Patent Application
20140057800
The present invention pertains to different genetic markers of importance to the molecular mechanism involved in insulin resistance. A number of SNPs (single nucleotide polymorphisms) that are associated with insulin resistance have been located in the gene vesicle associated membrane protein-associated protein A (VAPA). Individual responses to a dietary challenge are expected to vary among individuals. Individuals with either a weak or strong response in insulin resistance upon dietary changes in glycemic load showed distinct genotype profiles. These markers have been extensively screened and connections between variation in SNPs and changes in insulin resistance in response to diets with different glycemic load have been identified.
An association between genetic variability in VAPA and insulin resistance has been found where several specific SNPs on identified quantitative trait loci (QTLs) are pinpointed.
Susceptibility loci traits for insulin resistance and SNPs which are involved in the molecular mechanism of the VAPA genetic interactions with insulin resistance have been identified. The protein encoded by this gene is a type IV membrane protein. It is present in the plasma membrane and intracellular vesicles. It may also be associated with the cytoskeleton. This protein may function in vesicle trafficking, membrane fusion, protein complex assembly and cell motility. Alternative splicing occurs at this locus and two transcript variants encoding distinct isoforms have been identified.
One aspect of the present invention is directed to specific SNPs as new markers of candidate QTLs related to genetic aspects of developing insulin resistance. Another aspect of the present invention involves the use of VAPA and plasma protein inhibitor of activated STAT-1 (PIAS1) as candidate genes for molecular mechanisms involved in insulin resistance. Yet another aspect of the present invention involves a specific marker SNP in the GIP (gastric inhibitory polypeptide) gene, a candidate expressional QTL (eQTL) affecting plasma plasminogen activator inhibitor-1 (PAI-1) concentrations related to insulin resistance.
The identified genetic markers can be used in the diagnosis of insulin resistance correlated with dietary diseases, especially glycemic loads. Furthermore such markers can be used in developing suitable drugs for regulating glycemic response in people with such diseases.
Furthermore, such markers associated with insulin resistance can be used to explain individual physiological responses to dietary glycemic load. SNP typing can be used to provide concrete dietary advice to persons genetically predisposed to type II diabetes (T2D).
Type 2 diabetes (T2D) is defined as chronic hyperglycemia, manifested when insulin production is overwhelmed by insulin resistance in target cells, leading to a decreased ability of glucose uptake (Tripathy and Chavez, Curr Diab Rep, 2010, 10(3): pp. 184-91, incorporated herein by reference). Insulin resistance, however, precedes the onset of T2D by many years (Pagel-Langenickel et al., Endocr Rev, 2010, 31(1): pp. 25-51, incorporated herein by reference), and in addition to be a risk factor for T2D it is also an independent predictor for e.g. hypertension, coronary heart disease (CHD), stroke, and cancer (Facchini et al., J Clin Endocrinol Metab, 2001, 86(8): pp. 3574-8, incorporated herein by reference). Even though obesity is associated with increased insulin resistance, individuals of normal weight do also experience variable sensitivity to insulin (McLaughlin et al., Metabolism, 2004, 53(4): pp. 495-9, incorporated herein by reference).
Already 30 years ago it was stated that the prevalence of T2D could be reduced by lifestyle changes, but so far the incidence of T2D has only been increasing, and the expansion is now called a modern epidemic (Meigs, Diabetes Care, 2010, 33(8): pp. 1865-71, incorporated herein by reference). There are at least two plausible explanations for this: Firstly, the dietary guidelines may be underestimating the influence of dietary glycemic load on hyperinsulinemia (Ludwig, Jama, 2002, 287(18): pp. 2414-23, incorporated herein by reference). Secondly, the same guidelines may be too general. The capability to study the complex genetics behind interindividual metabolic differences (Lairon et al., Public Health Nutr, 2009, 12(9A): pp. 1601-6, incorporated herein by reference) has been developed only recently, revealing benefits of personalized nutrition among high-risk persons (Kaput, J., Curr Opin Biotechnol, 2008, 19(2): pp. 110-20; Martinez et al., Asia Pac J Clin Nutr, 2008, 17 Suppl 1: p. 119-22; both incorporated herein by reference).
