Non-canonical WNT-signaling controls differentiation of lymphatics and extension lymphangiogenesis via RAC and JNK signaling

Development of lymphatics takes place during embryogenesis, wound healing, inflammation, and cancer. We previously showed that Wnt5a is an essential regulator of lymphatic development in the dermis of mice, however, the mechanisms of action remained unclear. Here, whole-mount immunostaining shows that embryonic day (ED) 18.5 Wnt5a-null mice possess non-functional, cyst-like and often blood-filled lymphatics, in contrast to slender, interconnected lymphatic networks of Wnt5a+/− and wild-type (wt) mice. We then compared lymphatic endothelial cell (LEC) proliferation during ED 12.5, 14.5, 16.5 and 18.5 between Wnt5a−/−, Wnt5a+/− and wt-mice. We did not observe any differences, clearly showing that Wnt5a acts independently of proliferation. Transmission electron microscopy revealed multiple defects of LECs in Wnt5a-null mice, such as malformed inter-endothelial junctions, ruffled cell membrane, intra-luminal bulging of nuclei and cytoplasmic processes. Application of WNT5A protein to ex vivo cultures of dorsal thoracic dermis from ED 15.5 Wnt5a-null mice induced flow-independent development of slender, elongated lymphatic networks after 2 days, in contrast to controls showing an immature lymphatic plexus. Reversely, the application of the WNT-secretion inhibitor LGK974 on ED 15.5 wt-mouse dermis significantly prevented lymphatic network elongation. Correspondingly, tube formation assays with human dermal LECs in vitro revealed increased tube length after WNT5A application. To study the intracellular signaling of WNT5A we used LEC scratch assays. Thereby, inhibition of autocrine WNTs suppressed horizontal migration, whereas application of WNT5A to inhibitor-treated LECs promoted migration. Inhibition of the RHO-GTPase RAC, or the c-Jun N-terminal kinase JNK significantly reduced migration, whereas inhibitors of the protein kinase ROCK did not. WNT5A induced transient phosphorylation of JNK in LECs, which could be inhibited by RAC- and JNK-inhibitors. Our data show that WNT5A induces formation of elongated lymphatic networks through proliferation-independent WNT-signaling via RAC and JNK. Non-canonical WNT-signaling is a major mechanism of extension lymphangiogenesis, and also controls differentiation of lymphatics.


Results
Wnt5a-null mouse embryos show early signs of edema formation. Most of the multiple defects of Wnt5a-null mouse embryos have been shown quite some time ago 27 . Defects of the dermal lymphatic vascular system were demonstrated exclusively in embryonic day (ED) 18.5 embryos 25 . Since the number of progeny of Wnt5a +/− -mice was very low and Wnt5a-null embryos could hardly be recovered especially in late stages of development (there appeared to be a constant loss over time), we regularly crossed Wnt5a +/− -mice (BL/6 background) with normal C57BL/6 mice. This recovered the number of progenies significantly. We observed edema formation already in ED 14.5 Wnt5a −/--embryos ( Fig. 1). Dermal edema was still present at ED 17.5 and the embryos presented with petechial intra-lymphatic bleedings ( Fig. 2A-D), as confirmed by immunostaining with the lymphatic marker Lyve1 (Fig. 2E). Using interstitial injection of 2000kDa FITC-dextran into the paw, Buttler et al. showed that lymphatics of ED 18.5 Wnt5a +/− embryos are functional 25 . Here we observed that at ED 16.5 (not shown) and ED 17.5 transport of lymph from the paw is functional in both wild-type (wt)-mice and Wnt5a +/−mice (Suppl. Fig. 1). Thereby, superficial lymphatic collectors were seen to connect the inguinal with the axillary region. In contrast, Wnt5a-null mice never presented a functional lymphovascular system (Suppl. Fig. 1D).

Wnt5a-null mice possess numerous isolated dermal lymphatic cysts. Previous immunofluores-
cence studies of cryo-sections revealed dilated lymphatics in ED 18.5 Wnt5a −/--mice, and normal lymphatics in wt-and Wnt5a +/− -mice 25 . Here, we studied the lymphatic pattern using whole-mount immunostaining of the dermis. At ED 18.5 we found regular lymphatic networks in wt-and Wnt5a +/− -embryos (Fig. 3A,B), and immature sinusoidal or cysts-like lymphatics in Wnt5a-null embryos (Fig. 3C). Obviously, there is a failure of elongation and connectivity of lymphatic anlagen in Wnt5-null embryos. Blood capillaries appeared to be less frequent in these embryos (Fig. 3). Collectively, our data show a massive failure of the development of dermal lymphatic networks in the absence of Wnt5a.
