A Four-Dimensional Organoid System to Visualize Cancer Cell Vascular Invasion

Simple Summary Using vascular organoid culture with collagen microfiber, we have established a method for culturing organoids that recapitulates the vascular invasion process of cancer cells. This culture model made it possible to four-dimensionally evaluate the dynamics of cancer cells infiltrating into blood vessels. Abstract Vascular invasion of cancer is a critical step in cancer progression, but no drug has been developed to inhibit vascular invasion. To achieve the eradication of cancer metastasis, elucidation of the mechanism for vascular invasion and the development of innovative treatment methods are required. Here, a simple and reproducible vascular invasion model is established using a vascular organoid culture in a fibrin gel with collagen microfibers. Using this model, it was possible to observe and evaluate the cell dynamics and histological positional relationship of invasive cancer cells in four dimensions. Cancer-derived exosomes promoted the vascular invasion of cancer cells and loosened tight junctions in the vascular endothelium. As a new evaluation method, research using this vascular invasion mimic model will be advanced, and applications to the evaluation of the vascular invasion suppression effect of a drug are expected.


Introduction
Metastasis is a complex, multistep process that begins with cancer cells within the primary tumor that gain the ability to invade through the basement membrane that separates the epithelium from the underlying stroma [1]. Vascular invasion of cancer cells is an essential step in the mechanism of tumor

Establishment and Four-Dimensional (4D) Evaluation of the Model of Cancer Cell Invasion into the Vascular Lumen
Next, a cancer cell invasion model was created using vascular organoid techniques in a fibrin gel with CMF. First, to examine whether vascular cancer cell invasion into the vascular lumen occurred, two types of gels were used: those containing vascular organoids and those containing cancer cells. However, the cancer cells were not confirmed to have invaded the gap between the gels (see Figure S1A), suggesting that the distance between each blood vessel and cancer cell was too large

Establishment and Four-Dimensional (4D) Evaluation of the Model of Cancer Cell Invasion into the Vascular Lumen
Next, a cancer cell invasion model was created using vascular organoid techniques in a fibrin gel with CMF. First, to examine whether vascular cancer cell invasion into the vascular lumen occurred, two types of gels were used: those containing vascular organoids and those containing cancer cells. However, the cancer cells were not confirmed to have invaded the gap between the gels (see Figure S1A), suggesting that the distance between each blood vessel and cancer cell was too large to induce cancer cell invasion into vessels. Thus, a coculture of cancer cells, normal human dermal fibroblast cells (NHDFs), HUVECs, as well as CMF plus thrombin and fibrinogen, was devised, which formed the fibrin network in the gel. A large number of cancer cells (HCT116), labeled with red CMTPX dye, were deformed along the luminal side in the blood vessel, where cells were labeled with green CMFDA dye and a part of the cell body was entering the blood capillary (Figure 2A,B). To exclude any possibility that cancer cells were attached randomly to the blood vessel wall, the samples were observed by confocal microscopy, which allowed 3D analysis. HT29 cells marked in red were located inside the area of the capillaries labeled in green (see Video S2A). In sliced samples, immunohistochemical analysis, with the combination of fluorescent immunostaining, clearly confirmed that cancer cells invaded the lumen ( Figure 2C and Figure S2A). On every slide that was observed in the histological examination, there was no sign of fibroblast invasion into the lumen (data not shown). Although the same cells were cocultured on a dish (2D culture), no vascular network formation or cancer cell invasion was observed (see Figure S1B). To determine whether HUVECs might be involved in the growth of HT29 cells in the fibrin gel, the growth of a single culture of HT29 and the coculture was compared; HUVECs did not affect the proliferation of HT29 cells (see Figure S1D). Next, the 4D culture in the fibrin gel was observed using CLSM in a time-lapse series every hour from 48-96 h after culture. Multiple cancer cells moved into the blood vessel regions (see Video S2B). To observe the detailed mechanism of invasive cell dynamics, time-lapse photography was performed using CLSM (Fluoview FV10i, Olympus) with a 60× objective lens every hour during a culture period of 24-72 h. After cancer cell clusters adhered to the vessel capillaries in the gel, a part of the cancer cell clusters invaded and separated into the capillaries. Additionally, it was confirmed that the cancer cell clusters moved into vessel capillaries, along with proliferation and division characteristics (see Video S3A). By observing the same region from a different angle, it was confirmed that cancer cells were moving inside the blood vessels but not outside the vessel capillaries (see Video S3B). The observation of the sample revealed that there was a difference in cell kinetics between cancer cells invading the capillary and cells distant from the capillary; the migration distance of the cells in the gel was measured using Imaris software (see Video S3C). Although cancer cells outside of the blood vessels remained in the fibrin gel without moving around, the cells that had completely entered the vessel capillary continued to extend their distance after detachment from the cancer cell cluster ( Figure 2D and Figure S1E, Video S3D ). cells in the gel was measured using Imaris software (see Video S3C). Although cancer cells outside of the blood vessels remained in the fibrin gel without moving around, the cells that had completely entered the vessel capillary continued to extend their distance after detachment from the cancer cell cluster ( Figures 2D and S1E, Video S3D).

