Blood vascular cells and lymphatic endothelial cells (BECs and LECs, respectively) form two separate vascular systems and are functionally distinct cell types or lineages with characteristic gene expression profiles. Interconversion between these cell types has not been reported. Here, we show that in conventional in vitro angiogenesis assays, human BECs of fetal or adult origin show altered gene expression that is indicative of transition to a lymphatic-like phenotype. This change occurs in BECs undergoing tubulogenesis in fibrin, collagen or Matrigel assays, but is independent of tube formation per se, because it is not inhibited by a metalloproteinase inhibitor that blocks tubulogenesis. It is also reversible, since cells removed from 3D tubules revert to a BEC expression profile upon monolayer culture. Induction of the lymphatic-like phenotype is partially inhibited by co-culture of HUVECs with perivascular cells. These data reveal an unexpected plasticity in endothelial phenotype, which is regulated by contact with the ECM environment and/or cues from supporting cells.
Higher animals possess two separate circulatory networks, the blood and lymphatic vascular systems, which have distinct but interdependent functions. In the developing mammalian embryo, the lymphatic vasculature appears after the formation of the blood vasculature and is believed to be of venous origin. In fact, blood vascular cells and lymphatic endothelial cells (BECs and LECs, respectively) share a common progenitor (Jain and Padera, 2003; Kubo and Alitalo, 2003). A subset of cells in the embryonic cardinal vein transdifferentiate and take on a lymphatic identity via the stepwise expression of lymphatic regulatory genes (Oliver and Harvey, 2002; Wigle et al., 2002). The Sox18 transcription factor specifies target cells in which expression of the Prox-1 transcription factor is induced (Francois et al., 2008), which, in turn, commits cells to differentiation into the lymphatic lineage such that this subset of cells gives rise to the entire lymphatic vascular network. This view is supported by data from mouse knockout studies, which show that the absence of Prox-1 expression leads to a failure to differentiate into the lymphatic endothelial cell lineage (Johnson et al., 2008). Once the two endothelial cells lineages are established, BECs and LECs are considered to be mature, non-interchangeable cell types that display characteristic profiles of gene expression (Hirakawa et al., 2003; Kriehuber et al., 2001; Podgrabinska et al., 2002).
In this study, we have identified an unexpected plasticity in the phenotype of human endothelial cells when analysed in several conventional in vitro models of angiogenesis. We show that blood vascular endothelial cells are able to adopt an expression profile resembling that of lymphatic endothelial cells in response to particular in vitro ECM environments, and this behaviour is reversible and independent of tubulogenesis. These studies shed new light on the environmental cues that regulate endothelial cell differentiation and lineage maintenance.
Human umbilical vein endothelial cells adopt a ‘lymphatic’ gene expression profile when cultured in 3D matrices
Endothelial cells form tube-like structures when cultured in 3D matrix environments and stimulated with pro-angiogenic growth factors (Fig. 1A–C). We performed a microarray analysis to compare gene expression in human umbilical vein endothelial cells (HUVECs) undergoing tube formation in a 3D fibrin matrix (Fig. 1B) with monolayer cultures (Fig. 1A). This analysis identified 169 genes that were upregulated and 554 genes that were downregulated during tubulogenesis (supplementary material Table S1). Unexpectedly, we observed altered expression of a number of genes that had been previously identified to be mainly or exclusively expressed in either BECs or LECs (referred to henceforth as BEC or LEC markers). A number of marker genes were examined further using quantitative real-time RT-PCR (qRT-PCR). These included the LEC markers: lymphatic vascular endothelial hyaluronan receptor (LYVE1) (Banerji et al., 1999); PROX1 (Hong et al., 2002; Wigle and Oliver, 1999); vascular endothelial growth factor-3 (FTL4) (Adams and Alitalo, 2007; Kaipainen et al., 1995); integrin α9 (ITGA9) (Huang et al., 2000; Mishima et al., 2007) and podoplanin (PDPN) (Hirakawa et al., 2003; Kriehuber et al., 2001; Petrova et al., 2002; Schacht et al., 2003). The BEC markers selected were: Lamin-B1 (LMNB1) (Podgrabinska et al., 2002), VEGFC (Hirakawa et al., 2003; Kriehuber et al., 2001; Makinen et al., 2001) and CD44 (Cao et al., 2006; Hirakawa et al., 2003).
