Introduction

Global health

Over the last 100 years, scientific innovation has led to a revolution in our understanding of the causes of chronic diseases and our ability to prevent and treat disease.1, 2, 3 This revolution has primarily improved health in industrialized countries. In developing countries, the situation is markedly different. A global health approach is needed to tackle inequalities in health between the industrialized and developing world.4, 5 A global health strategy has to take into account political, epidemiological, environmental, infrastructural and genetic aspects. The aim is to accomplish the greatest benefit for the most people over the longest period of time and that our significant, but finite, resources are used most effectively.

Type II diabetes: a global problem

Prevention of diabetes is a global investment.6, 7.In conjunction with environmental and behavioural factors, genetic susceptibility is an integral part of the pathogenesis of type II diabetes mellitus (T2D) and the metabolic syndrome.8, 9 Parallel with globalization, pronounced changes in the human environment and in human behaviour and lifestyle have resulted in escalating rates of both obesity and diabetes.10 There is evidence that adipose tissue functions not only as an energy reservoir, but also produces and secretes several cytokines that modulate energy metabolism and glucose homeostasis.11, 12, 13, 14 The identification of functionally relevant genetic polymorphisms influencing glucose homeostasis may be important in future treatment or preventative strategies.15, 16, 17 Therefore, it is becoming increasingly necessary to identify population-specific effects of novel genomic targets relevant to drug development and individualized drug action,18, 19 and ultimately the development of global health strategies.

ADIPOQ gene and its relevance to type II diabetes

In the investigation of quantitative trait loci involved in the susceptibility towards the Metabolic Syndrome (MIM 605552), it was determined that a locus on 3q27 was strongly linked to six traits fundamental to the expression of this cluster of disorders.20 These traits include body mass index (BMI), body weight, waist circumference, hip circumference, fasting plasma insulin and the ratio of glucose to insulin. The ADIPOQ (adipocyte, C1q and collagen domain containing) gene encoding adiponectin has been mapped to chromosome 3q2721 and is expressed in adult adipocyte tissue.22 Both reduced expression of adiponectin and lower serum adiponectin levels were detected in patients with obesity, T2D and coronary artery disease.23, 24, 25, 26, 27 Moreover, treatment of diabetic mice with adiponectin was shown to induce a marked improvement in insulin sensitivity.28, 29, 30, 31, 32, 33

Any alteration on a genetic level, which affects the production of the adiponectin protein, may thus result in impaired insulin action and hence result in T2D.26 Various single-nucleotide polymorphisms (SNPs) within the adiponectin gene have in turn been associated with T2D disease risk as well as increased BMI.34, 35, 36 The SNPs may induce their effects by altering the binding of transcription factors and disrupting regulatory elements, whereas the missense alterations may affect or disrupt the formation of specific isoforms as well as the higher order structure of this protein.34 In an investigation by Vasseur et al.34 in the French population, an alteration of a cytosine to a guanine at nucleotide position −11377 within the promoter region of the ADIPOQ gene was associated with lowered adiponectin levels and thus increased risk towards T2D. This alteration therefore strongly affects the correct functioning of cellular metabolism.

Population-specific metabolomic effects

In certain investigations based upon organellar genome involvement in cellular metabolism, it has been determined that various alterations within these genomes result in different phenotypes, depending on the genetic background upon which expression of the gene variant occurs.37 Furthermore, it has been determined that certain loci undergo adaptation, thus altering the metabolism of an individual in response to the external environment38 influencing the therapeutic relevance of a genetic variant. It would be intuitive that the nuclear genome is undergoing similar effects, and thus different populations must have altered basal metabolic rates due to evolutionary adaptation brought about by various environmental factors. Future metabolomic research should therefore undertake the elucidation of population-based metabolic variation.39 Current knowledge about the aetiology and pathogenesis of T2D has led to the description of a multifactorial model for disease susceptibility with strong genetic and environmental components,40, 41 and, therefore, it is a prime candidate for these types of population-specific effects. It, thus, becomes necessary to investigate the association of described alterations such as C-11377G with T2D, in various populations.

