BRIEF REPORT
Epigenetics 6:1, 20-28; January 2011; © 2011 Landes Bioscience
DNA methylation profiling identifies epigenetic differences between diabetes patients with ESRD and diabetes patients without nephropathy Carmen Sapienza,1,2,* Jean Lee,3 Jasmine Powell,1 Oluwatoyin Erinle,1 Faahud Yafai,3 James Reichert,3 Elias S. Siraj3 and Michael Madaio4 Fels Institute for Cancer Research and Molecular Biology; Departments of 2Pathology and Laboratory Medicine; 3Medicine; Temple University Medical School; Philadelphia, PA USA; 4Department of Medicine; Medical College of Georgia; Augusta, GA USA
1
Key words: DNA methylation, end-stage renal disease, diabetic nephropathy, dialysis
We identified potential epigenetic biomarkers for chronic kidney disease progression by comparing site-specific DNA methylation levels in more than 14,000 genes between African American and Hispanic diabetes patients with end stage renal disease (ESRD) and diabetes patients without nephropathy. We identified 187 genes that are differentially methylated between the two groups on at least two CpG sites in each gene in DNA extracted from saliva. Of the 187 genes whose mean methylation levels differed between the two groups, 39 genes or closely related gene family members, have been reported to be involved in kidney development or diabetic nephropathy, per se, or have been associated with dialysis-induced changes in gene expression in peripheral blood cells. The fact that such a substantial fraction (21%) of the 187 candidate genes have been implicated previously through genome association or transcription profiling studies suggests strongly that the DNA methylation differences we observe are associated with disease predisposition and/ or treatment. The fact that these nephropathy and/or dialysis-associated differences between patients were identified in DNA extracted from saliva offers proof-of-principle that inter-individual epigenetic differences may prove useful as predictive biomarkers of disease susceptibility.
©201 1L andesBi os c i enc e. Donotdi s t r i but e. Introduction
Personalized medicine has been defined as “the application of genomic and molecular data to better target the delivery of health care, facilitate the discovery and clinical testing of new products, and help determine a person’s predisposition to a particular disease or condition.” 1 While there have been some notable successes in using genetic data to tailor treatment of disease, particularly in the area of cancer genetics, the utility of genetic information to stratify individuals according to disease predisposition, especially predisposition to common disease, has been limited. The limited utility of genetic information for predicting an individual’s risk for type 2 diabetes, for example, stems from the fact that multiple loci are involved,2,3 the relative risk associated with carrying any particular disease-predisposing allele is small3 and the influence of environmental factors is large.2,3 In fact, it has been argued that for many common diseases, precise genotype information adds no greater predictive power than information on family history alone. If the goals of personalized medicine are to be advanced, scientists and physicians must find ways to add significant predictive value to genotype information. Molecular profiling technologies,
such as proteomic profiling, metabolomic profiling and epigenetic profiling, offer the most likely options for obtaining additional relevant biomarker information. One difficulty with the use of such technologies is that, unlike genotype, which does not vary between the tissues of an individual, an individual’s proteomic, metabolomic or epigenetic profiles may differ between different tissues. The identification of biomarkers that have predictive value in disease predisposition requires that the biomarkers give information about the likelihood of pathology in a tissue that may not be available for analysis by assaying a tissue that can be accessed in a minimally invasive way. Biomarkers that predict the risk of dementia or kidney failure, for example, will only be useful clinically if they can be assayed by non-invasive techniques or in tissues such as blood or saliva, rather than brain and kidney. Diabetic nephropathy is a common complication of type 1 and type 2 diabetes and familial aggregation in multiple populations suggests that susceptibility is, in part, genetic.4-6 Linkage studies and association studies have identified candidate chromosomal regions and candidate genes that explain a small fraction of the nephropathy trait variance, but genes with major effect have not been identified.4-6 In this study, we assessed whether quantitative differences in CpG site-specific methylation in DNA
*Correspondence to: Carmen Sapienza; Email:
[email protected] Submitted: 05/07/10; Accepted: 08/17/10 Previously published online: www.landesbioscience.com/journals/epigenetics/article/13362 DOI: 10.4161/epi.6.1.13362 20
Epigenetics
Volume 6 Issue 1
BRIEF REPORT
BRIEF REPORT
Table 1. Characteristics of diabetes patients without nephropathy [>10 years of diabetes, estimated glomerular filtration rate >60 (by MDRDS formula) and urine microalbumin 10 years, Table 1; henceforth called “patients without nephropathy”) and 24 diabetes patients with ESRD who were being treated by hemodialysis (Table 2; henceforth called “ESRD patients on HD”) using DNA extracted from saliva of each patient. We used the Illumina HumanMethylation27 BeadChip array to assay the fraction of molecules in each sample that was methylated at each of 27,578 CpG sites—the “b-value,” which may range from 0–1 (www.illumina.com/documents/icommunity/article_2008_05_infinium_methylation.pdf). One African American ESRD on HD patient’s array results were of insufficient quality (as judged by the fraction of array CpGs detected, see Methods), and this sample and the sample of one African