Expression profiles of the markers were assessed in HUVECs undergoing tubulogenesis in either type I collagen or fibrin 3D angiogenesis assays, compared with HUVECs in standard monolayer culture conditions on plates coated with type I collagen. In both types of 3D matrix environment, the LEC markers LYVE1, PROX1 and FLT4 were upregulated, and the BEC markers LMNB1, VEGFC and CD44 were downregulated compared with monolayer culture (Fig. 1D–G). Another LEC marker, ITGA9, was upregulated in 3D collagen cultures, whereas no significant changes in expression were detected in 3D fibrin gels (data not shown). In contrast to the other LEC markers, expression of PDPN was not detected in either assay. The mRNA expression data were supported by the appearance of LYVE-1 protein consistently throughout the entire tubular network in HUVECs undergoing tube formation in 3D collagen gels, whereas only a few isolated cells were weakly positive for LYVE-1 in HUVECs cultured as a monolayer (Fig. 1H and supplementary material Fig. S1). We also observed VEGFR3 and nuclear Prox-1 staining in HUVECs that formed tubes (supplementary material Figs S2 and S5), and detected downregulation of Lamin-B1 by western blot (supplementary material Fig. S3). These data demonstrate that mature blood vascular endothelial cells alter their gene expression profiles upon culture in 3D matrices, where they transdifferentiate to a ‘lymphatic-like’ phenotype.
Transdifferentiation is not restricted to endothelial cells of fetal origin
We tested whether endothelial cells derived from other tissue sources showed gene expression changes in 3D culture. Committed adult human aortic endothelial cells (HAECs) showed similar changes in BEC or LEC marker gene expression during tubulogenesis in type I collagen as those observed in HUVECs (Fig. 2A,B). Once again, expression of the LEC marker podoplanin was not induced. Similar data were obtained with adult human dermal microvascular endothelial cells (HDMECs; data not shown), with the exception that podoplanin was expressed by these cells. However, data from these cells must be considered with caution because HDMEC preparations are known to contain variable proportions of LECs and BECs (Makinen et al., 2001). The finding that mature adult blood vascular endothelial cells are capable of adopting a lymphatic-like gene expression profile indicates that this plasticity is a property of many endothelial cell types.
Adoption of a ‘lymphatic-like’ phenotype is not dependent on a 3D environment
Induction of the lymphatic-like phenotype occurred in 3D culture in both matrix types tested, but not in cells in monolayer. We therefore investigated whether a 3D environment is essential for this process using the 2D Matrigel assay. In this assay, endothelial cells seeded on top of Matrigel migrated and aligned to form a network of cord-like structures within 24 hours (Fig. 2C). Analysis of expression of BEC and LEC marker genes showed that, similarly to the 3D tubulogenesis assays, HUVECs had upregulated expression of mRNA encoding LYVE-1, Prox-1, VEGFR3 (and integrin α9, data not shown) paralleled by downregulation of mRNA encoding lamin-B1, VEGF-C and CD44 over the course of network formation (Fig. 2D–E). Comparison of BEC and LEC marker expression in HUVECs cultured within 3D Matrigel versus cells seeded on Matrigel revealed similar patterns; indeed, induction of Prox-1 was strongest in cells forming alignments on top of this matrix. These data suggest that a 3D environment per se is not obligatory for induction of the ‘lymphatic’ phenotype, but the nature of the ECM is critically important.
Based on the results obtained when endothelial cells were grown on Matrigel, we next asked whether similar results could be obtained with cells grown in monolayer on top of collagen or fibrin gels. We observed that when seeded on top of collagen gels, HUVECs formed alignments in a manner comparable with cells seeded on Matrigel (Fig. 3A). By contrast, cells seeded on fibrin gels retained a cobblestone appearance resembling that of cells in standard culture (Fig. 3B,C). Expression analysis showed that cells seeded on collagen gels adopted the ‘LEC’ profile of gene expression, whereas cells seeded on fibrin retained the same profile as the standard 2D cells (Fig. 3D,E). The ‘LEC’ phenotype was induced more effectively when the cells were cultured in 3D collagen rather than on top of the same gel matrix (data not shown). These data provide confirmation that a 3D environment is not essential for the BEC to LEC transition, but they emphasise that the nature of the matrix is critical, because although monolayer growth on a dense collagen gel is permissive for the phenotypic change, growth on fibrin is not.
