• <tr id="yyy80"></tr>
  • <sup id="yyy80"></sup>
  • <tfoot id="yyy80"><noscript id="yyy80"></noscript></tfoot>
  • 99热精品在线国产_美女午夜性视频免费_国产精品国产高清国产av_av欧美777_自拍偷自拍亚洲精品老妇_亚洲熟女精品中文字幕_www日本黄色视频网_国产精品野战在线观看 ?

    Bidirectional Causality Between Immunoglobulin G N-Glycosylation and Metabolic Traits: A Mendelian Randomization Study

    2024-01-22 10:32:42XioniMengWeijieCoDiLiuIsintMrngElijhWeijiXingHifengHouXizhuXuMnshuSongYouxinWng
    Engineering 2023年7期

    Xioni Meng,Weijie Co,Di Liu,Isint Mrng Elijh,Weiji Xing,Hifeng Hou,Xizhu Xu,Mnshu Song*, Youxin Wng,b,*

    a Beijing Key Laboratory of Clinical Epidemiology, School of Public Health, Capital Medical University, Beijing 100069, China

    b Centre for Precision Health, Edith Cowan University, Perth, WA 6027, Australia

    c Centre for Biomedical Information Technology, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen 518055, China

    d School of Public Health, Shandong First Medical University & Shandong Academy of Medical Sciences, Jinan 250117, China

    e School of Medical and Health Sciences, Edith Cowan University, Perth, WA 6027, Australia

    Keywords:Mendelian randomization study Immunoglobulin G N-glycosylation Metabolic traits Quantitative trait loci Bidirectional causality

    ABSTRACT Although the association between immunoglobulin G(IgG)N-glycosylation and metabolic traits has been previously identified, the causal association between them remains unclear.In this work, we used Mendelian randomization (MR) analysis to integrate genome-wide association studies (GWASs) and quantitative trait loci (QTLs) data in order to investigate the bidirectional causal association of IgG Nglycosylation with metabolic traits.In the forward MR analysis,59(including nine putatively causal glycan peaks (GPs) for body mass index (BMI) (GP1, GP6, etc.) and seven for fasting plasma glucose (FPG)(GP1, GP5, etc.)) and 15 (including five putatively causal GPs for BMI (GP2, GP11, etc.) and four for FPG (GP1, GP10, etc.)) genetically determined IgG N-glycans were identified as being associated with metabolic traits in one- and two-sample MR studies, respectively, by integrating IgG N-glycan-QTL variants with GWAS results for metabolic traits (all P < 0.05).Accordingly, in the reverse MR analysis of the integrated metabolic-QTL variants with the GWAS results for IgG N-glycosylation traits,72(including one putatively causal metabolic trait for GP1 (high-density lipoprotein cholesterol (HDL-C)) and five for GP2(FPG,systolic blood pressure(SBP),etc.))and four(including one putatively causal metabolic trait for GP3(HDL-C)and one for GP9(HDL-C))genetically determined metabolic traits were found to be related to the risk of IgG N-glycosylation in one- and two-sample MR studies, respectively (all P < 0.05).Notably,genetically determined associations of GP11 →BMI (fixed-effects model-Beta with standard error (SE):0.106 (0.034) and 0.010 (0.005)) and HDL-C →GP9 (fixed-effects model-Beta with SE: –0.071 (0.022)and –0.306 (0.151)) were identified in both the one- and two-sample MR settings, which were further confirmed by a meta-analysis combining the one- and two-sample MR results (fixed-effects model-Beta with 95% confidence interval (95% CI): 0.0109 (0.0012, 0.0207) and –0.0759 (–0.1186, –0.0332),respectively).In conclusion,the comprehensively bidirectional MR analyses provide suggestive evidence of bidirectional causality between IgG N-glycosylation and metabolic traits, possibly revealing a new richness in the biological mechanism between IgG N-glycosylation and metabolic traits.

    1.Introduction

    Metabolic syndrome is a set of widely prevalent and multifactorial metabolic disorders, including obesity, insulin resistance, dyslipidemia, and hypertension [1,2].These conditions are interrelated and share underlying mediators, mechanisms,and pathways [2].However, the molecular mechanisms of metabolic disorders are not fully understood.Metabolic disorders are critical risk factors for fatal diseases (including cardiovascular disease,stroke,and cancer)[3–5],which escalate global public health problems and exacerbate the global burden of related fatal diseases.Therefore, exploring the pathogenesis of metabolic traits is of great importance for the prevention and intervention of metabolic disorders.

    Glycosylation, as a crucial and common posttranscriptional attachment of proteins, is involved in many biological and/or pathological processes [6–8].Previous studies have shown that more than half of all known proteins have a glycan structure,which affects the function of the protein [7,9].As the most common type of antibody,immunoglobulin G(IgG)plays a crucial role in nonspecific immune function.N-glycans are remarkably abundant and diverse, and are usually found on the outermost surface of the IgG binds to cellular and secreted macromolecules.The placement of N-glycans induces the effects of the Fc region of an IgG macromolecule in antibody-based immunity in humans.Alteration of N-glycosylation significantly influences IgG structure and function and, by extension, its immunological properties, thereby causing a shift between pro- and anti-inflammatory IgG functions[7,10,11].Altered IgG N-glycans have been shown to be involved in several metabolic disorders, including obesity [12], dyslipidemia[13], hypertension [14], and type 2 diabetes mellitus (T2DM)[15,16].However, the causality between the alteration of IgG Nglycosylation and metabolic traits remains to be determined.

    Recently,Mendelian randomization(MR)has been increasingly used to make causal inferences between risk factors and outcomes.More specifically,using a variety of instrumental variable(IV)analysis methods, MR studies rely on the natural, random assortment of genetic variants during gamete formation,which greatly reduces the impact of reverse causality and confounding factors [17–19].Nevertheless, due to genetic variation, the MR approach is usually biased by weak IVs that only explain a small part of the exposure variance[20–22].Genome-wide association studies(GWASs)have ascertained genetic variants related to quantitative exposure, and thus provide knowledge of quantitative trait loci (QTLs) [23–26].Integrating QTL data and GWAS analysis can increase the ability of GWASs to ascertain genetic loci controlling complex traits and improve the explanation of trait variance—suggesting that QTL variants could provide highly effective IVs for risk factors in MR studies [23,26,27].Previous studies have reported that the polymorphisms of B3GAT1, SLC9A9, MGAT5, FUT8, FUT6/FUT3, and

    HNF1A are linked with variation within the plasma N-glycome in European adults [28], and that these variants are involved in the regulation of glycoenzymes and play crucial roles in the process of IgG N-glycosylation among people of European descent [29].Consequently, we assumed that identifying IgG N-glycan-QTLs IVs and integrating them into disease (outcome)-related IVs(genetic variants) in a GWAS could help elucidate the underlying molecular mechanisms of genetic susceptibility to diseases,at least in part due to altered IgG N-glycans.

    In fact,it should be pointed out that all current MR approaches present the challenge of discriminating the mediation effect from horizontal pleiotropy [30,31], with the former being interpreted as results when the genetic variant affects the target outcome through multiple phenotypes (outside of the pathway of interest exposure).We hypothesized that the association between IgG N-glycosylation and metabolic traits might have bidirectional causality,which might contribute to bidirectional regulation based on independent pathogenic mechanisms or pathways.

    With the continuous development and optimization of glycoengineering techniques, glycans provide a novel dimension of medical science—namely, glycomedicine, in which glycomics approaches can be employed with the aim of better targeting disease diagnostics,drug discovery,and dosing based on individual glycomics profiles to enable preventive, predictive, and precision medicine[32].Therefore,we undertook a bidirectional MR analysis integrating GWASs for metabolic traits/IgG N-glycans and IgG N-glycan-QTLs/metabolic-QTLs data in order to assess the causation between IgG N-glycosylation and metabolic traits in East Asians.Our study further identified the mechanisms underlying genetic susceptibility to complex traits and explored the regulatory mechanisms of IgG N-glycosylation.This work may provide a certain basis for the development of glycomedicine targeting metabolic status or diseases and may contribute to facilitating timely healthy-lifestyle interventions and treatment decisions for metabolism-related diseases from the perspectives of prediction,prevention, and precision medicine.

    2.Materials and methods

    In this study, we used two approaches relying on two data sources,as follows:We used a one-sample MR approach,in which the IV-exposure and IV-outcome effects were evaluated in the same sample;and we applied a two-sample MR approach,in which the IV-exposure and IV-outcome effects were assessed in different samples.The overall design used for this study is delineated in Fig.S1(a) in Appendix A.

    2.1.One-sample MR study based on individual-level data

    2.1.1.Study participants

    This study was based on a community-based cohort crosssectional study using data from Xuanwu Hospital, Capital Medical University, Beijing, China.More detailed descriptions of the study design and evaluation methodologies have been previously reported[12,33].All participants met the following inclusion criteria: ①an age greater than or equal to 18 years; ②no medication history in the past 2 weeks;and ③having IgG N-glycosylation and GWAS data.In addition, the exclusion criteria were presented as follows: If the participants had a history of mental illness, infectious disease, malignant tumor, stroke, or other cerebrovascular diseases and/or presented incomplete data, they were excluded from this study.In total, 536 participants who aligned with the above inclusion and exclusion criteria were included in the study.Each participant was required to sign written informed consent before enrollment.The study was approved by the Ethics Committee of the Capital Medical University, Beijing, China (No.2009SY16), and was performed in accordance with the principles expressed in the Declaration of Helsinki.

