day 0 in Fig. 2a,   Fig. 3a) 3,13,23,24 . Formation of the perfusion defect was associated with the development of a tumor mass in a microenvironment that exhibited loss of vessels £ 50 mm in diameter and dilation of existing blood vessels > 50 mm in diameter, where the spatial resolution was > 30 mm. During the formation of the perfusion defect, LN volume did not change (Fig. 2b, Fig. 4a), and tumor angiogenesis was not induced (Fig. 4, Fig. 5). We consider the dilated blood vessels to be primarily congested venous vessels that become dilated because the tumor causes a circulatory disturbance within the LN. The pO 2 in the LN at the onset of perfusion defect development was the same as that in normal LNs (Fig. 1c).
Micro-CT-detectable perfusion defects are di cult to visualize with clinical CT and ultrasound modalities, which have a lower spatial and tissue resolution. Therefore, non-enlarged LNs would be diagnosed as clinical N0 (cN0) LNs even though they contained metastasized tumor cells. Additionally, non-enlarged LNs would not be considered to be in the early stage of metastasis from a histopathological standpoint because tumor cells have already grown out of the subcapsular sinus at this stage. Perfusion defects have been observed in metastatic LNs using clinical imaging modalities such as CT and ultrasonography and discussed the mechanisms of their formation with regard to cell type 12 , reactive LNs 28 , the fatty hilum 29 , tissue necrosis 26,27 , abnormal lymphatic drainage, keratinization 30 and biomarkers 31,32 . Importantly, perfusion defects in LNs were observed with each imaging modality, and the LNs were subjected to pathological analyses to determine whether they were cN0 or N1. The present nding was that in the early stage of metastasis, tumor cells replace microvascular areas to form a perfusion defect that is not pathologically necrotic tissue. During perfusion defect formation, replacement of parenchyma with tumor supplied by only a few microvessels occurs, intranodal pressure increased 19 , pO 2 remained constant and tumor neovascularization was not induced 20 . Perfusion defects, which are formed by the replacement of LN parenchyma with tumor cells, have also been observed with intranodal lymphangiography in non-enlarged LNs 25 . The characteristics of perfusion defect formation in LNs may explain why metastatic LNs show an inadequate response to hematologic systemic chemotherapy using small-molecule and macromolecular anticancer agents 33,34 .
The concept of LN-mediated hematogenous metastasis implies that tumor cells spread from metastatic LNs to distant vital organs via blood vessels. Our group has recently developed a lymphatic drug delivery system (LDDS) that aims to treat sentinel LNs and their downstream LNs by injecting anticancer drugs directly into the sentinel LNs 8,35,36 . This therapeutic system can treat metastatic LNs using smaller doses of anticancer drug than is required for systemic chemotherapy. Enhancing the sensitivity of clinical imaging modalities to detect perfusion defects in LNs will facilitate the identi cation of metastatic LNs that could be targeted with the LDDS, thereby substantially improving the quality of life of patients with cancer.

Implications for patient care
Expanding the roles of contrast-enhanced CT, ultrasound or MRI to the imaging of perfusion defects in LNs could improve the monitoring of early-stage LN metastasis and facilitate cancer management. New drug delivery systems could be developed to target the tumor foci that form perfusion defects, considering that there are many discrepancies in the rate of tumor growth, the microenvironment and the blood composition in human and mouse tumor tissues.

Cell lines
Keratinizing tumor cells have been detected in LNs 18 . Therefore, we used two types of non-keratinizing tumor cells to eliminate the effect of keratinization in the present study: malignant brous histiocytomalike (KM-Luc/GFP) cells, which express a fusion of the luciferase and enhanced green uorescent protein genes, and C3H/He mouse mammary carcinoma (FM3A-Luc) cells, which express the luciferase gene 38 .
Cell lines were incubated (37°C, 5% CO 2 /95% air) until 80% con uence was achieved. Cell growth rate The mouse was anesthetized by administering 2% iso urane in oxygen and rested on a heated stage throughout the imaging session 6,7 . The images were analyzed by ultrasound contrast agent detection (UCAD) software 39 to detect temporal changes in the density of microvessels in the PALN.

Measurement of oxygen partial pressure (pO 2 )
The pO 2 values of solid tumor (n = 2), control (tumor-free) SiLN (n = 14), control (tumor-free) PALN (n = 14), tumor-inoculated SiLN (n = 7), metastatic PALN (n = 7) and tumor-free tail vein (n = 7) were measured. The pO 2 values in the tumor-inoculated SiLN and metastatic PALN were measured when the luciferase activity of the PALN exceeded 1.0 × 10 6 photons/sec. The mice were anesthetized with 2% iso urane in oxygen and placed on a heated stage. The skin was incised to expose the target tissue, and a digital pO 2 monitor (Unique Medical Co., Ltd., Tokyo, Japan) was used to measure the pO 2 . A 24G indwelling needle (Terumo, Tokyo, Japan) was inserted into the center of the target tissue and then withdrawn, and a pO 2 monitor electrode (UOE-04T, Unique Medical Co., Ltd., Tokyo, Japan) was subsequently inserted. Values were recorded 3 min after electrode insertion using a data acquisition system (ML826 PowerLab 2/26, BioResearch Center, Aichi, Japan) and LabChart software (BioResearch Center).

