Background
Endothelial cells (ECs) are the building blocks of the vascular system and are characterized as the single-cell layer of epithelium that forms the inner cell lining of blood vessels and lymphatics [
1]. Vascular ECs were initially considered passive conduits for delivering oxygen and nutrients to all tissues [
2,
3]. The development of appropriate culturing systems for primary human organ-specific ECs provided the opportunity to identify EC heterogeneity in different organs as well as their functional properties under normal and pathological conditions [
4-
6]. Today, ECs have been implicated in several perfusion-independent processes including tissue regeneration, tumor growth and dormancy through secretion of angiocrine factors [
7-
14]. In addition, increasing evidence reveals that endothelial cellular identity is more plastic than previously thought [
15]. This plasticity results in phenotypical and functional modifications under different contextual conditions. A characteristic example of such phenotypic modification is endothelial-to-mesenchymal transition (EndMT), during which ECs lose their endothelial phenotype and acquire mesenchymal traits [
16-
18]. EndMT is implicated in tumor progression through complex modulation of the tumor and its stroma [
17]. It is likely that precise analysis of cellular transformation in tumor microenvironments will reveal subsets of additional cellular phenotypes that might be drug targets and/or biomarkers.
In this study, we aimed to investigate the role of tumor cells in promoting mesenchymal phenotype in ECs by setting up tumor-endothelial co-culture systems in the absence of serum or cytokine supplementations. We initially confirmed the induction of mesenchymal phenotype in Human Umbilical Vein Endothelial Cells (HUVECs) by breast tumor cells. Then, to overcome the barrier of endothelial sensitivity to starvation and tumor cell-induced cell death [
19], we continued our work with the previously described E4-ECs (that we here refer to as ECs) [
10,
20-
22]. ECs were produced through transfection of Primary Endothelial Cells (PECs) with adenoviral
E4ORF1 gene as described previously [
21]. While this transfection provides a low Akt activation allowing endothelial survival in a serum and cytokine-free condition, it does not modify the endothelial phenotype as has been widely used [
10,
20,
22]. Besides, activation of Akt in tumor endothelium has been previously reported [
23] and our model might thus be more optimal to mimic the crosstalk between ECs and cancer cells
in vivo without any background effect. Using breast cancer cells (BCCs), we showed that BCCs in co-culture with ECs stimulated transcriptomics modification of ECs partly represented by acquisition of mesenchymal phenotype. While a similar phenomenon (EndMT) has already been described in the developmental and pathological context, we were able to show that tumor cells were capable of stimulating mesenchymal phenotypes in ECs and the tumor-associated ECs retained their endothelial properties while gaining mesenchymal phenotypes. In addition, this transition was reversible and dependent on continuous contact between ECs and BCCs. Subsequently, we showed that the mesenchymal ECs were capable of constituting a pro-tumoral niche responsible for increasing BCC proliferation, mammary stem cell self-renewal, and pro-metastatic properties. Our results also suggest that tumor-promoted mesenchymal shift in ECs is regulated by Smad signaling through the synergistic stimulation of TGFβ and notch pathways.
Methods
Cell culture & reagents
Breast cancer cell lines MDA-MB231 (MDA-231), MCF-7, and HUVEC were purchased from American Type Culture Collection (ATCC, USA). GFP
+ECs (ECs) were developed as described previously [
21]. Human recombinant Jagged1 and TGFβ1 were obtained from R&D Systems and PeproTech, respectively. Υ-secretase inhibitors (GSI) and SB-431542 were purchased from Sigma (USA). Breast cancer cells (BCCs) were grown in DMEM/High glucose (HyClone, USA) supplemented with 10% FBS, L-glutamine, non-essential amino acids (NEAA), and penicillin/streptomycin in a humidified incubator with 5% CO
2. ECs were grown in M199 growth medium (Gibco, USA) supplemented with 20% FBS, 20 ng/ml β-Endothelial Cell Growth Factor (βECG), 20 units/ml heparin and penicillin/streptomycin. The co-cultures were prepared by mixing one part BCCs with 10 parts GFP
+ECs (1:10 ratio) and cells were grown in 1:1 ratio of DMEM/High and M199 media in the absence of serum and growth factors (complete starvation). Co-cultivation of BCCs and ECs was performed over 3–5 days under adherent condition.
