Assessment of the Causal Effect of IgG N-Glycosylation Level on Risk of Dementia: A 2-Sample Mendelian Randomization Study

Abstract

Background:

Previous prospective studies highlighted aberrant immunoglobulin G (IgG) N-glycosylation as a risk factor for dementia [such as Alzheimer’s disease (AD) and vascular dementia (VaD)]. It is unclear whether this association is causal or explained by confounding or reverse causation.

Objective:

The aim is to examine the association of genetically predicted IgG N-glycosylation with dementia using 2-sample Mendelian randomization (MR).

Methods:

Independent genetic variants for IgG N-glycosylation traits were selected as instrument variables from published genome-wide association studies (GWAS) among individuals of European ancestry. We extracted their corresponding summary statistics from large-scale clinically diagnosed AD GWAS dataset and FinnGen biobank VaD GWAS dataset. The inverse variance weighted (IVW) was performed to calculate the effect estimates. Meanwhile, multiple sensitivity analyses were used to assess horizontal pleiotropy and outliers.

Results:

There were no associations of genetically predicted IgG N-glycosylation traits with the risk of AD and VaD using the IVW method (all Bonferroni corrected p > 0.0013). These estimates of four additional sensitivity analyses methods were consistent with the IVW estimates in terms of direction and magnitude. Additionally, the MR-PRESSO global test and the intercept of MR-Egger regression indicated no evidence of horizontal pleiotropy. Meanwhile, the heterogeneity test showed no significant heterogeneity using the Cochran Q statistic. The leave-one-out sensitivity analyses also did not detect any significant change.

Conclusion:

Our MR study did not support evidence for the hypothesis that IgG N-glycosylation level may be causally associated with the risk of dementia.

Keywords

Alzheimer’s disease IgG N-glycosylation mendelian randomization vascular dementia

INTRODUCTION

Dementia represents a major and increasing significant public health challenges in the 21st century, with approximately 40–50 million people currently living with dementia worldwide [1]. Alzheimer’s disease (AD) and vascular dementia (VaD) are two of the most common types of dementia, accounting for ∼85% of all dementia cases [2, 3]. Although the pathophysiologic mechanism of dementia is still largely unknown, accumulating evidence has suggested that inflammatory responses may make a valuable contribution towards dementia [4 –6].

Glycosylation is a ubiquitous post-translational modification in more than half of all mammalian proteins [7 –9]. It can lead to severe or fatal diseases if genes mutate in modification of glycan antennae [10]. Glycosylation of immunoglobulin G (IgG) influences IgG effector function by modulating binding to Fc receptors and acts as a switch between pro- and anti-inflammatory [11 –17]. Hence, aberrant in IgG N-glycosylation may contribute to the promotion of inflammation [13 , 19]. Previous observational studies aiming to study the relationship between IgG glycosylation profile and dementia have mostly yielded inconsistent results. A lower abundance of complex galactosylation and sialylation played a role in the development of AD in European ancestry individuals [20]. Another study found that the fucose and bisecting-GlcNAc structures were significantly increased while disialylated N-glycans were decreased in AD patients’ serum when compared with the normal control group [21]. These conflicting findings may be due to possible confounding factors and reverse causality, which is methodological limitations for observational studies [22]. Furthermore, it remains unclear whether IgG N-glycome composition is casually associated with AD and VaD.

Mendelian randomization (MR), utilizing genetic variations as instrumental variables, is an epidemiological approach that enables stronger causal inference between an exposure and risk of diseases [22]. As genetic material is randomly allocated and fixed from parents to offspring at conception, MR approach is less prone to be influenced by potential confounders and reverse causation [23]. Hence, we perform a two-sample MR analysis to comprehensively elucidate the potential causal role of IgG N-glycosylation level on the risk of AD and VaD, using publicly available genome-wide association studies (GWASs) data.