Insulin resistance is a pathophysiological trait characterised by an aberrant blood lipid profile, endothelial dysfunction, increased plasma concentration of procoagulant factors, and markers of inflammation (Goldberg, R. B., J Clin Endocrinol Metab, 2009, 94(9): pp. 3171-82, incorporated herein by reference). The etiology of insulin resistance is complex and unlikely to be the same in every individual. A major determinant, though, seems to be cytokine induced activation of proinflammatory pathways in insulin target cells, reducing insulin sensitivity. This activates and attracts immune cells, and establishes a feed forward loop resulting in macrophage infiltration of the tissue, and additional cytokine secretion (Olefsky and Glass, Annu Rev Physiol, 2010, 72: pp. 219-46, incorporated herein by reference). The inflammatory origin can be retraced to cellular stress, caused by metabolic imbalance, hence, called metaflammation (Hotamisligil, Nature, 2006, 444(7121): pp. 860-7, incorporated herein by reference). Prolonged malnutrition leads to chronic metaflammation, and may eventually cause degeneration of tissue, and onset of disease (Kushner et al., Arthritis Care Res (Hoboken), 2010, 62(4): pp. 442-6, incorporated herein by reference). Hyperglycaemia and hyperinsulinemia following a meal rich in easily digested carbohydrates are associated with cellular stress and increase of inflammatory markers (O'Keefe et al., J Am Coll Cardiol, 2008, 51(3): pp. 249-55, incorporated herein by reference). Diets with low glycemic load and glycemic index are suggested to silence metaflammation, and subsequently increase insulin sensitivity (Barclay et al., Am J Clin Nutr, 2008, 87(3): pp. 627-37; McKeown et al., Diabetes Care, 2004, 27(2): pp. 538-46; and Qi and Hu, Curr Opin Lipidol, 2007, 18(1): pp. 3-8; all incorporated herein by reference).
Current evidence suggests that insulin resistance and the associated abnormalities constitute complex phenotypes, explained by both environmental and genetic factors. The genetic makeup underlying these traits consists of several quantitative trait loci (QTL), whereof each QTL only explains a small fraction of the phenotype. The limited effects of these individual QTL make them difficult to identify, but the list of allelic variants associated with susceptibility to T2D development, in terms of single nucleotide polymorphisms (SNPs), is growing (Voight, B. F., et al., Twelve type 2 diabetes susceptibility loci identified through large-scale association analysis. Nat Genet, 2010. 42(7): p. 579-89, incorporated herein by reference). Also SNPs associated directly with insulin resistance have been found, but this line of research is in an early phase. (See Kantartzis et al., Clin Sci (Lond), 2009, 116(6): pp. 531-7; Liu et al., J Clin Endocrinol Metab, 2009, 94(9): pp. 3575-82; Palmer et al., Diabetes, 2004, 53(11): pp. 3013-9; Richardson et al., Diabetologia, 2006, 49(10): pp. 2317-28; Ruchat et al., Diabet Med, 2008, 25(4): pp. 400-6; and Smith et al., Diabetes, 2003, 52(7): pp. 1611-8; all incorporated herein by reference.)
The expression of a gene is the most basic phenotype in an organism. The genotype determines complex phenotypic traits through expression of several genes: expressional QTL (eQTL) (Jansen and Nap, Trends Genet, 2001, 17(7): pp. 388-91; and Schadt et al., Nature, 2003, 422(6929): pp. 297-302, both incorporated herein by reference). eQTL provide a direct link between genotype variation and gene- or pathway activities. The motivation to study how SNPs associated with a disease or a phenotypic trait may affect gene expression is to gain a direct understanding of the molecular mechanisms affected by the allelic variation (Rockman and Kruglyak, Nat Rev Genet, 2006, 7(11): pp. 862-72, incorporated herein by reference).
