Ancestral Vascular Lumen Formation via Basal Cell Surfaces

The cardiovascular system of bilaterians developed from a common ancestor. However, no endothelial cells exist in invertebrates demonstrating that primitive cardiovascular tubes do not require this vertebrate-specific cell type in order to form. This raises the question of how cardiovascular tubes form in invertebrates? Here we discovered that in the invertebrate cephalochordate amphioxus, the basement membranes of endoderm and mesoderm line the lumen of the major vessels, namely aorta and heart. During amphioxus development a laminin-containing extracellular matrix (ECM) was found to fill the space between the basal cell surfaces of endoderm and mesoderm along their anterior-posterior (A-P) axes. Blood cells appear in this ECM-filled tubular space, coincident with the development of a vascular lumen. To get insight into the underlying cellular mechanism, we induced vessels in vitro with a cell polarity similar to the vessels of amphioxus. We show that basal cell surfaces can form a vascular lumen filled with ECM, and that phagocytotic blood cells can clear this luminal ECM to generate a patent vascular lumen. Therefore, our experiments suggest a mechanism of blood vessel formation via basal cell surfaces in amphioxus and possibly in other invertebrates that do not have any endothelial cells. In addition, a comparison between amphioxus and mouse shows that endothelial cells physically separate the basement membranes from the vascular lumen, suggesting that endothelial cells create cardiovascular tubes with a cell polarity of epithelial tubes in vertebrates and mammals.


Introduction
It has been suggested that the cardiovascular system of bilaterians evolved from a common ancestor [1][2][3]. This is because the heart and major blood vessels develop as tubes along the anterior-posterior (A-P) axes in both vertebrates and invertebrates [4,5]. In addition, several genes have been identified in both vertebrates and invertebrates that have similar expression domains and functions in cardiovascular development. For example, the homeotic gene tinman and its homologue NKX2.5 are expressed in cardiac mesoderm in Drosophila and mouse, respectively, and these genes are required for proper cardiac development in both animals [1,5]. Despite conservation of several genes involved in cardiovascular development, new features evolved in the vertebrates. For example, in vertebrates, endothelial cells line the lumen of the heart and of all blood vessels [6]. In contrast, in invertebrates, endothelial cells either are not present or do not form a continuous vascular wall [3], showing that endothelial cells are not a conserved feature of cardiovascular tubes. Therefore, in order to understand the ancestral and conserved part of cardiovascular tube formation, we investigated developing vessels in the invertebrate amphioxus and compared these vessels with the homologous ones in mouse. We used the cephalochordate amphioxus, Branchiostoma lanceolatum, because its body plan is similar to the one of vertebrates [7,8], and because it forms a monophyletic clade with vertebrates and urochordates [9].

Localization of basement membrane in amphioxus and mouse vessels
As shown in Fig. 1, both the amphioxus aorta ( Fig. 1A-1C, asterisk) and the larger mouse aorta (Fig. 1D-1F, asterisk) are located on top of, or dorsal to, the adjacent intestine. Based on previous morphological studies, an electron dense layer, morphologically similar to a basement membrane, has been observed on the luminal side of invertebrate blood vessels [3,[10][11][12]. However, it remained to be shown whether invertebrate vessels were lined by a real basement membrane or, alternatively, by an apical ECM that is morphologically similar, but molecularly different from a basement membrane [13][14][15][16].
To clarify this issue, we localized the basement membrane protein laminin on transverse sections through amphioxus ( Fig. 1A-1C). As shown in Fig. 1B, this basement membrane protein (shown in red color) surrounds the lumen of the aorta in amphioxus (asterisk). By contrast, in mouse, endothelial cells (shown in green color in Fig. 1D and 1E) surround the aortic lumen (asterisk in Fig. 1E) and separate it from the laminincontaining basement membrane (shown in red color in Fig. 1E).
We also localized acetylated tubulin (Ac tub) as an apical marker that labels primary cilia in amphioxus (as shown in green color in Fig. 1A and 1B). We found that this apical marker was exclusively localized to the surface of the intestinal epithelium that faced the gut lumen (lower part of Fig. 1A and 1B). Since we observed laminin, but not acetylated tubulin, in the aortic lumen (asterisk in Fig. 1B), we conclude that the intestinal epithelium is polarized. It faces the gut lumen with its apical cell surface harboring cilia (green color in Fig. 1B) and the aortic lumen with its basal cell surface and basement membrane (arrow in Fig. 1B). Therefore, even though in both animals the aortae form on the basement membrane of the intestinal epithelium ( Fig. 1B and 1E), endothelial cells introduced a change in mouse. In amphioxus, the basement membrane of the intestinal epithelium directly lines the vascular lumen (arrow in Fig. 1B and 1C), whereas in mouse, endothelial cells (shown in green color in Fig. 1E) separate the intestinal basement membrane (arrowhead in Fig. 1E) from the aortic lumen (asterisk in Fig. 1E).

