Differential gene expression patterns in vein regions susceptible versus resistant to neointimal hyperplasia

Abstract

Arteriovenous hemodialysis graft (AVG) stenosis results in thrombosis and AVG failure, but prevention of stenosis has been unsuccessful due in large part to our limited understanding of the molecular processes involved in neointimal hyperplasia (NH) formation. AVG stenosis develops chiefly as a consequence of highly localized NH formation in the vein-graft anastomosis region. Surprisingly, the vein region just downstream of the vein-graft anastomosis (herein termed proximal vein region) is relatively resistant to NH. We hypothesized that the gene expression profiles of the NH-prone and NH-resistant regions will be different from each other after graft placement, and analysis of their genomic profiles may yield potential therapeutic targets to prevent AVG stenosis. To test this, we evaluated the vein-graft anastomosis (NH-prone) and proximal vein (NH-resistant) regions in a porcine model of AVG stenosis with a porcine microarray. Gene expression changes in these two distinct vein regions, relative to the gene expression in unoperated control veins, were examined at early (5 days) and later (14 days) time points following graft placement. Global genomic changes were much greater in the NH-prone region than in the NH-resistant region at both time points. In the NH-prone region, genes related to regulation of cell proliferation and osteo-/chondrogenic vascular remodeling were most enriched among the significantly upregulated genes, and genes related to smooth muscle phenotype were significantly downregulated. These results provide insights into the spatial and temporal genomic modulation underlying NH formation in AVG and suggest potential therapeutic strategies to prevent and/or limit AVG stenosis.

INTRODUCTION

Synthetic arteriovenous grafts (AVGs) are the choice of vascular access for patients who require hemodialysis and yet are not candidates for arteriovenous fistula creation surgery. AVGs are, however, prone to stenosis and subsequent thrombosis and occlusion, leading to primary failure rates that are reported to be 50 and 75% at 1 and 2 yr after implantation, respectively (39, 43). Localized neointimal hyperplasia (NH) is the predominant cause of AVG failure and forms most frequently at the vein-graft anastomosis [hereafter referred to as the venous anastomosis (VA)] and less commonly at the artery-graft anastomosis (arterial anastomosis) (2). Examination of human specimens from failed AVGs and tissues from animal models reveals that the NH lesions are highly cellularized with excessive deposition of extracellular matrix (ECM) proteins, infiltration of inflammatory cells, and formation of microneovessels (21, 37). These changes in the vascular wall can be observed as early as 7 days after graft placement and increase progressively over time (21, 37). There are currently no clinically used pharmacotherapies to prevent or treat NH formation to achieve long-term AVG patency. Development of effective therapies is impeded largely by our limited understanding of the mechanisms by which the vein remodels in response to stimuli. It is postulated that the venous anastomosis has inherent properties that, upon graft placement, may play causal roles in forming NH, including 1) excessive wound healing, fibrotic, and inflammatory responses, 2) sensitivity to the presence of nonphysiological hemodynamic stress, and 3) possibly the peculiar sensitivity of the protective adventitia and endothelium of the vein to damage (11, 12, 22, 29, 33, 34). As such, NH formation at the venous anastomosis may be the result of a complex interplay among many pathogenic factors, although the underlying cellular and molecular events are poorly understood.

In our porcine model of AVG stenosis, NH formation occurs predominantly at the venous anastomosis and is rarely observed in the vein regions immediately downstream of the venous anastomosis (i.e., the vein region within 1 cm of the venous anastomosis, hereafter referred to as the proximal vein), comparable to what is observed in humans (21, 22). Because pathogenic factors operating early on in the process of NH formation are expected to trigger a spectrum of gene expression profiles, we hypothesized that the gene expression profiles of the NH-prone venous anastomosis and nearby NH-resistant proximal vein regions will be different from each other after graft placement and that analysis and comparison of the gene expression patterns in these two vein regions will provide clues regarding the key pathogenic mechanisms and potential therapeutic targets to prevent NH. We employed a porcine global gene expression microarray to investigate the distinctive genomic features of the NH-prone and the NH-resistant regions relative to an unoperated, reference control vein from the same animal. Gene expression features were investigated at day 5 and day 14 time points following AVG placement, representing “pre-NH” and “active NH” phases of NH development, respectively, to determine the important cellular pathways at each stage of NH development. The porcine model was chosen because 1) large vessels are needed to accommodate graft placement, 2) it better replicates human hemodynamic flow conditions than rodent models, and 3) the cardiovascular system is similar to humans. To our knowledge, this is the first investigation of the differential global genomic profiles in these distinct vein regions in a porcine AVG model.

MATERIALS AND METHODS

Materials

All reagents used for RNA isolation including RNAlater, QIAzol Lysis Reagent and the miRNeasy RNA extraction kit were purchased from Qiagen (Valencia, CA). Reagents used for microarray labeling and hybridizations were provided by the Microarray Core Facilities at the University of Utah. The porcine 44K (v1) oligonucleotide microarray slides representing 43,803 probes were purchased from Agilent Technologies (Palo Alto, CA). All reagents used for RT-PCR were obtained from Applied Biosystems (Foster City, CA).

Porcine Model of AVG Stenosis

Animals.

The porcine AVG stenosis model employed in our laboratory has been previously described (20, 21). Surgeries were performed in accordance with a protocol that was approved by the Institutional Animal Care and Use Committees at the Veterans Affairs Salt Lake City Health Care System and the University of Utah. Female Yorkshire-cross pigs weighing 30–35 kg (Sigma Livestock, Salt Lake City, UT) were used to implant expanded polytetrafluoroethylene grafts unilaterally between the ipsilateral external jugular vein (EJV) and the common carotid artery, creating both graft-end-to-vessel-side arterial and venous anastomoses. Whether the graft was to be implanted on the right or left side EJV was randomly determined before each surgery. Sodium heparin was administered during the surgery and aspirin (Pharmaceutical Formulations, Edison, NJ), clopidogrel (Bristol-Myers Squibb, New York, NY) and enrofloxacin (Bayer, Pittsburgh, PA) were administered perioperatively. Graft patency was monitored by Doppler ultrasound until the animals were euthanized at 5 or 14 days postgraft placement. A total of eight operated animals (four for each time point) were used to perform gene expression studies.

