Association of TLR2 and TLR9 gene polymorphisms with atopic dermatitis: a systematic review and meta-analysis with trial sequential analysis

Abstract Atopic dermatitis (AD) is a chronic, inflammatory skin disease. The mechanism was complex. Genetic mutations of Toll-like receptor (TLR) may be associated with AD, yet still unclear. We aim to provide specific evidence of the association of TLR2, TLR9 gene polymorphisms with AD. Publications were selected according to the criteria. Newcastle-Ottawa Scale was applied to evaluate the quality. The value of ORs and 95%CIs were applied to measure the associations. According to the heterogeneity, the effects model of fixed or random was selected in data combination. For TLR2 gene rs5743708 polymorphism, under allele and recessive contrasts, the pooled data showed a significant correlation, which was A vs a, OR = 0.51 (95%CI: 0.30, 0.86); AA vs Aa + aa, OR = 0.54 (95%CI: 0.33, 0.88). For TLR2 gene rs4696480 polymorphism, under allele, homozygous, heterozygous, and dominant contrasts, the pooled data showed a significant correlation, which was A vs a, OR = 0.79 (95%CI: 0.64, 0.97), AA vs aa, OR = 0.65 (95%CI: 0.43, 0.97), Aa vs aa, OR = 0.68 (95%CI: 0.48, 0.97), AA + Aa vs aa, OR = 0.67 (95%CI: 0.49, 0.93). There are significant associations of TLR2 gene rs5743708, rs4696480 polymorphisms with atopic dermatitis, while no associations are found in TLR9 gene rs5743836, rs187084 polymorphisms.


Introduction
Atopic dermatitis (AD) is an inflammatory skin disease, with clinical manifestation of chronic or relapsing eczema-like lesions plus pruritus [1]. Prevalence of AD in children is about 20%, while 10% in adults [2,3]. In recent years, AD has become a global health issue, for the severe disease burden and increasing prevalence, especially in developing countries [4,5]. The mechanism of AD is complex. The defection of skin barrier function, the immune dysregulation, environment, as well as genetics are all thought to be vital factors in AD [4,6,7].
About 80% of AD patients have Staphylococcus aureus colonized [8]. In the recognition of microbial products, the immune system plays an important role. Down to the cellular level, various transmembrane proteins participate in immunity reaction, through inducing and activating immune cells [9,10]. Toll-like receptor (TLR) is one of those, which act as pattern recognition receptors [11]. The main function of TLRs is to mediate the innate immune response caused by tissue damage or infections [12,13].
Genetic mutations of TLR families are associated with various diseases, such as AD [14], asthma [15], and cancers [16]. In the past few years, increasing studies have paid attention to the gene polymorphisms of TLR2 and TLR9, including rs5743708, rs4696480, rs5743836 and rs187084, with the risk of atopic dermatitis. TLR2 and TLR9 are all belonged to TLRs family. TLR2 is an extracellular receptor which mainly recognizes the antigens from microbial, such as peptidoglycans, lipopeptides, while TLR9 is intracellular, which recognizes nucleic acids from microbial or the own.
However, in a single study, the sample size, the enrollment of the subject, and other factors are often limited. Besides, the statistical power may be insufficient. Under those situations, the results may be sometimes varied or even opposite among different studies. Hence, we conduct further investigation through meta-analysis, and we hope to provid more accurate results of the relationship between TLR2 rs5743708, rs4696480 polymorphisms, TLR9 rs5743836 and rs187084 gene polymorphisms and risk to atopic dermatitis.
The search strategies were in Supplementary material.

Selection of the publications
The selection of the publications was based on the criteria we wet. To be included, the publications should meet the conditions: (1) studies on TLR2 rs5743708, rs4696480 gene polymorphisms, TLR9 rs5743836 and rs187084 gene polymorphisms and atopic dermatitis (AD) which were already published; (2) case-control studies or cohort studies; (3) the diagnose of AD was under certain criteria; (4) healthy subjects or individuals without related diseases as control subjects; (5) genotype frequencies were shown in the text or could be calculated; (6) pvalues, ORs, 95%CIs were all shown or could be calculated from the provided data.
Publications with those conditions were excluded: (1) case reports, conference posters or abstracts, reviews of the literature, commentary or studies with animals; (2) Duplicate published or overlapping data; (3) the data could not be obtained or calculated from the information in the publications.
Two researchers (Boyang Zhou and Shuai Shang) completed the selection of these literatures separately and independently. Once disagreements were met, another researcher (Surong Liang) would check the whole process and make the final decision.

