| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Cardiovascular Research Institute, Comprehensive Cancer Center, and Department of Anatomy, University of California, San Francisco, California 94143-0452
Submitted 13 November 2002 ; accepted in final form 27 June 2003
 |
ABSTRACT |
|---|
 TOP ABSTRACT MATERIALS AND METHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
blood vessels; cationic liposomes; immunohistochemistry; gene profiling; oligonucleotide arrays
One major challenge in gene profiling of endothelial cells in organs or tumors is separating the cells from other cell types. Selective culture conditions (68, 70) and lectins or antibodies attached to magnetic beads (20, 22, 43, 44, 51, 74, 106) have proven useful in this regard. However, incomplete specificity for endothelial cells and removal of cell surface epitopes during dissociation are limitations of this approach. Another challenge is to minimize changes in expression that occur during the isolation process. Here, time is an important factor. Methods using multiple levels of selection with antibodies conjugated to magnetic beads involve multiple steps that may last 12 h or longer (22, 106). Cell culture requires even more time and imposes in vitro conditions that increase the likelihood of changes in gene expression. Another factor is the number of cells needed for gene expression measurements dictated by the sensitivity of the assay. When only a small number of cells are available, methods used for gene expression profiling may not be sufficiently sensitive to detect basal levels of expression.
The purpose of the present study was to identify genes expressed in normal microvascular endothelial cells of the lung. To deal with the issues of specific cell labeling, speed of isolation, and number of cells isolated, we developed a method based on the natural capacity of the endothelial cells of lung capillaries to bind and internalize fluorescent cationic liposomes from the bloodstream (76). Because the liposomes do not cross the endothelium, cells outside the vasculature are not labeled (76, 114). In the past, this property has been used to target substances selectively to certain types of endothelial cells, while sparing extravascular cells (76, 114). However, we reasoned that the same property could be used to label endothelial cells in mouse lungs for subsequent isolation and purification. After dissociation of the tissue, the fluorescent endothelial cells could be purified by using fluorescence-activated cell sorting (FACS) to yield a large number of endothelial cells rapidly and without the need for cell culture. Although intravascular leukocytes may also take up cationic liposomes (76), surface markers could be used to remove these cells by negative selection during FACS purification.
RNA from freshly isolated endothelial cells can be analyzed on oligonucleotide microarrays covering tens of thousands of genes or expressed sequence tags (ESTs) (64). In the present study, performed through a collaboration with Eos Biotechnology (South San Francisco, CA), we used Affymetrix (Santa Clara, CA) oligonucleotide microarrays covering 39,000 genes and ESTs in the mouse genome to investigate gene expression in endothelial cells isolated from normal lung capillaries. Genes with greater expression in endothelial cells were identified by relating expression in purified endothelial cells to the expression in unpurified lung cells. Expression of a representative selection of genes in endothelial cells was validated by immunohistochemistry on lung sections. Based on stringent selection criteria, the study revealed a profile of 234 known genes and 321 ESTs enriched in lung endothelial cells that contribute to their distinctive functions.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
Preparation and injection of fluorescent cationic liposomes. Fluorescent cationic liposomes were prepared by mixing 1,2-dioleoyl-3-trimethylammonium-propane with cholesterol (Avanti Polar Lipids; Alabaster, AL) and 1,2-dipalmitoyl-N-glycero-3-phosphoethanolamine-lissamine-rhodamine-B-sulfonyl chloride (Molecular Probes; Eugene, OR) in chloroform at a molar ratio of 55:44:1. Small unilamellar vesicles containing 10 mM total lipid in 5% glucose were prepared by sonication, filtered to sterility, and stored under argon as previously described (76, 114). Rhodamine-labeled liposomes (
1 mM total lipid in 100 µl) were injected via a femoral vein into anesthetized mice (25 mg/kg ip pentobarbital sodium, Nembutal; Abbott Laboratories, N. Chicago, IL) and circulated for 20 min (76).
Verification of labeling of pulmonary endothelial cells. In initial validation studies, mice received an intravenous injection of rhodamine-labeled liposomes followed 18 min later by FITC-labeled Lycopersicon esculentum lectin (5 mg/kg iv; Vector Laboratories, Burlingame, CA) to mark pulmonary vascular endothelial cells in which liposomes were taken up (76). Two minutes later, the lungs were fixed by vascular perfusion of 1% paraformaldehyde in PBS (pH 7.4) for 1 min at a pressure of 30 mmHg via a cannula in the right ventricle (76). The lungs were inflated through the trachea with 3% SeaPlaque warm low-melting-point agarose (BioWhittaker Molecular Applications, Rockland, ME), excised, and cut into 100-µm sections with a Vibratome (The Vibratome; St. Louis, MO). Tissue sections were mounted in antifading mounting medium (Vectashield; Vector Laboratories) on glass microscope slides and examined with a Zeiss Axiophot fluorescence microscope or Zeiss LSM-410 confocal microscope. Although effects of liposome uptake on gene expression in endothelial cells cannot be excluded, no consistent differences in expression were found in pilot experiments by using whole lung RNA on 8000-oligonucleotide arrays to compare mRNA from mice that received an intravenous injection of liposomes 20 min earlier with mRNA from mice that had not received liposomes (data not shown).
Endothelial cell isolation. In mice used for gene expression studies, fluorescent cationic liposomes were injected intravenously, and 20 min later, the lung vasculature was perfused in situ for 2 min with 1% collagenase D (Boehringer-Mannheim, Indianapolis, IN) in HBSS containing 2 mM Ca2+ and 2 mM Mn2+, supplemented with 0.5% BSA at 37°C at a pressure of 30 mmHg, to begin the enzymatic digestion in situ. In each of three identical experiments, the lungs of two mice were excised, minced, and digested further in collagenase (1%, RNase-free, Boehringer-Mannheim) and DNase (1 U/ml, Boehringer-Mannheim) for 20 min at 37°C. Cell clumps, debris, and undigested pieces of lung were removed by filtration through a 40-µm mesh. The cells were washed twice with cold HBSS and then incubated in FITC-conjugated leukocyte specific anti-mouse CD18 (LFA-1/
2 integrin) antibody (Pharmingen; San Diego, CA) for 15 min at 4°C, and washed twice. The CD18 antibody was used for negative selection by FACS to remove intravascular leukocytes that took up liposomes.