Proliferation of LECs is unaltered in Wnt5a-null mice. In previous studies it was shown that there is no significant difference in proliferation of LECs of ED 18.5 Wnt5a −/− embryos as compared to controls 25 . Here, we studied if earlier stages of development might be affected. We therefore compared proliferation of LECs in jugular lymph sacs and dermal lymphatics with anti-Ki67 and anti-Prox1 double immunostaining in ED 12.5, ED 14.5, ED 16.5, and ED 18.5 embryos of all three genotypes (Fig. 4). There were no differences in the proliferation rates between Wnt5a −/− , Wnt5a +/− and wt-embryos in any of the developmental stages. Highest proliferation rates were consistently found at ED 14.5. Our data show that Wnt5a does not regulate LEC proliferation. www.nature.com/scientificreports www.nature.com/scientificreports/ Ultrastructure of Wnt5a −/− lymphatics is abnormal. Next, we studied the ultrastructure of lymphatics in Wnt5a −/− , Wnt5a +/− and wt-embryos at ED 18.5. Thereby, the fine-structure of initial dermal lymphatics of wt-embryos displayed typical characteristics such as thin, flat LECs with overlapping junctions, functioning as delicate valves (Fig. 5). We could not detect any obvious alterations in ED 18.5 Wnt +/− embryos (data not shown). However, in Wnt5a −/− embryos the lymphatics displayed clear alterations. The endothelial lining was ruffled and occasionally discontinuous (Fig. 6). The valve-like, overlapping junctions were not existent. We often observed wide inter-endothelial spaces (Fig. 7). There was conspicuous bulging of LEC nuclei and cytoplasm into the lumen, indicating that flattening and cell-cell communication of LECs was abnormal.
WNT5A protein rescues lymphatic morphogenesis in Wnt5a −/− mice. Next, we isolated small pieces of dorsal thoracic skin from ED 15.5 Wnt5a −/− mouse embryos and treated the specimens with recombinant human WNT5A protein vs. medium controls. After two days, anti-Lyve-1 whole-mount staining of controls revealed immature lymphatic sinusoids with large vessel diameter and short inter-branching segments ( Fig. 8A-C). Application of WNT5A rescued formation of lymphatic networks with slender, elongated vessels; indistinguishable from the lymphatic networks of normal ED17.5 mouse dermis (Fig. 8D-F). Quantification with AngioTool 0.5a software revealed a significant decrease in the vessel-covered area in WNT5A-treated specimens, without a significant change in the number of branching points (Fig. 8G,H). The data show that WNT5A controls the differentiation of sinusoidal lymphatic plexuses into slender, elongated lymphatic networks. Thereby, a change in morphology of LECs became visible. Whereas in the control-treated specimens LECs had a wide sweeping  www.nature.com/scientificreports www.nature.com/scientificreports/ morphology, in the WNT5A-treated dermis they were smaller and oriented longitudinally in the direction of the vessels (Fig. 8I,J).

Inhibition of Wnt-secretion inhibits lymphatic maturation.
In the invers experiment, we treated ED 15.5 wt-mouse dermis with the Wnt-secretion inhibitor LGK974. This compound is a potent inhibitor of porcupine (PORCN), a membrane-bound O-acyltransferase, which palmitoylates WNT ligands. Inhibition of PORCN prevents the export of WNT from the endoplasmic reticulum (ER) and inhibits WNT signaling 28 .
LGK974 completely blocked WNT5A secretion in human LECs after 3 days (Suppl. Fig. 2; complete original blot see: Suppl.  www.nature.com/scientificreports www.nature.com/scientificreports/ LGK974, and a mild, significant increase in the number of branching points (Fig. 9G,H). The data provide evidence for the inhibition of lymphatic network maturation after blocking of Wnt-secretion.

WNT5A promotes horizontal migration of LECs in vitro. The above data have shown that
Wnt5a-signaling regulates morphogenesis and differentiation of lymphatics. We then wanted to know, which intracellular signaling pathways are involved. Here, we used the scratch-assay, which can be performed in large numbers to test various activators and inhibitors that affect cell shape and migration of cells in a monolayer.