Homeostatic model assessment (HOMA) is a method for assessing surrogate measures of pancreatic β-cell function, insulin sensitivity, and insulin resistance derived from fasting blood glucose and insulin, alternatively insulin connecting peptide (C-peptide) concentrations (Wallace et al., Diabetes Care, 2004, 27(6): pp. 1487-95, incorporated herein by reference). The model was first proposed in 1985 (Matthews, et al., Diabetologia, 1985, 28(7): pp. 412-9, incorporated herein by reference), and an updated computer model (HOMA2) was published in 1998 (Levy et al., Diabetes Care, 1998, 21(12): pp. 2191-2, incorporated herein by reference). The calculation of insulin resistance designated as HOMA2 IR, is calibrated to a reference population, where the value 1 is set as normal (Wallace et al., 2004). HOMA2 IR was found to be a significant determinant of insulin resistance (Mojiminiyi et al., Clin Chem Lab Med, 2010, incorporated herein by reference).
In the present study, performed on modestly overweight but otherwise healthy individuals, associations between variation in SNPs and changes in insulin resistance in response to diets with different glycemic load were examined. SNPs were linked to genes and biological functions to develop an understanding of the molecular mechanisms potentially involved in onset of insulin resistance.
A randomized, controlled cross-over diet intervention trial was conducted on thirty-two young and healthy women and men, with body mass index (BMI, in kg/m2) between 24.5 and 27.5. Iso- and normocaloric meal replacement diets (MRDs) constituted all nutrients consumed during the study periods of two times six days with an eight day wash-out period in-between. Fasting blood samples were collected before and after each diet period, and effects of dietary intake on leukocyte gene expression profiles and insulin resistance were analyzed, as described previously ((Arbo I, Brattbakk H R, Langaas M et al. A balanced macronutrient diet induces changes in a host of pro-inflammatory biomarkers, rendering a more healthy phenotype; a randomized cross-over trial, 2010, (manuscript submitted), incorporated herein by reference). The two MRDs were: a high-carbohydrate diet (AHC) composed of 65:15:20 energy percent (E %) of carbohydrates, proteins, and fats; and a moderate-carbohydrate diet (BMC) with 27:30:43 E % of carbohydrates, proteins, and fats. The glycemic load of the AHC diet was calculated to be 2.71 times higher than the BMC diet.
Data extracted from samples were grouped and coded, according to diet and time of sampling. The abbreviations AHC0, AHC6, BMC0, and BMC6 denote before (day 0) and after (day 6) the AHC and the BMC diet intervention, respectively. Pair-wise analyses of data were performed for four different comparisons, which will be referred to throughout this paper: 1) AHC6-AHC0 and 2) BMC6-BMC0 identified responses to the AHC and the BMC diets, respectively, during six days on the respective diets. The comparison 3) BMC6-AHC6 identified the differences between the end-point responses to diet AHC and BMC after six days on diet, and finally, 4) (BMC6-BMC0)-(AHC6-AHC0) identified differences between the responses to AHC and BMC dieting. Complementary and more detailed information about subject recruitment, exclusion criteria, subject baseline characteristics, MRD compositions, and sampling techniques were described previously (Arbo I, Brattbakk H R, Langaas M et al. A balanced macronutrient diet induces changes in a host of pro-inflammatory biomarkers, rendering a more healthy phenotype; a randomized cross-over trial, 2010, (manuscript submitted), incorporated herein by reference).
Microarray analysis and preprocessing of microarray data was performed as previously described (Arbo I, Brattbakk H R, Langaas M et al. A balanced macronutrient diet induces changes in a host of pro-inflammatory biomarkers, rendering a more healthy phenotype; a randomized cross-over trial, 2010, (manuscript submitted), incorporated herein by reference). Briefly, leukocyte gene expression profiling was done on the HumanHT-12 Expression BeadChip v3.0 (Illumina). After removal of two outlier samples, background correction based on negative controls, quantile-quantile normalization, signal log2-transformation, and removal of not detected or bad probes, 27 372 unique probes were left in the “gene expression dataset”. The paired analyses of AHC6-AHC0 and BMC6-BMC0 identified 3225 and 1370 differentially expressed genes, respectively, where 843 genes overlapped between the analyses. For the paired groups BMC6-AHC6 and (BMC6-BMC0)-(AHC6-AHC0), no differentially expressed genes were identified. Microarray data were submitted to ArrayExpress (www.ebi.ac.uk/arrayexpress, accession number: E-TABM-1073).
