Differential Expression of Profibrotic and Growth Factors in Chronic Allograft Nephropathy

Abstract

Background. 

Chronic allograft nephropathy (CAN) is a multifactorial process, where both immunological and nonimmunological factors play roles. Microarrays detect thousands of genes simultaneously.

Methods. 

We have analyzed gene expression profiles of 16 kidney transplant biopsy samples with CAN by high-density oligonucleotide microarrays, comparing to six normal transplant biopsies. Eight CAN biopsies showed nodular arteriolar hyalinization and one was positive for C4d staining.

Results. 

Hierarchical clustering analysis of the 22 biopsies revealed differential gene expression patterns in CAN versus the control biopsies. However, microarray analysis did not reveal differential gene expression patterns in patients with or without arteriolar hyalinization. Fifty percent of the 100 genes with highest hybridization intensities in a C4d positive sample were related to cellular and humoral immune response. Although 212 genes were upregulated a minimum of 1.5-fold, 112 genes were downregulated in CAN samples. There was differential expression of profibrotic and growth factors that while transforming growth factor-β induced factor, thrombospondin 1, and platelet derived growth factor-C were up-regulated, vascular endothelial growth factor, epidermal growth factor, and fibroblast growth factors 1 and 9 were downregulated. Selected differentially expressed genes were confirmed in microdissected samples by real-time quantitative PCR. Immunopathologic examination of biopsies revealed strong TGF-β but decreased glomerular VEGF expression in CAN.

Conclusion. 

Microarrays might be an important tool to uncover the mechanisms of multifactorial diseases, such as CAN.

Chronic allograft nephropathy (CAN) is a multifactorial process in which immunologic (e.g., clinical and subclinical acute rejection episodes, and existence of donor specific alloantibodies) and nonimmunologic factors (e.g., delayed graft function, older donor age, and calcineurin inhibitor nephrotoxicity) contribute to the progressive demise of renal graft function (¹). Although these factors are not mutually exclusive and CAN likely results from the contribution of multiple factors, it is often difficult to determine the contribution of each factor. There is no documented treatment for CAN and it is difficult to decide whether to increase or to decrease the doses of immunosuppressive medications based on Banff classification by light microscopy. Histopathologic features of CAN are nonspecific, including interstitial fibrosis, tubular atrophy, and fibrous intimal thickening of arteries. Two pathologic features, transplant glomerulopathy and splitting and lamination of the peritubular capillary basement membrane detected by electron microscopy, may be more specific for immune-mediated CAN; while the presence of nodular hyaline arteriolar changes is more specific for calcineurin inhibitor nephrotoxicity (^1–3). Moreover, the humoral immune response plays a role in CAN as recent studies have shown positive C4d staining in biopsy samples of some CAN patients (^{3, 4}).

The pathogenesis of CAN is complex, involving many cell types and making it difficult to predict which gene products are reliable markers of CAN. Microarray technology measures the expression of thousands of genes simultaneously, and is a promising tool to understand the mechanisms of allograft rejection and tolerance (^{5, 6}). Currently two types of methods, oligonucleotide and cDNA microarrays, have been successfully used to analyze the gene expression patterns of different types of cancers, autoimmune and inflammatory diseases. Microarrays are created by in situ synthesis of oligonucleotides, or robotic deposition of cDNA spots onto a glass surface. Recent studies have applied this technology in human kidney transplantation (^7–12). We previously reported that a set of genes are persistently up-regulated in kidney biopsy samples from patients with acute rejection compared to controls using oligonucleotide microarrays (⁷). Sarwal et al. demonstrated the molecular heterogeneity of acute rejection, with at least three possible distinct subtypes indistinguishable by light microscopy, using cDNA microarrays (⁸). However, 19 CAN biopsies were clustered together without demonstrating any subsets. Scherer et al. studied the gene expression profiles of 17 protocol kidney biopsies with normal histopathology at 6 months after transplantation by oligonucleotide microarrays, and found that a set of 10 genes might predict the development of chronic rejection at 1 year after transplantation (⁹). The same group compared the CAN biopsies with normal renal allografts and demonstrated different gene expression profiles (¹⁰). Donauer et al. demonstrated two distinct subsets of chronically rejected transplants in 13 transplant nephrectomy samples by cDNA microarrays (¹¹). A more recent article by Flechner et al. showed significant up-regulation of genes responsible for immune/inflammation and fibrosis/tissue remodeling in protocol biopsies of patients with CAN in a randomized prospective trial comparing cyclosporin and sirolimus treatment at 2 years after transplantation by oligonucleotide microarrays (¹²).

