Transcriptional analysis of the molecular basis of human kidney aging using cDNA microarray profiling

Kidney International, Vol. 68 (2005), pp. 2667–2679

Transcriptional analysis of the molecular basis of human kidney aging using cDNA microarray profiling ANETTE MELK,1 ELAINE S. MANSFIELD,1 SZU-CHUAN HSIEH, TINA HERNANDEZ-BOUSSARD, PAUL GRIMM, DAVID C. RAYNER, PHILIP F. HALLORAN, and MINNIE M. SARWAL Division of Nephrology and Transplantation Immunology, Department of Medicine, University of Alberta, Edmonton, Alberta, Canada; Children’s Hospital, University of Heidelberg, Heidelberg, Germany; Department of Pediatrics, Stanford University School of Medicine, Stanford, California; Department of Genetics, Stanford University, Stanford, California; Department of Pediatrics, University of California, San Diego, California; and Department of Pathology and Laboratory Medicine, University of Alberta, Edmonton, Alberta, Canada

Transcriptional analysis of the molecular basis of human kidney aging using cDNA microarray profiling. Background. The molecular basis of renal aging is not completely understood. Methods. We used global gene expression monitoring by cDNA microarrays to identify age associated genes in human kidney samples. Our samples included young (8 weeks–8 years, N = 4), adult (31–46 years, N = 7), and old kidneys (71–88 years, N = 9). Results. Old kidneys had more glomerulosclerosis, tubular atrophy, interstitial fibrosis, and fibrous intimal thickening in small arteries. We identified ∼500 genes that were differentially expressed among the three age groups. Old kidneys appeared to have increased extracellular matrix turnover and a nonspecific inflammatory response, combined with a reduction in processes dependent on energy metabolism and mitochondrial function. Quantitative supervised bioinformatics analyses of adult and old kidney expression data correlated the expression of 255 gene profiles with renal pathology scores. Microarray class prediction analysis (PAM) identified 50 unique genes that segregated old kidneys into two distinct clusters: those more similar within age class (OO, N = 5) versus old kidneys more similar to adult kidneys (OA, N = 4). The expression of six functionally significant genes was further validated by quantitative reverse transcription-polymerase chain reaction (RT-PCR) (FN1, MMP7, TNC, SERPIN3A, BPHL, CSPG2) in the experiment group and, subsequently, confirmed independently in 17 additional old and adult age-stratified test kidney samples. The p53 inducible gene, CSPG2, performed best in separating OO kidneys from adults and OA samples in this analysis. Conclusion. The method described in this study using independent validation samples can be envisioned to test utility of the identified genes in assessing age-related changes that contribute to decline in renal function. 1 A.M.

and E.S.M. are joint first authors on this paper.

Key words: aging, kidney, microarray, expression profile. Received for publication February 5, 2005 and in revised form June 15, 2005 Accepted for publication July 20, 2005
C

2005 by the International Society of Nephrology

Age-associated changes in kidney are important not only because normal aging alters renal function but also because of the high frequency of end-stage renal disease (ESRD) in the elderly [United States Renal Data System Report (USRDS) 2003]. The incidence of acute renal failure is higher in older subjects (USRDS Report 2003), and the clinical outcome of renal failure is significantly worse with regard to mortality, requirement of dialysis, and complete recovery [1]. Old kidneys also perform poorly when transplanted, and donor age remains a major determinant for donor selection and of graft survival [2]. Interactions between aging and diseases may also contribute to these problems. Understanding the mechanisms underlying age-related decline in organ function and the role of stress in accelerating these changes may be relevant to the mechanisms of progression of chronic renal insufficiency. Kidney aging is also of interest as a general model for organ aging because renal function can be assessed with relative ease in clinical practice, and has been quantified in longitudinal studies [3]. The phenotype of human renal aging is a phenotype of loss: loss of mass, particularly in cortex [4, 5], and loss of function, with increased renal vascular resistance, reduced renal plasma flow, and increased filtration fraction [5, 6]. Estimates of glomerular filtration rate (GFR) using multiple equations such as Cockcroft-Gault [7] and Modified Diet in Renal Disease (MDRD) [8] equation show declines in GFR with age. Studies in selected populations ([9] and Baltimore Longitudinal Study of Aging [4]), excluding all renal diseases, hypertension, and heart failure, demonstrated mean loss of glomerular filtration rate of 0.75 mL/min per year [4]. They also found that one [4] to two [9] thirds of the elderly manifested little if any decline in glomerular filtration rates. The principal histologic features of renal aging are glomerulosclerosis, tubular atrophy, interstitial fibrosis, and fibrous thickening of the intimal of arteries [10]. Although it is unclear whether these changes are primary or secondary events,

2667

2668

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

they do reflect degeneration and loss of nephrons. In the presence of age-related diseases, such as hypertension and heart failure, these changes can be accelerated [9]. The molecular basis of renal aging is not known. The decrease in renal hemodynamics could be explained by either potentially reversible functional changes due to an imbalance between factors controlling vasodilatation and vasoconstriction, or they could be the result of the irreversible morphologic changes. We have shown previously that kidney aging is associated with changes that are characteristic of somatic cell senescence and irreversible cell cycle arrest in vitro [11–13]. We found telomere shortening and increased expression of the cell cycle regulator p16INK4a in kidneys from older individuals with normal renal function and histopathologic features within the range expected for age, suggesting that renal function under normal conditions is not affected by these features [12, 13]. This is supported by the clinical observation that ESRD, although greatly increased in the elderly, is still uncommon, and that the majority of the elderly without other age-related diseases have normal renal function (USRDS Report 2003). We have interpreted these findings to indicate that telomere shortening and p16INK4a expression results in a limited ability of the aged nephron to replicate and thereby lead to the diminished capacity of older kidneys to cope and withstand normal and abnormal stresses [14, 15]. The objective of the presented study was to extend our knowledge on the molecular basis of renal aging. We were looking for molecular markers that add to the existing measurements of renal age, such as histologic findings and chronologic age, and might predict how a kidney will function under certain stresses (e.g., transplantation or renal disease).

METHODS Kidney samples We studied a total of 20 human kidney cortex samples that reflected three different age groups: young (mean age 3–1 years; N = 4), adult (mean age 38.1 years; N = 7), and old (mean age 78.5 years; N = 9; Table 1). All individuals were Caucasian, except for two children that were of Native Canadian descent. None of the patients had diabetes mellitus or uncontrolled arterial hypertension. Samples were derived from total nephrectomies (N = 18) and autopsies (N = 2). In the case of tumor nephrectomies, normal renal tissue remote from the tumor was chosen for analysis. The histology of all 20 samples was within normal limits of changes expected for age [16–20]. All tissue samples were snap-frozen in liquid nitrogen and stored at −70◦ C for transcriptional studies, and paraffin embedded for histopathologic assessment. The study was approved by the Health Research Ethics

Table 1. Demographic data and histological grading Young (N = 4) Age years Mean ± SD 3.1 ± 3.5 Range 8 weeks–8 years Gender f/m 1/3 Creatinine lmol/L 28.0 ± 11.4 Systolic blood 84.8 ± 10.1 pressure mm Hg Diastolic blood 41.0 ± 5.9 pressure mm Hg Proteinuria None History of hypertension 1 (25%) Glomerulosclerosis % Mean ± SD 0.33 ± 0.58 Range 0–1 Interstitial fibrosis % Mean ± SD 0 Range 0 Tubular atrophy % Mean ± SD 0 Range 0 Fibrous intimal thickening % Mean ± SD 0 Range 0 Arteriolar hyalinosis score Mean ± SD 0 Range 0