Data extraction
The following information was extracted for the included studies: (1) first author's name and the publication year, as Study ID; (2) country (or region) where the patients from; (3) the diagnostic criteria of atopic dermatitis; (4) ethnicity, gender and age of the patients (cases) and the controls; (5) numbers and genotype frequencies of cases and controls; (6) p-value of Hardy Weinberg Equilibrium (HWE).
Two researchers (Boyang Zhou and Surong Liang) finish the extraction of the information. The third researcher (Shuai Shang) would check the input result, correct errors and handle the inconsistency.

Methodological quality assessment
The assessment was through Newcastle-Ottawa Scale (NOS) [17], for case-control studies and cohort studies were included. The total score in NOS is 9, which consists scoring of selection (4), comparability (2) and exposure (3). For each study, score in total !6 can be considered as high-quality. Two researchers (Boyang Zhou and Shuai Shang) conducted the assessment. Once inconsistency was met, another researcher (Surong Liang) would re-assess the quality, and the final decision was made after the discussion.

Statistical analysis
Firstly, we used STATA 16 to calculate the Hardy-Weinberg equilibrium (HWE) for the control group [18].
Three exposure groups (AA, Aa, aa) could be generated in case groups and control groups, based on two alleles (A, a). A total of five genetic models were applied, including allele contrast (A vs a), homozygous contrast (AA vs aa), heterozygous contrast (Aa vs aa), dominant contrast (AA þ Aa vs aa), and recessive contrast (AA vs Aa þ aa). The associations between TLR gene polymorphisms and AD were exhibited and evaluated in the value of OR and 95%CI. For each contrast, p-value < .05 was significant in statistics.
Based on between-study heterogeneity, either fixed or random effects model could be selected in the combination of data. I 2 test were applied to evaluate the heterogeneity between the included studies. When I 2 >50% and p-value < .10, the between-study heterogeneity was significant in statistics, therefore the random effects model was selected. In contrast, the fixed effects model was used.
Subgroup analysis was introduced based on different characteristics, when heterogeneity between studies was significant. In order to test the influence of a single study to the overall effect of the comparison, analysis of sensitivity was performed through omitting each study seriatim in every comparison. For comparisons included five and more studies, publication bias should be assessed by funnel plot. The analysis and the construction of forest plots and funnel plots was all through RevMan 5.3 software [19].

Trial sequential analysis (TSA)
TSA was conducted through the application based on Java. The aim of TSA is to estimate if the case number is adequate to make a valid result in each comparison. In our study, risk of type I error (false positive) was set 5% and 30% of type II error (false negative) [20,21].

Search and selection of the publications
Initially, 277 publications from five databases were collected in total. Then, 69 duplicated ones were removed, and there were still 208 publications. After screening the tittle and abstract, 61 publications were first included, with 147 excluded in this procedure. Next, the whole content was checked, based on the criteria, we then excluded 53 publications. Finally, eight studies were included for meta-analysis [9,10,14,[22][23][24][25][26]. The flowchart of the search and selection procedure was in Figure 1.

Main information of the included studies
A total of eight studies were included in our metaanalysis [9,10,14,[22][23][24][25][26]. Among these studies, seven reported TLR2 gene rs5743708 polymorphism, with 636 AD patients included [9,10,14,22-25]; three reported TLR2 gene rs4696480 polymorphism, with 393 patients [9,10,14]; two reported TLR9 gene rs5743836 and rs187084 polymorphisms, with 407 and 408 patients [14,26]. The included studies were all case-control studies. In the study groups, the patients varied from 13 to 272, and the control subjects varied from 6 to 247. In two studies [22,23], only the frequencies of patients carrying recessive gene was provided, as a result, the amounts of alleles could not be obtained, also in the two studies, the HWE could not be test. All other studies provided precise number of gene types, which the amounts of alleles could be calculated. The HWE could also be test in those studies, and all the study groups complied with HWE (Table 1, sTable 1 in Supplementary material).
For TLR2 gene rs4696480 polymorphism, a total of three studies were included [9,10,14]. According to the between-study heterogeneity, fixed effect model was applied in each contrast. Under allele contras, homozygous contras, heterozygous contrast, and dominant contrast, the pooled data showed significant relationship between TLR2 gene rs4696480 polymorphism and AD, which was A vs a, OR For TLR9 gene rs5743836 polymorphism, two studies were included [14,26]. According to the between-study heterogeneity, fixed effect model was applied in each contrast. The pooled data did not show significant relationship between TLR9 gene rs5743836 polymorphisms and AD under each contrast (p > .05, Figure 4).
For TLR9 gene rs187084 polymorphism, two studies were included [14,26]. According to the between-study heterogeneity, fixed effect model was applied in each contrast. The pooled data did not show significant relationship between TLR9 gene rs187084 polymorphisms and AD under each contrast (p > .05, Figure 5).