Owing to the fact that the greatest increase in affected individuals will occur in the so-called developing countries,42 it is necessary to investigate the genetic structures that increase T2D susceptibility in these populations. Thus, in this investigation, a meta-analysis was undertaken by comparing a genetic risk factor at the ADIPOQ locus between African and Cuban cohorts, and in turn comparing this to a cohort from a developed country, that is, Germany. This investigation underscores the importance of using a population-specific case–control approach for the comparative investigation of genetic susceptibility loci in T2D, and possibly in all metabolomic endeavours.

Results

The basic clinical data of the cohorts investigated are presented in Table 1. As depicted in Table 1, the Cuban and German cohorts presented with much lower BMI levels than the black South African cohorts. It has been determined that the majority of female diabetic patients in the South African population tends to present with high BMI values.43 However, the control group presented with a similar overall BMI measurements, thus negating the need for possible stratification with regard to this criterion. The association of high BMI and T2D is not present in the Cuban cohort investigated as both the diabetic and control cohorts presented with similar BMI measurements, thus negating the effect this has on disease risk in this population.

Table 1 Basic clinical data of the individuals included in the study (mean±s.d.)
Full size table

Both the Cuban and black South African cohorts were similarly investigated, whereas the German cohort has been extracted from a similar study undertaken in the German population.44 No significant difference in genotype frequencies was present between the South African diabetic and control cohorts upon χ2 analysis, as presented in Table 2. A significant difference was determined upon comparison of the genotype distributions of the two cohorts at the C-11377G locus (P-value <0.00) within the Cuban cohort investigated, as presented in Table 2. The homozygote of the variant allele was in association with increased risk as evidenced by the high odds ratio (OR) value calculated (OR=2.54, 95% confidence interval (95% CI): 1.00–6.46) as highlighted in Table 2. Conversely, the 1,2 (heterozygous) genotype was associated with a protective factor due to the calculated OR (OR=0.68, 95% CI: 0.48–0.96) as depicted in the lighter shaded block within Table 2. The German cohort presents a significant difference in its genotype distribution; however, the source of this variation is not easily determined upon analysis of the OR values.

Table 2 χ2-Analysis for the comparison of the genotype distribution at the C-11377Ga locus in the three populations under investigation
Full size table

Discussion

The fact that there is no significant difference between the black South African control and diabetic cohorts allows for the hypothesis that C-11377G is not a significant risk factor within the South African population. The low OR (OR=0.19, 95% CI: 0.02–1.68) for the homozygote of the variant allele may be indicative of a possible association of this genotype with a protective factor against T2D susceptibility. However, as it is at a relatively low frequency within these cohorts (frequency=0.00 and 0.02 in the diabetic and control cohorts, respectively), it may ultimately be determined not to be a significant factor in mitigating disease risk.45, 46

Alternatively, the association between the 2,2 (homozygous genotype of the risk allele) homozygote and increased risk towards T2D alludes to the possible role of this alteration within the Cuban cohort. However, due to the fact that the 95% CI does include unity, it is possible that this association is spurious. The relatively wide interval may be due to the fact that the homozygote for the variant allele is relatively rare in the cohorts investigated (frequency=0.06 and 0.02 in the patient and control cohorts, respectively). It will be useful in the future to investigate a larger cohort to narrow the interval.

The association of the heterozygous genotype with a protective effect towards T2D is a surprising result but is by no means unique. The higher levels of this genotype in the control cohort may be due to heterozygote advantage caused by its association with a protective factor. There are numerous examples of heterozygote advantage such as the protection that an individual heterozygous at the locus for sickle cell anaemia has against malaria.47 Functional analysis is required in the future to determine the exact nature of this association.

Meta-analysis of the ADIPOQ C-11377G variant

Meta-analytical investigation of the C-11377G locus within the black South African, Cuban and German populations resulted in the elucidation of a possible global association of the 1 allele with protection against T2D, in the context of fixed effects as presented in Table 3. This is due to the fact that under the dominant model, the OR of 0.64 (95% CI: 0.40–1.00) indicates that individuals harbouring the 1,1 (homozygous genotype of the wild-type allele) and 1,2 genotypes may be protected against disease risk as compared with homozygotes of the variant allele. A similar trend is differentiable upon comparison of the 1,1 and 1,2 genotypes separately to the 2,2 genotype as indicated by the lighter shaded blocks.