In this study, our aims were to characterize the gene expression patterns of transplant kidneys with CAN in comparison to the expression patterns of transplant kidneys with normal histopathology. We further aimed to correlate the gene expression profiles of allograft biopsies with specific histopathologic features of CAN (C4d staining, nodular arteriolar hyalinization, transplant glomerulopathy, or peritubular capillary multilayering), as well as the presence of donor specific antibodies.

MATERIALS AND METHODS

All renal transplant patients undergoing kidney biopsy since July 2002 at Mount Sinai Medical School were eligible for this IRB approved study. The indication for kidney biopsy was rising creatinine and/or proteinuria. Renal allograft tissue was obtained using an 18-gauge biopsy needle. One biopsy core was immediately snap-frozen in liquid nitrogen and stored at −80°C. All biopsies were evaluated according to the Banff criteria. Additional staining was performed on paraffin sections of each biopsy for C4d using rabbit anti-C4d antibody (Biomedica, Vienna, Austria). Donor-specific anti-HLA antibodies were studied by Flowbeads (Luminex). Biopsy samples from kidney transplant recipients with normal histopathology were used as controls.

RNA Preparation and Microarray Hybridization

Total RNA was isolated using QIAGEN RNeasy kit. A complete protocol for converting total RNA into biotin-labeled RNA target that is suitable for Affymetrix GeneChip hybridization can be found in our Microarray Facility website (http://www.mssm.edu/research/resources/microarray). Briefly, total RNA was reverse-transcribed using T7-polydT primer and converted into double-stranded cDNA using Superscript Choice System (Life Technologies), with templates being used for an in vitro transcription reaction at 37°C for 8 hr to yield biotin-labeled antisense cRNA (BioArray High Yield RNA Transcript Labeling Kit, Enzo Diagnostics, Farmingdale, NY). The labeled cRNA was chemically fragmented and made into a hybridization cocktail according to the Affymetrix GeneChip protocol, which was then hybridized to U133A GeneChip probe arrays containing 16,000 genes (Affymetrix, Santa Clara, CA). The array image was generated by a high-resolution GeneArray Scanner (Agilent) and initially reviewed by Affymetrix MAS 5.0 software.

Data Analysis

Microarray data were summarized at probe level for background subtraction, normalization and log2 transformation of the expression values using the RMA (multi-array analysis) algorithm, available as part of the GeneTraffic Data Analysis Software (Iobion, CA) (¹³). Differential expression was calculated using the method of “Significance Analysis of Microarrays” (SAM) (¹⁴). Genes were ranked according to a score assigned by SAM on a basis of its change in gene expression relative to the standard deviation of repeated measurements (multiple data points obtained from different patients in defined groups) for that gene. False discovery rate (FDR) was estimated by analyzing permutations of the measurements in the dataset. A minimum of 1.5 fold change was added to the criteria for selection (mean (CAN) / mean (Control) ≥1.5). A range of smaller to larger sets of genes was reviewed by adjusting the threshold (Δ value). A cutoff of FDR < 0.05 was applied to the selection of differentially expressed genes.

Clustering analysis was done using the hierarchical algorithm provided in the GeneSight Data Analysis Software (BioDiscovery, CA). In order to reduce the data complexity, the most variable probe sets were chosen for clustering. Since the variability increases as the intensity goes lower, we first binned the microarray data into 4 subgroups based on the average of the expression values of all measurements for each gene. A conventional t test is applied to each subgroup and results in a probability rank of genes that are differentially expressed between ‘normal’ and ‘diseased’ samples after a sort on p-value. A total of 3475 transcripts comprised from transcripts with p-value less than 0.05 from each intensity group was selected and used in the clustering analysis.

