Clin Exp Nephrol (2014) 18:197–200 DOI 10.1007/s10157-013-0874-9
REVIEW ARTICLE
WCN 2013 Satellite Symposium ‘‘Kidney and Lipids’’
Genetic associations in diabetic nephropathy Antien L. Mooyaart
Received: 25 August 2013 / Accepted: 17 September 2013 / Published online: 16 October 2013 Ó Japanese Society of Nephrology 2013
Abstract Diabetic nephropathy is a complex disease, caused by both environmental and genetic factors. As in most complex diseases, genetic association studies in diabetic nephropathy showed inconsistent results. In retrospect, studies with small sample sizes, given what are now known to be small odds ratios, were partially responsible for this poor replication record. Furthermore, the low prior probability in complex genetics and multiple testing played a role. Results become more consistent when one only considers those that were replicated. In a large meta-analysis study including only replicated associated genetic variants, 24 genetic variants in 16 genes were found to be associated with diabetic nephropathy. These genetic variants may provide novel biological insight. In particular, rare variants with a large effect found by hypothesis-free approaches (genome-wide association scans, next-generation sequencing) may open new avenues of discovery. Keywords Diabetic nephropathy Genetics Complex disease Meta-analysis
Introduction Control of blood pressure and blood glucose has been shown to reduce the risk of developing diabetic nephropathy (DN) [1]. However, there seems to be a difference between diabetes patients who develop DN and those who do not, which cannot be fully explained by environmental factors [2]. In both type 1 and type 2 diabetes familial A. L. Mooyaart (&) Department of Pathology, Leiden University Medical Center, Bldg.1, L1-Q, PO Box 9600, 2300 RC Leiden, The Netherlands e-mail: [email protected]
aggregation has been found and some ethnicities are more at risk than others [3, 4]. In type 1 diabetes, it has been reported that 30 % of patients develop DN within the first 15 years after the development of diabetes mellitus; however, if no signs of DN occur during that period the chances that they will ever develop DN are small [2]. This indicates that DN is not entirely due to environmental factors, but also some genetic predisposition seems to play a role. DN is therefore considered a complex disease, i.e., a disease caused by both environmental and genetic factors. Here, we present possible explanations for discrepancies in the field of genetics on complex diseases such as DN together with an overview and discuss future perspectives of genetic associations in DN.
Inconsistencies in genetic associations in complex diseases Over the last decades, many studies on candidate genes and genome-wide association scans (GWAS) have been published. A candidate gene is a gene which is suspected to be involved in a disease based on the literature. GWAS are association studies of a genome with common single nucleotide polymorphisms (SNPs) used as genetic markers. Since many SNPs are present in the human genome, this gives good coverage of the human genome. It maps the genome with common SNPs either directly involved in disease pathology or indirectly through tagSNPs, which are potentially linked with risk alleles. The field of genetic association studies (GWAS and candidate gene studies) in any complex disease has been plagued by inconsistencies in the published literature [5]. In retrospect, studies with small sample sizes, given what are now known to be small odds ratios (ORs), were
123
198
partially responsible for this poor replication record [6]. To give an indication on what sample sizes are needed, a study of 1,000 cases and 1,000 controls provides only 1 % power to detect a 20 % allele frequency that increases risk by a factor of 1.3 (= OR), but a study of 5,000 cases and 5,000 controls provides 98 % power [7]. Apart from underpowered studies, the prior probability of candidate genes and GWAS is low. Therefore, the chance of a false positive association being generated by chance increases [8]. For GWAS, it is clear that the prior probability is low as it is a hypothesis-free approach. However, we also know that candidate genes have a low prior probability because we have learned from monogenetic diseases that most of the genes that were found were completely unexpected [8]. Furthermore, often more than one hypothesis is investigated, leading to the problem of multiple testing, which increases the chance of a false positive. This applies to GWAS where thousands of SNPs are tested (even though some correction occurs) as well as to case–control studies where many different candidate genes are tested in one cohort. The low power, low prior probability as well as multiple testing make studies in the field of complex genetics susceptible for false negatives and false positives [7]. Therefore, independent replication of the detected association and large sample sizes (combined in a meta-analysis) are essential.
