Transcription Profiles of Endothelial Cells in the Rat Ductus Arteriosus during a Perinatal Period

Endothelial cells (ECs) lining the blood vessels serve a variety of functions and play a central role in the homeostasis of the circulatory system. Since the ductus arteriosus (DA) has different arterial characteristics from its connecting vessels, we hypothesized that ECs of the DA exhibited a unique gene profile involved in the regulation of DA-specific morphology and function. Using a fluorescence-activated cell sorter, we isolated ECs from pooled tissues from the DA or the descending aorta of Wistar rat fetuses at full-term of gestation (F group) or neonates 30 minutes after birth (N group). Using anti-CD31 and anti-CD45 antibodies as cell surface markers for ECs and hematopoietic derived cells, respectively, cDNAs from the CD31-positive and CD45-negative cells were hybridized to the Affymetrix GeneChip® Rat Gene 1.0 ST Array. Among 26,469 gene-level probe sets, 82 genes in the F group and 81 genes in the N group were expressed at higher levels in DA ECs than in aortic ECs (p<0.05, fold change>2.0). In addition to well-known endothelium-enriched genes such as Tgfb2 and Vegfa, novel DA endothelium-dominant genes including Slc38a1, Capn6, and Lrat were discovered. Enrichment analysis using GeneGo MetaCore software showed that DA endothelium-related biological processes were involved in morphogenesis and development. We identified many overlapping genes in each process including neural crest-related genes (Hoxa1, Hoxa4, and Hand2, etc) and the second heart field-related genes (Tbx1, Isl1, and Fgf10, etc). Moreover, we found that regulation of epithelial-to-mesenchymal transition, cell adhesion, and retinol metabolism are the active pathways involved in the network via potential interactions with many of the identified genes to form DA-specific endothelia. In conclusion, the present study uncovered several significant differences of the transcriptional profile between the DA and aortic ECs. Newly identified DA endothelium-dominant genes may play an important role in DA-specific functional and morphologic characteristics.


Introduction
A wide variety of vascular biology studies in the past several decades have verified that the vascular endothelium plays a critical role in the homeostasis of the cardiovascular system. Endothelial cells (ECs) lining blood vessels form monolayers to cover the inner lumen of all types of the vessels in the normal state. The vascular endothelium regulates vasomotor tone by secretion of several vasoactive substances such as nitric oxide. It also mediates leukocyte trafficking and monocyte activation to control platelet adhesion and coagulation. In addition, it interacts with vascular smooth muscle to mediate proliferation and differentiation of smooth muscle cells (SMCs). ECs respond diversely to a variety of external or internal stimuli to alter membrane permeability, transcellular transport systems, membrane adhesive molecules, various growth factor secretions, and so on [1,2]. These responses occur in order to fulfill the needs of tissues and maintain the homeostasis of the circulatory system. Endothelial phenotypic heterogeneity also plays an important role in the remodeling of the cardiovascular system where specific ECs are localized [3,4].
The ductus arteriosus (DA), a fetal shunt artery between the pulmonary artery and the aorta, closes promptly after birth, although its connecting arteries remain open. The DA exhibits characteristics that are distinct from those of its connecting arteries. For example, the DA is more sensitive to the change in circulating oxygen concentration and prostaglandin E 2 [5][6][7]. Since the changes in circulating oxygen concentration and prostaglandin E 2 directly transduce the intravascular lumen where ECs surround its surface, ECs of the DA must play an important role in regulating these distinct characteristics. Accordingly, several studies have demonstrated the endothelium-dependent or independent vasoreaction of the DA [8][9][10][11]. Nonetheless, the majority of previous studies investigating molecular events in the DA utilized the whole DA tissue or cultured SMCs. Therefore, the role of DA ECs remains largely unknown. We hypothesized that the ECs of the DA exhibit a unique gene profile involved in DA-specific vasoconstriction and vascular remodeling. Recently, Weber et al. reported that they successfully isolated ECs from fetal rat DA using the immunomagnetic cell separation method and harvested the isolated ECs to further confirm their purity by flow cytometry analysis [12]. In the present study, we investigated gene expression differences in ECs between the DA and the aorta by using a combination of fluorescence-activated cell sorter (FACS) and DNA microarray experiments followed by further enrichment analysis using MetaCore GeneGo software.