2 METHODS

We performed two-sample MR analysis to estimate the causal effect of the IgG N-glycosylation on the risk of AD and VaD. MR relies on three key assumptions as shown in Fig. 1: I) the genetic variants are robustly associated with the exposure; II) the genetic variants are independent of potential confounders of the exposure-outcome association; and III) the genetic variants are only associated with the outcome through the exposure of interest (no horizontal pleiotropy) [22].

Fig. 1

MR Study design overview and key assumptions. Solid lines are theorized to exist; Dashed lines are theorized to be nonsignificant based on MR assumptions. IgG, immunoglobulin G; GP, glycan peak; MR, mendelian randomization; SNP, single nucleotide polymorphism.

All data analyses for MR were performed by “TwoSampleMR” package (Version 0.5.6) [24] and “MR-PRESSO” package (Version 1.0) in the R environment (R version 4.1.3, R Project for Statistical Computing). The “TwoSampleMR” package harmonizes exposure and outcome data sets including information on SNPs, alleles, effect sizes, standard errors, p values, and effect allele frequencies for selected instrumental variables. All results are presented as odds ratios (ORs) and corresponding 95% confidence intervals (CIs) of the outcomes with per genetically predicted increase in IgG N-glycosylation level. In order to account for multiple testing, Bonferroni correction was used to adjust the thresholds of significance level. The significance threshold was suggested for p < 0.0013 (twenty IgG N-glycosylation traits and two outcomes). Reporting guidelines follows the STROBE-MR statement [25]. The ethical approval included in this MR analyses can be found in the original articles.

Data sources and instruments

Outcome data sources

Summarized data on AD were obtained from publicly available GWAS meta-analysis of the Phase 1 (n_cases = 47,793; n_controls = 328,320) and Phase 2 (n_proxycases = 47,793; n_{proxycontrols} = 328,320) [26].

Genetic associations with diagnosed VaD were obtained from the GWAS of the FinnGen biobank with 1,118 cases and 251,154 controls of European ancestry (release 6, https://finngen.gitbook.io/documentation/data-download) [27].

Instrumental variable selection

We used genetic variants robustly associated with glycosylation of IgG identified in Meta-analysis of genome-wide association study (GWAS) conducted in 8,090 individuals from European population [28]. SNPs associated with IgG N-glycosylation traits that reached genome-wide significance (p < 5×10^–8) were selected as instrumental variables (IVs). We selected independent genetic variants using the cutoff of linkage disequilibrium (LD) value (threshold set at r² = 0.001) to ensure that the IVs for N-glycosylation traits were independent [24].

Statistics analysis

A multiplicative random effects inverse-variance weighted (IVW) model was used as principal statistical model (for analysis with ≥3 SNPs) [29, 30]. The causal estimates were obtained from meta-analysis of SNP-specific Wald ratio estimates (i.e., the beta coefficient for the effect of the SNP on the outcome divided by the beta coefficient for the effect of the SNP on the exposure) based on inverse variance weighting with multiplicative random effects that weights each ratio by SNP-outcome standard error [31].

Sensitivity analyses were also conducted including WM (weighted median), PWM (penalized weighted median), MR-Egger, and MR-PRESSO (Mendelian Randomization Pleiotropy RESidual Sum and Outlier). Combining data on multiple genetic variants into a single causal estimate, the weighted median could generate consistent estimates even when up to 50% of selected genetic variants are invalid instruments [32]. The MR-PRESSO method was performed to identify horizontal pleiotropic outliers in multi-instrument summary-level MR testing [33].

MR-Egger regression can detect some violations of the standard instrumental variable assumptions, and provide an effect estimate which is not subject to these violations [34]. Besides, the p-value of the MR-Egger intercept >0.05 indicates no horizontal pleiotropic effects. Cochran’s Q test were used to evaluate heterogeneous effects among instrumental SNPs in IVW method [32]. The Leave-one-out analysis sequentially removes a single variant from the analysis and evaluates whether a single genetic variant disproportionately affects the over MR estimate [35].