NH index.

Serial sections from the center of the venous anastomosis from grafts explanted at various postoperative days were stained with hematoxylin and eosin. An NH index was defined as the neointimal area within the native venous wall and graft divided by the sum of the medial area and graft area, as described in detail in our previous study (21).

Collection of specimens for RNA analysis.

The VA and proximal vein (PV) segments (each 1 cm in length) were retrieved according to the scheme shown in Fig. 1. A vein segment from the unoperated control EJV on the contralateral side of each operated animal was also collected. At the time of tissue collection, vein samples were immediately immersed in RNAlater solution. Immersed vein samples were then kept at 4°C for no longer than 2 wk before samples were processed for total mRNA extraction as described below.

Fig. 1.

Diagrammatic representation of the neointimal hyperplasia (NH)-prone venous anastomosis (VA) region and downstream NH-resistant proximal vein (PV) segment of an operated external jugular vein (EJV) in the porcine arteriovenous graft (AVG) stenosis model and time-course of the progression of neointimal hyperplasia. A: the VA and the PV segments (shaded regions, 1 cm length each) were collected at day 5 or day 14 following graft placement. A segment (1 cm length) from the contralateral unoperated EJV of each animal was collected as a common reference control sample (not shown). Arrows indicate the direction of the blood flow within the vein and AVG. The black dashed segment represents the expanded polytetrafluoroethylene (ePTFE) graft. B: NH index in the venous anastomosis region increases over time in a porcine model. The NH index values plotted here are from a previous study (21); NH is defined as the neointimal area within the native venous wall and graft divided by the sum of the medial area and graft area. Animal numbers were 3, 2, 7, and 5 at days 7, 14, 21, and 28, respectively. Representative images of the AV graft cross section are shown from tissues collected at 7 and 14 days. These images were collected in conjunction with the previous study (21). L, vessel lumen; G, graft; H, hyperplasia.

RNA Extraction, Microarray Hybridization, and Data Analysis

Extraction of total mRNA and microarray hybridization.

Total mRNA was extracted from vein sample lysates with the miRNeasy Mini Kit. After the removal of any graft and suture material left in the tissue, a portion (≤50 mg) of each vein sample, including any tissue that might have grown inside the vessel lumen, was minced with razor blades. Samples were further disrupted with a TissueLyser (Qiagen). The quality of mRNA was determined by an Agilent Bioanalyzer 2100 (Agilent Technologies) and mRNA integrity was assessed by the RNA integrity number. The Agilent Two-Color Quick Amp Labeling Kit was used to generate fluorescently-labeled cRNA targets. Total mRNA primed with an oligo d(T)-T7 primer was converted into dsDNA. The cRNA was transcribed with all control vein (EJV) samples labeled with Cy3 and were paired with Cy5-labeled PV or VA samples from the same animal at the same time point. Microarray hybridizations were performed using Agilent SureHyb Hybridization chambers, and microarray slides were scanned in an Agilent G2505C Microarray Scanner at 5 μm resolution. Finally, scanned data were loaded into Agilent Feature Extraction Software (v10.5) for initial data reduction and analysis. Microarray data have been deposited in the Genome Expression Omnibus database repository (accession number GSE97377).

Data normalization and batch adjustment.

All arrays passed the quality metrics in Agilent’s QC report. Microarray data were filtered with the Agilent Feature Extraction software to remove values from control features and any features flagged as nonuniform or population outliers. Remaining features from each microarray probe were averaged to yield a single value for each probe. Resulting intensity values were log2-transformed and normalized by median-centered Lowess normalization. Microarray data were processed in multiple batches over the course of 2 yr, and clustering of normalized data revealed a batch effect that correlated with the date on which the data were generated (data not shown). To control for this, batch-adjustment analysis was performed with ComBat analysis software (16). ComBat first estimated factors that correlated with batch number (1–4), origin of tissue, and time of tissue retrieval and then subtracted these batch-related factors while leaving the effects of other covariates. Differential gene expression between EJV (control) and PV or VA samples was tested with limma version 3.30.13 (36), a linear modeling analysis package designed for complex gene expression data sets from microarrays or RNA-Seq experiments. Gene expression data were also analyzed using GeneSifter Analysis Edition v3.7 (http://www.genesifter.net, Geospiza, Seattle, WA).

Bioinformatics analyses of gene expression profiles.

In both the limma and GeneSifter analyses we applied a Benjamini and Hochberg adjustment to P values, and required adjustment to P ≤ 0.05 and a log2 fold-change of ≥4 between EJV (unoperated control vein) and either the PV or VA sample groups. Transcripts meeting these selection threshold criteria are expected to eliminate normal gene expression variations based on normal variations reported in the literature (7, 19, 28). Volcano plots showing differential expression between EJV (control vein) and PV or VA samples at day 5 and day 14 are shown in Fig. 2. These plots show the log2 (PV/EJV) or log2 (VA/EJV) expression ratio on the x-axis, and the Benjamini and Hochberg-adjusted P value of the test of differential expression using Student’s t-test on the y-axis. There are 43,803 probes in this porcine microarray, and any unannotated transcripts without accession numbers were excluded from further analyses. For incompletely annotated porcine transcripts with accession numbers, we used nucleotide BLAST (https://blast.ncbi.nlm.nih.gov/Blast.cgi) to determine similar sequences in other species with sequence matches above a specified cut-off threshold (≥50% matching query-subject coverage and E-value ≤10⁻⁴). In most cases, the annotated “hit” sequence with the highest total score that fulfilled the cut-off threshold criteria was selected. There were 356 transcripts that were identified this way, and they were added to those that were fully annotated for subsequent analyses. To extract biological information from the patterns of gene expression associated with NH, we used the Gene Annotation Tool to Help Explain Relationships (GATHER) gene annotation for identification of enriched biological themes (6). Transcripts significantly altered in each of the two vein regions were separately subjected to GATHER analysis. Also, up- or downregulated transcripts were separately analyzed. A Bayes factor cut-off of at least 6 was applied to all analyses (6).