Animals
Timed-pregnant Wistar rats were purchased from Japan SLC, Inc. (Shizuoka, Japan). Rat fetuses at the 21st day of gestation as full-term were divided into two groups: fetuses before breathing (F group) and neonates obtained 30 minutes after breathing (N group). Animals in both groups were delivered by cesarean section. All animals were cared for in compliance with the American Physiological Society. The experiments were approved by the Ethical Committee on Animal Experiments of Waseda University.

Fluorescence Activated Cell Sorter (FACS)
Pooled tissues from the DA or the aorta were obtained from three litters of timed-pregnant Wistar rats, which accounted for approximately thirty fetuses. Tissues were treated with collagenase-dispase enzyme mixture as described previously [13]. Approximately 1.0610 6 cells were obtained from combined whole DA tissues from the three litters. These cells were reacted with FITC-conjugated anti-CD31 and APC/Cy7-conjugated anti-CD45 antibodies as cell surface markers for EC and hematopoietic derivation cells, respectively. In order to confirm the nonspecific binding of antibodies to cells, we also prepared the cells reacted with a fluorescence conjugated anti-control IgG antibody. When dead cells reacted with PI solutions (Dojindo, Kumamoto, Japan), we found that about 30% of the cells (3.0610 5 cells) stained with PI solutions had died during isolation. The dead cells were then removed from further analysis. All cells were detected and sorted using BD FACSAria TM II (Becton Dickinson, San Jose, CA, USA). In order to obtain as many cells as possible without affecting their purity, we set the FACS's sorting mode on ''purity'' in our cell sorter, which allowed the machine to automatically abort any possible forms of contamination, such as doublet cells, and to yield a purity of approximately more than 99%. In addition, we used the 100 mm size of the nozzle to decrease its sorting speed, which may contribute to an increase in accuracy. Therefore, we did not check the purity of the sorted endothelial cells by re-run, because the amount of ECs from rat DA was extremely small. The sorted cells were received by a 1.5 ml centrifuge tube containing 500 ml of DPBS (Wako, Osaka, Japan) and 0.4 ml of RNase inhibitor (Roche, Meylan, France).

Quantitative real-time PCR (qRT-PCR)
The cell suspensions sorted by FACS were centrifuged at 3006g for 15 minutes, and the precipitation was quickly frozen in liquid nitrogen. Total RNA was extracted from the collected cells using an RNeasy micro kit (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions. Total RNA was reverse-transcribed to cDNA using a High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA). For quantitative RT-PCR analysis, sequences for PCR primers are listed in Table 1. qRT-PCR was performed using a Step One Real-time PCR System (Applied Biosystems) with Fast SYBR Green Master Mix (Applied Biosystems). The abundance of each gene was determined relative to an internal control using 18S rRNA. For each qRT-PCR experiment, which included an RTnegative control, we confirmed there was no non-specific amplification in any reaction.

DNA microarray procedure
We repeated FACS sorting ten times for each developmental group (30 litters used in total) in order to accumulate enough total RNA (,100 ng). Then, cDNA was generated using the WT Expression Kit (Ambion, Austin, TX, USA) in accordance with the manufacturer's protocol. Briefly, a total of 100 ng of total RNA was reverse-transcribed to cDNA, which was subsequently used as a template for an in vitro transcription reaction. Sensestrand cDNA that contains dUTP was synthesized by amplified cRNA. We used the Affymetrix GeneChipH WT Terminal Labeling Kit (Affymetrix, Santa Clara, CA, USA) to recognize the dUTP and to fragment the cDNA with uracil-DNA glycosylase   (UDG) and apurinic/apyrimidinic endonuclease 1 (APE1). These fragmented cDNAs were then labeled through a terminal deoxytransferase reaction and hybridized to the Affymetrix GeneChipH Rat Gene 1.0 ST Array (Affymetrix). The hybridization experiments were performed in triplicate (approximately 180 litters were needed in total), and the intensities were averaged.

