Background
Formation and maintenance of new vascular vessels supporting tumor growth is a process which involves complex communications among different cell types. Interactions between endothelial cells and supporting pericytes and vascular smooth muscle cells lead to active remodeling during angiogenesis [
1,
2]. It is now well established that pericytes participate in angiogenic signal transduction and direct control of endothelial cell proliferation and thus play an important role in vascular morphogenesis and function [
3]. Paracrine pericyte-endothelial interactions through VEGF, PDGF, and angiopoitin signaling promote tumor angiogenesis, pericyte recruitment and pericyte coverage of tumor vessels [
4‐
7]. It is also known that pericytes shield endothelial cells from apoptotic signals [
8]. As such, pericytes, in conjunction with endothelial cells, present a logical target for antiangiogenic therapeutic strategies.
Several approaches are currently employed to target tumor angiogenesis. Inhibition of VEGF signaling by specific antibodies or small molecule agents directed against VEGFR2 tyrosine kinase has been shown to inhibit formation and outgrowth of new blood vessels and several angiogenesis inhibitors have been used as anticancer therapy in patients [
9]. In addition to inhibition of angiogenesis, recent studies showed that pharmacological VEGF inhibition, especially with anti-VEGF antibody, normalizes vessel structure and function through pericyte recruitment to disorganized tumor vessels [
10,
11]. Since normalization of tumor vasculature improves their perfusion and oxygenation, cytotoxic drugs were given in combination with anti-VEGF antibody causing the best response to radiation or cytotoxic therapy [
12]. On the other hand, roles of pericytes in resistance to antiangiogenesis therapy have emerged, since most of cancer patients acquire resistance and become refractory to anti-VEGF therapy [
13].
Another approach which is used to selectively target abnormal tumor vasculature is to apply agents which show vascular-disrupting activity to already formed blood vessels. Tumor vascular disrupting agents (VDA) cause selective induction of apoptosis in endothelial cells and induce vascular damage in already formed tumors [
14]. Several specific VDAs are now undergoing clinical evaluation [
15], although no VDA has been approved for cancer therapy.
Tubulin-binding agents were known to show antivascular effects represent both antiangiogenesis and vascular disrupting activity, which were well reported for paclitaxel and combrestatin analogs, respectively [
16,
17]. It is apparent that microtubules play an important role in angiogenesis and maintenance and stability of tumor vessels. Eribulin mesylate (eribulin), a non-taxane inhibitor of microtubule dynamics, belongs to the halichondrin class of antineoplastic drugs [
18]. Eribulin’s novel mode of inhibiting microtubule dynamics differs from most other tubulin-targeting agents, and involves inhibition of the microtubule growth phase without affecting the shortening phase, together with sequestration of tubulin into non-productive aggregates [
19]. This results in blockage of normal mitotic spindle formation, irreversible mitotic block and cell death by apoptosis [
18,
20,
21]. At the biochemical level, eribulin achieves these results by specifically binding with high affinity to a small number of sites on the plus ends of microtubules [
22].
In this paper we examined effects of eribulin on HUVECs and HBVPs in vitro. We analyzed effects of eribulin on global gene expression in HUVECs and HBVPs in a comparison with paclitaxel, which is a stabilizer of microtubule dynamics and has a distinct mechanism from eribulin. We determined effects of these two drugs on cell proliferation in mono-cultures of HUVECs and HBVPs, and employed a newly developed capillary network formation assay, in which HBVPs co-cultured with HUVECs to promote network formation, in order to assess antivascular activities of both drugs in the context of physiologically relevant cell-cell interactions.
Materials and methods
Cell cultures and compounds
Primary human umbilical vein endothelial cells (HUVECs) were either purchased from Lonza (Walkersville, MD) or isolated from a single umbilical cord by a method described previously [
23] and maintained in endothelial cell growth medium EGM-2 supplemented with EGM-2 SingleQuots except for hydrocortisone (Lonza,). Human brain vascular pericytes (HBVPs) were obtained from ScienCell Research Laboratories (Carlsbad, CA), and were grown in Pericyte Medium (ScienCell). To confirm their authenticity, cultured HBVPs were examined for expression levels of 6 key pericyte markers grown on plastic (Additional file
1). Both HUVECs and HBVPs were grown on collagen type I-coated plastic ware. For cell proliferation and gene expression experiments, cells were used at <5 passages.
