Author + information
- Received April 14, 2003
- Revision received September 12, 2003
- Accepted September 15, 2003
- Published online February 4, 2004.
- Cheol Whan Lee, MD*,
- Eugenio Stabile, MD*,
- Timothy Kinnaird, MD*,
- Matie Shou, MD*,
- Joseph M. Devaney, PhD*,†,
- Stephen E. Epstein, MD* and
- Mary Susan Burnett, PhD*
Objectives We sought to understand the genomic program leading to collateral vessel formation.
Background Recently, technology has advanced to the point that it is now possible to elucidate the large array of genes that must be expressed, as well as the temporal expression pattern, for the development of functionally important collateral vessels. In this investigation, we used deoxyribonucleic acid array expression profiling to determine the time course of differential expression of 12,000 genes after femoral artery ligation in C57BL/6 mice.
Methods Ribonucleic acid was extracted from the adductor muscle, which showed no signs of ischemia. Sampling was at baseline, 6 h, and 1, 3, 7, and 14 days after femoral artery ligation or sham operation.
Results Femoral artery ligation caused the differential expression (>2-fold) of 783 genes at one or multiple time points: 518 were induced and 265 were repressed. Cluster analysis generated four temporal patterns: 1) early upregulated (6 to 24 h)—immediate early transcriptional factors, angiogenesis, inflammation, and stress-related genes; 2) mid-phase upregulated (day 3)—cell cycle and cytoskeletal and inflammatory genes; 3) late upregulated (days 7 to 14)—angiostatic, anti-inflammatory, and extracellular matrix-associated genes; and 4) downregulated—genes involved in energy metabolism, water channel, and muscle contraction. Microarray data were validated using quantitative reverse transcription polymerase chain reaction.
Conclusions This study documents the large number of genes whose differential expression and temporal functional clustering appear to contribute to collateral formation. These results can serve as a genomic model for arteriogenesis and as a database for developing new therapeutic strategies.
The development of collateral vessels, referred to as arteriogenesis, is a complex process requiring the action of multiple genes expressed in an appropriate time-dependent manner. The process is undoubtedly tightly regulated by multiple factors, some of which include the appropriately timed expression of both arteriogenic and arteriostatic factors, the relative balance of which changes over the course of time (1).
Understanding the genomic program leading to collateral formation is of fundamental importance to develop insights into what factors are responsible for the highly variable ability that different patients have to form collaterals, as well as optimal therapeutic strategies designed to enhance collateral formation. Despite its importance, little is known about the array of genes that must be expressed, or their temporal expression pattern, for functionally important collaterals to develop.
Recent advances in microarray technology provide the tools to perform comprehensive, quantitative comparisons at the transcriptional level of thousands of genes simultaneously (2). In this regard, a gene chip is commercially available for transcriptional profiling that contains ∼12,000 mouse genes. Fortunately, one of the better models commonly used for arteriogenesis research is a murine model of acute hindlimb ischemia. Thus, the animal model and technology are available that make possible sophisticated gene expression studies. Therefore, we investigated the temporal profiles of global gene expression in muscle surrounding developing collaterals during the response to acute hindlimb ischemia in the mouse, with the goal of obtaining a more comprehensive understanding of the genomic program of arteriogenesis.
Twelve-week-old male C57BL/6 mice were used for the experiments. Animals either underwent left femoral artery ligation or a sham operation. Mice were anesthetized with a mixture of ketamine (40 mg/kg) and xylazine (100 mg/kg). A skin incision was performed on the medial aspect of the left thigh. After careful dissection of the vein and nerve, the femoral artery was ligated and cut immediately distal to the inguinal ligament and proximal to the popliteal bifurcation site (Fig. 1). No branches were ligated along the length of the excised segment. In the sham group, the femoral artery was dissected free but not ligated.
After surgery, all animals were closely monitored and sacrificed at five different time points. The entire adductor muscle was used for array studies and Western blotting. The study protocol was approved by the Animal Care and Use Committee of the MedStar Research Institute.