Sphere forming assay was used to enrich mammary stem cells (mammospheres) as previously described by Dontu [
24]. We slightly modified that protocol and co-cultured mammospheres with GFP
+ECs at 1:10 ratio under non-adherent condition to obtain mammo-angiospheres. Mammo-angiospheres were therefore composed of both tumor and GFP
+ endothelial colonies mingling together. Spheres were grown in a so-called “
3D media” as described by Dontu and colleagues by using DMEM-F12 (HyClone, USA) supplemented with 2% B27, 20 ng/mL basic fibroblast growth factor (bFGF) and epidermal growth factor (EGF), and 5 μg/mL insulin. In order to prevent the formation of cellular aggregates, a highly viscose
3D media was prepared by addition of 0.2% methylcellulose (Sigma, USA). Stem cell enrichment was evaluated by measuring the perimeter of mammospheres or angiospheres with NIH ImageJ 64 software or by quantifying the number of spheres. A GFP filter was used to distinguish angiospheres.
Cell proliferation assay
MDA-231 or MCF-7 cells were co-cultured with GFP+ECs (1:10 ratio) under starvation and ECs survival was assessed at different intervals by trypsinization and repeated manual counting by hemacytometer. A GFP filter was used to distinguish the GFP+ECs from unstained BCCs. In this study, ECs that have been pre-exposed to BCCs are referred to as ECsMes, whereas ECsNorm are normal ECs with no prior contact with BCCs. To see the effect of ECsMes on BCC proliferation and survival, GFP+ECs were directly co-cultured with MDA-231 and MCF-7 cells for three to five days to obtain GFP+ECsMes prior to initiating a proliferation assay. Next, we started a proliferation assay with ECsMes while still growing with BCCs and newly established co-cultures of GFP+ECsNorm and BCCs for seven more days under complete starvation. BCCs either in mixture with GFP+ECsNorm or GFP+ECsMes were then counted by trypsinization and manual counting excluding ECs by GFP filter.
Flow cytometry & cell sorting
Antibodies to human PE-CD31 (560983), AF647-VE-cadherin (561567), and fibronectin (FN1, 610077) were purchased from BD Biosciences (USA). AF633-F-actin (phalloidin, 22284) is a product of Invitrogen (USA), vimentin (5741) and α-SMA (ab5694) are from Cell Signaling Technologies and Abcam, respectively (USA). The secondary antibodies were purchased from Invitrogen. GFP+ECs were either cultured alone or co-cultured with BCCs. To stain ECs in mono or co-cultures, cells were initially trypsinized and washed with PBS. For labeling intracellular proteins, cells were initially fixed then permeabilized on ice in freshly prepared 3.7% paraformaldehyde and 0.1% TritonX-100 for 10 minutes/each prior to incubation with primary antibodies (permeabilization by TritonX-100 was not carried out for cell surface proteins). Briefly, cells were resuspended at 1 × 106 cells/100 μL density in a staining buffer containing 5% FBS, 1% BSA, 0.2 mM EDTA in PBS. To enhance the specificity of staining, FcR blocking (Miltenyi Biotec, USA) was added at 5 μL/1 × 106 cells prior to incubation with primary antibodies. Primary antibodies were then added according to the instructions provided by the manufacturers and incubation was done for 1 hour at 4°C. After washing, cells were stained with secondary antibodies for 30 minutes at 4°C followed by washing. Fluorescent light (FL) was quantified using Fluorescence Activated Cell Sorting (FACS) on a SORP FACSAria II (BD Biosciences), eGFP fluorescence was acquired using a 488 nm blue laser and 510/50 nm emission, Phycoerythrin fluorescence (PE) was acquired using a 498 nm blue laser and 575/26 nm emission. Alexa Fluor® 647 fluorescence was obtained with a 650 nm red laser and 660/20 nm emission, while Alexa Fluor® 633 was obtained with 633 nm red laser and 647 nm emission. The figures display the median of fluorescence intensity (MFI) relative to controls. Doublets were excluded by FSC-W × FSC-H and SSC-W × SSC-H analysis, and single stained channels were used to compensate. Fluorescent minus one was used for gating. 10,000-30,000 events were acquired per sample. Finally, data were processed with FACSDiva 6.3 software (BD Biosciences) or Summit 4.3 (Dako).