Genetic associations in diabetic nephropathy In 2011, Mooyaart et al. [9] published an overview of all genetic variants up to April 2010 which were replicated at least once. These were subsequently analyzed in a meta-
Fig. 1 Scheme of the large meta-analysis study performed by Mooyaart et al. Two examples are given—APOe being reproducibly associated with diabetic nephropathy and finally significant in the metaanalysis (right), and ADIPOQ which was found to be associated with diabetic nephropathy in at least two studies, but was finally, after including all studies of this gene, not associated in the metaanalysis
123
Clin Exp Nephrol (2014) 18:197–200
analysis study (Fig. 1). A total of 671 genetic association studies were found, finally leading to 24 genetic variants in 16 genes. Variants in or near ACE, ALR2 (2 variants), APOC1, APOe, EPO, eNOS (2 variants), HSPG2, VEGF, FRMD3 (2 variants), CARS (2 variants), UNC13B, CPVL/ CHN2, and GREM1 (as well as four variants not near known genes) were associated with DN and ELMO1, CCR5 and CNDP1 in a subgroup (Asians and type 2 diabetes, respectively). Of special interest are the genetic variants in the lipid metabolism, APOe and APOC1. The exact role of lipids in DN is unclear. However, it is believed that dyslipidemia might lead to glomerulosclerosis through oxidation of low-density lipoprotein cholesterol, increased uptake in the tubular epithelium and formation of mesangial foam cells [10]. ‘Diabetic dyslipidemia’ is mostly seen in type 2 diabetes and it is therefore interesting that APOe seems to be mainly associated with DN in type 2 diabetes patients. In our study, the effect sizes are relatively high, ranging from 1.1-1.7 (OR), when compared to other complex diseases. One explanation for this is the ‘winner’s curse’ phenomenon, in which the first publication shows the largest effect [11]. This is supported by a large study by Williams et al. [12] in 2012 in which they investigated the genetic variants from this meta-analysis study in three relatively large cohorts together with a sample size of 6,366 patients, and none of these associations could be reproduced. Considering a reasonably low prior probability (0.001) and a minor allele frequency of 10 %, this study would have at least 80 % power to detect associations with an OR of 1.5 [7]. Therefore, several of the variants (if there is no winner’s curse phenomenon) would have been found. In other complex diseases such as type 1 and 2 diabetes, and coronary heart disease, etc., much larger samples ([10 thousands) [13] were analyzed to be able to replicate
Clin Exp Nephrol (2014) 18:197–200
findings. Therefore, even larger studies are probably needed in DN to replicate these findings and determine exact effect sizes. To be realistic, probably not all of these candidate genes will be replicated in these kind of large samples, as seen in the other above-mentioned complex diseases. The discrepancy is probably due to publication bias, in which not only positive results with a large effect, but positive results in general, are more likely to be published [14]. This finally leads to false positive results in meta-analysis studies, even when only replicated genes are investigated. Larger samples are being collected in the SUMMIT consortium, and these might give more insight into which variants will remain to be associated with DN (http://www.imi-summit.eu/). The possible relevance of these genetic associations in DN can be roughly divided into two goals—genotypic screening for diabetes patients to be able to predict the risk of developing DN and novel biological insight in the etiology and pathogenesis of DN, potentially leading to new therapies [15]. For screening of individuals with an increased risk of disease, the measurement of the area under the curve (AUC) of the receiver operating characteristic curve (ROC) is often used. An AUC of 1 is a perfect prediction and an AUC of 0.5 is similar to tossing a coin, and therefore not predictive. It has been suggested that an AUC[0.75 can be considered of predictive value [16]. Studies on screening with genetic variants for DN have not yet been performed; however, this has been performed for screening for type 2 diabetes. A total of 18 genes have been established in type 2 diabetes and these genes when combined have been shown to reach an AUC of 0.6 in predicting onset of type 2 diabetes. This is low when compared to traditional risk factors as only body mass index, age and gender which have been shown to have an AUC of 0.78, and risk scores such as the QDS score (a scoring algorithm based on traditional risk factors without the need for laboratory tests) reach an AUC of 0.85 in predicting onset of type 2 diabetes [17]. In a different study on type 2 diabetes, the addition of genetic risk factors to clinical risk showed only a slight increase of the AUC (from 0.74 to 0.75), which seems to increase with duration of follow-up [18]. This indicates that genetic personalized medicine in type 2 diabetes will not easily be achieved in the near future. Even though much of the heritability in type 2 diabetes is unexplained, it is unlikely that the prediction by genetic markers will ever prove to be better than the use of environmental factors. Even though this is not yet investigated in DN, prediction might not seem to be the direction which is likely to be relevant. On the other hand, however, genetic association studies can also lead to new insights on pathogenesis which can then lead to the development of new therapeutic targets, biomarkers and opportunities for disease prevention.