Tissue collection and ribonucleic acid (RNA) preparation
The temporal expression profiles were analyzed at five time points (baseline, 6 h, and 1, 3, 7, and 14 days after surgery), because the majority of flow recovery occurs during this period. Left adductor muscles were harvested from groups subjected to femoral artery ligation and from sham-operated groups (n = 4 animals per group per time point). Tissue samples were immediately frozen in liquid nitrogen and stored at −80°C. Total RNA was extracted from pools of four mice using Trizol reagent (Invitrogen, Carlsbad, California), according to the manufacturer's instructions. The RNA was cleaned using a RNeasy mini-kit (Qiagen, Valencia, California) and stored at −80°C. The RNA concentrations were obtained by measuring absorbance at 260 nm, and its integrity was verified with a 2% agarose mini-gel.
Double-stranded complementary deoxyribonucleic acid (cDNA) was synthesized from 8 μg total RNA. For the first cDNA strand synthesis, oligo(dT) primers were annealed to the RNA, and extension by reverse transcriptase was performed in the presence of deoxyoligonucleotides. The second strand was synthesized using deoxyribonucleic acid polymerase I and purified using phase-lock gels-phenol/chloroform extraction, followed by ethanol precipitation.
In vitro transcription, using double-stranded cDNA as a template in the presence of biotin-labeled ribonucleotides, was performed by using an Enzo in vitro transcription kit (Enzo Diagnostics, Inc., Farmingdale, New York). Biotin-labeled complementary RNA was purified, fragmented, and hybridized to Affymetrix Murine Genome U74Av2 chips (Affymetrix, Santa Clara, California), which contains ∼12,000 mouse genes. Hybridization, washing, antibody amplification, staining, and scanning of probe arrays were performed according to the Affymetrix Technical Manual.
Scanned raw data were processed with Affymetrix GeneChip version 5.0 software. The average intensity value for each probe set, which directly correlates to messenger RNA abundance, was calculated as an average of fluorescence differences for each perfectly matched probe versus single-nucleotide-mismatched probe. This software also gives each gene a qualitative assessment of “absent” or “present” calls. Data sets on each GeneChip were normalized by scaling total chip fluorescence intensities to a common value of 800 before comparison. All data were imported into GeneSpring version 5.0 software (Silicon Genetics, Redwood City, California) for further analysis. The fold changes of each gene at different time points were calculated based on the normalized values and represented as relative to baseline (day 0). We selected the genes with a >2-fold change over baseline in at least one time point. A value of differential gene expression >2-fold at one or multiple time points between the femoral artery ligation and sham-operated groups was considered as significant. Genes that were induced or repressed at similar levels in both groups were excluded from the analysis. Cluster analysis was performed to identify families of genes with a similar time-dependent expression.
Real-time reverse transcription polymerase chain reaction (RT-PCR)
To validate the microarray data, expression of three selected genes (monocyte chemoattractant protein-1 [MCP1], metallothionein-1 [MT1], and metalloelastase [or metalloproteinase; MMP12]) were analyzed using quantitative real-time RT-PCR. Complementary DNA was synthesized from 200 ng total RNA in a 20-μl reaction using TaqMan reverse transcription reagents (Applied Biosystems, Branchburg, New Jersey). Real-time RT-PCR was performed with an ABI PRISM 7700 Sequence Detection System instrument and software (PE Applied Biosystems Inc., Foster City, California), according to the manufacturer's recommendation. The principle of realtime RT-PCR has been described in detail elsewhere (3). Primer and probe sequences were chosen using sequences in the GenBank: MCP1 forward (5′-GAGCATCCACGTGTTGGCT-3′), reverse (5′-TGGTGAATGAGTAGCAGCAGGT-3′), probe ([6-FAM]-AGCCAGATGCAGTTAACGCCCCACT-[TAMRA-FAM]); MT1 forward (5′-CCTGCTCCACCGGCG-3′), reverse (5′-GCAGACACAGCCCTGGG-3′), probe ([6-FAM]-CTGCTGCTCCTGCTGTCCCGTGTG-[TAMRA-FAM]; and MMP12 forward (5′-GAGGCAGAAACGTGGACTAAAAGT-3′), reverse (5′-GTTCATGAACAGCAACAAGGAAGA-3′), probe [6-FAM]-TTTCAAGGCACAAACC-[TAMRA-FAM]). Quantitative PCR was performed in 96 sample plates. The cDNA equivalent of 100 ng total RNA/tube containing TaqMan PCR Universal Master Mix (Applied Biosystems), 100 nmol/l probe, and 200 nmol/l of each primer was used. As a control for RNA integrity and for assay normalization, 18S ribosomal RNA was amplified using a TaqMan ribosomal RNA control reagents kit (Applied Biosystems). Gene expression levels were compared using the Ctmethod, as follows: ΔCt1= [Ctof target gene in RNA from ligated mice] − [Ctof 18S in RNA from ligated mice]; and ΔCt2= [Ctof target gene in RNA from sham mice] − [Ctof 18S in RNA from sham mice]. Expression levels were calculated by 2exp(−[ΔCt1 − ΔCt2]).