Fig. 2.

Volcano plots showing overall differential expression of transcripts between NH-prone (VA) or NH-resistant (PV) regions compared with control vein (EJV) at days 5 and 14. A: NH-prone (VA) at day 5; B: VA at day 14; C: NH-resistant (PV) at day 5; and D: PV at day 14. Genes highlighted in red were identified by limma as being significantly differentially expressed with Benjamini and Hochberg-adjusted P values, using a paired model that controls for any animal or batch effects in the data. At day 5, limma identified 1,306 and 805 differentially expressed transcripts in VA and PV tissues, respectively, when compared with the EJV control. At day 14, limma identified 274 and 0 differentially expressed transcripts in VA and PV tissues, respectively, when compared with the EJV control.

Quantitative RT-PCR

The expression of six genes of interest was assessed by quantitative RT-PCR using the same mRNA samples used for microarray. Conversion from mRNA to cDNA was conducted using the High Capacity RNA-to-cDNA kit (Applied Biosystems). All primer and probe sets were obtained from the inventoried TaqMan Gene Expression Assay for Sus scrofa (Applied Biosystems). PCR amplification efficiencies were similar for all genes tested. Relative gene expression was normalized to the expression of the housekeeping gene, beta-2 microglobulin (B2M), and the comparative C_T method was used to quantify the fold-differences in gene expression.

RESULTS

Spatial and Temporal Gene Expression Differences between the NH-prone and NH-resistant Regions at Day 5 and Day 14 Postgraft Placement

We compared the global gene expression profile differences between NH-prone (VA) and NH-resistant (PV) regions against a control vein (unoperated contralateral external jugular vein, EJV) from the same animal after filtering the transcripts that satisfied our selection threshold criteria (adjusted P ≤ 0.05 and a log2 fold-change of ≥4 compared with EJV) at day 5 and day 14 post-AVG placement. The day 5 time point was chosen to capture early pre-NH events, and the day 14 time point was selected to capture later events after NH had developed (active NH). Analysis with limma identified 1,463 and 425 transcripts that met our selection threshold criteria from four independent microarray experiments at day 5 and day 14, respectively. These transcripts included unannotated transcripts with no accession numbers. Using the same threshold criteria, GeneSifter identified 1,392 and 787 transcripts at day 5 and 14, respectively. Comparison of genes identified by GeneSifter, but not by limma, indicates the genes not identified by limma were typically those near the threshold values of adjusted P value and fold-change. Genes identified as significantly changed by both limma and GeneSifter had similar adjusted P values and fold-changes compared with control. Figure 3 illustrates the spatial and temporal global gene expression profiles of the filtered transcripts by limma presented as a microarray heat map with the clustered gene expression profiles for the 50 genes with smallest Benjamini and Hochberg-adjusted P values. A similar clustered heat map showing all gene transcripts identified by GeneSifter is shown in Fig. 4. Clustering indicates that gene expression patterns in the NH-prone (VA) and the NH-resistant (PV) regions are more similar to each other than to the control vein (EJV) segment, suggesting a common response of the tissues at the NH-prone and NH-resistant regions to surgical perturbation and sequela. The direction of transcript expression (up- or downregulation) at either time point was, with a few exceptions, similarly regulated in both the NH-prone and NH-resistant regions. At day 14, the gene expression in the NH-prone region remained, in general, more significantly different from the gene expression in the control vein, whereas the gene expression in the NH-resistant region had become similar to that in the control vein, suggesting that the NH-resistant region is returning to the preoperative status while the NH-prone region remains stimulated and responsive to that stimulation. Complete lists, with annotations, of genes identified by GeneSifter as significantly changed in expression compared with control vein at day 5 (Supplementary File 1) and day 14 (Supplementary File 2) are provided in the Supplementary Materials. A complete list of genes identified by limma as being significantly changed compared with control at both time points is provided in Supplementary File 3. (The online version of this article contains supplemental material.)

Fig. 3.

limma analysis of spatial and temporal gene expression profiles of the NH-prone (VA) and NH-resistant (PV) segments, compared with control vein (EJV) at day 5 (A) and day 14 (B) post-graft placement. Analysis with limma identified 1463 and 425 transcripts that met our selection threshold from four independent microarray experiments at day 5 and day 14, respectively. A complete list of significantly altered annotated genes identified by limma for both time points is provided in Supplementary File 3 in the Supplementary Information. Clustering indicates that gene expression patterns in the NH-prone (VA) and the NH-resistant (PV) regions are more similar to each other than to the control vein (EJV) segment. The microarray heat map shows the clustered gene expression profiles for the 50 genes with the smallest Benjamini and Hochberg-adjusted P values that also showed at least a 4-fold differential expression compared with control vein. The color-coded scale for log-transformed data is shown at top left (red indicates increased expression, green decreased expression), with coloring based on Z-score (number of standard deviations away from the mean expression of the row). In the color key and histogram at top left of each panel, Z-scores for all 600 gene spots are divided into 20 bins, each with a different color ranging from green (positive Z score) to black to red (negative Z score). The blue line shows the distribution of the 600 gene spots across the 20 bins. Genes were clustered in R by Pearson correlation as the distance metric and complete linkage for dendrogram construction. Heat maps were produced with the gplots library.

Fig. 4.

GeneSifter analysis of spatial and temporal gene expression profiles of the NH-prone (VA) and NH-resistant (PV) segments, compared with control vein (EJV) at day 5 (A) and day 14 (B) postgraft placement. GeneSifter identified 1,392 and 787 transcripts that met our selection threshold (P ≤ 0.05 and a fold-change of ≥4 between EJV (unoperated control vein) and either the NH-prone (VA) or NH-resistant (PV) regions from four independent microarray experiments at day 5 and day 14, respectively. These transcripts included unannotated transcripts with no accession numbers. Clustering of average sample expression indicates that gene expression patterns in the NH-prone and the NH-resistant regions are more similar to each other than to the control vein, suggesting a common response of the tissues at the NH-prone and NH-resistant regions to surgical perturbation and sequela. The microarray heat map shows segments of the NH-prone and NH-resistant cluster analyses that had the greatest differences when compared with each other. The color-coded scale (Z-scale units) for log-transformed data is shown at bottom. Red indicates increased expression; green indicates decreased expression. Complete lists of significantly altered genes identified by GeneSifter for day 5 (Supplementary File 1) and day 14 (Supplementary File 2) are provided in the Supplementary Materials.