Adult (N = 7)

Old (N = 9)

38.1 ± 5.1 75.8 ± 5.3 31–46 years 71–88 years 4/3 5/4 91.9 ± 21.9 91.9 ± 16.9 124.6 ± 20.4 145.9 ± 23.4 76.6 ± 9.1

78.4 ± 15.0

None 1 (14%)

None 11 (44%)

2.57 ± 1.72 0–5

8.22 ± 3.07 0–11

0.57 ± 1.13 0–3

3.56 ± 1.94 2–8

1.21 ± 2.31 0–6

3.78 ± 1.86 1–6

5.29 ± 5.19 17.63 ± 14.92 0–14 0–51 0.17 ± 0.41 0–1

0.88 ± 0.64 0–2

Detailed information on histopathologic chronicity scores for each individual study subject can be found in Supplemental Table 1.

Board of the University of Alberta, Edmonton, and the Institutional Review Board at Stanford University. Pathology Histopathologic assessment of age-related renal changes was performed for all four compartments of the kidney: glomeruli, tubules, interstitium, and vessels, in a blinded manner by a renal pathologist (D.C.R.) These scores have been listed as mean pathology scores for age (Table 1), as well as individual pathology scores in the study subjects analyzed (Supplemental Table 1 at http://microarray-pubs.stanford.edu/renalaging/). Interstitial fibrosis and tubular atrophy were expressed as percentage of parenchymal involvement based on examination of at least 10 high power fields (magnification 400×). In addition, interstitial fibrosis was assessed using Sirius Red staining (see below). Glomerulosclerosis was expressed as percentage of globally sclerosed glomeruli based on all glomeruli present in the section. We counted a mean of 143 glomeruli per sample with a range from 18 to 345 glomeruli present in the cortex. Fibrous intimal thickening was assessed as percentage of occlusion of arterial lumen across all vessels present in the section. We assessed between 3 to 18 cross-sections per specimen. Arteriolar hyaline thickening was assessed based on the Banff classification [16], with 0 = no PAS positive

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

hyaline thickening; 1 = mild-to-moderate PAS-positive hyaline thickening in at least one arteriole; 2 = moderateto-severe PAS-positive hyaline thickening in more than one arteriole; 3 = severe PAS-positive hyaline thickening in many arterioles.

Sirius red staining Unstained paraffin-embedded sections were baked at 60◦ C for one hour. Slides were soaked in xylene for 24 hours followed by acetone for 24 hours to reduce paraffin contamination. Then, slides were taken through xylene and graded ethanols into distilled water and stained overnight in saturated picric acid with 0.1% Sirius Red F3BA (Aldrich Chemicals, Oakville, Ontario, Canada). Slides were washed in hydrochloric acid (0.01N) and rapidly dehydrated and mounted. Image analysis was performed by a technician blinded to the source of the sample. The slides were examined with a Nikon E600 microscope (Nikon, Mississauga, Ontario, Canada), and a Hitachi analogue 3 CCD camera (Hitachi, Mississauga, Ontario, Canada) was used to capture gray scale 256-bit images that were archived as TIFF files. A background image was initially obtained and background correction was performed in real time while the images were being acquired using the 40× objective. Images of the kidney cortex were obtained in a serpentine fashion starting at one end of the tissue and working toward the other. Large glomeruli, vessels larger than the size of adjacent tubules, and the medulla were not included in the image acquisition. Image analysis was performed using an automated macro (available from P.G.) specially written for the software package NIH Image (National Institutes of Health, Bethesda, MD, USA). Automated analysis of the images was performed with operator supervision. Data are expressed as cortical fractional interstitial fibrosis volume (V IntFib ); a value of 1 would reflect 100% fibrosis.

RNA extraction and microarray hybridization Total RNA was extracted from tissue samples as described previously [11, 17]. RNA quality was checked by Northern blotting. Isolation of mRNA from total RNA was done using Invitrogen Kit (Invitrogen Canada, Inc., Burlington, Ontario, Canada). A “common reference” RNA pool [18] was used as an internal standard for microarray experiments. The microarray platform was the human 42,000 cDNA array, containing 29,043 unique human genes and 13,103 genes with known function [19]. Standard protocols for amplification, labelling, and processing of cDNA microarrays were used as previously described [20] and can also be found in further detail at http://microarray-pubs.stanford.edu/renalaging/.

2669

Microarray analysis All microarray data were stored in the Stanford Microarray database [21, 22] and gene lists filtered at retrieval. A subset of 2778 cDNA clones were found to be differentially expressed using a cutoff of to log2 >1 in at least one array and 80% representative data. The expression data were subjected to statistical analysis to identify differentially expressed genes, specific for age class, using the statistical analysis of microarray (SAM) [23], predictive analysis of microarray (PAM) [24], and singular value decomposition (SVD; [25]) programs. Cluster (version 3.0) generated hierarchic clusters by measuring similarity of expression of the genes and similarity across arrays that were displayed with the TreeView program [26]. Real-time reverse transcription (RT) polymerase chain reaction (PCR) Transcription into cDNA was done using MMLV reverse transcriptase and random primers (Life Technologies, Burlington, Ontario, Canada) [27, 28]. cDNA was amplified in an ABI PRISM 7700 Sequence Detector (Applied Biosystems, Foster City, CA, USA). Assay on Demand reagents (Applied Biosystems) were purchased. Quantification of gene expression was performed using the delta Ct method (Applied Biosystems). Each gene of interest was normalized to a housekeeping gene (hypoxanthine guanine phosphoribosyl transferase, HPRT). All samples were analyzed in duplicate. Immunohistochemistry Immunoperoxidase staining for CD79 was performed using 2 lm sections of paraffin-embedded tissue using standard protocols based on antigen retrieval. Kidney tissue sections were cooked in citrate buffer, immersed in 3% H 2 O 2 in methanol, blocked with 20% normal goat serum, incubated for two hours at room temperature with the primary antibodies (CD79a; Dako, Mississauga, Ontario, Canada) and rinsed with phosphate-buffered saline (PBS). Following 30 minutes of incubation with the Envision monoclonal system (Dako), sections were washed again in PBS. Visualization was performed using the DAB substrate kit (Dako). The slides were counterstained with hematoxylin and mounted. Statistical data analysis The SAS statistical analysis package and Multitest procedure (SAS Institute, Inc., Cary, NC, USA) were used to determine statistical significance of PCR and immunohistochemistry data. Individual PCR measurements were averaged and groups compared using Student t test. An F-statistic cut-off of P < 0.05 was used as threshold for determining significance. Statistical utilities in Excel (t test,

2670

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

Cell maintenance 14%

Cellular transport 6%

ESTs 33%

Cell proliferation 7% Cell adhesion, cell-cell interaction 3% Apoptosis 2% Other known genes 6% Immune/Inflammatory response 5% Stress response 2%

Growth factors 2% Extracellular matrix components and modulators 3% Energy metabolism and mitochondrial genes 17%