Protein concentration analyses and assessment of insulin resistance were performed as previously described (Arbo I, Brattbakk H R, Langaas M et al. A balanced macronutrient diet induces changes in a host of pro-inflammatory biomarkers, rendering a more healthy phenotype; a randomized cross-over trial, 2010, (manuscript submitted), incorporated herein by reference), using fasting EDTA-plasma samples. Bio-Plex Diabetes Panel assays (Bio-Rad Laboratories Inc., Hercules, Calif., USA) were performed using Luminex xMAP™ technology, with a Bio-Plex 200 suspension array reader, and the data was extracted with the Bio-Plex Manager 5.0 software (Bio-Rad Laboratories Inc.). Briefly, analysis of the paired groups showed decreased (P<0.05) plasma concentrations of visfatin (nicotinamide phosphoribosyltransferase, Nampt), and increased plasma concentration of resistin (P<0.05) during AHC dieting. Likewise, during BMC dieting the analysis showed decreased plasma concentrations of insulin, C-peptide, glucagon, plasminogen activator inhibitor-1 (PAI-1), glucagon-like peptide-1 (GLP-1), tumor necrosis factor-α (TNF), interleukin-6 (IL-6), and visfatin, and increased plasma concentrations of resistin. Gastric inhibitory polypeptide (GIP), ghrelin, and leptin did not respond to any of the diet interventions.
The HOMA2 calculator version 2.2.2® (Diabetes Trials Units, University of Oxford, www.dtu.ox.ac.uk/homacalculator/index.php) (Matthews et al., 1985) was used to determine changes in insulin resistance in terms of HOMA2 IR. There was an average decrease in HOMA2 IR during both the AHC diet and the BMC diet, but the downregulation was only significant during BMC dieting.
DNA was extracted from EDTA-blood using E.Z.N.A Blood DNA Kit (D3392, OMEGA Bio-Tek, Inc., Norcross, Ga., USA). The subjects were genotyped using the ˜200 K Cardio-MetaboChip (Metabochip) SNP array, an Infinium iSelect HD Custom Genotyping BeadChip (Illumina, San Diego, Calif., USA), designed by the Cardio-MetaboChip Consortium (Broad Institute, Cambridge, Mass., USA), and analyzed according to the Infinium HD Assay Ultra, Manual Experienced User Card. The Metabochip consists of SNPs associated with diseases or traits relevant to metabolic and atherosclerosis-cardiovascular endpoints, including T2D and hyperglycemia. The BeadChips were read by a BeadArray™ reader, and data were exported to GenomeStudio™ V2009, Genotyping V1.1.9 (Illumina), for visual quality control of genotype clustering, and extraction of quality measures (ChiTest100 and GenTrain Score) (Illumina, GenomeStudio™ Genotyping Module v1.0 User Guide. 2008, Illumina, Inc: San Diego, Calif., incorporated herein by reference). The ChiTest100 is a p-value calculated for each SNP, reflecting the deviation of that SNP to the genotype distribution according to the Hardy-Weinberg Equilibrium (HWE), using the χ2 statistic, normalized to 100 subjects. GenTrain Score is a measure of SNP clustering performance indicated by a number increasing with cluster quality, form 0 to 1.
A set of 22 transcription regulators and seven ligand-dependent nuclear receptors central to insulin resistance development (Olefsky and Glass, 2010; Hotamisligil, 2006; and Wymann and Schneiter, Nat Rev Mol Cell Biol, 2008, 9(2): pp. 162-76; all incorporated herein by reference) were selected. The selected candidate genes were uploaded to the Ingenuity Pathway Analysis 8.7 (IPA Ingenuity Systems®, Redwood City, Calif., USA, www.ingenuity.com) to find the upstream activators and inhibitors, and downstream target genes of the transcription regulators and the nuclear receptors. No filters were applied in IPA regarding species, tissues or cell lines, but an upper limit of 150 upstream and 150 downstream genes was defined. The SNPs linked to the extended selected list of 276 candidate genes were extracted from the dbSNP database (www.ncbi.nlm.nih.gov/projects/SNP, National Center for Biotechnology Information, U.S. National Library of Medicine, Bethesda), and matched with 469 SNPs on the Metabochip. These 469 SNPs (linked with 276 candidate genes) were uploaded to the web server FASTSNP (Yuan et al., Nucleic Acids Res, 2006, 34 (Web Server issue): pp. W635-41, incorporated herein by reference) to prioritize the SNPs that were most likely to have functional effect on the expression of the linked gene. According to a decision tree, each SNP was assigned a risk score between 0 and 5. Risk score 0 means that the SNP has no known effect (e.g. located in a downstream or upstream untranslated region, nearby the gene), and 5 means that the SNP has a functional effect (e.g. introduces a stop codon and hence premature translational termination). Basically all SNPs with risk score lower than 2 were discarded. Since several SNPs with risk score 2 or higher were linked to a single gene, we defined an upper limit of seven SNPs per gene. That was done by increasing the risk score claim one factor at the time, until the number of SNPs was at most seven. The result was a list of 190 SNPs.
