Expression profiling reveals functionally important genes and coordinately regulated signaling pathway genes during in vitro angiogenesis

C. N. Hahn1, Z. J. Su1, C. J. Drogemuller1, A. Tsykin1,3, S. R. Waterman1, P. J. Brautigan1, S. Yu1, G. Kremmidiotis2, A. Gardner2, P. J. Solomon3, G. J. Goodall1,4, M. A. Vadas1,4 and J. R. Gamble1,4

1 Human Immunology, Hanson Institute, Adelaide, Australia
2 Bionomics Limited, Thebarton, Australia
3 School of Mathematics, Adelaide, Australia
4 Department of Medicine, The University of Adelaide, Adelaide, Australia


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
Angiogenesis is a complex multicellular process requiring the orchestration of many events including migration, alignment, proliferation, lumen formation, remodeling, and maturation. Such complexity indicates that not only individual genes but also entire signaling pathways will be crucial in angiogenesis. To define an angiogenic blueprint of regulated genes, we utilized our well-characterized three-dimensional collagen gel model of in vitro angiogenesis, in which the majority of cells synchronously progress through defined morphological stages culminating in the formation of capillary tubes. We developed a comprehensive three-tiered approach using microarray analysis, which allowed us to identify genes known to be involved in angiogenesis and genes hitherto unlinked to angiogenesis as well as novel genes and has proven especially useful for genes where the magnitude of change is small. Of interest is the ability to recognize complete signaling pathways that are regulated and genes clustering into ontological groups implicating the functional importance of particular processes. We have shown that consecutive members of the mitogen-activated protein kinase and leukemia inhibitory factor signaling pathways are altered at the mRNA level during in vitro angiogenesis. Thus, at least for the mitogen-activated protein kinase pathway, mRNA changes as well as the phosphorylation changes of these gene products may be important in the control of blood vessel morphogenesis. Furthermore, in this study, we demonstrated the power of virtual Northern blot analysis, as an alternative to quantitative RT-PCR, for measuring the magnitudes of differential gene expression.

quantification; gene expression; mitogen-activated protein kinase pathway


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
THE FORMATION OF NEW BLOOD VESSELS from preexisting vessels (angiogenesis) is an important process in development and, in the adult, in the female reproductive cycle, where angiogenesis is exquisitely controlled. The process of angiogenesis involves endothelial cell proliferation, migration, lumen formation, and maturation and is highly coordinated by soluble factors and the local cellular and extracellular environment. Angiogenesis is also observed in pathologies such as cancer, diabetic retinopathy, rheumatoid arthritis, and macular degeneration, where negative control mechanisms seem to be dysfunctional (10, 18).

Approaches to identify genes that are differentially regulated in angiogenesis include colony arrays, cDNA filter arrays, microarrays, gene-calling, SAGE, differential display, and suppression subtractive hybridization (2, 21, 22, 27, 48). These approaches have been successful in identifying genes involved in endothelial cell proliferation, migration, or differentiation, but have been limited by the fact that the majority of studies have utilized cells cultured in a restricted two-dimensional (2-D) milieu.

We utilized our in vitro three-dimensional (3-D) collagen gel model of capillary tube formation, in which the time course of morphological stages is very well characterized (20) and which recapitulates many of the critical individual processes in angiogenesis (36). Importantly, because these stages are highly synchronized, whereby the majority of cells at any one time are undergoing the same morphological rearrangements, crucial transcriptional events in the angiogenic process are demonstrable. We produced an "angiogenic" microarray chip from libraries generated using suppression subtractive hybridization. This allowed enrichment for regulated cDNAs relevant to the angiogenic process and amplification, via normalization, of a number of rare and lowly regulated mRNAs.

To date, the majority of microarray-based analyses of gene expression in angiogenic models have performed single or duplicate hybridizations per experiment and then applied a typical twofold or more cutoff (1, 2, 26, 56). However, this results in the identification of only highly regulated genes. By increasing the number of hybridizations, performing time course analyses, and applying rigorous statistical analyses, we identified lowly differentially expressed genes and entire signaling pathways, which have proven biologically relevant. Furthermore, we demonstrated the coordinated regulation of successive members of signaling pathways, in particular the mitogen-activated protein kinase (MAPK) and leukemia inhibitory factor (LIF) pathways, as an additional strength of this approach. In addition, we also showed that "virtual Northern" (VN) blot analysis is an alternative approach to quantification, being up to 2,500 times more saving on RNA than Northern blots and having benefits over quantitative RT-PCR (Q-RT-PCR) without jeopardizing sensitivity.


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
Cell culture: in vitro capillary tube model.
Human umbilical vein endothelial cells (HUVECs) were isolated as previously described (20) and cultured in gelatin-coated flasks in medium 199 (M199) with Earles' balanced salts and 0.68 mM glutamine, 20 mM HEPES, 20% FCS, 15 µg/ml endothelial cell growth supplement (ECGS; BD Biosciences; Bedford, MA), 50 U/ml penicillin, 50 U/ml streptomycin, and 15 µg/ml heparin (Sigma; St. Louis, MO). The protocol for the harvesting of HUVECs was performed with approval from the Royal Adelaide Hospital Research Ethics Committee.

The capillary tube formation assay was performed as previously described (20). Briefly, 6.4 x 104 cells/160 µl HUVEC medium were plated onto 100 µl gelled bovine type I collagen (Celtrix Laboratories; Palo Alto, CA) in 96-well flat-bottomed microtiter plates. Capillary tube formation was stimulated by the addition of 20 ng/ml phorbol myristate acetate (PMA) and an antibody against {alpha}2ß1-integrin (RMAC11), which promotes the formation of complex multicellular tubes.

Generation of subtracted cDNA libraries.
HUVECs from three different umbilical cords were plated onto collagen gels to induce capillary tube formation (15 wells/cord). At times of 0, 0.5, 3, 6, and 24 h after plating, total RNA was isolated using the TRIzol procedure (Invitrogen; Carlsbad, CA). These times were chosen because they approximate time points at which distinct angiogenic morphological events are known to occur (20, 36). TRIzol reagent was added directly to each well and gently agitated to aid in dissolution. RNA was extracted according to the manufacturer's protocol except that two TRIzol extractions and a further two phenol-chloroform-isoamyl alcohol (25:24:1) extractions were required to completely remove the collagen. The RNA samples, derived from tubes from each of the three HUVEC lines, corresponding to the same time points, were pooled to minimize between-subject variation. An additional precipitation was performed in which 3 vol of 4 M sodium acetate were added to the RNA, and it was pelleted at 13,000 g for 30 min. This removed other nonproteinaceous contaminants from the RNA. Yields of 7–13 µg of total HUVEC RNA were obtained. To achieve enough mRNA for subtraction experiments, the "switching mechanism at the 5'-end of RNA transcript" (SMART) PCR cDNA synthesis procedure (Clontech; Palo Alto) was used to convert the mRNA to cDNA and to preferentially amplify full-length cDNAs. The PCR Select cDNA Subtraction Kit (Clontech) was then used to perform suppression subtractive hybridization between each adjacent time point in both the forward and reverse orientation. Four forward and four reverse libraries were then generated by digesting the subtracted cDNAs with EagI and cloning them into the NotI site in pBluescriptIISK+ plasmids (Stratagene; La Jolla, CA). These libraries were used to transform XL-10 Gold ultracompetent bacteria (Stratagene).