Adoption of a ‘lymphatic-like’ phenotype is independent of tubulogenesis and is reversible
We asked whether the ‘LEC’ phenotype was functionally linked with tubulogenesis by analysing marker expression in HUVECs cultured in 3D collagen in the presence of the broad-spectrum matrix metalloproteinase inhibitor BB-94 (Fig. 4A–D). This inhibitor blocks tube formation by inhibition of key enzymes required for ECM digestion and remodelling, such as MT1-MMP (Hiraoka et al., 1998; Lafleur et al., 2002). We observed that upregulation of expression of LEC markers and downregulation of BEC markers was largely unaffected in cultures in which tubulogenesis was inhibited, although a reduction in the level of FLT4 induction was detected. These data show that HUVECs undergo the transition to a ‘lymphatic’ phenotype when cultured in conditions conducive to tube formation, even when the tubulogenesis is inhibited.
We also tested whether induction of the ‘LEC’ expression profile might be reversed after transfer of the cells from a 3D environment back to monolayer culture. HUVEC were suspended within 3D collagen gels for 24 hours and then the gels were digested with collagenase and the cells returned to monolayer cultures and maintained for a further 48 hours. As previously observed, 3D culture led to induction of expression of the LEC markers LYVE1, PROX1 and FLT4, and the downregulation of the BEC markers LMNB1 and VEGFC. Following replating in monolayer culture, expression of each of these markers demonstrated full reversibility, returning to levels that were comparable with those observed in control monolayer samples at the assay end point (Fig. 4E,F). These data demonstrate that the pattern of BEC and LEC marker gene expression is directly regulated by the extracellular environment of the cells.
Adoption of the ‘lymphatic-like’ phenotype is partially inhibited by co-culture with perivascular cells
Blood vessels in vivo have an extensive basal lamina and are in close contact with smooth muscle cells or pericytes, whereas lymphatic vessels have little basement membrane and largely lack pericyte coverage (Pepper and Skobe, 2003). We hypothesised that the culture of isolated BECs in direct contact with the ECM in the absence of perivascular cells might create a microenvironment more similar to that associated with lymphatic vessels and thus encourage the cells to take on LEC-type behaviour; conversely, co-culture of HUVECs with perivascular cells might favour the maintenance of the ‘BEC’ phenotype. We therefore compared marker gene expression levels in HUVECs cultured in 3D collagen gels either as monocultures or co-cultured with murine perivascular cells (Brachvogel et al., 2005; Brachvogel et al., 2007). By using species-specific qRT-PCR probes we were able to determine expression levels in the human cells only. HUVECs cultured in the presence of perivascular cells showed markedly reduced expression of mRNA encoding LYVE-1 and Prox-1 compared with monocultures, whereas VEGFC mRNA expression increased (Fig. 5A,B). These results demonstrate a reversal of the expression of several BEC and LEC marker genes compared with the pattern seen upon monoculture in 3D matrices, such that induction of the ‘lymphatic’ phenotype was at least partially inhibited. LYVE-1 protein was clearly expressed throughout the tubular networks in HUVEC monocultures, and strongly downregulated in HUVEC and perivascular cell co-cultures (Fig. 5A,B; supplementary material Fig. S4). However, not all the BEC and LEC markers followed the same pattern, because co-culture with perivascular cells resulted in increased expression of FLT4 (Fig. 5B; supplementary material Fig. S5), and decreased expression of LMNB1 (Fig. 5B) compared with levels in HUVEC monocultures, suggesting that these genes respond to the matrix environment rather than the presence of pericytes. Interestingly, we saw a similar overall trend in the effects on BEC and LEC marker expression (namely, decreased LYVE1 and increased VEGFC and FLT4) when we used siRNA to reduce induction of PROX1 expression in endothelial cells seeded on Matrigel (supplementary material Fig. S6).