Contrast-enhanced ex vivo micro-CT imaging
Ex vivo contrast-enhanced imaging was performed using a micro-CT scanner speci cally developed to image small laboratory animals (scanXmate/E090, Comscantecno Co., Kanagawa, Japan) 25 . A gelatinbased barium contrast agent (nontoxic due to its insolubility; 1.0 ± 0.3 mm in size) was prepared as previously described 40 . Brie y, the left ventricle of the heart was perfused rst with PBS to remove the blood and then with 4% paraformaldehyde (FUJIFILM Wako Pure Chemical Co., Osaka, Japan) to x the tissues. Pre-heated gelatin-based barium contrast agent was injected to better visualize blood vessel branches including small capillaries. Mice were refrigerated (4°C) for at least 2 h after perfusion to solidify the gelatin-based contrast agent and prevent wash-out during xation. PALNs were harvested and scanned at resolutions of 8 -20 μm for plain angiography and a slice thickness of 100 μm for CT. Acquired slice data were rendered as 3D images (Amira, Maxnet Co., Ltd, Tokyo, Japan). Samples were obtained on day 4 (n = 5), day 7 (n = 6) and day 10 (n = 5) for the KM-Luc/GFP group and on day 0 (n = 6), day 14 (n = 5), day 21 (n = 6) and day 28 (n = 5) for the FM3A-Luc group. The PALNs resected on each of these days were then used for histological analyses.

Histological analysis
Some LNs were xed overnight in 10% formalin at 4°C, dehydrated, embedded in para n, serially sectioned (3-5 μm), and either stained with hematoxylin and eosin (HE) or immunostained for CD31positive cells. Histological images were captured using a digital camera (DP72; Olympus). To measure vessel density in each section, the 5 most vascularized elds (hotspot areas) were manually selected under low magni cation (´40 or ´100) by different researchers to avoid technical problems, and the percentage of each eld stained for CD31 (´200) was calculated using ImageJ software (National Institutes of Health, Bethesda, MD, US). This hotspot selection process may have lacked reproducibility, so the selected areas represented high vascularized regions and different tumor localizations.

Figure 2
Contrast-enhanced high-frequency ultrasound imaging a, Representative heatmaps of the PALN on days 0 and 10 after inoculation of KM-Luc/GFP cells into the SiLN and on days 21, 28 and 35 after inoculation of FM3A-Luc cells. The heatmaps were generated from contrast-enhanced high-frequency ultrasound images using UCAD software. Arrows: perfusion defect. b, Changes in normalized vascular area ratio and normalized volume of the PALN. KM-Luc/GFP or FM3A-Luc cells were inoculated into the SiLN to induce metastasis to the PALN. KM-Luc/GFP cells: day 4 (n = 6), day 7 (n = 6) and day 10 (n = 6). FM3A-Luc cells: day 21 (n = 10), day 28 (n = 9) and day 35 (n = 10). There were no signi cant changes in normalized LN volume or normalized vascular area ratio in the metastatic PALN (one-way ANOVA and Tukey's post-hoc test). Data are shown as mean ± S.E.M. ns: not signi cant.  Characteristics of the blood vessels in the metastatic PALN KM-Luc/GFP cells or FM3A-Luc cells were inoculated into the SiLN to induce metastasis to the PALN. a, Changes in the normalized blood vessel volume and normalized LN size measured by contrast-enhanced micro-CT. Blood vessel volume for the KM-Luc/GFP group, one-way ANOVA, * P < 0.05, day 0 vs day 7, ** P < 0.005, day 0 vs day 10; blood vessel volume for the FM3A-Luc group, one-way ANOVA, * P < 0.05, day 0 vs day 21, ** P < 0.005, day 0 vs day 28. Data are shown as mean ± S.E.M. b, Total blood vessel length measured by contrast-enhanced micro-CT. Data are shown as mean ± S.E.M. There were no signi cant differences among the groups. ns: not signi cant. c, Distribution of blood vessel diameter in the PALN after the inoculation of KM-Luc/GFP cells. Blood vessels with a diameter < 50 µm accounted for more than 80% of blood vessels on day 0. The proportion of blood vessels with a diameter < 50 µm decreased with time, and the proportion of blood vessels with diameters in the range 50-60 µm increased by more than 50%. KM-Luc/GFP cells: day 4 (n = 5), day 7 (n = 6) and day 10 (n = 5). Blood vessel diameter distribution, < 50 μm, one-way ANOVA, **** P < Figure 5 Histological evaluation of perfusion defects in PALNs a, Serial sections of metastatic PALNs on day 10 after inoculation of KM-Luc/GFP cells and day 35 after inoculation of FM3A-Luc cells into the SiLN. The sections were stained with HE or immunostained for CD31. KM-Luc/GFP cells formed a tumor mass in the LN and proliferated expansively, and the boundary with the surrounding normal tissue was quite well de ned. There were no mature blood vessels containing blood in the tumor mass. There were blood vessels around the tumor mass that appeared to be compressed, and some dilated blood vessels were also evident. FM3A-Luc cells proliferated invasively in the LN, and the boundary with surrounding normal tissue was poorly de ned. No mature blood vessels containing blood were seen in the tumor tissue.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.