* Correspondence:
Zhang Hui and Zhirong Yao, Institute of Dermatology, Shanghai Jiaotong
University School of Medicine, 1665 Kongjiang Road, Shanghai 200092,
China,
E-mail: c_zhanghui@sina.comandyaozhirong@xinhuamed.com.cn
Author Contributions: Zhenquan Xuan,
Hui Zhang and Zhirong Yao conceived
the project. Zhenquan Xuan and Jiawen Chen was responsible for the data
analysis. Zhenquan Xuan, Zhen Zhang, Yifeng Guo, Hui Zhang and Zhirong
Yao wrote and revised the manuscript. Hui Zhang and Zhirong Yao were
responsible for the study concept and design, study supervision and
funding. All authors reviewed and approved the manuscript.
To the Editor,
Atopic dermatitis (AD) and
allergic contact dermatitis (ACD)
are both common inflammatory skin diseases, having an immune
pathogenesis. There has been
substantial interest about their inter-relationships with respect to
altered susceptibility. Some studies have indicated that patients with
AD were more likely to suffer from ACD, especially caused by nickel,
preservatives and cosmetics,
probably due to the compromised skin
barrier abnormality of AD [1-3]. However, there is compelling
evidence in favor of reduced risk of ACD in patients with AD, especially
severe patients. A systematic review showed that the prevalence of ACD
was significantly lower among children with AD, compared with controls
[4]. Additionally, in patients with AD, severe ones were less
sensitive to dinitrochlorobenzene than mild or moderate ones [5, 6].
A register-based clinical study also found that patients with severe AD
have an overall lower prevalence of ACD when compared with controls,
whereas mild-to-moderate AD does not suppress ACD [7]. Since severe
cases may have a higher genetic load for the disease-causing mutations
and help reduce the interference of confounding factors [8].
Therefore, we presumed that there might be a negative causal effect from
AD to ACD at the genetic level.
Mendelian randomization (MR) is a genetic epidemiological tool which
uses single-nucleotide polymorphisms (SNPs) as instrumental variables
(IVs) to infer the potential causality of exposure and outcome. The MR
tool could conquer confounders and reverse causation bias, which is due
to the fact that an individual’s genotype is determined at conception
and cannot be altered. Considering this, we performed a two-sample MR
analysis for the association of AD with ACD using large-scale
genome-wide association study (GWAS) data sheets of Europeans. For AD,
the instrumental variables (IVs) were generated from the wide summary
statistics from a GWAS of 10788 cases and 30047 controls from 20 studies
of European ancestry, excluding the 23andMe study [9]. For ACD, we
used the summary statistics (finn-b-L12_ALLERGICCONTACT) from FinnGen
biobank analysis including 2,204 cases and 198,740 controls of European
ancestry. The participants of AD and ACD data sheets, to our knowledge,
did not overlap. The “TwoSampleMR”
package of the R software (version 4.1.2) was used to performed
two-sample MR analysis. SNPs associated (P < 5e-08)
with AD were selected, and the clumping process
(r 2 < 0.001, clumping distance =
10,000 kb) was carried out to eliminate the linkage disequilibrium
between the included SNPs, also the palindromic SNP rs10790275 with
intermediate allele frequencies was excluded. Eventually, 11 SNPs were
chosen as IVs (Table S1). Inverse variance weighting (IVW) was used as
the main analysis, and the MR Egger, weighted median, simple mode, and
weighted mode as supplementary methods. And the effect of each SNP locus
on ACD was shown in Figure 1A and 1B. Analysis based on IVW suggested a
significant casual effect of AD on ACD (OR 95% confidence interval
[CI] = 0.73 [0.63-0.89]; P = 3.02e-05) (Figure 1A; Table
S2). Moreover, weighted median analysis also showed a remarkable casual
effect of AD on ACD (OR 95% CI =
0.75 [0.64-0.89]; P = 8.76e-04) (Figure 1A, Table S2), while
the results of analysis based on simple mode (OR 95% CI = 0.75
[0.57-1.00]; P = 0.08) (Figure 1A, Table S2) and weighted
mode (OR 95% CI = 0.77 [0.62-0.97]; P = 0.05) (Figure 1A,
Table S2) were slightly significant.
Leave-one-out analysis further
revealed that there was no fundamental impact on ACD (all lines were on
the left side of 0), regardless of which SNP was removed, suggesting
that the MR result was robust (Figure 2). MR Egger regression was used
to detect the horizontal pleiotropy of our analysis, and no significant
pleiotropy was determined (P = 0.53) (Table S3), indicating the
basic assumption of MR analysis was satisfied. The P -values for
heterogeneity tests using the
MR-Egger and IVW methods were 0.15 and 0.12, respectively, which
suggests that there is no heterogeneity (Table S3).
To our knowledge, this study is the first to analyze the causal
relationship between AD and ACD based on two-sample MR. Surprisingly,
the present results found that AD had negative causal relationships with
overall ACD with respect to genetic effects. The altered risk of ACD in
AD patients is a controversial topic in the field as mentioned above
[1-7], and indeed, some showed the decreased risk of ACD in patients
with AD, in especial, severe cases [4-7]. However, these
observational studies could not draw a causal link between the two
conditions, therefore, we utilized MR to detect it. The present study
extends prior research by showing a causal relationship between AD and
ACD, although the pathophysiological mechanisms underlying this
relationship are not entirely clear. One possible explanation for this
relationship is, under genetic control, the relative weak Th1-mediated
cellular immunity in AD inhibit the onset of ACD [10]. The data used
for this study were limited to participants of European ancestry,
therefore, a comprehensive and clear database is required for further
research. Furthermore, it is of importance to explore the causal effect
between AD and diverse
allergen-specific-ACD subtypes in the
future. Finally, the exposure factors in this study were the effects of
mutated genes, which may differ from the actual exposure in terms of
time and dose.