Avoiding selection bias in metabolomics studies: a tutorial.

BACKGROUND: Metabolomics techniques are increasingly applied in epidemiologic research. Many available assays are still relatively expensive and therefore measurements are often performed in small patient population studies such as case series or case-control designs with strong participant selection criteria. Subsequently, metabolomics data are frequently used to assess secondary associations for which the original study was not explicitly designed. Especially in these secondary analyses, there is a risk that the original selection criteria and the conditioning that takes place due to this selection are not properly accounted for which can lead to selection bias. AIM OF REVIEW: In this tutorial, we start with a brief theoretical introduction on the issue of selection bias. Subsequently, we demonstrate how selection bias can occur in metabolomics studies by means of an investigation into associations of metabolites with total body fat in a nested case-control study that was originally designed to study effects of elevated fasting glucose. KEY SCIENTIFIC CONCEPTS OF REVIEW: We demonstrate that standard analytical methods, such as stratification or adjustment in regression analyses, are not suited to deal with selection bias and may even induce the bias when analysing metabolite-phenotype relationships in selected groups. Finally, we show that inverse probability weighting, also known as survey weighting, can be used in some situations to make unbiased estimates of the outcomes.

IgG is elevated in obese white adipose tissue but does not induce glucose intolerance via Fcγ-receptor or complement.

BACKGROUND/OBJECTIVES: In obesity, B cells accumulate in white adipose tissue (WAT) and produce IgG, which may contribute to the development of glucose intolerance. IgG signals by binding to Fcγ receptors (FcγR) and by activating the complement system. The aim of our study was to investigate whether activation of FcγR and/or complement C3 mediates the development of high-fat diet-induced glucose intolerance. METHODS: We studied mice lacking all four FcγRs (FcγRI/II/III/IV-/-), only the inhibitory FcγRIIb (FcγRIIb-/-), only the central component of the complement system C3 (C3-/-), and mice lacking both FcγRs and C3 (FcγRI/II/III/IV/C3-/-). All mouse models and wild-type controls were fed a high-fat diet (HFD) for 15 weeks to induce obesity. Glucose metabolism was assessed and adipose tissue was characterized for inflammation and adipocyte functionality. RESULTS: In obese WAT of wild-type mice, B cells (+142%, P<0.01) and IgG (+128% P<0.01) were increased compared to lean WAT. Macrophages of FcγRI/II/III/IV-/-mice released lower levels of cytokines compared to wild-type mice upon IgG stimulation. Only C3-/- mice showed reduced HFD-induced weight gain as compared to controls (-18%, P<0.01). Surprisingly, FcγRI/II/III/IV-/- mice had deteriorated glucose tolerance (AUC +125%, P<0.001) despite reduced leukocyte number (-30%, P<0.05) in gonadal WAT (gWAT), whereas glucose tolerance and leukocytes within gWAT in the other models were unaffected compared to controls. Although IgG in gWAT was increased (+44 to +174%, P<0.05) in all mouse models lacking FcγRIIb, only FcγRI/II/III/IV/C3-/- mice exhibited appreciable alterations in immune cells in gWAT, for example, increased macrophages (+36%, P<0.001). CONCLUSIONS: Lack of FcγRs reduces the activity of macrophages upon IgG stimulation, but neither FcγR nor C3 deficiency protects against HFD-induced glucose intolerance or reduces adipose tissue inflammation. This indicates that if obesity-induced IgG contributes to the development of glucose intolerance, this is not mediated by FcγR or complement activation.

Obesity is associated with an altered autonomic nervous system response to nutrient restriction.