Fig. 1. Pie diagram for genes associated with renal age. Statistical Analysis of Microarray (SAM) multiclass response analysis identified 505 transcripts in kidney samples from three age groups (young, N = 4; adult, N = 7; old, N = 9) to be significantly differentially expressed (q<5%). The cellular function of these genes is reflected in this pie diagram. See supplemental Table 2 (http://microarray-pubs.stanford.edu/renalaging/) for full description of the genes, significance scores, and average fold differences among the three sample groups.

correlation, linear regression, coefficient of variation and variance; Microsoft Corporation, Redmond, WA, USA) were also used to rank significance of genes. The SAM quantitative response was used to correlate gene expression measurements with clinical parameters reported in Supplemental Table 1 (http://microarray-pubs. stanford.edu/renalaging/) to generate the pathologyassociated gene lists. The two-class (unpaired) and multiclass response utilities in SAM (version 1.12) were used to assess gene expression fold-differences and significance of genes correlating with age groups. Functional enrichment of genes based on gene ontology annotations were assessed using the EASE expression analysis explorer software tool ([29]; http://apps1.niaid.nih.gov/david/). All supplemental tables and figures can be accessed at http://microarray-pubs.stanford.edu/renalaging/. RESULTS Histopathology analyses Old kidneys had significantly greater chronicity scores for all parameters studied when compared to adult kidneys: the mean values for sclerosed glomeruli were 8% (P < 0.05 vs. adult), for interstitial fibrosis and tubular atrophy 4% (P < 0.05 vs. adult), and for fibrous intimal thickening 18% (P < 0.05 vs. adult). These values were similar to published histopathologic changes in

the kidney with age [30–34], though glomerulosclerosis was lower in our old kidney cohort (Table 1) because up to 20% of glomerular sclerosis is considered within the normal range for age [30–32]. As 18 samples were derived from tumor nephrectomy’s “normal” renal tissue remote from tumor, a potential problem of excess global sclerosis existed. Although this caveat cannot be excluded, given the low glomeruloclerosis scores, it does not seem likely. Consistent with age, the 8-week-old infant showed some immature juvenile glomeruli with prominent cuboidal epithelial cells. Interstitial fibrosis and tubular atrophy never exceeded 10% of the total area analyzed in any of the samples studied. Mean and standard deviation of scores for each pathologic parameter are shown in Table 1. Individual chronicity scores for each sample are also shown in Supplemental Table 1 (http://microarray-pubs. stanford.edu/renalaging/). Microarray analyses Hierarchic unsupervised clustering of the expression data from the 20 human renal cortex samples did not cluster samples by age groups (Supplemental Figure 1A and B; http://microarray-pubs.stanford.edu/renalaging/), suggesting that age specific transcriptional changes are dominated by expression differences relating to patient heterogeneity. Age-related transcriptional changes were

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

thus queried by supervised multiclass analysis across the three age groups using SAM. This identified a renal-aging gene set of 505 genes with high significance scores (q score <5%). Functionally characterizing these genes by ontology groups suggests that the age-specific changes largely relate to changes in cell integrity, cellular proliferation, cell transport, and energy metabolism (Fig. 1; for details of all genes see Supplemental Table 2; http://microarraypubs.stanford.edu/renalaging/). Separation of genes specific to kidney age-groups is discussed below and shown in Figure 2. Renal transcriptional changes with age In this relative comparison of transcriptional changes relating to age, many cellular and metabolic processes showed declining expression with age, with 67% of the 505 gene-set down-regulated in old kidneys. Significant genes within this old kidney dataset suggested a reduction in energy metabolism via the respiratory chain [electron transfer flavoprotein-b (ETFB) and NADH dehydrogenase 1a (NDUFA10)], altered glucose and lipid metabolism (for example, aldolase-B and -C and apolipoprotein-D, -E, and -M), as well as reduced protein, amino acid, and nucleotide turnover. Examples of the latter genes include prenyl-cysteine oxidase 1 (PCL1), kynureninase (KYNU), formiminotransferase cyclodeaminase (FTCD), phenylalanine hydroxylase (PAH), aromatic L-amino acid decarboxylase (DDC), and cytosine-5-methyltransferase (DNMT2). In keeping with increased fibrosis, glomeruloclerosis, and tubular atrophy in old kidneys, there was altered expression of cytoskeletal genes [including kelch-like 3 (KLHL3), dynein light chain (DNALI1), claudin-8 (CLDN8), kinesin family member 13B (KIF13B), developmentally down-regulated 5 (NEDD5), microtubuleassociated protein tau (MAPT), and phosphatidylinositol-4-phosphate 5-kinase (PIP5K1B), which is thought to be involved in the organization of the actin cytoskeleton]. In addition, there is evidence of increased extracellular matrix synthesis and turnover [genes include matrix metalloproteinase 7 (MMP7), fibulin-1 and -2, fibronectin 1 (FN1), tenascin (TNC), chondroitin sulfate proteoglycan 2 (CSPG2), and alpha-1-antichymotrypsin, or serine proteinase inhibitor A3 (SERPINA3)], as well as reduced collagen catabolism via the PEPD peptidase-D gene. The insulin/insulin-like growth factor pathway has previously been implicated in aging; we found decreased expression of the tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein YWHAZ that interacts with insulin receptor substrate proteins IRS1 and IRS2, which are known to be involved in insulin signaling. Altered tubular function in old kidneys is likely due to a combination of tubular loss, altered energy consumption (as shown above), and a reduction in the tubu-

2671

lar transport of electrolytes, ions, glucose, amino acids, and organic acids manifested decreased expression of 12 different solute carrier family genes in addition to the lysosomal ATPase, H+ transporting genes ATP6V1G3 and ATP6V1A1, the potassium voltage-gated channel KCNAB2, and glucose-6-phosphate transporter G6PT1 (see also Supplemental Table 2, http://microarray-pubs. stanford.edu/renalaging). An age-related loss of glutathione synthesis had been described [35], leading to an increased susceptibility to oxidative stress because of decreased antioxidative capacity. In old kidneys, decreased GSH levels were due to decreased c-GCS activity [36]. Confirming this finding, we found reduced expression in the old kidneys for the glutathione-related enzymes GSTT1 and GSTA3, which play a protective role against oxidative stress, and aldehyde oxidase 1 (AOX1), which produces hydrogen peroxide and can catalyze the formation of superoxide. A surprising finding was evidence of an increased immune and inflammatory response in old kidneys [immunoglobulin genes: IgLa, IgKC, IgJ, IgLJ3, the T cell receptor delta locus (TRD), high-affinity IgE receptor FCER1A, immediate early protein gene ETR101, and the natural killer cell activator PRG2]. These changes may be a nonspecific inflammatory response to chronic tissue injury with age. Genes whose products affect telomere function and maintenance such as POT1, TERF1, and TERF2 were not altered. P16INK4a , the gene that has previously been shown to have increased expression with renal aging [11, 13], was not present on the array platform for study. The somatostatin receptor, protein serine/threonine kinase HAK, and Thy-1 cell surface antigen (THY1) genes all show increased expression in senescent fibroblasts [37] and were increased in old kidney samples, with the greatest difference between the young and old kidneys. These genes are among a mitogen-activated protein kinase cascade recently implicated in experimental mesangioproliferative glomerulonephritis [38]. Unlike old kidneys, adult kidney samples showed robust expression for genes associated with epithelial transport (glucose-6-phosphate transporter G6PT1; H+ transporter ATP6VIG3, cytochrome P8B1), mitochondrial function (PCK2, MIRO-2, electron transporter ETFB, NDUFA10), hormone synthesis (corticotrophin and thyrotrophin proteins, and TRHDE), tubular function, and urinary concentration (natriuretic peptide receptor, NPR2, SLC6A12). Young kidney samples, ranging in age between 8 weeks and 8 years, showed considerable intergroup variability in expression profiles. Fetal collagen COL3A1 and cyclin B2 and CDC25, the latter promoting the cell cycle at the G2/M transition [33], were significantly increased in young relative to the adult kidneys (SAM significance q = 1.3%). Young kidneys, unlike the adult kidneys, showed evidence of increased extracellular matrix synthesis and