American patient without nephropathy (a case-matched female with the lowest fraction of CpGs detected among the patients without nephropathy) were excluded from the analysis, leaving 23 individuals (21 African Americans and two Hispanics) in each group. Patients with diabetic nephropathy do not cluster separately by epigenotype but ~10% of CpG sites are differentially methylated. Unsupervised hierarchical cluster analysis using b-values for all CpGs did not cluster ESRD patients on HD separately from diabetic patients without nephropathy (Sup. Fig. 1). However, the clustering of most females separately from most males (Sup. Fig. 1) argues for robust detection of interindividual methylation differences between males and females, most likely at genes on the X chromosome (see Methods). When the 1,092 CpGs present on the X and Y chromosomes are removed from the analysis (Fig. 1), the ESRD on HD patients cluster somewhat better, in that eight of the 11 patients on the deepest branches (i.e., greatest differences from the rest of the group) are from the ESRD on HD group. However, as there is significant intermingling of patient groups within branches, we compared mean beta-values at each CpG between the 23 ESRD patients on HD and the 23 patients without nephropathy. A significant difference (a “DiffScore” of +20; see Methods)
www.landesbioscience.com Epigenetics
21
between mean beta-values in each group was detected for 2,870 CpGs (Sup. Table 1). Remarkably, 2,693 of these CpGs exhibited lower mean methylation levels in the ESRD on HD patients while only 177 CpGs were methylated at higher mean values in the dialysis patients. Epigenetic differences identify candidate genes related to candidate genes identified on the basis of genetic differences. Many of the CpG sites exhibiting significant methylation differences between the groups are found in genes that have been associated with diabetic nephropathy in more than one study (i.e., TIMP3, BCL2; Sup. Table 1). Because we have assayed methylation differences in only a small number of patients, we applied the additional more robust selection criterion that candidate genes must have significant differences in methylation at two or more CpG sites within the same gene in order to reduce the number of false-positive candidates. When this selection was applied, 389 CpGs in 187 genes [n.b.: using the “2 significant CpGs in the same gene” selection criterion and a significant “DiffScore” of +20 (see Methods for relationship to p value), we expect only five to six false positives in a scan of this size] were methylated significantly differently between the two groups (Table 3). Of the 187 genes in Table 3, 156 (83%) are assayed at only two CpGs on the array and both differ significantly between groups. Twenty-three of the remaining 32 genes have more than three CpGs assayed on the array but only two are significantly different. The remaining nine genes all have more than three CpGs that differ significantly. Twenty-one of the genes in Table 3 or related gene family members (Table 4), are among the 115 genes tested for linkage or association with diabetic nephropathy by Ewens et al.6 Although our analysis is not based on known differences in genotype between the two groups, it is interesting that a significant fraction of genes or like-gene family members, appear to exhibit epigenetic differences between the two groups (Table 4). This observation suggests that disease progression may be influenced by genetic and/or epigenetic differences in genes that operate in closely related or parallel pathways. Among the most interesting candidate genes (Table 4) are myosin light chain 9 (MYL9), matrix metalloprotease 10 (MMP10), tissue inhibitor of metalloprotease 4 (TIMP4) and methylenetetrahydrofolate reductase (MTHFR). Genetic variants in myosin heavy chain 9 (MYH9) have been reported to explain a significant fraction of focal glomerulosclerosis in African Americans,7 who also make up the majority of the patient population examined in this study. It is interesting that genetic variation in MYH9 is associated with nephropathy caused by hypertension8 but does not appear to be associated with diabetic nephropathy.9 Our identification of an association between epigenetic variation in a regulatory MYL9 and diabetic nephropathy is highly suggestive that similar pathways may operate in both diabetic and hypertensive nephropathy but that the defects involve different parts of the pathway. MMP10 is part of a cluster of nine matrix metalloprotease genes spread over approximately 500 kb on chromosome 11q (www.ensembl.org/Homo_sapiens/Location/View?g=ENSG
Table 2. Characteristics of ESRD patients on HD Sex
Age
Ethnicity
Type of diabetes
female
55
African American
2
female
57
African American
2
female
59
African American
2
female
63
African American
2
female
67
African American
2
female
71
African American
2
female
72
African American
2
female
72
African American
2
female
79
African American
2
female
80
African American
2
female
87
African American
2
female
38
African American
2
female
61
African American
2
male
41
African American
2
male
50
African American
2
male
54
African American
2
male
58
Hispanic
2
male
61
African American
2
male
61
Hispanic
2
male
63
African American
2
male
64
African American
2
male
66
African American
2
male
70
African American
2
male
74
African American
2
©201 1L andesBi os c i enc e. Donotdi s t r i but e.