Four different selections of SNPs were used in the analyses:
For all analyses a two-stage strategy was performed. In the first stage, analysis of variance (ANOVA) was performed to test the null hypothesis, whether there was no difference in either HOMA2 IR, gene expression (log2-ratio), or protein concentration (loge-ratio) change between the genotypes. Genotype was used as covariate, and changes as response variables. P-values were adjusted for multiple testing using the Benjamini-Hochberg step up algorithm (Benjamini and Hochberg, Journal of the Royal Statistical Society. Series B (Methodological), 1995, 57(1): pp. 289-300, incorporated herein by reference) to control the false discovery rate (FDR). In the second stage, a one-sample, two-sided t-test was assigned to test if the change in HOMA2 IR, gene expression, or protein concentration, was different from zero for any of the genotypes. For the ref-SNP selection and the gene-SNP selection, the second stage was performed only for the 100 best ranked entries, according to the ANOVA p-values. Hence, eight Top100 lists were generated, one for each comparison, the ref-SNP selection and the gene-SNP selection separately (see Supplementary tables 1-8). Within these lists the t-test p-values were adjusted for multiple testing using the Benjamini-Hochberg step up algorithm.
Unsupervised hierarchical clustering analyses were performed, using Manhattan distance measures and complete linkage. PCA was performed with discrete data, where the three possible genotypes were represented by numerical values (0, 1, 2). Analyses were performed using the R statistical analysis framework (R Development Core Team, R: A Language and Environment for Statistical Computing, 2010; Available from: www.r-project.org, incorporated herein by reference).
Functional analysis to identify biological functions and diseases significantly associated with gene lists were performed using IPA 8.7 (Ingenuity). Since the Metabochip is custom made, biased by SNPs associated with metabolic and cardiovascular traits, a custom reference set was also used in all analyses. This was composed of all the 10 515 genes that according to the dbSNP database were linked to the 71 061 SNPs on the Metabochip. P-values were adjusted for multiple testing using the Benjamini-Hochberg step up algorithm.
SNPs Associated with HOMA2 IR
Insulin resistance is a complex trait and the contribution of each single locus to the phenotype is small. To propose loci involved in the manifestation of this trait, the environmental homeostasis was challenged by introducing the subjects to two different diets. The responding change in HOMA2 IR for the four comparisons was related to SNPs in the ref-SNP selection. The biological relevance to insulin resistance was examined for all SNPs with FDR<0.2, a cut-off used in larger cohorts (>3000 subjects) earlier (Povel et al., Int J Obes (Lond), 2010, 34(5): pp. 840-5, incorporated herein by reference).