Generation and probing of microarrays.
Generation of microarrays was performed by the Australian Genome Research Facility (Melbourne, Australia). Colonies from the forward subtracted libraries (0–0.5 h, 3,200 colonies; 0.5–3 h, 3,000 colonies; 3–6 h, 2,800 colonies; and 6–24 h, 1,000 colonies) were robotically picked, and glycerols stocks were generated. A total of 10,000 clones was picked. Each clone was then PCR amplified using primers to the T7 and T3 promoters in the vector that flank the cloning site. The PCR products were microarrayed onto glass slides in duplicate along with controls known to be regulated in this angiogenic model [e.g., prostaglandin endoperoxidase synthase 2 (PTGS2), matrix metalloproteinase 10 (MMP10), and connective tissue growth factor (CTGF)] together with genes predicted not to be regulated [thymosin-ß4 (TMSB4X) and heat shock protein 150 kDa (HSPH1)]. A dilution series of a pool of the 10,000 clones was also incorporated, along with cDNAs from other species as negative controls. We refer to these slides as "angiogenic" microarray chips.

Because of the small amounts of RNA available, the Message Amp aRNA Kit (Ambion; Austin, TX) was used to convert 0.5 µg of the total RNA from each time point to cDNA and then linearly amplify the mRNA ~400-fold. The amplified RNA (aRNA) was generated incorporating 5-(3-aminoallyl)-UTP and subsequently coupled to Cy3 and Cy5 dyes (Amersham Biosciences; Buckinghamshire, UK).

aRNA (3.5 µg) was desiccated and resuspended in 0.1 M sodium bicarbonate (pH 9.0) for 15 min. Solubilized aRNA was coupled to fluorescent CyDyes and incubated at room temperature for 1 h. Uncoupled CyDyes were removed using a PCR Purification Kit (Qiagen). aRNA was eluted in RNase-free water (60 µl). Blocking solutions [60 µl human Cot-1 DNA (1 mg/ml, Invitrogen); 0.96 µl yeast tRNA (25 mg/ml, Sigma); 1.5 µl poly dA (8 mg/ml, Sigma); 1.5 µl of 50x Denhardt's solution; and 4 µl salmon sperm DNA (10 mg/ml, Invitrogen)] were added to the aRNA and desiccated to almost complete dryness. aRNA was resuspended in hybridization buffer (16 µl of 100% deionized formamide and 16 µl of 12.5x SSC) and denatured at 100°C for 3 min. Denatured aRNA was chilled on ice, and 0.4 µl of 10% SDS were added. The hybridization mix was then added to slide arrays that had prehybridized at 42°C in 25% formamide, 5x SSC, 0.1% SDS, and 10 mg/ml BSA in a Coplin jar for 45 min. Hybridizations were performed in a hybridization chamber (Corning) at 42°C for 16 h. The slides were then washed in 0.5x SSC and 0.01% SDS for 1 min, 0.5x SSC for 3 min, and 0.06x SSC for 3 min and rinsed once in filtered MQ water. All slides were dried in a bench centrifuge at 750 rpm for 5 min. A probing strategy in which 3.5 µg of labeled aRNA at each time point was compared with the zero time point was used. All comparisons were performed in quadruplicate including duplicate dye swaps and two independent probe amplifications.

Analysis of microarray data and gene identification.
The microarray slides were scanned using a GenePix 4000B Scanner driven by GenePix Pro 4.0 (Axon Instruments; Foster City, CA). All analyses were performed using freely available statistical programming and graphics environment R (http://cran.r-project.org). Image processing was done using the "SPOT" software package (7) with a seed-growing algorithm for spot image segmentation and a morphological opening algorithm for background subtraction. The extracted information was normalized using the "Statistics in Microarray Analysis" package (http://www.stat.berkeley.edu/~terry/zarray/software/smacode.html). Within-slide normalization was based on intensity-dependent Lowess, performed for each print tip group. The data from four individual arrays used for each time point were scaled to make them directly comparable with each other. Bayesian analysis was used on the data from quadruplicate slides and duplicate spots to identify and rank clones showing consistent differential expression both within and between arrays (33). For confirmation, "moderated t-statistics" were calculated (46). In addition, profiles of fold induction versus time were generated for all pairs of spots. With the use of Bayesian analysis, moderated t-statistics, and profile analysis, the clones most likely to be regulated were selected, and the inserts were sequenced by the Australian Genome Research Facility.

Analysis of the gene sequences was performed using two programs from the freeware Staden Package (http://www.mrc-lmb.cam.ac.uk/pubseq/staden_home.html), PreGAP4 and GAP4. With the use of PreGAP4, the gene sequences were first converted from the ABI file format to SCF format. The quality of the gene sequences were analyzed (Phred), and the sequences were stripped of vector and adapter sequences. This information was then stored in a file output. GAP4 was then used to analyze and assemble the various sequences into contigs. The analysis from GAP4 produces five potential different categories including contigs, singlets (if not assembled), vector (no inserts), Escherichia coli sequences, and failed read (bad sequences). Each analysis was stored in a separate file. All the files were then merged into a single FASTA file, and BlastAll [http://www.ncbi.nlm.nih.gov/; National Center for Biotechnology Information (NCBI)] was used to blast all the sequences against the RefSeq database (NCBI). For sequences unable to be verified in this way, a BLAT search (http://genome.ucsc.edu/cgi-bin/hgBlat; University of California-Santa Cruz Genome Bioinformatics) was used to position the sequences onto chromosomes and, where possible, to link these sequences to a neighbouring gene by overlapping expressed sequence tags. NCBI UniGene nomenclature was used for naming the genes. The MIAME compliant data is available at the NCBI Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) [platform: GPL555; series: GSE779 (all experiments), GSE767 (0.5 h), GSE771 (3 h), GSE777 (6 h), GSE778 (24 h); sample: GSM11959–11962, 12034–12035, 12191–12194, and 12428–12435].

VN blot analysis.
VN blots were performed using the SMART PCR cDNA synthesis procedure (Clontech). In brief, full-length double-stranded cDNAs were generated from 1 µg of total RNA by the SMART procedure according to the manufacturer's protocol, ensuring optimization of the number of PCR cycles (typically 18–21) to remain within the linear range of amplification. Approximately 1 µg of cDNA from each time point was run on a 1% agarose gel. The cDNA was transferred by capillary transfer to nylon membranes (Hybond-N, Amersham Biosciences) and cross-linked (Stratalinker, Stratagene) to the membrane. For probe generation, cloned cDNA fragments were PCR amplified using T7 and T3 primers, and the products were labeled with [{alpha}-32P]dATP using either a MegaPrime Kit (Amersham) or StripEZ Kit (Ambion). Hybridization was performed in ExpressHyb solution (Clontech) for 2 h, and blots were washed in 2x SSC and 0.1% SDS at 25°C (twice) and 0.1x SSC and 0.1% SDS at 60°C for 20 min. Blots were visualized and quantified using a Typhoon 9410 PhosporImager and Imagequant 3.3 software (Molecular Dynamics and Amersham Biosciences). All blots were exposed for the appropriate time to ensure that the signal was within the linear range of the machine for quantification.

Q-RT-PCR.
Total RNA (3 µg) was converted to first-strand cDNA using an oligo(dT) primer (T18VN) and SuperscriptII reverse transcriptase (Invitrogen) according to the manufacturer's protocol. Residual contaminating DNA was removed using a DNA-free Kit (Ambion). Q-RT-PCR was performed on cDNA equivalent to 40 ng of starting total RNA with gene-specific primers and 0.1x SYBR green I nucleic acid gel stain (Molecular Probes) using AmpliTaq Gold DNA polymerase (Applied Biosystems) according to the Applied Biosystems' protocol. The amplification parameters were optimized for each primer pair so that appropriate-size single gene products were generated. After each real-time PCR amplification, melt curves were examined to confirm the generation of single product amplicons. Amplification parameters were typically 95°C for 10 min and cycling at 95°C for 20 s, 60–65°C for 20 s, and 72°C for 30 s for a maximum of 40 cycles using a RoboCycler thermal cycler (Corbett Research; Mortlake, New South Wales, Australia). Data were analyzed using Rotor-Gene Analysis Software version 4.6. Primer pairs were designed within the 3'-untranslated region using PrimerExpress 1.5 software (Applied Biosystems) or Primer3 freeware and synthesized by GeneWorks (Adelaide, Australia). The sequences are given in Table 1.


