The Causal Associations between Adipokines and Alzheimer’s Disease: A Two-Sample Mendelian Randomization Study

INTRODUCTION

Alzheimer’s disease (AD) is a typical neurodegenerative disease and the most common form of dementia, which accounts for nearly 80% of all kinds of dementia [1]. The earliest phase of AD starts with the accumulation of Aβ and the induction of the spread of tauopathy [2]. Several common mechanisms are responsible for AD, including imbalance of lipid metabolism, oxidative stress and inflammatory responses induced by protein accumulation [3–5]. Moreover, these pathways are also believed to be deeply involved in obesity, by participating in metabolic imbalance [6].

Obesity is a major health problem across the world. It is closely related to white matter intensity and disruption and progressive brain atrophy [7]. Some studies have demonstrated that anti-obesity therapy provides benefits for patients at risk of AD [8]. Therefore, it is important to demonstrate the potential link between obesity and AD. Studies have indicated that individuals with obesity are likely to present a chronic inflammatory status [9], with the activation of immune cells surrounding adipose tissues and the production of adipokines [10, 11]. Adipokines include a series of components mainly produced by adipose tissues. The dysregulation of adipokines has been consistently accompanied with the adverse outcomes in obesity, leading to a persistent inflammation [12]. Recent studies have indicated that the operation of adipokines could be affected during neurodegeneration and feedback to contribute to the neurodegenerative process. For instance, adiponectin interacts with blood-brain barrier [13], and regulate neurogenesis and synaptic function in hippocampus, cortex and hypothalamus [14], through its respective binding to adiponectin receptor 1 and 2 [14]. Resistin is another widely studied adipokine, associated with the integrity of brain structure. It is identified in the hypothalamus and cortex in mouse. The neuroprotective role of resistin was stated in AD mouse model by Liu et al. [15]. Therefore, it is believed that adipokines offer a window to explain the relationship between obesity and AD. However, current reports only provide observational findings concerning their associations. The causality between adipokines and AD have not been established.

Mendelian randomization (MR) approach provides a tool to estimate the causal effects of certain exposure on an outcome [16]. It uses genetic variants as instrumental variables (IVs) of exposure, and pools the effects of variants for effect estimation [16]. The approach diminishes the effects of potential confounders in observational level and overcomes the influences of reverse causation. We here used the two-sample MR analysis to determine the causal effects of adipokines on risk of AD.

MATERIALS AND METHODS

Study design

A two-sample MR analysis was done to test the causality between adipokines (adiponectin and resistin) and risk of AD, using summary data from GWAS consortia studies. Data sources were shown in Table 1. To detect the potential reverse causality, we performed an MR analysis to test the effects of AD on circulating adipokines. All participants have given informed consent in the respective center. An overview of the study assumptions was displayed in Fig. 1.

OPEN IN VIEWER

Fig. 1

The key assumptions of the MR study. First, the selected SNPs (IVs) should be significantly associated with the exposure (adipokine); second, the included SNPs for the exposure (adipokine) were not associated with any confounder; third, the selected SNPs should only affect the outcome (risk of AD) though the exposure (adipokine). AD, Alzheimer’s disease; IV, instrumental variable; MR, Mendelian randomization; SNP, single-nucleotide polymorphism.

OPEN IN VIEWER

Table 1

Characteristics of the data sources in the study

Variables	Data source	Authors	Sample size	Ancestry
adiponectin	ADIPOGen	Dastani et al. [17]	39,883	European
resistin	INTERVAL	Sun et al. [18]	3,301	European
AD	IGAP	Kunkle et al. [19]	63,926	European

AD, Alzheimer’s disease; IGAP, International Genomics of Alzheimer’s Project.

RESULTS

Causal effects of adiponectin on AD

The analysis results of the relationship between adiponectin and AD were displayed in Table 2. IVW mode indicated a detrimental effect of the decreased level of circulating adiponectin on risk of AD (OR: 0.850; 95% CI: 0.731-0.990, p = 0.037). However, such trend was not confirmed in the estimates in MR-Egger, weighted median, simple mode and weighted mode methods. The statistical power to detect the causal effects of adiponectin on AD was 100%.

Cochrane’s Q test and MR-PRESSO global test did not find obvious pleiotrophic effects (Table 3). The MR-Egger intercept analysis did not find significances (p = 0.982), indicating no evidence of potential pleiotrophy (Table 3).

OPEN IN VIEWER

Table 3

Results of the Cochrane’s Q test, MR-Egger intercept test and MRPRESSO Global test in the MR study

	Cochrane’s Q test		MR-Egger intercept test		MR-PRESSO Global test
	Q-value	P-Q test	intercept	P-intercept	P-Global
Adiponectin versus AD
	8.502	0.580	–0.0002	0.982	0.376
Resistin versus AD
	2.010	0.570	–0.008	0.761	0.606

AD, Alzheimer’s disease; MR, Mendelian randomization. *p < 0.05.

Scatter plots, funnels plots, forest plots and leave-one-out analysis of the MR analysis were shown in Fig. 2. rs17366568 was discovered to be a possible driving factor for the significant association of adiponectin with risk of AD. Therefore, we re-analyzed by excluding rs17366568 (Supplementary Table 3). It indicated no significant associations between adiponectin and AD in the IVW model.