199
Hypothesis-free approaches, such as GWAS, are most promising in this regard. To date, there are not many examples, if any, of new therapeutic strategies which came from genes found by GWAS in complex diseases. This, however, might be more a question of time.
The prospect of different genetic disease models Four models have been proposed to explain the genetic component in complex diseases—the common diseasecommon variant model (CD-CV model), the rare alleles of major effect (RAME) model, the infinitesimal model and the broad-sense heritability model [19]. The CD-CV model assumes that common variants contribute to risk, each explaining a small proportion of disease liability. Due to their high frequencies, these variants may explain a large part of the population risk. The GWAS, based on the CD-CV model, has proven to be successful [8]. An example of a common disease variant, which was consistently replicated, is APOe in Alzheimer’s disease and heart disease [20]. In contrast, the RAME model postulates that common diseases are in fact genetically heterogeneous and caused by de novo mutations. A rare variant in each individual (or family) with large effect may contribute to the risk. Support for this model comes from studies of rare genetic variants in the high-density lipoprotein C (HDL-C) gene [21, 22]. Rare variants are more often present in patients with low levels of HDL-C (\fifth percentile) than in those with high levels of HDL-C ([95th percentile). These findings were replicated in an independent study population [21]. A strong argument for rare variants comes from evolutionary theory. As disease is deleterious to survival of the fittest, variants that increase disease risk would be selected against; this is called purifying selection. This keeps these variants at a low frequency and therefore these variants stay rare. The fact that GWAS provide solid evidence for novel gene associations does not ultimately mean that the CD-CV model explains the genetic risk in complex diseases. Common SNPs identified by GWAS could be associated with rare mutations. Therefore, some common SNPs may be a proxy for a rare variant with large effect. The infinitesimal model assumes that liability for disease follows an asymptotic distribution, as shown for the continuous trait height. Hundreds of genes would be involved, covering a wide range of frequencies [23]. This model has gained support in GWAS, where ORs of 1.2 may each account for a small fraction of the liability. The broad-sense heritability model does not consider genetic variance to be additive, but rather multiplicative and consisting of interactions between genotype-environment, genotype–genotype and epigenetics. In order to
123
200
Clin Exp Nephrol (2014) 18:197–200
investigate this, even larger samples are required, even though it seems likely that there is a potential contribution from this model. The test burden will probably be larger than its relevance. As the main goal of finding genetic variants is novel insight into disease, it seems wise to focus on hypothesisfree approaches leading to rare variants with a relatively large effect on disease as these provide clear information on etiology compared to variants with small effects. Many genes with small effects might together give an increased risk of disease, but it remains unclear how. Rare variants (mutations) might lead to either upregulation, downregulation or a defect protein, which provides information about how this gene seems to affect disease etiology. Even though not all DN patients have a mutation in this gene, they might benefit from the information that a certain protein/gene is involved in DN and new biomarkers and therapeutic modalities can be developed.
Concluding remarks In the field of complex genetics, one has to consider the likelihood of a large number of false positive and false negative associations in the published literature. Recently, a large meta-analysis study showed 24 genetic variants in 16 genes which are the most likely to be associated with DN. Although the predictive role of these variants in DN still needs to be elucidated, its main value may be that it presents a novel biological insight. Conflict of interest
6. 7.
8. 9.
10. 11.
12.
13.
14. 15.
16.
17.
None to declare. 18.