22
00000166670). Another member of this group, MMP20, has been associated with human kidney aging and age-related decline in glomerular filtration rate.10 Regulation of matrix metalloproteases is likely to be an important factor in response to renal injury (reviewed in ref. 11) and ESRD patients on HD have significant kidney fibrosis. In this regard, the identification of epigenetic variation in both a matrix metalloprotease gene and a matrix metalloprotease inhibitor gene (TIMP4) may be significant. We also believe it is significant that MTHFR is identified as a differentially methylated gene in ESRD patients on HD. MTHFR is a central node in the one-carbon metabolism pathway,12,13 responsible for generating methyl groups for all cellular processes, including DNA methylation. Stenvinkel et al.14 have found that overall levels of DNA methylation are increased in patients with chronic kidney disease and inflammation (a secondary consequence of dialysis) and five of the six CpGs assayed in MTHFR are significantly less methylated in dialysis patients (Sup. Table 1), suggesting upregulation of this gene. The degree to which the ESRD patients on HD and the patients without nephropathy differ in methylation at two sites within each gene is illustrated for four representative genes in Figure 2; ESRD patients on HD have lower mean methylation in three of the genes (MMP10, MTFR and TIMP4; Figs. 2A–C, respectively) and higher mean methylation in one of the genes (CLIC6; Fig. 2D).
Epigenetics
Volume 6 Issue 1
©201 1L andesBi os c i enc e. Donotdi s t r i but e.
Figure 1. Cluster diagram using Pearson correlation and the 26,486 CpG probes that are not located on the X and Y chromosomes. All patients are African American except M ESRD on HD 19, M ESRD on HD 16, M PWON 12 and M PWON 3, who are Hispanic. All patients have type 2 diabetes except F PWON 19, F PWON 15 and M PWON 13, who have Type 1 diabetes. M ESRD on HD, male end stage renal disease on dialysis patient; F ESRD on HD, female end stage renal disease on dialysis patient; M PWON, male patient without nephropathy; F PWON, female patient without nephropathy.
Differentially methylated candidate genes are involved in pathways important in diabetic nephropathy and dialysis treatment. We also used the 187 ESRD/HD-associated candidate genes as the input core data for Ingenuity Pathway Analysis (www.ingenuity.com/) to determine whether specific metabolic pathways might be associated with the methylation differences in dialysis patients. Overall, this analysis suggests that dialysis patients have alterations in pathways involved in inflammation, oxidative stress, ubiquitination, fibrosis and drug metabolism (Table 5). These gene-specific methylation differences also
appear to occur non-randomly in genes involved in cancer (and we note that dialysis patients have a higher incidence of renal cell carcinoma; reviewed in ref. 15), as well as genes associated with cardiotoxicity, hepatotoxicity and nephrotoxicity (Table 5). Wilflingseder et al.16 have performed transcriptional profiling on peripheral blood from dialysis patients and found that more than 200 genes were upregulated in dialysis patients. Of the 187 candidate genes identified as differentially methylated in ESRD on HD patients (Table 3), 18 genes or closely related family members (Table 6) are among the genes identified as upregulated
www.landesbioscience.com Epigenetics
23
Table 3. Genes in which two or more CpGs are differentially methylated between ESRD patients on HD and diabetes patients without nephropathy ACSL6
APC
DLX3
JUN
ADAMTS10
CDKN1C
DNAJC18
AKR1B1
CFL1
DNM2
SLC2A4
KIAA0922
PCSK4
SLC4A3
KIAA1199
PDLIM3
SOAT2 SOCS1
ALDH3B2
CHPF
DTL
KIAA1838
PFKFB4
ANKRD15
CHRNB3
EDA2R
KIF1B
PFTK1
SOLH
ANXA13
CLIC3
EIF4B
KLK10
PHF2
SOSTDC1
ARHGAP22
CLIC6
EMG1
LAPTM4A
PHLDA2
SPINLW1
ARL4
CNAP1
EMX1
LGALS3BP
PHYH
SPINT2
ASPA
CRLF1
EPHA3
LMO4
PIP5KL1
SPRR2A
ATP10A
CSNK1E
FAM105A
LOC284837
PKNOX1
SPRY4
ATP5I
CTBP1
FBXO32
LRRC3
PLCL1
SRF
BAIAP2
CTSZ
FBXL16
LYST
POLE
STK6
BIK
CYBASC3
FEN1
MAF1
POLR2F
STMN1
BOC
CYLN2
FGFRL1
MAP3K3
POLR3D
SURF5
BPHL
CYP2S1
FHIT
MBNL3
POMGNT1
TCF15
BRD3
CYP8B1
FLJ12118
MCAM
PSME2
TFAP2E
BTNL2
DAAM1
FLJ21103
MEST
PTGER3
TIMP4 TMEM79
C1orf142
DAB2IP
FLJ32569
MKNK2
RAP1B
C1orf149
CMTM2
FZD2
MMP10
RARRES2
TNXB
C20orf100
COX8C
GAS1
MSX1
RASSF1
TSPAN14
GNAS
GLOXD1
MTHFR
RASSF5
TUBA6
HIST2H4
GMFB
MYL9
RBM14
TUSC4
UBE2J1
C20orf29 C5 C8orf34 C9orf127 C9orf26 CACNB3
©201 1L andesBi os c i enc e. Donotdi s t r i but e. HIST3H3
GRB10
MYST4
RBM35B
LILRB5
GSDML
NARG1
RGS19
UGDH
RB1
HM13
NDUFB7
RNF166
USMG5
ZIM2
IGF2AS
NIP
RPL7A
VKORC1
CBX7
ZPBP2
INHBA
NSD1
RPRML
WDR37
CCND1
DCAKD
ISYNA1
NTN4
RRAS2
WSB2
CCND2
DDX5
ITGB4
OVOL1
SFRS9
ZDHHC3
CDC16
DHRS4
JAG1
PCAF
SH2D3A
ZFP36L2
CDKL3
DHX9
JUB
SHB
ZNF688
DIRAS2
SLC1A1