Cells were sorted on a FACS Vantage flow cytometer (Becton, Dickinson; San Jose, CA). FITC was detected in the fluorescence 1 (FL1) channel and rhodamine, in the FL2 channel. Forward scatter (FSC) and side scatter (SSC) were measured by using logarithmic and linear amplifiers, respectively. Initially, 50,000 events were collected and analyzed by using CellQuest Software (Becton Dickinson). Two identical regions, R1 and R2, were set to delimit two cell populations (Fig. 1). Gate R1 defined the entire population of rhodamine-liposome labeled cells. Gate R2 defined the subset of population R1 that did not bind the CD18 antibody. Both gates were set to include cells with red fluorescence values >10 intensity units and green fluorescence <25 units. Most events with green fluorescence values greater than this threshold, regardless of their red fluorescence, were CD18-immunoreactive cells, autofluorescent cells, or debris.
|
Rhodamine-positive/FITC negative cells gated in R2, which represented one population in the FSC/SSC profile, was the putative endothelial cell population. In each experiment, this population consisted of 700,000 cells that had an estimated diameter of 4–14 µm as measured with calibrated microspheres. The cells were isolated for gene expression profiling by sorting directly into TRIzol (Life Technologies; Rockville, MD) over a 1-h period. Subsequent samples were obtained for morphological analysis. Average preparation time from perfusion of the lungs until the isolated cells were in TRIzol was 3.5 h.
Viability, number, and morphology of isolated cells. At the end of the sort, a sample of cells in population R2 (10,000 cells) was sorted into PBS and then diluted 1:40 to assess cell viability by Trypan blue exclusion. On average, 62 ± 5% of the cells excluded Trypan blue. On the basis of this value, an average of 440,000 ± 30,000 viable cells (700,000 total cells) were obtained from the lungs of each pair of mice (n = 3 groups of two mice each, means ± SE). Another sample was saved for staining with Diff-Quik (Dade Behring; Deerfield, IL) for examination of cell morphology. Sorted cell populations R1 and R2 were deposited onto glass poly-L-lysine-coated glass microscope slides with a Cytospin centrifuge (Shandon; Pittsburgh, PA) and stained with Diff-Quik to show the cytoplasm and nucleus. Neutrophils were identified by their polymorphic nucleus, and monocytes were identified by their large size and kidney-shaped nucleus.
Identification of isolated cells by immunohistochemistry. To assess the FACS parameters, samples of isolated cells, gated for regions R1 or R2, were allowed to adhere to poly-L-lysine-coated glass microscope slides, fixed with 1% paraformaldehyde in PBS, and examined for liposome fluorescence or immunohistochemical markers by fluorescence and confocal microscopy. Uptake of fluorescent cationic liposomes was examined in specimens mounted in aqueous Gel/Mount (Biomeda; Foster City, CA) without further staining. In immunohistochemical studies, endothelial cells were identified by using a rat monoclonal anti-mouse CD31 antibody (platelet endothelial cell adhesion molecule, Clone MEC 13.3; Pharmingen) (62) and rabbit polyclonal vascular endothelial (VE)-cadherin antibody (Dr. Elisabetta Dejana, Institute Negri, Milan, Italy) (17). Before VE-cadherin immunohistochemistry, cells were permeabilized with 0.1% Triton X-100 in HEPES-buffered saline to expose the cytoplasmic tail of the molecule (17). Leukocytes were identified by using a rat monoclonal anti-mouse CD45.2 antibody (Clone 104, Pharmingen). Slides were incubated in primary antibodies (1:200 dilution) at room temperature for 15 min, washed 3 times with HEPES-buffered saline, incubated in Cy3-goat anti-rabbit IgG or Cy2-goat anti-rat IgG secondary antibody (1: 800 dilution; Amersham Life Science, Pittsburgh, PA) at room temperature for 60 min, and mounted in Vectashield (Vector Laboratories).
Uptake of acetylated LDL. As a functional assay, samples of population R2 were tested for uptake of fluorescent acetylated LDL (acetyl-LDL). Sorted cells were incubated for 3 h at room temperature in the presence of 10 mg/ml BODIPY-labeled acetyl-LDL (Molecular Probes) in HBSS supplemented with 10% FBS. After incubation, the cells were washed twice, allowed to adhere to poly-L-lysine-coated slides, fixed with 1% paraformaldehyde in PBS, mounted, and examined by fluorescence microscopy.
RNA purification, cDNA synthesis, in vitro transcription and array hybridization. TRIzol lysates of cells in FACS population R2 (endothelial cell fraction) and the corresponding dissociated but unpurified lung cell fraction were treated with phenol, and total RNA was precipitated with isopropanol. The integrity of the RNA was confirmed by running on an agarose gel 1 µg of total RNA extracted from the endothelial cell fraction. To construct 3'-directed cDNA libraries, cDNA was synthesized with a SuperScript II cDNA synthesis ki (Invitrogen Life Technologies; Carlsbad, CA) by using a T7: (dT)24 oligonucleotide to prime the first strand and purified by phenol extraction and ethanol precipitation. cDNA was used as a template for in vitro transcription by using the MEGAscript system (Ambion; Austin, TX) with the inclusion of biotinylated CTP and UTP. The in vitro transcription product was separated from free nucleotides by using an RNeasy column (Qiagen; Valencia, CA) and was fragmented with 150 mM magnesium chloride at 95°C for 35 min. Fragmented cRNA was hybridized in 0.2 ml of a mixture of 120 mM sodium chloride, 10 mM Tris buffer (pH 7.4), 0.005% Triton X-100, 1 mg/ml BSA, 0.1 mg/ml herring sperm DNA, and bacterial transcripts spiked at known concentrations. For each of the three experiments, 10–15 µg of biotinylated cRNA were hybridized on an oligonucleotide array (see Oligonucleotide array) for 12–16 h at 40°C with rotation. The arrays were incubated at 50°C for 1 h, washed at room temperature in PBS containing 2.5 mM EDTA diluted 1:1 with water, stained with streptavidin-phycoerythrin, followed by amplification with biotinylated anti-streptavidin antibody, washed, and given a second round of streptavidinphycoerythrin. The arrays were scanned with a Hewlett-Packard/Agilent GeneArray Scanner (Agilent Technology; Palo Alto, CA).