    2.1.2.Data collection

    All participants completed routine physical examinations that contained anthropometric measurements and biochemical tests;detailed information is described in our previous report[33].Blood pressure(BP),including diastolic blood pressure(DBP)and systolic blood pressure(SBP),was measured(with at least 5 min of rest)on the right arm with a standard mercury sphygmomanometer by trained nurses.The height, weight, and body mass index (BMI) of the participants were measured with the participants wearing light indoor clothing and no shoes.After an overnight fast, peripheral blood from the subjects were collected in the morning in ethylenediaminetetraacetic acid (EDTA) anti-coagulated tubes for the detection of biochemical indexes.Biochemical traits related to metabolism (triacylglycerol (TG), high-density lipoprotein cholesterol (HDL-C), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and fasting plasma glucose (FPG)) were assayed as previously described [13].Since the metabolic indicators were not normally distributed,we used zero-mean normalization(z-score normalization)to process the metabolic trait data and then continued the subsequent analysis.In this MR study,we integrated GWAS and QTL data to assess the causality between IgG Nglycosylation and metabolic traits.Therefore, the metabolic traits included in this study were BMI (kg?m-2), SBP (mmHg,1 mmHg = 0.133 kPa), DBP (mmHg), TG (mmol?L–1), HDL-C(mmol?L–1), TC (mmol?L–1), LDL-C (mmol?L–1), and FPG (mmol?L–1),which were used as continuous variables for subsequent statistical analysis.

    2.1.3.Analysis of the IgG N-glycome

    We performed IgG N-glycans detection on all participants, as previously described [34].In brief, IgG was separated from the human plasma samples of all subjects utilizing pretreated 96-well protein G monolithic plates.The isolated IgG was dried and denatured according to standard experimental procedure; then,the sample was incubated for 18–20 h (37 °C) to obtain the released IgG N-glycans.

    Next, 2-aminobenzamide labeling solution was added to each released IgG N-glycans sample,incubated for 3 h at 65°C,and then filtered to obtain the labeled N-glycans.Hydrophilic interaction chromatography-ultraperformance liquid chromatography was used to separate and detect the IgG N-glycan profiles.The separated chromatograms were first processed by the automatic processing method with a traditional integration algorithm; then,the chromatograms were manually corrected to ensure that all the analytic samples maintained the same intervals of integration.Finally,24 glycan peaks(GPs;GP1–GP24)were obtained according to peak position and retention time, and the amount of glycans in each peak was expressed as a percentage(%)of the total integrated area.Detailed structures of the N-glycans in each peak have been previously described [34].To control the experimental variability,we used batch correction and z-score transformation methods to deal with the original glycan data in order to make all data comparable [13,34].

    2.1.4.Genotyping and genotype imputation

    The IgG N-glycan-QTL genetic variants were applied as the IVs from the IgG N-glycans GWAS, which was conducted in 536 participants with both genetic and IgG N-glycosylation data from Xuanwu Hospital, Capital Medical University, Beijing, China [33].We used Illumina Omni Zhonghua chips(Illumina,USA)to perform the genotyping procedures.The quality control process was performed as previously described in detail [35].Genotypes were imputed using the East Asian component of the 1000 Genomes Project panel phase 3 via the Michigan Imputation Server [36].Single nucleotide polymorphisms (SNPs) with an imputation quality ratio>0.30 and a minor allele frequency(MAF)>0.01 were retained, leaving 7 108 659 imputed SNPs that were subsequently used for IgG N-glycan-QTLs mapping.We did not correct the principal component (PC), as we did not encounter the problem of population stratification.

    2.1.5.Statistical analyses

    With the individual-level data from one sample, we used a series of MR analyses to investigate the relationship between IgG N-glycosylation and metabolic traits.A genome-wide IgG N-glycan-QTL study was performed to identify genetic variants known to influence IgG N-glycosylation.In total, 7 108 659 imputed genetic variants against each of the 24 IgG N-glycans were tested in 536 participants.Adjusting for age and gender, IgG Nglycans and SNPs were used as dependent variables and independent variables, respectively, and the correlation between each SNP and the IgG N-glycans was tested by means of linear regression.We then performed GWASs for metabolic traits, adjusted for the same covariates of age and gender.Next, using SNPs as independent variables and each metabolic trait as a dependent variable, the correlation between each SNP and each metabolic trait was estimated by means of linear regression.To screen the IgG N-glycan-QTLs as IVs for IgG N-glycans, we performed a relatively conservative Bonferroni correction to select IVs (i.e.,P<1×10–5).Previous studies have suggested that a weak standard of significance level of 1×10–5can be used for causal inference in MR studies, especially in studies with a small sample size [37,38].Considering that our GWAS was based on 536 individuals, a relaxed standard 1 × 10–5was used to filter out candidate SNPs as instrumental variables(IVs)in the one-sample forward MR analysis.Since there is high linkage disequilibrium(LD)between many significant IgG N-glycan-QTLs and metabolic-QTLs, only the independent IgG N-glycan-QTLs/metabolic-QTLs for each IgG N-glycan and/or metabolic trait was retained (the r2of LD < 0.001).We selected the SNPs with the lowest P values from the genetic variants with LD.Moreover, LD proxies were defined based on 1000 genomes from East Asian samples.To reduce the bias of vertical pleiotropy, overlapping SNPs as candidate IVs for both IgG N-glycans and metabolic traits were removed [39].The retained SNPs were then used as final IVs.We estimated the variance in exposures explained by each SNP (R2= (2 × Beta2× MAF × (1 –MAF))/(2 × Beta2× MAF × (1 – MAF) + SEB2eta× 2 × N × MAF ×(1 – MAF)) [40] and summed them to calculate the overall R2and F-statistics (R2× (N – 2)/(1 – R2)).A higher R2and F-statistic suggested a lower risk of weak instrument bias.

    To evaluate the bidirectional causality between IgG N-glycans and metabolic traits, the inverse-variance weighted (IVW)method was adopted in the main MR analyses; this method was performed using the mr_ivw function based on the R package MendelianRandomization (version 0.6.0).We used the observed correlation between IgG N-glycosylation and metabolic traits to explain the relationship between genetic association with IgG N-glycosylation and genetic association with metabolic traits due to sample overlap.An IVW Q test was conducted to test the heterogeneity of the IVs.If heterogeneity existed (P < 0.05),the random-effect IVW method was applied; otherwise, the fixed-effect IVW method was used.We also implemented sensitivity analyses to determine potential violations of valid IV assumptions using the weighted median (WM) and the MREgger regression.The intercept of the MR-Egger analysis can be used to evaluate pleiotropy.If a pleiotropic effect was not observed, meaning that the threshold of the MR-Egger intercept was more than 0.05, then the results of the IVW MR method were retained.The MR pleiotropy residual sum and outlier(MR-PRESSO) was also performed, in which the global test was used to examine horizontal pleiotropy [31] (R package MR-PRESSO, version 1.0).As some weak IVs were included in the one-sample MR analysis, the MR robust adjusted profile score(MR.RAPS) method (R package mr.raps, version 0.2) was further used to make the results more robust [41].In addition, we performed leave-one-out sensitivity analyses, in which SNPs were excluded in turn for analyses, in order to investigate SNPs that could potentially bias the causal association.Moreover,a previous study reported that an inflammatory marker—namely, C-reactive protein (CRP)—was associated with the level of IgG N-glycans[42]; therefore, the study further investigated whether CRP in fact mediates the effect of IgG N-glycans on metabolic traits and vice versa, by means of a multivariable MR analysis performed using the IVW, WM, and MR-Egger regression approaches.The results were recorded as the Beta and standard error (SE) of the outcomes per genetically predicted increase in each exposure.The power of the MR studies was determined using mRnd online?? https://cnsgenomics.com/shiny/mRnd/; accessed in December 2021..

    The study data were cleaned by means of R language 4.0.0,and statistical analysis was conducted using R language 4.0.0 and PLINK 1.9.To adjust for multiple testing, Benjamini–Hoch berg’s (BH) procedure was used to control the false discovery rate (FDR).The above findings showed limited pleiotropic bias(MR-Egger intercept: P > 0.05; MR-PRESSO global test:PFDR> 0.05).

    2.2.Two-sample MR study based on summary-level data

    2.2.1.Data source

    The GWAS results for the IgG N-glycans based on the above onesample study and GWAS results for metabolic traits based on summary statistics obtained by Biobank Japan?? http://jenger.riken.jp/en/result; accessed in April 2021.? http://www.phenoscanner.medschl.cam.ac.uk/; accessed in September 2022.?? https://biit.cs.ut.ee/gprofiler/snpense; accessed in December 2021.[43]were used to perform two-sample MR analyses.The GWASs incorporated gender, age, the top 10 PCs,and the affection status of 47 diseases as covariates.The sample size of the summary statistics of the GWAS about metabolic traits based on Japanese ancestry populations was 158 284 for BMI,93 146 for FPG, 136 597 for SBP, 136 615 for DBP, 105 597 for TG,128 305 for TC, 70 657 for HDL-C, and 72 866 for LDL-C.The GWAS summary data for CRP used in the above one-sample MR study was also based on Biobank Japan,with a sample size of 75 391.

