Gene Expression Profiling of Human Stent-Induced Neointima by cDNA Array Analysis of Microscopic Specimens Retrieved by Helix Cutter Atherectomy
Detection of FK506-Binding Protein 12 Upregulation
Background—Restenosis due to neointima formation is the major limitation of stent-supported balloon angioplasty. Despite abundant animal data, molecular mechanisms of neointima formation have been investigated on only a limited basis in patients. This study sought to establish a method for profiling gene expression in human in-stent neointima and to identify differentially expressed genes that may serve as novel therapeutic targets.
Methods and Results—We retrieved tissue specimens from patients with symptomatic in-stent restenosis using a novel helix cutter atherectomy device. cDNA samples prepared from neointima (n=10) and, as a control, from the media of normal arteries (n=14) were amplified using a novel polymerase chain reaction protocol and hybridized to cDNA arrays. Immunohistochemistry characterized the atherectomy material as neointima. cDNA arrays readily identified differentially expressed genes. Some of the differentially expressed genes complied with expected gene expression patterns of neointima, including downregulation of desmin and upregulation of thrombospondin-1, cyclooxygenase-1, and the 70-kDa heat shock protein B. Additionally, we discovered previously unknown gene expression patterns, such as downregulation of mammary-derived growth inhibitor and upregulation of FK506-binding protein 12 (FKBP12). Upregulation of FKBP12 was confirmed at the protein level in neointimal smooth muscle cells.
Conclusions—Gene expression patterns of human neointima retrieved by helix-cutter atherectomy can be reliably analyzed by cDNA array technology. This technique can identify therapeutic targets in patients, as exemplified by the findings regarding FKBP12. FKBP12 is the receptor for Rapamycin (sirolimus), which in animal models reduced neointima formation. Our study thus yields a rationale for the use of Rapamycin to prevent restenosis in patients.
Restenosis is the most important limitation of percutaneous angioplasty procedures. Although stent implantation reduces the risk of restenosis compared with other percutaneous treatment modalities, angiographic restenosis rates continue to stay at ≈30%.1 More than 90% of the late lumen loss after stent implantation is caused by neointima formation.2 Neointima formation is considered an arterial healing response that is initiated by dedifferentiation of vascular smooth muscle cells (SMCs), followed by emigration and proliferation with subsequent elaboration of abundant extracellular matrix.3 4
Because of the limited availability of human neointima, our current understanding of neointima formation is based almost exclusively on animal models. However, therapeutic concepts for the prevention of neointima formation derived from animal models have not been successful in clinical practice.5 This suggests major differences in neointima formation between animals and humans. Accordingly, molecular studies in patients are needed to develop novel treatment strategies.
This study sought to establish a method for profiling gene expression in human in-stent neointima and to identify differentially expressed genes that may serve as novel therapeutic targets. We applied differential gene expression screening using cDNA array technology to probe microscopic specimens of human neointima. A previous hurdle of this method was the need for micrograms of mRNA from samples that are usually composed of 105 to106 cells. Here, we used a novel polymerase chain reaction (PCR) technology allowing the generation of representative cDNA amplificates from basically a single cell in quantities sufficient for comprehensive cDNA array hybridization.
Patients and Sample Preparation
Our study group included 16 patients who were treated for symptomatic in-stent restenosis with a novel atherectomy device (Xsizer, Endicor). This 6-French device was advanced to the lesion over conventional guidewires using conventional 8-French catheters. The device consists of a helix cutter, which protrudes out of a protecting sheath by 0.4 mm (<1 turn). When activated, the cutter rotates at a speed of 2100 rpm and transports the excised material under continuous suction into a filter assembly with 100-μm pores. In all study patients, we reached the in-stent restenosis and retrieved material. Thirteen specimens were used for cDNA analysis and 3 were used for immunohistochemistry. All patients gave informed consent to the procedure.
The control group consisted of 7 specimens of gastrointestinal arteries from 7 patients and 7 specimens of coronary arteries from 4 patients who underwent cardiac transplantation. Before mRNA preparation, we examined the control arteries for atherosclerotic changes. Specimens (≈1 mm3) without visible atherosclerotic changes were used as controls. The Table⇓ shows pertinent characteristics of the study and control groups. Immunohistochemistry of FK506-binding protein-12 (FKBP12) was performed on neointima samples from carotid restenotic arteries (n=3) that were obtained by surgical atherectomy.