Oligonucleotide array. Gene expression was measured by using an oligonucleotide microarray manufactured by Affymetrix to a custom design specified by Eos Biotechnology. The array design was based on the five commercial Affymetrix mouse gene arrays Mu11K A and -B and Mu19K A, -B, and -C. A subset of perfect-match probes was selected that had a high degree of cocorrelation as gene expression levels change over diverse samples based on a large quantity of experimental data. With this approach, eight oligonucleotides for each of 39,000 potentially expressed sequences were tiled in a single array of >300,000 oligonucleotides. After hybridization, gene intensities were represented as the Tukey's trimean of the eight oligonucleotide intensities. Oligonucleotide intensities were normalized across samples by fitting the intensities onto a distribution derived from a large set of experimental data. The assumption of this normalization scheme developed by Eos Biotechnology is that the distribution of all RNA species changes little across matched samples, and the scheme attempts to ascertain changes in the position within this distribution of individual RNA species from sample to sample. The normalization scheme is more robust to subtle alterations in hybridization conditions, array manufacture, and other variables than normalization schemes that match only a single statistic (for example, the 70th percentile) using a linear scaling factor.
Analysis of gene expression. Gene expression data from the oligonucleotide microarrays were mined and analyzed by using bioinformatics approaches and tools developed and implemented at Eos Biotechnology. The fluorescence intensity for each gene was calculated based on the trimean of the hybridization intensity of each probe. Hybridization intensities correspond only approximately to absolute expression levels because the protocol relies on an amplification scheme that may not be strictly quantitative for all transcripts. However, relative expression levels of the same transcript across different samples are preserved. The resulting average differences in intensity were used to represent expression levels and to calculate the ratio of the value for the endothelial cell fraction to the value for unpurified lung cells.
For each gene tiled on the microarray, the statistical significance of observed differences in the expression level between purified endothelial cells and unpurified lung cells was evaluated by using a three-factor fully crossed ANOVA design. The factors specified for each gene were 1) purification (purified cells in population R2 versus unpurified lung cells); 2) replication of the experiment (replication 1, 2, or 3), and 3) interrogating chip oligonucleotide (8 oligonucleotides tiled per gene). Thus each data point represented one of two purification conditions, one of three replications of the experiment, and the fluorescence intensity (expression) value for one of eight oligonucleotides representing a specific gene or EST. F statistics and the corresponding P values were calculated for each of the three main factor effects. The three-way interaction term (among purification, replication, and oligonucleotide) was used as the error term, because, for this design, there was a single entry per cell.
Candidates for genes/ESTs enriched in the endothelial cell fraction were selected by three criteria. First, the ratio of expression in the endothelial cell fraction to expression in the unpurified lung cell fraction (R = enrichment ratio) was 
1.5 (geometric mean) for the three experiments. Of the 39,000 genes/ESTs represented on the microarrays, 1,623 met this criterion. Second, the difference in fluorescence intensity between the two fractions (D = expression difference) was 50 units. A total of 3,388 genes/ESTs met this criterion, but only 868 had an enrichment ratio 1.5 and an intensity difference 50. Third, the P value (P = probability of no difference) was <0.05 for the three-factor ANOVA test between the two fractions.1 Only 555 of the 868 genes/ESTs met the third criterion as well as the other two.
Despite the large number of genes/ESTs examined, most were not relevant to the statistical analysis, because the first two selection criteria (ratio 1.5 and expression difference 50) excluded all but 868 of the original 39,000 genes/ESTs, thereby controlling multiple comparisons (10). With the statistical test applied to 868 genes/ESTs at a per comparison error rate of 0.05, 43 genes/ESTs would by chance alone be expected to be false positives. Because a total of 555 genes/ESTs met all three criteria, the estimated false discovery rate would be 43/555. Thus an estimated 8% of the 555 genes/ESTs designated as endothelial cell-associated could be false positives.
The accession number of each of the 555 genes/ESTs was evaluated by using standard nucleotide-nucleotide BLAST (Basic Logic Alignment Search Tool, http://www.ncbi.nlm.nih.gov:80/blast), and the bit score S and E value of the best match were identified. The 234 transcript-considered genes matched named genes with an E value < e-100; the remaining 321 were designated ESTs.
Validation of gene expression by immunohistochemistry. Endothelial cell expression of representative genes identified on microarrays was validated in sections of mouse lungs stained by immunohistochemistry. Genes were selected by their novelty and availability of antibodies for immunohistochemical staining.2 The antibodies tested included as positive controls two well-established endothelial cell markers (CD31 and VE-cadherin). Other controls included E-cadherin, which is expressed by lung epithelial cells but not endothelial cells, and omission of the primary antibody. Mice anesthetized with ketamine (100 mg/kg ip) and xylazine (5 mg/kg ip) were perfused with PBS through the aorta for 1 min at 120 mmHg, perfused with PBS through the pulmonary artery for 20 s at 20 mmHg, and then fixed by perfusion with 1% paraformaldehyde in PBS through the aorta for 2 min at 120 mmHg. Alveoli were expanded by infusion of warm 2% SeaPlaque agarose in PBS into the lungs via the trachea with a 20-gauge needle and allowed to solidify. Inflated lungs were removed, fixed in 1% paraformaldehyde for 1 h at 4°C, rinsed with PBS, infiltrated with 30% sucrose for 12–15 h, and frozen. Cryostat sections 80 µm in thickness were dried on slides for 12–15 h, permeabilized with 0.3% Triton X-100 in PBS, and incubated in 5% normal goat, hamster, or mouse serum (Jackson ImmunoResearch; West Grove, PA) in PBS+ (PBS containing 0.3% Triton X-100, 0.2% BSA, Sigma, and 0.01% thimerosal, Sigma) for 1 h to block nonspecific antibody binding. Sections were double-labeled by incubation for 12–15 h in humidified chambers with CD31 antibody and one of nine other primary antibodies2 diluted in 5% normal goat, hamster, or mouse serum in PBS+. After being rinsed with PBS containing 0.3% Triton X-100, the slides were incubated for 5 h with fluorophore (FITC or Cy3)-conjugated secondary antibodies (goat anti-rat, goat anti-rabbit, donkey anti-goat, or mouse anti-goat, 1:400; Jackson ImmunoResearch) diluted in 5% normal goat, hamster, or mouse serum in PBS+. All incubations were at room temperature. Slides were rinsed with PBS containing 0.3% Triton X-100, fixed briefly in 4% paraformaldehyde, rinsed with PBS, mounted in Vectashield (Vector Laboratories), and imaged with a Zeiss LSM 510 confocal microscope (x40 oil Plan Apochromat objective, x2 zoom). Two-dimensional projections were made from stacks of 0.7-µm confocal optical slices of 80-µm physical sections of lung alveoli.
|
RESULTS |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
Isolation of endothelial cell fraction by FACS. For the gene expression profiling experiments in which fluorescent liposomes were injected but lectin was not, lung cells were dissociated and then divided into two fractions: one was used to isolate cells by FACS (Fig. 1, C–F), and the other served as the unpurified cell reference for gene profiling.