    2.2.2.Statistical analyses

    The IVs for the IgG N-glycans in the two-sample MR were similar to those in the one-sample MR (significance threshold at P<1×10–5).For each metabolic trait,we selected SNPs associated with the metabolic traits at genome-wide significance thresholds P <5 × 10–8in the two-sample MR study.The process of selecting IVs in the two-sample MR was the same as that in the one-sample MR.Furthermore,in the two-sample MR analysis,the PhenoScanner tool was applied to check and exclude any of the selected SNPs(IVs)associated with other potential phenotypes affecting the outcomes?? http://jenger.riken.jp/en/result; accessed in April 2021.? http://www.phenoscanner.medschl.cam.ac.uk/; accessed in September 2022.?? https://biit.cs.ut.ee/gprofiler/snpense; accessed in December 2021..In the main MR analyses, the IVW method was used to assess the bidirectional causality between IgG N-glycosylation and metabolic traits based on the twosampleMR R package (version 0.5.6).For the two-sample MR analysis of IVW,it was not necessary to consider the problem of sample overlap.The other analyses followed the same processes as those used for the one-sample MR analyses.The BH procedure was used to control the FDR in the MR analysis.The above findings were of limited pleiotropic bias (MREgger intercept: P > 0.05; MR-PRESSO global test: PFDR> 0.05).

    2.3.Meta-analysis and the comparison of the one-and two-sample MR analyses

    If one or both of the one-or two-sample MR results for the same exposure associated with an outcome were statistically significant,we then used IVW meta-analytic technology to combine the IVW method results from the one-and two-sample MR studies(R package meta, version 4.18–2).The combined results are recorded as Beta with a 95% confidence interval (95% CI).Random-effect IVW models were used if heterogeneity existed (P < 0.05); otherwise,a fixed-effect IVW model was used.The use of Bonferroni correction for multiple testing adjusted the significance level to P < 0.05/24 (0.0021) in the meta-analysis.The overlapping potential causal associations of metabolic traits with IgG N-glycans were also compared in the one- and two-sample forward MR studies,and vice versa.

    In this study,a P value of less than 0.05 was regarded as suggestive evidence for a potential association.A P value corrected by the BH procedure to be less than 0.05 or a P value corrected by the Bonferroni method to be less than 0.0021 was regarded as evidence for a potential association.All statistical analyses were implemented using R version 4.0.0 and PLINK 1.9.

    2.4.Bioinformatics analyses

    The annotations of the SNPs of the corresponding located genes were obtained online g:Profiler tool??? http://jenger.riken.jp/en/result; accessed in April 2021.? http://www.phenoscanner.medschl.cam.ac.uk/; accessed in September 2022.?? https://biit.cs.ut.ee/gprofiler/snpense; accessed in December 2021..To functionally annotate putative genetic variants, a functional enrichment analysis for the identified genes was applied using the functional annotation tool Metascape [44] for the genes harboring the IgG N-glycan-QTLs.Enrichment analyses, including the gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes pathway,were performed using gene symbols, with putative genes (P < 0.05) as the input.

    2.5.Explanations for the findings

    As presented in Fig.S1(b) in Appendix A, the causal relationships between IgG N-glycosylation and metabolic traits were depicted as four possible conditions, which were investigated through bidirectional MR analysis.Based on the screened SNPs as IVs,the bidirectional MR analysis(i.e.,forward and reverse MR settings) did not overlap between IgG N-glycosylation and metabolic traits.Four possible explanations were determined from our results:

    ? Explanation 1.One or more genetic variants (SNPs) affect metabolic traits, mediated by IgG N-glycosylation.

    ? Explanation 2.One or more genetic variants (SNPs) affect metabolic traits through alternative biological mechanisms,which then play a downstream role in IgG N-glycosylation.

    ? Explanation 3.There is bidirectional causality between IgG N-glycosylation and metabolic traits through independent biological pathways.

    ? Explanation 4.There is no causal association between IgG N-glycosylation and metabolic traits, and the association between IgG N-glycosylation and metabolic traits may be caused by confounders.

    In Fig.1, we provide an overview of the MR approach used to explore these four explanations.A forward MR study was implemented to evaluate whether the level of IgG N-glycosylation influenced the metabolic trait (Explanation 1).To robustly test Explanation 2, reverse MR analysis was performed to appraise whether the metabolic trait affected the IgG N-glycosylation levels of interest.A P value> 0.05 indicated that Explanation 2 was unlikely in each instance.If there was bidirectional causality between IgG N-glycosylation and metabolic traits, with a P value < 0.05 under both forward and reverse MR settings, then these findings were interpreted as supporting Explanation 3—that is, that there is significant causality between IgG N-glycosylation and metabolic traits in both the forward and reverse MR, and that bidirectional regulation is carried out through different pathogenic pathways.When the P value was > 0.05 in both the forward and reverse MR, these results were identified as supporting Explanation 4.

    3.Results

    3.1.Characteristics of genetic variants as IVs for IgG N-glycans and metabolic traits

    Fig.1.Analysis pipeline to evaluate explanations for observed associations between IgG N-glycosylation and metabolic traits.This flowchart provides an overview of the analysis plan for evaluating four different explanations that might explain trait-associated glycan-QTLs.

    IgG N-glycan-QTLs analysis was conducted to explore the relationship between SNPs and IgG N-glycans in 536 participants with available genetic variants and IgG N-glycosylation data.The mean age of the participants was 48 (range, 43–51) years, comprising 169 males (31.53%) and 367 females (68.47%) (Table S1 in Appendix A).A total of 5661 IgG N-glycan-QTL SNPs and metabolic-QTL SNPs at P < 1 × 10–5were identified, and overlapping SNP as IVs for either IgG N-glycans or metabolic traits were not found.Finally,1167 IgG N-glycan-QTL SNPs were used as IVs for IgG N-glycans in one-sample MR studies, and the mean ± SD of R2and the F-statistics for each QTL were found to be (4.30 ± 0.86)% and 24.04 ± 5.15, respectively (Table S2 in Appendix A).Notably, the number of selected IgG N-glycan-QTL SNPs was reduced from 1167 to 348 after IVs-exposure matching with IVs-outcome in the two-sample MR setting (Table S2).In the one-sample study,647 metabolic-QTL SNPs were used as IVs for metabolic traits,and the mean ± SD of R2and the F-statistics for each QTL were found to be (4.49 ± 0.98)% and 25.18 ± 5.85, respectively(Table S3 in Appendix A).In the two-sample study, 231 metabolic-QTL SNPs were used as IVs for metabolic traits, and the mean ± SD of R2and the F-statistics for each QTL were found to be (0.11 ± 0.28)% and 106.04 ± 240.93, respectively (Table S4 in Appendix A).More details about the number of IgG N-glycan-QTL/metabolic-QTL SNPs used as IVs in the analysis of each phenotype (IgG N-glycans or metabolic traits) in the bidirectional MR analyses are presented in Fig.2 and Tables S2–S4.Furthermore,the mean ± SD proportions of power for the one- and twosample forward MR analyses were found to be (41.11 ± 33.76)%and(21.04±19.34)%,respectively.In addition,the mean±SD proportions of power for the one- and two-sample reverse-direction MR analyses were found to be (50.48 ± 36.31)% and (12.29 ±10.00)%, respectively.More information is available in Table S5 in Appendix A.Moreover, QTL SNPs were used as IVs for IgG Nglycans,metabolic traits,and CRP in the multivariable MR analysis and are presented in Tables S6 and S7 in Appendix A.

    3.2.Inferring putative forward causal associations

    We evaluated the causal estimates of IgG N-glycosylation with metabolic traits through bidirectional MR analysis.For forward MR analysis, IgG N-glycan-QTL SNPs were used as instruments for IgG N-glycans to test the putative causal influences of IgG Nglycosylation on metabolic traits.In the main MR analyses (IVW method), we observed that 66 and 15 genetically determined IgG N-glycans were associated with metabolic traits in the one- and two-sample forward MR studies, respectively (all P < 0.05).The mean±SD proportions of power based on the aforementioned significance statistical results in the one- and two-sample studies were (82.56 ± 15.87)% and (71.42 ± 15.32)%, respectively(Table S5).We appraised the pleiotropic bias and retained only those IgG N-glycans with little evidence of pleiotropy for further investigation (MR-Egger intercept: P > 0.05; MR-PRESSO global test: PFDR> 0.05).After controlling for pleiotropic bias, 59 genetically determined IgG N-glycans were identified to be associated with metabolic traits in the forward one-sample MR analysis (all P < 0.05), including nine putatively causal GPs for BMI (GP1, GP7,GP11, etc.), seven for FPG (GP1, GP5, GP12, etc.), six for SBP(GP10, GP13, GP14, etc.), seven for DBP (GP2, GP12, GP13, etc.),ten for HDL-C (GP1, GP2, GP6, etc.), seven for LDL-C (GP3, GP4,GP12, etc.), seven for TG (GP2, GP5, GP10, etc.), and six for TC(GP3, GP4, GP12, etc.) (Figs.3(a)–(g) and Tables S8–S15 in Appendix A).In parallel, 15 genetically determined IgG N-glycans were associated with metabolic traits in the forward two-sample MR study (all P < 0.05), including five putatively causal GPs for BMI(GP2, GP11, GP13, GP14, and GP24), four for FPG (GP1, GP10,GP16, and GP18), two for SBP (GP6 and GP23), one for DBP(GP6), one for LDL-C (GP11), one for TG (GP6), and one for TC(GP6) (Fig.3(h) and Tables S8–S15).

