| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Editorial |
Pediatric Endocrinology and Institut National de la Santé et de la Recherche Médicale Unite 561 Hôpital Saint-Vincent de Paul-Cochin University Paris V René Descartes 75014 Paris, France
Address all correspondence and requests for reprints to: Pierre Bougnères, M.D., Ph.D., Pediatric Endocrinology and INSERM U561, Hôpital Saint-Vincent de Paul, University Paris V René Descartes, 75014 Paris, France.
In humans, the region 5' flanking the insulin gene contains a variable number of tandem repeats (INS-VNTR), also called a minisatellite, made of 26–209 G-rich repeats whose number and sequence show extensive variation, generating more than 475 different VNTR alleles in the European population. Transfection of a handful of these alleles in fetal islets (1) or ß-cell lines (2) suggested that the VNTR polymorphism modulates the transcription of the insulin gene, possibly through differences in the ability of VNTR alleles to bind the transcription factor Pur1 (1). Class I (short) alleles seemed to confer more transcriptional activation than class III (long) alleles. Because the insulin gene is only expressed in the ß-cells and the thymus, the INS-VNTR is expected to have primary functional effects only in these two cell types. ß-Cells produce insulin to regulate metabolism and growth, and thymocytes produce insulin to make the immune system tolerant to their own ß-cells. For both functions, the quantity of insulin produced is critical. The study of a few autopsy specimens of adult or fetal pancreas (3, 4) and thymus (5) suggested that class I VNTR alleles are associated with more insulin gene transcripts in the pancreas and less transcripts in the thymus.
Human genetic studies found that the INS-VNTR associates with susceptibility to several common diseases (type 1 and type 2 diabetes, obesity, polycystic ovaries) (6, 7, 8, 9, 10) and with the variability of quantitative traits (QT) (birth size, insulin levels, adiposity in infancy or adolescence) (11, 12, 13, 14), the common factor of these diseases and traits being insulin secretion. These observations in given populations were not always replicated in following studies (15, 16). Several possibilities can account for the association of the INS-VNTR with a phenotype.
An Artifact?
A problem endangering all association studies is false positivity created when a cohort is demographically stratified. If a subgroup of a cohort bears VNTR or HLA genotypes different from other subjects and eats more carbohydrates, the whole cohort could show a misleading statistical association of insulin secretion with VNTR or HLA polymorphisms. What is true for continuous traits is true for susceptibility to type 1 or type 2 diabetes if case and/or control populations are recent admixtures of people of European, African, and Asian ancestries (8). Self-identified race/ethnicity of participants (17), grandparents’ place of birth (not in the United States), patronymics, or better than all, use of ethnically sensitive genomic markers (genomic control) (17) help safeguard association studies from stratification. Approaches evaluating the transmission of VNTR alleles from parents to offspring affected by a disease [transmission disequilibrium test (TDT)] or having a given value of the studied trait (quantitative TDT) escape stratification problems.
Simply a Regional Marker?
Even if they seem associated with a phenotype, VNTR alleles can be no more than neutral genomic markers linked to functional polymorphisms that are the true actors of the association (the VNTR alleles are said to be in ‘linkage disequilibrium’ with the functional polymorphisms if their association on the same haplotype is nonrandom). Because people inherit the functional polymorphic alleles together with the VNTR alleles, the effect appears to be linked to the VNTR, but the true actors are close by, neighboring the INS-VNTR at a short distance, like the –23Hph1 A/T SNP, or nearby, within the flanking regions of the insulin gene, possibly relatively far from the VNTR or within the neighboring IGF-2.