RESULTS

Supplementary Table 1 showed the characteristics and associations of these harmonized instruments with the corresponding N-glycosylation traits.

3.1 Causal effect of IgG N-glycosylation trait on AD

There was no association of genetically predicted IgG N-glycosylation trait with the risk of AD using the IVW method (all p > 0.0013) (Table 1). These estimates of four additional sensitivity analyses methods were consistent with the IVW estimates in terms of direction and magnitude including WM, PWM, MR-Egger, and MR-PRESSO (Supplementary Table 2). Supplementary Figure 1 showed the MR estimates about the causal effect of IgG N-glycosylation traits on the risk of AD in GWAS datasets based on the IVW method.

Table 1

Causal effect of IgG N-glycosylation on risk of Alzheimer’s disease via MR IVW analyses

Phenotypes	Method	IVs	OR (95% CI)	Beta (SE)	p	p for Q statistic
GP2	IVW	4	1.007 (0.992–1.023)	0.007 (0.008)	0.345	0.576
GP4	Wald ratio	1	0.990 (0.948–1.033)	–0.010 (0.022)	0.632	–
GP6	IVW	6	1.014 (0.996–1.032)	0.014 (0.009)	0.137	0.452
GP7	IVW	5	0.996 (0.982–1.010)	–0.004 (0.007)	0.581	0.452
GP8	IVW (fixed effects)	2	1.033 (1.006–1.061)	0.032 (0.014)	0.018	0.173
GP9	IVW	4	0.982 (0.956–1.009)	–0.018 (0.014)	0.196	0.039
GP10	IVW	8	0.992 (0.980–1.005)	–0.008 (0.006)	0.229	0.481
GP11	IVW (fixed effects)	2	1.026 (0.998–1.054)	0.026 (0.014)	0.066	0.280
GP12	IVW	3	1.009 (0.986–1.031)	0.009 (0.011)	0.449	0.300
GP13	IVW (fixed effects)	2	1.003 (0.984–1.022)	0.003 (0.010)	0.753	0.107
GP14	IVW	5	0.995 (0.980–1.009)	–0.005 (0.007)	0.471	0.479
GP15	IVW	7	0.992 (0.974–1.011)	–0.008 (0.009)	0.413	0.962
GP16	IVW	5	0.986 (0.970–1.002)	–0.014 (0.008)	0.089	0.266
GP17	Wald ratio	1	0.994 (0.961–1.027)	–0.006 (0.017)	0.706	–
GP18	IVW	5	0.997 (0.974–1.021)	–0.003 (0.012)	0.806	0.057
GP19	Wald ratio	1	1.049 (1.009–1.091)	0.048 (0.020)	0.016	–
GP20	IVW	1	1.008 (0.971–1.046)	0.008 (0.019)	0.680	–
GP22	IVW	3	1.009 (0.976–1.043)	0.009 (0.017)	0.592	0.142
GP23	Wald ratio	1	1.004 (0.982–1.028)	0.004 (0.012)	0.703	–
GP24	Wald ratio	1	1.016 (0.972–1.062)	0.016 (0.023)	0.480	–

Beta is the estimated effect size. IgG, immunoglobulin G; GP, glycan peak; CI, confidence intervals; IVs, instrumental variables; IVW, inverse-variance weighted; MR, mendelian randomization; OR, odds ratio; SE, standard error; SNP, single-nucleotide polymorphism.

There was no indication of horizontal pleiotropy using MR-Egger (all p > 0.0013). Besides, the MR-PRESSO Global test identified no horizontal pleiotropic outliers (all p > 0.0013) (Supplementary Table 2). Meanwhile, the heterogeneity test from IVW showed no significant heterogeneity using the Cochran Q statistic (all p > 0.0013) (Table 1). Meanwhile, the leave-one-out sensitivity analysis also did not detect any significant change (Supplementary Figure 2).