    Moreover,upon further controlling for pleiotropic bias and FDR,we found that 36 genetically determined IgG N-glycans were associated with metabolic traits in the forward one-sample MR analysis(PFDR< 0.05), including four putatively causal GPs for BMI (GP1,GP12, GP17, and GP22), four for FPG (GP1, GP5, GP13, and GP20),three for SBP (GP13, GP17, and GP20), six for DBP (GP2, GP12,GP13, GP17, GP19, and GP20), four for HDL-C (GP7, GP17, GP20,and GP22), five for LDL-C (GP3, GP4, GP12, GP21, and GP22), five for TG (GP2, GP5, GP10, GP14, and GP22), and five for TC (GP3,GP4, GP12, GP21, and GP22) (Table 1 and Tables S8–S15).No glycan-metabolic trait association was observed in the twosample MR study (PFDR> 0.05) (Tables S8–S15).Effect estimates were broadly consistent between the IVW method and the sensitivity analyses based on robust MR methods in the one-sample MR (WM method, MR-Egger regression, and MR.RAPS method)and two-sample MR settings (WM method and MR-Egger regression) (Table 1 and Tables S8–S15).Furthermore, after controlling for the effect of pleiotropic bias and FDR, the results of the multivariable MR analysis showed that the number of genetically determined IgG N-glycans associated with metabolic traits decreased from 36 (the results of the above univariate MR analysis) to 26 in the forward one-sample setting(Fig.S2 and Table S16 in Appendix A).We further observed that 14 new IgG N-glycans and metabolic traits associations were identified in the multivariable MR analysis compared with the univariate MR results,including one putatively causal GPs for BMI (GP19), four for SBP (GP1, GP5, GP12, and GP15), two for DBP (GP14 and GP15), three for FPG (GP6, GP12,and GP15), two for TC (GP5 and GP7), one for HDL-C (GP4), and one for LDL-C(GP5)(Fig.S3(a)in Appendix A).The effect estimates and more details on specific glycans associated with metabolic traits are presented in Fig.S3 and Table S16 in Appendix A.

    Fig.2.The number of SNPs was included in the analysis of each phenotype in the bidirectional one-and two-sample MR analyses.(a)Forward MR analyses;(b)reverse MR analyses.

    The results of the meta-analysis between one-and two-sample MR identified ten suggestive pieces of evidence for a potential association between IgG N-glycans and metabolic traits(P < 0.05), while only one potential association (GP23 →SBP)was observed (P < 0.0021) in the forward MR analysis (Table S17 in Appendix A).Furthermore, it is worth noting that there were three overlapping suggestive causal associations of metabolic traits with IgG N-glycans (GP11 →BMI, GP13 →BMI, and GP1 →FPG)between the one- and two-sample forward MR studies (P < 0.05)(Table 2).Of these, the results of the meta-analysis further confirmed a positive causal association between GP11 and BMI(fixed-effects model-Beta with 95% CI: 0.0109 (0.0012, 0.0207))(Table S17).Genetically determined GP1 was identified to be associated with FPG in both the one-and two-sample forward MR settings(fixed-effects model-Beta with SE:–0.046(0.016)and–0.006(0.002)); the findings were robust using the WM method and the MR-Egger regression in the one-sample MR analysis (Beta with SE: –0.063 (0.021) and –0.078 (0.030)) and using the WM method in the two-sample MR analysis (Beta with SE: –0.008 (0.003)).In addition, the leave-one-out analysis for the association between GP1 and FPG showed that no outlying genetic variant had a significant influence on the estimate (Figs.S4 and S5 in Appendix A).

    3.3.Inferring putative reverse causal associations

    Taking these putative associations forward, we assessed the potential causal associations by conducting a reverse MR of the metabolic measurements against the IgG N-glycosylation levels by selecting the metabolic-QTLs as IVs for metabolic measurements.For the reverse MR analysis, in the main MR analyses(IVW method),82 and four genetically determined metabolic traits were identified as being related to the risk of IgG N-glycosylation in the one- and two-sample MR studies, respectively (all P < 0.05).The mean ± SD proportions of power based on the aforementioned significance statistical results in the one-sample and two-sample studies were found to be (87.77 ± 15.32)% and(46.87 ± 9.87)%, respectively (Table S5).After application of the pleiotropic bias test (MR-Egger intercept: P > 0.05 ; MR-PRESSO global test: PFDR> 0.05), 72 metabolic trait–glycan associations were retained in the reverse one-sample MR study (all P < 0.05),including one putatively causal metabolic trait for GP1 (HDL-C),five for GP2 (FPG, SBP, HDL-C, etc.), two for GP3 (LDL-C and TC),four for GP4 (FPG, LDL-C, etc.), two for GP5 (FPG and SBP), five for GP6 (FPG, HDL-C, etc.), two for GP7 (FPG and DBP), two for GP8 (HDL-C and TG), three for GP9 (FPG, HDL-C, and LDL-C), two for GP10(FPG and TG),three for GP11(HDL-C,LDL-C,and TG),four for GP12 (FPG, DBP, etc.), four for GP13 (FPG, SBP, etc.), four for GP14 (FPG, HDL-C, etc.), one for GP15 (FPG), four for GP16 (FPG,TG, etc.), five for GP17 (BMI, FPG, etc.), five for GP18 (FPG, HDL-C,etc.), two for GP19 (BMI and TG), three for GP20 (FPG, DBP, and HDL-C), one for GP21 (TG), five for GP22 (BMI, DBP, etc.), one for GP23 (TC), and two for GP24 (HDL-C and TG) (Figs.4(a)–(g) and Tables S18–S25 in Appendix A).In addition,four genetically determined metabolic traits were retained that were associated with IgG N-glycans in the reverse two-sample MR study (all P < 0.05),including one putatively causal metabolic trait for GP3 (HDL-C),one for GP9 (HDL-C), one for GP16 (TC), and one for GP20 (LDLC) (Fig.4(h) and Tables S18–S25).

    Fig.3.Causal associations of IgG N-glycosylation with metabolic traits from the primary analysis (IVW) in the forward MR estimates.In one-sample MR analysis for(a) glycans and TC, (b) glycans and HDL-C, (c) glycans and LDL-C, (d) glycans and TG, (e) glycans and BMI, (f) glycans and BP, and (g) glycans and FPG; in two-sample MR analysis for (h) glycans and metabolic traits.P < 0.05 was considered statistically significant.

    Furthermore, to further control the FDR, the number of metabolic trait–glycan associations was reduced to 57 in the reverse one-sample MR analysis (PFDR< 0.05), including four putatively causal metabolic traits for GP2 (FPG, HDL-C, etc.), one for GP3(LDL-C), four for GP4 (FPG, LDL-C, etc.), two for GP5 (FPG and SBP), four for GP6 (FPG, HDL-C, etc.), two for GP7 (FPG and DBP),two for GP8 (HDL-C and TG), two for GP9 (FPG and HDL-C), one for GP10 (TG), two for GP11 (HDL-C and TG), four for GP12 (FPG,DBP, etc.), four for GP13 (FPG, SBP, etc.), four for GP14 (FPG,HDL-C, etc.), one for GP15 (FPG), three for GP16 (DBP, HDL-C,and TG), three for GP17 (BMI, FPG, and LDL-C), five for GP18(FPG, HDL-C, etc.), one for GP19 (TG), three for GP20 (FPG, DBP,and HDL-C), one for GP21 (TG), one for GP22 (TG), one for GP23(TC), and two for GP24 (HDL-C and TG) (Tables S18–S25).Similar to the condition of the forward two-sample MR, no metabolic trait–glycan association was observed in the reverse two-sample MR study (PFDR> 0.05) (Table 3 and Tables S18–S25).Effect estimates were broadly consistent between the IVW method and the sensitivity analyses based on robust MR methods in the onesample MR (WM method, MR-Egger regression, and MR.RAPS method) and two-sample MR settings (WM method and MREgger regression) (Table 3 and Tables S18–S25).

    In addition,in the reverse one-sample MR setting,we observed that the number of metabolic trait–glycan associations was reduced from 57 (i.e., the aforementioned univariate MR results)to 41 based on the multivariable MR analysis (Table S16 and Fig.S6 in Appendix A).More specifically,upon controlling for pleiotropic bias and FDR, the results based on the multivariate MR method identified 12 new metabolic trait–glycan associations compared with those of the univariate MR analysis, including two putatively causal metabolic traits for GP7 (SBP and HDL-C),two for GP11 (FPG and TC), one for GP12 (SBP), one for GP13(HDL-C), one for GP14 (TG), two for GP15 (SBP and HDL-C), one for GP17 (HDL-C), one for GP20 (SBP), and one for GP22 (FPG)(Fig.S3(b)).Notably, the number of meaningful results between IgG N-glycans and metabolic traits obtained by the multivariate MR method was generally lower than that of the aforementioned univariate analysis in the bidirectional one-sample MR settings,but most of them could be replicated (Figs.3 and 4 and Figs.S2 and S6).