2672

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

Table 2. cDNA clones identified that differentiate kidney samples with high- and low- histopathology scores from elderly patients Symbol CSPG2 IGLJ3 CDW52 TNC TNFRSF12A IGJ CDC25C ESPL1 FN1 COL16A1 LTBP1 DGKA SLC31A1 GDA PLGL CALML3 CDR1 BPHL ASS AK3 CYP8B1 BHMT2 PDCD8 PRKWNK1 ALDH6A1 SLC13A3 GPD1 FABP1 PLG LOC113251 HCN3 DDC APOE PSA RPC8 DIO1 NEO1 PCYT1B SLC7A7 G6PT1 HNF4A NEDD5 SGK2

Definition Chondroitin sulfate proteoglycan 2 Immunoglobulin lambda joining 3 CAMPATH-1 antigen Tenascin C (hexabrachion) Tumor necrosis factor receptor 12A Immunoglobulin J peptide (alpha and mu) M-phase inducer phosphatase 3 Extra spindle pole like Fibronectin 1 Collagen, type XVIa1 Latent transforming growth factor binding Diacylglycerol kinase alpha Copper transporter member 1 Guanine deaminase Plasminogen-like Calmodulin-like 3 Cerebellar degeneration-related Serine hydrolase Argininosuccinate synthetase Adenylate kinase 3 P450 Sterol 12-alpha-hydroxylase Betaine-homocysteine methyltransferase Programmed cell death 8 Protein kinase, lysine deficient 1 Aldehyde dehydrogenase 6A1 Sodium-dependent dicarboxylate transporter Glycerol-3-phosphate dehydrogenase Fatty acid binding protein Plasminogen c-Mpl binding protein Nucleotide-gated potassium channel Dopa decarboxylase Apolipoprotein E Phosphoserine aminotransferase RNA polymerase III subunit RPC8 Deiodinase, iodothyronine Neogenin homolog 1 Phosphate cytidylyltransferase Cationic amino acid transporter Glucose-6-phosphate transporter Hepatocyte nuclear factor 4 Neural precursor cell expressed Serum/glucocorticoid regulated kinase

Fold

Accession

3.64 3.55 3.33 3.24 3.20 3.17 3.04 2.74 2.38 2.33 2.31 2.10 −4.21 −4.03 −3.87 −3.51 −3.37 −3.04 −2.95 −2.88 −2.77 −2.75 −2.73 −2.65 −2.58 −2.53 −2.50 −2.50 −2.50 −2.47 −2.47 −2.36 −2.35 −2.33 −2.31 −2.29 −2.28 −2.23 −2.16 −2.13 −2.12 −2.10 −2.08

AA056022 T67053 AI826477 AW075585 R33355 T70057 W95000 AI816969 R62612 AA088202 AW025698 AA026130 AA608729 R12516 H90507 AA055522 AA985354 AA169798 AA676466 AA947132 AA928708 AA745686 AA932997 H89604 AA196160 H18950 AI300352 T53219 R91117 AI240426 H77541 AI261686 AA947607 AI015679 H99699 N74025 AA447658 AI247953 T98393 AA490159 AI301528 T39376 AA045601

Clone IMAGE:377701 IMAGE:66560 IMAGE:2417330 IMAGE:2577230 IMAGE:135791 IMAGE:80948 IMAGE:415102 IMAGE:2409751 IMAGE:139009 IMAGE:488258 IMAGE:2516260 IMAGE:366944 IMAGE:950924 IMAGE:25984 IMAGE:241537 IMAGE:377348 IMAGE:1629478 IMAGE:610097 IMAGE:882522 IMAGE:1553394 IMAGE:1553245 IMAGE:1322845 IMAGE:1554541 IMAGE:240174 IMAGE:627687 IMAGE:51406 IMAGE:1901097 IMAGE:68557 IMAGE:195040 IMAGE:1856516 IMAGE:234522 IMAGE:2028959 IMAGE:1590208 IMAGE:1636108 IMAGE:262932 IMAGE:296702 IMAGE:784959 IMAGE:1848986 IMAGE:122063 IMAGE:839980 IMAGE:1899437 IMAGE:60565 IMAGE:509610

Age Yes Yes Yes Yes No Yes Yes Yes Yes No No No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes No Yes Yes Yes Yes Yes No

Fold expression differences between adult and high-histopathology old sample groups are reported.

turnover across a similar gene-set, as seen in old kidneys (MMP7, FN1, TNC, CSPG2, SERPINA3, OSF-2) and growth (LCN2). c-FOS and JUN, genes downstream in the TGFb-signaling pathway, were over 3-fold higher in the young than adults and up to 5-fold higher in the young than in the old (SAM multiclass response significance scores q = 1.5% and 2.36%, respectively). Immunohistochemistry for CD79 To better understand the foundation of differences observed in inflammatory response genes within the kidneys, immunohistochemistry for the B cell and plasma cell surface antigen CD79 was performed on all samples. A strong correlation was seen between CD79 staining (number of cells/mm2 by blinded biopsy analysis) and interstitial fibrosis scores (R2 = 0.79), with a weaker correlation of CD79 density and age (R2 = 0.55). This suggests

that the increased expression of immunoglobulin genes seen in the old kidneys may be secondary to increased number of activated B cells and/or plasma cells in aging kidneys (data not shown). Correlation of gene expression signatures with histopathology The degree of glomerular sclerosis, interstitial fibrosis, tubular atrophy, and fibrous intimal thickening were correlated with gene expression levels across adult and old renal samples (see Methods for data filtering criteria used). Using quantitative response analysis for the four pathology groups, SAM identified total of 255 genes; 163 genes correlated with interstitial fibrosis, 83 with fibrous intimal thickening, 38 genes with glomerular sclerosis, and 8 genes with tubular atrophy (FDR threshold of q<12.5% used for all datasets; the functional

2673

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

Young Y1-Y4

Adult A1-A7

Old O1-O9

>1.75

1:1

<1.75

Growth - EGR1, MYC, JUNB, GADD45B, ATF3 Cell adhesion - CYR61, SELE

Cell maintenance - UGT2B7, UGT2B15, UGT2B10, UGT2B17, CYP8B1, CYP4A11 - GSTA3 Glutathione S-transferase Energy metabolism - ALDOB, MIPEP, GLYAT Plasminogens - PLG, PLGL - BPHL Biphenyl hydrolase-like - ASS Argininosuccinate synthetase Ion and Sugar Transporters - SLC13A3, G6PT1, SLC9A2, SLC2A5, SLC7A7

- GSTT1 Glutathione S-transferase theta - HAK Heart alpha-kinase B-cell - IGJ, IGLα, IGKC, IGLJ3, IGHG3 ECM - FBLN2, Collagens, FN1, MMP7 Cell adhesion - TNC, ITGAM, CSPG2 Fig. 2. Kidney samples from three age groups. Expression measurements of 505 differentially expressed genes identified by multiclass analysis using the Statistical Analysis of Microarray program (SAM version 3, 2003) were clustered by expression similarity. The color in each cell reflects the expression level of the genes in the corresponding sample, relative to its mean expression level across the entire set of biopsy samples. Gray represents missing or excluded data. The log2 fluorescence ratio scale extends from of 0 to 2 relative to the mean level for all samples and the positions of major groups of genes are indicated. Names of the genes confirmed by real-time PCR in this study are underlined. All expression data in this cluster are available on the web supplement at http://microarray-pubs.stanford.edu/renalaging/.