Microarray data analysis
Of the 26,469 genes on the microarray, 14,944 were excluded based on aberrant low signals as determined by the poly-A spike of lys (probe set ID: 10700066) expression, the smallest composition out of the poly-A RNA control cocktail, which was added in each total RNA sample. All remaining gene probes were analyzed for their differential expression between the DA and the aorta at each developmental stage. Initially, we calculated the p value by Student's t-test across each group, and the data were cut off at  Table 2 and Table 3, respectively. The color scale is based on their expression intensities. The green or red color indicates the lowest or the highest expression levels, respectively. The midpoint shown as a dark color represents 235 since it is the average of whole gene expression. doi:10.1371/journal.pone.0073685.g003   p,0.05. For a more robust differential analysis between the DA and the aorta, we selected the genes that had more than a 2.0-fold change (|FC| §2.0). Genes that went through these analyses were considered significant. Genes were further analyzed for enriched biological themes and pathways using the MetaCore program (GeneGo, a division of Thomson Reuters, St. Joseph, MI, USA). The program ranked the significant ontology and pathways dominant in DA ECs in each developmental stage by importing whole expression results (excluding aberrantly low expressed genes) from the microarray. MetaCore is an established program that includes a manually annotated database of gene interactions and metabolic reactions obtained from scientific literature. The enrichment analysis of the biological process was based on the hypergeometric distribution algorithm and relevant pathway maps were then prioritized according to their statistical significance [14].
The complete data set of the DNA microarray is available in the GEO database (accession number: GSE40500).

Statistical treatment
Data are presented as mean 6 standard error (SEM) or independent experiments. Statistical analyses were performed between two groups by unpaired two-tailed t test or unpaired t test with Welch correction, and among multiple groups by one-way analysis of variance (ANOVA) followed by Neuman-Keuls multiple comparison test. A p value of ,0.05 was considered significant.

Endothelial cells were purely isolated from rat DA tissues
At least 10,000 of the cells (approximately 1% of the initially isolated cells) were sorted in anti-CD31 positive and anti-CD45 negative areas (CD31 + /CD45 2 ) from the pooled DA tissues of three litters of timed-pregnant Wistar rats ( Figure 1A). No cell in the CD31 + area reacted with an anti-IgG antibody ( Figure 1B), indicating that no false positive cells were contained in the CD31 + /CD45 2 cells believed to be ECs. We also assumed that CD31 2 /CD45 2 cells mainly consisted of SMCs. The detailed gating strategies of FACS sorting are shown in Figure S1. To confirm the purity of FACS isolation, we examined the expression levels of EC-specific and SMC-specific genes by qRT-PCR. The expression levels of Tie2 mRNA, an EC-specific gene, were significantly higher in CD31 + /CD45 2 cells than in CD31 2 / CD45 2 cells (p,0.05, n = 5) (Figure 2A). The expression levels of Sixty five genes in the F group and 52 genes in the N group were expressed more than 2-fold in ECs of the aorta than in ECs of the DA (p,0.05). Among these aorta dominant genes, 43 genes were expressed more than 2-fold in ECs of the aorta in both groups(above the thick line). F: fetuses before breathing; N: neonates obtained 30 minutes after breathing. doi:10.1371/journal.pone.0073685.t003 c2-actin and endothelin-1 receptor Ednra mRNAs, SMC-specific genes, were significantly lower in CD31 + /CD45 2 cells than in CD31 2 /CD45 2 cells (p,0.001, n = 3,5) ( Figures 2B and 2C, respectively). Therefore, we concluded that a FACS could isolate pure ECs in the CD31 + /CD45 2 area without contamination.

Identification of DA-specific genes in ECs
Among over 26,469 gene-level probe sets, we found that 82 genes in the F group and 81 genes in the N group were expressed more than 2-fold in ECs of the DA than in ECs of the aorta (p,0.05) (Table 2, Figure 3A). Among these DA dominant genes, 71 genes were expressed more than 2-fold in ECs of the DA in both groups. On the other hand, 65 genes in the F group and 52 genes in the N group were expressed more than 2-fold in ECs of the aorta than in ECs of the DA (p,0.05) (Table 3, Figure 3B). Among these aorta dominant genes, 43 genes were expressed more than 2-fold in ECs of the aorta in both groups. Importantly, the majority of the genes in Table 2 and 3 have never been reported previously as DA-related genes. We found only a limited number of the well-known endothelium-related genes such as transforming growth factor-beta 2 (Tgfb2) and vascular endothelial growth factor A (Vegfa). Validation of the results from the DNA microarray, qRT-PCR was performed with Slc38a1 and Lrat, the genes with the most significant difference between the DA and the aorta at both developmental stages ( Figure S2).
Surprisingly, the present results showed remarkably low variations of transcription profiles before and after birth. Although there were 178 genes of which expression levels significantly differed between both developmental stages (p,0.05), arrestin domain containing 3 (Arrdc3) and TBC1 domain family, member 30 (Tbc1d30) were the only two genes that had more than a 2.0fold change (|FC| §2.0) between both developmental stages. Among 178 genes, Table 4 shows 25 genes of which the p values were less than 0.01 (p,0.01). Among these 25 genes, connective tissue growth factor (Ctgf) and Tbc1d30 are listed in Table 2 (DA-dominant) and Table 3 (aorta-dominant), respectively.