    According to the results of the meta-analysis between the oneand two-sample MR, 45 potential associations of metabolic traitswith IgG N-glycans were identified (P < 0.0021) in the reverse MR analysis (Table S26 in Appendix A).Notably, there was only one overlapping potential causal association of metabolic traits with IgG N-glycans between the one-and two-sample reverse MR studies (P < 0.05; Table 2).Genetically determined HDL-C was identified to be negatively associated with GP9 (fixed-effects model-Beta with SE: –0.071 (0.022) and –0.306 (0.151)) in both the one- and two-sample MR settings, which was confirmed by the results of the meta-analysis (fixed-effects model-Beta with 95%CI: –0.0759 (–0.1186, –0.0332)) (Table S26).

    Table 2 Overlapping potential causal associations of metabolic traits with IgG N-glycans between one- and two-sample MR studies.

    Fig.4.Causal associations of metabolic traits and IgG N-glycosylation from the primary analysis(IVW)in the reverse MR estimates.In one-sample MR analysis for(a)TC and glycans, (b) HDL-C and glycans, (c) LDL-C and glycans, (d) TG and glycans, (e) BP and glycans, (f) BMI and glycans, and (g) FPG and glycans; in two-sample MR analysis for(h) lipids and glycans.P < 0.05 was considered statistically significant.

    3.4.Functional informatics

    Although it is widely hypothesized that genetic variants influence glycosylation, the molecular regulatory mechanism of IgG N-glycosylation is not fully understood.In this research, IgG N-glycan-QTL SNPs were annotated to 553 unique located genes(Tables S2 and S27 in Appendix A).A GO enrichment analysis,which was intended to identify molecular function, biological process, and cellular component pathways, suggested that the identified genes were involved in 20 GO terms (Fig.5), such aschemical synaptic transmission, regulation of transmembrane transport, cell junction organization, receptor localization to synapse, and regulation of membrane potential.

    Table 3 (continued)

    Some of the tagged genes are known to encode glycosylation transferases, including EXT2, GMDS, ST6GAL1, B4GALT1, and UTS2,which are involved in the N-glycan biosynthesis/complex type.In Table S28 in Appendix A, the results show that most of the genes are involved in positive regulation of chemical synaptic transmission, cellular component morphogenesis, cell junction organization, and actin filament-based processes (7.62%, 7.62%, 7.06%, and 7.25%, respectively), indicating that these genes—such as MCTP2,

    RIMS1, ITSN1, GRIN3A, DLGAP2, DLGAP1, TTN, ACTC1, ANTXR1,HOMER1, PTK2B, MYCBP2, PRKN, IFT172, and PTPRD—might play important roles in modifying IgG glycosylation.Further functional experiments are needed to reveal the molecular regulatory mechanism of the above pathways in IgG N-glycosylation.Our research might prove valuable in helping subsequent studies prioritize candidate genes of IgG N-glycosylation for in-depth functional followup studies.

    4.Discussion

    This study adopted a novel analytical framework in which bidirectional one- and two-sample MR analyses integrating GWASs and IgG N-glycan-QTLs data enabled the identification of causal associations between IgG N-glycosylation and metabolic traits in East Asians.In fact, a one-sample MR study is usually underpowered in a setting with small effect sizes and may result in a type II error.Although a two-sample MR study with a large sample size is unlikely to suffer from the aforementioned bias, it may present bias in terms of population stratification.To address these issues,one-and two-sample MR methods were used in the current study in order to comprehensively investigate causality.Suggestive evidence for bidirectional causality between IgG N-glycans and metabolic traits in the current study might have been supported by the one-and two-sample MR results based on the data from a previous glycan-related study [33] and a large-scale GWAS in a Japanese population [43].However, the relationship between each pair of IgG N-glycans and metabolic traits was not completely consistent in the different settings of MR analyses.In the forward MR analysis,the one-sample MR provided potential evidence that IgG N-glycans were associated with BMI,FPG,BP,and lipid traits.This result was confirmed in the two-sample setting,further evidencing the causal relationship between IgG N-glycans and metabolic traits.Furthermore, the number of meaningful associations was generally found to be lower in the multivariate MR analysis than in the univariate analysis in the bidirectional one-sample MR setting; however,most of them could be replicated.The multivariate MR analysis further identified some new metabolic trait–glycan/(glycan–metabolic trait) associations, indicating that an inflammatory molecule (CRP) may mediate the causality between IgG N-glycans and metabolic traits to some extent.

    The main function of IgG—which is largely affected by glycosylation—is regulation of the systemic inflammation balance at multiple levels,and the alteration of IgG N-glycans plays a crucial role in the pro- or anti-inflammatory process [6,45].An imbalance between pro- and anti-inflammatory activities leads to the activation of multiple signal transduction pathways, the production of inflammatory cytokine chemokines, and cell migration, which in sequence cause the development of metabolic dysfunction[46,47].In addition, due to the key role of IgG N-glycans in signal transduction [48], the levels of IgG N-glycans directly and objectively reflect the metabolism status.A previous study showed that the plasma N-glycome could represent alterations in metabolic activity in humans [48], and our forward MR results further demonstrate that IgG N-glycans may have causal effects on metabolic processes.

    Fig.5.Characteristics of the enriched GO terms.

    Significant associations between metabolic traits and IgG Nglycans were also observed in the reverse MR analysis in both the one- and two-sample settings, which indicates that metabolic traits may also have causal effects on the IgG N-glycosylation process.Glycosylation is reflected at the molecular level in a wide range of genetic influences and environmental exposures[29,49,50].Alternative IgG N-glycosylation plays a prominent role in complement activation by modulating complement-dependent cytotoxicity and antibody-dependent cellular cytotoxicity (ADCC),which indicates that altered IgG N-glycosylation could be a part of the molecular mechanism underlying pro-inflammation [7,45].Chronic inflammation—an important hallmark of metabolic dysfunction—can regulate metabolic function and trigger metabolic disorders [51–53].Metabolic dysfunction involves insulin resistance, abnormal lipid metabolism, an impaired immune response affecting both innate and adaptive immunity,and regulation of cytokine secretion, which can produce and secrete a variety of inflammatory molecules, including interleukin-1 receptor-associated kinase 1 and tumor necrosis factor-α [51,53].Inflammatory molecules might in turn regulate the IgG N-glycosylation process.The findings of the reverse MR suggest that metabolic traits might affect IgG N-glycosylation,which influences the immune and inflammatory function of IgG.

    The study showed that three pairs of glycan–metabolic trait associations (i.e., GP11 →BMI, GP13 →BMI, and GP1 →FPG)and one pair of metabolic trait–glycan associations (i.e.,HDL-C →GP9) were statistically significant in the bidirectional one- and two-sample MR estimates, respectively.Interestingly,the results of the meta-analysis provided further evidence of potential bidirectional causal associations between glycans and metabolic traits (i.e., GP11 →BMI and HDL-C →GP9).The results showed that the effect size of the aforementioned one-sample MR was generally larger than that of the two-sample MR, which might be caused by the fact that there were more IVs included in the one-sample MR than in the two-sample MR setting.To be specific, the findings showed completely consistent evidence for higher levels of IgG N-glycans (GP11) leading to higher levels of BMI and for higher levels of IgG N-glycans (GP1) leading to lower levels of FPG in the forward one- and two-sample MR analyses.In the reverse one- and two-sample MR analyses, completely consistent evidence was presented showing that lower levels of HDL-C lead to higher levels of IgG N-glycans(GP9).IgG N-glycans contain typical biantennary complexes—namely, fucose, galactose, bisecting GlcNAc, and sialic acid—which can inhibit or activate IgG Fcγ receptors by regulating the binding affinity of IgG, thereby decisively modulating the pro- and anti-inflammatory activities of IgG[54,55].Fucosylation(GP1,GP9,and GP11)and galactosylation(GP9 and GP11) play vital roles in the inflammatory activities of IgG, and highly fucosylated IgG presents a reduced magnitude of ADCC leading to an anti-inflammatory status of the body.The inverse correlation found between galactosylated and/or corefucosylated IgG glycoforms (GP1 and GP9) and metabolic traits(FPG and HDL-C) in the present study was somewhat expected and was consistent with the alterations of IgG N-glycosylation presented in studies on hypertension [14], T2DM [16], and dyslipidemia [13].

    Metabolic trait components include BMI, FPG, BP, and lipids,with the corresponding abnormal conditions being obesity,T2DM,hypertension,and dyslipidemia.These abnormal conditions have been documented as modifiable risk factors for cerebrovascular and cardiovascular diseases (CVDs) and are related to an increased risk of CVDs [56].In addition, abundant evidence suggests that the inflammatory function of IgG N-glycans is a risk factor for many chronic diseases and accompanies the development of specific health issues, including obesity, hypertension, T2DM,metabolic syndrome, dyslipidemia, and ischemic stroke [12–15,42,48,57].The forward MR in the present study confirmed that GPs have a potential causal effect on metabolic traits.Conversely,the causal associations between metabolic traits and GPs found in the reverse MR suggest that IgG N-glycans are likely to be driven by metabolic traits.Therefore,the effects of the bidirectional association that was found between IgG N-glycans and metabolic traits might be attributed to Explanation 3—that is,that there is bidirectional causality between IgG N-glycosylation and metabolic traits through independent biological pathways—further suggesting that reciprocal regulation and coexistence between IgG N-glycans and metabolic traits might accelerate the progression of CVDs.