2674

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

IF 163 genes

6

GS 38 genes

32

126 3 5 TA 8 genes

2 81

FIT 83 genes Fig. 3. Genes associated with four renal pathology measurements. Renal tissue adjacent to that sampled for mRNA expression analysis were scored for glomerulosclerosis (Gs), tubular atrophy (TA), interstitial fibrosis (IF), and fibrous intimal thickening (FIT). Correlation of these pathology scores to gene expression was performed by SAM quantitative response and a significance threshold of q<12.5% set for each parameter. The number of cDNA clones associated with each parameter is listed, as well as the number of genes in common between each group.

gene ontology groups are also shown for this dataset in Supplemental Table 2; http://microarray-pubs.stanford. edu/renalaging/). About half of this renal-pathology gene set (N = 128/255) overlaps with the renal-aging gene set (N = 505 genes). Similar ontology groups (multiple genes involved in apoptosis, cell adhesion, ion and glucose transport, cell proliferation, extracellular matrix, energy metabolism/mitochondrial genes, growth factors, stress response, and inflammatory response) were represented in both gene sets. Figure 3 shows that genes specifically altered in expression in kidneys with glomerulosclerosis and interstitial fibrosis have the greatest degree of overlap, and this is also reflected in the high degree of correlation between these two parameters across the full data set (R2 = 0.72). Hierarchic clustering across the 255 renal-pathology gene set was performed to examine the expression sim-

ilarity across adult and old samples. The adult samples clustered together while old samples split into two groups (Fig. 4A). Similarly, hierarchic clustering of the renalaging gene set across only the adult and old samples showed identical groupings of the old samples (Fig. 4B). To understand which histopathologic parameter might be driving separation of the old samples, samples were reclustered across genes specific for each of the four chronicity scores (data not shown). Interstitial fibrosis was the only parameter that groups the old samples consistent with both the age-related genes and the composite pathology set, suggesting that interstitial fibrosis drives the clustering of both sets of genes. To identify the most highly contributing genes, we used the class prediction tool PAM to identify a minimum set of 79 cDNA clones (∼37% are ESTs) (Fig. 4C), and the high confidence class prediction scores correlate with this grouping (Fig. 4D). In addition, there is significant overlap between this set of genes and ones associated with renal age or pathology (Fig. 4E). Among the most highly predictive set of clones are 43 known genes with expression differences of two-fold or greater, as reported in Table 2. In addition to scoring all kidney samples for interstitial fibrosis (%IF), quantitation of cortical fractional interstitial fibrosis volume (Vint) was performed with computerized image analysis on coded sections stained using Sirius red [39] and correlation to expression measurements tested. As previously observed, the two scoring methods for measuring fibrotic changes in the kidney positively correlate [40], although not strongly (R = 0.37; data not shown). The expression of none of the genes analyzed statistically correlate with Vint scores (minimum FDR 23%) although several collagens (COL4A4, 4A2, 5A2 and 3A1) and FN1 are among the top 200 ranked transcripts. These genes were also found to be age related in this study and are known to specifically stain with the Sirius red dye. Further, the top ranked genes also segregate AO and OO samples by cluster analysis (data not shown). Real-time PCR analysis of selected genes Expression differences of six age-related candidate genes were confirmed by real-time PCR analysis. Of

−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−→ Fig. 4. Hierarchic clustering of samples based on genes that correlate with renal histopathology scores. (A) Quantitative response analysis using SAM identified 163 genes whose expression levels correlate with IF, 38 genes with GS, 8 genes with TA, and 83 with FIT (false discovery rate threshold of q<12.5%). Combined this renal-pathology set of 255 genes were used to group kidney samples using complete hierarchic clustering of both genes and arrays. The dendrogram indicates that 4 of the old samples (O1, O6, O8, O9) cluster together with the adult samples. (B) Hierarchic clustering of the old and adult samples across the renal-aging gene set (N = 505 genes) revealed that the same old samples cocluster with the adults. (C) Hierarchic clustering of 16 nonpediatric kidney samples on the basis of 67 cDNA transcripts identified by PAM comparing adults with five kidneys from old donors (O2, O3, O4, O5, and O7) in the test set and the remaining other old kidney samples (O1, O6, O8, and O9) as the learning set. (D) The four OA samples in the learning set all phenotype as similar to adults and cocluster with the adult kidney samples. (E) A total of 52 genes of known function are represented in the PAM classification analysis and there is substantial overlap between these genes with both the human renal-aging and renal-pathology associated genes. The most significant percentage overlap in this analysis is highlighted. All expression data in these clusters are available on the web supplement at http://microarray-pubs.stanford.edu/renalaging/.

2675

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

B

O4 O2 O7 O5 O3 O8 A5 A1 A3 A2 A4 A6 O1 O9 A7 O6

A7 O7 O5 O3 O4 O2 O9 O1 O8 A6 A4 A2 A3 A1 A5 O6

A

D

O4 O2 O7 O3 O5 A7 A6 A5 A1 A4 A2 A3 O1 O8 O9 O6

C

100%

50%

0% - BPHL

>1.8

>.9

- CSPG2

<.9

<1.8

Old

- FN1 - TNC

Adult

E

Fig. 4.

1:1

# in each category

% of 52 PAM genes

% Enrichment in SAM lists (N)

45 in Age

86.5%

8.9% (505)

26 in Path

50.0%

10.2% (255)

0 in TA

0%

0% (8)

1 in FIT

1.9%

1.2% (81)

9 in Gs

17.3%

23.7% (38)

24 in IF

46.2%

14.5% (165)