Causal effect of IgG N-glycosylation trait on VaD

The result of IVW analysis showed no evidence of causal relationship of IgG N-glycosylation trait with the risk of VaD (all p > 0.0013) (Table 2). A similar pattern of results was observed using the WM, PWM, MR-Egger, and MR-PRESSO (Supplementary Table 3). Supplementary ">Figure 3 presented the MR estimates about the causal effect of IgG N-glycosylation traits on the risk of VaD in GWAS datasets using the IVW method.

Table 2

Causal effect of IgG N-glycosylation on risk of vascular dementia via MR IVW analyses

Phenotypes	Method	IVs	OR (95% CI)	Beta (SE)	p	p for Q statistic
GP2	IVW	4	1.097 (0.699–1.721)	0.093 (0.230)	0.687	0.152
GP4	Wald ratio	1	0.937 (0.230–3.821)	–0.065 (0.717)	0.928	–
GP6	IVW	6	1.333 (0.943–1.885)	0.287 (0.177)	0.104	0.715
GP7	IVW	5	1.184 (0.797–1.759)	0.169 (0.202)	0.402	0.120
GP8	IVW (fixed effects)	2	0.771 (0.398–1.495)	–0.260 (0.338)	0.442	0.252
GP9	IVW	4	1.493 (1.017–2.193)	0.401 (0.196)	0.041	0.547
GP10	IVW	8	0.944 (0.634–1.404)	–0.058 (0.203)	0.775	0.075
GP11	IVW (fixed effects)	2	1.208 (0.644–2.266)	0.189 (0.321)	0.556	0.192
GP12	IVW	3	1.020 (0.550–1.888)	0.019 (0.314)	0.951	0.175
GP13	IVW (fixed effects)	2	1.340 (0.876–2.050)	0.293 (0.217)	0.177	0.403
GP14	IVW	5	1.125 (0.799–1.584)	0.118 (0.175)	0.500	0.446
GP15	IVW	7	0.857 (0.519–1.417)	–0.154 (0.256)	0.548	0.037
GP16	IVW	5	1.160 (0.827–1.628)	0.148 (0.173)	0.391	0.892
GP17	Wald ratio	1	1.253 (0.589–2.666)	0.225 (0.385)	0.559	–
GP18	IVW	5	0.661 (0.456–0.957)	–0.414 (0.189)	0.028	0.822
GP19	Wald ratio	1	0.698 (0.255–1.909)	–0.360 (0.514)	0.483	–
GP20	IVW	1	1.435 (0.616–3.343)	0.361 (0.432)	0.403	–
GP22	IVW	3	1.488 (0.771–2.873)	0.397 (0.336)	0.236	0.236
GP23	Wald ratio	1	0.672 (0.402–1.125)	–0.397 (0.263)	0.131	–
GP24	Wald ratio	1	0.828 (0.312–2.200)	–0.188 (0.498)	0.705	–

Beta is the estimated effect size. IgG, immunoglobulin G; GP, glycan peak; CI, confidence intervals; IVs, instrumental variables; IVW, inverse-variance weighted; MR, mendelian randomization; MR-PRESSO, Pleiotropy Residual Sum and Outlier; OR, odds ratio; SE, standard error; SNP, single-nucleotide polymorphism.

The intercept of MR-Egger regression indicated no evidence of horizontal pleiotropy (all p > 0.0013). Additionally, no significant potential outlier was identified by MR-PRESSO Global test (all p > 0.0013) (Supplementary Table 3). The heterogeneity test from IVW method showed Cochran’s Q statistics p > 0.0013 (Table 2). Moreover, the leave-one-out analysis illustrated that the overall effect was not driven by any individual SNP (Supplementary Figure 4).

DISCUSSION

Using two-sample MR with genetic instruments selected from summary-level data of IgG N-glycosylation GWAS, we found no evidence supporting causal relationship of IgG N-glycosylation trait with risk for AD and VaD in European population. Sensitivity analyses presented almost similar results based on different statistical models.