    The comprehensive IgG N-glycan-QTL resources utilized in our study present plentiful information about genetic effects on IgG N-glycans.Apart from some identified IgG N-glycan-QTL tagging genes that are known to be glycosylation enzymes regulating translational initiation,some novel loci tagging genes are involved in glycosylation via metabolic mechanisms rather than directly.For example, INSR, which encodes a member of the receptor tyrosine kinase family of proteins, regulates glucose uptake and release and activates the insulin signaling pathway[58].In addition,some genes are involved in the regulation of the lipid metabolism pathway,such as elongation of very long-chain fatty acid(ELOVL)elongase 6 gene (ELOVL6) [59] and lipase C, hepatic type gene (LIPC)[60].These novel genes are very biologically relevant in metabolic disorders,which suggests mechanisms for the regulation of IgG Nglycans and indicates that certain network connections need to be validated in in vitro functional studies.

    The present bidirectional MR further dissected the directionally causal association between IgG N-glycans and metabolic traits,which may allow networks of related phenotypes.Because QTLs were shown to have a robust effect on exposure, our bidirectional MR analysis integrating a GWAS and QTL data may have been relatively less affected by bias, and the use of multiple independent QTL variants as IVs may have increased the proportion of the variance in exposures (i.e., IgG N-glycans or metabolic traits).Nevertheless, this study has several limitations, and the findings should be interpreted with caution.Firstly,there is inevitable concern that the small sample size in the one-sample MR analysis is likely to underpower the detection of genetic associations for IgG N-glycosylation,as judged by the latest studies in European populations.Admittedly, the one- and two-sample MR have their own strengths and limitations.Thus,one-and two-sample MR analyses were performed to confirm our findings,and the results were complementary,suggesting that the finding is robust.In addition,compared with recently published studies [61,62], the statistical powers of the present study were acceptable, at 82.56% and 71.42% in the one- and two-sample forward MR analyses, respectively,and 87.77%and 46.87%in the one- and two-sample reverse MR settings, respectively.The number of IgG N-glycan-QTL SNPs used in the one-sample MR analysis(n=1167)was larger than that used in the two-sample MR analysis(n = 348), which might result in the statistical power being greater in the one-sample MR study than in the two-sample MR study.Secondly,for the limited sample size of IgG N-glycan-QTLs (n = 536), a relatively relaxed threshold of 1×10–5was used for the selection of IVs in the one-sample MR analysis, which may have led to biased estimates—especially a weak IV bias.To reduce weak IVs, the MR.RAPS results were also reported to make the results more robust [41].Thirdly, the research data used in this study were obtained from Xuanwu Hospital, Capital Medical University, Beijing, China and Biobank Japan, which were a community-based survey and a patientbased cohort study, respectively.This choice of data may have led to potential selection bias when subjects were included in this study, which may have affected the results of the study to some extent.

    In this study, only the suggestive causal associations of three metabolic traits with four IgG N-glycans detected in the onesample MR round were considered to be statistically significant in the two-sample MR, and two pairs of causality were confirmed by meta-analysis.A possible explanation for the phenomenon that the one-sample MR results were not fully confirmed by the twosample MR results is that bias estimation was caused by population stratification in the two-sample MR study or by false positives in the one-sample MR study.Although we analyzed data based on the same studies in obtaining both SNP-IgG N-glycosylation and SNP-metabolic trait estimates while adjusting for the same covariates, other unrecognized or unmeasured confounding factors still cannot be completely excluded.Moreover, although pleiotropy was assessed by MR-Egger and MR-PRESSO analysis,the possibility of residual pleiotropic effects cannot be completely excluded.In addition, we only involved East Asian participants in the current study, so population stratification had little effect on the result.Nevertheless, our findings may be limited in their generalization to other ancestry groups.Therefore, it is necessary to conduct larger MR studies in multiethnic populations in order to examine causal inferences to further validate and replicate the association between IgG N-glycans and metabolic traits for personalized health monitoring and the prevention of metabolic-associated diseases.Finally, one of the challenges presented by all MR analyses is to completely eliminate an alternative direct causal effect—in this case, for IgG N-glycosylation and metabolic traits determined by both diverse environmental exposures and genetic variants.

    5.Conclusions

    In summary,the comprehensively bidirectional MR results presented here suggest potential bidirectional causality between IgG N-glycans and metabolic traits, and vice versa.According to our result, this causality might apply bidirectional regulation based on independent pathogenic mechanisms or pathways.The comprehensive IgG N-glycan-QTL resources used in this study and the potential causal metabolic traits uncovered herein delineate for the first time a richness of detail in relation to the regulation of IgG N-glycosylation, which may provide a theoretical basis for the future development and application of glycomedicine in metabolic diseases.A large-scale genome-wide scan for genetic variants of IgG N-glycan-related metabolites and further investigation to understand the molecular regulatory mechanism of IgG Nglycan-dependent metabolites in the development of CVDs are required.

    Acknowledgments

    The study was supported by grants from the National Natural Science Foundation of China(81872682),the Young Taishan Scholars Program of Shandong Province of China (tsqn20161046), the Academic Promotion Programme of Shandong First Medical University (2019RC010), the Shandong Province Higher Educational Young and Innovation Technology Supporting Program(2019KJL004), and the Doctoral Scientific Research Foundation of Shandong First Medical University.The funding organization had the role in the design and conduct of the study and the collection,management, analysis, and interpretation of the data.

    We thank all the research subjects for their participation.The Riken group reported the summary data.We would also like to acknowledge the skillful work of all medical staff at Xuanwu Hospital, Capital Medical University, Beijing, China.

    Authors’ contribution

    Youxin Wang and Manshu Song contributed to the conception and design.Xiaomi Meng,Weijie Cao,Di Liu,Isinta Maranga Elijah,Weijia Xing,Haifeng Hou,Xizhu Xu,and Manshu Song contributed to the acquisition and analysis of the data.Xiaomi Meng, Weijie Cao,and Di Liu drafted the manuscript.All authors made important contributions to editing and critically revising the manuscript for important intellectual content.All authors have approved the final version to be published.Youxin Wang and Manshu Song guarantee this work,have full access to all of the data and take responsibility for the integrity of the data.

    Compliance with ethics guidelines

    Xiaomi Meng, Weijie Cao, Di Liu, Isinta Maranga Elijah, Weijia Xing, Haifeng Hou, Xizhu Xu, Manshu Song, and Youxin Wang declare that they have no conflict of interest or financial conflicts to disclose.

    Appendix A.Supplementary data

    Supplementary data to this article can be found online at https://doi.org/10.1016/j.eng.2022.11.004.