Subgroup analysis
The countries and the age ranges were all varied in the studies, so we still performed subgroup analysis despite the between-study heterogeneity. For TLR2 gene rs5743708 polymorphism, under allele contrast and recessive contrast, the total effects were significant (p < .05), while in subgroup analysis by location, no significance was shown in Germany, Ukraine and Turkey groups, respectively (p > .05). Also, in age range subgroups, under allele contrast, no significance was shown in adults' group, and under recessive contrast, no significance was shown in children and adults' groups, respectively (p > .05). For TLR2 gene rs4696480 polymorphism, under allele contrast, homozygous contrast, heterozygous contrast, and dominant contrast, the total effects were significant (p < .05), while in subgroup analysis by location, no significance was shown in Germany and Turkey groups, respectively. Under allele contrast, homozygous contrast, and heterozygous contrast, no significance was shown in children and adults' groups, and under dominant contrast, no significance was shown in adults' group (p > .05). For TLR9 gene rs5743836 and rs187084 polymorphisms, the two included studies were both conducted in Germany and investigated in adults, so subgroup analysis was not need (Table 4).

Sensitivity analysis and evaluation of publication bias
For TLR2 gene rs5743708 polymorphism, under allele contrast (A vs a) and recessive contrast (AA vs Aa þ aa), with the study Salpietro 2011 ignored, statistical significance of the relationship disappeared. For TLR2 gene rs4696480 polymorphism, under allele contrast (A vs a), homozygous contrast (AA vs aa), and heterozygous contrast (Aa vs aa), when the study Oh 2009 was ignored, the statistical significance was showed. Also, under allele contrast (A vs a), homozygous contrast (AA vs aa), heterozygous contrast (Aa vs aa), and    For TLR2 gene rs5743708 polymorphism, there were five studies under allele contrast (A vs a). Asymmetry was shown in the funnel plots, indicating the existence of publication bias. Besides, seven studies were included under recessive contrast (AA vs Aa þ aa). The funnel plots were symmetric, indicating that the bias was low ( Figure 6).