Previous observational studies concluded that aberrant IgG glycosylation was associated with risk of dementia. Gizaw et al found that the fucose and bisecting-GlcNAc structures were significantly increased while disialylated N-glycans were decreased in AD patients’ serum when compared with the normal control group [21]. Maguire et al found that a significant decrease in soluble sialyltransferase activity in serum was reported in a study comparing 12 AD patients with 12 age-matched controls [36]. Lundström et al reported an increased FA2 (GP4) and a reduced FA2G1 (GP8 or GP9), FA2G2 (GP14), FA2G2S1 (GP18) of IgG1 in 31 patients with AD compared to 23 age-matched controls in European ancestry [20]. However, our MR analysis supported no evidence that IgG N-glycosylation level had an effect on the risk of developing dementia using published GWAS on AD and VaD from European descent. Thus, it is possible that the original relationship resulted from confounding by one or more unmeasured/poorly measured confounders. Besides, previous observational studies might have overestimated dementia risk associated with aberrant IgG glycosylation and thus may have been subject to reverse causation. Therefore, confounders and reverse causation among these observational studies may partly explain the inconsistent with results of our MR study.

In our MR study, we used both proxy-cases and clinically diagnosed AD GWAS [37]. Given AD mostly occurs in old age, a GWAS including only of clinically diagnosed AD older participants may be more susceptible to bias due to selective survival (such as exposure and competing risk) violating the “exclusion-restriction” assumption [38]. In addition, the GWAS of IgG N-glycosylation consisted of participants aged 14–100 years [39]. Previous study suggested that IgG glycosylation appeared to correlate with both chronological and biological ages based on a community-based study [40]. By estimating SNP-IgG glycosylation associations among European participants with a wide age range, we might have underestimated the denominator effect of the ratio estimator [41]. Thus, when the effect of the instrument on exposure changes over time, the ratio estimator will represent a biased estimate of the lifetime effect of IgG N-glycosylation on AD and VaD.

This study has several notable strengths. The strength of this study was the design of 2-sample MR analysis which used genetic variants as instrumental variables to estimate the association between IgG N-glycosylation level and dementia based on data from large-scale GWAS databases. This design technique helped circumvent bias from reverse causation and residual confounding. In addition, a comprehensive series of sensitivity analyses was performed to assess whether the results were robust. Lastly, our results were unlikely to be affected by population stratification bias because we confined the studied population to individuals of European ancestry.

There are several limitations in this study. First, our data were restricted to participants having European ancestry, and therefore, the conclusions might not be extended to Asian and African populations. Second, this might have resulted in insufficient statistical power to detect associations with some of IgG N-glycosylation traits because the number of VaD cases was still relatively small compared with other outcomes used in MR studies. Third, because of the unavailability of individual data, we could not conduct stratified analyses or analyses adjusted for other covariates using the available summary statistics dataset.

In conclusion, our MR study did not find evidence for the hypothesis that IgG N-glycosylation might be causally associated with the risk of AD and VaD. Further MR studies for large-scale GWAS data with a sufficient number of IVs are necessary to validate the causality of IgG N-glycosylation in the risk of dementia.

Footnotes

ACKNOWLEDGMENTS

This work was supported by grants from the National Natural Science Foundation of China (81872682), Beijing Postdoctoral Research Foundation, the Application and Evaluation of Active Health Cloud Platform in China, National Key R&D Program of China (2018YFC2000704).

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

(2019) Global, regional, and national burden of Alzheimer’s disease and other dementias, 1990–2016: A systematic analysis for the Global Burden of Disease Study 2016. Lancet Neurol 18, 88–106.

Qiu

, Kivipelto

, von Strauss

(2009) Epidemiology of Alzheimer’s disease: Occurrence, determinants, and strategies toward intervention. Dialogues Clin Neurosci 11, 111–128.

Iadecola

(2013) The pathobiology of vascular dementia. Neuron 80, 844–866.

King

, O’Brien

, Donaghy

, Morris

, Barnett

, Olsen

, Martin-Ruiz

, Taylor

, Thomas

(2018) Peripheral inflammation in prodromal Alzheimer’s and Lewy body dementias. J Neurol Neurosurg Psychiatry 89, 339–345.