Green fluorescent protein (AcGFP)-expressing HUVECs were established by infection with a retrovirus for gene transfer of AcGFP followed by collecting high level AcGFP-expressing HUVECs by fluorescence activated cell sorting. Cells were maintained at 37°C in a humidified atmosphere containing 5% CO2.
Paclitaxel was purchased from Sigma-Aldrich (Saint Louis, MO) and Wako Pure Chemical (Osaka, Japan). Eribulin mesylate was manufactured by Eisai Co., Ltd (Ibaraki, Japan). Both compounds were dissolved in DMSO to yield a stock concentration of 1 mmol/L.
Cell proliferation assay
HUVECs and HBVPs were plated at 3000 cells per well in 96-well plates. Three hours later serial dilutions of tested compounds were added. Control wells were treated with 0.1% DMSO. Cell growth was assessed 4 days later using the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, WI). Three experiments were performed, each in triplicate. The mean of the half maximal inhibitory concentration (IC50) value and 95% confidence interval (CI) were calculated based on IC50 values generated from separate sigmoidal curves representing the growth inhibition activity versus the eribulin and paclitaxel concentration of three independent experiments. Statistical analyses were performed using the GraphPad Prism version 5.02 (GraphPad Software, San Diego, CA).
HUVEC/HBVP co-culture assay and measurement of pericyte-covered capillary network length reduction
HBVPs and AcGFP-expressing HUVECs were diluted and mixed to densities of 1.87 × 105 cells/mL and 1.3 × 104 cells/mL with medium, respectively. Cell suspensions were dispensed at 100 μL per well in black-walled, clear-bottomed, collagen type I-treated 96-well plates (Greiner Bio One, Frickenhausen, Germany) and incubated for 10 days with culture medium changes every 2 days.
To measure effects of compounds on capillary network lengths in HUVEC/HBVP co-cultures, 100 μL solutions of eribulin, paclitaxel or 0.1% DMSO (vehicle control) were exchanged into each well and incubated for 5 days. Three independent experiments were performed in triplicate. Fluorescence microscopy was performed every day for 5 days using an IN Cell Analyzer 1000 (GE Healthcare, Piscataway, NJ) to obtain images of pericyte-covered capillary networks formed by AcGFP-expressing HUVECs under co-culture conditions. Images were negative/positive inverted and high-contrasted by Irfan View software version 3.61 (Irfan Skiljan, Wierner Neustadt, Austria), followed by analysis using Angiogenesis Image Analyzer software version 2.0 (Kurabo, Osaka, Japan) to measure lengths of pericyte-covered capillary networks. Data for pericyte-covered capillary network lengths were expressed as means + SEM.
IC50’s of 5-day treatments in three independent experiments were analyzed by nonlinear regression analysis, and means of three IC50 values were determined. Statistical analyses were performed using GraphPad Prism version 5.04 (GraphPad Software).
Measurement of cell viability in co-culture
Assessment of cell viability in the co-culture assay was performed by modification of the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide colorimetric assay [
24]. After measurement of pericyte-covered capillary network lengths in the HUVEC/HBVP co-culture assay, medium in wells was exchanged with 100 μL of 11-fold media-diluted WST-8 reagent (Cell Counting Kit-8, Dojindo Laboratories, Kumamoto, Japan), followed by incubation for one additional hour. The optical density (OD) at 450 nm of each well was measured with a microplate reader (SpectraMax 250, Molecular Devices, Sunnyvale, CA), with a reference wavelength of 650 nm. Data for cell viability were expressed as means + SEM.