References 1. Diabetes Control and Complications Trial (DCCT). Update. DCCT Research Group. Diabetes Care. 1990;13:427–33. 2. Krolewski AS, Warram JH, Rand LI, Kahn CR. Epidemiologic approach to the etiology of type I diabetes mellitus and its complications. N Engl J Med. 1987;317:1390–8. 3. Pettitt DJ, Saad MF, Bennett PH, Nelson RG, Knowler WC. Familial predisposition to renal disease in two generations of Pima Indians with type 2 (non-insulin-dependent) diabetes mellitus. Diabetologia. 1990;33:438–43. 4. Seaquist ER, Goetz FC, Rich S, Barbosa J. Familial clustering of diabetic kidney disease. Evidence for genetic susceptibility to diabetic nephropathy. N Engl J Med. 1989;320:1161–5. 5. Lohmueller KE, Pearce CL, Pike M, Lander ES, Hirschhorn JN. Meta-analysis of genetic association studies supports a
123
19. 20.
21.
22.
23.
contribution of common variants to susceptibility to common disease. Nat Genet. 2003;33:177–82. Zeggini E, Ioannidis JP. Meta-analysis in genome-wide association studies. Pharmacogenomics. 2009;10:191–201. Wacholder S, Chanock S, Garcia-Closas M, El Ghormli L, Rothman N. Assessing the probability that a positive report is false: an approach for molecular epidemiology studies. J Natl Cancer Inst. 2004;96:434–42. Altshuler D, Daly MJ, Lander ES. Genetic mapping in human disease. Science. 2008;322:881–8. Mooyaart AL, Valk EJ, van Es LA, et al. Genetic associations in diabetic nephropathy: a meta-analysis. Diabetologia. 2011;54: 544–53. Rosario RF, Prabhakar S. Lipids and diabetic nephropathy. Curr Diab Rep. 2006;6:455–62. Molendijk ML, Bus BA, Spinhoven P, Kaimatzoglou A, Oude Voshaar RC, Penninx BW, van IJzendoorn MH, Elzinga BM. A systematic review and meta-analysis on the association between BDNF val(66)met and hippocampal volume—a genuine effect or a winners curse? Am J Med Genet B Neuropsychiatr Genet. 2012;159B(6):731–40. doi:10.1002/ajmg.b.32078. Williams WW, Salem RM, McKnight AJ, et al. Association testing of previously reported variants in a large case-control meta-analysis of diabetic nephropathy. Diabetes. 2012;61: 2187–94. Welcome Trust Consortium. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature. 2007;447:661–78. Ioannidis JP. Why most published research findings are false. PLoS Med. 2005;2:e124. McCarthy MI, Abecasis GR, Cardon LR, et al. Genome-wide association studies for complex traits: consensus, uncertainty and challenges. Nat Rev Genet. 2008;9:356–69. Janssens AC, Moonesinghe R, Yang Q, Steyerberg EW, van Duijn CM, Khoury MJ. The impact of genotype frequencies on the clinical validity of genomic profiling for predicting common chronic diseases. Genet Med. 2007;9:528–35. Hippisley-Cox J, Coupland C, Robson J, Sheikh A, Brindle P. Predicting risk of type 2 diabetes in England and Wales: prospective derivation and validation of QDScore. BMJ. 2009;338:b880. doi:10.1136/bmj.b880.:b880. Lyssenko V, Jonsson A, Almgren P et al. Clinical risk factors, DNA variants, and the development of type 2 diabetes. N Engl J Med. 2008;359:2220–32. Gibson G. Rare and common variants: twenty arguments. Nat Rev Genet. 2012;13:135–45. Payami H, Kaye J, Heston LL, Bird TD, Schellenberg GD. Apolipoprotein E genotype and Alzheimer’s disease. Lancet. 1993;342:738. Cohen JC, Kiss RS, Pertsemlidis A, Marcel YL, McPherson R, Hobbs HH. Multiple rare alleles contribute to low plasma levels of HDL cholesterol. Science. 2004;305:869–72. Romeo S, Pennacchio LA, Fu Y, et al. Population-based resequencing of ANGPTL4 uncovers variations that reduce triglycerides and increase HDL. Nat Genet. 2007;39:513–6. Goldstein DB. Common genetic variation and human traits. N Engl J Med. 2009;360:1696–8.