Western blot analysis of tissue hypoxia-inducible factor (HIF)1-alpha levels
Adductor and calf muscles were lysed in ice-cold buffer, and 40 μg protein was separated using 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis. After transferring onto nitrocellulose membrane, the blots were incubated with a goat anti-mouse HIF1-alpha antibody (Santa Cruz Biotechnology, Santa Cruz, California) and visualized with enhanced chemiluminescence (Pierce, Rockford, Illinois).
A total of 783 genes showed at least twofold differential expression between the femoral artery ligation and sham groups at one or multiple time points: 518 were induced and 265 were repressed. The largest number of differentially regulated genes was observed seven days after surgery; representative genes are shown in Table 1. Complete lists of upregulated and downregulated genes are presented as data supplements. (For Appendices 1 and 2, please see the February 4, 2004, issue of JACCat http://www.cardiosource.com/jacc.html.) Genes were classified into eight groups according to their predominant function, although some genes were difficult to classify because they exhibit multiple dominant functions.
Differential gene expression according to function
A number of angiogenesis-related genes, including cysteine-rich protein-61, hepatoma-derived growth factor, MCP1, placental growth factor, and transforming growth factor-beta, were differentially upregulated in the femoral artery ligation group as compared with the sham group (Table 1). However, some genes known to be related to angiogenesis (angiopoietin 1, fibroblast growth factor 1, endothelial nitric oxide synthase, FMS-related tyrosine kinase 1, hypoxia inducible factor 1 alpha, tyrosine kinase with Ig and EGF homology domains, and vascular endothelial growth factor B and C) were undetectable or not significantly changed, and other such genes (vascular endothelial growth factor A and D) were similarly upregulated in both groups.
It is interesting to note a significant upregulation of genes thought to exert angiostatic activities(interferon-gamma–inducible protein-10, monokine induced by interferon-gamma [MIG], and MMP12) during the late time points after femoral artery ligation. Stress-related genes, including heme oxygenase-1 (Hmox), heat shock protein (HSP)70-3, and MT1, were highly upregulated during the early time points.
Inflammatory response-related genescomprised the largest gene cluster that was upregulated. Although upregulated in both groups, the genes exhibited higher expression levels for longer times in the femoral artery ligation versus sham group. The epithelial neutrophil activating protein (ENA)-78 profile showed a peak level at one day after surgery. Interleukin (IL)-1-beta, IL-6, MCP1, macrophage inflammatory protein (MIP)1-alpha, and MIP2 expression also demonstrated early peak levels after femoral artery ligation, and then slowly declined. Markers for inflammatory cell infiltration were also detected: neutrophil markers (CD14, MRP8) appearing at 6 h, followed by those for macrophages (CD14, CD68), mast cells (Gp49, Gp49b), and lymphocytes (Lcp2). The most highly expressed cytokines (>10-fold induction) were ENA-78, IL-1-beta, IL-6, MCP1, MCP-3, and MIP2.