2676

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

8

A

2 Adult

P=0.072

Old

Fold difference

6

P=0.001

4

P=0.014

1

P=0.009 P=0.021

2 P=0.021 0 Fold old vs. adult: Gene array Taqman PCR

SERP

TNC

CSPG2

FN1

MMP7

-2.2 0.7

2.2 4.8

2.4 3.7

2.9 3.9

2.0 1.8

2.1 4.7

B Adult Old

P=0.826 6 Fold difference

C 1

2

P=0.005 P=0.001

4

1

P=0.003 P=0.002 2

0.5

0.25 P=0.315

0

0 SERP

P=0.006

0.75

P=0.109 BPHL

0

P=0.029 Fold difference

8

BPHL

TNC CSPG2 FN1

MMP7

0

Adult Old Old all Old N = 7 with adult N = 9 separate N=4 N=5

Fig. 5. Taqman-based analysis of gene expression in adult and old kidneys. Six genes, serine proteinase inhibitor (SERPIN3A), biphenyl hydrolaselike (BPHL), tenascin (TNC), chondroitin sulfate proteoglycan 2 (CSPG2), fibronectin (FN1), and matrix metalloproteinase 7 (MMP7), were chosen based on high significance scoring by SAM with renal aging. Four genes were also found among the renal-pathology associated genes (BPHL, TNC, CSPG2, FN1). (A) Taqman PCR results confirmed increased or decreased expression of all the genes in the 16 nonpediatric samples tested in the gene array analysis. The magnitude of the mean expression differences are indicated as fold differences for either gene array or Taqman PCR and are generally higher for the Taqman PCR results. The differences between adult and old samples were significant for all and approached significance for BPHL. (B) A separate test group of 17 kidneys (3 adult and 14 old kidneys) was used to validate differential expression of these genes with age. Increased significance in the differential expression by age was observed for all genes, except SERPIN3A and BPHL. (C) Expression of the proteoglycan CSPG2, whose array-based expression data was ranked with highest significance both by PAM and by SAM, was not only significantly different between adult and old kidneys, but differences in expression within two identified groups of old samples (OO vs. OA) is also confirmed.

these, four are also found in the SAM analysis for renal pathology. Three are related to fibrosis and growth response: FN1, TNC, and MMP7. MMP7 has also been implicated in tissue remodeling and wound healing and is know to regulate the activity of various defensin genes [41]. CSPG2 is both a component basement membrane [42] and was shown to be a direct target of p53 [43]. The SERPIN3A, which may be a causative of atherothrombotic infarction [44], was examined for decreased expression in the older age groups. Similarly, biphenyl hydrolase-like (BPHL), primarily expressed in liver and

kidney, has a potential enzymatic role in detoxification processes, and was proposed to be reduced with age. Collectively, differences in the expression of these genes could be envisioned to play a key role in renal aging. There was good correlation between array and realtime PCR for all of these genes (SERPIN3A R = 0.71, BPHL R = 0.83, TNC R = 0.83, CSPG2 R = 0.67, FN1 R = 0.63, MMP7 R = 0.79) (Fig. 5A). The magnitude of the mean differences between old and adult were generally higher for the Taqman PCR results, consistent with previous array-based expression studies [45, 46]. The

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

comparison between old and adult samples were significant for all genes and approached significance for BPHL; P = 0.072. A separate test group of 17 kidneys (3 adult and 14 old kidneys) was used to validate the expression of these genes in age (Fig. 5B). Increased significance in the differential expression by age was observed for all genes, except for SERPIN3A and BPHL, for which significance was no longer reached. Among the genes tested using quantitative PCR, only CSPG2 differentiated OO and OA kidneys. Expression of the proteoglycan CSPG2 is, thus, higher in old kidneys than adult kidneys, and differentiated the OO and OA kidneys (P = 0.029; Fig. 5C).

DISCUSSION We studied the gene expression profiles of kidney from individuals of different ages and observed expression signatures in old kidney consistent with the known changes of aging, such as fibrosis, cellular aging, and low-grade focal inflammation. The major change of human renal aging—loss of mass—would of course produce no transcripts in its established state; there are no transcripts in nephrons that are not there. The change would be detectable as functional units in the process of being lost and as interstitium remodeling with age. The results are compatible with the concept that not only is there loss of mass, but the existing tissue is altered, perhaps contributing to reduced capacity of the aged nephron to withstand stress. These expression changes, however, are subtle and only account for only 2% of the initial gene set of 13,930 genes and less than 18% of the differentially expressed genes analyzed, which explains why the samples do not cluster by age across these gene sets. Another reason for the small overall gene expression differences seen between the samples may result from the rather mild chronicity scores in all the old kidneys (which are ∼1/3 lower than renal chronicity scores, with age that were reported in some other studies [30–34]). Nevertheless, appreciable transcript differences are observed when adult kidneys are compared to high-functioning old kidneys. In young adults, renal architecture and function are at their peak, while chronic tubular and interstitial changes are known to accrue with age, resulting in loss of glomerular and single nephron filtration after 40 years of age [47]. Selection of gene expression specific for age by supervised analysis results in clustering of all adults together, with exception of the oldest adult in this group (aged 46), who clusters with old samples. In contrast, the old kidneys separate into two distinct groups: old kidneys that segregate together with adults (N = 4, termed OA in this study) or away from the adults (N = 5, termed OO in this study). Class prediction analysis to evaluate which genes that drive these differences narrows the differential genes to a list of 50 known unique genes, of which

2677

86% overlap with the renal-aging genes and 50% overlap with the renal-pathology genes. This confirms that age-related pathology correlates with age-related transcriptome changes, but also indicates that the age-related transcriptome changes go beyond what the pathology can assess. A significant enrichment of genes associated with the number of sclerotic glomeruli (24%) and the amount of interstitial fibrosis (14%) is observed, while fibrous intimal thickening and tubular atrophy play little or no role in this segregation, as evidenced by insignificant enrichment (0% and 1.2%, respectively). CSPG2 is expressed at highest levels in the OO kidneys in this study and expression of this gene alone segregates the OO and OA old kidneys. It is possible that the signature of tubular atrophy may be faint because these tubules may be transcriptionally inactive, or that the signature of arteries may be too infrequent to contribute the overall kidney signal. Aging kidney signatures, as expected in a crosssectional study, were not linearly related to age, gender, or individual histopathologic parameters. Old kidneys show reduced tubular transport, active cellular processes including cell cycling, energy metabolism, and recovery from oxidative stress and tissue injury. The altered balance between expression of growth factor and extracellular matrix components in the old kidneys likely results in the incremental tissue chronicity with age. In this context, reduced expression of key genes in energy pathways, ion transport, and response to oxidative stress in renal epithelial cells could contribute to reduced capacity of the aged nephron to withstand stress [1]. In contrast, the increased expression of genes involved in extracellular matrix restructuring and the inflammatory response are more likely derived from either resident fibroblasts or from infiltrating immune cells, respectively. A significant loss of tubular mass is seen with age and it is likely that cumulative environmental stress (e.g., oxidative damage) associated with the normal process of aging may evoke an inflammatory response linked to nephron loss and matrix remodeling. Aged kidneys develop mild mononuclear cell infiltrates similar to those in any kidney with low-grade fibrosis and atrophy and, thus, the inflammatory genes are expected. Thus, the increased expression of immune-specific genes (especially chemokines and immunoglobulin genes) is consistent with the increased infiltration of B cells in old kidneys in general and in this study. Chronic organ injury has been shown to have a focal increase in B cell content. The basis for the link between nephron loss with fibrosis and atrophy and mononuclear infiltration is not known, although they are probably a consequence rather than a cause of the fibrosis and atrophy. A paucity of cell turnover and metabolism is seen in the old kidneys in this study, consistent with the normal functioning of the surviving nephrons in healthy old kidneys [6, 9]. An increase of apoptosis-related genes