Enrichment analysis of DA dominant genes using GeneGo MetaCore software
In the MetaCore systems, there are about 110 cellular and molecular processes whose content is defined and annotated by Twenty five genes shows a significant difference between F and N groups in the DA (p values were less than 0.01. Among these 25 genes, Ctgf and Tbc1d30 are listed in Table 2 and GeneGo. The top 10 ranked regulatory biological processes were listed in each stage of the DA ECs based upon their p values ( Table 5). Most of the categories indicate morphogenesis and development. Four processes (anatomical structure morphogenesis, cardiovascular system development, circulatory system development, and locomotion) are ranked in both the F and N groups. Interestingly, excluding processes related to morphogenesis and development, regulation of phosphatidylinositol dephosphoryla-tion is an enriched process that is listed only in the F group. On the other hand, response to external stimulus, response to vitamin A stimulus, and axon guidance were listed only in the top 10 ranked biological processes in the N group. In these GeneGo biological processes, 322 and 172 genes were listed in the F and N groups, respectively. The genes included in each category are shown in Figure 4. There are a considerable number of overlapping genes in each process. Table 6 shows the 30 genes that frequently  Table 5). The color scale table imitating heat maps was created manually based on the genes in GO biological processes. A) The genes in all the top processes except the development or morphogenesis processes that emerged in both F and N. B) The genes categorized in the processes related to the development and morphogenesis in both F and N. C) The genes categorized only in cardiovascular or circulatory specific development processes. The color scale is the same as that used in Figure 3. All heat maps were created manually based on the genes in the GeneGo biological processes. a) The genes in all the processes in Table 4, except the development or morphogenesis processes which emerged in both F and N. b) The genes categorized in the processes related to development and morphogenesis in both F and N. c) The genes categorized only in cardiovascular or circulatory specific development processes. doi:10.1371/journal.pone.0073685.g004 appeared in more than five processes of the top 10 ranking as active genes. These genes are likely to be involved in the network by potential interactions with many of the identified genes to form DA-specific endothelium. Furthermore, there are over 1200 pathway maps in MetaCore, comprehensively covering signaling and metabolism, selected diseases and some drug targets mechanisms. All maps are accurately drawn by GeneGo annotators and manually curated and edited. The canonical pathway maps and GeneGo process networks, validated by statistical values, were evaluated by MetaCore and are listed in Table 7 together with the top 10 ranking for each pathway significantly worked in the DA ECs. As we found that the gene expression profiles exhibited remarkably low variations at both time points, nine of the top 10 ranked pathway maps were listed in both F and N groups. These categories are related to regulation of epithelial-to-mesenchymal transition (EMT), cell adhesion, and retinol metabolism.