more than 2-fold in dialysis patients.16 Seven of the genes identified by methylation profiling (in bold in Table 6) appear associated with both diabetic nephropathy (Table 4) and dialysis (Table 6). Discussion At first sight, our results appear to be in contrast to those of Stenvinkel et al.14 who found that overall levels of DNA methylation were higher in patients with chronic kidney disease and inflammation, whether they were being treated by dialysis or not. However, these investigators measured relative levels of methylation at HpaII sites (CCGG), genome-wide, by the “LUMA” assay.17 Although both CpG sites and HpaII sites are enriched in gene promoters, the majority of HpaII sites in the genome are present in repeated sequences, such as the Alu family of SINE elements. Any inflammation-based global or genome-wide
24
PCGF1
increases in methylation may be more likely to affect repeated sequences than the largely gene promoter-based CpGs assayed on the Illumina BeadChip used in our study. We also note that Wilflingseder et al.16 found that dialysis was associated, overwhelmingly, with upregulation of gene expression and we found that genes that were differentially methylated in ESRD patients on HD were significantly less methylated in more than 90% of cases. Gene promoter methylation is, in general, inversely correlated with transcript level18 so the overall results of our study and Wilflingseder et al.16 are in agreement. Unfortunately, we were unable to determine whether the between-group methylation differences we observed were associated with differences in transcript level between the two groups because appropriate samples were unavailable. However, we have shown previously19 that 25–40% of methylation differences between groups are associated with moderate cis effects on mean transcript level. Even if only a fraction of the methylation
Epigenetics
Volume 6 Issue 1
Table 4. Genes differentially methylated in ESRD on HD patients that have been implicated in diabetic nephropathy or kidney development/ disease3,9-11 Genes implicated in diabetic nephropathy or kidney disease AKR1B1
Genes associated with ESRD or HD by methylation profiling
Genes expressed in kidney, identified by methylation profiling
AKR1B1 ANKRD15
CTSD, CTSL
CTSZ
CYBA
CYBASC3, CYP2S1, CYP8B1, COX8C
ANKRD15
FZD2
FZD2
ITGA1, ITGA3
ITGB4
ITGB4
IGF1
IGF2AS
LGALS3
LGALS3BP
MMP1, MMP2, MMP3, MMP9
MMP10
MTHFR
MTHFR
MYH9
MYL9
SLC2A2, SLC12A3 TIMP2, TIMP3 UBA52
NTN4
NTN4
PTGER3
PTGER3
SLC1A1, SLC2A4, SLC4A3
33% (data not shown). Although only two tissues are compared in our study, the two tissues diverge early in development (extraembryonic vs. embryonic) and might be expected to differ greatly from one another. Although our estimate would be expected to rise by comparing additional tissues, both our estimate and that of Rakyan et al.18 suggest that the majority of CpG sites should be methylated similarly between most tissues in most individuals. With respect to their utility as biomarkers, the most important variable is not the level of variability between tissues within an individual but the level of variability in the same tissue between individuals of different disease state or disease predisposition. We note that when Benjamini-Hochberg correction for multiple testing is applied to our data set, the number of CpGs identified as significantly different falls dramatically. For instance, the number of CpGs that have a DiffScore of greater/less than 20 falls to 30, from 2,870 detected without correction. However, we were well aware that nearly 50% of the original 2,870 CpGs were likely false-positives (1,379 expected at p < 0.05), and chose to apply the additional criterion of identifying candidate genes by selecting only those with two or more CpGs significantly different. This “common sense” biological criterion resulted in the elimination of 87% of the original CpGs, leaving only 389 of the 2,870. As a measure of the relative difference between a strict statistical selection and the selection used to identify the genes in Table 3, if we take all Benjamini-Hochberg corrected CpGs with DiffScores of greater/less than 15, there are 212 CpGs that identify 30 of the same genes identified by the “2 CpGs in the same gene” procedure. We note, also, that our sample comparison groups contain mostly African American patients, with only a small number of Hispanic patients. If the Hispanic patients are removed from the analysis, under the possibility that they introduce additional variance into the data set because of genetic heterogeneity, a larger number of candidate genes in which two CpGs are significantly different is identified. However, 146 of the original 187 candidates genes (76%) and 2,359 of the 2,870 original significantly different CpGs (82%) are the same when the Hispanic patients are excluded from the analysis, arguing that our conclusions on the original data set are conservative. We hope to address these and other concerns in a larger study in which patient groups may be subdivided by appropriate clinical considerations (comparison of patients with diabetic nephropathy who are pre-dialysis with diabetic nephropathy dialysis patients, for example) and in which environmental factors are taken into consideration.