Immunopathology

Paraffin sections were dewaxed and stained by immunoperoxidase using an Envision kit (Dako, Carpinteria, CA). Sections were stained using antibodies directed against human vascular endothelial growth factor (VEGF), (Santa Cruz Biotechnology, Santa Cruz, CA), and transforming growth factor (TGF)-β1 (R&D Systems, Minneapolis, MN), or control IgG. The specificity of cytokine staining was assessed by prior absorption of each antibody with the respective cytokine (TGF-β, R&D) or peptide (VEGF, Santa Cruz) used for immunization. Sections were counterstained with hematoxylin and analyzed in a blinded manner.

Real-time Quantitative PCR

We did not have adequate RNA left after microarray analysis of biopsy samples. For validation of microarray results, gene expression of selected targets were analyzed in an independent cohort of biopsies with quantitative PCR (QPCR), using microdissected tubulointerstitial segments of renal allograft biopsies available from the ERCB (European Renal cDNA Bank) (¹⁵). Human biopsies were obtained from patients after informed consent and in compliance with approval of the local ethical committees.

Microdissection of renal biopsies were stored in a commercially available RNase inhibitor (RNAlater, Ambion, Austin, TX) and were performed manually on ice under a stereomicroscope using two sterile dissection needle holders. This method offers a reliable and fast dissection of nephron segments into glomeruli and tubulointerstitial fragments (¹⁵). Effective tissue separation was verified by nephron segment specific gene expression pattern. For real-time RT-PCR only tubulointerstitial tissue was analyzed. Total RNA was isolated following the respective protocol (RNeasy mini, Qiagen, Hilden, Germany). QPCR was performed on an ABI PRISM 7700 Sequence Detection System (TaqMan, Applied Biosystems, Weiterstadt, Germany) using heat-activated TaqDNA polymerase (Amplitaq Gold, Applied Biosystems) as described previously (¹⁵). The oligonucleotide primers and probes used for human VEGF (NM_003376), matrix metalloproteinase (MMP)-7 (NM_002423), and platelet derived growth factor (PDGF)-C (AF244813) are available upon request. Commercially available TaqMan gene-expression assays were used for human epidermal growth factor (EGF) (NM_001963), fibroblast growth factor (FGF)-1 (NM_000800), thrombospondin-1 (NM_018676), TGF-β1 (NM_000660) and fibronectin-1 (NM_002026), as well as the housekeeper genes human GAPDH and 18S rRNA. All primers and probes were obtained from Applied Biosystems. All measurements were performed in duplicates. Ct values of duplicates were very similar (below 1.5 Ct) or identical.

RESULTS

Patient Characteristics

Sixteen kidney transplant recipients with CAN and 6 with normal histopathology were included in the study. Among the patients with CAN, there were 6 women and 10 men, ages between 25 and 64 (mean age 43±3 years), 9 cadaver, and 7 living transplant recipients (^{Table 1}). Fifteen patients were on cyclosporin microemulsion and prednisone, along with mycophenolate mofetil (n=9) and rapamune (n=1) and one patient was on tacrolimus, rapamycin, and prednisone. Biopsies were performed between 12–192 months after kidney transplantation (mean 73±11 months). Patient serum creatinine levels ranged from 1.1 mg/dL to 4.0 mg/dL (mean 2.9±0.2 mg/dL), with varying amounts of proteinuria, ranging from a negative dipstick in 5 patients to >300 mg/dL in 4 patients. Calculated glomerular filtration rate by Modified Diet in Renal Disease (MDRD) Formula showed patients renal function between 14–62 ml/min (26.2±3.3 ml/min). Histopathologic examination of the biopsies revealed that 3 patients had grade 1, and the remaining 13 had either grade 2 or grade 3 CAN per Banff classification. Eight patients (50%) had findings of nodular arteriolar hyalinization. Only one biopsy with CAN was positive for C4d staining. None of the biopsies demonstrated transplant glomerulopathy. The data was available for donor specific anti-HLA antibodies for 15 of the 16 patients with CAN. Two patients were positive, one with antibodies to DR8 and the other to A2. The patient with anti-DR8 antibodies was the only patient who had positive staining for C4d.

TABLE 1: Patient demographic characteristics

Among the six control patients with normal histopathology, there were two women and four men, ages between 28 and 58 (mean age 41±3 years), five living and one cadaver kidney transplant recipients. Biopsies were performed between 1–28 months (mean 10±3 months) posttransplant for increase in creatinine levels in four patients, and as a protocol biopsy in two patients. Patient serum creatinine levels ranged from 1.1 mg/dL to 1.7 mg/dL (mean 1.4±0.1 mg/dL).