Induction of vessels filled with luminal ECM
Since a laminin-containing basement membrane lines the vessels in amphioxus, we asked whether Matrigel, a reconstituted basement membrane, or basal ECM, was able to induce similar kinds of vessels in vitro (Fig. 2). Matrigel induces tube-like structures in several cell types, including endothelial cells and smooth muscle cells [17,18], and we used an immortalized endothelial cell line, Mile Sven 1 (MS1), for most experiments described in this study. A branched network of vascular tubes formed in 24-48 hrs after Matrigel overlay ( Fig. 2A). The average length of MS1 tubes between intersections was 112647 mm, and the average lumen width was 3.862.5 mm (n = 20). As shown by light microscopy ( Fig. 2A to 2C) and electron microscopy (Fig. 2D), the Matrigelinduced multicellular vessels had a visible lumen (asterisks). Importantly, the luminal cell surface was relatively smooth (Fig. 2D), indicative of a basal cell surface, whereas the abluminal cell surface possessed microvilli (open arrowheads), indicative of an apical cell surface. Finally, we detected an electron-dense material inside the vessel lumen (asterisk in Fig. 2D), which resembled the electron-dense material observed in developing vessels in amphioxus [19]. Therefore, our data show that MS1 vascular tubes induced by Matrigel overlay do not reflect vertebrate blood vessels, but instead resemble the vessels observed in invertebrates, such as in amphioxus.

Molecular composition of the luminal ECM in vessels formed by MS1 cells
Next we characterized the composition of the luminal ECM localized inside vessels formed by MS1 cells after Matrigel overlay. Laminin-a1 chain is a basement membrane protein present in Matrigel [20], and endothelial cells do not produce it [21]. In contrast, endothelial cells produce laminin-a4 and -a5 chains, which are not present in Matrigel [21,22]. Using laminin-a-chain specific antibodies, we found that the laminin-a1 chain was uniformly present within the vascular lumen (Fig. 3A), whereas the endothelial cell-produced laminin-a4 and -a5 chains were found close to their luminal cell surface ( Fig. 3B and 3C). These data show that Matrigel fills the vascular lumen and suggest that it induces MS1 cells to deposit basement membrane proteins towards the developing vascular lumen. A similar situation was found after overlaying human umbilical vein endothelial cells (HUVEC) with Matrigel ( Fig. 3D-3F). Based on the localization of basement membrane proteins, the resulting vessels molecularly resemble the vessels observed in amphioxus (compare Fig. 3B and 3C with Fig. 4E).