REVIEW ARTICLE
WCN 2013 Satellite Symposium ‘‘Kidney and Lipids’’
Genetic associations in diabetic nephropathy Antien L. Mooyaart
Received: 25 August 2013 / Accepted: 17 September 2013 / Published online: 16 October 2013 Ó Japanese Society of Nephrology 2013
Abstract Diabetic nephropathy is a complex disease, caused by both environmental and genetic factors. As in most complex diseases, genetic association studies in diabetic nephropathy showed inconsistent results. In retrospect, studies with small sample sizes, given what are now known to be small odds ratios, were partially responsible for this poor replication record. Furthermore, the low prior probability in complex genetics and multiple testing played a role. Results become more consistent when one only considers those that were replicated. In a large meta-analysis study including only replicated associated genetic variants, 24 genetic variants in 16 genes were found to be associated with diabetic nephropathy. These genetic variants may provide novel biological insight. In particular, rare variants with a large effect found by hypothesis-free approaches (genome-wide association scans, next-generation sequencing) may open new avenues of discovery. Keywords Diabetic nephropathy Genetics Complex disease Meta-analysis
Introduction Control of blood pressure and blood glucose has been shown to reduce the risk of developing diabetic nephropathy (DN) [1]. However, there seems to be a difference between diabetes patients who develop DN and those who do not, which cannot be fully explained by environmental factors [2]. In both type 1 and type 2 diabetes familial A. L. Mooyaart (&) Department of Pathology, Leiden University Medical Center, Bldg.1, L1-Q, PO Box 9600, 2300 RC Leiden, The Netherlands e-mail: [email protected]
aggregation has been found and some ethnicities are more at risk than others [3, 4]. In type 1 diabetes, it has been reported that 30 % of patients develop DN within the first 15 years after the development of diabetes mellitus; however, if no signs of DN occur during that period the chances that they will ever develop DN are small [2]. This indicates that DN is not entirely due to environmental factors, but also some genetic predisposition seems to play a role. DN is therefore considered a complex disease, i.e., a disease caused by both environmental and genetic factors. Here, we present possible explanations for discrepancies in the field of genetics on complex diseases such as DN together with an overview and discuss future perspectives of genetic associations in DN.
Inconsistencies in genetic associations in complex diseases Over the last decades, many studies on candidate genes and genome-wide association scans (GWAS) have been published. A candidate gene is a gene which is suspected to be involved in a disease based on the literature. GWAS are association studies of a genome with common single nucleotide polymorphisms (SNPs) used as genetic markers. Since many SNPs are present in the human genome, this gives good coverage of the human genome. It maps the genome with common SNPs either directly involved in disease pathology or indirectly through tagSNPs, which are potentially linked with risk alleles. The field of genetic association studies (GWAS and candidate gene studies) in any complex disease has been plagued by inconsistencies in the published literature [5]. In retrospect, studies with small sample sizes, given what are now known to be small odds ratios (ORs), were
123
198
partially responsible for this poor replication record [6]. To give an indication on what sample sizes are needed, a study of 1,000 cases and 1,000 controls provides only 1 % power to detect a 20 % allele frequency that increases risk by a factor of 1.3 (= OR), but a study of 5,000 cases and 5,000 controls provides 98 % power [7]. Apart from underpowered studies, the prior probability of candidate genes and GWAS is low. Therefore, the chance of a false positive association being generated by chance increases [8]. For GWAS, it is clear that the prior probability is low as it is a hypothesis-free approach. However, we also know that candidate genes have a low prior probability because we have learned from monogenetic diseases that most of the genes that were found were completely unexpected [8]. Furthermore, often more than one hypothesis is investigated, leading to the problem of multiple testing, which increases the chance of a false positive. This applies to GWAS where thousands of SNPs are tested (even though some correction occurs) as well as to case–control studies where many different candidate genes are tested in one cohort. The low power, low prior probability as well as multiple testing make studies in the field of complex genetics susceptible for false negatives and false positives [7]. Therefore, independent replication of the detected association and large sample sizes (combined in a meta-analysis) are essential.