Discussion
To date, the characteristic features of the DA endothelium have remained largely unknown. Several studies have demonstrated the endothelium-dependent or independent vasomotor reaction of the DA [8][9][10][11]. Rabinovitch et al. made great efforts to identify the role of the DA endothelium in vascular remodeling of the DA. They found that the increase in the expression of transforming growth factor-beta (Tgfb) in the DA endothelium promoted the synthesis of glycosaminoglycan such as hyaluronan that is a critical regulator of neointimal formation of the DA [15,16]. The present comprehensive gene expression analysis identified a DA endothelium-dominant rat gene profile during a perinatal period for the first time. It should be noted that we collected DA ECs from more than 30 litters to obtain a sufficient amount of mRNA for one sample. Additionally, the hybridization experiments were performed in triplicate and the intensities were averaged. To avoid an unexpected artificial bias, we did not use cultured ECs or any amplification method to increase mRNA from the endothelium. Therefore, the present study represents the transcription profile of the freshly isolated endothelium from the rat DA. Importantly, most of the genes that were expressed greater or lower in the DA endothelium than in the aortic endothelium have not yet been investigated in the DA. In addition to the up-regulated or downregulated genes that met the 2-fold threshold in Table 2 and 3, one may be interested in genes that showed a statistical significance but a lower than 2.0 difference. We therefore also listed the genes with a statistical significance (p,0.001) in Table  S1 and Table S2. Since the endothelium plays a critical role not only in vascular tone but also in vascular remodeling, the newly identified genes should be of great interest for further investigation of the molecular mechanisms of DA-specific differentiation and function.
Although two studies, including our previous one, have identified DA-dominant genes using DNA microarray analysis [17,18], the present study revealed that the transcription profile of DA ECs are quite different from those of DA whole tissues of which the majority is composed of SMCs. These data suggest that the transcription profiles of the DA endothelium are tightly regulated in a cell-specific manner. Furthermore, local interaction between ECs and SMCs may contribute to establishing each unique transcription profile in the DA. It would be beneficial to further investigate how ECs and SMCs interact locally with each other. To our surprise, the present study also demonstrated that the transcription profile of DA ECs did not change much before and after birth, although the DA does dramatically alter its morphology during the perinatal period. After birth, the change in oxygen and PGE 2 content in circulating blood induces functional closure of the DA. Costa et al. also demonstrated that the transcription profile of DA tissues significantly differs before and after birth [18]. In their experiment, DA samples were collected 3 hours after spontaneous delivery. We used neonatal DA ECs 30 minutes after delivery by cesarean section, because we aimed to detect an initial change in the transcription profile of DA ECs after birth. This period, however, may not be long enough to investigate the alternation in its transcription profile, although functional closure of the rat DA had mostly occurred in our previous studies [19,20].
It is very important to investigate the roles of the newly identified genes in the morphology and function of the DA. Unfortunately, an in vitro experiment using rat DA ECs is technically very difficult because of the limited amount of tissue Table 6. Thirty overlapping genes that appeared in more than five processes of the top ten ranking as active genes.