The culture of endothelial cells within or on top of 3D matrices such as type I collagen, fibrin and Matrigel provides simple, popular models of angiogenesis (Auerbach et al., 2003; Goodwin, 2007; Phung and Dass, 2006; Ucuzian and Greisler, 2007). The present work reports that when blood vascular endothelial cells such as HUVECs and HAECs are cultured under such conditions, they show gene expression changes that are characteristic of a shift to a lymphatic phenotype. The lymphatic-like phenotype was independent of tubulogenesis and fully reversible upon transfer of cells from 3D to monolayer culture on collagen-coated tissue culture plastic, suggesting that interactions between endothelial cells and matrix account for the alterations in gene expression observed. This was further reinforced by the observation that HUVECs adopt the lymphatic-like phenotype when cultured on top of Matrigel or a dense collagen gel, but not when grown on top of a fibrin gel. Also, co-culture of HUVECs with pericytes partially prevented the development of the LEC-like phenotype. These findings indicate that the concept of LECs and BECs as committed, differentiated lineages needs to be revised, and demonstrate the importance of contact with different extracellular matrices and perivascular cells in regulating endothelial cell phenotype and lineage specification.
Endothelial cells of different tissue origins represent different functional cell types, with characteristic gene expression profiles (Chi et al., 2003). We considered the possibility that, because HUVECs are of fetal origin, they might show increased phenotypic plasticity compared with adult endothelial cells. However, blood vascular cells of adult origin such as HAECs and HDMECs showed the same induced changes in gene expression, indicating that ability to adopt a lymphatic-like phenotype is a common property of cultured BECs. We also considered it possible that the HUVECs used in our initial experiments might contain a subset of LECs, which came to be the dominant cell type in the culture conditions we used. However, lymphatic vessels are absent from the umbilical cord, and moreover, the lack of expression of podoplanin in all our angiogenesis models using HUVECs and HAECs confirms the absence of a fully differentiated LEC subpopulation. By contrast, HDMECs that are known to contain an LEC subpopulation (Makinen et al., 2001) did indeed express podoplanin. Rather, these data, and in particular the induction of expression of Prox-1, imply an ability of BECs in 3D culture as a whole to undergo a type of transdifferentiation in which cells adopt a more lymphatic-like phenotype. But it is also clear from the absence of podoplanin expression in HUVECs undergoing 3D tubulogenesis that the cells do not undergo full transdifferentiation from a BEC to an LEC phenotype. This might imply that the cells adopt an intermediate phenotype that shows characteristics of both LECs and BECs, or it might reflect the fact that podoplanin expression is a late marker of lymphangiogenesis, which is consistent with the proposed role of podoplanin in later stages of lymphatic vascular patterning (Schacht et al., 2003). It is possible that HUVECs commence BEC-to-LEC transdifferentiation in the 3D culture models, but the process is not complete in the timeframe of the assays.
Adoption of the ‘LEC’ phenotype could be correlated with the tendency of the endothelial cells to form tube or cord structures. However, our findings show that induction of the ‘LEC’ phenotype is neither a consequence of tubulogenesis per se, nor dependent upon it. Rather, the nature of the cell's extracellular environment determines the induction of the ‘lymphatic’ phenotype. Importantly, the phenotype is reversible upon returning cells from 3D to standard monolayer culture. This reversibility shows that the ‘LEC’ expression profile is not the result of simple culture-induced de-differentiation, a possibility that needs to be considered because cultured endothelial cells are known to lose their lineage specificity (Amatschek et al., 2007), and low-level LYVE-1 expression has been reported in late-passage HUVECs (Banerji et al., 1999).
One of the most interesting observations from the present study is that inclusion of pericytes in the HUVEC tubulogenesis cultures blocked the changes in expression of key LEC and BEC markers such as LYVE-1, Prox-1 and VEGF-C. This indicates that signals provided by perivascular supporting cells help to reinforce the BEC phenotype. It is interesting that co-culture of HUVECs with perivascular cells resulted in a largely similar effect on marker expression to that observed in cells following knockdown of Prox-1. This suggests that signals elicited by interactions between endothelial cells and pericytes act via Prox-1 to regulate endothelial cell lineage specification. The converse might also apply, namely that Prox-1 expression in endothelial cells influences endothelial cell to pericyte communication. Supporting this notion, ectopic expression of Prox-1 in isolated blood vascular endothelial cells has been shown to convert the transcriptional programme of the cells toward the lymphatic phenotype by upregulating expression of LEC-specific marker genes, as well as downregulating expression of a large number of BEC-specific transcripts (Mishima et al., 2007; Petrova et al., 2002). Furthermore, constitutive Prox-1 expression is required for maintenance of LEC lineage specificity because conditional inactivation of Prox-1 in mice resulted in loss of phenotypic identity in lymphatic vessels and adoption of blood vessel characteristics, including abnormal coverage by pericytes (Johnson et al., 2008).