Table 3 Meta-analysis of genotypes at the C-11377Ga locus for the black South African, Cuban and German cohorts
Full size table

The most significant result from this analysis is the fact that this alteration seems to have different effects in the different populations. In a reported French Caucasian population, the variant allele was associated with decreased levels of adiponectin production.34 In the Cuban cohort presented in this investigation, the heterozygous genotype is associated with a protective effect, whereas the black South African cohort did not present with any significant association to disease risk. A possible explanation for this variable pattern of disease risk is the fact that these populations have experienced very different environments during their evolution.48 It has been determined that adaptation affects the functioning of metabolism by natural selection within the genome of a cytoplasmic organelle.38 Similar effects must be at play within the nuclear environment, which alters the profile of disease risk of a specific alteration dependent upon the genomic background upon which an alteration is expressed.37 It is therefore essential that the ethnicity of an affected individual be taken into consideration before a treatment strategy can be implemented.

Limitations of the study

The protective effect determined upon meta-analysis is interesting, because it is evident that the 1,1 and 1,2 genotypes have a similar effect, that is, there is no increased protection associated with harbouring two 1 alleles as compared with harbouring one allele. Caution must, however, be exercised with regard to any conclusions derived from this analysis as the 95% CI for all the OR values discussed do span unity. It is possible therefore that this association is spurious. However, the fact that similar effects were determined under the dominant model as determined under the two additive models comparing the 2,2 genotype to the 1,1 and 1,2 genotypes separately indicates that this association is unlikely to be a Type I error.

This association is, however, not present under the random-effects model indicating that the interpopulation variance has an effect on the association of this alteration and disease risk. As discussed under the previous heading, this is the most likely scenario, which strengthens the hypothesis that there are strong population-specific effects affecting disease pathogenesis, which cannot be ignored. χ2-analysis for heterogeneity did not result in the determination of any significant interpopulation variability in the genotype distributions, which further indicates that population-specific effects are responsible. Future studies should be undertaken to investigate the functional role of this polymorphism in these specific populations to understand its molecular pathogenesis in disease susceptibility.

Ethnicity and disease risk

The significance of ethnicity is an important, but often overlooked factor in disease risk. In future, the use of systems biological approaches will be crucial for the investigation of disease. This investigation presents evidence that ethnicity is a major source of extraneous variability in the effects of a single alteration. The effect of it on a metabolomic approach wherein it has not been corrected for will thus be extremely large. Development of standardization procedures is thus integral to the future success of most future biological endeavours.

In previous investigations,34 it has been determined that the alterations within the promoter region of the adiponectin gene are associated with disease progression. Lower adiponectin levels are associated with increased risk towards T2D; however, as determined in the Cuban cohort investigated, this locus was associated with protection towards T2D. Protection would imply that there is slower disease progression or ideally no disease development at all. Thus, it would seem that an alteration may have differential functional effects dependent on the ethnic group investigated.

The means by which this may occur can be explained by discussing a fundamental assumption in the use of a meta-analytical investigation. In this investigation, the individuals involved were selected according to diabetic status and tested for association to the C-11377G locus. To be able to make this comparison, however, it is necessary to assume that the populations are exposed to similar environments or at the very least that the alteration has a similar effect irrespective of the environmental influences, to ignore the effects of epistasis. However, this cannot be assumed within this investigation as we know that these different populations have undergone adaptation as evolution has occurred.38 Thus ultimately, the use of a retrospective case–control strategy may not be most appropriate for the elucidation of complex disease risk, although it is often the initial method used.

Recently, we have shown that adiponectin may have far-reaching implications in diabetes management.26, 49 Therefore, its population-specific relevance needs to be elucidated. It may therefore be necessary that the effects of numerous genes as well as environmental factors are taken into consideration. Furthermore, the use of a case–control study design is no longer appropriate. It is thus necessary to use a method that records disease progression within a population that is well described at the clinical and environmental levels. Thus, a transdisciplinary prospective study in all three environments would be the most effective strategy to investigate the role of genetic susceptibility in disease progression, as it will be possible to differentiate the size of certain environmental and epistatic contributions towards disease risk. By understanding the effect of population-specific factors on genomics, metabolomics and disease risk, it will assist greatly in the development of global health initiatives.