The change in HOMA2 IR during the AHC diet was associated (FDR<0.1) with four SNPs, with identical allele distribution between the subjects (
The three other SNPs,
rs29095 (tccctcgaataaaggtgaaatttttaaaat[a/c]tcagtgaataggaatgtgcaaagactaag) (SEQ ID NO: 3) (Chr18:9957549),
rs7237794 (ctgcccctcccgacacagcacatacacaca[c/t]tgacgttttgctactacagcatatagcctt) (SEQ ID NO: 4) (Chr18:9951304), and
rs917688 (ttctctcatgcttaatatttggaactataa[a/c]gctaaaggccattgacgtagctaaaaatct) (SEQ ID NO: 5)(Chr18:9962736), (Chr18:9951304 . . . 9962736, C→A, T→C, and C→A, respectively) were closely linked, and rs7237794 and rs917688 were located in an intron region, and in the untranslated region of the 3′end of the gene vesicle-associated membrane protein-associated protein A, 33 kDa (VAPA), respectively. The 29 subjects homozygous for the consensus allele had an average downregulation of HOMA2 IR during the AHC diet (estimate of average change (
Another association (FDR<0.02) was found between HOMA2 IR change during the AHC diet and the SNP rs10803976 (
The same procedure was followed for the candidate gene-SNP list. HOMA2 IR change during the BMC diet was associated with the SNP rs6494711 (
SNP in the GIP Gene is Associated with PAI-1 Protein Concentration in Plasma
To screen for cis- and trans-regulating eQTLs that affected expression of genes central in pathogenesis of T2D, we tested the association between the SNPs in the diabetes panel-SNP and HOMA2 IR. Association was detected for the SNP rs2291726 (
SNPs Associated with HOMA2 IR Change are Related to Type 2 Diabetes
Since insulin resistance is manifested by numerous QTL, we wanted to explore how the genotype profile in each individual correlated with the change in HOMA2 IR in response to the diets. Heatmaps were generated showing hierarchical clustering of the ref-SNP selection Top100 SNPs associated with HOMA2 IR. SNPs were clustered according to allele distribution, and the subjects were sorted according to HOMA2 IR differences (increasing from left to right,
Since the subjects with the strongest increase in HOMA2 IR during the AHC diet had such a distinct genotype profile, these were suspected to be involved in our most significant associations between SNP and HOMA2 IR. To determine if this was due to technical artifacts we generated a dendrogram and a PCA plot (showing the three first principal components) based on allele information from the 71 061 SNPs to examine whether we had outlier individuals (
To explore functional and biological information about the SNPs that showed the highest association with HOMA2 IR, we assessed IPA's Functional Analysis. We extracted the genes that according the dbSNP database were linked to all the SNPs in the ref-SNP selection Top100 lists. We found 366 unique SNPs in the four lists, and these were linked with 150 unique genes. The gene set was significantly associated with T2D, displaying an FDR-value equal to 6.39×10−13 for the sum of all four Top100 lists (Table 1). The SNPs included in the ref-SNP selection Top100 lists were also associated with several traits that usually co-exist with insulin resistance, like cardiovascular disorder, hypertension, and immunological disorder.
To identify potential insulin resistance eQTLs, we matched the genes of the Top100 pairs from the gene-SNP lists for the four comparisons, with the genes related to insulin resistance in the literature, using the following search in PubMed (NCBI, NIH, USA): (“Diabetes Mellitus, Type 2”[MeSH] OR “Insulin resistance”[MeSH] OR “Hyperglycemia”[MeSH] OR “Insulin-Secreting Cells”[MeSH]). None of the pairs of SNPs and genes related to insulin resistance showed significant association between genotype and expression changes, but several genes showed significant genotype specific expression changes (FDR<0.05) in response to diet AHC and BMC (Table 2). This suggests that genotype is a considerable variable, contributing to interindividual gene expression variability.
In this study we have defined a method to relate SNPs to phenotypic changes in response to an intervention, and applied this method to identify potential susceptibility loci for insulin resistance. The method should also be applicable on larger cohorts. We observed distinctive genotype profiles among strong responders to high and low glycemic load, concerning increase and decrease of insulin resistance, respectively. Several eQTL were found linked to genes related to insulin resistance, showing inter-genotype variability. On a limited number of subjects, we successfully applied statistical and bioinformatical methods new to this area of genetic research.