True Actor of Associations
The more prevalent possibility nowadays is that the INS-VNTR itself affects insulin gene transcription in ß-cells and thymus, as suggested by in vitro experiments. Through transcription, the VNTR could affect insulin biosynthesis. A difficulty is that multiple environmental, time-dependent, and polygenic factors also influence the ß-cells and the thymocytes. Another problem is that insulin production by ß-cells or, even more difficult, by thymocytes of infants (the age for acquiring tolerance to self insulin) cannot be measured in vivo. Effects of INS-VNTR polymorphism can thus only be tested on more complex traits that are more distant from insulin gene expression (Fig. 1
).
|
The concentration of insulin in the peripheral venous circulation seems to be the in vivo measurable QT that is the more proximal to insulin gene transcription in ß-cells and thus the more likely to be influenced by INS-VNTR (Fig. 1
).
How much of the variance of fasting insulin could be attributed to the INS-VNTR? The answer stems from the following equation: variance of insulin levels = genetic variance + environmental variance + interactions. Genetics seems to account for 25% of fasting insulin variance in adults (18). Based on views prevailing among quantitative geneticists, the larger effect size for a single gene variant is not expected to exceed 2–10% of the genetic variance of a complex trait (19). Remember, after all, that the genetic variation of phenotypes has not been designed by evolution to allow large individual contrasts. If the VNTR is responsible at the most for 10% of the genetic variance of fasting insulin (<5% of general variance?), genotypic effects of the VNTR are not easy to demonstrate. It is likely for example, that the detection of an INS-VNTR effect would be difficult if not impossible in situations where both the mean value and the variance of insulin production are so small [for example, in slim healthy subjects (20)] that documenting VNTR effects would require tens of thousands of people. If the simple math is forgotten, association studies will be underpowered and carry a high risk of false negativity, sometimes called "lack of replication."
Physiological Caveats
Nongenetic covariables influence variations of circulating insulin, including duration of fasting, diet, and physical activity for the preceding days, with additional confounding factors such as ovulatory cycle, subject’s age, and metabolic history. When measured several times in well-controlled conditions in the same individual, insulin shows reasonably reproducible values after overnight fasting and less reproducible but consistent values after glucose ingestion (21). In that sense, insulin concentration is a QT, like height or weight, that characterizes an individual. But if insulin concentration is measured in uncontrolled physiological conditions, a background erratic "noise" will partly obscure the search for genetic effects. For pragmatic purposes, subjects are often sampled on the "battlefield" of real life, far from controlled conditions. The investigators are thus forced to bet that the effect size of the VNTR will overcome the noise. But because metabolic phenotypes are versatile, I think that more physiological control should be applied when phenotypes are to be quantified for genetic analyses. Reading the Subjects section of published association studies usually reveals the physiological quality of the phenotypic data.
Can we rely instead on direct measurements of insulin gene expression in ß-cells or thymus to know more about quantitative effects of the INS-VNTR on insulin production? The answer would certainly be "yes" if ß-cells or thymus could be collected in controlled conditions (sampling of the insulin-producing tissues does not eliminate physiological noise) and in a sufficient number of individuals to allow the rules of quantitative genetics to be reliably applied.
Complex Traits and Metabolic Disorders
Beyond insulin levels, which traits could be influenced by the INS-VNTR? As a general rule, the more complex the studied trait is, the more difficult will be the demonstration of genotypic effects at a single locus (Fig. 1). Santoro et al. (22) have studied insulin levels and metabolic traits in juvenile obesity, a clinical situation where insulin secretion is overstimulated, allowing evaluation of individual differences in amplified conditions (although not reflecting normal physiology). Adult obesity also stimulates insulin secretion through many mechanisms including insulin resistance, but can also alter, on the other side, insulin production because of long-term deleterious effects of the obesity status and aging on ß-cell function.