For sorting GFP
+ECs, GFP fluorescence was acquired using 488 nm blue laser and 510/50 nm emission and sorting was done using purity masks [
13]. For sorting BCCs, cancer cells were stained with a PE-conjugated dye called PKH26 (Sigma, USA) prior to co-culture and PE fluorescent was acquired using 496/566 nm blue laser and 576 nm emission to separate them from GFP
+ECs. Control GFP
+ECs or PKH
+BCCs monocultures were processed and sorted to normalize the cellular stress caused by cell sorting.
Wound healing assay
GFP
+ECs and PKH26
+BCCs were co-cultured under starvation for 3–5 days, and then the cells were sorted as described in the previous section. Sorted ECs or cancer cells were immediately plated and grown at 100% confluence in complete medium to recover overnight. Next, they were continued to culture under complete starvation for 6 hours to impede cellular growth before a wound healing assay was initiated [
25]. Finally, the migration capability of cells to close the wound (scratch) was evaluated after 48 hours using NIH ImageJ 64 software.
Growth factor reduced Matrigel (BD Biosciences, USA) was thawed at 4°C overnight, and added to each well of a 48-well plate (120 μL/well) and allowed to solidify for 30 minutes at 37°C. GFP+ECs were sorted from BCCs and immediately plated on matrigel at subconfluent density (2.5 × 104 cells/well). The formation of capillary-like structures was examined with an inverted microscope after 24 hours and the number of capillary junctions was quantified by analyzing the digitized images.
Immunocytochemistry
Antibodies against PE-CD31 (560983), AF647-VE-Cadherin (561567), and FN1 (610077), CD44 (555478) and desmin (550626) were purchased from BD Biosciences. F–actin (AF633-phalloidin, 22284) is a product of Invitrogen, whereas vimentin (5741) and α-SMA (ab5694) antibodies are products of Cell Signaling Technologies and Abcam, respectively. Anti-fade gold DAPI and secondary antibodies were purchased from Invitrogen. Cells were grown, stained, and imaged on glass chamber slides (Lab-Tek®). Briefly, the adherent cells were washed once with PBS and fixed in 3.7% formaldehyde, then permeabilized in 0.1% Triton X-100 for 20 minutes (no permeabilization required for cell surface proteins). After one wash, the cells were blocked for 30 minutes in a buffer containing 3% FBS and 1% BSA for one hour. Primary antibodies were prepared according to the instruction provided by the manufacturers and incubation was done for two to three hours on a shaker at normal temperature. After washing, the cells were incubated with secondary antibodies for 30 minutes. The fluorescent signals were acquired on a Zeiss Confocal Laser Scanning Microscope 710 (Carl Zeiss).
Immunohistochemistry
All antibodies are listed in the previous section. Formalin-fixed paraffin-embedded (FFPE) sections of neoplastic human breast biopsies were deparaffinized by dipping the slides in xylene for 15 minutes. The sections were rehydrated by immersing them in serial dilution of ethanol for 5 minutes followed by rinsing. Antigen retrieval was performed by boiling the slides in citrate buffer (pH 6.0) for 15 minutes. Snap frozen sections of human xenograft tumors were thawed briefly, fixed and permeabilized as described above. Primary antibody incubation was carried out overnight at 4°C in a moisture chamber after a 30-minute blocking period. Secondary antibodies were incubated for one hour followed by several washes. Slides were then mounted with DAPI.
shRNA transfection
Human shJagged1, scrambled lentiviral particles, and polybrene were purchased from Santa Cruz Biotechnology (USA). In summary, cells were cultured up to 50% confluence and were then treated with polybrene and lentiviral particles containing shRNA against Jagged1 or scrambled particles. Transfected cells were then selected using puromycin, and the down-regulation of Jagged1 was assessed by qPCR.
Total RNA was extracted with RNeasy Mini Kit (250) from Qiagen according to the manufacturer’s instructions. The RNA concentration was measured with Nanodrop 8000 spectrophotometer (Thermo Scientific) and 1 μg of RNA was used to produce cDNA with the ProtoScript M-MuLV
Taq RT-PCR kit using the oligo dT primers (New England BioLabs). Semi-quantitative real-time analysis (qPCR) was done with a 7500 qPCR System (Applied Biosystems, USA) using a Go
Taq 2-step RT-qPCR System (Promega) to amplify the gene of interest following the instructions provided. Primer sequences are listed in Additional file
1: Table S1.