View this table:
[in this window]
[in a new window]
 
Table 1. Primer sequences for selected genes

 
Western blot analysis.
HUVECs were plated as described above for the capillary tube assay on either a 2-D or 3-D collagen matrix and then treated without or with 20 µM U-0126 (Sigma), a MEK inhibitor. The cells were lysed in ice-cold lysis buffer [50 mM Tris·HCl (pH 7.4) with 1% NP-40, 150 mM NaCl, 2 mM EGTA, 1 mM NaVO4, 100 mM NaF, 10 mM Na4P2O7, and protease inhibitor cocktail]. Protein concentrations were assayed using Bradford reagent (Bio-Rad; Hercules, CA). Lysates relating to equivalent numbers of cells were loaded onto 10% acrylamide gels, separated by SDS-PAGE, and transferred to polyvinylidene fluoride membranes. The membranes were blocked with 5% skim milk powder and 0.1% Triton X-100 in PBS and probed with rabbit anti-pT183 MAPK pAb, anti-ERK1/2 pAb (Promega), or anti-actin (Chemicon Australia; Boronia, Victoria, Australia). After being washed, membranes were incubated with either goat anti-rabbit IgG (H+L)-horseradish peroxidase (HRP) or goat anti-mouse IgG (H+L)-HRP (Pierce; Rockford, IL) secondary antibody, and reactive bands were detected by autoradiography using chemiluminescence (ECL+ Western Blotting Detection System, Amersham Pharmacia Biotech). The individual protein bands were quantified by densitometry using a ScanMakerX12USL scanner (Microteck) and Imagequant 3.3 software (Molecular Dynamics).


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
Selection of differentially regulated genes using Bayesian analysis, microarray profiling, and moderated t-statistics.
The angiogenic microarray chip was established as detailed in MATERIALS AND METHODS and contained cDNA fragments of genes potentially upregulated during capillary tube formation. For probing of this, to identify those regulated genes, RNA isolated from cells taken at 0, 0.5, 3, 6, and 24 h after plating onto a 3-D collagen matrix was linearly amplified and subsequently labeled with Cy3 or Cy5 dyes. Labeled aRNA from each time point was hybridized against that at time 0. In an attempt to limit the number of genes falsely identified as regulated, we chose to perform at least two independent probe amplifications and to hybridize 4 slides/time point including duplicate dye swaps. After scanning, background subtraction, local and global normalization, and ratio determination, Bayesian analysis was used to rank clones according to the likelihood of their corresponding genes being truly regulated (Fig. 1) (56). In addition, microarray profiles were generated for all 10,000 clones spotted. To do this, eight data points, taken from duplicate spots on four replicate dye swap slides, were taken for each clone at each time point after plating. The fold induction, with respect to time 0, versus time was plotted. Clones were chosen for sequencing based on regulation of expression being clearly seen in at least one time point with respect to time 0 or where there was a trend of up- or downregulation of at least two consecutive time points despite relatively low confidence for individual estimates; 1,728 clones were thus selected. When data were highly consistent, this selection process has permitted reliable identification of genes induced only 1.2–1.3 times. BLASTN (NCBI) was used to identify genes relating to microarrayed cDNA fragment sequences. Whereas some genes were identified only once, others, including known highly regulated genes, were identified multiple times because of the enriching subtractive process.



View larger version (37K):
[in this window]
[in a new window]
 
Fig. 1. Likelihood of differential expression: Bayesian plot analysis. Bayesian plots were generated by comparison of the level of gene expression at each time with respect to the zero time point. Plotted is the likelihood of true regulation (log2 odds) on the y-axis versus fold induction (log2) on the x-axis (upregulated clones are to the right of zero and downregulated ones are to the left).

 
Reproducibility of cDNA microarray profiles.
Analysis was performed comparing the microarray profiles of multiple cDNA clones derived from the same gene. Figure 2 shows the combined and separate profiles of three genes. These profiles show two important points. First, each gene has a unique expression pattern that is consistently portrayed in each profile. Individual spot values have been plotted in individual profile graphs to demonstrate the reproducibility achieved (Fig. 2B). Despite this reproducibility, the maximum magnitude in the level of differential expression varied considerably. For example, in the case of MMP10, the range was from 1.2- to 24-fold, whereas for PTGS2 the range was from 1.3- to 13-fold. Therefore, having multiple cDNA fragments represented on each microarray slide decreases the likelihood of significantly regulated gene transcripts being ignored. Second, the generation of time course information enables assignment of confidence to the selection of a regulated clone, because even if a set of data from one time point are somewhat variable, the data from an adjacent time point can be considered.



View larger version (40K):
[in this window]
[in a new window]
 
Fig. 2. Reproducibility and sensitivity of microarray profiles. Microarray profiles were generated for each clone on the microarray slide. Plotted is the fold induction (log2) with respect to time 0 (y-axis) versus time (in h; x-axis). A: profiles from cDNA fragments relating to the same gene were overlayed. Those profiles relating to the 4 individual profiles in B are in bold. The total number of profiles plotted (n) of individual clones relating to each gene is shown in the top right corner. B: individual profiles show consistency in the pattern of regulation but variation in the magnitude of induction. The magnitude of the maximal fold induction in each profile is shown. MMP10, matrix metalloproteinase-10; PTGS2, prostaglandin endoperoxidase synthase 2; EGR1, early growth response-1.

 
Thus performance of these studies in quadruplicate with duplicate spots of each subtracted clone on each slide and combining this with time course measurements greatly enhances the probability of identifying regulated genes, particularly lowly regulated genes or genes that are only transiently expressed within a process.

VN blot analysis as a powerful and informative alternative to Q-RT-PCR for confirmation of microarray data. As with all microarray data, it is important and routine to verify the regulation using alternative more quantifiable methods such as Northern blotting, RNase protection, or Q-RT-PCR. This is particularly important in light of the data above demonstrating a wide range of magnitudes of induction even for the same gene. In the in vitro HUVEC 3-D collagen model of angiogenesis, cell numbers are limited, and this model is difficult to scale up while maintaining the morphological synchrony of the cells. This left Northern blot analysis out of the question, and Q-RT-PCR and VN analysis as the only options. Here, we investigated the feasibility of using VN blots as a method for confirming microarray results, particularly because no prior knowledge of the gene sequence is necessary. VN blots have been used previously to confirm differential expression (15, 32, 43, 44, 58), although, in the majority of cases, no numerical quantification of the levels of regulation were determined.

We chose five genes to analyze in detail as they were expressed to varying degrees in our libraries and had different regulation profiles. These were MMP10, hairy/enhancer-of- split related with YRPW motif 1 (HEY1), early growth response 1 (EGR1), PTGS2, and CTGF, all of which are well known to be involved in angiogenesis (5, 8, 16, 23, 45, 52). A comparison between microarray profiles (Fig. 3A), Q-RT-PCR (Fig. 3B), and VN blot analysis (Fig. 3, C and D) is shown. Microarray profiles of clones representing each gene have been overlaid. The maximal fold of regulation for each gene using all three methods of analysis is compiled in Fig. 3E. An excellent correlation was seen between the VN profiles and those derived by microarray and Q-RT-PCR in terms of profile shape. Over 100 VN blots were performed on genes identified as differentially expressed on microarrays, and their profiles closely resembled those obtained by microarray analysis (data not shown).