Cell population R1 isolated by FACS had red fluorescence from internalized cationic liposomes (Fig. 1C). Cell population R2, a subset of population R1, lacked green fluorescence from FITC-conjugated CD18 antibody (Fig. 1D). The remaining cells either lacked rhodamine fluorescence or had FITC fluorescence from CD18 immunoreactivity or autofluorescence. Forward and SSC plots showed the subpopulations of R1 (Fig. 1E), and the endothelial cell population showed R2 (Fig. 1F).
Characterization of endothelial cell fraction. Several steps were taken to confirm the identity and assess the homogeneity of cell populations isolated by FACS (Table 1). After Diff-Quik staining, population R1 was found to consist of abundant small round cells with little cytoplasm (Fig. 2A), interspersed by scattered larger cells, some of which had abundant cytoplasm and a large round nucleus typical of monocyte macrophages (Fig. 2B). When viewed by confocal microscopy, population R2 cells had punctate rhodamine fluorescence, were uniform in size, and had a diameter averaging 7 µm (Fig. 2C).
|
|
Immunohistochemical staining for endothelial cell and leukocyte markers revealed that 96% of population R2 cells had VE-cadherin immunoreactivity (Fig. 3, A and B), but only 74% were CD31 positive (Table 2). Fewer than 1% of population R2 cells had CD45.2 immunoreactivity (Fig. 3, C and D and Table 2). Approximately 80% of population R2 cells took up BODIPY-acetyl-LDL (Table 2), which appeared as a punctate cytoplasmic staining (Fig. 3, E and F). By comparison, 49% of population R1 cells had VE-cadherin immunoreactivity, 82% had CD31 immunoreactivity, and 47% were CD45.2-positive (Table 2).
|
|
Evaluation of selection criteria for genes enriched in the cell fractions. Gene expression profiles for the three replicates of the endothelial cell fraction were compared with their counterpart for the unpurified fraction of lung cells. The approach of comparing the two fractions was used to minimize the effect of the isolation process on the selection of candidate endothelial cell genes, because the fractions were exposed to similar conditions (liposomes, collagenase, and mechanical separation). A total of 555 genes/ESTs met the three criteria for endothelial cell enrichment. These genes/ESTs, which are described in Genes expressed in endothelial cell fraction, represented 1.4% of the 39,000 transcripts on the microarrays.
As one test of the selection criteria, we examined 270 genes/ESTs (0.7% of the transcripts on the array) that were at the opposite end of the spectrum from the endothelial cell candidates. This group had more than twofold greater expression in the unpurified lung cells fraction (R < 0.5, D < –250, P < 0.05). Consistent with expression in nonendothelial cells, the group included many epithelial cell genes, including cytokeratin 19, E-cadherin, and surfactant proteins A–D, or leukocyte/macrophage genes, including C-fms protooncogene (colony stimulating factor receptor), f-Met-Leu-Phe receptor, interleukin-1, lymph node homing receptor MEL-14, macrophage inflammatory proteins 1 and -2, mast cell high affinity IgE receptor, P-selectin ligand, and tumor necrosis factor (Table 3).
|
As another test of the isolation process and selection criteria, we assessed the expression of genes that have little or no known association with endothelial cells or other lung cells. No significant expression (intensity value <50) was detected in either fraction for the mammary gland-specific genes 
-lactalbumin or 
-casein, -D-crystallin gene of the lens, green cone pigment gene of the retina, neuronal clock gene involved in circadian rhythms, odorant-binding protein gene (OBP-I) of olfactory cells, preproinsulin I and II genes of pancreatic -cells, or the gene for tyrosinase, an enzyme involved in melanin synthesis in melanocytes of skin, retina, and tumors.
A further test of the effect of the isolation process and efficacy of the selection criteria involved a comparison of the expression profile of a group of housekeeping genes,3 which would be expected to be expressed in both cell fractions. This comparison showed expression of these genes in both cell fractions and no significant difference in amount of expression between the fractions.
The question of whether the isolation process itself perturbed gene expression in the endothelial cell fraction was also addressed by examining expression of immediate early genes that respond rapidly after activation. None of the genes for heat shock proteins or fos-related antigens 1 and 2 that were examined showed a significant difference in expression in the two fractions. Expression of c-Jun was higher in the endothelial cell fraction (Table 4), but c-Fos, FosB, Krox-20, and mitogen-activated protein p38 kinase were significantly enriched in the unpurified cell fraction.
|
Overall, by the selection criteria used, only 2.1% of genes/ESTs were found to be enriched in endothelial cells or nonendothelial cells. Most (97.9%) of the transcripts tiled on the microarrays did not meet the selection criteria for enrichment in either population, because they were expressed relatively equally in both fractions or did not have significant expression in either fraction.
Genes expressed in endothelial cell fraction. Many of the 234 genes that met the three selection criteria for endothelial cell enrichment have a well-documented association with endothelial cells (Tables 4,5,6,7). Among the familiar genes enriched in this fraction were key growth factors and their receptors: angiopoietin-2, and Tie1 and Tie2 receptors, VEGF and two of its receptors, VEGFR-1 (Flt-1), and VEGFR-2 [fetal liver kinase-1, kinase insert domain-containing receptor tyrosine kinase (KDR)] (Table 5). Well-documented endothelial cell junctional molecules included VE-cadherin, CD31, claudin 5, Zona occludens 1 (ZO-1), and gap junction protein connexin37 (Table 6). Endothelin-1, endothelin-B receptor, endothelin converting enzyme, leptin receptor, endomucin, and L-selectin-binding sialomucin CD34 were among the other familiar enriched genes (Tables 5 and 6).
|
|
|
Two integrin subunits, 6 and 1, which are known to be expressed in endothelial cells (59), were enriched in the endothelial cell fraction (Table 6). Integrin subunit 61, which is a laminin receptor, is associated with the attachment of endothelial cells and other cells to their basement membrane (45). Expression of 4- and 7-integrin subunits, which are associated with leukocytes (25, 90), were enriched in the unpurified cell fraction. Other integrin subunits were expressed about equally in both fractions.