Gene Expression

The amount of total RNA extracted from each sample was between 2.5–16.0 microgram (mean 5.44±0.87 microgram). Among 22,283 probe sets representing about 16,000 genes per chip, 37–59% of genes were present in all samples. Data reproducibility was confirmed by re-hybridizing two samples onto different U133A GeneChips three weeks apart. A high degree of correlation reflecting by R² of 0.983 and 0.989 was obtained from both sets of replicates. Less than 1% of 22,283 probes showed greater than 1.5-fold change within the replicates, and only 2.7% among them were overlapping, indicating the fluctuations randomly occurred.

Hierarchical clustering analysis of the 22 biopsies is shown in Figure 1. Two replicate samples of chronic allograft biopsies (D4 and D5) are clustered side to side. Control biopsies (C2, C3, C4 & C5) were clustered together, away from all the CAN biopsies. Interestingly, two control biopsies (C1 and C6) demonstrated a global gene expression pattern different from CAN samples, as well as the other control samples. Of note patient C1 had end stage renal disease due to systemic lupus erythematosus, and developed recurrent lupus nephritis at 2 months after that biopsy. C6 was taken 3 weeks after the transplantation, which was 2 weeks after Thymoglobulin induction treatment that could significantly change gene expression pattern by its strong anti-T cell, as well as broad range anti-inflammatory effects. Those four control biopsies that were clustered together were used in comparison analysis. However, using six control biopsies for comparison analysis revealed similar differential gene expression patterns with more than 80% overlapping in differentially expressed genes.

FIGURE 1.: Hierarchical clustering analysis of the 22 biopsies. Two replicate samples of CAN biopsies (D4 and D5) are clustered side to side. Control biopsies (C2, C3, C4, and C5) were clustered together, away from all the CAN biopsies (D1–D16). Two control biopsies (C1 and C6) demonstrated the global gene expression pattern different from CAN samples, as well as the other control samples.

Half of the patients with CAN had nodular arteriolar hyalinization, which may represent calcineurin inhibitor toxicity. However, microarray analysis did not reveal differential clustered gene expression pattern in patients with or without arteriolar hyalinization. While patients with lower grade fibrosis (grade 1 or 1/2, n=5), and less proteinuria (negative or 1+, n=9), compared to higher grade fibrosis (grade 2 or 3, n=11) and more proteinuria (more than 1.0 gram/day or 3+, n=7), there were no differences in clustering gene expression patterns. Only one patient with CAN had both C4d staining and donor-specific antibodies. Interestingly, gene expression data from this patient revealed that more than 50% of the upregulated genes were related to immunoglobulin structures, complement (C3, C4a, C4b, C1q, complement factor H and I), B cells (CD20 and CD48), T cell receptor, nuclear factor of activated T cells, natural killer cell, cytokine receptors (interleukin-2, 7, and 10 receptor), chemokines (monokine induced by gamma interferon, monocyte chemotactic protein-1, RANTES), chemokine receptors (CCR2 and CCR7), and B cell chemoattractant, indicating the possible importance of cellular and humoral immune mediated mechanisms in this particular patient. None of the CAN biopsies demonstrated transplant glomerulopathy, or peritubular capillary multilayering.

Four hundred and fifty-five probe sets representing 324 genes were differentially expressed in CAN biopsies compared to controls. Although 212 genes were upregulated a minimum of 1.5-fold, 112 genes were downregulated in CAN samples. The vast majority of these genes were related to cellular metabolism (53%), cell communication (32%), growth/maintenance (32%), and signal transduction (24%). Only 31 genes (9%) were related to the immune response, and of those five genes were related to the humoral immune response and five to complement activation (2%). These genes included the TGF-β induced factor, thrombospondin-1, and PDGF-C, which play important roles in the pathogenesis of fibrosis (Table 2). In addition, significant number of upregulated genes were related to fibrosis and extracellular matrix deposition, such as versican, integrins-beta 3 and 6, MMP-7, 9 and 10, laminin, fibronectin, tenascin, glypicans 3 and 4, collagen type IVα2, and disintegrin. The chemokine CXCL6, and the adhesion molecules vascular cell adhesion molecule (VCAM-1), activated leukocyte cell adhesion molecule, and selectin P were up-regulated, which might play roles in recruitment of cells into allografts.