View larger version (62K):
[in this window]
[in a new window]
 
Fig. 3. Comparison of angiogenesis time course profiles from microarray, quantitative RT-PCR (Q-RT-PCR), and virtual Northern (VN) blot analysis. Human umblical vein endothelial cells (HUVECs) were plated on three-dimensional (3-D) collagen in a tube formation assay. The cells were harvested at times of 0, 0.5, 3, 6, and 24 h, and total RNA was purified. The expression profiles for five genes [MMP10, hairy/enhancer-of-split related with YRPW motif 1 (HEY1), PTGS2, EGR1, and connective tissue growth factor (CTGF)] were determined using microarray (A), Q-RT-PCR (B), and VN blot analysis (C and D). Plotted is the fold induction (log2) with respect to time 0 (y-axis) versus time (in h; x-axis). Quantification of the data is shown in E. A: microarray profiles generated from individual clones [number (n)] relating to each gene were overlaid. B: Q-RT-PCR was performed in triplicate for each gene, and the average value was standardized to peptidylprolyl isomerase A (PPIA; cyclophilin A) control and plotted. C and D: VN blots were generated for each gene and visualized by phosphorimaging. Each time point was quantified and standardized to a time-matched PPIA control (inset at far right), and the results were plotted above each blot. AC are from the same batch of RNA (pool of 3 HUVEC lines), whereas D is from a separate experiment (pool of 4 HUVEC lines). E: results from A–D. For A, the median peak fold induction or repression is shown with the range of peak fold induction in parentheses. The differential expression of each of the 3 genes compared with time 0 peaks at different times: MMP10 (6 h), HEY1 (0.5 h), PTGS2 (3–6 h), EGR1 (0.5 h), and CTGF (24 h). The value for each gene profile at this specific time point was determined, and the greatest fold change was referred to as the maximum fold. The median of these profile maxima is also given. For B, the maximum fold change and standard deviation (n = 3) was determined by Q-RT-PCR. For C and D, the maximum fold change of these duplicate experiments was determined by VN analysis. F: VN analysis and Q-RT-PCR were used to measure the levels of gene expression for 75 samples (representing 13 different genes) in the capillary tube formation assay. These were standardized to PPIA control and expressed as fold induction with respect to time 0. A scatter plot is shown plotting fold induction (log2) as determined by VN blot versus Q-RT-PCR. Linear regression was used to determine the slope of the line of best fit. G: VN blot of testis-enhanced gene transcript (TEGT) showing alternatively processed and regulated transcripts.

 
The fold induction at each time point from the microarray was lower than that obtained by the more quantitative VN analysis or Q-RT-PCR, a phenomenon that is well known. Comparison between VN analysis and Q-RT-PCR showed closely parallel results (Fig. 3F), with a linear regression slope of 1.05 ± 0.06 and a r2 = 0.78, suggesting that VN analysis may be a useful technique for confirming microarray data.

VN blots show a number of advantages over other quantification procedures. First, they enable visualization of different mRNA transcripts. One example is testis enhanced gene transcript (TEGT; Fig. 3G), which shows two differentially expressed transcripts due to utilization of alternative polyadenylation sites (54). Second, VN blots provide superior sensitivity and resolution compared with Northern blots (data not shown). Finally, they can be reused multiple times (up to 8 in our hands).

The comparisons between methods also underscore interesting results. The MMP10 cDNA fragment that gave a 1.2-fold maximal induction by microarray (Figs. 2B and 3A) showed 80- to 120-fold maximal induction when measured by VN analysis and Q-RT-PCR, respectively (Fig. 3E). Thus lowly regulated genes, as determined by microarray, can in fact be highly regulated when measured by these alternative methods.

Gene expression profiles during capillary tube formation.
The sequenced regulated genes were placed, where possible, into ontological groups (Fig. 4). A number of interesting and previously unreported observations was found.




View larger version (130K):
[in this window]
[in a new window]
 
Fig. 4. Genes regulated during capillary tube formation. Differentially regulated genes have been listed in ontological groupings and identified using Unigene nomenclature and RefSeq numbers. The number of independent cDNA clones spotted on the array is given. The average fold increase with respect to time 0 at each time point is given along with its t-statistic value (brackets). The peak (either increase or decrease) fold change is in bold. Where more than one microarrayed cDNA fragment represents the same gene, the time course profile having the highest absolute t-statistic is shown.

 
In this in vitro model assay of capillary tube formation, the major processes are migration and cellular remodeling, with minimal proliferation over the 24-h period (20, 36). Consistent with this, we identified numerous RhoGTPase and RhoGTPase regulator genes that are involved in the continual dynamics of cellular movement and shape change. A number of GTPase-activating proteins (GAP) and guanine exchange factor (GEF) proteins that regulate the molecular switch GTPases were found. These included VAV3 (a RhoGEF) and DOCK9/Zizimin1 (a Cdc42GEF), proteins known to be involved in cytoskeletal rearrangements and transcriptional alterations, and filopodia induction, respectively (35, 37). Interestingly, we identified ARHGAP5 (p190-B), known for its involvement in cell migration, signaling, and differentiation (11, 47), and two additional novel RhoGAP proteins, ARHGAP18 and -24, each of whose expression peaked at different times during tube formation, implying roles at different stages of this process. We analyzed the functional characteristics of one of these genes, ARHGAP24, in detail (49). Expression profiling using Q-RT-PCR demonstrated a peak of 4-fold (compared with 1.76-fold by microarray) at 3 h. ARHGAP24 controls the formation of actin stress fibers via its regulation specifically on Rho but not Rac or Cdc42. Furthermore, knockdown of ARHGAP24 inhibits endothelial cell migration and capillary tube formation in vitro and angiogenesis in vivo (49). This is a powerful confirmation of our approach, showing the biological importance of a gene whose level of differential expression is less than twofold when measured by microarray, as is the case for many of the genes in this ontological group.

The generation of the lumen requires extensive vesicle/vacuole formation, fusion, and directional targeting to the apical surface (13, 19, 36). In keeping with this, several RabGTPases and an ArfGTPase (ARF6), known for roles in vesicle trafficking and membrane fusion in the cell, were upregulated. Expression of these genes peaked at times where vacuole expansion and fusion is leading to lumen formation. RAB5A promotes circular ruffle formation that is required for 3-D migration rather than lamellipodia, which is associated with cell motility/spreading on 2-D matrixes (50). RAB6C, which is phosphorylated by PKC (17) and plays a role in vesicle trafficking and secretion, was induced from 6 to 24 h. RAB11A and RAB21 are involved in endosomal and vesicular trafficking to the apical surfaces of epithelial cells, respectively (38), the membrane surface to which fused vacuoles travel in the forming lumen (41). Also upregulated were TBC1D4 (a RabGAP) and DDEF1 (an ArfGAP for ARF1, -5, and -6) (6), GAPs linked to vesicle trafficking, the focal adhesion protein paxillin, and fibronectin-dependent migration, respectively (39). In all, 19 of 26 genes from the cytoskeleton/vesicular trafficking ontological group peaked in expression at 3–6 h, consistent with the majority of cell migration and vesicular expansion and trafficking occurring during this time.

PTGS1 and PTGS2 (cyclooxygenase 1 and 2) are involved in lipid metabolism, both catalyzing similar reactions in the conversion of arachidonic acid to prostaglandins. PTGS2 is well known to be differentially expressed during angiogenesis, mitogenesis, and inflammation, whereas PTGS1 is generally considered to be a constitutively expressing gene (52, 55). Whereas PTGS2 expression peaked at 3–6 h and then returned to basal levels by 24 h, PTGS1 expression remained constant for 6 h before dropping (>4-fold) by 24 h, indicating that overall cyclooxygenase activity rises dramatically early in capillary tube formation and then drops markedly upon maturation of the endothelial cells. In one study (52) of colon cancer, PTSG2 modulated the production of growth factors in the cancer cells, whereas PTGS1 regulated angiogenesis in endothelial cells. The finding that PTGS1 and PTGS2 were differentially expressed in endothelial cells is an intriguing one, suggesting that both may play a role in capillary tube formation and are coordinately regulated.