Cell Proliferation-associated Genes were More Significantly Upregulated in the NH-prone Region at Day 5 Postgraft Placement Compared with NH-resistant Region

To begin analysis of the functional changes in gene expression, we used the GeneSifter data set. The unannotated transcripts with no accession numbers were excluded from further data analyses, leaving 801 and 447 filtered transcripts at day 5 and day 14, respectively. In Fig. 5, Venn diagrams of the 801 and 447 filtered transcripts at day 5 and day 14, respectively, indicate that a substantial number of transcripts were up- or downregulated in common in the NH-prone and NH-resistant regions relative to the control vein at both time points (i.e., the overlapping regions in Venn diagrams). However, the NH-prone tissue had a greater number of genes that were significantly up- or downregulated compared with the control vein, than in the NH-resistant region. For instance, at 5 days postgraft, 327 genes were upregulated in the NH-prone region compared with 182 genes in the NH-resistant region; at 14 days, 217 genes were downregulated in the NH-prone region vs. 55 in the NH-resistant region.

Fig. 5.

Venn diagrams indicating the degree of overlap in transcript expression levels in the NH-prone (VA) and the NH-resistant (PV) segments that were spatially and temporally regulated compared with control vein (EJV). The degree of overlap of 801 and 447 filtered transcripts, which were significantly different from control vein at day 5 and day 14, respectively, as assessed by GeneSifter analysis is shown. A: at both 5 days (top row) and 14 days (bottom row) following AVG placement, large numbers of transcripts whose expressions were up- or downregulated (indicated by arrows) are regulated in common between the NH-prone and the NH-resistant regions. B: the numbers of up- or downregulated (indicated by arrows) transcripts that temporally overlap between day 5 and day 14 in each vein region are shown. Overall, a greater number of transcripts was differentially regulated in the NH-prone than in the NH-resistant segment, and a greater number was differentially regulated at 5 days than at 14 days postgraft placement.

We performed GATHER analysis to investigate functional relationships of significantly altered transcripts in each of the two vein regions. All annotated GeneSifter transcripts (a total of 625 and 387 transcripts from the NH-prone and the NH-resistant regions, respectively) were submitted to GATHER for the day 5 time point. The most significant (Bayes factor ≥6) functional annotations of upregulated transcripts in the NH-prone region (compared with the control vein) represented cell cycle regulation gene ontology (GO) terms that putatively participate in promoting cell proliferation and survival (Table 1A). Genes contributing to these terms included cyclins and cyclin-dependent kinases (e.g., CCNA2, CCNB1, CCNF, CDC2), nuclear components of mitosis and cytokinesis (e.g., BRRN1, CENPF, KIF23, KNTC2), and transcription factors known to play roles in regulating cell survival (e.g., E2F8, WT1). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis showed that the cell cycle pathway was a significant pathway activated in the NH-prone region at day 5, further supporting the GO analysis results. At day 5, a number of transcripts with cell cycle regulation GO terms (e.g., mitosis, M phase, mitotic cell cycle, nuclear division, cell proliferation) were upregulated in the NH-prone region. Several of these GO terms were also identified as being significantly enriched in the NH-resistant region, but with lower Bayes factors compared with the NH-prone region (Table 1B). Since fewer transcripts were identified by GATHER analysis in the NH-resistant region, it indicates that the association of the cell cycle regulation terms is less significant in the NH-resistant region than in the NH-prone region. No KEGG pathway terms with a significant Bayes Factor were identified in the NH-resistant region at day 5.

Table 1.

Overrepresented Gene Ontology terms for significantly upregulated genes at 5 days postgraft placement in the NH-prone VA and the NH-resistant PV segments

Annotation		PV	Bayes Factor
A. Upregulated transcripts in the NH-prone VA region
Gene Ontology ID
1. GO:0007067 [8]	mitosis	<0.0001	25
2. GO:0000087 [7]	M phase of mitotic cell cycle	<0.0001	24
3. GO:0000279 [6]	M phase	<0.0001	24
4. GO:0000278 [6]	mitotic cell cycle	<0.0001	23
5. GO:0000280 [7]	nuclear division	<0.0001	22
6. GO:0008283 [4]	cell proliferation	<0.0001	17
7. GO:0007049 [5]	cell cycle	<0.0001	15
8. GO:0000910 [5]	cytokinesis	<0.0001	8
9. GO:0000074 [6]	regulation of cell cycle	0.0001	6
10. GO:0001501 [5]	skeletal development	0.0001	6
KEGG Pathway
1. path:hsa04110	cell cycle	<0.0001	8
B. Upregulated transcripts in the NH-resistant PV region
Gene Ontology ID
1. GO:0007067 [8]	mitosis	<0.0001	11
2. GO:0000087 [7]	M phase of mitotic cell cycle	<0.0001	10
3. GO:0000278 [6]	mitotic cell cycle	<0.0001	9
4. GO:0000280 [7]	nuclear division	<0.0001	8
5. GO:0000279 [6]	M phase	<0.0001	8
6. GO:0009628 [5]	response to abiotic stimulus	0.0001	6

A: at day 5 in the neointimal hyperplasia (NH)-prone region, Gene Annotation Tool to Help Explain Relationships (GATHER) analysis of the 283 upregulated transcripts representing 261 genes relative to the control vein is shown. Annotations with Bayes factor values of at least 6 predominantly denote transcripts associated with mitosis, proliferation, and cell cycle-related events. B: at day 5 in the NH-resistant segment, GATHER analyzed 162 upregulated transcripts representing 151 genes relative to the control vein. A number of annotations with significant Bayes factor values were also associated with mitosis, proliferation, and cell cycle-related events. PV, proximal vein; VA, venous anastomosis.