Isolation of mRNA and Global Reverse Transcription PCR
Atherectomy specimens were snap-frozen and kept in liquid nitrogen. mRNA isolation, cDNA synthesis, and PCR amplification were performed as described by Klein (C.A. Klein, MD, unpublished data, 2000). Frozen tissue was lysed, and Dynabeads Oligo (dT)25 were added for 30 minutes. Afterward, beads were alternately washed in wash buffer-1 (50 mmol/L Tris-HCl [pH 8.3], 75 mmol/L KCl, 3 mmol/L MgCl2, 10 mmol/L dithiothreitol [DTT], and 0.5% Igepal) and wash buffer-2 (50 mmol/L Tris-HCl [pH 8.3], 75 mmol/L KCl, 3 mmol/L MgCl2, 10 mmol/L DTT, and 0.5% Tween-20). mRNA was reverse-transcribed in First Strand Buffer (Gibco), 0.01 mol/L DTT, 0.25% Igepal, 50 μmol/L CFL5c-Primer [5′-(CCC)5 GTC TAG A (NNN)2-3′], 0.5 mmol/L dNTP, and 200 U Superscript II (Gibco) in 20-μL reactions. After 45 minutes at 44°C, beads were washed with tailing wash buffer (50 mmol/L KH2PO4 [pH 7.0], 1 mmol/L DTT, and 0.25% Igepal), and the tailing reaction was performed in a volume of 10-μL containing 4 mmol/L MgCl2, 0.1 mmol/L DTT, 0.2 mmol/L dGTP, 10 mmol/L KH2PO4, and 10 U terminal deoxynucleotide transferase (MBI Fermentas).
cDNA was amplified by PCR in buffer 1 (Expand Long Template, Roche), 3% deionized formamide, 1.2 μmol/L CP2-Primer [5′-TCA GAA TTC ATG (CCC)5-3′], 350 μmol/L dNTP, and 4.5 U DNA-Polymerase-Mix (Roche) in a total volume of 50 μL. The PCR reaction was performed for 20 cycles of 94°C for 15 s, 65°C for 30 s, and 68°C for 2 minutes; followed by 20 cycles of 94°C for 15 s, 65°C for 30 s, 68°C for 2 minutes and 30 s plus 10 s per cycle; and a final extension of 68°C for 7 minutes.
Labeling of cDNA Probes and Hybridization to cDNA Arrays
Aliquots of 25 ng of each amplified cDNA were labeled with digoxigenin-11-dUTP (Roche) during PCR in the presence of 50 μmol/L digoxigenin-11-dUPT, 300 μmol/L dTTP, and other dNTPs at a final concentration of 350 μmol/L.
Atlas human cancer 1.2, human 1.2, cardiovascular, and stress arrays (Clontech) were prehybridized overnight in the presence of 50 μg/mL Escherichia coli DNA, 50 μg/mL pbluescript, and 15 μg/mL herring sperm DNA in DigEasyHyb buffer (Roche) at 44°C. Denatured, labeled probes were added to the hybridization solution and incubated for 48 hours. Arrays were washed according to the manufacturer’s protocol and then adding 2 final washes in 0.1×saline-sodium citrate/0.1% sodium dodecyl sulfate for 30 minutes at 68°C. Detection of filter-bound probes was performed according to the digoxigenin detection system (Roche).
Developed films were scanned and analyzed using the Vision software array (Imaging Research Inc). Background was subtracted, and signals were normalized to 9 housekeeping genes present on each filter, whereby the average of the signals of the housekeeping genes was set to 1 and the background to zero.
A selection of differential hybridization signals was confirmed by gene-specific PCR. Amplification was performed using 2.5 ng of each cDNA in a 25-μL reaction containing PCR buffer (Sigma), 200 μmol/L dNTPs, 0.1 μmol/L of each primer, and 0.75 U Taq Polymerase (Sigma). PCR products were subjected to electrophoresis on a 2% agarose gel containing ethidium bromide (0.5 μg/mL).
For histology and immunohistochemistry, the coronary atherectomy specimens were fixed in 4% formaldehyde (pH 7.0) and embedded in paraffin. Serial paraffin sections (3 μm) were deparaffinized, dehydrated and, for antigen retrieval, pressure-cooked for 4 minutes in citrate buffer (10 mmol/L; pH 6.0); this was followed by blocking endogenous peroxidase (1% H2O2/methanol for 15 minutes) and preincubation with 4% dried skim milk in Antibody Diluent (Dako). Immunostaining employed the streptavidin-alkaline phosphatase technique for α-actin and the streptavidin–horseradish peroxidase technique (Dako ChemMate Detection Kit) for the lymphocyte marker CD3, the monocyte marker MAC387, and FKBP12. Primary antibodies against smooth muscle actin (M0635, Dako, 1:300), CD3 (A0452, Dako, 1:80), and MAC387 (E026, Camon, 1:20) were used. Frozen sections (3 μm) of carotid neointima specimens were fixed in 4% formalin (pH 7.0) for 4 minutes at 4°C and blocked. FKBP12 was detected using the anti-FKBP12 antibody SA-218 (Biomol, 1:20).