The second arm of the PCP-pathway involves Rac family small GTPase (RAC) and JNK. To gain deeper insight into this pathway, we used 4 different drugs to inhibit either RAC (NSC23766 and EHT-1864) or JNK (SP600125 and JNK-In-8). All drugs inhibited migration of HD-LECs very efficiently (

WNT5A induces transient phosphorylation of JNK.
Our studies showed that migration of HD-LEC is mediated via RAC and JNK. To examine if WNT5A is capable of activating the JNK pathway, we used LGK974-pre-incubated HD-LECs, stimulated them with recombinant WNT5A and investigated the phosphorylation status of JNK. Since the supplementation of the growth medium with serum led to an overall phosphorylation of JNK, we starved the cells for 24 h in serum-free medium. Our data show that WNT5A induces strong phosphorylation of JNK after 15 min and 30 min, while after 2 h the values decreased to control levels ( Fig. 12; complete original blot see: Suppl. Fig. 5). This proves that WNT5A is a potent activator of JNK. We then tested if the induction of JNK phosphorylation can be inhibited by the RAC-inhibitor EHT1864 and the JNK-inhibitor SP600125 (Fig. 13). After 30 min. stimulation with WNT5A, JNK phosphorylation could be detected, which was inhibited by pretreatment with each of the two inhibitors. (Complete original blot see: Suppl. Fig. 6). The data strongly suggest that WNT5A acts via RAC and JNK. www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
Important functions for Wnt-signaling during angiogenesis have been shown quite some time ago. Abnormal development of blood vessels was noted in mice that are negative for the ligands Wnt2a, Wnt5a, Wnt7a/b, and the receptors Frizzled-4 (Fzd4), and Fzd5 15 . Thereby, endothelium-derived Wnts are important regulators of vascular density, endothelial cell (EC) survival and proliferation, as shown in mice with EC-specific deletion of the pan-Wnt-secretion factor Wls/Evi 29 . Here, we observed hypoplasia of blood capillaries in the dermis of ED 18.5 Wnt5a-null mice, which is in line with recent studies showing that Wnt5a promotes vascular network formation in human adipose-derived microvasculature 30 and induces hypersprouting in mouse retina 31 . www.nature.com/scientificreports www.nature.com/scientificreports/ Regarding the lymphatic vascular system, macrophage-derived Wnts have been studied during inflammation-induced lymphangiogenesis in the cornea 32 . Earlier, regulatory functions for the Wnt-planar cell polarity (PCP) protein Cadherin EGF LAG seven-pass G-type receptor 1 (Celsr1) have been observed in the development of valves in lymphatic collectors 33 . The canonical Wnt pathway seems to be involved in lymphatic valve formation, too, as demonstrated by LEC-specific knock-out of β-catenin 34 . However, mutual interactions between canonical and non-canonical Wnt signaling must be considered. In line with previous models 23 , our studies show that canonical Wnt-signaling inhibits the non-canonical pathway. In LECs, the activators of β-catenin-signaling, BIO and IM-12, inhibit LEC migration via the PCP pathway significantly, showing that proliferation and morphogenesis can be regulated by intracellular interactions of the WNT pathways. Studies by Buttler and coworkers 25 already indicated a morphogenesis defect of lymphatic networks in the dermis of ED 18.5 Wnt5a-null mice, whereas lymphatics of inner organs appeared to be normal. Similarly, abnormal dermal lymphatics were seen in mice after myeloid-cell-specific knock-out (ko) of the pan-Wnt-secretion factor Wls/ Evi 26 . However, these abnormalities are very mild. This may be due to the fact that myeloid-cells (including dermal macrophages) represent just one out of numerous cell types in the dermis expressing Wnts. Wnt5a is highly expressed in epidermis and dermis of murine embryos (see: http://www.genepaint.org/cgi-bin/mgrqcgi94),  www.nature.com/scientificreports www.nature.com/scientificreports/ suggesting concerted activity of both paracrine and autocrine Wnt5a signaling. Since it is hardly possible to knock out Wnt5a in each of the cell types specifically, we chose to study dermis of global Wnt5a-knock-out-mice. Previous data have shown that embryonic lymphangiogenesis is controlled by Wnt-signaling, but detailed studies on the activities of Wnt-ligands, receptors and co-receptors in LECs are missing.