Lumen formation via basal cell surfaces in Matrigel
Our data suggest that basal cell surfaces and their ECM are the ancestral components of blood vessels. To provide evidence for vascular lumen formation via basal cell surfaces, we analyzed the localization of apical and basal cell surface markers in vascular tubes formed by MS1 cells. In most epithelial tubes (or cysts), b1integrin is preferentially localized to the abluminal or basal cell surface next to the basement membrane [23], whereas podocalyxin (gp135) is exclusively localized to the luminal or apical cell surface [24].
In stark contrast to the localization of apical and basal proteins in epithelial tubes, we observed strong b1-integrin expression on the luminal cell surface (arrow in Fig. 3I) rather than on the abluminal cell surface of MS1 cells (arrowhead in Fig. 3I). Conversely, podocalyxin (gp135) was expressed on the abluminal cell surface (arrowhead in Fig. 3J and 3L) rather than on the luminal cell surface (arrow in Fig. 3J and 3L). Given the abundance of b1-integrin on the basal cell surface of tube-forming cells (Fig. 3I), we asked whether integrins are involved in this type of tube formation. Thus, we applied a function-blocking antibody to interfere with b1-integrin signaling and quantified the numbers of open and closed tubes after 2 days of culture. In assays of MS1 cells treated with b1-integrin function blocking antibody, the frequency of open tubes ( Fig. S1) was four times higher compared with control antibody treatment (Fig. S1). These data show that b1-integrin is involved in tube formation of MS1 cells upon Matrigel overlay.
These findings demonstrate that the basal cell surface of MS1 cells is capable of forming a vascular lumen. This atypical tube formation may have been due to unique properties of MS1 endothelial cells used in this assay. Therefore, we performed the Matrigel overlay assay with primary human umbilical vein endothelial cells (HUVEC), which are frequently used for vascular tube formation studies [25,26]. Similar to MS1 cells, HUVEC formed tubes, which contained both exogenous (Fig. 3D) and endogenous ( Fig. 3E) laminins inside the lumen. In addition, the luminal plasma membrane of HUVEC had a more intense b1integrin expression (arrows in Fig. 3F) compared to the abluminal membrane (arrowheads in Fig. 3F).
To test whether Matrigel as a basal ECM was responsible for inducing tubes via basal cell surfaces, we used an assay in which HUVEC cells form lumenized vascular sprouts in fibrin gel [25]. When we analyzed the localization of laminin in these sprouts we found it on the abluminal cell surface (arrowhead in Fig. S2A, B). The lumen (asterisk in Fig. S2A and S2B) as well as the luminal cell surface (arrows in Fig. S2A) was laminin-free. In addition, we performed an EM analysis of HUVEC vascular tubes formed in fibrin gel and we detected a basement membrane on the abluminal surface of HUVEC ( Fig. S2C and S2D). Therefore, these results indicate that Matrigel ECM, rather than fibrin, can induce endothelial cells to form tubes with intraluminal ECM.

Blood cells-mediated removal of luminal ECM
Blood vessels of early amphioxus larvae are filled with an electron-dense material [19,27]. In adult amphioxus, however, this material has disappeared [19]. Since blood cells in amphioxus are phagocytotic and share functional similarities to macrophages in vertebrates [28,29], we investigated whether an ECM-free lumen coincided with the localization of blood cells. To this end, we focused on the subintestinal vessel, which is the heart and the largest vessel in this animal [30]. As shown in Fig. 4A to 4D, the subintestinal vascular lumen contained areas completely filled with laminin (asterisks in Fig. 4A and 4C) as well as areas, which were only lined by laminin ( Fig. 4B and 4D). Importantly, the localization of blood cells (B) inside the subintestinal vessel coincided with the presence of a patent vascular lumen (dashed bracket in Fig. 4E and 4F). In contrast, areas with no blood cells were filled with laminincontaining ECM (small bracket in Fig. 4E and 4F). To test whether phagocytotic cells were able to remove luminal ECM, Matrigel overlay was used to induce HUVEC vessels in the presence of the monocyte/macrophage cell line THP-1 (labeled with ''M'' in Fig. 4G-4J). Following this co-culture we detected an absence of luminal ECM in regions, where these blood cells were incorporated into the vascular lumen (Fig. 4G, and a cross-section through a vascular lumen in Fig. 4H-4J). We also detected cleared areas of lumen, which these cells appeared to have left (data not shown). There are several possibilities how macrophages can remove basal ECM: phagocytosis, extracellular proteolysis, or both phagocytosis and extracellular proteolysis. To determine the mechanism of macrophage-mediated Matrigel removal we used DQ collagen IV as a substrate that becomes fluorescent upon proteolytic cleavage. Collagen IV is a component of Matrigel and can be detected intraluminaly in tubes formed by HUVEC (Fig.  S3A, S3B). We incubated THP-1 monocyte/macrophage cells with Matrigel containing DQ collagen IV. After 24 hours we analyzed the cells microscopically and we observed the fluorescent cleavage product inside THP-1 cells (arrows in Fig. S3C, S3D) as well as extracellularly around THP-1 cells (arrowheads in Fig.  S3C, S3D). This finding demonstrates that THP-1 macrophages digest the Matrigel by extracellular proteolysis as well as by phagocytosis. The results therefore suggest that phagocytotic blood cells are involved in the formation of a patent vascular lumen during vessel formation via basal cell surfaces.