An unexpected result of both co-culture of HUVECs with pericytes and knockdown of Prox-1 was the marked induction of VEGFR3 expression. This is surprising given that VEGFR3 is directly regulated by Prox-1 (Hong et al., 2002; Mishima et al., 2007; Petrova et al., 2002). However, VEGFR3 is expressed in the absence of Prox-1 before lymphatic vascular formation (Kaipainen et al., 1995) and is re-expressed in blood vascular endothelium in some conditions (Tammela et al., 2008; Valtola et al., 1999). Furthermore, in our system, in both PROX1 siRNA-treated samples and under co-culture conditions, expression of the VEGFR3 ligand VEGF-C was increased relative to controls, which could create an autocrine stimulatory circuit.
Recently, the concept of endothelial plasticity, and lymphatic endothelial plasticity in particular, is receiving increasing attention (Bixel and Adams, 2008). Our findings add a novel and unexpected aspect to this discussion because we show that normal, committed blood vascular cells have the capacity to undergo aspects of a BEC to LEC transition in response to their ECM environment or the presence of perivascular supporting cells. The effects that we have observed on the expression profiles of endothelial cells grown in specific matrix environments strongly indicate a role for adhesion molecules, particularly integrins, in the control of the endothelial cell phenotype. Mechanotransduction of forces applied to cells by ECM environments of differing stiffness is known to have a profound impact on the specification of cell identity (Butcher et al., 2009), and it will thus be interesting to explore the effects of matrix density on the BEC and LEC expression profiles.
The question arises as to whether the artificial in vitro ECM conditions that we have studied relate to environments that EC experience in vivo. The 3D fibrin assay models an early phase of capillary formation in vivo, when the parent vessel dilates and extravasated fibrinogen is converted to fibrin, forming a provisional scaffold for assembly of the new vessel (Senger, 1996). The type I collagen model is more representative of later stages when EC in vivo migrate away from the parent vessel invading collagen-rich tissue stroma. By contrast, Matrigel might reflect the complex matrix environments of pathologies such as cancer and thus be relevant to the sprouting of tumour neovessels. The BEC-LEC plasticity that we have seen might be reflected in other recent observations from both in vitro and in vivo systems. For instance, expression of lymphatic marker genes in BEC has been seen in vitro in response to treatment with inflammatory cytokines such as interleukin-3 (Groger et al., 2004) or lysophosphatidic acid (Lin et al., 2008). Another recent paper described the presence of a subset of dermal blood capillaries expressing LYVE-1 and podoplanin in chronically inflamed skin (Groger et al., 2007). Likewise, the LEC marker VEGFR3 is seen on tumour neovessels and its inhibition attenuates blood vessel formation (Tammela et al., 2008).
In conclusion, our study has revealed an unexpected plasticity in the phenotype of endothelial cells when cultured in ‘traditional’, well-established in vitro angiogenesis assays, which suggests that these types of monoculture assays might in fact represent processes involved in lymphangiogenesis, rather than blood vessel formation.