Results of the experimental studies are reported as median expression values of the examined samples of each patient group. Differences between the 2 groups were analyzed by Mann-Whitney-U test (SPSS version 9.0). A descriptive P<0.01 was regarded as relevant.
Cellular Composition of Retrieved Material
We analyzed 3 samples of atherectomized tissue by immunohistochemistry (Figure 1⇓). Elastica–van-Gieson staining showed that most of the material was composed of loose extracellular matrix-like collagen (Figure 1A⇓). Cells with spindle-shaped nuclei and a yellow/brown-stained cytoplasm were most abundant. By smooth muscle α-actin staining, we identified these cells as SMCs (Figure 1B⇓). In some areas of the atherectomy specimens, we also found infiltrates of macrophages and, to a minor degree, scattered T lymphocytes, but no B lymphocytes were detectable.
Comparative Gene Expression Profiling
To confirm successful cDNA preparation and amplification, we performed PCR for the housekeeping genes β-actin and elongation factor-1α, as well as for the SMC marker α-actin (Figure 2⇓). In all controls and in 10 of 13 atherectomy specimens, we obtained PCR products of the correct size in equivalent amounts. These samples were hybridized to cDNA arrays analyzing the expression of 2435 known genes. Figure 3⇓ shows representative arrays for atherectomized material and for control media. Spots of human genomic DNA, which were used as a positive control, always showed strong hybridization signals (Figure 3⇓). Likewise, spots of 8 housekeeping genes consistently gave comparably positive signals (Figures 3⇓ and 4⇓). However, 3 negative control spots never hybridized. Except for 2 weak spots, we did not obtain any hybridization signal when a biological sample was omitted from cDNA synthesis and PCR amplification.
Visual inspection of the hybridization patterns readily identified a number of signals differentially expressed between normal and diseased tissue (Figure 3⇑). Analysis of the median densitometric signal intensity revealed that 201 genes differed between atherectomy specimens and control media by a factor of 2.5 at a descriptive P≤0.01 (complete information is available from the authors upon request). Figure 4⇑ shows the gene expression pattern for 6 differentially expressed genes with putative relevance to the pathogenesis of neointima. Among those, upregulation was found for thrombospondin-1 (TSP-1; P=0.003), the 70-kDa heat shock protein B (P<0.001), cyclooxygenase-1 (P<0.001), and FKBP12 (P<0.001), whereas desmin (P<0.001) and mammary-derived growth inhibitor (MDGI; P=0.01) were downregulated.
When comparing media specimens from coronary arteries and gastrointestinal arteries, 23 of the 2435 examined genes (0.9%) met the criteria for differential expression. We also performed separate analyses of differences between coronary media and neointima or gastrointestinal media and neointima, which essentially confirmed the aggregate analysis. In these analyses, only 22 of the 201 genes with differential expression in the aggregate analysis did not reach a descriptive P≤0.05; the highest P was 0.154.
Validation of cDNA Array Data by Gene-Specific PCR
For validation of hybridization signals through PCR using gene-specific primers, we selected 6 genes with putative relevance to the pathogenesis of neointima and β-actin. All PCR signals obtained had the predicted size (Figure 5⇓). When comparing the 168 gene-specific PCR signals (Figure 5⇓) with hybridization signals obtained from cDNA arrays (Figure 4⇑), we found that 160 signals matched with respect to intensity. This corresponds to a 95% fidelity of hybridization signals from cDNA arrays.
Validation of cDNA Array Data at the Protein Level
cDNA array analysis of carotid neointima retrieved by surgical atherectomy revealed similar gene expression profiles with respect to genes of interest as coronary atherectomy specimens. Specifically, we found robust upregulation of FKBP12 in carotid neointima (n=3). As shown in Figure 6⇓, we detected FKBP12 protein in the cytoplasm of neointimal SMCs (Figure 6B⇓), whereas no FKBP12 was detectable in SMCs from control media (Figure 6C⇓).