Materials and methods

Subjects and methods

Four hundred and fifty-three individuals from the black South African population were included in this investigation. Individuals were, however, from linguistically heterogeneous backgrounds. Owing to the current paucity of data linking linguistic background to genetic ancestry in the black South African population, it was not possible to substructure the study cohorts accordingly. The cohorts investigated consisted mainly of individuals from the Tswana and Northern Sotho language groups.

The black South African patient (sample size, n=227) and control (n=226) cohorts used in this investigation have been enrolled from various diabetic and outpatient clinics within the Gauteng and Northwest provinces of South Africa. All patients were ascertained to have been diagnosed via the criteria stipulated by the WHO (World Health Organization)50 consortium through investigation of the individual patient records. A random glucose reading was taken of each control individual at the initiation of the consultation. Control individuals were enrolled according to the following exclusion criteria: patients undergoing treatment with any known antidiabetic drug or a random plasma glucose level of greater than 7 mmol l−1.

Five hundred and seventy-five individuals from the Cuban population were enrolled. The diabetic (n=324) and control individuals (n=251) were enrolled according to the criteria stipulated in the previous paragraph. No specific ethnic group was preferentially selected for in either cohort.

Within the German cohort, 365 unrelated subjects with T2D and 323 unrelated control individuals, confirmed to be nondiabetic, were enrolled in various centres in Saxony, Germany as described in Schwarz et al.7 Therefore, a total of six hundred and eighty-eight German Caucasian individuals were included.

This research programme has been approved by the Ethics Committees of the North-West University (Potchefstroom Campus), the University of Pretoria and the Faculty of Medicine Carl Gustav Carus, Technical University Dresden. Written informed consent was obtained for each subject prior to their inclusion in the project. This was followed by the collection of blood samples for biochemical analyses, extraction of genomic DNA and genetic analysis.

Genetic analysis

Genetic screening was achieved via the use of a real-time PCR strategy employing the use of the LightCycler (LC) Real Time PCR machine (Roche Molecular Biochemicals, Indianapolis, IN, USA). The various genotypes were elucidated through the use of hybridization probe technology and melting curve analyses. The LC reaction consisted of the following components: 10 pmol of each of the forward and reverse primers (presented in Table 4); 3 pmol of the anchor probe (presented in Table 4); 1.5 pmol of the sensor probe (presented in Table 4), 3.75 μM MgCl2 and 2 μl of the FastStart Reaction Mix Hybridisation Probes (10 × reaction buffer) containing the LC-FastStart enzyme. Amplification was achieved via the following thermal cycling conditions: an initial denaturation at 95 °C for 10 min was followed by 36 cycles consisting of a denaturation step at 95 °C for 10 s, an annealing step at 58 °C for 7 s and an elongation step at 72 °C for 9 s.

Table 4 Primers and probes used in the amplification and detection of the C-11377G alteration in the adiponectin gene
Full size table

Melting curve analysis was achieved by decreasing the temperature to 30 °C, which was increased by 0.5 °C s−1 up to a final temperature of 90 °C. Fluorescence was detected continuously throughout this process and was plotted graphically as fluorescence vs temperature by the LightCycler software. The melting peak of each SNP was detected by determining the negative derivative of the fluorescence level and plotting it against temperature (−d(F)/d(T) vs T).

Population-specific analysis

Association of a specific genetic structure was analysed initially by determining if the populations under investigation were in Hardy–Weinberg equilibrium. Subsequently, comparison of the allele and haplotype frequencies between the patient and control populations was achieved using contingency table analysis. OR was calculated via Woolf's method.51

Global associations of specific genotypes were calculated via the use of a meta-analysis of the black South African, Cuban and a reported German population. These investigations were combined under two separate models, namely that of fixed effects using the Mantel–Haenszel method52 and random effects via the method described by DerSimonian and Laird.53 Furthermore, within each model, five possible types of genotypic association were investigated namely a recessive model (1,1 vs 1,2 and 2,2), a dominant model (1,1 and 1,2 vs 2,2) and three different additive models (1,1 vs 1,2; 1,2 vs 2,2; and 1,1 vs 2,2). Calculations were performed using the Review Manager 4.2 software.54