Other ontological groups of genes that were regulated include the nuclear transport genes RAN, RANBP2/4/9, KPNA4, and KPNB1. A highly regulated group of genes, with respect to magnitude, were the proteases/protease inhibitors, indicating the importance of extracellular matrix modulation in the generation of capillary tubes. Other groups such as the transcription factors, growth factors/hormones and receptors, and transporters/channels also had numerous highly regulated genes. Not surprisingly, the majority of these highly regulated genes have been reported previously in other in vitro and in vivo angiogenic model systems. However, we also identified a host of more lowly and subtly differentially expressed genes, which may shed new light on important processes or pathways in angiogenesis.

Transcriptional control of MAPK signal transduction pathway genes.
Classifying the regulated genes according to their functional role highlighted that the genes of several signal transduction pathways contained members that were significantly regulated. Figure 5A shows the microarray profiles of four sequential members of the MAPK signal transduction pathway. In this model system of angiogenesis, the levels of MAP3K1 (MEKK1) peak at 3 h and then return toward uninduced levels. Q-RT-PCR demonstrates that this level of induction is 13-fold (Fig. 5B). MAP2K1 (MEK1) and MAPK1 (ERK2) levels peak at 3–6 h, both rising only approximately twofold, but remaining elevated even out to 24 h. In marked contrast, the MAPK1/2 phosphatase (MKP1) dual specificity phosphatase 1 (DUSP1), which is a negative regulator of the MAPK pathway, peaked very early and rapidly declined to well below basal levels by 3 h. (Note that all other genes identified that peaked at 0.5 h are transcription factors; Fig. 4.) Within this time period, there was an 18-fold repression of the DUSP1 gene expression from its peak to lowest level. This highly dynamic regulation of the MAPK pathway members appears to be specific for the tube forming milieu because HUVECs, plated on a flat 2-D matrix of collagen and stimulated in the same way with PMA and anti-{alpha}2ß1-integrin antibody but where cells do not form tubes, did not show significant regulation of MAPK pathway members (Fig. 5C).



View larger version (44K):
[in this window]
[in a new window]
 
Fig. 5. Gene expression profiles showing regulation of the mitogen-activated protein kinase (MAPK) and leukemia inhibitory factor (LIF) signaling pathways during capillary tube formation. Microarray profiles for members of the MAPK (A) and LIF (F) signaling pathways were generated throughout the 3-D collagen tube formation assay. Multiple profiles are shown where the genes were represented on the array more than once. The maximal fold induction is given. For dual specificity phosphatase 1 (DUSP1), the maximal fold induction at time 0.5 h and repression at 3–24 h compared with time 0 are shown. B and C: pattern and magnitude of the MAPK pathway regulation in the 3-D versus 2-D collagen setting. The levels of gene expression in a 3-D setting (A) were confirmed using Q-RT-PCR (B) and compared with those from a 2-D setting (C). Q-RT-PCR was performed in triplicate for each gene, and the average value was standardized to PPIA control and plotted. D and E: Western blot analyses of MAPK1 during capillary tube formation. The levels of active phospho-MAPK1 (pERK2) and total MAPK1 (ERK2) in HUVECs at various times after the addition of PMA and RMAC11 plated on 3-D and 2-D collagen gel are shown. Actin was visualized as a loading control. E: levels of phospho-MAPK1 were standardized to total MAPK1 in the 3-D (dotted line and triangles) and 2-D (bold lineand squares) collagen setting, and the fold change was compared with time 0 plotted.

 
Such concerted changes in the mRNA levels of these consecutive MAPK pathway members in the 3-D tube-forming environment might be expected to result in an increase in the relative levels of active phosphorylated MAPK1 compared with the 2-D setting. To determine whether this is indeed the case, Western blots were performed. Figure 5, D and E, shows an increase in active phosphorylated MAPK1 during tube formation. This increase is highest at 0.5 h and then rises again at 6 h, demonstrating a bimodal sustained response. The levels of activated MAPK2 (ERK1) closely paralleled those of MAPK1. An inhibitor of MAPK1/2 phosphorylation (U-0126) that completely blocked MAPK1/2 phosphorylation in the 3-D setting inhibited even the earliest events of tube formation (data not shown), consistent with a previous report (14). Attempts to demonstrate coordinated changes in protein levels for MAP3K1, MAP2K1, and DUSP1 were unsuccessful because of the presence of large amounts of collagen and serum components in the 3-D matrix. It was estimated that the amount of cell lysate able to be analyzed by Western blot was ~50 times below the limits of detection for these antibodies.

A number of downstream target genes of the MAPK pathway were also identified, including the immediate-early genes Fos, JUNB, and EGR1 as well as phospholipase A2 (PLA2), signal transducer and activator of transcription 3 (STAT3), and the c-Myc binding partner MAX. Here, we demonstrated that the level of gene expression at the mRNA level of consecutive members of the MAPK signaling pathway is coordinately regulated in addition to its well-documented phosphorylation status.

Transcriptional control of the LIF signaling pathway.
LIF is a cytokine involved in the maintenance of embryonic stem cell pluripotency and inhibits angiogenesis and endothelial cell proliferation and migration (30). Microarray analysis has revealed the upregulation of members of this pathway during capillary tube formation (Fig. 5F). LIF, together with the LIF, oncostatin M, and IL-6 common receptor chain (IL6ST or gp130), peaked at 3–6 h. The levels of the downstream effectors JAK1 and STAT3 transcription show a sustained increase from 3 h. Targets of STAT3 that have been identified in this screen include Fos and IL-8, thus showing differential gene expression in the whole LIF signal transduction cascade from the extracellular ligand to the nucleus including targets in the nucleus.


    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
DNA microarray has proven to be a powerful tool for studying gene regulation and in some instances global patterns. Here, we demonstrated the power of combining RNA amplification and suppression subtractive hybridization with microarray and comprehensive multiple statistical analyses to 1) identify lowly regulated genes that are known to be highly biologically significant, 2) demonstrate the importance of ontological grouping of genes to indicate crucial biological processes, and 3) discover previously unrecognised entire pathways that are regulated at the transcript level.

We were able to identify on a global scale genes not only highly regulated but also many whose transcript levels vary by only 2- to 4-fold, representing a change of only 1.2- to 2-fold in microarray terms. In the vast majority of microarray studies reported to date on angiogenesis, a standard cutoff of greater than twofold has been used (1, 2, 26, 56). From our studies, this precludes the identification of a host of potentially important genes. For instance, at this cutoff, we would have listed only 7 rather than the 28 genes in the cytoskeleton/vesicular trafficking ontological group. Our recent report (49) on the role of ARHGAP24 in endothelial cell function in vitro and angiogenesis in vivo highlights the significance of identifying lowly regulated but biologically important genes. A meager twofold difference in gene expression is well recognized for numerous genes including VEGF and apolipoprotein B, where heterozygous knockout mice expressing half the wild-type levels of gene transcript are either nonviable or significantly affected (9, 24).