OBJECTIVE: Heart rate variability (HRV) reflects the balance of activities of sympathetic and parasympathetic components of the autonomic nervous system. We compared HRV parameters in response to a prolonged fast in obese versus normal weight humans. In addition, the effect of weight-loss was evaluated in obese individuals. DESIGN: Intervention study. PATIENTS: The study subjects included 14 nondiabetic obese (12 females/2 males, aged 30 ± 3 years, Body Mass Index (BMI) 35·2 ± 1·2 kg/m(2) ) and 12 lean subjects (10 females/2 males, aged 27 ± 3 years, BMI 23·3 ± 0·5 kg/m(2) ). MEASUREMENTS: HRV was examined 75 min after standardized breakfast and after a 48-h fast in 14 nondiabetic obese and 12 lean subjects. The postprandial measurement was repeated in 12 obese subjects after weight-loss. RESULTS: In lean subjects, fasting decreased high-frequency (HF) power by 43% (P < 0·05) and decreased low-frequency (LF) power by 37% (P = 0·1), leaving the LF/HF ratio unchanged (P = 0·7). In the obese group, autonomic nervous system tone shifted to sympathetic dominance as the LF/HF increased from 0·61 to 1·14 (P = 0·03). After an average weight-loss of 13·8 kg in obese subjects, a trend for sympathetic dominance was found; the LF/HF ratio increased by 56% (P = 0·06). CONCLUSION: Our data show that a 48-h fast leaves autonomic nervous system balance unaltered in lean subjects. In contrast, a 48-h fast, as well as weight-loss, induces sympathetic dominance in obese humans.

Insulin-resistance and metabolic syndrome are related to executive function in women in a large family-based study.

While type 2 diabetes is well-known to be associated with poorer cognitive performance, few studies have reported on the association of metabolic syndrome (MetS) and contributing factors, such as insulin-resistance (HOMA-IR), low adiponectin-, and high C-reactive protein (CRP)-levels. We studied whether these factors are related to cognitive function and which of the MetS components are independently associated. The study was embedded in an ongoing family-based cohort study in a Dutch population. All participants underwent physical examinations, biomedical measurements, and neuropsychological testing. Linear regression models were used to determine the association between MetS, HOMA-IR, adiponectin levels, CRP, and cognitive test scores. Cross-sectional analyses were performed in 1,898 subjects (mean age 48 years, 43% men). People with MetS had significantly higher HOMA-IR scores, lower adiponectin levels, and higher CRP levels. MetS and high HOMA-IR were associated with poorer executive function in women (P = 0.03 and P = 0.009). MetS and HOMA-IR are associated with poorer executive function in women.

Menopause impacts the relation of plasma adiponectin levels with the metabolic syndrome.

OBJECTIVE: Plasma adiponectin is negatively correlated with metabolic syndrome (MetS) components obesity and insulin sensitivity. Here, we set out to evaluate the effect of menopause on the association of plasma adiponectin with MetS. DESIGN: Data on plasma adiponectin and MetS were available from 2256 individuals participating in the Erasmus Rucphen Family study. Odds ratios for MetS were calculated by logistic regression analysis using plasma adiponectin quartiles. The discriminative accuracy of plasma adiponectin for MetS was determined by calculating the area under the curve (AUC) of receiver operator. Analyses were performed in women and men, pre- and postmenopausal women and younger and older men. RESULTS: Virtually all determinants of MetS differed significantly between groups. Low plasma adiponectin showed the highest risk for MetS in postmenopausal women (odds ratio = 18.6, 95% CI = 7.9-44.0). We observed a high discriminative accuracy of age and plasma adiponectin for MetS not only in postmenopausal women (AUC = 0.76) but also in other subgroups (AUC from 0.67 to 0.87). However, in all groups, the discriminative accuracy of age and body mass index (BMI) for MetS was similar to the discriminative accuracy of age and plasma adiponectin. CONCLUSIONS: Low plasma levels of adiponectin are associated with increased prevalence of MetS, especially in postmenopausal women. Age and BMI have similar discriminatory accuracies for presence of MetS when compared with age and plasma adiponectin. Thus, we conclude that the association of plasma adiponectin with MetS is significantly affected by menopause but challenge the additional value of adiponectin for the discriminatory accuracy for presence of MetS.