Cells were lysed in RIPA buffer (Sigma) containing protease and phosphatase inhibitors. For each sample, 40 μg of total protein were analyzed by Western blot. Proteins were separated on 10% SDS polyacrylamide gels and electroblotted at 4°C onto polyvinylidene difluoride (PVDF) membranes for one hour. The membranes were blocked in 5% nonfat dry milk or bovine serum albumin (BSA) in 0.1% Tween 20 in Tris-buffered saline prior to incubation with primary antibodies at 4°C overnight. The antibodies included phospho-Smad5 (1:500, Abcam, ab76296), phospho-Smad3 (1:500, Bioss, bs-3425R), Smad5 (1:1000, Cell Signaling, 9517), Smad3 (1:1000, Cell Signaling, 9523), Hes1 (1:200, Millipore), and β-actin (1:3000, Sigma, A2228). Blots were developed using HRP and chemiluminescence peroxidase substrate (ImmunoCruz) (Santa Cruz Biotechnology) and FluroChem HD2 (Cell Biosciences).
RNA was isolated as explained above. Two quality control measures were carried out: (1) a spectrophotometric analysis and (2) a size fractionation procedure using a microfluidics instrument (Agilent Technologies). Total RNA (200 ng) was analyzed on Affymetrix GeneChip Human Genome U133 Plus 2.0 Array. Data were analyzed using Partek Software (V6.09.1110-6; Affimetrix), Venny online software (BioinfoGP; CNB-CSIC) and Ingenuity Pathway analysis (Ingenuity Systems, Redwood City, CA). Class comparisons between ECsNorm and ECsMes (three biological replicates of each) were performed to identify gene expression changes with significant expression differences (p < 0.05) and two-fold increase or decrease in expression.
All animal procedures were approved by the Ethics Committee for animal experimentation of Weill Cornell University (New York, USA). Six-week old female NOD/SCID mice were purchased from Jackson Laboratories. MDA-231 cells were injected (2 × 105) with or without 2 × 106 ECs (1:10 ratio) in the mice mammary fat pad of NOD/SCID mice. Four mice were assayed for each group. Each mouse received an injection of tumor cells on the left and co-injection of tumor and endothelial cells on the right side. The mice were euthanized and checked for tumor formation 18 and 30 days after inoculation. The extracted tumors were snap frozen for histological analysis.
Ingenuity pathway analysis
We used Ingenuity Pathway Analysis software (IPA) (Ingenuity Systems, Redwood City, CA) for network analysis of EC genes that were differentially regulated upon co-culture. A global gene list was defined representing IPA keywords: “Metastasis”, “Proliferation of cell lines” and “Cell death of tumor cell line”. All edges are supported by at least one reference from the literature, textbooks, or canonical information stored in the Ingenuity Pathways knowledge database. P-values for enrichment of biological functions were generated based on hypergeometric distribution and calculated with the right-tailed Fisher's exact for 2 × 2 contingency tables as implemented in Ingenuity.
Statistical analysis
All quantitative data are expressed as mean ± standard error of the mean (SEM). Statistical analysis and graphical presentation were performed using SigmaPlot 12 (Systat Software Inc., Chicago, IL) or Excel (Microsoft Corporation). A Shapiro-Wilk normality test, with a p = 0.05 rejection value, was used to test normal distribution of data prior to further analysis. All pairwise multiple comparisons were performed using one-way ANOVA followed by Holm-Sidak posthoc tests for data with normal distribution or, in case of a failed normality test, by Kruskal-Wallis analysis of variance on ranks followed by Tukey posthoc tests. Paired comparisons were performed using Student's t-tests or Mann–Whitney rank sum tests in case of unequal variance or failed normality test. Statistical significance was accepted for p < 0.05 (*), p < 0.01 (**) or p < 0.001 (***). All experiments were performed in triplicate and repeated three times (n = 3).
Discussion
Our main finding is that the tumor cells promote the acquisition of a transient contact-dependent mesenchymal phenotype in ECs contributing to the generation of a pro-tumoral niche. Transforming growth factorβ (TGFβ) and notch pathways seem to be determinant inducers of tumor-fostered mesenchymal phenotype in ECs.