Genetic associations in diabetic nephropathy In 2011, Mooyaart et al. [9] published an overview of all genetic variants up to April 2010 which were replicated at least once. These were subsequently analyzed in a meta-
Fig. 1 Scheme of the large meta-analysis study performed by Mooyaart et al. Two examples are given—APOe being reproducibly associated with diabetic nephropathy and finally significant in the metaanalysis (right), and ADIPOQ which was found to be associated with diabetic nephropathy in at least two studies, but was finally, after including all studies of this gene, not associated in the metaanalysis
123
Clin Exp Nephrol (2014) 18:197–200
analysis study (Fig. 1). A total of 671 genetic association studies were found, finally leading to 24 genetic variants in 16 genes. Variants in or near ACE, ALR2 (2 variants), APOC1, APOe, EPO, eNOS (2 variants), HSPG2, VEGF, FRMD3 (2 variants), CARS (2 variants), UNC13B, CPVL/ CHN2, and GREM1 (as well as four variants not near known genes) were associated with DN and ELMO1, CCR5 and CNDP1 in a subgroup (Asians and type 2 diabetes, respectively). Of special interest are the genetic variants in the lipid metabolism, APOe and APOC1. The exact role of lipids in DN is unclear. However, it is believed that dyslipidemia might lead to glomerulosclerosis through oxidation of low-density lipoprotein cholesterol, increased uptake in the tubular epithelium and formation of mesangial foam cells [10]. ‘Diabetic dyslipidemia’ is mostly seen in type 2 diabetes and it is therefore interesting that APOe seems to be mainly associated with DN in type 2 diabetes patients. In our study, the effect sizes are relatively high, ranging from 1.1-1.7 (OR), when compared to other complex diseases. One explanation for this is the ‘winner’s curse’ phenomenon, in which the first publication shows the largest effect [11]. This is supported by a large study by Williams et al. [12] in 2012 in which they investigated the genetic variants from this meta-analysis study in three relatively large cohorts together with a sample size of 6,366 patients, and none of these associations could be reproduced. Considering a reasonably low prior probability (0.001) and a minor allele frequency of 10 %, this study would have at least 80 % power to detect associations with an OR of 1.5 [7]. Therefore, several of the variants (if there is no winner’s curse phenomenon) would have been found. In other complex diseases such as type 1 and 2 diabetes, and coronary heart disease, etc., much larger samples ([10 thousands) [13] were analyzed to be able to replicate
Clin Exp Nephrol (2014) 18:197–200
findings. Therefore, even larger studies are probably needed in DN to replicate these findings and determine exact effect sizes. To be realistic, probably not all of these candidate genes will be replicated in these kind of large samples, as seen in the other above-mentioned complex diseases. The discrepancy is probably due to publication bias, in which not only positive results with a large effect, but positive results in general, are more likely to be published [14]. This finally leads to false positive results in meta-analysis studies, even when only replicated genes are investigated. Larger samples are being collected in the SUMMIT consortium, and these might give more insight into which variants will remain to be associated with DN (http://www.imi-summit.eu/). The possible relevance of these genetic associations in DN can be roughly divided into two goals—genotypic screening for diabetes patients to be able to predict the risk of developing DN and novel biological insight in the etiology and pathogenesis of DN, potentially leading to new therapies [15]. For screening of individuals with an increased risk of disease, the measurement of the area under the curve (AUC) of the receiver operating characteristic curve (ROC) is often used. An AUC of 1 is a perfect prediction and an AUC of 0.5 is similar to tossing a coin, and therefore not predictive. It has been suggested that an AUC[0.75 can be considered of predictive value [16]. Studies on screening with genetic variants for DN have not yet been performed; however, this has been performed for screening for type 2 diabetes. A total of 18 genes have been established in type 2 diabetes and these genes when combined have been shown to reach an AUC of 0.6 in predicting onset of type 2 diabetes. This is low when compared to traditional risk factors as only body mass index, age and gender which have been shown to have an AUC of 0.78, and risk scores such as the QDS score (a scoring algorithm based on traditional risk factors without the need for laboratory tests) reach an AUC of 0.85 in predicting onset of type 2 diabetes [17]. In a different study on type 2 diabetes, the addition of genetic risk factors to clinical risk showed only a slight increase of the AUC (from 0.74 to 0.75), which seems to increase with duration of follow-up [18]. This indicates that genetic personalized medicine in type 2 diabetes will not easily be achieved in the near future. Even though much of the heritability in type 2 diabetes is unexplained, it is unlikely that the prediction by genetic markers will ever prove to be better than the use of environmental factors. Even though this is not yet investigated in DN, prediction might not seem to be the direction which is likely to be relevant. On the other hand, however, genetic association studies can also lead to new insights on pathogenesis which can then lead to the development of new therapeutic targets, biomarkers and opportunities for disease prevention.