Koyama

, O’Brien

, Weuve

, Blacker

, Metti

, Yaffe

(2013) The role of peripheral inflammatory markers in dementia and Alzheimer’s disease: A meta-analysis. J Gerontol A Biol Sci Med Sci 68, 433–440.

Wood

(2018) Dementia: Peripheral inflammation could be a prodromal indicator of dementia. Nat Rev Neurol 14, 127.

Skropeta

(2009) The effect of individual N-glycans on enzyme activity. Bioorg Med Chem 17, 2645–2653.

Arnold

, Wormald

, Sim

, Rudd

, Dwek

(2007) The impact of glycosylation on the biological function and structure of human immunoglobulins. Annu Rev Immunol 25, 21–50.

Marek

, Vijay

, Marth

(1999) A recessive deletion in the GlcNAc-1-phosphotransferase gene results in peri-implantation embryonic lethality. Glycobiology 9, 1263–1271.

10.

Jaeken

, Matthijs

(2007) Congenital disorders of glycosylation: A rapidly expanding disease family. Annu Rev Genomics Hum Genet 8, 261–278.

11.

Ohtsubo

, Marth

(2006) Glycosylation in cellular mechanisms of health and disease. Cell 126, 855–867.

12.

Vidarsson

, Dekkers

, Rispens

(2014) IgG subclasses and allotypes: From structure to effector functions. Front Immunol 5, 520.

13.

Novokmet

, Lukić

, Vucković

, THurić

, Keser

, Rajsl

, Remondini

, Castellani

, Gasparović

, Gornik

, Lauc

(2014) Changes in IgG and total plasma proteinglycomes in acute systemic inflammation. Sci Rep 4, 4347.

14.

Kaneko

, Nimmerjahn

, Ravetch

(2006) Anti-inflammatory activity of immunoglobulin G resulting from Fc sialylation. Science 313, 670–673.

15.

Anthony

, Ravetch

(2010) A novel role for the IgG Fc glycan: The anti-inflammatory activity of sialylated IgG Fcs. J Clin Immunol 30 Suppl 1, S9–14.

16.

Nimmerjahn

, Ravetch

(2008) Fcgamma receptors as regulators of immune responses. Nat Rev Immunol 8, 34–47.

17.

Ferrara

, Stuart

, Sondermann

, Brünker

, Umaña

(2006) The carbohydrate at FcgammaRIIIa Asn-162. An element required for high affinity binding to non-fucosylated IgG glycoforms. J Biol Chem 281, 5032–5036.

18.

de Jong

, Selman

, Adegnika

, Amoah

, van Riet

, Kruize

, Raynes

, Rodriguez

, Boakye

, von Mutius

, Knulst

, Genuneit

, Cooper

, Hokke

, Wuhrer

, Yazdanbakhsh

(2016) IgG1 Fc N-glycan galactosylation as a biomarker for immune activation. Sci Rep 6, 28207.

19.

Plomp

, Ruhaak

, Uh

, Reiding

, Selman

, Houwing-Duistermaat

, Slagboom

, Beekman

, Wuhrer

(2017) Subclass-specific IgG glycosylation is associated with markers of inflammation and metabolic health. Sci Rep 7, 12325.

20.

Lundström

, Yang

, Lyutvinskiy

, Rutishauser

, Herukka

, Soininen

, Zubarev

(2014) Blood plasma IgG Fc glycans are significantly altered in Alzheimer’s disease and progressive mild cognitive impairment. J Alzheimers Dis 38, 567–579.

21.

Gizaw

, Ohashi

, Tanaka

, Hinou

, Nishimura

(2016) Glycoblotting method allows for rapid and efficient glycome profiling of human Alzheimer’s disease brain, serum and cerebrospinal fluid towards potential biomarker discovery. Biochim Biophys Acta 1860, 1716–1727.