Categorizing regulated genes into ontological groupings is a useful way of visualizing the types of genes involved in a particular biological situation. The MAPK pathway is known to be crucial in the angiogenic process, and inhibitors of it prevent capillary tube formation and proliferation of endothelial cells (3). Activation of each of the kinases in this pathway is via rapid and cascading phosphorylation, whereas DUSP1/MKP-1 transcription and protein stability is upregulated by MAPK1 (ERK2) (and MAPK2/ERK1) (4, 25). DUSP1 then acts back on MAPK1/2 to dephosphorylate and inactivate the pathway as well as playing a possible role in the anchoring of MAPK1/2 within the nucleus (40). In most systems and experimental models, the levels of total proteins for these kinases remain quite consistent even over periods of up to 24 h. Here, we showed, for the first time, that the level of gene expression of multiple consecutive members of this pathway is regulated in a 3-D setting. The timing of this regulation takes hours, in contrast to the rapid signal transduction of this pathway by phosphorylation, which occurs in minutes. Interestingly, the three kinases of the MAPK signal transduction cascade peak at a similar time, with MAP2K1 and MAPK1 remaining elevated throughout the whole 24-h differentiation process. At least for MAP3K1 (MEKK1), this peak in expression occurs at a time when endotheial cells are actively migrating into the collagen gel. In fibroblasts, MAP3K1 orchestrates rear-end detachment of migrating cells through activation of calpain (12), and, in a human breast cancer cell line (T47D), protein levels increase threefold during mitosis and microtubule disruption, consistent with its role as a regulator of cell responses with respect to changes in microtubule integrity (57).

In contrast to the other members of the MAPK pathway, the expression profile of DUSP1, which inactivates MAPK1/2, is markedly downregulated after an initial spike of transcription. This may indicate the importance of ensuring that the MAPK pathway remains activated or activatable throughout the angiogenic process. A previous study (42) in rat glomerular mesangial cells demonstrated that transient or weak activation of MAPK1 leads to proliferation, whereas chronic activation, as might be expected here, results in differentiation. In other studies (28, 29), DUSP1 is differentially regulated at the transcript and protein levels in HUVECs, peaking at 1 h after adhesion to fibronectin or treatment with the cardiovascular hormone atrial natriuretic peptide. This results in a shortened activation of MAPK1/2 and p38 MAPK, respectively.

Long-term changes in the expression of ERK1 have been observed during neuronal regeneration (31), whereas upregulation of the murine Erk2 promoter occurs during differentiation of pC19 endothelial cells with retinoic acid (51). Upregulation of both the murine Erk1 and Erk2 genes was also reported during osteoblast differentiation from mesenchymal progenitors with BMP2 (34). Therefore, our results for upregulation of ERK2 transcript and phosphorylation levels during endothelial differentiation, as occurs in our 3-D model but not in a 2-D system, extend the observations of long-term ERK1/2 changes occurring during differentiation processes. Consistent with this are increases in markers of differentiation or maturation in this angiogenic model such as FLT1, TFPI2, ITGB1, ANGPT2, CXCR4, STC1, and HEY1. Each of these genes have previously been identified as important in angiogenesis, and their differential regulation at the mRNA level has been shown by others (2, 27).

In a similar fashion to the MAPK pathway, we found members of the LIF signaling pathway that were coordinately regulated. In this instance, differential expression is seen in members of the entire pathway from ligand through receptor and into the nucleus. This pathway is implicated in the inhibition of angiogenesis and endothelial cell proliferation and migration (30). Individual members of each of these pathways have been reported previously to be differentially expressed in other systems including the oncostatin M, LIF, and IL-6 common receptor (IL6ST). Consistent with our results, JAK1 was also regulated in a similar 3-D collagen HUVEC model (2). Interestingly, both LIF and oncostatin M, but not IL-6, promote angiogenesis in vivo (53).

A significant technical finding from this study was the large variation in fold induction seen for different spots representing not only the same gene but also the same cDNA fragment. This variation may be due to target fragment length, GC content, concentration, distance from the polyA+ tail, or ability to be PCR amplified before being spotted onto glass slides. It may also be due to abundance of transcript, alternatively processed variants, probe amplification, or signal-to-noise ratio. Preliminary studies showed a trend where target fragments most distant to the polyA+ tail resulted in low magnitudes, although this was not true for all genes. We are currently pursuing which of these factors plays a role in such variation, as this may facilitate future design of target sequences on microarrays. What is clear is that, until the rules for generating optimal cDNA target sequences are better understood, it remains crucial in microarray design to provide multiple targets and target concentrations for each gene transcript. As demonstrated here, a subtractive hybridization approach is one way to achieve this while providing the advantage of identifying as-yet-unrecognized novel cDNA sequences or splice variants.

Q-RT-PCR has become the most commonly used method to confirm differential expression of genes identified using microarray. In our system, sample size is small with total RNA available being <10 µg. Ten micrograms of total RNA is sufficient for 1 Northern blot, 5–10 RNase protection assays, 500 Q-RT-PCRs, or 2,500 VN blots. In addition, VN blots are 10–50 times more sensitive than Northern blots and, because they contain cDNA rather than labile RNA, can be probed multiple times without a significant loss of signal. We find that both VN analysis and Q-RT-PCR display similarly sensitivity (data not shown) and give similar magnitudes of differential expression. Whereas Q-RT-PCR requires knowledge of the gene sequence, the design and synthesis of primer pairs, and may require optimization of conditions, VN analysis needs no knowledge of gene sequence, utilizes simple probe preparation by excision of inserts or their PCR amplification using common vector-derived primers, and requires no expensive equipment. Hence, VN blot analysis provides a reproducible, sensitive, rapid, scalable, and cost-effective approach to quantify gene transcript levels.

The combination of precision and utilization of multiple statistical analysis in this study has uncovered biologically relevant but lowly regulated genes and entire pathways that are coordinately regulated at the gene transcript level. Thus, in a situation where a whole system is being altered, as takes place during angiogenesis, analysis of entire pathways may be crucial to our ability to both understand and ultimately manipulate such processes.


    GRANTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
This study was generously supported by a National Health and Medical Research Council Program Grant, a Cancer Council of South Australia Fellowship, and Bionomics Limited.


    ACKNOWLEDGMENTS
 
We thank Jenny Drew, Anna Sapa, and Christopher Carter for the excellent technical help and Gary Glonek for the assistance with statistical analysis. We also sincerely thank the Burnside War Memorial Hospital and Adelaide Women's and Children's Hospital for the constant supply of human umbilical cords.

Present address of C. J. Drogemuller: Applied Genetic Technology Corporation, 12085 Research Dr., Alachua, FL 32615.

Present address of S. R. Waterman: Dept of Microbiology, Univ. of Texas Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 73590-9048.


    FOOTNOTES
 
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).

Address for reprint requests and other correspondence: C. N. Hahn, Vascular Biology Laboratory, Hanson Institute, Frome Rd., Adelaide SA 5000, Australia (E-mail: chris.hahn{at}imvs.sa.gov.au).


