Paracellular and transcellular migration of metastatic cells through the cerebral endothelium

Abstract Breast cancer and melanoma are among the most frequent cancer types leading to brain metastases. Despite the unquestionable clinical significance, important aspects of the development of secondary tumours of the central nervous system are largely uncharacterized, including extravasation of metastatic cells through the blood‐brain barrier. By using transmission electron microscopy, here we followed interactions of cancer cells and brain endothelial cells during the adhesion, intercalation/incorporation and transendothelial migration steps. We observed that brain endothelial cells were actively involved in the initial phases of the extravasation by extending filopodia‐like membrane protrusions towards the tumour cells. Melanoma cells tended to intercalate between endothelial cells and to transmigrate by utilizing the paracellular route. On the other hand, breast cancer cells were frequently incorporated into the endothelium and were able to migrate through the transcellular way from the apical to the basolateral side of brain endothelial cells. When co‐culturing melanoma cells with cerebral endothelial cells, we observed N‐cadherin enrichment at melanoma‐melanoma and melanoma‐endothelial cell borders. However, for breast cancer cells N‐cadherin proved to be dispensable for the transendothelial migration both in vitro and in vivo. Our results indicate that breast cancer cells are more effective in the transcellular type of migration than melanoma cells.


| INTRODUC TI ON
Brain metastases are devastating complications of lung or breast carcinoma, melanoma and other cancer types, characterized by challenging treatment options. 1,2 Melanoma is the third most common source of brain metastases, 3 but has the highest risk to spread into the central nervous system (CNS). 4 Brain metastatic lesions can be found in approximately three of four patients dying of melanoma. 5 Breast cancer is the second most frequent cause of CNS metastases. 3 Among different subtypes, triple negative (ie negative for estrogen and progesterone hormone receptors and also for Her2/human epidermal growth factor receptor 2) and Her2-enriched (ie negative for hormone receptors and positive for Her2) mammary tumours are the most prone to give secondary brain tumours. 6 Incidence proportion of brain metastases was found to be 11.37% and 11.45% in patients having metastatic triple negative or metastatic Her2-enriched breast tumours, respectively. 7 In brain metastatic melanoma, novel systemic targeted therapies and immunotherapies significantly increased the median overall survival of the patients; however, it is still below 2 years. 8,9 Among patients with breast cancer brain metastasis, the median overall survival time was reported to be 6 months for the triple negative subtype and 10 months for the Her2-enriched subtype. 7 Considering this poor prognosis, new treatment and prevention strategies are urgently needed, 10 which imply understanding of the mechanisms involved. Development of metastases of the CNS depends on the unique interaction between tumour cells and cells of the neurovascular unit (NVU). 11 It has been suggested that both bloodbrain barrier (BBB)-forming tightly interconnected endothelial cells and glial cells have a Janus-faced attitude towards cancer cells, that is killing the vast majority of brain invading metastatic cells, but protecting those which are able to overcome the detrimental mechanisms. 12 Here we addressed the first unique step of brain metastasis formation, that is diapedesis of tumour cells through the cerebral endothelium. Interconnected by continuous tight junctions (TJs), brain endothelial cells form the tightest endothelial barrier in the organism 13 ; therefore, extravasation of metastatic cells through this highly restrictive vasculature is a key step in the development of secondary brain tumours. After adhesion to the luminal surface of endothelial cells, cancer cells may incorporate into the endothelial monolayer, 14 followed by the transmigration step.
Transendothelial migration of tumour cells might involve disruption of the TJs and consequently the paracellular movement of the tumour cells. We have previously demonstrated the ability of melanoma cells to breach the junctional complex of cerebral endothelial cells (CECs) through direct contact and secretion of soluble factors. 15,16 Not only TJs, but adherens junctions (AJs) might also be involved in this process, especially the transmembrane cadherin proteins. In non-brain endothelial cells, N-cadherin was observed to participate in the formation of heterocellular contacts between tumour cells and endothelial cells during transendothelial migration of melanoma cells. 17 N-cadherin is up-regulated in response to transforming growth factorβ (TGFβ) released by metastatic cells, leading to an increase in the ability of melanoma cells to attach to and to migrate through CECs. 18 Besides the paracellular route, brain-invading cells might also take the transcellular way, through individual endothelial cells. Our previous results raised the possibility of transcellular migration of breast cancer cells 16 ; however, no direct evidence of the movement of cancer cells through the cell body of CECs has existed so far.
By using high resolution transmission electron microscopy (TEM), here we aimed to investigate the transmigration routes of melanoma and triple negative breast cancer cells through brain endothelial cells. In addition, involvement of N-cadherin in these processes was also assessed.  blood-brain barrier, brain metastasis, breast cancer, cerebral endothelial cell, incorporation, intercalation, melanoma, N-cadherin, paracellular, transcellular Medium-2 (EGM-2) Endothelial Bullet Kit including basal medium and supplements (Lonza) and 2.5% FBS (Sigma Aldrich). All cell lines were routinely tested for mycoplasma infection.