199
Hypothesis-free approaches, such as GWAS, are most promising in this regard. To date, there are not many examples, if any, of new therapeutic strategies which came from genes found by GWAS in complex diseases. This, however, might be more a question of time.
The prospect of different genetic disease models Four models have been proposed to explain the genetic component in complex diseases—the common diseasecommon variant model (CD-CV model), the rare alleles of major effect (RAME) model, the infinitesimal model and the broad-sense heritability model [19]. The CD-CV model assumes that common variants contribute to risk, each explaining a small proportion of disease liability. Due to their high frequencies, these variants may explain a large part of the population risk. The GWAS, based on the CD-CV model, has proven to be successful [8]. An example of a common disease variant, which was consistently replicated, is APOe in Alzheimer’s disease and heart disease [20]. In contrast, the RAME model postulates that common diseases are in fact genetically heterogeneous and caused by de novo mutations. A rare variant in each individual (or family) with large effect may contribute to the risk. Support for this model comes from studies of rare genetic variants in the high-density lipoprotein C (HDL-C) gene [21, 22]. Rare variants are more often present in patients with low levels of HDL-C (\fifth percentile) than in those with high levels of HDL-C ([95th percentile). These findings were replicated in an independent study population [21]. A strong argument for rare variants comes from evolutionary theory. As disease is deleterious to survival of the fittest, variants that increase disease risk would be selected against; this is called purifying selection. This keeps these variants at a low frequency and therefore these variants stay rare. The fact that GWAS provide solid evidence for novel gene associations does not ultimately mean that the CD-CV model explains the genetic risk in complex diseases. Common SNPs identified by GWAS could be associated with rare mutations. Therefore, some common SNPs may be a proxy for a rare variant with large effect. The infinitesimal model assumes that liability for disease follows an asymptotic distribution, as shown for the continuous trait height. Hundreds of genes would be involved, covering a wide range of frequencies [23]. This model has gained support in GWAS, where ORs of 1.2 may each account for a small fraction of the liability. The broad-sense heritability model does not consider genetic variance to be additive, but rather multiplicative and consisting of interactions between genotype-environment, genotype–genotype and epigenetics. In order to
123
200
Clin Exp Nephrol (2014) 18:197–200
investigate this, even larger samples are required, even though it seems likely that there is a potential contribution from this model. The test burden will probably be larger than its relevance. As the main goal of finding genetic variants is novel insight into disease, it seems wise to focus on hypothesisfree approaches leading to rare variants with a relatively large effect on disease as these provide clear information on etiology compared to variants with small effects. Many genes with small effects might together give an increased risk of disease, but it remains unclear how. Rare variants (mutations) might lead to either upregulation, downregulation or a defect protein, which provides information about how this gene seems to affect disease etiology. Even though not all DN patients have a mutation in this gene, they might benefit from the information that a certain protein/gene is involved in DN and new biomarkers and therapeutic modalities can be developed.
Concluding remarks In the field of complex genetics, one has to consider the likelihood of a large number of false positive and false negative associations in the published literature. Recently, a large meta-analysis study showed 24 genetic variants in 16 genes which are the most likely to be associated with DN. Although the predictive role of these variants in DN still needs to be elucidated, its main value may be that it presents a novel biological insight. Conflict of interest
6. 7.
8. 9.
10. 11.
12.
13.
14. 15.
16.
17.
None to declare. 18.