Wnt5a does not affect proliferation of LECs.
Data regarding Wnt5a and endothelial proliferation are inconsistent. A weak but statistically not significant proliferative effect of WNT5A has been reported for HUVECs (human umbilical vein endothelial cells) 35 , while the same group observed an inhibitory effect on bovine aortic endothelial cells 36 . In previous studies we observed a small, though statistically not significant, reduction of LEC numbers in the dermis of ED 18.5 Wnt5a-null mice as compared to controls consisting of a pool of both wt-and heterozygous Wnt5a-ko-mice 25 . Here, we followed up on this topic and compared mice of all three genotypes separately from ED 12.5 -ED 18.5. Highest LEC proliferation rates are consistently present in ED14.5 mice, but there are no differences in Ki67 labeling of LECs between the three genotypes. In line with this, application of recombinant WNT5A protein to human LECs, which were treated for 3d with the PORCN-inhibitor LGK974 to prevent autocrine WNT-signaling, does not affect proliferation. Loss of all Wnt-secretion from myeloid cells significantly increases LEC proliferation in mouse embryos 26 , but the involved components have not been characterized. Regulation of lymphangiogenic growth factors by Wnts in myeloid cells may be a likely hypothesis.
Complex networking of WNTs is also seen in proliferative malignancies in the human. Thereby, significant effects of the canonical pathway (β-catenin, Adenomatous-polyposis-coli, axin) have been well documented e.g. for colon cancer 37 . For WNT5A the data are less clear, however, the majority of studies point to a motility enhancing effect of WNT5A, e.g. by epithelial-mesenchymal transformation 38,39 . Our studies on murine and human LECs strongly support the view that non-canonical Wnt5a signaling regulates aspects of cell shape, motility and differentiation, but not proliferation.

Wnt5a controls differentiation of lymphatics.
We observed that both initial lymphatics and lymphatic collectors are not properly developed in Wnt5a-null mice. Interstitially injected FITC-dextran is not transported centripetally in these mice. In contrast, Wnt5a +/− mice (like wt-mice) take up and transport the marker normally. Thereby, functional lymphatics can be observed in the limbs of ED 16.5 embryos (data not shown), and a very efficient transport is found at ED 17.5. Thereby, lymphatic collectors can be seen accompanying the main blood vessels of the limbs. Additionally, superficial lymphatic collectors are present in the dermis of the trunk and connect the inguinal with the axillary region. They are obviously the reason why metastases of tumors injected into the dermis of the dorsal hip of nude mice are regularly found in axillary lymph nodes. In the human, such collectors do not exist.
Differentiation of initial lymphatics is normal in ED 18.5 Wnt5a +/− mice. Like in wt-mice, the ultrastructural characteristics display typical overlapping junctions, which control the influx of interstitial fluid. In Wnt5a-null mice, these junctions are not present. LECs are ruffled and display protrusions, both luminally and abluminally. The lymphatics are dilated (oedemic) and often contain blood cells. These data strongly suggest that Wnt5a is a crucial regulator of differentiation of lymphatic vessels. Similarly, important functions of Wnt-signaling for the specification of lymphatic vessels have been shown in studies on zebrafish. There, the development of lymphatics from a specific niche in the cardinal vein was demonstrated to be dependent on Wnt5b. Astonishingly, Wnt5b thereby activates the canonical, β-catenin-dependent pathway 40 .
Extension: an underestimated mechanism in lymphangiogenesis. Our studies show that Wnt5a regulates the morphogenesis of dermal lymphatic vessels without influencing proliferation of LECs. The defects of the dermal lymphatics in Wnt5a-null embryos can be described as: (i) failure to form elongated lymphatics with slender lumen and (ii) failure of lymphatic anlagen to connect into networks, resulting in immaturity of lymphatic plexuses and frequent formation of isolated lymphatic cysts. This shows that non-canonical Wnt-signaling in lymphangiogenesis has very similar functions as during early embryogenesis, e.g. during gastrulation. There, the effects of non-canonical Wnt-signaling have been summarized as 'convergent extension' 41,42 .