Cytoskeletal genes, including cysteine-rich protein 2 (double LIM protein-1) and tubulin were slowly induced and remained at high levels throughout the study period. Genes encoding contractile proteinswere either enhanced (Myhse, Myh8, Myla) or repressed (Mlc2, Tpm5).
The expression of extracellular matrix-associated genesoccurred at late time points. The genes in this category included those encoding several types of collagen, biglycan, Ctss (cathepsin S), Mglap (matrix Gla protein), and matrix metalloproteinases (MMP-3, MMP12, MMP14). The genes encoding CPX-1 (metallocarboxypeptidase), osteoblastic-specific factor-2, and Spp1 (osteopontin) manifest similar time trends of expression.
Genes encoding proteins involved in energy metabolism(Acadm, Cox8b, Glut-4, Lp1, Pdha-1) were generally repressed across the time course of the experiment, whereas the genes involved in cholesterol transport(ABCA1 and ApoE) were slightly upregulated. There were a number of genes corresponding to expressed sequence tags that were differentially expressed (data not shown).
Clustering gene expression patterns
The expression profiles were clustered according to similarity of temporal expression patterns, using k-means cluster analysis (Fig. 2).
Genes that were rapidly induced at early phase (peak levels at 6 to 24 h) included immediate early transcriptional factors (Fos, Junb, Ler2), followed by genes associated with angiogenesis and inflammation, as well as stress-related genes.
Genes that were upregulated at mid-phase, defined as those genes whose peak expression occurred at day 3, included genes associated with cell cycle regulation, cytoskeletal-related genes, and additional inflammation-related genes.
Genes that were upregulated at the late phase, defined as those genes whose peak expression occurred at days 7 to 14, included genes associated with angiostatic, anti-inflammatory (lipo1), cytoskeletal, and extracellular matrix-associated effects.
Genes that were downregulated (<0.5-fold) after surgery included those genes involved in energy metabolism, water channels (Aq1, Aqp4), and muscle contractions.
Validation by real-time RT-PCR
Table 2summarizes the comparison of fold changes in expression for three representative genes after surgery, showing consistent changes between TaqMan quantitative PCR and microarray analysis.
Western blot analysis of tissue HIF1-alpha levels
The HIF1-alpha protein was examined in the adductor and calf muscles (Fig. 3). In calf muscle, HIF1-alpha protein levels were elevated during the first three days after femoral artery ligation, whereas there was no evidence of HIF1-alpha expression through the study period in adductor muscle.
In this investigation, we have identified dynamic global changes of gene expression that occur in a region of active collateral growth after induction of ischemia caused by femoral artery ligation. Compared with a sham-operated group, a total of 783 genes in the femoral artery ligation group changed expression levels by at least two-fold at one or multiple time points. Of course, we cannot be certain that all of these genes are related to collateral development and what portion of them are critical components of the response. Furthermore, it is also possible that some of the changes in gene expression we found resulted from a response to ischemia that has nothing to do with collateral development.
To further understand the potential role of ischemia in inducing the gene expression changes we found, we characterized the ischemic profile of the adductor muscle, the tissue used for RNA extraction. This was accomplished by analyzing the tissue for protein levels of HIF1-alpha, which is mainly regulated post-translationally and is a critically important cell sensor of hypoxia (4). Under normoxic conditions, HIF1-alpha protein is ubiquitinated, which targets the molecule for proteasome-mediated degradation. When the cell is exposed to hypoxia, ubiquitination no longer occurs, and levels of HIF1-alpha protein are stabilized. Therefore, tissue levels of HIF1-alpha provide a marker of hypoxia/ischemia. Using this marker, we found that HIF1-alpha protein was not detectable in the adductor muscle, which lies proximally in the thigh. However, HIF1-alpha was elevated in the calf muscle during the first three days after femoral artery ligation. In addition, HIF1-alpha target genes relating to anaerobic energy metabolism (aldolase A, enolase 1, lactate dehydrogenase 1, phosphofructokinase, and phosphoglycerate kinase 1) showed no difference in expression in the adductor muscle between the sham and ischemic groups (data not shown). Finally, Western blotting showed no change in LDH1 gene expression in the adductor muscle, compared with significant induction of LDH1 in calf muscle after femoral artery ligation (data not shown).