Also among the genes enriched in the endothelial cell fraction were the lymphangiogenic growth factor VEGF-C (98), LYVE-1, a hyaluronan receptor associated with lymphatic endothelial cells but also expressed in some types of vascular endothelial cells (7, 85), and stabilin-1, a fasciclin-like hyaluronan receptor present on sinusoidal endothelial cells (94) (Table 5). VEGFR-3 (Flt-4), the receptor for VEGF-C on lymphatic and vascular endothelial cells, had roughly equal expression in both fractions.
Genes involved in development and angiogenesis but less commonly associated with quiescent endothelial cells in the adult included members of semaphorin/neuropilin (semaphorin 3C, Nrp1), Delta/Notch (Hey1, Jagged 2, Notch1, Notch4, and Numb), and wingless/Wnt (Frizzled 4 and Tle1) signaling pathways (Table 5). Also enriched were endothelial differentiation gene 1 (Edg1), a sphingosine 1-phosphate receptor involved in PDGF signaling and essential for vascular maturation and smooth muscle cell/pericyte envelopment during development (66); four members of the ephrin/Eph family (ephrin A1, B1, B2, and EphB4); stromal cell-derived factor 1-, a chemotactic and antiviral CXC chemokine and ligand for CXCR-4 receptors that is expressed by endothelial cells and is required for cardiovascular development (100); activin receptor-like kinase-1 (Alk-1), a member of the transforming growth factor- (TGF-) receptor family and the site of missense mutations that cause hereditary hemorrhagic telangiectasia type 2 (1); and bone morphogenic protein receptor type II, another TGF- receptor family member and the site of mutations that cause primary pulmonary hypertension (4) (Table 5).
Proteases enriched in the endothelial cell fraction included ADAMTS-1, a disintegrin and metalloproteinase with thrombospondin motifs (117); carboxypeptidase D, a duck hepatitis B virus receptor homologue previously identified on liver sinusoidal endothelial cells (13); dipeptidylpeptidase IV (CD26) (36); and neurotrypsin (Table 7). Two protease inhibitors, tissue inhibitor of metalloproteinase 3 (Timp3) and serine protease inhibitor 1 (serpinh1), were also enriched in the endothelial cell fraction (Table 7).
Genes that have limited or no previous association with endothelial cells included five members of the tetraspanin family (Table 6) and many of the 15 transcription factors that were found to be enriched (Table 4). Other genes not usually linked to endothelial cells included Ryk, a receptor protein tyrosine kinase essential for normal craniofacial development (Table 5), three G protein-coupled receptors (CD97, Etl1, and Rdc1; Table 5), three ion channels (Kcnb1, Kcnn4, and Trpm7; Table 4), four cell junction-related proteins (afadin, nectin-3, protocadherin 1, and sorbs1; Table 6), multiple cytoskeletal proteins (dynein Dnahc8, kinesin Kif1b, midline 2, palmdelphin, pleckstrin 2, and septin 2; Table 6), as well as exostosin 2, lipocalin 7, phospholipid scramblase 2, scube 1 growth factor, and sucrose nonfermenting protein (SNF-1)-related kinase (Snrk) (Tables 4, 5, and 7).
Because of the conservative design of the three selection criteria, not all potentially important genes expressed by endothelial cells were detected. For example, von Willebrand factor VIII-related antigen, a glycoprotein in endothelial cell Weibel-Palade bodies involved in hemostasis, slightly missed meeting all three criteria (R = 1.61, D = 501, P = 0.055). Other genes that are known to be expressed in endothelial cells but did not meet the three criteria included angiotensin-converting enzyme (ACE; R = 1.6, D = 413, P = 0.08), endothelial constitutive nitric oxide synthase (ecNOS; R = 1.3, D = 58, P = 0.3), endoglin (CD105; R = 1.2, D = 333, P = 0.1), and thrombomodulin (R = 1.3, D = 204, P = 0.065).
Few of the 321 ESTs that met the criteria for enrichment in the endothelial cell fraction had previously been studied in relationship to their cell associations. Table 8 shows 20 ESTs that had the greatest enrichment in this fraction (R 2, D 50, P < 0.01).
|
Immunohistochemical validation of endothelial cell gene expression. Immunohistochemical staining verified the expression in lung capillaries of 10 representative genes enriched in the endothelial cell fraction. Confocal microscopic imaging of CD31 immunoreactivity, a standard marker of endothelial cells, showed the distinctive pattern of alveolar capillaries in 80-µm tissue sections (Fig. 4A). Consistent with an endothelial cell association, LYVE-1 immunoreactivity had a pattern matching that of CD31 (Fig. 4, A and B). Similarly, as expected, VE-cadherin colocalized with CD31 on lung capillaries (Fig. 4, C and D). No staining was present when the primary antibody was omitted (data not shown). E-cadherin, an epithelial cell marker, had a conspicuously different distribution from that of CD31 (Fig. 4, E and F). These contrasting results made it possible to distinguish pulmonary endothelial cells from alveolar epithelial cells and confirmed that visualization of the capillary pattern required staining of a protein expressed by endothelial cells. By using these staining patterns as a reference, we compared the distributions of Alk-1, bone morphogenic protein receptor type II, CD31, dipeptidylpeptidase IV, integrin-6, neuropilin-1, phospholipid scramblase 2, and VEGFR-2 (Fig. 5). All had generally similar patterns that fit with endothelial cells of pulmonary capillaries. The staining of LYVE-1 and Alk-1 was homogeneous, consistent with a uniform plasma membrane distribution (Figs. 4B and 5A), but dipeptidylpeptidase IV (Fig. 5D) and VEGFR-2 (Fig. 5H) immunoreactivities were granular. Antibodies to bone morphogenic protein receptor type II (Fig. 5B) and dipeptidylpeptidase (Fig. 5D) stained endothelial cells and a second cell type, possible type II alveolar epithelial cells. None of the antibodies tested other than E-cadherin stained predominately nonendothelial cells in lung.2
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
Isolation of endothelial cells. It has been challenging to isolate a sufficient number of endothelial cells from whole organs to permit studies of gene expression. In previous studies (22, 31, 57, 68, 70, 96), endothelial cells have been isolated from the lung or other organs and then grown in vitro to increase the cell number. Comparisons of gene expression profiles in endothelial cells, grown as capillary-like tubes as a model for angiogenesis with cells grown in monolayer culture, have shown differences in expression of matrix metalloproteinases, integrins, and extracellular matrix proteins (33, 52). There is a question, however, of how closely such in vitro models reflect the normal spectrum of endothelial cell functions that depend on interactions of endothelial cells, pericytes, extracellular matrix, and surrounding cells as well as blood pressure and flow.