©201 1L andesBi os c i enc e. Donotdi s t r i but e. TIMP4
UBE2J1
differences that we detect are associated with differences in transcript level, the metabolic networks likely to be affected (Table 5) appear to be those that are involved in diabetic nephropathy and/ or causes or effects of dialysis. At this point, it is not possible to distinguish whether any particular difference we observe between patient-groups is caused by the ESRD state, per se, treatments associated with ESRD, or are predisposing factors that lead to ESRD. However, the fact that we observe methylation differences in genes that are involved in kidney disease progression, and that we observe these differences in non-renal tissues, suggests that either the patients who develop ESRD have constitutional epigenetic differences (similar to the constitutional “loss of imprinting” observed at IGF2/H19 in colon cancer patients)20 that may, themselves, have a genetic basis or that the differences are the result of environmental factors that alter epigenetic marks in a similar way in all tissues exposed. In this regard we note that although there is a common perception that CpG sites are differentially methylated between tissues, the fraction of CpG sites that are actually methylated differently appears modest. Rakyan et al.18 have estimated that only 18% of the ~25,000 genes analyzed in 16 different tissues show “tissue-specific” patterns of methylation, with between tissue differences of more than 33%.18 We have corroborated this result by comparing cord blood and placenta samples from 71 children on either an Illumina GoldenGate array (15) or the Illumina Infinium array we have used in this study. Of the 58,228 CpGs queried between the two tissues, only 11% varied by more than
Methods Saliva sample collection and DNA extraction. Diabetes patients with ESRD were recruited from the DCI and Fresenius Dialysis Clinics in Philadelphia and diabetes patients without nephropathy were recruited through the Department of Endocrinology at Temple University Medical School. Those who gave informed consent (Temple University IRB approved protocol 11061) to participate in the study were provided with an “Oragene DNA” saliva collection kit (OG-100) (www.dnagenotek.com/index.
www.landesbioscience.com Epigenetics
25
©201 1L andesBi os c i enc e. Donotdi s t r i but e. Figure 2. Each plot shows methylation levels at two CpGs in ESRD/HD-associated candidate genes MMP10 (A), MTHFR (B), TIMP4 (C) and CLIC6 (D) in all 46 patients. For MTHFR, the graph shows the two CpGs with the largest difference between patient groups.
html) into which the patients expectorated the required amount (~2 ml) of saliva. All interactions with patients conformed to the principles of the Declaration of Helsinki. DNA was prepared from saliva according to the manufacturer’s protocol. Sample preparation and array processing. 500 ng of each DNA sample was treated with sodium bisulfite using the EZ DNA Methylation-Gold Kit D5005 (Zymo Research, Orange, CA). Completeness of conversion was monitored by bisulfite pyrosequencing of 12 CpG sites in the imprinted SNRPN gene, using a commercially available assay (EpigenDx, Worcester, MA). “HumanMethylation27 BeadChip” arrays were purchased from Illumina, Inc., (San Diego, CA) and samples were processed at the University of Pennsylvania Microarray Facility. Data analysis. Data analysis was performed using the Illumina Genome Studio 2008.1 software. Greater than 23,000 of the 27,578 CpGs on the array were detected in all samples but one. This ESRD on HD sample and a diabetes patient without nephropathy, matched for sex and having the lowest number of CpGs detected in this group (24,849), were excluded from the analysis. Unsupervised hierarchical clustering using all CpGs clustered samples largely on the basis of sex (Sup. Fig. 1),
26