Metabolomics (liver and blood profiling) in a mouse model in response to fasting: a study of hepatic steatosis.

A metabolomic approach was applied to a mouse model of starvation-induced hepatic steatosis. After 24 h of fasting it appears that starvation reduced the phospholipids (PL), free cholesterol (FC), and cholesterol esters (CE) content of low-density lipoproteins (LDL). In liver lipid profiles major changes were observed using different techniques. High performance thin layer chromatography (HPTLC)-measurements of liver-homogenates indicated a significant rise of FC with 192%, triacylglycerols (TG) with 456% and cholesterol esters (CE) with 268% after 24 h of starvation in comparison with the control group. Reversed phase liquid chromatography coupled to mass spectrometry measurements (LC-MS) of liver homogenate indicated that the intensity of Phosphatidylcholine (PC) in the 24-h starvation group dropped to 90% of the value in the control group while the intensity of CE and TG increased to 157% and 331%, respectively, of the control group. Interestingly, a 49:4-TG with an odd number of C atoms appeared during starvation. This unique triacylglycerol has all characteristics of a biomarker for detection of hepatic steatosis. These observations indicate that in mammals liver lipid profiles are a dynamic system which are readily modulated by environmental factors like starvation.

Specific and efficient targeting of adenovirus vectors to macrophages: application of a fusion protein between an adenovirus-binding fragment and avidin, linked to a biotinylated oligonucleotide.

BACKGROUND: The application of serotype 5 adenoviruses (Ad5) in macrophages is hampered by the absence of the endogenous coxsackie adenovirus receptor (CAR). METHODS: To overcome this limitation, we first generated a linker protein consisting of the virus-binding domain of CAR and the C-terminus of avidin. Second, to target macrophages, this linker protein was equipped with the biotinylated (bio) oligonucleotide dA6G10, which was previously shown to display a high affinity for the scavenger receptor A (SR-A). RESULTS: As compared to nontargeted virus, the linker protein equipped with bio-dA6G10 showed a 500-fold increased reporter gene expression in mouse macrophage RAW264.7 cells. A linker protein equipped with a bio-dA16 control oligonucleotide was inactive. Moreover, the bio-dA6G10-equipped linker showed a 390-fold increased luciferase expression in the macrophage cell line J774 and 276- and 150-fold increased reporter gene expression in primary peritoneal and bone marrow (BM)-derived macrophages, respectively. Using BM-derived macrophages from SR-A knockout mice, it was shown that the dA6G10-mediated uptake is predominantly SR-A-mediated. CONCLUSIONS: Thus, we have developed a novel tool to link biotinylated ligands to a virus-binding fragment of CAR and have exploited this linker protein to extend the applicability of Ad5 to infect transformed and primary macrophages.

Efficient targeting of adenoviral vectors to integrin positive vascular cells utilizing a CAR-cyclic RGD linker protein.

Vascular smooth muscle (VSMC) and endothelial cells (EC) are particularly resistant to infection by type 5 adenovirus (Ad) vectors. To overcome this limitation and target Ad vectors to ubiquitously expressed alpha(V)beta(3/5) integrins, we have generated a linker protein consisting of the extracellular domain of the coxsackie adenovirus receptor (CAR) connected via avidin to a biotinylated cyclic (c) RGD peptide. After optimization of CAR to cRGD and to Ad coupling, infection of mouse heart endothelial cells (H5V) could be augmented significantly, as demonstrated by 600-fold increased transgene expression levels. In EOMAs, a hemangioendothelioma-derived cell line, the fraction of infected cells was enhanced 4- to 6-fold. Furthermore, the fraction of infected primary mouse VSMC was increased from virtually 0% to 25%. Finally, in human umbilical vein endothelial cells, the number of GFP positive cells was enhanced from 2% to 75%. In conclusion, CAR-cRGD is a versatile and highly efficient construct to target Ad vectors to both transformed and primary VSMC and EC.