To avoid changes that occur when cells are grown in vitro, endothelial cells have been isolated freshly from organs (51, 106). Profiling of gene expression in endothelial cells isolated from high endothelial venules in mice led to the identification of genes involved in the distinctive role of these venules in lymphocyte homing to lymph nodes (51). Gene expression patterns have also been compared in endothelial cells isolated from normal colon and colon cancer by using serial analysis of gene expression (106). This approach led to the identification of multiple genes that have higher expression in endothelial cells of tumor vessels than in normal vessels (16, 106).
The purity of the endothelial cell fraction is obviously a key factor in gene profiling studies. On the basis of current methods, this purity depends on the selective labeling of the cells during the purification process. Usually endothelial cells are labeled with a specific lectin or antibody after enzymatic digestion of the tissue. The selectivity of the labeling is determined by the specificity of the label, presence of other cell types, and loss of cell-specific antigens during digestion. Another factor is the time required for cell isolation. The more prolonged the isolation procedure, the greater the likelihood of changes in gene expression occurring after the cells are removed from their normal environment. Cessation of blood pressure and flow, separation of intercellular junctions, detachment of pericytes, removal from extracellular matrix attachment, exposure to hypoxia, and changes in chemical environment are among the conditions that could trigger changes in gene expression.
The relatively simple isolation procedure used in the present studies made it possible to obtain within three and a half hours >100,000 viable endothelial cells from each mouse lung. With the use of labeling endothelial cells with fluorescent cationic liposomes (76, 114), the cells could be isolated by FACS. Because the liposomes do not extravasate, cells outside the vasculature were not labeled. Pericytes or smooth muscle cells do not come into contact with the intravascular liposomes although the cells are intimately associated with endothelial cells. Although intravascular macrophages and neutrophils do have access to and may take up cationic liposomes (76), these cells were removed by negative selection by using an anti-CD18 antibody (39). After purification, fewer than 1% of the cells in the endothelial cell fraction expressed the leukocyte marker CD45.2.
Several endothelial cell surface markers, CD31 being the most common (44) and VE-cadherin being the most specific (17), are routinely used to confirm the identity of established endothelial cell lines. The CD31 antibody we used uniformly labels intact endothelial cells of the pulmonary vasculature but labeled only 74% of the cells in the endothelial cell fraction. One explanation for the incomplete labeling is that the collagenase digestion reduced the amount of CD31 on the cell surface. Because of possible loss of such surface epitopes, we used for the main test of cell purity a VE-cadherin antibody that targets an intracellular epitope on the molecule (17). On the basis of VE-cadherin immunoreactivity, the population of isolated cells was at least 96% pure. The 4% that were VE-cadherin-negative may have been endothelial cells with subthreshold amounts of the marker. However, we cannot exclude the fact that the fraction contained some liposome-containing leukocytes that did not express sufficient CD18 to be removed by negative selection or lacked CD45.2 immunoreactivity and were thus not detected in the validation experiments.
As part of the cell isolation procedure, vascular perfusion of collagenase expedited the enzymatic digestion and made it possible to limit exposure to the enzyme at 37°C to only 20 min. The cells were kept on ice thereafter. This approach, coupled with the use of a high-speed cell sorter, shortened the isolation time. The endothelial cell fraction from the lungs of two mice provided sufficient RNA to prepare cRNA (10–15 µg) for the oligonucleotide arrays without the need to expand the cell population in vitro.
Evaluation of gene expression. Multiple levels of comparison were included in the experimental design to identify candidate genes/ESTs: 1) expression in the endothelial cell fraction was compared with that in unpurified lung cells; 2) endothelial cell isolation experiments, each involving both lungs of two mice, were replicated on three separate days; and 3) gene expression measurements by using RNA from the isolated cells were made on microarrays that contained 8 oligonucleotides for each of 39,000 genes/ESTs. The evaluation of differences in gene expression took all three of these factors into consideration.
The 555 genes/ESTs that were considered significantly enriched in the endothelial cell fraction met three criteria. First, the geometric mean of the amount of enrichment calculated from the ratio of expression in the endothelial cell fraction to that in the unpurified fraction was at least 1.5 for the three replications. Second, the difference in the absolute expression of a given gene/EST in the two fractions was at least 50. Third, the ANOVA test comparing expression values for each gene/EST in the two fractions had a P value <0.05. These criteria were determined by several factors. Because endothelial cells constitute 40% of lung cells (37), the maximal enrichment would be expected to be 2.5, assuming that the isolated cells had the same proportions as the intact lung, and the purification procedure was completely efficient and preserved numerical relationships. An enrichment ratio of 1.5 took into account the presumption that endothelial cells would not be as efficiently isolated as some other lung cells. Genes/ESTs meeting the selection criteria had a mean enrichment ratio of 1.92. The difference of 50, which is equivalent to an estimated five copies per cell, is just above the threshold for detecting expression (fluorescence) on the microarray.
About half of the genes that met all three selection criteria are known to be expressed in endothelial cells. VE-cadherin, CD31, VEGFR-1, VEGFR-2, Tie1, Tie2, and angiopoietin-2 were among the familiar ones that had significantly greater expression in the endothelial cell fraction than in unpurified lung cells. Others included ZO-1, claudin 5, connexin37, neuropilin-1, Edg-1, endothelin-1, endothelin converting enzyme, endothelin-B receptor, Notch4, and focal adhesion kinase. Expression of Alk-1 and bone morphogenic protein receptor type II (Bmpr-2) in lung endothelial cells is significant because of their roles in hereditary hemorrhagic telangiectasia and primary pulmonary hypertension respectively (1, 4). The consistency of finding such familiar examples in the endothelial cell fraction favors the likelihood that other genes meeting the same criteria, but not previously linked to endothelial cells, do indeed have greater expression in endothelial cells than in other lung cells.