    国产精品一区二区免费欧美| 久久中文字幕一级| 亚洲 欧美一区二区三区| 黑人猛操日本美女一级片| 国产单亲对白刺激| 夜夜夜夜夜久久久久| xxxhd国产人妻xxx| 88av欧美| 最好的美女福利视频网| 欧美日韩av久久| 99香蕉大伊视频| 亚洲欧美日韩另类电影网站| 国产一区二区三区在线臀色熟女 | 99精国产麻豆久久婷婷| 亚洲在线自拍视频| 手机成人av网站| 91大片在线观看| 欧美中文综合在线视频| √禁漫天堂资源中文www| 亚洲色图综合在线观看| 在线观看舔阴道视频| 国产精品一区二区免费欧美| 夫妻午夜视频| 免费在线观看影片大全网站| 亚洲 欧美一区二区三区| 窝窝影院91人妻| 50天的宝宝边吃奶边哭怎么回事| 色综合欧美亚洲国产小说| 欧美成人免费av一区二区三区| 国产区一区二久久| 精品国产一区二区久久| 日韩有码中文字幕| 午夜精品在线福利| 亚洲黑人精品在线| 超色免费av| 波多野结衣一区麻豆| 男人操女人黄网站| 国产精品偷伦视频观看了| av网站免费在线观看视频| 中亚洲国语对白在线视频| 又大又爽又粗| 如日韩欧美国产精品一区二区三区| 久久精品亚洲av国产电影网| 午夜免费观看网址| www国产在线视频色| 18禁裸乳无遮挡免费网站照片 | 欧美日韩瑟瑟在线播放| 国产亚洲精品一区二区www| 一本大道久久a久久精品| 国产成+人综合+亚洲专区| 高清欧美精品videossex| 亚洲午夜精品一区,二区,三区| 侵犯人妻中文字幕一二三四区| 亚洲五月色婷婷综合| 精品少妇一区二区三区视频日本电影| 国产成人欧美| 天堂俺去俺来也www色官网| 国产aⅴ精品一区二区三区波| 免费看十八禁软件| 亚洲欧美一区二区三区黑人| 一个人免费在线观看的高清视频| 日韩成人在线观看一区二区三区| 激情视频va一区二区三区| 国产一区二区三区视频了| 国产欧美日韩一区二区三区在线| 精品久久久久久,| 午夜福利欧美成人| 久久久久久久久中文| 成年人黄色毛片网站| 99国产精品一区二区蜜桃av| 大型av网站在线播放| 亚洲精品粉嫩美女一区| 国产精品爽爽va在线观看网站 | 亚洲欧美日韩无卡精品| 另类亚洲欧美激情| 两个人免费观看高清视频| 午夜a级毛片| 亚洲久久久国产精品| 美女高潮到喷水免费观看| 男女之事视频高清在线观看| 啪啪无遮挡十八禁网站| 一边摸一边抽搐一进一小说| 一级片免费观看大全| 亚洲国产欧美一区二区综合| 久久香蕉激情| 亚洲第一欧美日韩一区二区三区| 亚洲 欧美一区二区三区| 在线观看www视频免费| 亚洲国产看品久久| cao死你这个sao货| 亚洲欧洲精品一区二区精品久久久| 国产成+人综合+亚洲专区| 99久久人妻综合| 丝袜美腿诱惑在线| 亚洲一码二码三码区别大吗| 亚洲av片天天在线观看| 淫秽高清视频在线观看| 欧美激情极品国产一区二区三区| 欧美老熟妇乱子伦牲交| 他把我摸到了高潮在线观看| 免费在线观看完整版高清| 国产午夜精品久久久久久| 亚洲性夜色夜夜综合| 国产精品久久久久久人妻精品电影| 在线看a的网站| 亚洲五月天丁香| 久久久久九九精品影院| 人人妻人人澡人人看| 亚洲成a人片在线一区二区| 亚洲一区高清亚洲精品| 欧美另类亚洲清纯唯美| 亚洲在线自拍视频| 精品国产国语对白av| 正在播放国产对白刺激| 最近最新中文字幕大全电影3 | 手机成人av网站| 九色亚洲精品在线播放| 国产99白浆流出| 女人高潮潮喷娇喘18禁视频| 国产不卡一卡二| 88av欧美| 丝袜美腿诱惑在线| 黄片大片在线免费观看| 欧美一级毛片孕妇| 欧美久久黑人一区二区| 成人精品一区二区免费| 久久精品亚洲熟妇少妇任你| 久久精品91蜜桃| 国产三级在线视频| 亚洲av美国av| 丝袜美腿诱惑在线| 日韩人妻精品一区2区三区| 国产精品久久久久久人妻精品电影| 大陆偷拍与自拍| 亚洲七黄色美女视频| 一级,二级,三级黄色视频| 母亲3免费完整高清在线观看| 高清毛片免费观看视频网站 | 一级a爱视频在线免费观看| 欧美日韩av久久| 免费在线观看影片大全网站| 国产精品香港三级国产av潘金莲| 少妇被粗大的猛进出69影院| 老汉色av国产亚洲站长工具| 欧美成人性av电影在线观看| 超色免费av| 97超级碰碰碰精品色视频在线观看| 在线播放国产精品三级| 亚洲中文日韩欧美视频| 岛国在线观看网站| 12—13女人毛片做爰片一| 国产欧美日韩一区二区三| 黄片大片在线免费观看| 在线观看免费高清a一片| 亚洲全国av大片| 91精品三级在线观看| 久9热在线精品视频| 男女高潮啪啪啪动态图| 精品国产超薄肉色丝袜足j| 国产精品一区二区三区四区久久 | 男人舔女人的私密视频| 日日干狠狠操夜夜爽| 久久久久国产一级毛片高清牌| 亚洲一区中文字幕在线| 国产91精品成人一区二区三区| 91国产中文字幕| 亚洲欧美日韩高清在线视频| 乱人伦中国视频| 免费人成视频x8x8入口观看| 成人影院久久| 波多野结衣高清无吗| 久久99一区二区三区| 欧洲精品卡2卡3卡4卡5卡区| 老司机在亚洲福利影院| 一二三四社区在线视频社区8| 亚洲人成电影免费在线| 9191精品国产免费久久| 男人舔女人下体高潮全视频| 国产精品久久视频播放| 丁香六月欧美| 91成人精品电影| 久久精品成人免费网站| 亚洲avbb在线观看| 制服人妻中文乱码| 一区二区三区国产精品乱码| 亚洲午夜精品一区,二区,三区| 亚洲一区二区三区色噜噜 | 黄色女人牲交| 久久久久久大精品| 久久精品91无色码中文字幕| 国产麻豆69| 久久国产精品人妻蜜桃| 曰老女人黄片| www.自偷自拍.com| 99re在线观看精品视频| 欧美日韩中文字幕国产精品一区二区三区 | 中文字幕人妻熟女乱码| 免费一级毛片在线播放高清视频 | 一二三四社区在线视频社区8| 两人在一起打扑克的视频| 久久 成人 亚洲| 女人高潮潮喷娇喘18禁视频| 这个男人来自地球电影免费观看| 日韩 欧美 亚洲 中文字幕| av超薄肉色丝袜交足视频| 欧洲精品卡2卡3卡4卡5卡区| 757午夜福利合集在线观看| 午夜福利一区二区在线看| bbb黄色大片| 欧美日韩国产mv在线观看视频| 波多野结衣av一区二区av| 一级,二级,三级黄色视频| av片东京热男人的天堂| 国产99久久九九免费精品| 男女午夜视频在线观看| 久久久久久久精品吃奶| 淫妇啪啪啪对白视频| 亚洲精品成人av观看孕妇| 日本欧美视频一区| 夜夜爽天天搞| 国产单亲对白刺激| 欧美人与性动交α欧美软件| 男女床上黄色一级片免费看| 成年女人毛片免费观看观看9| 午夜激情av网站| 欧美av亚洲av综合av国产av| 国产av又大| 精品卡一卡二卡四卡免费| 日本欧美视频一区| 欧美日韩一级在线毛片| 亚洲av五月六月丁香网| 欧美黄色淫秽网站| 天堂俺去俺来也www色官网| 欧美日韩亚洲综合一区二区三区_| 制服人妻中文乱码| 村上凉子中文字幕在线| 国产色视频综合| 韩国av一区二区三区四区| 亚洲中文字幕日韩| 在线观看免费视频日本深夜| 美女大奶头视频| 91麻豆精品激情在线观看国产 | 可以免费在线观看a视频的电影网站| 淫妇啪啪啪对白视频| 一边摸一边做爽爽视频免费| 国产在线观看jvid| 黑丝袜美女国产一区| 国产精品秋霞免费鲁丝片| 午夜免费鲁丝| 麻豆成人av在线观看| 国产在线观看jvid| 成在线人永久免费视频| 欧美日韩福利视频一区二区| av天堂久久9| 亚洲色图av天堂| 精品人妻1区二区| 久久香蕉国产精品| 91麻豆av在线| 大型黄色视频在线免费观看| 亚洲一区二区三区欧美精品| 国产伦一二天堂av在线观看| 亚洲 国产 在线| 日韩欧美三级三区| 老鸭窝网址在线观看| 两性夫妻黄色片| 成熟少妇高潮喷水视频| 国产亚洲精品一区二区www| 亚洲精品中文字幕在线视频| 国产不卡一卡二| 免费观看人在逋| 成年人黄色毛片网站| 高清av免费在线| 国产亚洲精品综合一区在线观看 | 女同久久另类99精品国产91| 亚洲全国av大片| 婷婷丁香在线五月| 亚洲va日本ⅴa欧美va伊人久久| 亚洲第一av免费看| 99国产综合亚洲精品| 黄色片一级片一级黄色片| 国产人伦9x9x在线观看| 免费日韩欧美在线观看| 久久欧美精品欧美久久欧美| 亚洲片人在线观看| 国产伦人伦偷精品视频| 精品久久久久久电影网| 老司机午夜十八禁免费视频| 欧美成人免费av一区二区三区| 国产免费男女视频| 村上凉子中文字幕在线| 丝袜美腿诱惑在线| 两人在一起打扑克的视频| 精品高清国产在线一区| 亚洲全国av大片| 国产亚洲欧美在线一区二区| 日本一区二区免费在线视频| 在线播放国产精品三级| 色婷婷久久久亚洲欧美| 天天影视国产精品| 十八禁人妻一区二区| 亚洲欧美激情综合另类| 成人精品一区二区免费| 黑人巨大精品欧美一区二区蜜桃| 久久香蕉国产精品| 美女 人体艺术 gogo| 亚洲国产欧美网| 婷婷六月久久综合丁香| 中出人妻视频一区二区| 亚洲欧美日韩无卡精品| 