NDT Advance Access originally published online on January 8, 2007
Nephrology Dialysis Transplantation 2007 22(3):687-689; doi:10.1093/ndt/gfl815
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Lumping, splitting and mapping: assessing linkage in different ethnic groups for albuminuria and glomerular filtration rate in the HyperGen study
Division of Cardiovascular Medicine, Center for Human Genetics Research, Vanderbilt University, Nashville, TN 37232, USA
Correspondence and offprint requests to: Scott M. Williams, Division of Cardiovascular Medicine, Center for Human Genetics Research, Vanderbilt University, Nashville, TN 37232, USA. Email: smwillams{at}cghr.mc.vanderbilt.edu
Keywords: albuminuria; ethnic differences; genetics; glomerular filtration rate
The ability to map genes that predispose to complex disease has been an ongoing area of research for more than a decade, but results have often been less than satisfying in terms of discovering genes of major importance. One likely reason is that the phenotypes used in many linkage analyses have heterogeneous genetic causes, so that different loci may confer susceptibility in different families. Such a situation can lead to the inability to identify loci in studies, or alternatively, loci found in one study may not replicate in others. An example of this kind of situation is found in the genetic study of hypertension, where multiple genome screens have identified a variety of loci, several of which do not replicate [1,2]. How can investigators deal with such heterogeneity?
In their article in the current issue of NDT, Leon et al. [3] use two potentially important methods to help address the problem. They report on linkage mapping for two traits that may be related to each other and to hypertension: albuminuria (measured as albumin to creatinine ratio, ACR) and glomerular filtration rate (GFR) in two large family-based samples from the HyperGEN cohort. One sample is African-American (AA) and the other is European-American (EA). This study is important because of the approaches and because of the large size of each of the samples (1251 AA and 1129 EA). Another interesting feature of this study is that it attempted to identify loci that affect both traits simultaneously. This last feature is interesting in that it tests for pleiotropy, a type of genetic variation rarely assessed in linkage or association studies.
First, Leon et al. [3] chose presumably more homogenous populations in terms of geographic origin and with presumably more similar genetic risk factors. Such analyses involve defining ethnicity prior to analyses and analysing groups by this criterion. However, this approach may still fail because even within a homogeneous population, the complexity of the factors that predispose to disease may be dependent on a variety of factors that are left unmeasured. These unmeasured factors can provide the critical context in which the susceptibility genes function; context being defined broadly as variations in other genes and/or environmental parameters. Failure to assess context may make results difficult to reproduce across studies [4–6]. Second, they mapped presumably more homogenous phenotypes that underlie at least some of the risk of clinical disease. It is possible that this approach will be better at identifying loci that can replicate across studies, because the loci identified are likely to be closer to underlying gene action.
 |
Splitting and lumping by geographic origin |
|---|
 Top Splitting and lumping by... Lumping of phenotypic categoriesReferences |
|---|
In linkage analyses it is common practice to pool all samples collected, regardless of ethnicity. This is done because of the need to increase power by increasing sample size. This approach is understandable when sample sizes are limited, but in the case of Leon et al. [3] the sample size of both AA and EA cohorts is large enough that it is possible to analyse the cohorts separately (splitting), and they do so. Of importance is the fact that they identify linkage signals in both of the cohorts, and the signals do not overlap. For GFR, evidence for linkage in AA was found for chromosomes 7, 14 and 19. Similar regions on the same chromosomes have been observed in previous linkage results for hypertension, although the previous results are for samples of different ethnic origins (reviewed in [2]). No significant findings were obtained for GFR in EA. For ACR several interesting signals were observed for both AA (chromosomes 8, 16 and 17) and EA (chromosomes 18 and 19). A previous analysis of some HyperGEN samples, as well as Family Blood Pressure Program samples, also detected linkage at the same site on chromosome 19 for ACR in hypertensive families [7]. It is of note that there was no overlap between AA and EA signals. This reinforces the potential importance of assessing these groups independently.
The results of Leon et al. can also be compared with recent work by Krolewski et al. [8] and Turner et al. [9], who assessed linkage in traits related to kidney disease. Krolewski et al. studied ACR, and detected a linkage peak on chromosome 7. Turner et al. examined both estimated GFR and urine ACR in both AA and EA, using the GENOA cohort, with sample sizes comparable with that of Leon et al. They found a linkage signal for eGFR in AA on chromosome 7, in a location close to that reported by Leon et al. for ACR. Additionally, Turner et al. found a peak for UACR on chromosome 7, but not near the locus found by Leon et al. As with the results of Leon et al., the results of Turner et al. indicate different linkages in AA and EA.
In an alternative analysis, Leon et al. pooled the AA and EA samples and did another linkage analysis. In the combined analysis, additional signals to those found for each group separately were detected, for ACR (chromosome 19) and for GFR (chromosomes 14, 15 and 16). Although the chromosome 16 and 19 results are the same as reported in the separate analysis, the other two chromosomes are not found in either of the samples alone. Therefore, although there are likely to be some loci that differ among groups, there is also evidence for common signals that can only be found in the larger analyses. Therefore, by both splitting and lumping, different potentially important linkage signals can be detected.
|
Lumping of phenotypic categories |
|---|
|
TopSplitting and lumping by...Lumping of phenotypic categoriesReferences |
|---|
In the above descriptions, the two phenotypes were analysed separately (univariate). In a separate bivariate analysis, Leon et al. looked for loci that simultaneously affect both traits. This approach was also used by Turner et al. [9]. Loci that are identified in this way are said to exhibit pleiotropy, or multiple effects of a single gene. In this analysis, perhaps the locus of the most significance is chromosome 7. This region of chromosome 7 was identified in the univariate analyses for GFR in AA and for UACR in AA by Turner et al. and for GFR in AA by Leon et al. [3]. No loci were identified that were not found in at least one of the univariate analyses. These results indicate that this approach will be fruitful in detecting loci for genes that affect separate but related phenotypes. Such an approach may become commonplace as large datasets, such as the HyperGEN study, become available.