Convergence of angioblasts and extension of the primitive vascular plexus into an organotypic network are important morphogenetic mechanisms in both hemangiogenesis and lymphangiogenesis. It has been shown that hemangioblasts emerge from various extra-and intra-embryonic compartments and assemble (converge) into vascular tubes, which form a dense primitive vascular plexus. The lumen of these vessels is very wide, so they have been assigned as sinusoidal, primary or 'primitive' . Further development of the vascular plexus takes place by a variety of mechanisms, such as the further integration of angioblasts (convergence), fusion of vessels, sprouting, splitting and intussusception [43][44][45][46] . However, elongation of vessels (extension) is a widely observed mechanism, which has often been neglected, or obscured by terminology such as pruning or vessel maturation 44 . Thereby, the term pruning reflects the cutting away of branches from a tree, however, this is not what actually happens during vascular maturation. Instead, vessel extension takes place, which is immediately visible in flat mounts of the retinal vasculature, where the primitive vascular plexus in the periphery of the retina matures into the elongated vascular network in central areas 47 .
Similarly, convergence and extension take place during lymphangiogenesis. Lymphangioblasts, either originating from the venous system or from mesenchymal sources 13,48,49 , assemble (converge) into primary lymphatic networks, which mature (elongate, extend) into lymphatic capillaries and collectors. Thereby, the functional significance of Wnt-signaling has been noted earlier 15,25,40 , but has not yet been clearly assigned to convergent extension mechanisms.

Wnt5a controls morphogenesis of lymphatics via RAC and JNK. Our studies have shown that
Wnt5a is a highly potent morphogen for the lymphatic vascular system. The ex vivo application of WNT5A (2019) 9:4739 | https://doi.org/10.1038/s41598-019-41299-7 www.nature.com/scientificreports www.nature.com/scientificreports/ protein to dermis of Wnt5a-null mice rescues the development of slender, elongated lymphatic networks. This obviously takes place without any functional flow in these vessels. Later during development, flow appears to be a morphogenic factor for lymphatic collectors 50 . This may involve flow-induced expression of the transcription factor FOXC2, an essential regulator of lymphatic (and venous) valve formation 51 . In tube formation assays, WNT5A significantly improves network formation by increasing the number and length of tubes. However, it was surprising to see that the complete removal of WNT-signaling in this assay induced network formation of LECs. This obviously shows that, although WNT signaling does have an important function in this process, LECs possess alternative signaling pathways to coordinate homophilic cellular interaction. These pathways might involve the glycogen synthase kinase-3β (GSK3β), a highly versatile kinase, which is not only centrally positioned in the WNT-pathway, but also involved in energy metabolism, neurogenesis, and pattern formation 20,52 .
Our systematic analyses of scratch assays with activators and inhibitors of WNT-signaling provided evidence that inhibition of the RHO-GTPase RAC, or the c-Jun N-terminal kinase JNK significantly reduces migration of LECs, and of note, WNT5A potently induces transient phosphorylation of JNK, which can be inhibited by RACand JNK-inhibitors. In contrast, inhibition of the protein kinase ROCK has no effect on LEC migration. Together, these data suggest that WNT5A regulates morphogenesis of lymphatics via WNT-PCP-signaling along the RAC and JNK path.
WNT signaling is an important regulator of the development and differentiation of lymphatics, however, nature seems to have provided LECs (and probably other cell types, too), with an alternative program, which steps in in case of complete absence of WNTs. We have speculated that GSK3β might be involved in this pathway, but this needs to be studied in detail. Then, we observed that both initial lymphatics and lymphatic collectors are maldeveloped in Wnt5a-null mice. It is likely that divergent pathways regulate the differentiation of these vessel types, and again the involvement of WNTs and their interaction with other pathways have to be elucidated. Therefore, the impact of WNTs on lymphangiogenesis and lymphangiogenesis-dependent diseases represents an interesting field for future studies.

Materials and Methods
All methods were carried out in accordance with relevant guidelines and regulations as listed below. Additionally, all experimental protocols on murine embryos and tissues were approved by the animal welfare officer of the University Medical School Goettingen (UMG).