References 1. Diabetes Control and Complications Trial (DCCT). Update. DCCT Research Group. Diabetes Care. 1990;13:427–33. 2. Krolewski AS, Warram JH, Rand LI, Kahn CR. Epidemiologic approach to the etiology of type I diabetes mellitus and its complications. N Engl J Med. 1987;317:1390–8. 3. Pettitt DJ, Saad MF, Bennett PH, Nelson RG, Knowler WC. Familial predisposition to renal disease in two generations of Pima Indians with type 2 (non-insulin-dependent) diabetes mellitus. Diabetologia. 1990;33:438–43. 4. Seaquist ER, Goetz FC, Rich S, Barbosa J. Familial clustering of diabetic kidney disease. Evidence for genetic susceptibility to diabetic nephropathy. N Engl J Med. 1989;320:1161–5. 5. Lohmueller KE, Pearce CL, Pike M, Lander ES, Hirschhorn JN. Meta-analysis of genetic association studies supports a
123
19. 20.
21.
22.
23.
contribution of common variants to susceptibility to common disease. Nat Genet. 2003;33:177–82. Zeggini E, Ioannidis JP. Meta-analysis in genome-wide association studies. Pharmacogenomics. 2009;10:191–201. Wacholder S, Chanock S, Garcia-Closas M, El Ghormli L, Rothman N. Assessing the probability that a positive report is false: an approach for molecular epidemiology studies. J Natl Cancer Inst. 2004;96:434–42. Altshuler D, Daly MJ, Lander ES. Genetic mapping in human disease. Science. 2008;322:881–8. Mooyaart AL, Valk EJ, van Es LA, et al. Genetic associations in diabetic nephropathy: a meta-analysis. Diabetologia. 2011;54: 544–53. Rosario RF, Prabhakar S. Lipids and diabetic nephropathy. Curr Diab Rep. 2006;6:455–62. Molendijk ML, Bus BA, Spinhoven P, Kaimatzoglou A, Oude Voshaar RC, Penninx BW, van IJzendoorn MH, Elzinga BM. A systematic review and meta-analysis on the association between BDNF val(66)met and hippocampal volume—a genuine effect or a winners curse? Am J Med Genet B Neuropsychiatr Genet. 2012;159B(6):731–40. doi:10.1002/ajmg.b.32078. Williams WW, Salem RM, McKnight AJ, et al. Association testing of previously reported variants in a large case-control meta-analysis of diabetic nephropathy. Diabetes. 2012;61: 2187–94. Welcome Trust Consortium. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature. 2007;447:661–78. Ioannidis JP. Why most published research findings are false. PLoS Med. 2005;2:e124. McCarthy MI, Abecasis GR, Cardon LR, et al. Genome-wide association studies for complex traits: consensus, uncertainty and challenges. Nat Rev Genet. 2008;9:356–69. Janssens AC, Moonesinghe R, Yang Q, Steyerberg EW, van Duijn CM, Khoury MJ. The impact of genotype frequencies on the clinical validity of genomic profiling for predicting common chronic diseases. Genet Med. 2007;9:528–35. Hippisley-Cox J, Coupland C, Robson J, Sheikh A, Brindle P. Predicting risk of type 2 diabetes in England and Wales: prospective derivation and validation of QDScore. BMJ. 2009;338:b880. doi:10.1136/bmj.b880.:b880. Lyssenko V, Jonsson A, Almgren P et al. Clinical risk factors, DNA variants, and the development of type 2 diabetes. N Engl J Med. 2008;359:2220–32. Gibson G. Rare and common variants: twenty arguments. Nat Rev Genet. 2012;13:135–45. Payami H, Kaye J, Heston LL, Bird TD, Schellenberg GD. Apolipoprotein E genotype and Alzheimer’s disease. Lancet. 1993;342:738. Cohen JC, Kiss RS, Pertsemlidis A, Marcel YL, McPherson R, Hobbs HH. Multiple rare alleles contribute to low plasma levels of HDL cholesterol. Science. 2004;305:869–72. Romeo S, Pennacchio LA, Fu Y, et al. Population-based resequencing of ANGPTL4 uncovers variations that reduce triglycerides and increase HDL. Nat Genet. 2007;39:513–6. Goldstein DB. Common genetic variation and human traits. N Engl J Med. 2009;360:1696–8.