The tumor microenvironment is implicated in the propagation and metastasis of several tumor types [
36-
38]. The role of endothelial cells (ECs) -as one the components of tumor stroma- in cancer development were merely thought to involve angiogenesis [
26]. Recently, our team demonstrated a novel role for tumor endothelium in tissue repair, self-renewal of HSCs as well as tumor growth and stemness by
angiocrine factors [
7,
8,
11,
13]. In addition, ECs were previously described as cells that demonstrated a high degree of plasticity under different conditions, a feature that is implicated in tumor development [
1,
6]. Tumor-associated endothelial plasticity may be explained in the context of spatiotemporal plasticity (i.e., to change phenotype and function) and reciprocity (i.e., by processing signals received from the environment) that have been earlier explained by Bissell’s group to be fundamental in step-wise changes in both tumor cells and their microenvironment [
39-
41]. Accordingly, plasticity and reciprocity account for the morphologic and functional heterogeneity driven by mechanisms such as cell-to-cell signaling to allow cells to cope with altering environmental conditions [
42-
44]. These mechanisms might be modulated by tumor and stromal cells crosstalk to co-evolve in a dynamic microenvironment [
44,
45]. Hence, our work emphasizes a crosstalk mechanism that is absolutely dependent on cell-to-cell contact between tumor and endothelium. The focal points of interaction between tumor and vasculature might not be abundantly present within the tumor bulk, but may potentially serve as miniature sites within the tumor microenvironment that may enhance neovascularization leading to increased tumor growth and metastasis.
In this study, we demonstrated that tumor cells are capable of promoting mesenchymal phenotype in their neighboring endothelium. In return, the ECs
Mes significantly contributed to tumor development. Although we initially observed this phenomenon in HUVECs, we continued our work with the widely used E4-ECs [
21] to circumvent the hurdle of tumor-induced HUVEC apoptosis in co-cultures that was also reported previously by Kebers et al. [
19]. The mesenchymal transformation in tumor endothelium in conjunction with loss of endothelial phenotype has been previously described in EndMT phenomenon as a mechanism for generation of CAFs [
17,
18]. However, the acquisition of mesenchymal phenotype with maintenance of endothelial trait and its significance for tumor propagation has never been explained earlier. The importance of this phenomenon may be explained by enhanced survival, mobility and angiocrine properties as well as the angiogenic ability of ECs
Mes. Our transcriptomics data further validate our hypothesis by showing up-regulation of pathways involved in cell development, signaling, and movement in addition to vascular system expansion and blood vessel formation.
The present work also focused on looking into the role of mesenchymal endothelium (ECs
Mes) in breast tumor progression.
In vivo, human xenograft tumor formation was enhanced by co-injection of ECs and tumor cells in NOD/SCID mice showing the up-regulation of mesenchymal markers in tumor-associated ECs. Also, by developing adherent (2D) and non-adherent (3D) co-culture systems, the role of ECs
Mes in cancer proliferation, stemness and invasiveness was evaluated. Based on a study by Maffini et al., the preliminary target for a carcinogen is tumor stroma and mutations in mammary epithelial cells are not sufficient for tumor initiation [
46]. Campbell’s group earlier demonstrated that a modified stroma preferentially promotes the outgrowth of abnormal epithelial cells [
44,
47]. Also, a study conducted by Moses and colleagues demonstrated that cell signaling abnormalities in stromal fibroblasts promoted mammary tumorigenesis in a non-cell-autonomous manner [
48,
49]. Therefore, it is primordial to study tumor microenvironmental changes that occur during cancer progression. In addition to influencing tumor initiation and progression, these changes significantly impact the efficacy of cancer therapy specifically when targeting stroma-regulated pathways [
50]. In accordance with these reports, our work highlighted the importance of tumor contexture in fostering phenotypic changes in ECs
Mes and how this alteration impact tumor proliferation, survival, stemness and pro-metastatic properties. Therefore, elucidation of the mechanisms underlying microenvironment alteration will be beneficial in targeting stroma to combat cancer.