| Cell culture and in vitro models
Primary rat brain endothelial cells (RBECs) were isolated from 2-to 3-week-old rats. 20 After the removal of meninges, cerebral cortices were cut into small pieces and digested with 1 mg/mL collagenase type 2 (Sigma Aldrich) for 75 minutes at 37°C. After separation of myelin by centrifugation in 20% BSA, a second digestion was performed using 1 mg/mL collagenase/dispase (Roche) for 50 minutes at 37°C. Microvessel fragments were collected after centrifugation on Percoll (Sigma Aldrich) gradient (10 minutes on 1000g) and plated onto fibronectin/collagen-coated dishes. Endothelial cells growing out of the microvessels were cultured in DMEM Nutrient F-12 Ham (DMEM/F12, Thermo Fischer Scientific), 10% plasma-derived serum (PDS, First Link), insulin-transferrin-sodium selenite (ITS) supplement (Sigma Aldrich), heparin (Sigma Aldrich) and basic fibroblast growth factor (bFGF, Sigma Aldrich). In the first 2 days, 4 μg/mL puromycin was added to the culture medium to remove contaminating cells.
Endothelial-tumour cell co-cultures were prepared as previously described. 15,16 Briefly, brain endothelial cells were cultured until confluence in filter inserts (Corning-Costar Transwell Clear, Corning, NY, USA; for electron microscopy), microscope slides (ibidi, for immunofluorescence), culture dishes (for Western blot) or E-plates (for impedance measurements). Tumour cells were seeded upon the endothelial monolayer in a density of 0.5-1.5 × 10 5 cells/cm 2 surface and left for 5-24 hours. CellTracker Red CMTPX (Thermo Fisher Scientific) staining was performed according to the manufacturer's instructions.

| Preparation of ultrasections and transmission electron microscopy (TEM)
Whole brains were embedded in 10% gelatin and 100 μm sections were prepared using a Leica VT1000 S vibratome. Sections were examined under a fluorescence microscope. Sections containing EmGFP-4T1 cells were further used.
The filter inserts or the selected brain slices were fixed for 2.5 hours in 2.7% glutaraldehyde and post-fixed for 75 minutes in 2% osmium tetroxide. After dehydration in graded ethanol baths, the samples were immersed in graded ethanol-Epon baths and then embedded in Epon 812. The blocks were cut with a Leica EM UC7 ultramicrotome, and the 50 nm thick sections were stained with uranyl acetate and lead citrate, then analysed with a Tecnai 12 Biotwin TEM. Brain sections were placed in plates and subjected to antigen retrieval using 100% methanol for 30 minutes. Permeabilization was performed with 0.5% TritonX-100 for 30 minutes at room temperature, followed by blocking with 3% BSA in PBS. The first antibody was applied in a dilution of 1:100 in 1% BSA overnight at 4°C. After washing in PBS, the secondary antibody (STAR RED anti-mouse IgG; Abberior, Göttingen, Germany) was applied in a dilution of 1:500 in PBS for 1 hour at room temperature.

| Immunofluorescence and fluorescence microscopy
After three further washing steps, samples were mounted with FluoroMount-G media (SouthernBiotech). Nuclear staining was performed with Hoechst 33342 (0.66 μg/mL) during the second washing step. Fluorescent signals were examined with a Leica SP5 confocal laser scanning microscope. HRP-conjugated anti-rabbit IgG (1:1000, Cell Signalling Technology) or HRP-conjugated anti-mouse IgG (1:4000, BD Transduction Laboratories). After washing, immunoreaction was visualized using the Clarity Chemiluminescent Substrate (Bio-Rad) in a ChemiDoc MP imaging system (Bio-Rad). Image lab software version 5.2 (Bio-Rad) was used for the quantification of the blots by densitometry.

| Real-time impedance monitoring
To monitor the effects of tumour cells on RBECs in real time, we measured the electrical impedance using the xCELLi-

| Interactions of melanoma cells with brain endothelial cells in vitro
Since our previous results indicated that melanoma cells have increased ability to attach to and to migrate through brain endothelial cells than breast cancer cells, we aimed to investigate these phenomena at ultrastructural level.
We first focused on the adhesion step, which precedes transmigration of tumour cells through endothelial cells. We observed several melanoma cells attached to brain endothelial cells in close proximity to the interendothelial junctions ( Figure 1A), but also in regions distant from endothelial-endothelial contacts ( Figure 1B).  Figure 2C) or, more often, were seen in the neighbourhood of the damaged endothelial cells ( Figure 2D).

| Interactions of breast cancer cells with brain endothelial cells in vitro
Similar to melanoma cells, we could also identify breast cancer cells attached to cerebral endothelial junctions ( Figure 3A and B), although less in number. In the proximity of these cells, filopodia-like endothelial protrusions could be seen, similar to those observed in the vicinity of melanoma cells.
We have also detected several breast cancer cells completely