Evaluation by 3D Construction
To study the detailed mechanism of the 3D positional relationship, as well as the cell morphology of cancer cells, confocal microscopic observation was performed with fluorescence signal data using Imaris software, which was analyzed by 3D image reconstruction to observe the composition of capillary lumens. The data indicated that the blood vessels induced luminal formation, and cancer cells had clearly entered the lumen ( Figure 3A). Cancer cells entering the vessel capillary showed morphological changes, that is, the induction of a spindle-shape formation, reminiscent of the epithelial-to-mesenchymal transition (EMT) phenotype ( Figure 3B). Given that the occurrence of EMT in capillaries was hypothesized, EMT of the corresponding sample was studied by fluorescent immunostaining with an anti-E-cadherin antibody. The expression of E-cadherin was reduced in cancer cells within vessel capillaries compared to cancer cells outside vessel capillaries, suggesting that cancer cells induced EMT when entering the vessel capillary ( Figure 3C,D).
bar, 20 μm. (C) Immunohistochemical staining of vascular organoids. The first antibody is anti-CK20 (brown), which binds HT29 cancer cells. These cancer cells have invaded the vascular lumen. Scale bar, 50 μm. (D) The migration length on the vertical axis of cancer cells within the fibrin gel at the observation period of 48 h. Two types of cells, extravascular cells and infiltrating cells, are compared. The cells in the graph show morphological changes with time on the horizontal axis. The time axis progresses from right to left.

Evaluation by 3D Construction
To study the detailed mechanism of the 3D positional relationship, as well as the cell morphology of cancer cells, confocal microscopic observation was performed with fluorescence signal data using Imaris software, which was analyzed by 3D image reconstruction to observe the composition of capillary lumens. The data indicated that the blood vessels induced luminal formation, and cancer cells had clearly entered the lumen ( Figure 3A). Cancer cells entering the vessel capillary showed morphological changes, that is, the induction of a spindle-shape formation, reminiscent of the epithelial-to-mesenchymal transition (EMT) phenotype ( Figure 3B). Given that the occurrence of EMT in capillaries was hypothesized, EMT of the corresponding sample was studied by fluorescent immunostaining with an anti-E-cadherin antibody. The expression of E-cadherin was reduced in cancer cells within vessel capillaries compared to cancer cells outside vessel capillaries, suggesting that cancer cells induced EMT when entering the vessel capillary ( Figure 3C,D).