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 

  1. Abe M and Sato Y. cDNA microarray analysis of the gene expression profile of VEGF-activated human umbilical vein endothelial cells. Angiogenesis 4: 289–298, 2001.[CrossRef][Medline]
  2. Bell SE, Mavila A, Salazar R, Bayless KJ, Kanagala S, Maxwell SA, and Davis GE. Differential gene expression during capillary morphogenesis in 3D collagen matrices: regulated expression of genes involved in basement membrane matrix assembly, cell cycle progression, cellular differentiation and G-protein signaling. J Cell Sci 114: 2755–2773, 2001.[Abstract/Free Full Text]
  3. Berra E, Milanini J, Richard DE, Le Gall M, Vinals F, Gothie E, Roux D, Pages G, and Pouyssegur J. Signaling angiogenesis via p42/p44 MAP kinase and hypoxia. Biochem Pharmacol 60: 1171–1178, 2000.[CrossRef][ISI][Medline]
  4. Bhalla US, Ram PT, and Iyengar R. MAP kinase phosphatase as a locus of flexibility in a mitogen-activated protein kinase signaling network. Science 297: 1018–1023, 2002.[Abstract/Free Full Text]
  5. Brigstock DR. Regulation of angiogenesis and endothelial cell function by connective tissue growth factor (CTGF) and cysteine-rich 61 (CYR61). Angiogenesis 5: 153–165, 2002.[CrossRef][Medline]
  6. Brown MT, Andrade J, Radhakrishna H, Donaldson JG, Cooper JA, and Randazzo PA. ASAP1, a phospholipid-dependent arf GTPase-activating protein that associates with and is phosphorylated by Src. Mol Cell Biol 18: 7038–7051, 1998.[Abstract/Free Full Text]
  7. Buckley MJ. Spot User's Guide. Sydney: CSIRO Mathematical and Information Sciences, World Wide Web database services 2002.
  8. Burbridge MF, Coge F, Galizzi JP, Boutin JA, West DC, and Tucker GC. The role of the matrix metalloproteinases during in vitro vessel formation. Angiogenesis 5: 215–226, 2002.[CrossRef][Medline]
  9. Carmeliet P, Ferreira V, Breier G, Pollefeyt S, Kieckens L, Gertsenstein M, Fahrig M, Vandenhoeck A, Harpal K, Eberhardt C, Declercq C, Pawling J, Moons L, Collen D, Risau W, and Nagy A. Abnormal blood vessel development and lethality in embryos lacking a single VEGF allele. Nature 380: 435–439, 1996.[CrossRef][ISI][Medline]
  10. Carmeliet P and Jain RK. Angiogenesis in cancer and other diseases. Nature 407: 249–257, 2000.[CrossRef][ISI][Medline]
  11. Chakravarty G, Roy D, Gonzales M, Gay J, Contreras A, and Rosen JM. P190-B, a Rho-GTPase-activating protein, is differentially expressed in terminal end buds and breast cancer. Cell Growth Differ 11: 343–354, 2000.[Abstract/Free Full Text]
  12. Cuevas BD, Abell AN, Witowsky JA, Yujiri T, Johnson NL, Kesavan K, Ware M, Jones PL, Weed SA, DeBiasi RL, Oka Y, Tyler KL, and Johnson GL. MEKK1 regulates calpain-dependent proteolysis of focal adhesion proteins for rear-end detachment of migrating fibroblasts. EMBO J 22: 3346–3355, 2003.[Abstract/Free Full Text]
  13. Davis GE and Camarillo CW. An alpha 2 beta 1 integrin-dependent pinocytic mechanism involving intracellular vacuole formation and coalescence regulates capillary lumen and tube formation in three-dimensional collagen matrix. Exp Cell Res 224: 39–51, 1996.[CrossRef][ISI][Medline]
  14. Eliceiri BP, Klemke R, Stromblad S, and Cheresh DA. Integrin alphavbeta3 requirement for sustained mitogen-activated protein kinase activity during angiogenesis. J Cell Biol 140: 1255–1263, 1998.[Abstract/Free Full Text]
  15. Endege WO, Steinmann KE, Boardman LA, Thibodeau SN, and Schlegel R. Representative cDNA libraries and their utility in gene expression profiling. Biotechniques 26: 542–548 and 550, 1999.[ISI][Medline]
  16. Fahmy RG, Dass CR, Sun LQ, Chesterman CN, and Khachigian LM. Transcription factor Egr-1 supports FGF-dependent angiogenesis during neovascularization and tumor growth. Nat Med 9: 1026–1032, 2003.[CrossRef][ISI][Medline]
  17. Fitzgerald ML and Reed GL. Rab6 is phosphorylated in thrombin-activated platelets by a protein kinase C-dependent mechanism: effects on GTP/GDP binding and cellular distribution. Biochem J 342: 353–360, 1999.[CrossRef][ISI][Medline]
  18. Folkman J. Angiogenesis in cancer, vascular, rheumatoid and other disease. Nat Med 1: 27–31, 1995.[CrossRef][ISI][Medline]
  19. Gamble J, Meyer G, Noack L, Furze J, Matthias L, Kovach N, Harlant J, and Vadas M. B1 integrin activation inhibits in vitro tube formation: effects on cell migration, vacuole coalescence and lumen formation. Endothelium 7: 23–34, 1999.[Medline]
  20. Gamble JR, Matthias LJ, Meyer G, Kaur P, Russ G, Faull R, Berndt MC, and Vadas MA. Regulation of in vitro capillary tube formation by anti-integrin antibodies. J Cell Biol 121: 931–943, 1993.[Abstract]
  21. Gerritsen ME, Soriano R, Yang S, Ingle G, Zlot C, Toy K, Winer J, Draksharapu A, Peale F, Wu TD, and Williams PM. In silico data filtering to identify new angiogenesis targets from a large in vitro gene profiling data set. Physiol Genomics 10: 13–20, 2002.[Abstract/Free Full Text]
  22. Grove AD, Prabhu VV, Young BL, Lee FC, Kulpa V, Munson PJ, and Kohn EC. Both protein activation and gene expression are involved in early vascular tube formation in vitro. Clin Cancer Res 8: 3019–3026, 2002.[Abstract/Free Full Text]
  23. Henderson AM, Wang SJ, Taylor AC, Aitkenhead M, and Hughes CC. The basic helix-loop-helix transcription factor HESR1 regulates endothelial cell tube formation. J Biol Chem 276: 6169–6176, 2001.[Abstract/Free Full Text]
  24. Homanics GE, Smith TJ, Zhang SH, Lee D, Young SG, and Maeda N. Targeted modification of the apolipoprotein B gene results in hypobetalipoproteinemia and developmental abnormalities in mice. Proc Natl Acad Sci USA 90: 2389–2393, 1993.[Abstract/Free Full Text]
  25. Ingolia NT and Murray AW. Signal transduction. History matters. Science 297: 948–949, 2002.[Free Full Text]
  26. Jih YJ, Lien WH, Tsai WC, Yang GW, Li C, and Wu LW. Distinct regulation of genes by bFGF and VEGF-A in endothelial cells. Angiogenesis 4: 313–321, 2001.[CrossRef][Medline]
  27. Kahn J, Mehraban F, Ingle G, Xin X, Bryant JE, Vehar G, Schoenfeld J, Grimaldi CJ, Peale F, Draksharapu A, Lewin DA, and Gerritsen ME. Gene expression profiling in an in vitro model of angiogenesis. Am J Pathol 156: 1887–1900, 2000.[Abstract/Free Full Text]
  28. Kiemer AK, Weber NC, Furst R, Bildner N, Kulhanek-Heinze S, and Vollmar AM. Inhibition of p38 MAPK activation via induction of MKP-1: atrial natriuretic peptide reduces TNF-alpha-induced actin polymerization and endothelial permeability. Circ Res 90: 874–881, 2002.[Abstract/Free Full Text]
  29. Kim F and Corson MA. Adhesion to fibronectin enhances MKP-1 activation in human endothelial cells. Biochem Biophys Res Commun 273: 539–545, 2000.[CrossRef][ISI][Medline]
  30. Kimura K, Tsuda K, Moriwaki C, Kawabe T, Hamada M, Obana M, Baba A, Hano T, and Nishio I. Leukemia inhibitory factor relaxes arteries through endothelium-dependent mechanism. Biochem Biophys Res Commun 294: 359–362, 2002.[CrossRef][ISI][Medline]
