We have developed an in vitro angiogenesis system using immortalized endothelial cell lines that form branched networks of vessel-like structures that assume a normal inside-outside orientation in co-culture with EMCs, which form smooth muscle and pericytes, and appears suitable for scale-up to HTS. The cells retain their vessel-forming potential over many passages (>30). Moreover, both cell types were engineered to express a panel of nuclear- or cytoplasm-localized green (eGFP) or red (mCherry) fluorescent proteins that greatly facilitate time-lapse imaging and analysis to monitor cellular behavior during vessel formation. The characteristics of branched networks were quantified using MetaMorph, revealing that the in vitro assay can discriminate the effects of culture supplements and bioactive small molecules. Most assays for vessel formation use primary cells that are poorly suited for high throughput applications, and below we discuss the biological similarities of our system to primary cell assays and the added utility for HTS.
The HMEC-1 line, that we used for creating our fluorescent lines, was originally derived from human dermal microvascular endothelial cells and has already been extensively characterized and compared to both microvascular and macrovascular primary cells (Ades et al., 1992
; Xu et al., 1994
; Bouis et al., 2001
; Unger et al., 2002
; Bender et al., 2008
) and references therein). In previous studies HMECs were concluded to be one of the best characterized and also one of the most physiologically preserved (i.e. similar to primary endothelia) microvascular lines currently available (Bouis et al., 2001
; Unger et al., 2002
; van, Jr. et al., 2008
). Consistent with this conclusion, HMEC-1 line has been widely used in numerous assays related to microvascular endothelia functions with many of the results later confirmed in vivo, thus validating HMECs as an appropriate endothelial cell model (see the specific references in Supplemental Note 1
In our hands, fluorescent HMECs were found to express a repertoire of endothelial cells receptors similar to primary endothelial cells (HUVECs), such as VEGFR-2, VE-cadherin, Tie-2 and several Eph receptors (). Moreover, we have shown that HMECs respond to proangiogenic factors, as exemplified by the ability of VEGF and EphA2 receptors to become tyrosine phosphorylated (activated) in response to VEGF and TNFα treatment, respectively, consistent with the formation of branched networks in our and prior studies (Ades et al., 1992
; Meade-Tollin and Van Noorden, 2000
). The similar levels of VEGFR-2 activation obtained in parental and eGFP expressing HMECs, together with the robust network formation we have observed with the fluorescent HMEC-based stable lines suggest that the expression of fluorescent proteins does not alter their angiogenic properties.
We found that stably transduced HMECs expressing fluorescent proteins showed far less toxicity than CellTracker dye labeling over the time course of our studies (), and are therefore a convenient source of labeled cells for HCS. Similarly, fluorescent protein-labeled EMCs did not show significant toxicity or impair vessel network formation. Other endothelial lines might be similarly engineered, including three human microvascular endothelial cell lines that have been immortalized using human telomerase catalytic protein (Yang et al., 1999
; Venetsanakos et al., 2002
; Shao and Guo, 2004
). Differences between the endothelial cell lines might be useful to generalize and validate results obtained from one line, or pinpoint compounds that affect particular vascular beds.
Importantly, the vessel formation assay can detect activating and inhibiting factors. Maintenance of the HMEC vessel-like network depends on the presence of factors in the complete medium (that contains FGF, VEGF, IGF, EGF, ascorbic acid and FBS), and the immature networks quickly deteriorate in incomplete medium (). In addition to indicating a dependence on characterized angiogenic factors, this result also suggests that the basal system can be used to study angiogenic factors. In this regard, it is interesting that a low dose of PMA delayed network formation, while also prolonging the persistence of endothelial networks ().
Conversely, sunitinib, SU5416, suramin, and vinblastine inhibited vessel network formation. Taken together, our results indicate that the normal responsiveness of the assay to angiogenic factors, plus the ability to use high content imaging to dynamically monitor vessel formation for over 24 hours, rather than a single endpoint, in high throughput should enable screens to discern complex effects on angiogenesis. Indeed, the success of our semi-automated preliminary quantification of network parameters formed in 96 and 384-well formats yielded Z′ values between 0.3 and 0.9, indicating the high dynamic range needed for minimizing false positive and negative results with only one experiment (or well) per condition, as needed for large scale HCS. This evidence of suitability for scale-up has also led us to begin developing more robust algorithms to segment and automatically evaluate network morphogenesis and the “correctness” of the inside-outside cellular orientation.
HMEC-1 cells have previously been described to migrate and form branching structures when grown on Matrigel (Ades et al., 1992
; Meade-Tollin and Van Noorden, 2000
). However, here we present quantitative analyses of the dynamics and morphological parameters of networks formed by HMECs alone, as well as (for the first time) in co-culture with EMCs. Although co-culture of HUVECs and smooth muscle cells have already been proposed as a model for high content screening (Evensen et al., 2010
), we are the first to demonstrate that microvascular HMECs are not only capable of network formation in conjunction with EMCs, but also assume correct orientation within differentiated branches.
Inclusion of EMCs resulted in branching structures with normal inside-outside disposition (–), similar to normal vessels and primary assay co-cultures (Evensen et al., 2010
), and this should facilitate the automated study of factors that influence the formation of intermediate and large vessels. HMEC-EMC vessel-like structures () sprouted spike-like projections from the EMCs into the extracellular matrix. Similar spikes also form in HUVEC/pericyte co-cultures, and in vivo
such projections are not found on the normal vessels, but are associated with immature, leaky vessels (M. Komatsu, personal communication). Thus, while the co-culture model does not fully recapitulate the formation of normal vasculature, it might be useful to study vessel normalization, or screen and evaluate molecules that exert a normalizing effect on defective vasculature analogous to that associated with tumors. Specifically, defective vasculature has been considered a potential reason for poor delivery of anti-tumor drugs to within solid tumors, and inhibition of angiogenesis has been suggested to hinder drug delivery and promote cancer cell selection towards higher malignancy and metastatic potential (Paez-Ribes et al., 2009
). Thus, the approach of promoting or normalizing vasculature, instead of inhibiting tumor vasculogenesis, may prove to be a more helpful strategy by allowing more efficient delivery of chemotherapy drugs (reviewed in (Jain, 2005
). Thus, our experimental results should enable HCS assays based on functional vessel formation that may lead to drugs that normalize defective vasculature.
In conclusion, we have shown that our engineered HMEC fluorescent lines retain sufficient features of the angiogenesis phenotype to serve as a model of blood vessel formation that should enable large scale HCS. These features include retention of angiogenic potential and phenotypic stability, relative ease of maintenance, and suitability for creation of stable fluorescent protein cell lines that facilitate image analysis. Nuclear localized fluorescent proteins will enable future cellular tracking of each cell type, while cytoplasm-localized fluorescent facilitates image segmentation and analysis necessary for accurate quantification of dynamic aspects of network morphogenesis, and both will facilitate automated subcellular localization of mechanistically important signaling proteins. Use of a model characterized by formation of vessel-like structures will be an exciting advance over the migration-only surrogates that have been used thus far in large-scale screens. While it’s not necessary for an initial HC/HT screen to recapitulate normal biology perfectly because each can be followed by secondary screens using more temperamental primary cell models, a better initial screen is much more likely to generate pertinent hits. The novel features of our model enable both large scale primary screening and fluorescent time-lapse imaging, the latter of which will be critical for identifying the key time point for each mechanistic/assay goal and for identifying differences in the behavioral/phenotypic actions of molecular hits in follow-up studies without the need for fixing the cells at many different time points.