Involvement of Vascular Endothelium Exosomes in Vascular Invasion of Colorectal Cancer (CRC)
To determine how exosomes derived from CRC cells affect vascular endothelial cells during the invasion of tumor cells into the vascular lumen, exosomes were collected by ultracentrifugation from the culture medium of KM12-SM cells, a highly metastatic colon cancer cell line of humans [29]. The size of the collected microvesicles was measured using NanoSight (NS-10, Malvern, UK), and it was confirmed that each vesicle size was compatible with exosomes (see Figure S2A). Exosomes were labeled with PKH67 dye and were incorporated into the cytoplasm of HUVEC cells in the coculture of HUVECs and exosomes (see Figure S2B). Although the effect of exosomes exposed to the formation of vascular networks was studied, no significant difference was observed in vessel volume in the fibrin gel (see Figure S2C). Then, exosome-exposed HUVECs were cocultured with cancer cells in the gel, and the degree of invasion by the cancer cells was studied. The proportion of cancer cells that invaded vessels was significantly higher in the exosome-exposed group than in the nonexposed control group (Figure 4A,B). To study the enhancement mechanism of cancer cell invasion into the vascular lumen by exosomes, immunohistochemical analysis of p120 and VE-cadherin was performed in the vessels, which are reportedly tight-junction-related proteins in the vascular endothelium [30,31].
To study the involvement of cancer cells in the expression of these proteins, vascular organoids were cultured using HUVECs exposed to exosomes, but the HUVECs did not contain cancer cells in the gel; then, the expression of each protein was studied by fluorescence staining and signal volume calculation. The expression of VE-cadherin in the vascular endothelium was significantly lower in the exosome-exposed group than in the nonexposed control group (Figure 4C,D). Similarly, the expression of the p120 protein in the vascular endothelium was significantly lower in the exosome-exposed group than in the nonexposed control group (Figure 4E,F). These results suggested that cancer-derived exosomes reduced vascular endothelium tight junctions and increased the vascular permeability of cancer cells. Clinical open database analysis, using the PROGgeneV2 database and dataset GSE28814, confirmed that metastasis-free survival (MFS) was significantly higher in the p120 high-expression group than that in the low-expression group ( Figure 4G). This suggested that MFS might be affected by exosomes derived from cancer cells, following the alteration of tight junctions.

Discussion
CRC is the third most common cancer in the world and the fourth most common cause of cancer death, and it is expected to increase in the future [32,33]. Prognosis has improved due to developments in surgical techniques and chemotherapeutic approaches against CRC [34], although stage IV CRC patients with recurrence and metastasis have a 5-year survival rate of 20-30% and a poor prognosis [34,35]. Therefore, further treatment development is desired. Generally, cancer metastasis occurs in the advanced stages of tumors, of which aggressive behaviors are involved in the prognosis of cancer patients. To improve the outcome of colorectal cancer, further studies of the mechanism of cancer metastasis and invasion, as well as the development of new therapeutic approaches, are necessary. To the best of our knowledge, no drugs that directly inhibit the invasion of blood vessels have been developed for clinical settings. Given that it is difficult to achieve complete eradication of cancer metastases with standard chemoradiation treatments, the development of innovative methods to inhibit cancer metastasis, especially controlling the invasion of cancer cells into the vascular lumen, is required. Using the present invasion mimic model of cancer cells into the vascular lumen to conduct research, it will be possible to elucidate the mechanism of cancer cell invasion in detail, which can lead to the development of inhibitors for invasion.
Reportedly, 3D culture has cell-cell interaction and maintains a cell function that is closer to that of a living body compared to 2D monolayer culture [15]. In addition, procedures in 3D culture are easier to perform, and 3D culture has increased reproducibility compared to in-vivo animal experiments, showing the benefit of 3D culture. Considering the emerging surveillance of compliance rules for ethical issues in animal experiments [17], developing 3D organoid research would be more advantageous for cancer studies than other methods. Beyond the 3D structure, this method can be further applied to the novel mimic model that enables 4D evaluation of vascular invasion using time-lapse videos with a time axis. In the vascular invasion mimic model, it is useful to evaluate cell dynamics and functions that are difficult to evaluate at the cellular level in in-vivo experiments; eventually, the involvement of the p120 protein and the EMT was observed ( Figure 4H). In hepatocellular carcinoma, cancer-derived exosomes increase vascular permeability as an effect on the tumor microenvironment [35,36]. These results show the same phenomenon, which suggests that this is a reliable model for vascular invasion. The present mimic model enables the elucidation of the mechanism of vascular invasion and the application of drug tests on the vascular invasion.
Although previous studies have reported modeling invasion of cancer cells into the vascular lumen, those studies used endothelial cells in a monolayer manner, in which cancer cells pass through [37]; therefore, they are not suitable for the study of vascular infiltration of cancer cells utilizing vascular organoids. In this regard, in vitro pathological mimic models of the tumor microenvironment are easily constructed, highly reproducible, and easy to evaluate, compared to in-vivo experiments and microfluidic devices. Thus, in the present study, a new, important platform for promoting invasion research of cancer cells is created. In vivo imaging does not allow the study of detailed mechanisms of tissues in a real-time manner. In the case of 3D evaluation of biological tissues with a fluorescence optical microscope, it is essential to make the tissue transparent for observations; however, transparency requires a lot of time and labor. In sharp contrast, the 3D culture in the present study was performed in a fibrin gel with collagen microfibers. The sample was studied without clearing tissues, and the cell dynamics and histological positional relationship was studied in real-time. During cancer cell progression, cancer cells undergo various changes in their invasion, including EMT, to become circulating tumor cells. To study this, it is necessary to understand the dynamics of such cells and to detect genetic changes in each situation. To achieve this, time-lapse observation of cell dynamics is an important observation method.
The present mimic model has some limitations: the vessel capillary has a monolayer structure and lacks the backing of the basement membrane and pericytes. Therefore, the possibility that the barrier mechanism against infiltration may be weaker than the original vasculature must be considered. In the vascular invasion model, there is also no fluid flow in the vessel, and the mechanism of cancer cell movement in the vessel is unknown. Although E-cadherin expression was evaluated, it is unclear when cancer cells decrease the expression of E-cadherin during the vascular invasion. Nevertheless, this mimic model has merits. A vascular organoid and invasion model was easily made, with high reproducibility, and was used to clearly observe the samples in detail.