Microarray analysis
HUVECs and HBVPs were plated at 1 × 105 cells per well in 12-well plates. The next day, cells were treated with compounds at 10 × IC50 concentrations as determined in cell proliferation assays. Control wells were treated with 0.1% DMSO in culture media. Experiments were done in triplicates. Twenty four hours later, cells were collected and RNA was extracted using an RNeasy Mini kit (Qiagen, Valencia, CA) with DNase I on-column treatment according to the manufacturer’s protocol. RNA was quantified using Nanodrop ND-1000 (Thermo Scientific, Wilmington, DE). For each sample, 1 μg total RNA was used to prepare biotinylated fragmented cDNA for analysis on Human Exon 1.0 ST arrays (Affymetrix, Santa Clara, CA). Sample preparation was performed in accordance with manufacturer’s instructions using the WT Expression kit (Life Technologies, Carlsbad, CA) and the GeneChip WT (Whole Transcript) Terminal Labeling Kit (Affymetrix). Hybridizations were conducted according to the GeneChip Expression Analysis Technical Manual (Affymetrix). Arrays were washed and stained using Affymetrix Fluidics Station 450, and finally scanned using Affymetrix GeneChip Scanner 3000.
Gene chips were analyzed using Affymetrix Microarray Analysis Suite (MAS) version 5.0 to obtain raw data. Fluorescence intensities of scanned images were quantified, corrected for background noise, and RMA normalized using Refiner software (Genedata, Basel, Switzerland). Normalization procedure within RMA consisted of a GC background subtraction, quantile normalization and summarization. QC analysis of the microarrays was performed according to the standard protocol within Genedata software. All microarrays passed QC criteria. Statistical analysis was performed with Expressionist software (Genedata). We restricted our analysis to gene intensities >100, based on the detection limit of the Affymetrix gene chip. Gene expression levels of untreated samples were compared to those from treated samples using t-test analysis. Comparisons were limited to p-values of 0.05 in order to eliminate 95% of false positives from the data set, and to fold change levels of at least 1.5 in order to remove background effects.
Quantitative real time PCR analysis
Reverse transcription was carried out with 2.5 μg of total RNA using a Superscript VILO kit (Life Technologies). Synthesized cDNA were used as templates for quantitative polymerase chain reaction (qPCR) using custom TaqMan Low Density Array (TLDA) (Life Technologies) with ABI 7900HT (Life Technologies). Selected gene probes related to differentially expressed genes identified in the microarray experiment were used for TLDA (Additional file
2). Data were normalized using Expressionist (Genedata).
Hierarchical clustering was done using Genedata software. Genes differentially expressed between eribulin and control and between paclitaxel and control were uploaded into Ingenuity Pathway Analysis (IPA; Ingenuity Systems, Redwood City, CA), and only significantly affected signaling pathways were reported with the cut off value of p < 0.01 for pathway enrichment.
Discussion
In addition to normal development, angiogenesis plays an important role in tumor growth [
25]. However, tumor vasculature is largely distinct from normal vessels in a number of properties. Tumor vasculature is known to be grossly disorganized and tortuous. Furthermore, tumor vasculature is leakier than normal vasculature because it grows fast and does not have close interactions with pericytes. Interactions between endothelial cells and pericytes in the blood vessel wall are important processes in the regulation of vascular formation, stabilization, remodeling and function. Failure of such interactions are implicated in human pathological conditions, including diabetic microangiopathy [
26], ectopic tissue calcification [
27], stroke [
28] and cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) syndrome, one of the most common inherited small vessel diseases of the brain [
29]. Tumor vasculature is heterogeneous in its interaction with pericytes, and many reports indicate that those having a “mature” appearance, defined primarily as having close endothelial cell-pericyte interactions, show resistance to antiangiogenic therapies [
30].
To shed light on effects of eribulin on tumor angiogenesis, we studied HUVECs and HBVPs as models of endothelial cells and pericytes in tumor vasculature. We showed that these two cell types responded differently to eribulin and paclitaxel treatment by specific changes in expression of multiple genes. In HUVECs, both eribulin and paclitaxel led to down-regulation of most affected genes, with drug signatures greatly overlapping. Notably, eribulin and paclitaxel signatures in HBVPs were clearly different from those in HUVECs.