However, the bivariate analysis of these two traits poses some interesting limitations/opportunities. Although as noted by Leon et al., GFR and ACR may be negatively correlated, this is an oversimplification of the relationship between the two phenotypes. Specifically, it has been shown that although increased urinary albumin excretion (UAE) poses a risk for altered GFR, the nature of this alteration in GFR is not simple [10,11]. Specifically, small increases in UAE increase risk of both increased AND decreased GFR [11]. A likely scenario is that initial, small increases in UAE initially increases GFR and then over time leads to decreased GFR. If this is the case, then an analysis as performed runs the risk of confounding these two GFR outcomes [11]. Therefore, although similar genetic pathways may (or may not) lead to the two phenotypes, the relationship may be based on age or time after an increase in ACR. Interestingly, although this was not accounted for in the study of Leon et al., they still detected significant evidence for loci affecting both traits, likely justifying this exploratory analysis.
The report of Leon et al. provides an excellent example of how large scale family-based studies can be used to map putative loci for complex, but related phenotypes. Their study had several important features: first, the size of both the AA and EA cohorts. Clearly, size does matter in linkage and this study was among the largest so far reported. Second, they examined more than one phenotype related to clinical disease, but ones that are more likely to have a simpler relationship to the genes. Lastly, Leon et al. performed a series of analyses that used the entire dataset and appropriate subdivisions. This last feature may be very important in understanding ethnic disparities in common disease, as well as identifying common genetic risk factors across populations.
Conflict of interest statement. None declared.
(See related article by Leon et al. Genome scan of glomerular filtration rate and albuminuria: the HyperGEN study. Nephrol Dial Transplant 2007; 22: 763–771.)
|
References |
|---|
|
TopSplitting and lumping by...Lumping of phenotypic categoriesReferences |
|---|
- Lalouel JM. (2003) Large-scale search for genes predisposing to essential hypertension. Am J Hypertens 16:163–166.[CrossRef][Web of Science][Medline]
- Samani NJ. (2003) Genome scans for hypertension and blood pressure regulation. Am J Hypertens 16:167–171.[CrossRef][Web of Science][Medline]
- Leon JM, Freedman BI, Miller MB, et al. (2006) Genome scan of glomerular filtration rate and albuminuria: The HyperGen study. Nephrol Dial Transplant.
- Kardia SL, Rozek LS, Krushkal J, et al. (2003) Genome-wide linkage analyses for hypertension genes in two ethnically and geographically diverse populations. Am J Hypertens 16:154–157.[CrossRef][Web of Science][Medline]
- Moore JH and Williams SM. (2002) New strategies for identifying gene-gene interactions in hypertension. Ann Med 34:88–95.[CrossRef][Web of Science][Medline]
- Williams SM, Haines JL, Moore JH. (2004) The use of animal models in the study of complex disease: All else is never equal or why do so many human studies fail to replicate animal findings? Bioessays 26:170–179.[CrossRef][Web of Science][Medline]
- Freedman BI, Beck SR, Rich SS, et al. (2003) A genome-wide scan for urinary albumin excretion in hypertensive families. Hypertension 42:291–296.
[Abstract/Free Full Text] - Krolewski AS, Poznik GD, Placha G, et al. (2006) A genome-wide linkage scan for genes controlling variation in urinary albumin excretion in type II diabetes. Kidney Int 69:129–136.[CrossRef][Web of Science][Medline]
- Turner ST, Kardia SL, Mosley TH, Rule AD, Boerwinkle E, de AM. (2006) Influence of genomic loci on measures of chronic kidney disease in hypertensive sibships. J Am Soc Nephrol 17:2048–2055.
[Abstract/Free Full Text] - Pinto-Sietsma SJ, Janssen WM, Hillege HL, Navis G, de ZD, de Jong PE. (2000) Urinary albumin excretion is associated with renal functional abnormalities in a nondiabetic population. J Am Soc Nephrol 11:1882–1888.
[Abstract/Free Full Text] - Verhave JC, Gansevoort RT, Hillege HL, Bakker SJ, de ZD, de Jong PE. (2004) An elevated urinary albumin excretion predicts de novo development of renal function impairment in the general population. Kidney Int Suppl 66:S18–S21.
Accepted in revised form: 14.12.06
CiteULike 
Connotea 
Del.icio.us What's this?
Related articles in NDT:
- Genome scan of glomerular filtration rate and albuminuria: the HyperGEN study
- Joanlise M. Leon, Barry I. Freedman, Michael B. Miller, Kari E. North, Steven C. Hunt, John H. Eckfeldt, Cora E. Lewis, Aldi T. Kraja, Luc Djoussé, and Donna K. Arnett
NDT 2007 22: 763-771.[Abstract] [FREE Full Text]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||