Multiple Sclerosis Journal - October 2017 - 1465

J Devorak, LE Mokry

Figure 1. Mendelian randomization estimate of the association of adiponectin level with risk of MS. Estimates obtained
using a fixed-effects model.

gene ADIPOQ) and five from the IMSGC/WTCCC2
GWAS (rs2980879, near TRIB1; rs601339, near
GPR109A; rs7133378, within DNAH10; rs7955516,
near PDE3A; and rs3001032, near LYPLAL1) (Table
1). Therefore, 10 of the 12 ADIPOGen SNPs were
selected for this MR study. None of the 10 adiponectin increasing alleles were significantly associated with MS risk, accounting for multiple testing
(all ps > 0.05/10 = 0.005, Table 1 and Figure 1).

SNP validation
LD. None of the 10 adiponectin-associated SNPs
were found to be in LD (all pairwise r2 < 0.05) in the
UK10K European samples.25
Pleiotropy. MR-Egger regression analyses suggested that pleiotropy did not greatly influence the
results of the MR analyses (Egger intercept, p = 0.21;
95% confidence interval (CI) = −0.015-0.058). Additionally, a literature review failed to unearth pleiotropic mechanisms for any of the investigated SNPs,
with the exception of rs12922394. This SNP is
located within an intron of the CDH13 gene, which
encodes T-cadherin, a protein known to bind both
high molecular weight (HMW) adiponectin and lowdensity lipoprotein (LDL). It is thought that T-cadherin might function as a receptor for both these
ligands.33 Numerous epidemiological studies have
demonstrated associations between elevated serum
LDL and MS disease progression, as well as adverse
clinical and magnetic resonance imaging (MRI) outcomes.34 Based on these findings, the possibility that
CDH13 functions independently of adiponectin to
produce MS phenotypes could not be eliminated;
therefore, sensitivity analyses were undertaken to
exclude rs12922394 from MR analyses.
journals.sagepub.com/home/msj

Population stratification. A one-way ANOVA
revealed that serum-log-transformed adiponectin concentrations did not differ across the European subpopulations interrogated in ADIPOGen (F(5,
17) = 1.27, p = 0.32).
MR estimates
Employing a fixed-effects model including all 10
adiponectin-altering alleles revealed that a one-unit
increase in natural-log-transformed adiponectin,
which corresponds to a two-standard deviation
change on the absolute scale, was not associated with
a clear effect on the odds of MS (odds ratio
(OR) = 0.93; 95% CI = 0.66-1.33; p = 0.61) (Figure
1). The I2 estimate of heterogeneity was 0%, suggesting no heterogeneity of effect. Sensitivity analyses
excluding rs12922394 (CDH13) for possible pleiotropic effects did not influence these results
(OR = 0.93; 95% CI = 0.64-1.37; p = 0.72). Analysis
of the ADIPOQ variant rs6810075 alone revealed
that a one-unit increase in natural-log-transformed
adiponectin did not alter the odds of MS (OR = 0.60;
95% CI = 0.34-1.07; p = 0.08). Analysis of the pooled
non-proxy SNPs, rs1108842, rs12922394, rs1597466,
and rs2925979, revealed no evidence of an association with MS risk (OR = 1.06; 95% CI = 0.60, 1.88).
Discussion
In this MR study investigating the role of adiponectin
level upon MS risk, we have demonstrated that a large
(two-SD), lifelong genetic increase in adiponectin
level was not associated with a clinically relevant
change in the odds of MS. This finding does not support a substantial role for adiponectin in the causal
pathway of MS; however, given the wide CI, a small
1465
https://journals.sagepub.com/home/msj
Table of Contents for the Digital Edition of Multiple Sclerosis Journal - October 2017