Our most significant finding is association of insulin resistance to VAPA, a protein previously shown to play a role in the vesicle budding and fusion events involving protein transport in cells (Weir et al., Biochem Biophys Res Commun, 2001, 286(3): pp. 616-21, incorporated herein by reference). GLUT4 is translocated to the surface of myocytes and adipocytes in response to insulin binding to its receptor. Various proteins control this GLUT4 translocation, including VAMP2 and syntaxin-4. VAPA interacts with both of these proteins in skeletal myoblasts, and is suggested to be a regulator of VAMP2 availability in insulin-dependent GLUT4 translocation (Foster, et al., Traffic, 2000, 1(6): pp. 512-21, incorporated herein by reference). The effect of insulin on GLUT4 translocation in monocytes is discussed, but there are indications that systemic insulin resistance is indicated by the presence of GLUT4 receptors on the monocyte surface (Mavros et al., Diabetes Res Clin Pract, 2009, 84(2): pp. 123-31, incorporated herein by reference. There is a strong association between variation in the SNPs rs29095, rs7237794, and rs917688 (
There are not many known genes regulated by the transcription factor PIAS1, but three of them, myogenin (MYOG) (Hsu et al., J Biol Chem, 2006, 281(44): pp. 33008-18, incorporated herein by reference), actin, alpha 2, smooth muscle, aorta (ACTA2, member of F-actin) (Kawai-Kowase et al., Mol Cell Biol, 2005, 25(18): pp. 8009-23, incorporated herein by reference), and cyclin-dependent kinase inhibitor 1A (CDKN1A) (Megidish et al., J Biol Chem, 2002, 277(10): pp. 8255-9, incorporated herein by reference) are all mediators of insulin induced signalling, shown in a variety of cells, including neutrophils, adipocytes, myocytes, pancreatic islet cells, and intestinal endocrine cells. (See Chodniewicz and Zhelev, Blood, 2003, 102(6): pp. 2251-8; Inoue et al., J Biol Chem, 2008, 283(30): pp. 21220-9; Kaneto et al., Diabetologia, 1999, 42(9): pp. 1093-7; Lim et al., Endocrinology, 2009, 150(12): pp. 5249-61; Sumitani et al., Endocrinology, 2002, 143(3): pp. 820-8; and Yoshizaki et al., Mol Cell Biol, 2007, 27(14): pp. 5172-83; all incorporated herein by reference.)
The association we found between the SNP rs6494711 and insulin resistance showed that homozygotes for both the consensus and the alternative allele had a decrease in insulin resistance during the BMC diet, but the heterozygotes had no significant change. However, the genotype specific change was not reflected in the mRNA expression data of PIAS1, but the effect of the transcription factors could be controlled by post-transcriptional activation. The effect may also be mediated through gene expression responses in other cells more insulin sensitive than leukocytes.
Increased PAI-1 concentration in the liver is associated with insulin resistance in mice (Takeshita et al., Metabolism, 2006, 55(11): pp. 1464-72, incorporated herein by reference), and loss of affinity between GIP and GIP-receptor affect localization of PAI-1 to mouse plasma (Hansotia et al., J Clin Invest, 2007, 117(1): pp. 143-52, incorporated herein by reference). Since GIP-secretion is stimulated by glucose, this could explain why genetic variation in the GIP gene was associated with changes in PAI-1 protein concentrations in plasma.
Today the recommendation of daily intake of carbohydrates in Norway is 50-60 E % (Utviklingen i norsk kosthold, Vol. 2008, Utviklingen i norsk kosthold 2008, Oslo: Direktoratet, 2008, 27 s, incorporated herein by reference). Such a high fraction will contribute to a high dietary glycemic load, unless considerable caution is taken to choose carbohydrate sources with low glycemic index. With precaution, regarding the small sample size, our results suggest that some individuals are sensitive to high glycemic load, which is shown by an increase in insulin resistance during high-carbohydrate dieting (AHC) (
The use of genotyping data to link gene expression differences with phenotypes has increased markedly the last years. However, the use of genetic variation to stratify responses to a homeostatic challenge, like a diet intervention, has not been quite as common. The reason might be that the sample size required to gain significant results far exceeds what is easily manageable in an intervention study. We have shown that genotype is a source of interindividual variability in the response to a change in glycemic load, and suggest that genotype information can be integrated as an explanatory variable in microarray gene expression analysis.