Materials and Methods
ECM components, angiogenic factors, and other reagents
Plasminogen-depleted fibrinogen was obtained from Calbiochem (Merck). Growth-factor-depleted Matrigel™ and high concentration rat-tail type-I collagen solution were purchased from BD Biosciences. Calf skin type-I collagen for coating tissue culture surfaces was purchased from Sigma. VEGF and basic FGF were purchased from PeproTech EC. Antibodies used were a polyclonal goat anti-human LYVE-1 [R&D Systems (AF2089)], polyclonal rabbit anti-human antibodies to VEGR3 (Abcam, ab27278), Prox-1 (Abcam, ab11941) and Lamin-B1 (Abcam ab16048), and a mouse anti-human PECAM/CD31 (555444, BD Pharmingen). Polyclonal anti-GAPDH was from Cell Signaling Technology (Danvers, MA). Secondary antibodies included a FITC-conjugated polyclonal donkey-anti goat secondary antibody (Abcam), a goat anti-mouse Alexa-Fluor-488-conjugated secondary antibody (Invitrogen), donkey anti-mouse Cy2 or Cy3 (Jackson ImmunoResearch.) The MMP inhibitor batimastat (BB-94) was a gift from British Biotechnology.
Cells and cell culture
Primary human umbilical vein endothelial cells (HUVECs), primary adult human aortic endothelial cells (HAECs), and adult human dermal microvascular endothelial cells (HDMECs) were obtained from TCS Cellworks, except for the gene array studies for which the HUVECs were purchased from Clonetics. Cells were grown flasks coated with 60 μg/ml type-I collagen (Sigma), using medium and reagents provided by the manufacturer. Cells were grown at 37°C and 5% (v/v) CO2 and used between the first and fifth passages for all experiments. For experimental purposes, endothelial cells were grown in parallel with the 3D gel cultures in serum-containing endothelial cell culture medium, supplemented with 25 ng/ml VEGF and FGF-2. Murine MII perivascular cells have been described previously (Brachvogel et al., 2005; Brachvogel et al., 2007). These cells were originally identified from mice carrying a targeted insertion in Annexin V and were purified from heterozygous mice carrying the Anxa5-lacZ fusion gene (Brachvogel et al., 2005). The cells were routinely maintained on tissue culture plastic in Dulbecco's Modified Eagle's Medium (high glucose with GlutaMAX, Invitrogen), supplemented with 10% fetal bovine serum at 37°C and 5% (v/v) CO2. Pericytes were used between passages 35 and 40.
EC culture within or on top of 3D gels
Collagen gels (1.6 mg/ml) were prepared by mixing rat-tail type-I collagen solution (BD Biosciences) with 10× 199 medium (Sigma), 25 ng/ml VEGF and FGF-2 (PeproTech) and (where indicated) BB94 (a gift from British Biotechnology) or vehicle control (DMSO). The desired concentrations of Matrigel (BD Biosciences) and plasminogen-free fibrinogen (Calbiochem) were obtained by diluting stock concentrations with serum-free endothelial cell medium and supplementing with 25 ng/ml VEGF and FGF-2.
For 3D cultures, endothelial cells were resuspended in either 2.5 mg/ml plasminogen-depleted human fibrinogen (1.5×106 cells/ml), 5 mg/ml growth factor-depleted Matrigel or 1.6 mg/ml type-I collagen (both 1.25×106 cells/ml). Thrombin (0.5 U/ml; Sigma) was added to the fibrinogen mixture to enable polymerisation. Collagen gels were polymerised by adjusting the pH to 7.5 using 1 N sodium hydroxide. Cultures were prepared n 24-well plates and allowed to polymerise at 37°C and 5% (v/v) CO2 for 30 minutes. Where indicated, 200 μl growth-factor-depleted Matrigel (7.9 mg/ml) was polymerised in wells before addition of endothelial cells to upper surface of the matrix (2×104 cells/gel). Serum-containing culture medium was added to the top of the wells. VEGF and FGF-2 were added to the culture medium at a concentration of 25 ng/ml. Where indicated, BB94 or vehicle control were also added to bathing medium. Gels were then incubated for several days at 37°C and 5% (v/v) CO2.
For experiments in which cells were cultured on top of collagen and fibrin gels, the matrices were prepared as described above, and polymerised in six-well plates at a volume of 500−700 μl per well. 1.75×105 cells/gel were added to the upper surface of each gel, in 1 ml serum-containing culture medium supplemented with VEGF and FGF-2 at a concentration of 25 ng/ml. Gels were then incubated for several days at 37°C and 5% (v/v) CO2.
Collagen, fibrin and Matrigel gels were digested for cell harvesting using (respectively): 0.1% collagenase (Worthington), 5 mg/ml trypsin (Sigma) or Dispase (BD Biosciences) as per the manufacturer's instructions. Cells were then recovered from the digested gels by centrifugation at 1200 r.p.m. for 5 minutes.