Thus, the tissue we used for transcriptional profiling and in which collaterals are developing demonstrated no overt evidence of ischemia. These findings indicate that ischemia plays either no or at most a minor role in the gene expression results we obtained. These results also support an interesting previously described finding—that collateral vessels developing proximal to an arterial obstruction do not require the local expression of either VEGF or HIF1 (5).
It is important to emphasize that our model is one in which collateral development has been well documented. In the mouse model of hindlimb ischemia, immediately after femoral artery ligation, a dramatic reduction in distal hindlimb blood flow occurs, which progressively recovers over time (6). These findings have been observed using laser Doppler perfusion imaging to assess hindlimb blood flow perfusion in the distal hindlimb and are well correlated with microsphere-assessed perfusion, as recently reported (7). Thus, an increase in flow after femoral artery ligation occurs, and this increase in flow can only be due to an increase in collaterals. In this regard, in the same report, it was found that in the upper thigh, the number and size of second- and third-generation collateral branch arteries progressively increase after femoral artery ligation, changes that correlate with laser Doppler perfusion imaging-assessed flow. Moreover, in a recent a report from our laboratory (8), we evaluated the number of collaterals present within the upper thigh (between muscle bundles and fibers) of wild-type C57BL/6 at baseline and 28 days after femoral artery ligation (same model as in the present investigation). A significant increase in the density (number/surface area) of collaterals above baseline was observed in the operated leg of C57BL/6 mice (increase of 50%). Finally, Scholz et al. (9)demonstrated that the adductor muscle is a site of active arterial remodeling. Thus, we believe there exists compelling evidence clearly indicating that collateral vessels develop in the tissue we are examining for differential gene expression relating to collateral development (the adductor muscle), as shown by laser Doppler perfusion imaging for collateral perfusion, anatomic evidence of the presence of collaterals by several different techniques, and microsphere flow determinations of collateral flow.
Overall, temporal patterns of gene expression after femoral artery ligation can be schematically summarized as shown in Figure 4. It is interesting to note the apparently prominent role played by genes modulating inflammatory responses. We found numerous genes relating to inflammatory responses differentially regulated, with many showing extremely high levels of transcriptional activity. The ENA-78, indicative of early neutrophil infiltration (10), was upregulated early after ischemia onset. Maximum levels of MCP1, which recruits monocytes, were reached by day 1 and slowly declined thereafter. The interferon-gamma-inducible protein-10 and MIG are T-cell attractants, and their expression peaked at day 7. The time course of chemokine expression roughly correlated with the recruited leukocyte markers, consistent with a previous report that chemokines are sequentially expressed during wound healing (10). Interestingly, markers for mast cells (GP49, GP49b) were also detected, supporting a role for these cells in arteriogenesis (11). The anti-inflammatory lipocortin-1 gene was induced late after femoral artery ligation, which probably contributes to resolution of inflammation (12).
Although these finding relating to inflammation could be an artifact of our specific model, the results support the increasing evidence suggesting that components of inflammatory responses constitute critically important players in the development of collateral vessels (13,14). Our results are not only compatible with such a concept, but also the diversity of the response demonstrates the extraordinarily complex gene responses that are involved in arteriogenesis. Moreover, the sequential expression of pro-inflammatory followed by anti-inflammatory genes suggests that inflammation contributes to early processes, leading to the initiation of collateral formation, but resolves soon after induction begins.
One of the mechanistically important findings of this study which has not been previously described is the increased expression of angiostatic genes (including IP-10, MIG, and MMP12) (15,16)at late time points after arterial ligation. The MMP12 is the most efficient angiostatin-inducing MMP, and angiostatin is one of the most potent angiostatic factors (16). The relatively late appearance of IP-10, MIG, and MMP12 probably contributes to a late shift in the processes involved in arteriogenesis, possibly ushering in a different phase of collateral formation, such as a phase relating to collateral maturation.