Genes in Osteo-/chondrogenesis Pathways were Differentially Upregulated in the NH-prone and the NH-resistant Regions at Day 14 Postgraft Placement

At the molecular level, the spectrum of expressed genes at day 14 was expected to be different from that of the earlier 5 day period as tissue in and around the AVG began to heal and remodel. Based on the annotated GeneSifter data set, a total of 336 and 111 filtered transcripts for the NH-prone and NH-resistant regions, respectively, were altered and these were subjected to GATHER analysis for the day 14 time point. In the NH-prone tissue, GO analysis of upregulated transcripts revealed that genes associated with a group of vascular calcification pathway-related annotations (skeletal development, ossification, and bone remodeling) were significantly enriched (Table 2A). Upregulated genes contributing to these terms were also associated with less refined GO classes such as morphogenesis, organogenesis, and development (Table 2A). When each gene in these gene sets was explored, we found that several key components of the osteo-/chondrogenic pathways (e.g., CDH11, COL11A1, IBSP, SPP1) together with the upstream master transcriptional regulators of bone (RUNX2) and cartilage (SOX9) differentiation were significantly upregulated in the NH-prone region. Although RUNX2 was identified by GeneSifter analysis as being significantly upregulated at day 5 and day 14 in both regions (NH-prone, log2 ratio 4.8, P < 0.0004 and log2 ratio 4.2, P < 0.0001, respectively; NH-resistant, log2 ratio 3.5, P < 0.003 and log2 ratio 2.3, P < 0.004, respectively), this gene was not identified as being significantly changed at any time point or region by limma analysis. Of particular interest with regard to key targets for therapeutic interventions, several genes were significantly upregulated in the NH-prone region but not in the NH-resistant region at day 14. For example, the transcript for IBSP (integrin-binding sialoprotein) is still significantly upregulated in the NH-prone region, but has returned to control levels in the NH-resistant region (log2 ratio 6.5 vs. nonsignificant, respectively, by GeneSifter analysis, and log2 ratio 6.7 vs. nonsignificant by limma). In addition, GATHER identified ECM-receptor interaction as a significant pathway activated in the NH-prone region at day 14, and AP-1 transcription factor binding sites were identified in a large number of upregulated genes subjected to GATHER analysis. In the NH-resistant region, several osteo-/chondrogenic genes observed in the NH-prone region were also significantly upregulated at day 14 according to GeneSifter (Table 2B), but no genes were identified as being significantly upregulated at day 14 by the limma analysis. The greater magnitude of upregulated genes in vascular calcification-related pathways in the NH-prone vs. NH-resistant region implicates these pathways in the pathological development of NH following AVG placement.

Table 2.

Overrepresented Gene Ontology terms for significantly upregulated genes at 14 days postgraft placement in the NH-prone VA and the NH-resistant PV regions

Annotation		P Value	Bayes Factor
A. Upregulated transcripts in the NH-prone VA region
Gene Ontology ID
1. GO:0001501 [5]	skeletal development	<0.0001	10
2. GO:0009653 [3]	morphogenesis	<0.0001	10
3. GO:0007155 [4]	cell adhesion	<0.0001	10
4. GO:0009887 [4]	organogenesis	<0.0001	9
5. GO:0048513 [3]	organ development	<0.0001	9
6. GO:0007275 [2]	development	<0.0001	8
7. GO:0044237 [4]	cellular metabolism	<0.0001	7
8. GO:0001503 [5]	ossification	<0.0001	7
9. GO:0046849 [4]	bone remodeling	<0.0001	7
10. GO:0030216 [4]	keratinocyte differentiation	<0.0001	6
KEGG Pathway
1. path:hsa04512	ECM-receptor interaction	<0.0001	8
B. Upregulated transcripts in the NH-resistant PV region
Gene Ontology ID
1. GO:0001501 [5]	skeletal development	<0.0001	11
2. GO:0009887 [4]	organogenesis	0.0001	6
3. GO:0048513 [3]	organ development	0.0001	6

A: at day 14 in the NH-prone region, GATHER analysis of the 171 upregulated transcripts representing 154 genes relative to the control vein is shown. Osteo-/chondrogenic pathway gene transcripts were enriched and were associated with organ and tissue developmental terms. B: at day 14 in the NH-resistant region, GATHER analyzed 75 upregulated transcripts representing 70 genes relative to the control vein. Several of the organ developmental terms observed in the NH-prone region were also significantly enriched.

Downregulation of Genes Associated with Muscle Phenotype in the NH-prone and NH-resistant Regions at Day 5 Postgraft Placement

There were also many gene transcripts downregulated in both the NH-prone and NH-resistant regions compared with the control vein at day 5. These genes were largely represented by the morphogenesis, organogenesis, and development terms. GO terms related to muscle contractile phenotype (muscle development and muscle contraction) were also associated with many genes downregulated in the NH-prone and NH-resistant regions (Table 3). Downregulated genes represented by these GO terms revealed components of muscle cytoskeletal structure and muscle contractile function (e.g., ACTA2, ACTG2, CASQ2, DES, ITGB1BP2, MYOM1, and MYH1 were all identified by both GeneSifter and limma as being significantly downregulated); these genes were also associated with the less refined development process terms described above. Cell adhesion was a GO class identified as downregulated only in the NH-prone region, suggesting that some cells in this region might demonstrate increased cell motility and plasticity during NH pathogenesis. At day 14 in the NH-prone region, a few GO terms associated with development processes were also overrepresented, as many of the downregulated transcripts contributing to these terms at day 5 were also significantly downregulated at day 14 in this region (Table 4). There were no GO classes that satisfied our Bayes factor cut-off identified at day 14 in the NH-resistant region.

Table 3.