TABLE 2: Selected upregulated genes in biopsy samples with CAN

A significant number of downregulated genes were podocyte related, such as podocin, nephrin, Wilms tumor-1, podocalyxin, and synaptopodin; or renal-specific genes, such as renin, calbindin, adrenomedullin, stanniocalcin, and kininogen (Table 3). Interestingly, VEGF, EGF, and FGF-1 and 9, insulin-like growth factor binding proteins 3 and 5 were also downregulated in CAN samples compared to controls.

TABLE 3: Selected downregulated genes in biopsy samples with CAN

Immunopathology

We decided to investigate the VEGF expression by immunopathology, due to previous studies showing increased expression of VEGF in CAN biopsies, while our microarray study indicate downregulation of VEGF in CAN samples. We used TGF-β1as a positive control due to its proven up-regulation in previous studies. Intragraft expression of 2 genes, TGF-β1 and VEGF, in paraffin sections of CAN biopsies were assessed by immunoperoxidase staining. Although TGF-β1 showed increased protein expression in all 16 CAN biopsy samples, with considerable staining of glomerular capillary loops as well as tubulointerstitial area, all control biopsies were unstained (Fig. 2). In contrast, VEGF showed comparable labeling of interstitial and vascular smooth muscles cells in both groups, but a marked down-regulation of glomerular VEGF expression in all CAN biopsies. Sections labeled with IgG or absorbed antibodies were unstained.

FIGURE 2.: Immunoperoxidase staining of two transplant kidney biopsies, CAN and normal, with TGF-β1 and VEGF. Although TGF-β1 showed staining of glomerular capillary loops, as well as tubulointerstitial area of CAN biopsy, normal allograft was unstained. In contrast, VEGF showed comparable labeling of interstitial and vascular smooth muscles cells in both biopsies but lack of glomerular staining in CAN biopsy. Sections labeled with IgG or absorbed antibodies were unstained.

Real-time Quantitative PCR

The microdissected tubulointerstitial segments of 10 renal allograft biopsies with CAN and 5 with normal histopathology were studied for selected 4 up-regulated genes (PDGF-C, thrombospondin-1, fibronectin, and MMP-7), 3 down-regulated genes (VEGF, EGF, and FGF-1), and TGF-β. There were 10 CAN patients, 8 men, 2 women, ages between 21 and 57 (mean age 39±12 years), and all were cadaveric kidney donor transplant recipients. The biopsies were done 1–3 years after transplantation. Patient mean creatinine level was 4.8±2.6 mg/dL, with a varying range of proteinuria (3 patients had nephrotic range proteinuria, and 5 patients had less than 1.0 gram/day). There were 5 control biopsies from 4 male and 1 female, ages 49–68, and all cadaveric kidney donor transplant recipients. The biopsies were done 1-14 years after transplantation (mean 6.4±5.7 years). Patient creatinine levels were between 1.6 to 1.8 mg/dl at the time of biopsy.

Increased tubulointerstitial expression of TGF-β, thrombospondin-1, fibronectin, and MMP-7 mRNA levels were found in microdissected samples of CAN compared to normal biopsies, whereas VEGF mRNA showed similar expression levels between groups and increased EGF expression in normal biopsy samples (Fig. 3). However, there was no significant difference in PDGF-C and FGF-1 expression in both groups.

FIGURE 3.: QPCR quantification of mRNA expression of six genes (EGF, VEGF, TGF-β, thrombospondin, MMP-7, and fibronectin). The graphs show expression ratios of each gene to the housekeeping gene for normal (n=5) and CAN group (n=10). Increased tubulointerstitial expression of TGF-β, thrombospondin-1, fibronectin, and MMP-7 mRNA levels were found in microdissected samples of CAN compared to normal biopsies, whereas VEGF mRNA showed similar expression levels between groups and increased EGF expression in normal biopsy samples.