2678

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

was observed, although apoptosis-related genes account for only 2% of the genes associated with renal aging. CSPG2, a p53 regulated gene, is highly differentially expressed with renal aging and its expression is highest in the old kidneys. CSPG2 encodes versican, which is induced in a p53-dependant manner in response to specific genotoxic stress [43]. p53 is expressed ubiquitously in all cell types, including fibroblasts, and becomes active when cells are subjected to a variety of cellular insults, such as DNA damage, leading to the arrest of cell proliferation or apoptosis. Recently, we have shown that p53 is not transcriptionally regulated in renal aging [13], although its expression is increased in senescent fibroblasts. CSPG2 is a component of the basement membrane in tubular cells [42], and is also expressed in senescent fibroblasts ([37]; http://source.stanford.edu). This gene is also secreted into the extracellular matrix, where it is present in all 3 layers of veins and elastic arteries, with increased expression in atherosclerotically transformed vessel walls. It is noteworthy that CSPG2 is expressed in highest levels in the OO old kidneys in this study, which, although marginally higher in standard pathology scoring, cannot otherwise be differentiated on standard pathology scoring. This observation suggests that CSPG2 may play a role in increased tubular basement membrane matrix deposition, leading to tubular basement membrane thickening, interstitial fibrosis, and intimal thickening. A number of genes known to be associated with age or senescence in vitro in other studies were confirmed by the current study. Genes expressed in senescent fibroblasts (HAK, Thy1, transgelin) are significantly lower in young kidney samples. Another area of active research in aging is the influence of the insulin/insulin-like growth factor pathway signaling. In Caenorhabditis elegans, mutations in a number of genes in this pathway result in dramatic increases or decreases of lifespan [48, 49], and altered expression of several genes have been observed as a result these genetic changes [50]. Although we did not observe a difference in the expression of insulin/insulin-like growth factor or its receptor, we did find that genes known to be up-regulated in the long-lived daf-2 mutant also show altered expression in renal aging (cytochrome P450 family, aquaporin, glyceraldehyde-3-phosphate dehydrogenase, fibroblast growth factor, and genes involved in tertahydrofolate metabolism and UDP glycosylation). The present results are compatible with aging being not just loss of mass but alterations in the remaining tissue, which could be useful in defining genes whose expression could predict the performance of old kidneys after renal transplantation. Clinicians need molecular markers that will accurately predict whether a deceased donor kidney from an old donor has a high probability of recovering function in a new host following the stresses of brain death, preservation, implantation, etc. New validated transcriptosome-based criteria for predicting kidney per-

formance and functional reserve would permit more kidneys from older donors to be transplanted, expanding organ donor pools. Currently, kidneys with reasonable function in the old age group are considered acceptable for transplantation only on the basis of criteria such as glomerulosclerosis on biopsy that in fact have poor predictive value. We need to add predictive power, so that the problems of old kidneys can be estimated more accurately (e.g., the rate of primary nonfunction, delayed graft function, and graft survival) [51]. Improved methods of monitoring for “tissue senescence” are required, which results from a combination of age and/or environmental stress-dependent modification of cellular function. It is notable that the oldest kidney in this study belonged to an 88-year-old, but it coclustered with adult kidneys in all supervised analysis. We believe that the degree of tissue senescence, which may be more accurately predicted by molecular markers, such as CSPG2, identified in this study, will likely be inversely correlated with subsequent performance of the kidney after stress, such as transplantation. Testing this hypothesis will require prospective screening of kidney donors and correlating the results with immediate and late outcomes, with subsequent adaptations of technology to make the promising information available to clinicians in a cost-effective and timely manner. In addition, extending the patient population to a non-Caucasian cohort is merited. ACKNOWLEDGMENTS We are indebted to those who helped make the kidney samples available: Dr. Ronald Moore, Dr. Gerald Todd, and the team of technicians at the Department of Pathology, University of Alberta. We thank Dr. Mei-Sze Chua for her contribution analyzing the initial set of arrays in this study, and Dr. Orly for assistance with the singular value decomposition analysis. Assistance with statistical analysis of the Taqman real-time PCR data by Dr. Frank Hsieh (Veterans Administration Hospital, Palo Alto, CA) is also gratefully acknowledged. This study is funded by grants from Roche Organ Transplant Research Foundation, Canadian Institutes of Health Research, Kidney Foundation of Canada, The Muttart Foundation, The Royal Canadian Legion, NIH (NIH5P3-05 and NIH3P3-05S1), Clinical Center for Immunological Studies at Stanford University (CCIS), Packard Foundation, and Roche Pharmaceuticals. Reprint requests to Anette Melk, M.D., Ph.D., Children’s Hospital, University of Heidelberg, Im Neuenheimer Feld 150, D-69120 Heidelberg, Germany. E-mail: [email protected]

REFERENCES 1. BARALDI A, BALLESTRI M, RAPANA R, et al: Acute renal failure of medical type in an elderly population. Nephrol Dial Transplant 13(Suppl 7):25–29, 1998 2. TERASAKI PI, CECKA JM, GJERTSON DW: Impact analysis: A method for evaluating the impact of factors in clinical renal transplantation, in Clinical Transplants, edited by Ceck JM and Terasaki PI, Los Angeles, UCLA Tissue Typing Laboratory, 1998, pp 437–441 3. LINDEMAN RD, TOBIN J, SHOCK NW: Longitudinal studies on the rate of decline in renal function with age. J Am Geriatr Soc 33:278–285, 1985

Melk and Mansfield et al: Transcriptional analysis using cDNA microarray profiling