Discussion
Our study provides the first model of vessel formation in amphioxus and possibly in other invertebrates (Fig. 5A and 5B). Invertebrates account for more than 99% of living species, and a cardiovascular system already occurred in a bilaterian ancestor, most likely prior to the divergence of deuterostomes and protostomes [1]. The finding that the Drosophila heart is induced by expression of a homeotic gene (tinman), which was later found to have a homolog (NKX2.5) with a similar role in mouse, supports the notion that the cardiovascular system of most bilaterians developed from a common ancestor. To understand the ancestral mechanism of vessel formation, we investigated the invertebrate cephalochordate amphioxus and compared its developing vessels with the homologous ones in mouse (Fig. 5). We showed that vessel formation is initiated between the basal cell surfaces of endoderm and mesoderm (Fig. 5A). For instance, the heart, or subintestinal vessel, develops between the basal cell surfaces of the intestinal epithelium and the mesothelium that face with their apical cell surfaces the gut tube (upper part of Fig. 5A) and the coelomic cavity (lower part of Fig. 5A), respectively. Therefore, the vascular lumen must form between the basal cell surfaces of these two polarized epithelial cell layers. The vessels are initially filled with a laminin-containing ECM that we term ''basal ECM'' (Fig. 5A) to distinguish it from apical ECM.
Our in vitro experiments show that Matrigel, a reconstituted basement membrane or basal ECM, is sufficient for inducing vessel formation via basal cell surfaces. To our knowledge, this is the first demonstration that the lumen of multicellular tubes can form via basal cell surfaces with the apical cell surface oriented abluminally. Since the endoderm is the first polarized epithelium to form inside the developing embryo [31], its basement membrane may be the initial site of vessel formation. Subsequently, mesoderm-derived cells contribute to vessel formation and provide basal ECM and blood cells. In Drosophila, most blood cells are phagocytotic and produce as well as degrade ECM [2,29]. In support of a role of blood cells in the formation of a vascular lumen, incorporation of a macrophage/monocyte cell line into developing Matrigel-induced vessels led to the development of an ECM-free vascular lumen (asterisk in Fig. 5B). Recently, two reports were published on the mechanism of lumen formation in the dorsal vessel of Drosophila [32,33] that reveals similarities as well as differences compared to the mechanism of vascular tube formation in amphioxus that we described here. In Drosophila, the lumen is formed via repulsion of distinct plasma membrane domains having basal characteristics, showing that the dorsal vessel also forms via basal cell surfaces. In contrast to amphioxus, however, where the abluminal plasma membrane is apical, the abluminal cardioblast cell surface in Drosophila also has a basal character.
Interestingly, the mechanism of vessel formation via basal cell surfaces in invertebrates may also help to explain vasculogenic mimicry by which certain malignant cancer cells form primitive vascular channels within tumor tissue. Similar to the invertebrate blood vessels, laminin-containing ECM lines these vessels within the tumor [34]. Therefore, our findings warrant further research to find out whether cancer cells resume an ancestral mechanism of vessel formation as we described here (Fig. 5A and 5B).
Similar to amphioxus, blood vessel formation in mouse often happens on, or between, basal cell surfaces. However, in the case of the heart and aortae, mesoderm-derived endothelial cells develop between the basal cell surfaces of endoderm and mesoderm (Fig. 5C). When these vertebrate-specific cells form a vessel, they separate the basement membranes from the developing vascular lumen (Fig. 5D). Therefore, our data suggest that the evolution of endothelial cells in vertebrates led to the creation of blood vessels with a morphology and cell polarity similar to epithelial tubes (compare Fig. 5D with Fig. 5B). Moreover, it is likely that this event may have paved the way for a more modular cardiovascular and hematopoietic system in vertebrates and mammals.

Vascular tube formation assays with Matrigel
A modification of a previously published assay [17] was used. In brief, MS1 cells or HUVEC were plated either on 8-well chamber slides (BD Falcon) or on 35 mm glass-in-bottom Petri dishes (MaTek Corporation). After 6-8 hrs, the cells were washed with appropriate serum-free media and overlaid with Matrigel TM diluted 1:5 in serum-free media. After gel formation over 30 min at 37uC in the incubator, the gels were overlaid with 0.5 ml of culture media (in an 8-well chambered slide) or 3 ml of culture media (in a Petri dish). For MS1 cells the culture medium was supplemented with 40 ng/ml VEGF and 40 ng/ml FGF-b.