Discussion
The prevalence and incidence of atopic dermatitis have increased in these years. Now in developed countries, about 20% of children and 10% of adults are affected [27,28]. Generally, AD is associated with other atopic diseases, such as d food allergy, asthma, conjunctivitis and so on, and the condition is called 'atopic march' [4]. Therefore, the pathophysiology is rather complex. Currently, the mutation of filaggrin (FLG) gene, which leads to the water loss on skin, is thought as one of the key mechanisms [29]. However, there are also other factors involved in the occurrence and development of AD, including genetic background, skin barrier dysfunction, innate and adaptive immune response [27].
The treatments of AD include skin care, topical anti-inflammatory drugs, oral corticosteroids, cyclosporin, antihistamines [30], however, the traditional treatment methods are relatively slow and the effect is varied. Patients with AD are more possible to have psychopsychological disorders and sleep disturbances [31,32], thus patients are in urgent need of more effective and rapid treatment. Recently, with more researches on etiology mechanism, new methods are developed. For example, monoclonal antibodies can target a certain cytokine or receptor directly, which can modulate the inflammatory or immune response [33]. Dupilumab is a fully human monoclonal antibody that blocks the IL-4 and IL-13 pathway, which can inhibit type 2 inflammatory response. Dupilumab was the first monoclonal antibody approved for the treatment of AD, while there are still other antibodies are under clinical investigation, such as Tralokinumab, Lebrikizumab, Tezepelumab, and Etokimab [3].
Toll-like receptor (TLR) is one of the pattern recognition receptors, which can mediate the innate immune response [13]. Many studies focused TLRs family gene polymorphisms with AD [9,14,23,24,26]. However, up to now, no targeted therapy of AD is based on the TLR. For the research and development of biologics, a key basis is the relationship between the potential target and the disease. In this study, we evaluated the association between TLR2 & TLR9 gene polymorphisms and atopic dermatitis, which may provide a basis for further investigation of related treatment methods.
The association between TLRs family gene polymorphisms and atopic dermatitis have been widely investigated, including TLR2, TLR4, TLR6 and TLR9. For TLR2, a dual role was found. In infection diseases, TLR2 has been confirmed to play a protective role by triggering inflammatory response. However, in this procedure, excessive inflammation response may also be caused by TLR2, which will lead to tissue damages [34]. TLR2 can stimulate keratinocytes, and then IL-1, IL-6, IL-8 and TNF-a can be generated [35]. TLR2 can be expressed on the surface of T cells. CD4þ and CD8þ T cells can be activated. As a result, TLR2 enhances the activation and function of T cells [36]. Those factors play key roles in the mechanism of AD. Besides the act in    the pathogenesis, studies also found that TLR2 gene polymorphisms are associated with the severity of AD. Patients with TLR-2 gene rs5743708 polymorphism allele variant showed a distinct phenotype of severe to moderate AD [14,22]. Common TLR2 gene polymorphisms include rs5743708 and rs4696480. For rs5743708 polymorphism, studies also found the association with cytomegalovirus infection [37], colorectal cancer [38], and HIV infection [39]. In our study, under allele contrast (A vs a) and recessive contrast (AA vs Aa þ aa), the pooled data both showed significant correlations between TLR2 gene rs5743708 polymorphism and AD (p < .05). Meanwhile, the results of TSA showed that adequate cases were included for rs5743708 polymorphism. However, in sensitivity analysis, with the study Salpietro 2011 ignored, the significance disappeared under both contrasts. Also, we conducted the subgroup analysis despite the low between-study heterogeneity. Through subgroup analysis, we found the significance was only in Italy, not in Germany, Ukraine and Turkey; in age range subgroups, the only significance was shown in children under allele contrast (A vs a). In conclusion, from the total effects of the comparisons, a significant relationship between rs5743708 polymorphism and atopic dermatitis can be found. However, from the sensitivity analysis and subgroup analysis, the results were not robust and perfect enough. Although adequate cases were included for rs5743708 polymorphism, the case number in each subgroup was insufficient. In the future, further studies in multiple countries and focused on different age groups are still in need. TLR2 rs4696480 polymorphism is located at chr4:153685974, and found associated with bacterial meningitis [40], psoriasis [41], and so on. In our meta-analysis, for rs4696480 polymorphism, under allele contrast (A vs a), homozygous contrast (AA vs aa), heterozygous contrast (Aa vs aa) and dominant contrast (AA þ Aa vs aa), the results were similar to that of rs574370. Therefore, the conclusion of significant relationship between rs4696480 polymorphism and AD could be draw. Different from the rs574370, the recessive contrast of rs4696480 polymorphism demonstrated no significance in the total value and in subgroup analysis. Form the result of sensitivity analysis, this outcome was robust, nevertheless, from the results of trial sequential analysis, only under recessive contrast, the Z-curve failed to cross required information size (RIS) line, indicated that the cases were inadequate. Meanwhile, considering the difference in regions and age ranges under other contrasts, further studies for rs4696480 polymorphism were also need.
TLR9 is also a pattern recognition receptor belongs to TLR family. Currently, more than ten TLRs have been identified in mammals, which participate in multiple cell-extrinsic and -intrinsic responds [42,43]. Common polymorphisms of TLR9 gene are rs5743836, rs187084, rs352140, and rs2066807 [44,45]. In AD patients, rs5743836 and rs187084 polymorphisms were investigated. In our study, for both polymorphisms, two studies were included, yet no significant relationship with AD was found. Both the studies were conducted in Germany and in adults, thus the subgroup analysis was unnecessary, and based on the results of sensitivity analysis, the results of the relationship were robust. However, we further conducted the trial sequential analysis. From the results, the majority of the Z-curves failed to cross RIS line, which indicated the inadequate cases. Meanwhile, only two studies were included for TLR9 gene polymorphisms, as a result, more studies concerning the relationship between TLR gene polymorphisms and AD should be performed.
We investigated the association of TLR2 and TLR9 gene polymorphisms with AD through metaanalysis with trial sequential analysis in this study. Previously, an analysis of the association of TLR2&4 gene polymorphisms with AD risk was published [46]. However, we studied further in the following aspect. First, we included rs5743708 and rs4696480 polymorphism of TLR2, which was more comprehensive. Next, we set the inclusion criteria that diagnose of AD was under certain criteria, such as UK criteria, Williams criteria, and Hanifin & Rajka criteria. Clinically, the manifestation of AD was complex, as a result, the diagnosis would be confused without the criteria. Moreover, we conducted the trial sequential analysis, aiming to test if the case numbers were sufficient for the polymorphisms.
There were some limitations that could not be ignored in our study. Firstly, the location of the included studies was limited in Europe, and the Caucasian was the main ethnicity, so more studies in other regions are need. Secondly, under allele contrast of TLR2 gene rs5743708 polymorphism, the funnel plots were asymmetry, indicating the existence of publication bias. Besides, the results of TSA showed that the included patients were inadequate in some polymorphisms. To make a more reliable conclusion, the sample size should be enlarged.

Conclusions
There are significant associations of TLR2 gene rs5743708, rs4696480 polymorphisms with atopic dermatitis, while no associations are found in TLR9 gene rs5743836, rs187084 polymorphisms. Further studies with larger samples, in multiple countries and focused on different age groups are need.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Data availability statement
The data that support the findings of this study are available from the corresponding author, upon reasonable request.