22.

Lawlor

, Harbord

, Sterne

, Timpson

, Davey Smith

(2008) Mendelian randomization: Using genes as instruments for making causal inferences in epidemiology. Stat Med 27, 1133–1163.

23.

Davies

, Howe

, Brumpton

, Havdahl

, Evans

, Davey Smith

(2019) Within family Mendelian randomization studies. Hum Mol Genet 28, R170–r179.

24.

Hemani

, Zheng

, Elsworth

, Wade

, Haberland

, Baird

, Laurin

, Burgess

, Bowden

, Langdon

, Tan

, Yarmolinsky

, Shihab

, Timpson

, Evans

, Relton

, Martin

, Davey Smith

, Gaunt

, Haycock

(2018) The MR-Base platform supports systematic causal inference across the human phenome. Elife 7, e34408.

25.

Skrivankova

, Richmond

, Woolf

BAR

, Yarmolinsky

, Davies

, Swanson

, VanderWeele

, Higgins

JPT

, Timpson

, Dimou

, Langenberg

, Golub

, Loder

, Gallo

, Tybjaerg-Hansen

, Davey Smith

, Egger

, Richards

(2021) Strengthening the reporting of observational studies in epidemiology using mendelian randomization: The STROBE-MR statement. JAMA 326, 1614–1621.

26.

Jansen

, Savage

, Watanabe

, Bryois

, Williams

, Steinberg

, Sealock

, Karlsson

, Hägg

, Athanasiu

, Voyle

, Proitsi

, Witoelar

, Stringer

, Aarsland

, Almdahl

, Andersen

, Bergh

, Bettella

, Bjornsson

, Brækhus

, Bråthen

, de Leeuw

, Desikan

, Djurovic

, Dumitrescu

, Fladby

, Hohman

, Jonsson

, Kiddle

, Rongve

, Saltvedt

, Sando

, Selbæk

, Shoai

, Skene

, Snaedal

, Stordal

, Ulstein

, Wang

, White

, Hardy

, Hjerling-Leffler

, Sullivan

, van der Flier

, Dobson

, Davis

, Stefansson

, Pedersen

, Ripke

, Andreassen

, Posthuma

(2019) Genome-wide meta-analysisidentifies new loci and functional pathways influencing Alzheimer’sdisease risk. Nat Genet 51, 404–413.

27.

Lyon

, Andrews

, Elsworth

, Gaunt

, Hemani

, Marcora

(2021) The variant call format provides efficient and robust storage of GWAS summary statistics. Genome Biol 22, 32.

28.

Klarić

, Tsepilov

, Stanton

, Mangino

, Sikka

, Esko

, Pakhomov

, Salo

, Deelen

, McGurnaghan

, Keser

, Vucković

, Ugrina

, Kristić

, Gudelj

, Stambuk

, Plomp

, Pucić-Baković

, Pavić

, Vilaj

, Trbojević-Akmaci

, Drake

, DobrinićP , Mlinarec

, Jelusić

, Richmond

, Timofeeva

, Grishchenko

, Dmitrieva

, Bermingham

, Sharapov

, Farrington

, Theodoratou

, Uh

, Beekman

, Slagboom

, Louis

, Georges

, Wuhrer

, Colhoun

, Dunlop

, Perola

, Fischer

, Polasek

, Campbell

, Rudan

, Wilson

, Zoldos

, Vitart

, Spector

, Aulchenko

, Lauc

, Hayward

(2020) Glycosylation ofimmunoglobulin G is regulated by a large network of genespleiotropic with inflammatory diseases. Sci Adv 6, eaax0301.

29.

Burgess

, Butterworth

, Thompson

(2013) Mendelian randomization analysis with multiple genetic variants using summarized data. Genet Epidemiol 37, 658–665.

30.

Burgess

, Foley

, Zuber

(2018) Inferring causal relationships between risk factors and outcomes from genome-wide association study data. Annu Rev Genomics Hum Genet 19, 303–327.