  31. Kitahara T, Kiryu S, Ohno K, Morita N, Kubo T, and Kiyama H. Up-regulation of ERK (MAP kinase) and MEK (MAP kinase kinase) transcription after rat facial nerve transection. Neurosci Res 20: 275–280, 1994.[CrossRef][ISI][Medline]
  32. Lenzner S, Prietz S, Feil S, Nuber UA, Ropers HH, and Berger W. Global gene expression analysis in a mouse model for Norrie disease: late involvement of photoreceptor cells. Invest Ophthalmol Vis Sci 43: 2825–2833, 2002.[Abstract/Free Full Text]
  33. Lonnstedt I and Speed TP. Replicated microarray data. Statistica Sinica 12: 31–46, 2002.[ISI]
  34. Lou J, Tu Y, Li S, and Manske PR. Involvement of ERK in BMP-2 induced osteoblastic differentiation of mesenchymal progenitor cell line C3H10T1/2. Biochem Biophys Res Commun 268: 757–762, 2000.[CrossRef][ISI][Medline]
  35. Meller N, Irani-Tehrani M, Kiosses WB, del Pozo MA, and Schwartz MA. Zizimin1, a novel Cdc42 activator, reveals a new GEF domain for Rho proteins. Nat Cell Biol 4: 639–647, 2002.[CrossRef][ISI][Medline]
  36. Meyer GT, Matthias LJ, Noack L, Vadas MA, and Gamble JR. Lumen formation during angiogenesis in vitro involves phagocytic activity, formation and secretion of vacuoles, cell death, and capillary tube remodelling by different populations of endothelial cells. Anat Rec 249: 327–340, 1997.[CrossRef][ISI][Medline]
  37. Movilla N and Bustelo XR. Biological and regulatory properties of Vav-3, a new member of the Vav family of oncoproteins. Mol Cell Biol 19: 7870–7885, 1999.[Abstract/Free Full Text]
  38. Opdam FJ, Kamps G, Croes H, van Bokhoven H, Ginsel LA, and Fransen JA. Expression of Rab small GTPases in epithelial Caco-2 cells: Rab21 is an apically located GTP-binding protein in polarised intestinal epithelial cells. Eur J Cell Biol 79: 308–316, 2000.[ISI][Medline]
  39. Oshiro T, Koyama S, Sugiyama S, Kondo A, Onodera Y, Asahara T, Sabe H, and Kikuchi A. Interaction of POB1, a downstream molecule of small G protein Ral, with PAG2, a paxillin-binding protein, is involved in cell migration. J Biol Chem 277: 38618–38626, 2002.[Abstract/Free Full Text]
  40. Pouyssegur J, Volmat V, and Lenormand P. Fidelity and spatio-temporal control in MAP kinase (ERKs) signalling. Biochem Pharmacol 64: 755–763, 2002.[CrossRef][ISI][Medline]
  41. Prekeris R, Klumperman J, and Scheller RH. A Rab11/Rip11 protein complex regulates apical membrane trafficking via recycling endosomes. Mol Cell 6: 1437–1448, 2000.[CrossRef][ISI][Medline]
  42. Schramek H, Schumacher M, and Pfaller W. Sustained ERK-2 activation in rat glomerular mesangial cells: differential regulation by protein phosphatases. Am J Physiol Renal Physiol 271: F423–F432, 1996.[Abstract/Free Full Text]
  43. Scutt CP, Jenkins T, Furuya M, and Gilmartin PM. Male specific genes from dioecious white campion identified by fluorescent differential display. Plant Cell Physiol 43: 563–572, 2002.[Abstract/Free Full Text]
  44. Shi YH, Liu J, Xia JH, and Gui JF. Screen for stage-specific expression genes between tail bud stage and heartbeat beginning stage in embryogenesis of gynogenetic silver crucian carp. Cell Res 12: 133–142, 2002.[CrossRef][ISI][Medline]
  45. Shimo T, Nakanishi T, Nishida T, Asano M, Kanyama M, Kuboki T, Tamatani T, Tezuka K, Takemura M, Matsumura T, and Takigawa M. Connective tissue growth factor induces the proliferation, migration, and tube formation of vascular endothelial cells in vitro, and angiogenesis in vivo. J Biochem (Tokyo) 126: 137–145, 1999.[Abstract]
  46. Smyth GK. Linear models and empirical Bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol 3: Article 3, 2004.
  47. Sordella R, Jiang W, Chen GC, Curto M, and Settleman J. Modulation of Rho GTPase signaling regulates a switch between adipogenesis and myogenesis. Cell 113: 147–158, 2003.[CrossRef][ISI][Medline]
  48. St Croix B, Rago C, Velculescu V, Traverso G, Romans KE, Montgomery E, Lal A, Riggins GJ, Lengauer C, Vogelstein B, and Kinzler KW. Genes expressed in human tumor endothelium. Science 289: 1197–1202, 2000.[Abstract/Free Full Text]
  49. Su ZJ, Hahn CN, Goodall GJ, Reck NM, Leske AF, Davy A, Kremmidiotis G, Vadas MA, and Gamble JR. A vascular cell-restricted RhoGAP, p73RhoGAP, is a key regulator of angiogenesis. Proc Natl Acad Sci USA 101: 12212–12217, 2004.[Abstract/Free Full Text]
  50. Suetsugu S, Yamazaki D, Kurisu S, and Takenawa T. Differential roles of WAVE1 and WAVE2 in dorsal and peripheral ruffle formation for fibroblast cell migration. Dev Cell 5: 595–609, 2003.[CrossRef][ISI][Medline]
  51. Sugiura N, Suga T, Ozeki Y, Mamiya G, and Takishima K. The mouse extracellular signal-regulated kinase 2 gene. Gene structure and characterization of the promoter. J Biol Chem 272: 21575–21581, 1997.[Abstract/Free Full Text]
  52. Tsujii M, Kawano S, Tsuji S, Sawaoka H, Hori M, and DuBois RN. Cyclooxygenase regulates angiogenesis induced by colon cancer cells. Cell 93: 705–716, 1998.[CrossRef][ISI][Medline]
  53. Vasse M, Pourtau J, Trochon V, Muraine M, Vannier JP, Lu H, Soria J, and Soria C. Oncostatin M induces angiogenesis in vitro and in vivo. Arterioscler Thromb Vasc Biol 19: 1835–1842, 1999.[Abstract/Free Full Text]
  54. Walter L, Marynen P, Szpirer J, Levan G, and Gunther E. Identification of a novel conserved human gene, TEGT. Genomics 28: 301–304, 1995.[CrossRef][ISI][Medline]
  55. Williams CS, Smalley W, and DuBois RN. Aspirin use and potential mechanisms for colorectal cancer prevention. J Clin Invest 100: 1325–1329, 1997.[Free Full Text]
  56. Yang YH, Dudoit S, Luu P, Lin DM, Peng V, Ngai J, and Speed TP. Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation. Nucleic Acids Res 30: e15, 2002.[Abstract/Free Full Text]
  57. Yujiri T, Fanger GR, Garrington TP, Schlesinger TK, Gibson S, and Johnson GL. MEK kinase 1 (MEKK1) transduces c-Jun NH2-terminal kinase activation in response to changes in the microtubule cytoskeleton. J Biol Chem 274: 12605–12610, 1999.[Abstract/Free Full Text]
  58. Zhang QJ, Goddard M, Shanahan C, Shapiro L, and Bennett M. Differential gene expression in vascular smooth muscle cells in primary atherosclerosis and in stent stenosis in humans. Arterioscler Thromb Vasc Biol 22: 2030–2036, 2002.[Abstract/Free Full Text]




This Article
Abstract
Full Text (PDF)
All Versions of this Article:
22/1/57    most recent
00278.2004v1
Citation Map
Services
Email this article to a friend
Similar articles in this journal
Similar articles in ISI Web of Science
Similar articles in PubMed
Alert me to new issues of the journal
Download to citation manager
Search for citing articles in:
ISI Web of Science (1)
Google Scholar
Articles by Hahn, C. N.
Articles by Gamble, J. R.
Articles citing this Article
PubMed
PubMed Citation
Articles by Hahn, C. N.
Articles by Gamble, J. R.


HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Visit Other APS Journals Online
Copyright © 2005 by the American Physiological Society.