Overrepresented Gene Ontology terms for significantly downregulated genes at 5 days postgraft placement in the NH-prone VA and the NH-resistant PV regions

Gene Ontology ID.	Annotation	P Value	Bayes Factor
A. Downregulated transcripts in the NH-prone VA region
1. GO:0007275 [2]	development	<0.0001	18
2. GO:0009653 [3]	morphogenesis	<0.0001	17
3. GO:0009887 [4]	organogenesis	<0.0001	15
4. GO:0048513 [3]	organ development	<0.0001	15
5. GO:0007517 [5]	muscle development	<0.0001	11
6. GO:0007155 [4]	cell adhesion	<0.0001	9
7. GO:0045995 [4]	regulation of embryonic development	<0.0001	7
8. GO:0007399 [5]	neurogenesis	0.0001	6
9. GO:0030155 [4]	regulation of cell adhesion	0.0001	6
B. Downregulated transcripts in the NH-resistant PV region
1. GO:0009653 [3]	morphogenesis	<0.0001	18
2. GO:0007275 [2]	development	<0.0001	18
3. GO:0009887 [4]	organogenesis	<0.0001	17
4. GO:0048513 [3]	organ development	<0.0001	17
5. GO:0007517 [5]	muscle development	<0.0001	15
6. GO:0006936 [3]	muscle contraction	<0.0001	7
7. GO:0007399 [5]	neurogenesis	<0.0001	7
8. GO:0008016 [5]	regulation of heart contraction rate	<0.0001	7

A: at day 5 in the NH-prone region, GATHER analyzed 342 downregulated transcripts representing 303 genes relative to the control vein. Significant occurrences included GO terms related to organ morphogenesis, muscle physiology and cell adhesion. B: at day 5 in the NH-resistant segment, 225 downregulated transcripts representing 199 genes relative to the control vein were analyzed. Genes related to organ morphogenesis and muscle contractile phenotype were significantly enriched.

Table 4.

Overrepresented Gene Ontology terms for significantly downregulated genes at 14 days postgraft placement in the NH-prone VA region

Gene Ontology ID	Annotation	P Value	Bayes Factor
1. GO:0007275	development	<0.0001	9
2. GO:0009653	morphogenesis	<0.0001	9
3. GO:0009887	organogenesis	0.0001	6
4. GO:0048513	organ development	0.0001	6

At day 14 in the NH-prone region, GATHER analyzed 165 downregulated transcripts representing 148 genes relative to the control vein. No annotations that satisfied our Bayes factor cut-off were identified at day 14 in the NH-resistant region.

GO Analysis of the Genes that are Regulated Similarly in Both the NH-prone and NH-resistant Regions

Genes involved in regulating cell proliferation were upregulated in both the NH-prone region and NH-resistant region at day 5, but the relative enrichment of these genes was less prominent in the NH-resistant region compared with that in the NH-prone region (Tables 1A and 5A). In-common downregulated genes were associated with morphogenesis, development, and organogenesis GO terms at day 5 (Table 5B). At day 14, upregulated genes displayed a few overrepresented GO terms associated with development (Table 5C) that satisfied our Bayes factor cut-off. There were no significant GO terms associated with downregulated genes at this time point as indicated by either GeneSifter or limma analysis.

Table 5.

GO analysis of genes that are regulated in common between the NH-prone (VA) and the NH-resistant PV segments at 5 and 14 days postgraft placement

Gene Ontology ID	Annotation	P Value	Bayes Factor
A. Upregulated transcripts in the VA and PV regions at day 5
1. GO:0007067 [8]	mitosis	<0.0001	11
2. GO:0000087 [7]	M phase of mitotic cell cycle	<0.0001	11
3. GO:0000278 [6]	mitotic cell cycle	<0.0001	9
4. GO:0000280 [7]	nuclear division	<0.0001	9
5. GO:0000279 [6]	M phase	<0.0001	8
6. GO:0007049 [5]	cell cycle	<0.0001	6
B. Downregulated transcripts in the VA and PV regions at day 5
1. GO:0009653 [3]	morphogenesis	<0.0001	18
2. GO:0007275 [2]	development	<0.0001	17
3. GO:0009887 [4]	organogenesis	<0.0001	16
4. GO:0048513 [3]	organ development	<0.0001	16
5. GO:0007517 [5]	muscle development	<0.0001	13
6. GO:0007399 [5]	neurogenesis	<0.0001	8
C. Upregulated transcripts in the VA and PV regions at day 14
1. GO:0001501 [5]	skeletal development	<0.0001	9
2. GO:0009887 [4]	organogenesis	<0.0001	7
3. GO:0048513 [3]	organ development	<0.0001	7
4. GO:0009628 [5]	response to abiotic stimulus	<0.0001	7
5. GO:0009653 [3]	morphogenesis	<0.0001	7

At day 5 in the NH-prone and NH-resistant regions, GATHER analyzed 155 upregulated transcripts representing 142 genes (A) and 215 downregulated transcripts representing 189 genes (B) relative to the control vein. C: at day 14 in the NH-prone and NH-resistant segments, 62 upregulated transcripts representing 57 genes relative to the control vein were analyzed. At day 14 in the NH-prone and NH-resistant segments, no transcripts were downregulated in either the NH-prone or NH-resistant segments as assessed by GATHER analysis.

Quantitative RT-PCR Measurements of Selected Gene Expression

To validate our microarray results, we selected six genes of interest and performed RT-PCR using the same RNA samples that were used for microarray studies. The six genes are the key components of the osteo-/chondrogenic pathways that were significantly upregulated in the NH-prone region (IBSP, SPP1), muscle phenotype-specific genes significantly downregulated in the NH-prone region (CKM, ACTA2, CNN1), and a gene (ETS1) whose expression was not significantly altered compared with control vein in the NH-prone region or the NH-resistant region in the microarray study. The expression of the housekeeping gene B2M was used as a control gene for PCR. Figure 6 demonstrates that the quantitative PCR results generally agree well with the semiquantitative microarray studies in that IBSP and SPP1 were substantially upregulated in both the NH-prone and NH-resistant regions. The RT-PCR analysis indicated that CKM, ACTA2, and CNN1 were downregulated in the NH-prone region at day 5, but only CKM was significantly downregulated in this region at day 14. Expression of ETS1 was relatively unchanged at both time points. No substantial variation in B2M expression was detected across all samples.