4. GOURTSOYIANNIS N, PRASSOPOULOS P, CAVOURAS D, PANTELIDIS N: The thickness of the renal parenchyma decreases with age: A CT study of 360 patients. AJR Am J Roentgenol 155:541–544, 1990 5. LINDEMAN RD, GOLDMAN R: Anatomic and physiologic age changes in the kidney. Exp Gerontol 21:379–406, 1986 6. FLISER D, ZEIER M, NOWACK R, RITZ E: Renal functional reserve in healthy elderly subjects. J Am Soc Nephrol 3:1371–1377, 1993 7. COCKCROFT DW, GAULT MH: Prediction of creatinine clearance from serum creatinine. Nephron 16:31–41, 1976 8. LEVEY AS, BOSCH JP, LEWIS JB, et al: A more accurate method to estimate glomerular filtration rate from serum creatinine: A new prediction equation. Modification of Diet in Renal Disease Study Group. Ann Intern Med 130:461–470, 1999 9. FLISER D, FRANEK E, JOEST M, et al: Renal function in the elderly: Impact of hypertension and cardiac function. Kidney Int 51:1196– 1204, 1997 10. PANNU N, HALLORAN P: The kidney in aging, in Primer on Kidney Diseases (vol 3), edited by Greenberg A, San Diego, Academic Press, 2001, pp 377–381 11. MELK A, KITTIKOWIT W, SANDHU I, et al: Cell senescence in rat kidneys in vivo increases with growth and age despite lack of telomere shortening. Kidney Int 63:2134–2143, 2003 12. MELK A, RAMASSAR V, HELMS LM, et al: Telomere shortening in kidneys with age. J Am Soc Nephrol 11:444–453, 2000 13. MELK A, SCHMIDT BM, TAKEUCHI O, et al: Expression of p16INK4a and other cell cycle regulator and senescence associated genes in aging human kidney. Kidney Int 65:510–520, 2004 14. MELK A, HALLORAN PF: Cell senescence and its implications for nephrology. J Am Soc Nephrol 12:385–393, 2001 15. HALLORAN PF, MELK A, BARTH C: Rethinking chronic allograft nephropathy: The concept of accelerated senescence. J Am Soc Nephrol 10:167–181, 1999 16. RACUSEN LC, SOLEZ K, COLVIN RB, et al: The Banff 97 working classification of renal allograft pathology. Kidney Int 55:713–723, 1999 17. CHIRGWIN JM, PRZYBYLA AE, MACDONALD RJ, RUTTER WJ: Isolation of biologically active ribonucleic acid from sources enriched in ribonuclease. Biochemistry 18:5294–5299, 1979 18. PEROU CM, SORLIE T, EISEN MB, et al: Molecular portraits of human breast tumours. Nature 406:747–752, 2000 19. SCHENA M, SHALON D, DAVIS RW, BROWN PO: Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science 270:467–470, 1995 20. SARWAL M, CHUA MS, KAMBHAM N, et al: Molecular heterogeneity in acute renal allograft rejection identified by DNA microarray profiling. N Engl J Med 349:125–138, 2003 21. GOLLUB J, BALL CA, BINKLEY G, et al: The Stanford Microarray Database: Data access and quality assessment tools. Nucleic Acids Res 31:94–96, 2003 22. SHERLOCK G, HERNANDEZ-BOUSSARD T, KASARSKIS A, et al: The Stanford Microarray Database. Nucleic Acids Res 29:152–155, 2001 23. TUSHER VG, TIBSHIRANI R, CHU G: Significance analysis of microarrays applied to the ionizing radiation response. Proc Natl Acad Sci U S A 98:5116–5121, 2001 24. TIBSHIRANI R, HASTIE T, NARASIMHAN B, CHU G: Diagnosis of multiple cancer types by shrunken centroids of gene expression. Proc Natl Acad Sci U S A 99:6567–6572, 2002 25. ALTER O, BROWN PO, BOTSTEIN D: Singular value decomposition for genome-wide expression data processing and modeling. Proc Natl Acad Sci U S A 97:10101–10106, 2000 26. EISEN MB, SPELLMAN PT, BROWN PO, BOTSTEIN D: Cluster analysis and display of genome-wide expression patterns. Proc Natl Acad Sci U S A 95:14863–14868, 1998 27. GIBSON UE, HEID CA, WILLIAMS PM: A novel method for real time quantitative RT-PCR. Genome Res 6:995–1001, 1996 28. HEID CA, STEVENS J, LIVAK KJ, WILLIAMS PM: Real time quantitative PCR. Genome Res 6:986–994, 1996 29. DENNIS G, JR., SHERMAN BT, HOSACK DA, et al: DAVID: Database for Annotation, Visualization, and Integrated Discovery. Genome Biol 4:P3, 2003 30. SMITH SM, HOY WE, COBB L: Low incidence of glomerulosclerosis in normal kidneys. Arch Pathol Lab Med 113:1253–1255, 1989

2679

31. KAPLAN C, PASTERNACK B, SHAH H, GALLO G: Age-related incidence of sclerotic glomeruli in human kidneys. Am J Pathol 80:227–234, 1975 32. KAPPEL B, OLSEN S: Cortical interstitial tissue and sclerosed glomeruli in the normal human kidney, related to age and sex. A quantitative study. Virchows Arch A Pathol Anat Histol 387:271– 277, 1980 33. TRACY RE, VELEZ-DURAN M, HEIGLE T, OALMANN MC: Two variants of nephrosclerosis separately related to age and blood pressure. Am J Pathol 131:270–282, 1988 34. TRACY RE, BERENSON GS, CUETO-GARCIA L, et al: Nephrosclerosis and aortic atherosclerosis from age 6 to 70 years in the United States and Mexico. Virchows Arch A Pathol Anat Histopathol 420:479–488, 1992 35. SUH JH, SHENVI SV, DIXON BM, L et al: Decline in transcriptional activity of Nrf2 causes age-related loss of glutathione synthesis, which is reversible with lipoic acid. Proc Natl Acad Sci U S A 101:3381– 3386, 2004 36. LIU R, CHOI J: Age-associated decline in gamma-glutamylcysteine synthetase gene expression in rats. Free Radic Biol Med 28:566–574, 2000 37. DIEHN M, SHERLOCK G, BINKLEY G, et al: SOURCE: A unified genomic resource of functional annotations, ontologies, and gene expression data. Nucleic Acids Res 31:219–223, 2003 38. BOKEMEYER D, PANEK D, KRAMER HJ, et al: In vivo identification of the mitogen-activated protein kinase cascade as a central pathogenic pathway in experimental mesangioproliferative glomerulonephritis. J Am Soc Nephrol 13:1473–1480, 2002 39. GRIMM PC, NICKERSON P, GOUGH J, et al: Computerized image analysis of Sirius Red-stained renal allograft biopsies as a surrogate marker to predict long-term allograft function. J Am Soc Nephrol 14:1662–1668, 2003 40. SUND S, GRIMM P, REISAETER AV, HOVIG T: Computerized image analysis vs semiquantitative scoring in evaluation of kidney allograft fibrosis and prognosis. Nephrol Dial Transplant 19:2838–2845, 2004 41. ZUO F, KAMINSKI N, EUGUI E, et al: Gene expression analysis reveals matrilysin as a key regulator of pulmonary fibrosis in mice and humans. Proc Natl Acad Sci U S A 99:6292–6297, 2002 42. ERICKSON AC, COUCHMAN JR: Basement membrane and interstitial proteoglycans produced by MDCK cells correspond to those expressed in the kidney cortex. Matrix Biol 19:769–778, 2001 43. YOON H, LIYANARACHCHI S, WRIGHT FA, et al: Gene expression profiling of isogenic cells with different TP53 gene dosage reveals numerous genes that are affected by TP53 dosage and identifies CSPG2 as a direct target of p53. Proc Natl Acad Sci U S A 99:15632– 15637, 2002 44. TACHIKAWA H, TSUDA M, ONOE K, et al: alpha-1–Antichymotrypsin gene A1252G variant (ACT Isehara-1) is associated with a lacunar type of ischemic cerebrovascular disease. J Hum Genet 46:45–47, 2001 45. ERICKSON LM, PAN F, EBBS A, et al: Microarray-based gene expression profiles of allograft rejection and immunosuppression in the rat heart transplantation model. Transplantation 76:582–588, 2003 46. MANDEL M, GUREVICH M, PAUZNER R, et al: Autoimmunity gene expression portrait: Specific signature that intersects or differentiates between multiple sclerosis and systemic lupus erythematosus. Clin Exp Immunol 138:164–170, 2004 47. HOANG K, TAN JC, DERBY G, et al: Determinants of glomerular hypofiltration in aging humans. Kidney Int 64:1417–1424, 2003 48. LEE SS, KENNEDY S, TOLONEN AC, RUVKUN G: DAF-16 target genes that control C. elegans life-span and metabolism. Science 300:644– 647, 2003 49. WALKER GA, LITHGOW GJ: Lifespan extension in C. elegans by a molecular chaperone dependent upon insulin-like signals. Aging Cell 2:131–139, 2003 50. MURPHY CT, MCCARROLL SA, BARGMANN CI, et al: Genes that act downstream of DAF-16 to influence the lifespan of Caenorhabditis elegans. Nature 424:277–283, 2003 51. BERARDINELLI L, BERETTA C, RAITERI M, CARINI M: Early and longterm results using older kidneys from cadaver or living donors. Clin Transpl:157–166, 2001