We found that among 14 genes common for both drugs and both cell types, 4 genes were related to cell cycle progression (BTG2, CDC45, CSE1 and MCM5). The largest group of 6 common genes, all of which were down-regulated, was made up of histones, with 5 of these belonging to the H1 gene family. In addition, 1B histone chaperone gene ASF1B was also down-regulated by both drugs in HUVECs and HBVPs. There are 11 known H1 family members in the human genome, and 5 of those were found in this study to be down-regulated by the two tubulin-binding drugs under study. The H1 histones work as linkers for nucleosomal core particles and have an important role in establishing and maintaining chromatin structure and regulation of gene activity [
31]. Down-regulation of H1 histones causes de-condensation of chromatin and possibly up-regulation of gene transcription.
Our data showed that in HBVPs, gene expression changes caused by eribulin and paclitaxel were dramatically different. Only 12 percent of genes were similarly affected by both drugs, while the rest showing drug-specific changes. While paclitaxel induced transcriptional activity of a majority of affected genes, 61 percent of genes were down-regulated and by eribulin. Direct comparison of eribulin’s versus paclitaxel’s gene expression profiles showed a large number of differentially expressed genes. These data indicate that either the direction of altered expression was opposite for the two drugs, or the degrees of alteration were significantly different between eribulin and paclitaxel treatment.
A number of observed differentially expressed eribulin-specific genes (Additional file
3, highlighted in red) were previously reported to be associated with angiogenesis, pericyte biology and vascular remodeling (NOTCH3, PTX3,TNFAIP1, PGF, GREM1, TIMP4, LIPG, MYOCD) [
32‐
36]. Of particular interest is NOTCH3, a gene encoding one of the notch family members and known to be underlying cause of CADASIL [
37,
38]. NOTCH signaling is critical for maintenance of normal vascular structure, angiogenesis and vascular remodeling in both physiological and pathological conditions [
39,
40]. In our study we detected significant upregulation of NOTCH3 expression after eribulin treatment in HBVPs.
To evaluate specific effects of eribulin on signaling pathways, we performed pathway analysis using Ingenuity software. This analysis showed that in pericytes, genes in several pathways were selectively affected by eribulin compared to paclitaxel. In particular, cell cycle control of chromosomal replication including RAN signaling and mismatch repair pathways were among most significantly changed.
Most studies published to date have used in vitro endothelial cell-based vascular disruption assays to evaluate activities of compounds against newly formed capillary-like structures. In such assays, endothelial cells that are plated on thick layers of Matrigel form networks of cord-like structures, reminiscent of newly formed vessels. Treatment with antivascular compounds results in disruption of the integrity of such cord-like networks. However, such assays, using only endothelial cells, do not take into account the physiologically-relevant close interactions between endothelial cells and supporting pericytes within the context of tumor vasculature. To overcome this limitation, we developed a novel assay in which AcGFP-transfected endothelial cells grow and form capillary networks in co-culture with pericytes. The effects of two tubulin-targeting compounds, eribulin and paclitaxel, on network formation were evaluated in this assay by measuring the lengths of pericyte-covered capillary networks. We found that eribulin and paclitaxel behaved dramatically different in this assay. Eribulin, in contrast to paclitaxel, showed significant antivascular activity causing dramatic pericyte-covered capillary network shortening effects relative to its antiproliferative effects on HUVECs. On the other hand, pericytes seemed to protect endothelial networks under paclitaxel treatment conditions and thus network shortening activity was much weaker compared to its antiproliferative effects on HUVECs. Thus, eribulin appeared to impair interaction of pericytes with endothelial cells through the activity against pericytes differently from paclitaxel in this assay. Consequently, eribulin showed much higher activity as an anti-vascular agent than to paclitaxel in this co-culture assay. Interestingly, in the sandwich tube formation assay in which HUVECs can form capillary network without co-culturing with HBVPs, eribulin and paclitaxel showed similar IC50 values in shortening capillary network at 4 days after treatment. This strongly supports the above hypothesis. In future studies, it will be important to compare and further define this newly discovered antipericyte-based antivascular property of eribulin with other known tubulin-binding drugs both in in vitro and in vivo angiogenesis models.
Competing interests
SIA, SK, NT, JM, JC, YO and YF are full-time employees of Eisai. JO was a full-time employee of Eisai. The authors declare no other competing interests.
Authors’ contributions
All authors contributed to the design of the study, acquisition of data, analysis and interpretation of data or manuscript writing, and have read and approved the final manuscript.