suggesting methylation differences on the inactive X chromosome in females was the largest influence. In fact, 728 of the 1,085 X-linked CpGs exhibited significant differences in mean beta-value between males and females, again arguing for high quality of sample preparation and array processing. The 1,085 CpGs on the X chromosome and 7 CpGs on the Y chromosome were then excluded from the cluster analysis to eliminate the effect of sex chromosome differences in methylation (Fig. 1). Patient group mean b-value methylation at each site was compared and “DiffScores” computed. We note that all analyses were performed using the background normalized b-values to correct for background differences between arrays; however, additional normalization procedures were not employed because beta values are fractional and based on ratios of signals between different probes on the same array and are constrained between 0 and 1. In addition, normalization procedures have the effect of discounting small but significant differences between groups on the basis that they are unlikely to be of biological significance. However, we have found that some small absolute differences in methylation are correlated with between group differences in transcript levels19 and we did not wish to bias our selection of genes as
Epigenetics
Volume 6 Issue 1
Table 5. Top gene functions of the 187 ESRD/HD-associated gene candidates in ingenuity pathway analysis IPA functional category
Genes that are differentially methylated between diabetes patients without nephropathy and ESRD patients on HD
Renal fibrosis, renal proliferation, renal dysplasia
STMN1, CDKN1C
Cardiac hypertrophy, cardiac hemorrhaging
FBXO32, IL33, JUN, TCF14, CTBP1, INHBA
Liver proliferation, liver fibrosis, liver necrosis/cell death, liver hematopoiesis, liver dysplasia
JUN, CCND1, RB1, C5, SOCS1
Cell cycle, developmental disorder, genetic disorder
APC, AURKA, CCND2, CDC16, CDKN1C, Cyclin A, Cyclin D, Cyclin E, DAB2IP, ERK, FEN1, GC-GCR dimer, GRB10, Growth hormone, HIST3H3 (includes EG:8290), Histone h3, Histone h4, INHBA, JAG1, JUB, KAT2B, MEF2, Mhc ii, MKNK2, NSD1, NTN4, PFTK1, POLE, Rar, Rb, RB1, RPL7A, SOCS1, SPINT2, SWI-SNF
Amino acid metabolism, small molecule biochemistry, cancer
Actin, Akt, ALP, ARHGAP22, BAIAP2, CFL1, CYP8B1, DHX9, DTL, Fgf, FSH, GNAS, hCG, ITGB4, JUN, MCAM, Mmp, MMP10, MTHFR, Pdgf, PDGF BB, PDLIM3, Pka, PLC, PP2A, SLC1A1, SPRR2A, SPRY4, STMN1, Tgf beta, TIMP4, TUBA1C, Tubulin, UGDH, Vegf
Connective tissue development and function, tissue development, cancer
ADCY, ALDH3B2 (includes EG:222), Ap1, APC, Ck2, CSNK1E, Ctbp, CTBP1, DDX5, EIF4B, FBXO32, FZD2, GAS1, Gsk3, HIST2H4A, IgG, IL1, IL12 (complex), Interferon a, Jnk, LMO4, MAP3K3, MED22, MIR1, P38 MAPK, PHLDA2, Pias, Pkc(s), POLR2F, POLR3D, Proteasome, PTGER3, RGS19, RNA polymerase II, SFRS9
Cancer, respiratory disease, tumor morphology
ANXA13, BRD3, C5, Cbp/p300, CCND1, CLIP2, Creb, EDA2R, ERK1/2, FHIT, GMFB, Hat, IL33, Integrin, Laminin, MAP2K1/2, Mek, MYST4, NFκB (complex), Nuclear factor 1, PI3K, PLCL1, Rac, Raf, Rap, Rap1, RAP1B, Ras, Ras homolog, RASSF1, RASSF5, RRAS2, SRF, TCR
Organismal injury and abnormalities, cellular development, nervous system development and function
ADAMTS10 (includes EG:81794), AKR1B1, AKR1B10, ATP6V1B2, BOC, BTNL2, CDON, CLC, COX15, COX8C, CRLF1, CTSZ (includes EG:1522), CYP2S1, Cytochrome c oxidase, DDAH2, EPO, FGF2, FGFRL1, GDF3, IFNG, MC1R, MEST, NDUFB7, NDUFV1, NR3C1, PSMB9, PSME2, RARRES2, retinoic acid, SHB, STAT3, TNF, TNFAIP2, WDR37
Cellular growth and proliferation, hematological system development and function, dental disease
ALDH3A2, ARL4A, ATN1, BMP4, BTNL2, C9ORF127, CASP1, CASP14, CORO1C, CRIP2, DLX2, DLX3, DLX5, FURIN, HRH2, IL4, KIAA1199, KLK10, LEFTY1, LGALS3BP, LTA4H, LYST, MSX1, NODAL, OVOL1, PAX9, PCSK4, PCSK7, SOLH, SOSTDC1, SP1, TGFβ1, Timp, TPSB2, UBE2J1
Cellular compromise, neurological disease, organismal injury and abnormalities
ACOT2, ATP5I, ATP6AP1, ATP6V1D, CHRNB3, DCAKD, DIRAS2, DNM, DNM2, dopamine, FAM105A, G-protein beta, HM13, HTT, ISYNA1, JPH2, KANK1, LEPRE1, Mapk, MIR124, MOGS, MYO1E, NARG1, NME2, PKNOX1, PRKAR1A, REST, SFXN3, SLC2A4, SUCLG2, SYNGR2, TSPAN14, TUSC4, Ubiquitin, XBP1