Vascular Organoids in a Fibrin Gel with CMF
Collagen type 1 sponges from pigs were kindly provided by Nippon Ham (Osaka, Japan). The CMF-200 µm was fabricated from a collagen type 1 sponge after dehydration condensation, based on a previous study [38]. The collagen type 1 sponge was dehydrothermally treated by drying with heat under a vacuum at 200 • C for 24 h for crosslinking. The crosslinked collagen sponge was mixed with 10× phosphate-buffered saline (PBS) solution at a concentration of 10 mg/mL (pH = 7.4, 25 • C), and a 6-min homogenization (30,000 rpm) was performed using a Violamo VH-10 homogenizer (S10N-10G), with a probe 10 mm in diameter and 115 mm in length. The CMF-20 µm suspension was obtained with further ultrasonication (ultrasonic processor VC50, 50 W, 20 kHz) in an ice bath for 100 cycles (1 cycle included 20 s ultrasonication and 10 s cooling). The solution was transferred to a glass recipient after filtration (40 µm filter, microsyringe 25 mm filter holder, Merck), and the filtrate was freeze-dried for 48 h (freeze dryer FDU-2200, Eyela Co., Tokyo, Japan). The obtained CMF-20 µm was kept in a desiccator at room temperature (RT). For the preparation of 3D capillary tissue, NHDFs and HUVECs were trypsinized (5 min, 37 • C) and collected by centrifugation (5 min, 1000 rpm, RT). The culture solution used in the following operation was a mixture of EGM-2 and DMEM (FBS free, 1% antibiotics) in equivalent amounts. CMF-20 µm (0.15 mg), thrombin (0.15 Unit), and cells (1 × 10 5 NHDFs and 5 × 10 4 HUVECs) were mixed with 20 µL medium, and then 0.15 mg of fibrinogen was dissolved in 10 µL of medium at 37 • C for 30 min to prepare the fibrinogen solution; two types of solutions were mixed in a tube. The completed mixture (30 µL) was quickly dropped onto the culture dish or glass-bottom dish. In the vascular invasion model, 1 × 10 4 cancer cells were added and mixed with NHDFs and HUVECs, as described above. After 30 min of gelation at 37 • C, the gel samples were further cultured at 37 • C and 5% CO 2 in a humidified incubator.