Murine embryos and genotyping. We used Wnt5a [tm1Amc] B6 mice (The Jackson Laboratory; Bar Harbor, Maine, US) and C57BL/6 mice. All mice were bred in the Central Animal Facility of the University Medical Center Göttingen, with a 12 h dark/light cycle and water and food ad libitum. All prescriptions of the German Animal Welfare Act (TierSchG) and the German regulations on the welfare of animals used for experiments or for other scientific purposes (TierSchVersV) were kept. PCR genotyping was performed as described 25 . Immunohistology. Embryos between embryonic day (ED) 12.5 and 18.5 were fixed in 4% Ex vivo culture of embryonic mouse dermis. Dorsal thoracic skin of freshly isolated Wnt5a −/− mouse embryos was dissected at ED 15.5, cut into pieces of app. 0.3-0.5 mm and cultured on Millicell ® cell culture inserts (0.4 µm, 12 mm; Merck Millipore, Billerica, US) with Endothelial Cell Growth Medium MV 2 (with 1% penicillin-streptomycin; PromoCell, Heidelberg, DE) without (controls) or with 500 ng recombinant Wnt5a protein (R&D Systems, Minneapolis, US) at 37 °C and 5% CO 2 . After 24 h, 50 µl fresh culture medium with 100 ng recombinant Wnt5a was added to the experimental group (n = 5). In the controls, only the medium was supplemented with an equal amount of PBS plus 0.1% BSA (n = 5). After 48 h, the dermis was fixed in 4% PFA for 2 h at room temperature, washed 3 × 15 min and blocked with 1% BSA in PBS overnight. Then, samples were stained with anti-Lyve-1-antibodies (1:200; ReliaTech) for 3 days, followed by overnight washing and overnight staining with goat-anti-rabbit Alexa Fluor 594 (1:200; Life technologies) together with DAPI (1:10 000, Thermo Fisher Scientific, Waltham, US). All steps were done under constant shaking at 4 °C. Images were taken with Axio Imager.Z1 (Carl Zeiss), and 'vessel-covered area' as well as 'number of branching points' were determined with AngioTool 0.5a 55 .
www.nature.com/scientificreports www.nature.com/scientificreports/ Fluorescence microlymphangiography. To study the functionality of the lymphatics we used ED16.5-18.5 mouse embryos and injected 50-100 nl of 2000 kDa FITC-dextran (Sigma, Taufkirchen, DE; 8 mg/ml PBS) into the paws of control embryos and the most distal part of the legs of the homozygous Wnt-null embryos (n = 18). Embryos were studied with a fluorescence stereomicroscope (Leica Microsystems, Wetzlar, DE). Images were taken at regular intervals up to 15 min after injection.
Cell culture. We purchased several Human Dermal Lymphatic Endothelial Cells (HD-LEC) from PromoCell (Heidelberg, DE) and cultured them according to the supplier's instructions. Purity of the cells was controlled by CD31 and Prox1 immunostaining, and the three cell lines, which presented as almost 100% pure, were used 56 . All experiments were performed on these 3 cell-lines separately and the results pooled for statistical analyses.
In vitro cell proliferation. In vitro proliferation assays were performed as described previously 57 . 5 × 10 3 cells per well were seeded in 96-well culture plates with 100 µl culture medium. After 12 hours (h), treatment was applied to the cultures and the controls (t = 0) were fixed. Subsequently cells were fixed at 24 h intervals, stained with crystal violet and analyzed photometrically at 570 nm.
Cell migration/scratch assay. HD-LEC were seeded into 24-well-plates and grown to confluence. For better documentation, cells were labeled with 1.5 µM CellTracker ™ Green CMFDA (Thermo Fischer Scientific, Waltham, US) according to the manufacturer's instructions, before scratching. Scratches were made with a 100 µl pipette tip and fresh medium with test substances or solvents (control) was added. Photos were taken after with Leica DMI6000 B (Leica Microsystems) after 0 and 24 h. The gap closure was measured with GNU Octave, version 4.0.3 (https://www.gnu.org/software/octave/) and plotted in % (100% = complete closure).
Tube formation assay. Assays were performed on ibidi µ-Slide Angiogenesis slides (Ibidi, Martinsried, DE) and the protocol provided with the slides was optimized for LECs. 10 µl Matrigel (Corning Inc., Corning, US) were used as coating gel-matrix and allowed to polymerize overnight at 37 °C. 5,000 cells were seeded in each well.
The cells were pre-treated with LGK974 for 3 d and labeled with CellTracker ™ Green CMFDA (1.5 µM; Thermo Fischer Scientific) before the experiment. After 9 h, formation of networks was documented with Leica DMI6000 B (Leica Microsystems). Quantification of networks was performed with WimTube analysis software (Onimagin Technologies SCA, Córdoba, Spain).