To identify novel therapeutic targets for the prevention of neointima formation, this study compared gene expression profiles of human coronary in-stent restenosis with those of control media. To our knowledge, we show the following 4 results for the first time. (1) Neointima from human in-stent restenosis can be retrieved by helix cutter atherectomy. (2) The material is suitable for gene expression analysis by gene-specific PCR and cDNA array technology, and cDNA array technology identifies differentially expressed genes with high sensitivity and fidelity and can correctly predict expression of corresponding proteins. (3) Tissue retrieved from in-stent restenosis by helix cutter atherectomy exhibits known gene expression patterns of neointima. (4) We discovered previously unknown gene expression events of neointima, such as downregulation of MDGI and upregulation of FKBP12.
Characteristics of Atherectomized Material and Control Media
The helix cutter atherectomy device enables the retrieval of human in-stent restenotic tissue over conventional guiding catheters, without the risk of entrapment of stent struts that limits the use of directional atherectomy in this setting. Even in cases in which effective debulking was impossible, the material retrieved was suitable for gene expression analysis. In the 3 samples that we analyzed histologically, the retrieved material clearly exhibited the characteristics of human neointima as described by Komatsu et al,6 whereas characteristics of atherosclerotic plaque were missing. Because the cellular component of human neointima consisted predominantly of SMCs, the majority of gene expression signals of our neointima samples may be derived from SMCs.
For control tissue, we chose apparently normal media, as assessed by careful histological examination. Like neointima, normal media is mainly composed of SMCs (albeit resting ones). We obtained media from coronary arteries and from gastrointestinal arteries. Comparison of gene expression profiles between these 2 types of arteries revealed that <1% of the genes examined met the criteria for differential expression. This proportion is in the order of the probability of error, given that our criterion for putative differential expression was a descriptive P≤0.01. Therefore, we considered specimens from coronary and gastrointestinal arteries together as controls.
Feasibility of Gene Expression Analysis in Microscopic Samples
The amount of tissue retrieved was very scant and contained ≈5×103 to 10×103 cells. To overcome technical problems posed by the limited amount of mRNA in such a low number of cells, we employed a new method of cDNA amplification.
The high concordance between hybridization signals and gene-specific PCR signals and the homogeneity of signals from housekeeping genes indicates that differences in hybridization signal intensity reflect variations of gene expression rather than variations of the cDNA array procedure. In fact, the 95% fidelity of hybridization signals from cDNA arrays demonstrates that our gene expression profiling approach is comparable with respect to quality and sensitivity to gene-specific PCR.
Confirmation of Presumed Gene Expression Patterns in Neointima
Our approach is further validated by our ability to confirm presumed gene expression patterns of neointima. This includes downregulation of desmin and upregulation of cyclooxygenase-1, the 70-kDa heat shock protein B, and TSP-1.
Examinations of coronary arteries after PTCA for expression of desmin demonstrated that desmin is highly expressed in quiescent, differentiated SMCs, whereas its expression is reduced in dedifferentiated, proliferating SMCs.7 Cyclooxygenase-1 upregulation is a known characteristic of SMCs that have been exposed to mechanical stress.8 Additionally, PTCA has been shown to induce the expression of heat shock protein 70 in vascular SMCs.9
Moreover, TSP-1 is involved in SMC proliferation and migration.10 Accordingly, antibody blockade of TSP-1 reduced neointima formation in a rat model of restenosis.11 Therefore, our data provide a rationale to test antibody blockade of TSP-1 in humans.
Upregulation of FKBP12 and Downregulation of MDGI
To our knowledge, we describe for the first time the upregulation of FKBP12 at the mRNA and protein level of human neointima. FKBP12 is involved in controlling transforming growth factor (TGF)-β receptor I signaling.12 It binds to the TGF-β receptor I and inhibits receptor-mediated signaling.13 14 By this mechanism, FKBP12 may prevent TGF-β–mediated cell cycle arrest.15
Silencing MDGI in the neointima was also previously unknown. MDGI is a potent tumor suppressor16 whose expression is generally associated with terminally differentiated cells.17 Its silencing may be caused by hypermethylation leading to loss of transcription, as shown in human breast cancer.18 Our findings raise the possibility that inherited cell- and gene-specific hypermethylation may contribute to the variability of restenosis formation after coronary interventions.
In this study, we focused on the feasibility of gene expression profiling and on genes that are involved in the presumably principal mechanisms of neointima proliferation. In interpreting our findings, we must consider the limitation that perfect comparison between neointima and control are not achievable in a clinical study. Coronary artery specimens could only be obtained from patients with heart failure. We cannot exclude the possibility that the gene expression profiles of these controls were influenced by the cytokine and neuroendocrine changes of heart failure. These changes would not have affected the specimens from gastrointestinal arteries. With these controls, however, the difference in embryological origin may come into play. Moreover, with regard to the restenosis specimens, we cannot exclude the possibility that some of the gene expression patterns reflect the underlying atherosclerosis.