Some obvious limitations need to be acknowledged in our study. What is already mentioned is the limited sample size. Whereas the study of average responses to a dietary intervention in a controlled cross-over study has produced robust findings (Arbo I, Brattbakk H R, Langaas M et al., A balanced macronutrient diet induces changes in a host of pro-inflammatory biomarkers, rendering a more healthy phenotype; a randomized cross-over trial, 2010, (manuscript submitted), incorporated herein by reference), dividing the subjects into two or three groups based on genotype inevitably decreases the statistical power. We nevertheless used reasonable criteria to declare associations of SNPs and eQTL (FDR<0.05), while acknowledging that of course in largest studies much more significant association confidence can be obtained. We considered various quality criteria of the SNPs that could account for aberrant behaviour in our statistical tests. One quality criterion concerns the Hardy-Weinberg equilibrium (HWE). In a population, deviation from HWE may be indicative of selective pressure, but because most genes are not under selection it can also be used as an indicator of problems in the genotyping procedure leading to bias in the observed allele frequencies (Greene et al., Lect Notes Comput Sci, 2010, 6023(LNCS): pp. 74-85, incorporated herein by reference). Another reason why the allele distribution might deviate from HWE is the relatively small sample size of the current study, making it vulnerable to biased selection of subjects. Nevertheless, except where noted, none of the SNPs for which we found associations showed a gross skewness in allele frequencies that would significantly violate HWE, and indeed all passed the SNP call quality criteria of the Genotyping V1.1.9 software (Illumina). To ascertain the genotyping and HWE quality of each individual SNP is challenging, so we did carefully consider these quality criteria when interpreting the results of individual SNPs.
Hierarchical clustering and PCA revealed two genetical outliers in our sample size (
Leukocytes are an easy accessible source for transcriptome profiling, and an obvious choice to screen for inflammatory gene expression changes in response to food. However, the knowledge on insulin responsiveness is limited. The inflammatory properties of monocytes and macrophages are central in the development of insulin resistance in insulin target cells, like adipocytes and myocytes. But it is not known whether the established molecular mechanisms behind insulin resistance are the same in leukocytes. We have shown earlier that monocytes are insulin responsive in a dose dependent manner (Ingerid Arbo, Cathinka L Halle, Darshan Malik, et al. Insulin induces fatty acid desaturase expression in human monocytes, 2010, (manuscript submitted), incorporated herein by reference), inducing increased desaturase transcription. However, this does not guarantee that we can expect significant association between leukocyte gene expression and changes in insulin resistance, considering an earlier finding that gene expression profiles in leukocytes and adipocytes deviate (Brattbakk H R, Arbo I, Aagaard S, et al. Balanced caloric macronutrient composition downregulates immunological gene expression in human blood cells—adipose tissue diverges, 2010, (manuscript submitted), incorporated herein by reference). This demonstrates the need to investigate not only blood, but also additional parallel sampled biopsies of well established insulin target tissue, like adipose tissue.
Of the SNPs disclosed herein, it is seen that some are directly related to the genes for VAPA, Pias1 and GIP, while some are closely related thereto and can serve as “surrogate” markers. These SNPs are more specifically: rs16961756, rs1242483, rs29095, rs7237794, rs917688, rs6494711, rs1489595 and rs2291726.
The SNPs may serve as new markers of candidate QTL contributing to explain the genetic aspect of insulin resistance development. Also, VAPA and PIAS1 are new candidate genes involved in the molecular mechanisms behind insulin resistance. Finally, certain SNPs are candidate eQTL for plasma PAI-1 concentration, also related to insulin resistance. Our results have demonstrated the added value of incorporating genotype data in gene expression analysis to explain interindividual variability. A genotype profile of specific SNPs can distinguish weak and strong responders to glycemic load, with respect to insulin resistance. SNP typing may eventually be used to provide concrete dietary advice to persons genetically predisposed to T2D.
All references cited herein, are hereby incorporated by reference. Sequences of the single nucleotide polymorphisms cited by the accession numbers herein, are hereby incorporated by reference and can be found at www.ncbi.nlm.nih.gov/sites/entrez or snpper.chip.org/bio, among other sites, using the accession numbers provided.
0.036
<0.001
0.039
0.042
0.005
0.031
0.021
0.002
0.048
0.009
0.028
0.011
0.032
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/IB11/03038 | 12/13/2011 | WO | 00 | 10/16/2013 |
| Number | Date | Country | |
|---|---|---|---|
| 61424005 | Dec 2010 | US |