Fig. 6.

Quantitative RT-PCR measurement of six select genes at day 5 (A) and day 14 (B) postgraft placement. The expression of six genes of interest was assessed by quantitative RT-PCR to confirm the results observed by microarray analysis. The genes selected were identified by microarray analysis to be upregulated, downregulated, or unchanged in the NH-prone (VA, solid black) or NH-resistant (PV, gray) segments relative to the control vein (EJV); RT-PCR measurements were performed on the same mRNA samples used for microarray experiments. The box and whisker plots represent the RT-PCR data (n = 4 for each data point), and the corresponding histograms present the fold-change in the corresponding microarray data as determined by GeneSifter. Genes that are lacking histograms (i.e., ETS1 in both A and B, ACTA2 and CKM in B) are those that GeneSifter identified as being not significantly differentially expressed compared with control vein.

DISCUSSION

Incomplete understanding of the cellular and molecular mechanisms underlying the pathophysiology of venous NH formation in AVG motivated us to investigate different genomic signatures of NH-prone and NH-resistant vein segments in a porcine model of AVG stenosis. This animal model recapitulates the pathology of human AVG stenosis, with venous NH typically being observed as early as 7 days after the AVG placement, then rapidly progressing over time, eventually resulting in significant stenosis and occlusion of the graft within several months (Fig. 1; 21, 38, 39). Accelerated development of NH in pigs enables us to efficiently study early as well as later changes in the development of NH. In our present study, comparison of the NH-prone and the NH-resistant regions revealed markedly different gene expression profiles. We also observed differential gene transcription profiles in these regions at day 5 vs. day 14 following AVG placement, which spans a period that includes the earliest histological evidence of NH development in the pig AVG model (21).

Genomic studies involving animal models that are relatively less commonly studied have faced challenges with the incomplete and sometimes incorrect annotation of their genomes. The first draft sequencing of the domestic pig (S. scrofa) genome was accomplished by the International Swine Genome Sequencing Consortium in 2009, an effort led by the Wellcome Trust Sanger Institute. Revised and improved annotations have continued to be released since then with the most recent annotation produced by the National Center for Biotechnology Information (NCBI) eukaryote genome annotation pipeline being available through NCBI RefSeq (1, 9, 42). Although an important advancement for livestock science and biomedical research, it will take time until the pig genome is as fully and accurately annotated as human or other model organisms such as mouse. Because of the current incomplete status of the pig genome annotation, our efforts to comprehensively identify the porcine genes of interest in our microarray experiments, among the 43,803 probes represented in the Agilent porcine array, has been limited. Despite this relative paucity of annotated genes in the pig genome, there were a sufficient number of annotated genes among the genes identified as being up- or downregulated in our studies to permit identification of important pathways and cellular processes involved in NH development.

Analysis of the genomic signatures revealed that the most significant genes upregulated in the NH-prone and NH-resistant regions at day 5 following AVG placement are those with strong cell proliferation-associated GO annotations. These findings coincide with previous histological studies reporting the highly proliferative early stage of AVG NH development involving proliferation of smooth muscle cells (SMCs) and/or myofibroblasts (13, 18, 21, 22, 33, 37, 38). Our present results provide insights into the specific genes and pathways implicated in the cell proliferation processes. Knowledge of these key genes and pathways can guide therapeutic strategies targeting cell proliferation processes at early stages of NH development that may allow wound healing of the tissue at the graft site, while also effectively preventing NH. Figure 1B depicts a time-course change in porcine NH formation in the NH-prone region and at least two distinct phases of NH development, pre-NH from day 0 to around day 5, and active NH from day 7 onward. Figure 7 shows the pathways and specific genes that are up- and downregulated in the NH-prone and NH-resistant regions at these two phases of NH development. The pre-NH phase is characterized by an early proliferative phase transitioning into a later active NH phase characterized by loss of muscle phenotype and cell/matrix remodeling. Based on our microarray data, we have identified genes postulated to play important roles during each phase, and those with greater expression changes occurring in the NH-prone region. It is noteworthy that the NH-resistant region also demonstrates enrichment of cell proliferation-associated genes at the pre-NH time point, although upregulation of these genes in the NH-resistant region is less prominent compared with the NH-prone region.

Fig. 7.

Gene expression changes occurring at 5 and 14 days after graft placement in NH-prone VA and NH-resistant PV regions. Heat maps show the relative gene expression of genes that GeneSifter indicated as changing the greatest at day 5 (left) and 14 (right) in the NH-prone VA region. Genes are grouped according to pathways and processes. Gene expression analysis at day 5 of the NH-prone region indicate that the pre-NH phase is characterized predominantly by activation of proliferative/survival and osteo-/chrondrogenesis pathways/processes and downregulation of genes associated with muscle contractile phenotype, cell migration, and cell differentiation. By day 14, during the early active NH development phase, genes predominantly upregulated in the NH-prone region are associated with osteo-/chrondrogenesis pathway/processes, and genes associated with muscle contractile phenotype are the genes predominantly downregulated. The names of genes demonstrating marked changes in expression at each time point are shown.