ID
Gene Symbol mRNA-Description Number of overlapped processes F N or cells obtainable from small animals. Currently, bioinformatic technology has developed to the point that it is now possible to attribute functions to genes and their encoded proteins, and to identify the regulatory networks controlling metabolic, protein synthesis and signal transduction pathways. To facilitate the analysis of experiments using post-genomic technologies, newly developed knowledge-based gene set enrichment analysis provides a powerful analytical method to link the vast amount of raw data to biological pathways [14,21]. Pathway analysis by MetaCore is based on the concept that the function of a gene depends directly on the context in which it acts, and MetaCore correlates genes identified by DNA microarray with the cellular pathways that are hypothetically activated dominantly in the DA endothelium. The present study revealed possible regulatory factors involved in specific ''process networks'' such as the regulation of morphogenesis and development. We found a considerable number of overlapping genes in these processes that are likely to be involved in the network through potential interactions with many of the identified genes. The genes listed in Table 6 are considered the ''functional hubs'' of the DA endothelium-specific network. Eight transcription factors in Table 6 that have been known to be involved in formation of the cardiovascular system may play an important role in forming the endothelial phenotypic heterogeneity of the DA, although none of them has been intensively investigated in the DA. Although several previous studies have demonstrated that the transcription factors Tfap2b, Hif2a, and myocardin play an important role in ductal smooth muscle development [22,23], the transcriptional regulation of DA endothelium-specific differentiation remains largely unknown. The DA originates from the sixth pharyngeal arch artery that initially forms as protuberances of the dorsal aorta and the aortic sac [24]. The cells that comprise the pharyngeal arch arteries are of pharyngeal mesoderm origin. The mesodermal core of the arches is continuous with the mesoderm derived from the second heart field (SHF) [25]. To date, the majority of the cells that constitute the DA media are known to derive from cardiac neural crest cells (NCCs) at the somite 1 to somite 3 level [26,27]. The importance of this neural crest origin in understanding specific DA differentiation lies in the segmental nature of the pharyngeal arches themselves and of the origin of the NCCs that invade them. Accordingly, transcription factors related to NCCs such as Hoxa1, Hoxa4, and heart and neural crest derivatives expressed 2 (Hand2) [28][29][30] are listed in Table 6. Although a previous study suggested that Hoxb5 may be involved in DA differentiation [31], the expression level of Hoxb5 mRNA was not increased in the DA ECs in the present study. A recent study in humans revealed that mutations in Hoxa1 can cause severe cardiovascular malformations in patients with Bosley-Salih-Alorainy Syndrome [32]. Furthermore, Hoxa1 null mice show defects such as interrupted aortic arch, aberrant subclavian artery and tetralogy of Fallot, demonstrating that Hoxa1 is required for patterning of the great arteries and outflow tract of the heart [28].
In addition, MetaCore enrichment analysis revealed that the SHF-related transcription factors T-box (Tbx) 1, Tbx18, and Isl1, and the receptor ligand Fgf10 are enriched in the DA endothelium. It has also been shown that Isl1, Fgf10-positive mesoderm of the posterior arches forms the ECs of arterial blood vessels 25 . Although Rana et al. demonstrated that the endothelium of the pharyngeal arch arteries is largely negative for Tbx1 [33], its expression levels were greater in the DA ECs than in the aortic ECs in the present study. Therefore, the results indicated that the SHF-derived cells are more prevalent in the DA endothelium than in the descending aortic endothelium. To our knowledge, no study has reported that the SHF-derived cells contribute to DA differentiation. Isl1 and Tbx1 regulate Fgf10 transcription in the SHF during cardiac outflow formation [34,35]. It would be of value to further investigate the interaction between the NCCderived and SHF-derived cells in the DA. Furthermore, MetaCore enrichment analysis identified that the epithelial-to-mesenchymal transition (EMT) pathway functional in the DA ECs. The key genes Tgfb2, actin, alpha 2, smooth muscle, aorta (ACTA2), N-cadherin (cadherin 2, Cdh2), and met protooncogene (hepatocyte growth factor receptor, Met) are listed in the pathway. One of the characteristics of the DA vascular remodeling is physiological intimal thickness that is profound during the perinatal period. Within certain environments, endothelial-tomesenchymal transition (EnMT) plays a role in promoting arterial intimal hyperplasia [36,37]. EnMT shares a number of molecular signaling pathways with Tgfb1-induced EMT [38,39], although it has not been proven that migrated ECs of the DA into the subendothelium differentiate into smooth muscle-like (mesenchymal) cells. Gittenberger-de Groot at al. demonstrated that the invaginated ECs were located in the subendothelial region [40]. In addition, they found that a rat patent DA model exhibited an abundant subendothelial elastic lamina and failure of intimal formation [41]. Therefore, they suggest that some of the intimalmesenchymal cells are also derived from the invaginated ECs [42]. In agreement with their idea, the present data suggest that EnMTrelated signal pathways can be active in the DA endothelium and contribute to physiological intimal thickness of the DA.
MetaCore enrichment analysis also revealed that the retinol metabolism pathway worked in the DA ECs. The key genes lecithin:retinol acyltransferase (Lrat), aldehyde dehydrogenase 1 family, member A1 (Aldh1a1), and Aldh1a7 are listed in the pathway. It is noteworthy that the endogenous retinol signaling may play a role in inducing and maintaining smooth muscle differentiation in the DA [43] and that maternal vitamin A accelerates vascular maturation of the DA in premature fetuses, at least in terms of vascular contraction [44,45]. Furthermore, our previous study showed that maternally administrated vitamin A increased fibronectin production and intimal thickening in the rat DA [46]. Therefore, we may reconsider vitamin A as a candidate for treatment of patent DA, even though one study demonstrated that postnatal vitamin A therapy did not have a beneficial effect on ductal closure in premature infants [47]. Indeed, there are several studies demonstrating that vitamin A induces various embryonic developments via many different pathways such as Tgfb2, Cdh2, or Pitx2 [48]. Recently, Amengual at al. identified that Lrat is critical for cellular uptake of vitamin A from serum retinol-binding protein [49].
In conclusion, the present comprehensive transcription analysis identified the novel DA endothelium-dominant genes during a perinatal period that are highly related with biological processes involved in morphogenesis and development. Moreover, we found that regulation of epithelial-to-mesenchymal transition, cell adhesion, and retinol metabolism are the active pathways that form DA-specific endothelium. Newly identified DA endotheliumdominant genes may play an important role in DA-specific functional and morphologic characteristics.