Cancer, cellular compromise, cellular movement
ATP10A, BAT1, CDKL3, CNN1, COPB1, CRY1, CTSZ (includes EG:1522), DHRS4, EMG1, F9, FAM120B, GRIP1, HNF4A, KHDRBS1, KIF1B, LAPTM4A, NSF, PFKFB4, PSMB7, PSMC4, PTPRN, SFRS2, SLC25A5, SMARCA5, TMEM79, TRAF2, TRAIP, TUBB4, UBE2N, UBQLN1, UBQLN4, USMG5, VKORC1, ZDHHC3
Cellular development, cellular growth and proliferation, reproductive system development and function
Alpha actin, Bcl9-Cbp/p300-Ctnnb1-Lef/Tcf, CAPZA2, CCNE2, CHPF, COL5A2, CPSF6, DAAM1, EGFR, ELF3, EP300, ERBB2, HABP2, HNF1A, HSPB8, LAMC2, LTA4H, MACF1, MXI1, MYL9 (includes EG:10398), PHF2, PHYH, PIK3C2A, PIN1, PTGER1, RBM14, RIN2, SH2D3A, SLPI, SOAT2, SOLH, TFF3, TNXB, WSB2, ZFP36L2
Lipid metabolism, molecular transport, small molecule biochemistry
ACSL3, ACSL6, ACTL6A, APP, ASXL1, ASXL2, ASXL3, BCOR, BPHL, C1ORF149, C20ORF20, C8ORF34, CBX2, CBX7, CBX8, CDKN2A, CLIC6, DAG1, DCTN4, EFNA3, EMX1, EPHA3, ESRP2, FBXL16, L3MBTL2, MIR220A, MIRLET7E (includes EG:406887), MIRLET7G (includes EG:406890), oleic acid, PCGF1, PCGF6, PHF19, POMGNT1, RING1, RNF2
Cell cycle, cell signaling, molecular transport
AGT, AGTR2, ANXA3, AR, BCL2L10, b-estradiol, BIK, CACNA1C, CACNA2D1, CACNB3, CACNB4, CALD1, Caspase, CCND3, CLIC3, COX5A, CYTH3, F Actin, GHRHR, HBXIP, KIAA0922, LEFTY2, LIMA1, MAF1, MAPK15, phosphatidylinositol-3,4,5-triphosphate, PIP5KL1, RASGRP1, SLC4A3, SPINLW1, STK3, TCF15, TERT, TMOD3, TOX2
©201 1L andesBi os c i enc e. Donotdi s t r i but e.
candidates on this basis. DiffScores are related to p-values by the relationship: DiffScore = 10sgn(b-valueESRD - b-valuediabetes no nephropathy)log10p All CpGs with DiffScores greater than 20 (significantly more methylated in ESRD patients on HD) or less than -20 (significantly less methylated in ESRD patients on HD) were considered as potential candidates for genes associated with ESRD and/or HD. This selection resulted in 2,870 CpGs in more than 2,000 genes. However, the majority of these genes were significantly different at only a single CpG site (in fact, 2549 of the genes
on the array are represented by only a single CpG and 191 of these were among the more than 2,000 selected on the basis of DiffScores). We have used the criterion requiring two or more CpGs significantly different in a previous study19 and found that 25–40% of candidates selected adding this criterion show significant differences in transcript level. Applying this additional criterion, 389 CpGs in 187 genes were selected as candidates on the basis of strong epigenetic differences between groups. Calculation of fold-enrichment of genes involved in diabetic nephropathy or HD. 2,549 of the 14,495 genes on the
www.landesbioscience.com Epigenetics
27
HumanMethylation27 array are represented by only a single CpG and so cannot fulfill the “two or more CpGs” criterion. The 187 genes that do fulfill this criterion are, therefore, selected from a possible pool of 11,946 genes and represent 0.016 of the total. By contrast, the 21 genes known or suspected to be involved in diabetic nephropathy (Table 2) represent 0.11 of the 187 genes or a 6.9-fold enrichment.
Table 6. Comparison of differentially methylated genes found in dialysis vs. upregulated in Wilflingseder et al. 2008
Acknowledgements
Helen Elivera R.N., assisted with collection of the specimens and consenting of patients and Drs. Sridhar Reddy and Colleen Veloski were extremely helpful in selecting patients and reviewing chart data. We gratefully acknowledge the assistance of Lior Benjamin in preparing samples. Financial Support
This research was supported by a grant from the Paul Teschan Fund, Dialysis Clinic, Inc., Nashville, TN. Note
Supplementary materials can be found at: www.landesbioscience.com/supplement/Sapienza-EPI6-1-Sup. pdf www.landesbioscience.com/supplement/Sapienza-EPI6-1-Sup. xls
2.
3.
4.
5.
6.
7.
8.
28
Wilflingseder et al.