Endothelial cell survival is thought to be an essential mechanism during angiogenesis (17). We found that genes involved in cytoprotection signaling (Hmox1, HSP70-3, MT1) were highly upregulated early after femoral artery ligation, indicating that ischemic stress may trigger a genetic program for cell survival. In addition, evidence suggests that Hmox1 and MT1 are involved in angiogenesis (18,19), a conclusion compatible with our finding of the persistent expression of such genes into the mid-phase of gene expression after femoral artery ligation. Taken together, Hmox1 and MT1 may affect endothelial cell survival in collateral-forming tissues and contribute to arteriogenesis.
A number of genes associated with cell structure and extracellular matrix modification are differentially expressed during the late time points, as were several isoforms of collagen, cathepsin S, MMPs, osteoblastic-specific factor-2, and osteopontin. These genes may contribute to collateral development in that an altered balance between extracellular proteolysis and antiproteolysis is associated with growing collateral vessels (20).
Interestingly, aquaporins (Aq1, Aqp4) were significantly repressed throughout the study period. Aquaporins are water channels that regulate transcellular water permeability and have been implicated in the process of edema formation during inflammation (21). In our model, Aqp4 repression may help to prevent excessive edema in the collateral-forming tissues. However, their specific role in arteriogenesis remains unknown.
Finally, in the sham group, inflammation-related genes were also induced after surgery, suggesting that the sham operation itself can induce a number of changes in gene expression. This finding highlights the importance of carefully considering the optimal controls in gene expression profiling studies.
A potential limitation of the study needs to be addressed. Specifically, the adductor muscle, which contains the proximal portions of the developing collateral vessels, was used for RNA extraction. The specimen necessarily contains a far greater proportion of skeletal muscle tissue compared with collateral tissue. Therefore, we cannot at this time definitively distinguish between collateral-related differential gene expression and differential gene expression that relates exclusively to the wound healing response to our intervention.
Despite these limitations, we believe our study provides a preliminary picture of global temporal patterns of gene expression involved in the complex processes contributing to collateral formation. These findings can serve as a genomic model for a more complete understanding of arteriogenesis and as a data base for the development of new therapeutic strategies.
☆ This study was supported by an internal grant from MedStar Research Institute. Dr. Lee was supported by a grant from the Postdoctoral Fellowship Program of Korea Science and Engineering Foundation.
- complementary deoxyribonucleic acid
- epithelial neutrophil activating protein-78
- hypoxia-inducible factor-1
- heme oxygenase
- heat shock protein
- interferon-gamma-inducible protein
- monocyte chemoattractant protein-1
- monokine induced by interferon-gamma
- macrophage inflammatory protein
- metalloelastase (metalloproteinase-12)
- ribonucleic acid
- reverse transcription-polymerase chain reaction
- Received April 14, 2003.
- Revision received September 12, 2003.
- Accepted September 15, 2003.
- American College of Cardiology Foundation
- Gibson U.E.,
- Heid C.A.,
- Williams P.M.
- Wenger R.H.
- Deindl E.,
- Buschmann I.,
- Hoefer I.E.,
- et al.
- Stabile E.,
- Burnett M.S.,
- Watkins C.,
- et al.
- Engelhardt E.,
- Toksoy A.,
- Goebeler M.,
- Debus S.,
- Brocker E.B.,
- Gillitzer R.
- Hannon R.,
- Croxtall J.D.,
- Getting S.J.,
- et al.
- Heil M.,
- Ziegelhoeffer T.,
- Pipp F.,
- et al.
- Belperio J.A.,
- Keane M.P.,
- Arenberg D.A.,
- et al.
- Gorrin-Rivas M.J.,
- Arii S.,
- Furutani M.,
- et al.
- Chavakis E.,
- Dimmeler S.
- Penkowa M.,
- Carrasco J.,
- Giralt M.,
- et al.