Immunohistochemical Staining
The expression of CD31 was assessed by immunohistochemical staining of formalin-fixed and paraffin-embedded vascular organoids in a fibrin gel with CMF-20 µm. To liberate antibody-binding sites using L.A.B. solution (Polysciences Inc., Warrington, PA, USA) and to block endogenous peroxidase activity, blocking was performed for 20 min at room temperature using a VECTASTAIN Elite ABC kit (mouse IgG; #PK-6102; Vector Laboratories, Burlingame, CA, USA). Then, 3.5-µm thick sections were incubated overnight at 4 • C with the mouse monoclonal anti-CD31 antibody (dilution, 1:200; Wako, M0823, Osaka, Japan). Hematoxylin was used for nuclear staining for 1 min. Dehydration was performed using a graded ethanol series of 60%, 70%, 80%, 90%, and 95% ethanol for 1 min each, 100% ethanol for 2 min, twice, and xylene for 5 min, 3 times. Images were captured using a BZ-710 All-in-One Fluorescence Microscope (KEYENCE Corporation, Osaka, Japan).

Cell Tracker Labeling
HCT116 and HUVEC cells were labeled with either red CMTPX dye or green CMFDA dye (Carlsbad, CA, USA), according to the manufacturer's protocol. Confluent cells (~80% confluent) on a 10-cm dish were washed with PBS and then incubated with DMEM containing 1/1000 Cell Tracker dye at 37 • C for 30 min and kept in the dark. After incubation, the DMEM was removed, and cells were washed with PBS for subsequent experiments.

Exosome Isolation and Treatment
Exosomes were purified from CRC-derived conditioned medium (CM) by ultracentrifugation. CRC cell lines were cultured in DMEM supplemented with 3% exosome-depleted FBS (Exo-FBS-50A-1, SBI, Fremont, CA, USA). CM was collected after 48 h of cell culture and centrifuged at 500× g for 10 min at 4 • C, followed by centrifugation at 2000× g for 10 min at 4 • C. The supernatants were passed through a 0.22 µm filter (8020, IWAKI, Shizuoka, Japan) and ultracentrifuged at 174,900× g for 84 min at 4 • C. The exosomal pellets were washed with PBS, followed by a second ultracentrifugation at 174,900× g for 84 min at 4 • C, and then resuspended in PBS. An Optima WE-90 (Beckman, Brea, CA, USA) with SW32Ti as a swing rotor was used for ultracentrifugation. The amount of exosomes was measured as protein using the Qubit3.0 fluorometer (Thermo Fisher Scientific, Waltham, MA, USA). When exosomes were exposed to cells, cell culture was performed by adjusting the medium so that the exosome concentration was 20 µg/mL.
To examine the cellular internalization of exosomes, exosomes were labeled with PKH67GL (Sigma-Aldrich, St. Louis, MO, USA), added to HUVECs at 80% confluence and incubated for 24 h before imaging under CLSM (FV3000, Olympus, Tokyo, Japan). The nucleus was labeled with Hoechst 33324.

Ethic Committee Approval and Code
The experiments were approved by experimental ethical committee under code number 4305 in Osaka University, Japan.

Conclusions
As a new evaluation method, the constructed 3D tissue model will contribute to medical cancer care by suppressing the invasion of cancer cells and helping to control metastasis and cancer recurrence.
Supplementary Materials: The following are available online at http://www.mdpi.com/2079-7737/9/11/361/s1. Figure S1: The interaction between cancer cells and vascular endothelial cells, Figure S2: Effects of cancer-derived exosomes on vascular endothelial cells, Video S1: Time-lapse microscopy recordings of vascular network formation, Video S2: Three-dimensional evaluation of the positional relationship between vessels and cancer cells in vascular invasion model, Video S3: Time-lapse microscopy recordings of cancer vascular invasion with a CLSM every hour for 24 to 72 h of culture.