WSB2
WSB2
JUN
JUN
KIF1B
KIF11
ITGB4
ITGB1
LGALS3BP
LGALS9
RGS19
RGS1, RGS6, RGS18
ARHGAP22
RHGAP30
MAP3K3
MAP3K8, MAPK14
MYL9
MYLK
RASSF1, RASSF5
RASSF4
SOCS1
SOCS3
STK6
STK17B
AKR1B1
AKR1C3
SLC1A1, SLC2A4, SLC4A3
SLC27A3
LILRB5
LILRB3
Genes in bold may be associated with diabetic nephropathy, per se, as well as dialysis.
©201 1L andesBi os c i enc e. Donotdi s t r i but e.
References 1.
Our report
USA Senate bill S. 976: “Genomics and Personalized Medicine Act of 2007.” www.govtrack.us/congress/bill. xpd?bill=s110-976 Sladek R, Rocheleau G, Rung J, Dina C, Shen L, Serre D, et al. A genome-wide association study identifies novel risk loci for type 2 diabetes. Nature 2007; 445:828-30. Scott LJ, Mohlke KL, Bonnycastle LL, Willer CJ, Li Y, Duren WL, et al. A genome-wide association study of type 2 diabetes in Finns detects multiple susceptibility variants. Science 2007; 316:1341-5. Freedman BI, Bostrom M, Daeihagh P, Bowden DW. Genetic factors in diabetic nephropathy. Clin J Am Soc Nephrol 2007; 2:1306-16. Imperatore G, Knowler WC, Nelson RG, Hanson RL. Genetics of diabetic nephropathy in the Pima Indians. Curr Diab Rep 2001; 1:275-81. Ewens KG, George RA, Sharma K, Ziyadeh FN, Spielman RS. Assessment of 115 candidate genes for diabetic nephropathy by transmission/disequilibrium test. Diabetes 2005; 54:3305-18. Kopp JB, Smith MW, Nelson GW, Johnson RC, Freedman BI, Bowden DW. MYH9 is a major-effect risk gene for focal segmental glomerulosclerosis. Nat Genet 2008; 40:1175-84. Freedman BI, Hicks PJ, Bostrom MA, Cunningham ME, Liu Y, Divers J, et al. Polymorphisms in the non-muscle myosin heavy chain 9 gene (MYH9) are strongly associated with end-stage renal disease historically attributed to hypertension in African Americans. Kidney Int 2009; 75:736-45.
9. Kao WH, Klag MJ, Meoni LA, Reich D, BerthierSchaad Y, Li M, et al. MYH9 is associated with nondiabetic end-stage renal disease in African Americans. Nat Genet 2008; 40:1185-92. 10. Wheeler HE, Metter EJ, Tanaka T, Absher D, Higgins J, Zahn JM, et al. Sequential use of transcriptional profiling, expression quantitative trait mapping and gene association implicates MMP20 in human kidney aging. PLoS Genet 2009; 5:1000685. 11. Liu Y. Advanced oxidation protein products: a causative link between oxidative stress and podocyte depletion. Kidney Int 2006; 69:213-7. 12. Friso S, Choi SW. Gene-nutrient interactions in onecarbon metabolism. Curr Drug Metab 2005; 6:37-46. 13. Schwahn B, Rozen R. Polymorphisms in the methylenetetrahydrofolate reductase gene: clinical consequences. Am J Pharmacogenomics 2001; 1:189-201. 14. Stenvinkel P, Karimi M, Johansson S, Axelsson J, Suliman M, Lindholm B, et al. Impact of inflammation on epigenetic DNA methylation—a novel risk factor for cardiovascular disease? J Intern Med 2007; 261:488-99. 15. Bonsib SM. Renal cystic diseases and renal neoplasms: a mini-review. Clin J Am Soc Nephrol 2009; 4:19982007.
Epigenetics
16. Wilflingseder J, Perco P, Kainz A, Korbély R, Mayer B, Oberbauer R. Biocompatibility of haemodialysis membranes determined by gene expression of human leucocytes: a crossover study. Eur J Clin Invest 2008; 38:918-24. 17. Karimi M, Johansson S, Ekström TJ. Using LUMA: a Luminometric-based assay for global DNAmethylation. Epigenetics 2006; 1:45-8. 18. Rakyan VK, Down TA, Thorne NP, Flicek P, Kulesha E, Gräf S, et al. An integrated resource for genomewide identification and analysis of human tissue-specific differentially methylated regions (tDMRs). Genome Res 2008; 18:1518-29. 19. Katari S, Turan N, Bibikova M, Erinle O, Chalian R, Foster M, et al. DNA methylation and gene expression differences in children conceived in vitro or in vivo. Hum Mol Genet 2009; 18:3769-78. 20. Cui H, Cruz-Correa M, Giardiello FM. Loss of IGF2 imprinting: a potential marker of colorectal cancer risk. Science 2003; 299:1753-5.
Volume 6 Issue 1