Despite these limitations, several arguments suggest that our study yields valid information on the gene expression profiles of neointima. By histological examination, we demonstrated that the retrieved material exhibited the characteristics of human neointima, whereas changes characteristic of atherosclerotic plaques, like foam cells, were missing. Thus, our atherectomy specimens presumably reflect cellular changes due to neointima rather than a combination of neointima and atherosclerotic plaque formation. Likewise, the remarkable homogeneity in gene expression profiles between gastrointestinal and coronary arteries strongly suggests that the influence of embryological origin or of cytokine and neuroendocrine milieu was negligible. Therefore, we think the differences in the gene expression profiles of neointima and controls were largely specific for neointima. This interference is strengthened by our secondary analyses separating the 2 controls, which yielded results very similar to those from the primary aggregate analysis.
To identify the cellular source of differentially expressed genes, immunohistological detection of protein is needed. Because the cellular component of neointima consisted predominantly of SMCs, we assume that the majority of gene expression signals of our neointima samples is derived from SMCs. This was confirmed immunohistologically for 2 proteins, FKBP12 and α-actin.
Gene expression profiling of human neointima may break ground for novel therapeutic strategies. Proteins encoded by upregulated genes may serve as targets for the development of small-molecule, immunotherapeutic, or biological drugs. In this respect, FKBP12 deserves particular attention. Rapamycin (sirolimus) binds to FKBP1219 and thereby counteracts the TGF-β-inhibitory activity of overexpressed FKBP12. Accordingly, Rapamycin was shown to inhibit SMC migration and proliferation and intimal thickening after balloon angioplasty in a porcine model of restenosis.20 Our discovery of a significant upregulation of FKBP12 in human neointima may provide a rationale for the use of Rapamycin to prevent restenosis in patients with coronary stent placement.
Dr Zohlnhöfer is the recipient of a postdoctoral fellowship grant (Zo 104/1-1), and the study was supported by a grant (Br 1583/1-2) from the Deutsche Forschungsgemeinschaft, Bonn-Bad Godesberg, Germany. We gratefully acknowledge the large contribution made to this study through technical assistance by Renate Hegenloh.
- Received October 17, 2000.
- Revision received October 23, 2000.
- Accepted November 21, 2000.
- Copyright © 2001 by American Heart Association
Kastrati A, Schomig A, Elezi S, et al. Prognostic value of the modified American College of Cardiology/American Heart Association stenosis morphology classification for long-term angiographic and clinical outcome after coronary stent placement. Circulation. 1999;100:1285–1290.
Farb A, Sangiorgi G, Carter AJ, et al. Pathology of acute and chronic coronary stenting in humans. Circulation. 1999;99:44–52.
Schwartz RS. Animal models of human coronary restenosis. In: Topol EJ, ed. Textbook of Interventional Cardiology. Philadelphia: Saunders; 1999:358–378.
Komatsu R, Ueda M, Naruko T, et al. Neointimal tissue response at sites of coronary stenting in humans: macroscopic, histological, and immunohistochemical analyses. Circulation. 1998;98:224–233.
Ueda M, Becker AE, Tsukada T, et al. Fibrocellular tissue response after percutaneous transluminal coronary angioplasty: an immunocytochemical analysis of the cellular composition. Circulation. 1991;83:1327–1332.
Feng Y, Yang JH, Huang H, et al. Transcriptional profile of mechanically induced genes in human vascular smooth muscle cells. Circ Res. 1999;85:1118–1123.
Chen D, Asahara T, Krasinski K, et al. Antibody blockade of thrombospondin accelerates reendothelialization and reduces neointima formation in balloon-injured rat carotid artery. Circulation. 1999;100:849–854.
Huynh HT, Larsson C, Narod S, et al. Tumor suppressor activity of the gene encoding mammary-derived growth inhibitor. Cancer Res. 1995;55:2225–2231.
Huynh H, Alpert L, Pollak M. Silencing of the mammary-derived growth inhibitor (MDGI) gene in breast neoplasms is associated with epigenetic changes. Cancer Res. 1996;56:4865–4870.
Gallo R, Padurean A, Jayaraman T, et al. Inhibition of intimal thickening after balloon angioplasty in porcine coronary arteries by targeting regulators of the cell cycle. Circulation. 1999;99:2164–2170.