TGFβ was previously suggested as an important role player in the EndMT process during normal and pathological situations [
30,
51,
52]. In addition, notch signaling was shown to promote EndMT during both cardiac development and oncogenic transformation [
32,
33]. In this study, we showed a similar phenomenon through which tumor cells enhanced mesenchymal phenotypes in ECs while preserving the endothelial phenotypes. We showed that the tumor-stimulated processes leading to creation of EC
Mes are also mediated by phosphorylation of TGFβ/Smad1/5 in synergy with notch pathway activation. Notch involvement in the regulation of TGFβ signaling in ECs was previously reported [
32,
34,
53,
54]. Both synergy and antagonism between notch and TGFβ signaling were described in ECs in a context-dependent manner [
34,
53-
59]. Our data confirmed that tumor-activated TGFβ/Smad1/5 phosphorylation was regulated by synergistic activation of notch and TGFβ pathway by showing that TGFβ and Jag1 were capable of inducing Smad5 phosphorylation as well as Hes1 up-regulation. Simultaneous inhibition of notch and TGFβ pathways not only impaired Smad5 phosphorylation but also impaired the acquisition of mesenchymal traits by ECs.
Earlier reports by Karsan’s group emphasized the importance of Smad1/5 phosphorylation in promoting proliferation and migration of ECs [
34]. Also, phosphorylation of Smad1/5/8 was implicated in ECs migration [
60]. Consistent with these observations, we demonstrated that tumor interaction with ECs stimulated TGFβ/Smad1/5 phosphorylation possibly resulting in gain of functional advantages by ECs
Mes. Since the acquired mesenchymal phenotypes in ECs
Mes seem to be a reversible phenomenon, targeted inhibition of the molecular modulators of this process may therefore be considered as potential therapeutic approaches.
Our findings of the tumor-promoted mesenchymal phenotype in ECs
Mes might partly explain the molecular mechanisms that govern ECs plasticity. Interestingly, our transcriptomics data while confirming mesenchymal traits, demonstrated modifications of over 1000 genes that might be relevant to tumor endothelial biology and play roles in phenomena such as resistance to treatment and metastasis. As such, up-regulation of previously described angiocrine factors such as IL6, Jag1, and CXCR4 [
61] in ECs
Mes may have potentially participated in the constitution of a pro-tumoral niche; however, the exact determinants of the angiocrine niche need to be clarified in future studies. While in our settings a change to ECs
Mes found to be transient, its permanence has not yet been carefully addressed in any study. However, the constitution of a transient niche might offer a window for the constitution of residual or resistant disease. Additional investigations involving
in vivo approach are required to validate our data in order to design new drugs for impairing tumor-EC interaction as a mean to treat cancer.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.
The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.
Competing interests
There are declared that they have no competing interests.
Authors’ contributions
PG Primary researcher contributed to conception and design, acquisition of data, analysis and interpretation of data and drafting the manuscript. JJ Contribution to in vivo work. JP Assistance in data analysis. MM Technical assistance with data acquisition for immunofluorescence imaging. NA Technical assistance with data acquisition for molecular biology. NH Support for analysis and interpretation of genomic data. BG Assistance with acquisition of cells. SR Involvement in critical revision of the manuscript. AR Principal investigator, contribution to the conception and design, manuscript revision, and final approval for the manuscript submission.
PG- Post-graduate PhD student, 1Weill Cornell Medical College in Qatar, Doha, Qatar
JJ- Postdoctoral associate, 2Weill Cornell Medical College, NY, USA
JP- Postdoctoral associate, 1Weill Cornell Medical College in Qatar, Doha, Qatar; 2Weill Cornell Medical College, NY, USA
MM- Technical assistant, 1Weill Cornell Medical College in Qatar, Doha, Qatar
NA- Technical assistant, 1Weill Cornell Medical College in Qatar, Doha, Qatar
NH- Postdoctoral associate, 1Weill Cornell Medical College in Qatar, Doha, Qatar; 2Weill Cornell Medical College, NY, USA
BG- Postdoctoral associate, 1Weill Cornell Medical College in Qatar, Doha, Qatar; 2Weill Cornell Medical College, NY, USA
SR- Principal investigator, 2Weill Cornell Medical College, NY, USA
AR- Principal investigator, 1Weill Cornell Medical College in Qatar, Doha, Qatar; 2Weill Cornell Medical College, NY, USA.