久99久视频精品免费| 麻豆成人av在线观看| 亚洲av熟女| 国产精品野战在线观看 | 精品久久久久久成人av| 午夜影院日韩av| 一个人观看的视频www高清免费观看 | 五月开心婷婷网| 久9热在线精品视频| 黑人欧美特级aaaaaa片| 黑人巨大精品欧美一区二区mp4| 99re在线观看精品视频| 欧美日本中文国产一区发布| 欧美日韩乱码在线| 美女 人体艺术 gogo| 99久久99久久久精品蜜桃| 婷婷精品国产亚洲av在线| 一夜夜www| 三级毛片av免费| 中文字幕精品免费在线观看视频| 亚洲精品一区av在线观看| 欧美亚洲日本最大视频资源| 亚洲成人国产一区在线观看| 欧美日韩福利视频一区二区| 免费在线观看视频国产中文字幕亚洲| 久久亚洲精品不卡| 交换朋友夫妻互换小说| 国内久久婷婷六月综合欲色啪| 又紧又爽又黄一区二区| 99热只有精品国产| 日本一区二区免费在线视频| 美女福利国产在线| 成人三级黄色视频| 亚洲狠狠婷婷综合久久图片| 色婷婷av一区二区三区视频| 成人永久免费在线观看视频| 国产在线观看jvid| 日韩av在线大香蕉| 免费av中文字幕在线| 变态另类成人亚洲欧美熟女 | 亚洲成人久久性| 国产黄a三级三级三级人| 亚洲色图av天堂| 老熟妇仑乱视频hdxx| 成在线人永久免费视频| 久久久久国产一级毛片高清牌| 老汉色av国产亚洲站长工具| 国产不卡一卡二| 国产日韩一区二区三区精品不卡| 又黄又粗又硬又大视频| 亚洲精品一区av在线观看| av有码第一页| 国产麻豆69| 看黄色毛片网站| 亚洲成人免费av在线播放| 最近最新中文字幕大全免费视频| x7x7x7水蜜桃| 亚洲专区字幕在线| 天天影视国产精品| 国产精品久久视频播放| 搡老乐熟女国产| 女人被躁到高潮嗷嗷叫费观| 91成年电影在线观看| 老司机午夜福利在线观看视频| 国产精品99久久99久久久不卡| 久久天堂一区二区三区四区| tocl精华| 一区二区三区国产精品乱码| 亚洲国产毛片av蜜桃av| 亚洲国产欧美一区二区综合| 人人澡人人妻人| 午夜a级毛片| 在线看a的网站| 成人三级做爰电影| 在线观看66精品国产| 18禁国产床啪视频网站| 黄片大片在线免费观看| 久久国产精品影院| av在线天堂中文字幕 | 色尼玛亚洲综合影院| 亚洲五月色婷婷综合| 国产精品1区2区在线观看.| 一级黄色大片毛片| 国产精品国产av在线观看| 国产精华一区二区三区| tocl精华| 亚洲人成电影观看| 国产亚洲欧美精品永久| 99国产精品一区二区三区| 精品国产美女av久久久久小说| 国产欧美日韩一区二区精品| 欧美激情久久久久久爽电影 | 欧美+亚洲+日韩+国产| 久久香蕉国产精品| 很黄的视频免费| 最近最新免费中文字幕在线| 黄色毛片三级朝国网站| 两个人免费观看高清视频| 久久这里只有精品19| 久久香蕉精品热| 嫩草影院精品99| 多毛熟女@视频| 亚洲精品一区av在线观看| 母亲3免费完整高清在线观看| 国产成人精品久久二区二区免费| 免费在线观看完整版高清| 精品欧美一区二区三区在线| 色哟哟哟哟哟哟| 丝袜美腿诱惑在线| 国产在线精品亚洲第一网站| 亚洲午夜精品一区,二区,三区| 老司机在亚洲福利影院| 日本免费一区二区三区高清不卡 | 午夜免费成人在线视频| 亚洲一区二区三区欧美精品| 国产亚洲精品一区二区www| 久久天堂一区二区三区四区| 精品午夜福利视频在线观看一区| 妹子高潮喷水视频| 免费高清在线观看日韩| 黄色丝袜av网址大全| 亚洲伊人色综图| 亚洲色图 男人天堂 中文字幕| 91在线观看av| 久久精品国产清高在天天线| 91麻豆av在线| 97碰自拍视频| 青草久久国产| 日韩国内少妇激情av| 日韩欧美一区二区三区在线观看| 久99久视频精品免费| 亚洲欧美日韩另类电影网站| 中文字幕人妻丝袜制服| 久久青草综合色| 国产主播在线观看一区二区| 色婷婷av一区二区三区视频| 女同久久另类99精品国产91| 在线天堂中文资源库| 母亲3免费完整高清在线观看| 丝袜美腿诱惑在线| 国产精品成人在线| 自拍欧美九色日韩亚洲蝌蚪91| 一级毛片高清免费大全| 美女午夜性视频免费| 亚洲一卡2卡3卡4卡5卡精品中文| 制服人妻中文乱码| 亚洲一区中文字幕在线| 又黄又爽又免费观看的视频| 日本黄色日本黄色录像| 亚洲欧美精品综合久久99| 成年版毛片免费区| 国产午夜精品久久久久久| 两个人看的免费小视频| 不卡一级毛片| 免费看十八禁软件| 国产乱人伦免费视频| 精品免费久久久久久久清纯| 丝袜美足系列| 亚洲精品粉嫩美女一区| 精品久久久久久久久久免费视频 | 麻豆成人av在线观看| 日韩三级视频一区二区三区| 精品午夜福利视频在线观看一区| 又紧又爽又黄一区二区| 久久精品人人爽人人爽视色| 午夜成年电影在线免费观看| 亚洲欧美激情在线| 日韩人妻精品一区2区三区| 国产熟女xx| 亚洲,欧美精品.| 国产精品亚洲av一区麻豆| 国产成人欧美| 中亚洲国语对白在线视频| 啦啦啦在线免费观看视频4| 久久精品aⅴ一区二区三区四区| 国产欧美日韩综合在线一区二区| netflix在线观看网站| 老汉色av国产亚洲站长工具| 午夜成年电影在线免费观看| 亚洲伊人色综图| 51午夜福利影视在线观看| 在线国产一区二区在线| 1024香蕉在线观看| 久久久精品国产亚洲av高清涩受| 在线观看免费视频日本深夜| 亚洲狠狠婷婷综合久久图片| 欧洲精品卡2卡3卡4卡5卡区| 亚洲伊人色综图| 久久天躁狠狠躁夜夜2o2o| 黑人操中国人逼视频| 1024香蕉在线观看| 日本撒尿小便嘘嘘汇集6| 免费日韩欧美在线观看| 丁香欧美五月| 在线观看午夜福利视频| 免费在线观看视频国产中文字幕亚洲| 久久香蕉国产精品| 国产aⅴ精品一区二区三区波| 国产一卡二卡三卡精品| 丰满人妻熟妇乱又伦精品不卡| 午夜福利一区二区在线看| 99久久综合精品五月天人人| 亚洲国产精品sss在线观看 | 久久中文看片网| 正在播放国产对白刺激| 亚洲第一av免费看| 国产精品成人在线| 久久久久精品国产欧美久久久| 女人被躁到高潮嗷嗷叫费观| 免费看十八禁软件| 久久久久国产精品人妻aⅴ院| 天天添夜夜摸| 可以在线观看毛片的网站| 国产一卡二卡三卡精品| 午夜久久久在线观看| 精品人妻1区二区| 久久香蕉激情| 在线观看免费视频日本深夜| 18禁美女被吸乳视频| 69av精品久久久久久| 老司机午夜福利在线观看视频| 久久久久久人人人人人| 18美女黄网站色大片免费观看| 欧美日韩视频精品一区| 亚洲欧洲精品一区二区精品久久久| 日本黄色视频三级网站网址| 少妇裸体淫交视频免费看高清 | 国产成年人精品一区二区 | 女人精品久久久久毛片| 亚洲成人免费电影在线观看| 国产免费现黄频在线看| 高潮久久久久久久久久久不卡| 精品乱码久久久久久99久播| 婷婷丁香在线五月| 嫩草影视91久久| 高清毛片免费观看视频网站 | 欧美性长视频在线观看| 国产精品日韩av在线免费观看 | 国产黄色免费在线视频| www日本在线高清视频| 成在线人永久免费视频| 看片在线看免费视频| 国产精品免费一区二区三区在线| 亚洲精品国产色婷婷电影| 午夜福利欧美成人| 女警被强在线播放| 国产精品综合久久久久久久免费 | 999久久久国产精品视频| 亚洲精品国产一区二区精华液| 国产精品二区激情视频| 亚洲中文日韩欧美视频| 国产成人精品无人区| 无限看片的www在线观看| 一边摸一边抽搐一进一小说| 女人被躁到高潮嗷嗷叫费观| 狂野欧美激情性xxxx| 午夜福利一区二区在线看| 免费av中文字幕在线| 精品人妻1区二区| 丰满迷人的少妇在线观看| 久久精品人人爽人人爽视色| 757午夜福利合集在线观看| 99在线视频只有这里精品首页| 欧美日本亚洲视频在线播放| 免费av中文字幕在线| 免费在线观看完整版高清| 久久热在线av| 日本免费a在线| 在线观看66精品国产| 中文字幕人妻熟女乱码| 男女做爰动态图高潮gif福利片 | 婷婷丁香在线五月| 在线天堂中文资源库| 欧美精品亚洲一区二区| 日本vs欧美在线观看视频| 精品福利观看| 亚洲熟妇熟女久久| 曰老女人黄片| 久久这里只有精品19| 黄片小视频在线播放| 99国产精品一区二区三区| 精品熟女少妇八av免费久了| 国产精品久久久人人做人人爽| 国产熟女xx| 欧美日韩视频精品一区| 亚洲精品久久成人aⅴ小说| 精品国产国语对白av| 欧美乱妇无乱码| 国产精品香港三级国产av潘金莲| 热re99久久国产66热| 午夜免费激情av| 巨乳人妻的诱惑在线观看| 麻豆一二三区av精品| 1024视频免费在线观看| 日韩成人在线观看一区二区三区| 精品国产乱子伦一区二区三区| 一个人观看的视频www高清免费观看 | 亚洲精品久久午夜乱码| 国产成人精品久久二区二区91| 久久婷婷成人综合色麻豆| 精品久久久久久成人av| 日韩视频一区二区在线观看| 成人特级黄色片久久久久久久| 久久久国产一区二区| 天天添夜夜摸| 黄色 视频免费看|