31.

Burgess

, Small

, Thompson

(2017) A review of instrumental variable estimators for Mendelian randomization. Stat Methods Med Res 26, 2333–2355.

32.

Bowden

, Davey Smith

, Haycock

, Burgess

(2016) Consistent estimation in Mendelian randomization with some invalid instruments using a weighted median estimator. Genet Epidemiol 40, 304–314.

33.

Verbanck

, Chen

, Neale

, Do

(2018) Detection of widespread horizontal pleiotropy in causal relationships inferred from Mendelian randomization between complex traits and diseases. Nat Genet 50, 693–698.

34.

Bowden

, Davey Smith

, Burgess

(2015) Mendelian randomization with invalid instruments: Effect estimation and bias detection through Egger regression. Int J Epidemiol 44, 512–525.

35.

Burgess

, Bowden

, Fall

, Ingelsson

, Thompson

(2017) Sensitivity analyses for robust causal inference from Mendelian randomization analyses with multiple genetic variants. Epidemiology 28, 30–42.

36.

Maguire

, Gillian

, O’Mahony

, Coughlan

, Dennihan

, Breen

(1994) A decrease in serum sialyltransferase levels in Alzheimer’s disease. Neurobiol Aging 15, 99–102.

37.

Jansen

, Savage

, Watanabe

, Bryois

, Williams

, Steinberg

, Sealock

, Karlsson

, Hägg

, Athanasiu

, Voyle

, Proitsi

, Witoelar

, Stringer

, Aarsland

, Almdahl

, Andersen

, Bergh

, Bettella

, Bjornsson

, Brækhus

, Bråthen

, de Leeuw

, Desikan

, Djurovic

, Dumitrescu

, Fladby

, Hohman

, Jonsson

, Kiddle

, Rongve

, Saltvedt

, Sando

, Selbæk

, Shoai

, Skene

, Snaedal

, Stordal

, Ulstein

, Wang

, White

, Hardy

, Hjerling-Leffler

, Sullivan

, van der Flier

, Dobson

, Davis

, Stefansson

, Pedersen

, Ripke

, Andreassen

, Posthuma

(2020) Author Correction: Genome-widemeta-analysis identifies new loci and functional pathwaysinfluencing Alzheimer’s disease risk. Nat Genet 52, 354.

38.

Schooling

, Lopez

, Yang

, Zhao

, Au Yeung

, Huang

(2020) Use of multivariable Mendelian randomization to address biases due to competing risk before recruitment. Front Genet 11, 610852.

39.

Lauc

, Huffman

, Pucić

, Zgaga

, Adamczyk

, Muzinić

, Novokmet

, Polasek

, Gornik

, Kristić

, Keser

, Vitart

, Scheijen

, Uh

, Molokhia

, Patrick

, McKeigue

, Kolcić

, Lukić

, Swann

, vanLeeuwen

, Ruhaak

, Houwing-Duistermaat

, Slagboom

, Beekman

, de Craen

, Deelder

, Zeng

, Wang

, Hastie

, Gyllensten

, Wilson

, Wuhrer

, Wright

, Rudd

, Hayward

, Aulchenko

, Campbell

, Rudan

(2013) Loci associated with N-glycosylation ofhuman immunoglobulin G show pleiotropy with autoimmune diseases andhaematological cancers. PLoS Genet 9, e1003225.

40.

, Wang

, Kristic

, Dong

, Chu

, Ge

, Wang

, Fang

, Gao

, Liu

, Zhao

, Peng

, Pucic Bakovic

, Wu

, Song

, Rudan

, Campbell

, Lauc

, Wang

(2016) Profiling IgG N-glycans as potential biomarker of chronological and biological ages: A community-based study in a Han Chinese population. Medicine (Baltimore) 95, e4112.

41.

Labrecque

, Swanson

(2019) Interpretation and potential biases of mendelian randomization estimates with time-varying exposures. Am J Epidemiol 188, 231–238.