The results by Santoro et al. (22) are consistent with observations that class I VNTR alleles associate with greater insulin levels in juvenile obese Europeans (12, 21). The authors extend this observation to metabolic syndrome parameters. Whether or not the metabolic syndrome deserves individualization as an entity in juvenile obese remains to be debated, but it is clear that the parameters studied by Santoro et al. (22) are all insulin-dependent, and as such are candidates for undergoing the genetic influence of the INS-VNTR. Given the biology of the VNTR, its influence on metabolism starts primarily at the level of ß-cells, not of peripheral insulin resistance. For all the above reasons, I am not that much troubled a priori by the apparent discordance of findings reporting no association of INS-VNTR with insulin, metabolic syndrome, or other traits in 5646 individuals from northern Finland who were 31 yr of age and had a mean body mass index of 25 kg/m2 (healthy lean adults with a narrow distribution of QTs, and belonging to a genetically peculiar population) (15).
Type 2 Diabetes
The genetic story of the polymorphism in the 5' flanking region of the insulin gene started 25 yr ago at Washington University with the pioneering observations of Rotwein et al. (8) who described an association of what was then called "the insulin long polymorphic region (ILPR)" with type 2 diabetes. Since then, studies have or have not supported this association. This may not be entirely surprising if one considers type 2 diabetes to be a heterogeneous and complex combination of traits resulting in hyperglycemia. Among these traits, insufficient insulin secretion appears to play a role that validates the hypothesis of the INS-VNTR involvement. But the relative contribution of insufficient insulin secretion to the pathogenesis of type 2 diabetes is not consistent across all diabetic populations and so is expected to be the potential contribution of the INS-VNTR. An additional problem is again the demographic and genetic structure of the studied samples. Meta-analyses have become popular, and replications in various human groups are often required by referees and the scientific community. But the possibility that specific genetic influences are important in certain polygenic, environmental, and cultural contexts, not in others, should not be disregarded; lack of replication does not always signify lack of effect.
VNTR Diversity in Human Populations
The Jeffreys’ group (23), who provided us with invaluable information about the INS-VNTR polymorphism, developed a high-resolution system for analyzing variant repeat distributions applicable to all known VNTR alleles. This system revealed extremely low structural diversity in the minisatellite among English citizens, with all alleles belonging to one of only three highly diverged lineages called "I," "IIIA," and "IIIB." In contrast, the difference between levels of diversity in Africans and non-Africans was unusually large, with all 22 lineages identified in Africa compared with only three lineages seen in non-African populations (23). Stead and Jeffreys (23) also find evidence for overrepresentation of lineage I chromosomes in non-Africans. These data are consistent with a common out-of-Africa origin and an unusually tight bottleneck within the ancestry of all non-African populations, possibly combined with differential and positive selection for lineage I alleles in non-Africans. These data possibly have important implications for future disease-association or trait-association studies. VNTR polymorphism should no longer be defined by the linkage disequilibrium with the neighboring –23Hph1A/T single nucleotide polymorphism, as is done in more than 95% of published studies. The INS-VNTR has an extraordinary degree of molecular and physical diversity and complexity, with more than 475 different alleles. Using direct sequencing of VNTR alleles, we found that the association of class I alleles with insulin levels previously reported in French obese children (12) was in fact due to a limited lineage subgroup of class I alleles called "ID" (24). These alleles were not found in obese West African children, and therefore this specific association of VNTR alleles with insulin levels cannot be seen in such populations. What about "ID" INS-VNTR alleles in Finland, the United States, Asia, etc.?
We now need to revisit associations by digging into the genetic, evolutionary, and functional complexity of the INS-VNTR locus and into its interactions with other genetic and nutritional factors involved in the variance of insulin-related phenotypes.
Acknowledgments
I apologize to the colleagues whose important work I have not been able to quote because of space limitations. I thank my INS-VNTR co-workers Sophie Le Fur and Catherine Le Stunff.
Footnotes
Abbreviations: QT, Quantitative traits; VNTR, variable number of tandem repeats.
Received August 9, 2006.
Accepted August 18, 2006.
References
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Endocrinology | Endocrine Reviews | J. Clin. End. & Metab. |
| Molecular Endocrinology | Recent Prog. Horm. Res. | All Endocrine Journals |