Electron microscopy
Cells were fixed with 2.5% glutaraldehyde in 0.1 M Nacacodylate buffer for 1 h. Subsequently, cells were washed and stained with OsO 4 and uranyl acetate, dehydrated in graded solutions of ethanol, and embedded in Epon resin. Ultrathin sections (70 nm) were contrasted by using lead citrate and uranyl

Immunocytochemistry and immunohistochemistry
Cells and tissues were fixed with 4% paraformaldehyde in PBS (pH = 7.4) and used for immunocytochemistry and -histochemistry as recently described [36]. In brief, unspecific binding was blocked using 1% BSA, 5% normal serum, 0.1% Tween 20 or Triton-X-100 in PBS ( = blocking buffer) for 1 h. Primary antibodies were applied overnight at 4uC and, after washing, secondary antibodies were applied for 1 h at room temperature. Both antibodies were diluted in blocking buffer. Confocal imaging was performed using a LSM 510 laser scanning confocal system equipped with LSM 510 software (Zeiss). Plan-Apochromat 636/ 1.4 Oil DIC, 406/ 1.2 Apo W Corr and 106/ 0.3NA objectives (Zeiss) were used.

Immunohistochemical analysis of amphioxus
Juvenile specimens of amphioxus (Branchiostoma lanceolatum) fixed in formalin were obtained from Biologische Anstalt Helgoland of the Alfred-Wegener-Institute. Specimens were processed for immunohistochemistry on cryosections as described above.
Whole-mount immunolabeling was performed on 14 day-old larvae fixed in 4% paraformaldehyde and preserved in 70% ethanol. Larvae were rehydrated and permeabilized using 0.2% Triton-X-100. Incubations with antibodies were performed for 24 hrs at 37uC.

Vascular tube formation assay in co-culture with THP-1 cells
Vessel formation by HUVEC was induced in 24-well plates with coverslips at the bottom as described above, but with the following modification: THP-1 cells were mixed with Matrigel (10.000 cells/ well); HUVEC cells were overlaid with 180 ml of this mixture. Cocultures were kept for 72 hrs, fixed and labeled for laminin a-1 chain and actin.
Blocking of b1-integrin in Matrigel overlay tube formation assay MS1 cells were plated into 24-well plates that contained glass coverslips at the bottom of each well. After 10 hrs the cells were pretreated for 30 min with blocking or control antibody at a concentration 10 mg/ml in culture media or with media alone. Cells were washed with serum-free media and were overlaid with Matrigel TM diluted 1:5 in serum-free media without or with antibodies to obtain 10 mg/ml concentration. After gel formation over 30 min at 37uC in the incubator, the gels were overlaid with media without or with antibodies (10 mg/ml). After 48 hrs of culture, the cells were fixed and stained for CD31 to label the endothelial plasma membrane. Confocal z-stacks of tubes (n = 9-12 per assay from 3 different assays) were examined for the presence of openings. Tubes containing openings bigger than 1 mm were quantified as open tubes. Statistical difference was determined using Student's t-test.

Vascular tube formation assay using HUVEC in fibrin gel
A modification of an already published assay was used [25]. HUVEC on collagen-coated beads were embedded in the fibrin gel in 35 mm glass-in-bottom Petri dishes (MaTek Corporation). Human skin fibroblasts (100,000/ dish) were added on top of the gel in 3 ml of full EGM-2 media. After 12 days the lumenized vascular sprouts formed and the cells were processed for immunocytochemistry and electron microscopy. Visualization of matrix proteolysis by THP-1 cells using DQ TM collagen IV THP-1 cells were mixed with Matrigel diluted 1:5 with EBM-2 media containing 20 mg/ml DQ TM collagen IV, fluorescein conjugate (Invitrogen) and the 200 ml of the mixture was pipetted into glass-in-bottom Petri dishes. After 30 min at 37uC Matrigel gelled and 3 ml of EGM-2 media were added to each dish. After 24 hrs the cells were fixed, stained with phalloidine-rhodamine and DAPI.