HUVEC and perivascular cell co-culture
HUVECs were cultured in 3D type I collagen gels prepared as described above, either as monocultures (8.33×105 cells/ml) or in co-cultures with murine perivascular cells (1.6×105 cells/ml) at a 5:1 ratio. Cultures were stimulated with VEGF (10 ng/ml) and PDGF-B (10 ng/ml) added to the culture medium. Gels were cultured at 37°C and 5% (v/v) CO2 for 72 hours. For RNA expression analysis, gels were digested as described above, and HUVECs isolated from mixtures of HUVECs and perivascular cells using CD31-conjugated Dynabeads (Invitrogen) as per the manufacturer's instructions. Briefly, cell pellets isolated from collagen gel were mixed with 1 ml PBS containing 0.1% BSA and 50 μl Dynabeads-CD31, and incubated at 4°C for 20 minutes with gentle tilting and rotation. The HUVECs binding Dynabeads-CD31 were separated by magnetic force and RNA was extracted.
Total RNA was extracted from endothelial cells using High Pure RNA extraction kit (Roche), RNeasy mini columns (Qiagen), or TRIzol reagent (Life Technologies), all as per manufacturer's instructions and quantified using an ND-1000 spectrophotometer (Nanodrop Technologies). 0.1−1 μg total RNA per sample was reverse transcribed to cDNA using random hexamers (Amersham Pharmacia), Superscript II reverse transcriptase (Life Technologies), dNTP mix (Roche) and RNase inhibitor (Promega) as per manufacturer's instructions
Protein expression analysis by western blotting
Monolayer cultures were harvested by rinsing cells and scraping into 1 ml of PBS. Cells were harvested from Matrigel by digestion using Dispase, as described above. To obtain sufficient protein for western blotting, cells harvested from ten separate Matrigel assays were combined into one sample. The cells were pelleted by centrifugation and pellets were resuspended in lysis buffer (10 mM Tris-HCl, pH 7.6, 10 mM NaCl, 3 mM MgCl2 1% NP40, supplemented with Roche Complete Mini EDTA-free protease inhibitor tablets. Samples were left on ice for 1 hour with occasional mixing. Cellular debris was pelleted by centrifugation, and the protein content of the soluble fractions was measured with a BCA protein assay kit according to the manufacturer's instructions (Thermo Scientific, Rockford, IL). For western blotting, samples separated by SDS–PAGE on a 10% resolving gel were transferred to PVDF (polyvinylidene difluoride) membranes (Bio-Rad). Protein bands were detected by incubation with the appropriate primary antibody followed by horseradish peroxidase (HRP)-conjugated secondary antibody (DAKO).
Microarray analysis of differential gene expression during HUVEC tubulogenesis
HUVECs were embedded in 3D fibrin matrices and cultured as described above. For 2D cultures, 2.5×105 HUVECs were seeded on top of 1 ml fibrin gels (prepared as described above). For 3D cultures, a maximum of four 300 μl gels were pooled and homogenised per sample. The cDNA for gene array analysis was prepared from 20 μg of total RNA. For each sample, two cDNA preparations were made and coupled to either Cy3 or Cy5 dyes (Amersham Biosciences, Little Chalfont, UK) in 0.1 M Na2CO3 pH 9.0 for 1 hour at room temperature. The coupling reaction was then quenched with 2 M hydroxylamine for 15 minutes and the labelled cDNA purified with Amicon Microcon-30 concentrators. Quantification of labelled cDNA was performed with an ND-1000 nanodrop spectrophotometer with readings at A260, A550 and A650. 25 pmol of dye labelled cDNA was used for each microarray probe.