DISCUSSION

We investigated gene expression profiles of human kidney transplant biopsy samples with CAN compared to control allografts with normal histopathology by high-density oligonucleotide microarrays, and found that CAN biopsy samples displayed differential gene expression patterns compared to control samples. We hypothesized that patients with the histopathologic findings of transplant glomerulopathy and/or C4d deposition, along with serologic findings of donor-specific anti-HLA, representing immune-mediated mechanisms, and patients with nodular arteriolar hyalinization, which might represent calcineurin inhibitor nephrotoxicity, display differential gene expression patterns. However, hierarchical cluster analysis did not reveal any significant difference in 8 biopsies with nodular arteriolar hyalinization, compared to 8 biopsies without findings of arteriolar hyalinization. Previous studies demonstrated that 20–60% of patients with CAN have donor-specific antibodies with or without C4d deposition (^{3, 4}). However, only one of our patients had both donor-specific anti-HLA antibodies and C4d staining, but the gene expression pattern was very impressive with the majority of highest hybridization intensities related to the cellular and humoral immune response. None of the patients had transplant glomerulopathy or peritubular capillary multilayering findings in their biopsies.

Donauer et al. demonstrated two distinct subsets of chronic rejection by cDNA arrays in 13 transplant nephrectomies comparing to normal human kidneys (¹¹). We did not identify any subsets in our CAN patients. However, there were significant differences in patient population that transplant nephrectomies were done electively after patients initiated dialysis and their immunosuppressive medications were minimized. We used the transplant kidney biopsies of patients with normal histopathology but under similar immunosuppressive medications, instead of using normal kidney biopsies as controls, because immunosuppressive treatment itself has significant effects on gene expression patterns. Scherer et al. reported that a set of 10 genes (8 upregulated and 2 downregulated) from 6-month protocol biopsies with normal histopathology might predict the development of chronic rejection at 1 year after transplantation (⁹). Those genes were APRIL (acidic protein rich in leucins), OBCML (opiate-binding protein-cell adhesion molecule like), the tumor suppressor gene NPRL2, cytokeratin 15, homeobox gene B7, prolactin receptor, and guanine nucleotide binding protein gamma 7. However, the differentially expressed 10 genes in the study by Scherer et al. were not confirmed in our study, probably due to difference in timing of biopsies, where we used the biopsies after CAN developed, comparing to protocol biopsies with normal histopathology at 6 months after transplantation. The immunosuppressive treatment was also different in their study as patients were enrolled in a multicenter trial involving Certican (everolimus). Another reason for the differences between studies is the limited number of biopsy samples. We believe that in order for microarray studies to define certain subgroups of patients for diagnosis, prognosis, and management of CAN, or find candidate genes, hundreds of biopsies should be performed in multicenter collaborative studies to overcome false positive and false negative results. The discrepancy in the outcomes of the studies using human kidney samples is the tissue sampling differences, as biopsy samples contain a mixture of different cell and tissue types, involving varying proportions of muscle, capsule, cortex and medulla of the kidney. This problem can be overcome by using laser-capture microdissected of tissue subtypes (¹⁷), although it would be technically difficult to obtain an adequate amount of RNA in a small biopsy sample. Due to these limitations, we confirmed our results in microdissected biopsy samples with similar clinical features of CAN to our patient population.

TGF-β is the most studied fibrogenic cytokine in renal disease, and its role in the development of transplant kidney fibrosis has been well established (¹⁸). Our microarray analysis demonstrated up-regulation of TGF-β induced factor and thrombospondin-1, which plays a role in the TGF-β signaling pathway. Immunopathologic examination of all CAN biopsies also had strong staining for TGF-β. Tubulointerstitial mRNA expression of TGF-β and thrombospondin-1 was also significantly increased in our microdissected CAN biopsy samples. Thrombospondin-1 has been shown to be a major activator of TGF-β-promoted renal fibrosis in inflammatory renal disease, especially activating latent TGF-β (¹⁹).

There was a significant correlation between our results and Flechner et al., especially in genes related to fibrosis and extracellular matrix deposition, such as MMP-7, laminin, fibronectin, tenascin, integrin beta 6, and collagen type IVα 2 (¹²). We have confirmed the up-regulation of MMP-7 and fibronectin in tubulointerstitial part of microdissected CAN samples. Other notable upregulated genes in both studies were thrombospondin-1, PDGF, CXCL6, and VCAM-1. The upregulation of fibrogenic genes may only indicate non-specific fibrosis rather than any specific feature of CAN, as controls with fibrosis alone were not studied. Most of our patient's biopsies were taken later after transplantation, where CAN already developed, and microarray analysis of protocol biopsies taken before the development of CAN might give more definitive information about the mechanisms involving CAN.