The temporal difference in the number of significantly altered transcripts at the two time points in our experiments may reflect the complexity of very early molecular events (e.g., those occurring within a few days after graft surgery) that contribute to the initial stages of NH development compared with longer term healing and remodeling processes. It also shows that there are early transient gene expression changes occurring by day 5, as well as cellular processes that begin early on and that are sustained into day 14. At day 14, we identified cellular processes that have not been extensively reported in NH pathophysiology, namely skeletal development, ossification, and bone remodeling. Increased expression of genes associated with these processes was also detected in the NH-prone region at day 5, although the relative importance of this gene set enrichment was obscured by the upregulation of a large number of genes with cell proliferation-associated terms. These data indicate that osteo-/chondrogenic activities can take place early in NH initiation and are sustained for at least 2 wk after the initiating stimulus. Of note, vascular SMCs exhibiting osteo-/chondrocytic phenotypes that correlated with the enhanced expression of mineralization-regulating proteins in calcified atherosclerotic plaques has been reported by Tyson et al. (41). The markers of calcification they tested included SOX9, RUNX2 (CBFA1), and IBSP, along with several other osteo-/chondrocyte-specific proteins. We were also able to identify a number of significantly upregulated genes in the osteo-/chondrogenic pathways from our gene set that are known transcriptional targets of the master regulatory genes, SOX9 and RUNX2. These include SPP1, MMP13, COL11A1, and SPARC, genes that were not directly tested by Tyson et al. The precise role of vascular mineralization in NH development remains under investigation, although a few different hypotheses have attempted to explain its role in response to pathogenic stimuli in arteries (15). Given the knowledge that calcification is actively suppressed in fully contractile, differentiated arterial SMCs (4, 27, 41), our findings imply that venous NH in AVG may share fundamental similarities with osteo-/chondrogenic vascular remodeling that occurs frequently in arterial diseases such as coronary and peripheral artery disease. It is worth noting that WNT/β-catenin signaling has recently been shown to promote vascular smooth muscle osteogenic transdifferentiation and calcification by modulating RUNX2 expression (5, 24, 25). In addition, the gene product of IBSP, Integrin Bone Sialoprotein, has been shown to be implicated in angiogenesis as a mediator of endothelial cell adhesion and chemotactic migration (8), and SPP1 (Secreted Phosphoprotein 1; also commonly known as osteopontin) may be involved in promoting vascular smooth muscle cell proliferation (17). SPP1 is known to be secreted by activated macrophages and T cells (26), and its upregulation may indicate the importance of inflammatory cells during early NH development. Whether IBSP, SPP1, and other factors play a pivotal causal role in NH formation requires further investigation. Regardless of whether these factors are primary or secondary mediators of NH formation, they are attractive candidate targets for therapeutic interventions to prevent NH and other pathologies that involve vascular smooth muscle proliferation.

As part of the vascular structural remodeling that occurs in response to developmental and pathogenic cues, SMCs can adapt to perform various functions that are reflected in very different SMC phenotypes, specifically contractile vs. synthetic phenotypes, with a range of intermediate phenotypes (10, 35). Previous studies have proposed that SMCs in NH lesions may dedifferentiate to more efficiently respond to injury and subsequent wound-healing activities, with corresponding increases in proliferation and migration potentials. Concurrently, the downregulation of smooth muscle contractile gene expression would be expected to be a characteristic of the dedifferentiated state (3, 14, 35, 40). In fact, many downregulated genes in our gene sets (e.g., ACTG2, DES, FHL1, MYOM1, MYH11, ITGB1BP2) are putative transcriptional targets of serum response factor, a master regulator of SMC contractile protein genes (30). Our data demonstrating the downregulation of genes coding for muscle structural components and contractile proteins support the idea that SMCs dedifferentiate soon after AVG placement and point to a strong connection between SMC phenotypic conversion and NH development in AVG. It will be interesting to know whether interventions that inhibit SMC phenotypic plasticity can attenuate the progression of NH lesions in AVG.

We originally hypothesized that the gene expression profiles of the NH-prone and NH-resistant regions would be different after AVG placement. Because we are comparing the gene expression of the NH-prone region and the NH-resistant region with the same control reference (normal EJV), changes in gene expression in the control vein reflect systemic responses to the surgery and graft and allow us to look at the common and differential changes in the NH-prone and NH-resistant regions of the corresponding vein. Due to the highly localized nature of NH development, there can be relatively few factors that could cause this difference in NH susceptibility. Candidate causative factors include the presence of surgical injuries and the foreign graft material at the NH-prone region and the more disturbed hemodynamic environment at the anastomotic region. Although the gene expression profiles indicate that a number of cellular processes appear to occur in common between the NH-prone and NH-resistant regions, the regulation of a larger number of genes was significantly altered in the NH-prone region. The larger number of genes altered in the NH-prone region reflects not only a more robust response to the NH stimulus, but also a more diverse set of cellular pathways affected in the NH-prone region compared with the NH-resistant region. Understanding these underlying molecular changes will provide a mechanistic basis for the rational development of new therapeutic strategies that can prevent the formation of NH and improve the long-term performance of AVGs, as well as provide the rationale for therapies in other vascular hyperplasia-relevant pathways involving arterial NH and NH in other types of vein graft conduits (23, 29, 31, 32).

GRANTS

This work was supported by Bioengineering Research Partnerships Grant R01HL-67646 to A. K. Cheung and by National Institute of Diabetes, Digestive and Kidney Diseases Grant R01DK-100505 to Y. T. Shiu. Microarray data collection and analysis was supported by National Cancer Institute Award P30CA-042014.

DISCLOSURES

A. K. Cheung was a member of the Data and Safety Monitoring Board for a trial on vascular graft cosponsored by Humacyte, Inc. and the National Heart, Lung, and Blood Institute as well as a member of the Clinical Events Committee and Data Safety and Monitoring Board for the Novel Endovascular Access Trial sponsored by TVA Medical, Inc.

AUTHOR CONTRIBUTIONS

S.H.K., L.L., C.M.T., Y.-T.S., A.K.C., and D.K.B. conceived and designed research; S.H.K., L.L., and D.K.B. performed experiments; S.H.K., B.A.M., and D.K.B. analyzed data; S.H.K., L.L., C.M.T., Y.-T.S., P.J.M., B.A.M., A.K.C., and D.K.B. interpreted results of experiments; S.H.K., B.A.M., and D.K.B. prepared figures; S.H.K., A.K.C., and D.K.B. drafted manuscript; S.H.K., C.M.T., Y.-T.S., P.J.M., B.A.M., A.K.C., and D.K.B. edited and revised manuscript; S.H.K., L.L., C.M.T., Y.-T.S., P.J.M., B.A.M., A.K.C., and D.K.B. approved final version of manuscript.