Human 19K glass slide ‘oligo arrays’ were obtained from the Australian Genome Research Facility (AGRF, Parkville, Australia). Slides were first blocked in 0.1% SDS at 95°C for 1 minute, followed by a 1 minute wash in 5% ethanol and a further 1 minute wash in distilled H2O, and dried. Slides were then prehybridised with 10 mg/ml BSA, 25% formamide, 5× SSC and 0.1% SDS for 45 minutes at 42°C, and then hybridised with 25 pmol of each labelled cDNA. For each comparison (i.e. 2D vs 3D), sample A labelled with Cy3 was mixed with sample B labelled with Cy5. A dye-swap hybridisation control on a separate slide was also performed for each sample pair, with sample A labelled with Cy5 and sample B labelled with Cy3. 25 pmol of both A and B samples were mixed with Cot1 DNA (5 μg/μl), Poly A DNA (10 mg/ml), salmon sperm DNA (10 mg/ml) in hybridisation buffer (25% formamide, 5× SSC and 0.1% SDS) for a total reaction volume of 60 μl. Samples were heated at 100°C for 2 minutes and hybridised to glass slide microarrays overnight at 42°C. The next day, arrays were washed for 10 minutes in 2× SSC, 0.2% SDS, for 10 minutes in 2× SSC, and then for 10 minutes in 0.2× SSC. Glass slides were then dried by centrifugation at 500 g for 3 minutes and scanned with a GenePix 4000B scanner (Molecular Dynamics, Amersham). Gene array images were analysed using the GenePix Pro 4.0 software. Local background correction was performed and data normalised according to the median of ratios. A gene was considered differentially regulated if the ratio between sample pairs A and B was >1.5 or <0.67.
TaqMan real-time PCR analysis
Immunostaining of monolayer and 3D collagen cultures
Samples were washed in PBS, fixed with 80% methanol, 20% DMSO overnight at 4°C, rehydrated in 50% methanol in PBS, 20% methanol in PBS and PBS-T (PBS, 0.1% Tween-20) for 1 hour each and then were incubated with blocking solution (PBS, 10% fetal calf serum, 5% BSA) overnight at 4°C. Primary antibodies including goat anti-human LYVE-1 at 2 μg/ml), mouse anti-human PECAM/CD31 at 1 μg/ml, rabbit anti-human VEGFR3 at 4 μg/ml, rabbit anti-human Prox-1 at 10 μg/ml were added into the samples incubated in blocking buffer overnight at 4°C and then washed six times for 1 hour each in TBS-T (TBS, 0.1% Tween 20). Corresponding secondary antibodies conjugated with Cy2 or Cy3 (Jackson ImmunoResearch) were incubated in blocking buffer overnight at 4°C and washed as described above. For secondary antibody control, there were no primary antibodies added before incubation with same secondary antibodies. After nuclei staining, samples were mounted in Gelvatol and examined by fluorescence microscopy (SteREO LumarV12 and Axioplan2; Carl Zeiss, Germany). Pictures were captured with Axiovision software (version 4.5). Antibodies used in double staining procedures have been tested for minimal cross reactivity and spectral overlap.
Two pre-designed siRNA duplexes targeting PROX1 and a negative control siRNA duplex were purchased from Ambion (UK). HUVECs were seeded in collagen-coated six-well plates and transfected the following day with the various siRNAs at concentrations of 150–200 nM, as indicated. Oligofectamine and Opti-MEM (Invitrogen, UK) were used for delivery of the siRNAs as per manufacturer's protocol. Cells were cultured for a further 48 hours before seeding onto Matrigel layers (5 mg/ml). Serum-containing culture medium was added to the top of the wells. VEGF and FGF-2 were added to the culture medium at a concentration of 25 ng/ml, and cells cultured for a further 24 hours at 37°C and 5% (v/v) CO2. Samples were harvested as described above at the indicated time points.
TaqMan expression data were analysed using one-, two- or three-way ANOVA tests as described in the text. P<0.05 was considered significant.
This work was supported by funding from the Biotechnology and Biological Sciences Research Council (BBSRC), the Big C, Norfolk Fundraisers and the EU Framework Programme 6 Cancerdegradome Project LSHC-CT-2003-503297. L.S.C. was a recipient of a BBSRC PhD studentship. Z.Z. and E.P. were supported by the British Heart Foundation, Project PG/06/071/21115T. E.W.T. and M.A.L. were supported by the Victorian Breast Cancer Research Consortium, Australia.
↵* Present address: Amgen, 1 Amgen Center Drive, Thousand Oaks, CA 91320, USA
Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.064279/-/DC1
- Accepted July 16, 2010.
- © 2010.