The validity of the microarray findings was tested by immunohistochemical staining with antibodies against 13 proteins encoded by genes enriched in the endothelial cell fraction. Ten of the antibodies clearly labeled alveolar capillaries, and the other three gave little or no staining in lung. The complex geometry of alveolar capillaries and the proximity of endothelial cells to alveolar epithelial cells make it difficult to distinguish endothelial cells from other lung cells by conventional light microscopic immunohistochemistry. This problem was solved by using a confocal microscopic approach, whereby CD31 and VE-cadherin were used as standards, and immunoreactivity was examined in three-dimensional, 80-µm-thick sections of mouse lung. These standards illustrated the pattern of endothelial cell staining. Staining for E-cadherin documented the contrasting appearance of alveolar epithelial cells. With this perspective, we confirmed the expression in endothelial cells of the remaining genes, including two (dipeptidylpeptidase IV, and phospholipid scramblase 2) that have not, to our knowledge, been described as expressed in endothelial cells, as well as others (integrin-6, LYVE1, and neuropilin-1) that have received little or no attention in adult lung capillaries.
Further validation of the effectiveness of the isolation procedure was found in the significant enrichment of markers of nonendothelial cells in the unpurified cell fraction. Genes of this type included markers of lung epithelial cells such as E-cadherin and cytokeratin 19, markers of Type II alveolar epithelial cells such as surfactant-associated proteins A–D, as well as multiple markers of macrophages and leukocytes. These results indicate that the method used to isolate endothelial cells from lungs provided a relatively pure population that expressed many known genes in a predictable fashion.
Although the immunohistochemically validated genes enriched in the endothelial cell fraction were all expressed in normal lungs under baseline conditions, we cannot exclude the fact that the expression of some genes in endothelial cells was turned on by binding or uptake of cationic liposomes, exposure to collagenase, detachment from the basement membrane, or other steps of the isolation procedure. Indeed, the transcription factor Jun was enriched in the endothelial cell fraction, and several immediate early genes were enriched in the other fraction. However, the two fractions had the same treatment except for the FACS isolation step, and the test of enrichment involved the comparison of one fraction with the other. Any effects on gene expression should be reflected in both fractions, unless genes in certain cells are particularly sensitive to one of the steps. Changes that occur early might indicate a potential for rapid involvement in the injury/repair process in the lung.
Some genes (i.e., ACE, ecNOS, endoglin, and thrombomodulin) that are known to be present in endothelial cells did not meet the selection criteria for the endothelial cell fraction. There are at least four reasons for not detecting these or other genes by the approach we used. First, of particular relevance to the present study in which unpurified lung cells were used as a reference for gene enrichment in the endothelial cell fraction, genes that were expressed roughly equally in both fractions did not meet the selection criteria. This applied to 97.9% of the transcripts on the oligonucleotide arrays. In the case of ACE, ecNOS, endoglin, and thrombomodulin, expression tended to be higher in the endothelial cell fraction, but one or more of the other criteria for inclusion were not met. Endoglin and thrombomodulin are expressed by monocyte macrophages as well as by endothelial cells (75, 88). Second, the appropriate oligonucleotide for detecting a gene might not be present on the oligonucleotide arrays. For example, oligonucleotides for genes of endothelial cell junction adhesion molecules family (19) and tumor endothelial markers (16) were not in the microarrays. Third, gene expression cannot be measured when the amount of mRNA is below the limit of detection, is unstable, or is degraded by exonuclease during the isolation procedure. Finally, detection would be impaired by low specificity of the oligonucleotides for a particular gene, because the sensitivity of the detection depends on the specificity of the oligonucleotides on the microarray. Indeed, absolute expression values for different geens/ESTs were not used as selection criteria, because they are influenced by the affinity of the oligonucleotides for the corresponding transcripts as well as by the amount of RNA. For these reasons, lack of detection does not mean lack of expression, and the endothelial cell genes identified here are likely to represent only a small proportion of all genes expressed in normal endothelial cells.
Diverse genes expressed by pulmonary endothelial cells. Several genes found in the endothelial cell fraction are expressed in the vascular and nervous systems during embryogenesis, consistent with parallels that have been identified between these systems in development (104). Lung endothelial cells expressed semaphorin 3C and its receptor neuropilin-1 as well as Notch4, Jagged2, ephrin B2, and EphB4. Neuropilin-1 is not only involved in axonal guidance but also potentiates signaling of VEGF165 in endothelial cells by forming a complex with VEGF165 and VEGFR-2 (27, 34). As possible evidence of its role in angiogenesis, neuropilin-1 has been implicated in rheumatoid arthritis (50) and tumorigenesis (78). Notch4 receptors are involved in branching morphogenesis of the vasculature and other systems (116). Jagged is upregulated by factors that stimulate endothelial cell migration in vitro (125). Ephrin B2, the transmembrane ligand for EphB4, is expressed by arterial endothelial cells in the embryo and is thought to participate in the definition of boundaries between arteries and veins in the formation of the vasculature (2, 119). In the adult, ephrin B2 continues to be expressed mainly on the arterial side of the microcirculation, including the arterial end of some capillaries (29). Deletions or mutations of about a quarter of the genes lead to recognized embryonic defects or pathological conditions in the adult.4
Five members of the tetraspanin superfamily (TM4SF) of proteins, which have four transmembrane domains and form macromolecular complexes with other transmembrane proteins, were expressed in endothelial cells. Tetraspanins may act as linkers between extracellular domains of integrins and intracellular signaling molecules, and some are implicated in integrin-mediated endothelial cell migration in wound healing and angiogenesis (11, 21).
The expression of genes involved in vascular development, angiogenesis, axonal guidance, cell boundaries, and lateral inhibition suggests that endothelial cells in the adult lung undergo continuous remodeling or are predisposed to do so. However, the doubling time for endothelial cells in the normal lung appears to be long (estimated at 327 days in mice) (47). Therefore, these genes may have different functions in quiescent cells. Also, gene expression gauged by the amount of mRNA is not necessarily indicative of amount of protein synthesis.