Besides disruption of TJs, melanoma cells must open the AJs of
CECs during their paracellular migration from the apical to the basolateral side of the endothelium. N-cadherin-mediated interaction was shown to be involved in this process in non-brain endothelial cells. 17 Therefore, we investigated involvement of N-cadherin in When melanoma cells were seeded upon a confluent monolayer of CECs, tumour cells tended to rapidly intercalate among CECs. We observed the appearance of N-cadherin in the melanoma-melanoma and melanoma-endothelial contact regions ( Figure 5A). However, almost no N-cadherin was detected in endothelial-breast cancer cell co-cultures ( Figure 5B). Our Western blot results indicated that both A2058 and B16/F10 melanoma cells expressed high levels of N-cadherin, but no N-cadherin protein was detected in our breast cancer cell lines ( Figure 5C).
Therefore, as a next step we investigated the ability of N-cadherin-negative breast cancer cells to give brain metastases in vivo. As a unique feature of brain metastasis formation, tumour cells arrested in cerebral capillaries survive for long time (approximately 3-5 days) intravascularly before completing transmigration. 12 Therefore, the first timepoint studied in the in vivo setup was day 5 after the injection of the tumour cells into the circulation of mice. At this timepoint, we observed transmigrating tumour cells which were all N-cadherin negative ( Figure 6A). On the other hand, expression of N-cadherin was induced in the cerebral endothelium in the vicinity of some of the transmigrating cells ( Figure 6A). By day 12 after the inoculation, several micro-and macrometastatic lesions were formed in the brain parenchyma. Importantly, 4T1 cells remained N-cadherin negative throughout the metastatic process. N-cadherin was only detected in some vascular segments in the endothelium of tumour cell-bearing mice ( Figure 6B). These results suggest that N-cadherin is not necessarily needed by breast cancer cells to migrate through the brain vasculature and to form metastases in the CNS.

| Interactions of breast cancer cells with the brain endothelium in vivo
Finally, we examined interactions of metastatic breast cancer cells with the brain endothelium in vivo using TEM. In the brain sections obtained from mice injected with breast carcinoma cells, we observed active involvement of CECs in the metastatic extravasation process. Endothelial protrusions covering extravasating cancer cells were seen ( Figure 7A and B).

| D ISCUSS I ON
Development of brain metastases is largely dependent on the ability of the tumour cells to migrate through the tightest endothelium of the organism, which forms the BBB. Involvement of CECs in extravasation of cancer cells into the CNS is largely uncharacterized and might be both offensive and defensive at the same time with the invading cells. 12 By using TEM-a high resolution morphology technique-we assessed interactions of CECs with two of the most aggressive brain metastatic cells, that is melanoma and triple negative breast cancer cells.
Our in vitro and in vivo results indicate that CECs play an active role in the transendothelial migration of the tumour cells by extending filopodia-like processes, which might guide invading cells towards low resistance points. 22 Through this mechanism, endothelial cells may also incorporate the tumour cells or their extracellular vesicles, or isolate them from the circulating blood in vivo. Further studies are needed to understand whether this is a "friend or foe" reaction of endothelial cells, that is, whether endothelial protrusions facilitate transendothelial migration or engulf the tumour cells to protect the brain.
The observed ruffling of the endothelial plasma membrane is reminiscent of macropinocytosis, 23 which is the entry route for platelet-derived microparticles, 24  Among brain metastatic tumours, melanoma has the highest affinity towards the CNS. Earlier, this has been explained by the good capacity of melanoma cells to proliferate in the brain parenchyma. 27 However, our previous results suggested that melanoma cells might also have an increased ability to migrate through cerebral endothelial layers in comparison to breast cancer cells. 16 Especially, impairment of TJs of CECs was more pronounced in the presence of melanoma than mammary cancer cells.
Our present results are in line with these data, showing that melanoma cells can effectively use the paracellular route of transmigration. As a preceding step, melanoma cells intercalate between endothelial cells, which has previously been referred to as "incorporation". 14 However, based on the differences between diapedesis of melanoma and breast cancer cells through cerebral endothelial cells, presented here, we suggest using the term "intercalation" when tumour cells localize between two endothelial cells to proceed further to the paracellular transmigration. We use the term "incorporation" for describing tumour cells-independently whether intact or not-completely covered by endothelial cells. This phenomenon was mostly seen with breast cancer cells, most likely linked to the transcellular type of transendothelial migration.
To our best knowledge, we are the first to show direct evidence of transcellular migration of tumour cells through the BBB.
The transcellular route of migration has initially been recognized for leukocytes, 28 especially in the brain microvasculature. 29,30 As for tumour cells, the transcellular route of migration has only is directly involved in extravasation of tumour cells into the brain.
We also show that melanoma cells primarily utilize the paracellular route of transendothelial migration, while breast cancer cells are able to transcellularly migrate through the brain endothelial cell layer. During extravasation into the brain, triple negative breast cancer cells can migrate through the vessel walls in an N-cadherinindependent manner.

ACK N OWLED G EM ENTS
We acknowledge the help of Synnove N. Magnussen (UiT-The Arctic University of Norway, Tromso) in the preparation of EmGFP-4T1 cells. We also thank Gergely Groma (Department of Dermatology and Allergology, University of Szeged) for sorting the tdTomato-4T1 cells.

CO N FLI C T O F I NTE R E S T
None.