Our findings of downregulation of VEGF in CAN samples increased our interest to further investigate the role of VEGF in the pathogenesis of CAN and other renal pathologies. VEGF is an important regulator of angiogenesis, and promotes endothelial cell proliferation, differentiation, and survival (^{20, 21}). In the kidney, VEGF expression is most prominent in glomerular podocytes and tubular epithelial cells (²²). The lower mRNA expression levels of heme oxygenase-1 and VEGF were demonstrated in cadaveric donor kidneys compared to living donor kidneys (²³). Caramelo et al. summarized the protective role of exogenously administered or endogenous VEGF in cyclosporin nephrotoxicity in a recent review article (²⁴). The authors also reviewed the studies using anti-VEGF demonstrating enhanced cyclosporin nephrotoxicity. Those studies indicated that VEGF is critical for establishment and maintenance of the glomerular filtration barrier, and may contribute to cytoprotection by its anti-inflammatory and antiapoptotic properties. Similar to our results, VEGF was down-regulated in kidney biopsy samples with diabetic nephropathy using Affymetrix GeneChip and the results were validated by RT-PCR and immunopathology (²⁵). In contrast to these findings, VEGF was upregulated in heart (²⁶) and kidney allografts (²⁷) with chronic rejection, and anti-VEGF treatment markedly inhibited T cell infiltration of allografts and acute rejection in mice skin transplantation model (²⁸), indicating proinflammatory functions of VEGF in alloimmunity. VEGF is marginally upregulated (1.57-fold change by microarray) in CAN biopsies in study by Flechner et al. (¹²). In our study immunopathologic examination of biopsy samples showed marked downregulation of glomerular VEGF expression in CAN biopsies but comparable staining of interstitial and vascular smooth muscles cells in both groups. QPCR studies of microdissected CAN biopsy samples demonstrated similar tubulointerstitial expression of VEGF compared to controls. These results indicate that while VEGF may have proinflammatory properties in acute alloimmune response or early after transplantation, the lack of VEGF production by glomerular podocytes might diminish tissue repair capacity and contribute to the development CAN. Our studies also showed down-regulation of other growth factors, such as EGF, FGF-1 and 9, and insulin-like growth factor binding proteins 3 and 5, and the lack of EGF expression in CAN biopsy samples was confirmed with QPCR. Baelde et al. also showed down-regulation of FGF-1 and insulin-like growth factor binding protein 2 in microarray examination of human diabetic nephropathy biopsy samples (²⁵). FGF-1 and EGF play roles as angiogenic factors, and like VEGF might have cytoprotective roles involved in tubular regenerative response to injury. However, increased expression of FGF and EGF have been described in human transplant kidney biopsies with chronic rejection indicating the roles of these growth factors as fibrogenic cytokines (^{29, 30}). It is not clear if increased FGF and EGF expression is secondary to tissue repair mechanism to protect the kidney from further damage, or it mediates fibrosis.

Significant numbers of down-regulated kidney-specific or podocyte related genes are probably due to the decreased numbers of healthy glomerular and tubular cells in CAN biopsy samples. Schmid et al. investigated the gene expression profiles of podocyte-associated molecules as diagnostic markers in acquired proteinuric renal diseases (³¹).

In summary, microarray is a powerful technology that detects thousands of genes simultaneously and might be an important tool in elucidating patters for mechanism, diagnosis, prognosis, and treatment of complex, multifactorial diseases, such as CAN. This technology is hypothesis generating instead of hypothesis driven, and the genes of interest that were differentially expressed should be validated by QPCR or immunopathologic studies. Prospective studies starting microarray analysis of transplant kidney biopsies taken at implantation and follow-up protocol biopsies until CAN develops will not only define the known and/or novel pathways of graft rejection or acceptance, but also identify better surrogate markers of graft dysfunction in the future.

ACKNOWLEDGMENTS

We thank Dr. Jonathan S. Bromberg for his critical review of the manuscript.