VEGF and its receptors VEGFR-1 and VEGFR-2 were among the genes expressed in the endothelial cell fraction. The adult lung is a site of high VEGF expression, which may function as a survival factor for alveolar endothelial cells (56, 118). Clues that VEGF expression changes in lung injury, and may decrease under conditions leading to endothelial cell apoptosis, are beginning to emerge (56, 72, 73, 122).
In conclusion, by using the capacity of pulmonary capillary endothelial cells to take up fluorescent cationic liposomes in vivo, we isolated the cells with FACS and measured gene expression on microarrays, revealing many familiar genes as well as others that contribute to the distinctive functional and morphological properties of these cells. Some genes expressed in normal lung endothelial cells are linked to angiogenesis, neuronal guidance, boundary formation, or branching morphogenesis during development. Still unresolved is how gene expression suggestive of proliferation and remodeling can be reconciled with the limited capacity of the adult lung to regenerate in response to injury. The present study provides an approach and baseline data to address this issue by examining functionally altered endothelial cells at sites of angiogenesis in tumors, inflammation, or other pathological conditions.
|
DISCLOSURES |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
1 ANOVA was used as a statistical test based on the experimental design and consideration of several assumptions. Independence among samples is a potential issue for oligonucleotides, but the 8 oligonucleotides tiled per gene were designed to be nonoverlapping so they would give independent measures of gene expression. Outliers in the data would decrease the power of the statistical test, giving false negatives rather than overreporting. This was addressed in the design of the custom arrays by eliminating oligonucleotides that gave data inconsistent with their counterparts on commercial arrays. Outlier oligonucleotides were rare and usually showed promiscuous binding, giving saturated signals no matter what RNA was put on the array. In terms of the distribution of the data, ANOVA is quite robust with nonnormality when the design is balanced, as was the case in the present study. Although we did not test for nonnormality, the problem is more likely to produce loss of power than false positives. The issue of unequal variances was not readily addressed as there was only one value per cell. 
2 Primary antibodies used for validation of gene expression in lung endothelial cells in situ included: 1) Alk-1 (goat polyclonal, 1:1,000; R&D Systems, Minneapolis, MN); 2) bone morphogenic protein receptor type II (goat polyclonal, undiluted; R&D Systems); 3) CD31 [platelet endothelial cell adhesion molecule-1 (PECAM-1), rat monoclonal, clone MEC 13.3, 1:500; BD Pharmingen]; 4) dipeptidylpeptidase IV (CD26, rat monoclonal, clone H194–112, 1:1,000; BD Bio-Sciences); 5) integrin-6 (CD49f, rat monoclonal, clone GoH3, 1:1,000; BD Pharmingen); 6) LYVE-1 (rabbit polyclonal, 1:1,000; David Jackson, Oxford University, UK); 7) neuropilin-1 (rabbit polyclonal, 1:1,000; Oncogene Research, Cambridge, MA); 8) phospholipid scramblase 2 (rabbit polyclonal, 1:1,000; Oncogene Research); 9) VE-cadherin (rabbit polyclonal, 1:1,000; Elisabetta Dejana, FIRC Institute of Molecular Oncology, Milan, Italy); 10) VEGFR-2 (rabbit polyclonal, 1:2,000; Rolf Brekken and Philip Thorpe, University of Texas Southwestern Medical Center, Dallas, TX); and 11) E-cadherin (goat polyclonal, 1:500; R&D Systems). Three other antibodies that were tested gave faint staining of alveolar endothelial cells and epithelial cells (Notch4) or no staining of lung (ADAMTS-1, Notch1).
3 Housekeeping genes included alcohol dehydrogenase, asparagine synthetase, ATP synthase -subunit, 2-microglobulin, -glucoronidase, cytoplasmic -actin, DNA polymerase-, elongation factor 1-, glucose-6-phosphate dehydrogenase, glyceraldehyde-3-phosphate dehydrogenase, isocitrate dehydrogenase, lactate dehydrogenase, mitochondrial ribosomal protein S12, ornithine decarboxylase, RNA polymerase I subunit, RNA polymerase II subunit, and ubiquitin.
4 Enriched genes associated with developmental defects or diseases included activin receptor IIB (left-right axis malformations); Alk-1 (hereditary hemorrhagic telangiectasia type 2); ATPase, Ca2+ transporting-2 (Darier-White disease); bone morphogenic protein receptor type II (familial primary pulmonary hypertension); claudin 5 (Di-George syndrome); endothelin-converting enzyme-1 (Hirschsprung disease); frizzled-4 (exudative vitreoretinopathy); galactosamine (N-acetyl)-6-sulfate sulfatase (Morquio syndrome); Gata2 transcription factor (Mobius syndrome); growth hormone receptor (Laron's syndrome); hyaluronidase-1 (mucopolysaccharidosis IX); integrin-6 subunit (epidermolysis bullosa); kinesin heavy-chain member 1B (Charcot-Marie-Tooth disease type 2A); Kit ligand (Steel-Dickie mutation); laminin -2 (Walker-Warburg syndrome); lectin, mannose binding-1 (combined factor V-factor VIII deficiency); leptin receptor (obesity); multiple exostosis protein-2 (hereditary multiple exostoses type 2); neurofibromatosis-1 (neurofibromatosis vasculopathy); oligophrenin-1 (X-linked mental retardation); peripheral myelin protein-22 (Dejerine-Sottas syndrome); potassium intermediate/small conductance calcium-activated channel Kcnn4 (Diamond-Blackfan anemia); Ryk receptor-like tyrosine kinase (cleft palate); sarcoglycan epsilon (myoclonus-dystonia syndrome); semaphorin 3C (congenital heart defects); serpinh1 (ruptured blood vessels); sorbs1 (insulin resistance); Tie2 (venous malformations); Timp3 (Sorsby fundus dystrophy); and Tm4sf2 (X-linked mental retardation).
|
REFERENCES |
|---|
|
TOPABSTRACTMATERIALS AND METHODSRESULTSDISCUSSIONDISCLOSURESREFERENCES |
|---|
-actin movement in wounded human coronary artery endothelial cells exposed to shear stress. Exp Cell Res 270: 223–234, 2001.[ISI][Medline]