OPEN IN VIEWER

Fig. 2

Scatter plot (A), funnel plot (B), forest plot (C), and leave-one-out sensitivity analysis (D) of the associations of adiponectin with risk of AD. AD, Alzheimer’s disease.

Causal effects of resistin on AD

The OR for the associations of resistin with AD was 0.936 (95% CI: 0.851-1.029, p = 0.171) (Table 2). Null associations were also found in the MR-Egger, weighted median, simple mode and weighted mode methods. Moreover, we did not find significant results in Cochrane’s Q test and MR-PRESSO global test (Table 3). And the MR-Egger intercept test did not reveal pleiotrophy. Power analysis using the mRnd yielded 100% power to detect the effects of resistin on AD.

There was no significant change in the estimates of risk of diseases by genetically determined resistin level in the leave-one-out analysis, demonstrating that no specific SNP would change the results. Scatter plots, forest plots, funnels plots and leave-one-out sensitivity analysis were shown in Fig. 3.

OPEN IN VIEWER

Fig. 3

Scatter plot (A), funnel plot (B), forest plot (C), and leave-one-out sensitivity analysis (D) of the associations of resistin with risk of AD. AD, Alzheimer’s disease.

DISCUSSION

This MR study reported causal associations between genetically determined adiponectin and AD. Besides, the genetic susceptibility of resistin was not responsible for the development of AD.

It is believed that there is a connection between being overweight and AD [22]. Unwanted weight loss usually occurs among individuals with AD before the presence of clinical symptoms, even with unchanged eating patterns [23]. These suggest the alteration of adipose tissue metabolism along neurodegeneration [24]. Studies have indicated that the metabolic status is regulated by endocrine, paracrine and autocrine “adipokines” [25]. The agents can interact with blood brain barrier and participate in CNS homeostasis through their control of energy metabolism and immune responses [13]. Therefore, adipokines seem to be an intermediate player between obesity and AD.

Adiponectin is secreted by adipose tissues and participates in multiple cellular functions, including insulin responses, antioxidation and suppression of inflammation [14]. Accumulating evidence has suggested the presence of hypoadiponectinemia in metabolic disorders. Further, plasma adiponectin concentration appears to be inversely associated with the severity of adiposity and type 2 diabetes mellitus [26, 27]. Moreover, low level of adiponectin is found in AD dementia [28], which represents metabolic dysfunction in brain [29]. The inverse association between adiponectin and AD accords with our findings using IVW model that a descending adiponectin was associated with a higher risk of AD. Our findings indicated a neuroprotective potential of adiponectin against Alzheimer’s pathology. Interestingly, our findings were inconsistent with a previous MR study [30] revealing no causal effects of adiponectin on AD, using the same GWAS datasets with our study. In that study, all 14 SNPs significantly associated with adiponectin were included. However, we excluded 3 SNPs rs1597466, rs2980879 and rs7964945 for their unsatisfactory performances. Therefore, a relatively strict analysis process ensured the reliability of results. This may help find the potential role of promising markers in disorders.

Some studies have referred to the effects of adiponectin against amyloid pathology. A study used the SH-SY5YswAPP cell line to mimic amyloid burden in neuronal cells. And it found that adiponectin could improve cellular survival [31]. Mouse with adiponectin deficiency showed impaired memory and learning abilities, as well as increased Aβ depositions in brain, along with the over-production of pro-inflammatory cytokines [32]. Therefore, we proposed that adiponectin functioned against AD through its suppression of amyloid pathology and related inflammatory responses.

However, some studies have provided varied findings concerning adiponectin. Increase [33], or no change [34, 35] of adiponectin level has been found among AD individuals. Some factors seem to be responsible for the inconsistence. It is believed that increased weight is associated with decreased level of adiponectin [36]. The weight loss commonly seen in AD patients may induce the upregulation of adiponectin. However, the findings from the Framing Heart Study indicated elevated serum adiponectin was still associated with increased risk of AD in women, when adjusting for age, weight change and BMI [37]. Most studies evaluating adiponectin in AD patients did not take treatment details into consideration. A previous study reported increased adiponectin following donepezil application [38]. These may also explain the null association between adiponectin and AD without rs17366568 taking into analysis, for the sake that the findings may be interfered by underlying agents. Further exploration of the effects of adiponectin in AD should be done in the future.

Resistin is another adipokine protein derived from adipose tissue. It is involved in the production of pro-inflammatory cytokines, such as interleukin-6 (IL-6) and tumor necrosis factor-α (TNF-α) [12, 39]. It plays an important role in metabolic, inflammatory, and autoimmune conditions [39, 40]. The effects of resistin on AD has been reported in previous studies. Marcinnò et al. [41] found lower resistin concentrations in AD patients. Mooldijk et al. [42] proved that resistin level was inversely associated with the risk of AD. Resistin could suppress the production of oxidative factors in Neuro2a cells transfected with the Swedish amyloid precursor protein (swAPP) mutant and Presenilin exon 9 deletion mutant [43], indicating an anti-amyloidosis effect. Therefore, it is believed that resistin should also protect against AD, like adiponectin. However, our findings did not reveal causal effects of resistin on AD. This echoes a previous study that higher resistin is more likely to be associated with dementia with a vascular cause, but not with AD dementia [44]. Further studies indicated that resistin is involved in the regulation of mitochondrial function but not in APP metabolism [43].

This study explored the bidirectional relationship between circulating adipokines and risk of AD. GWAS datasets were used, and we identified association between adiponectin and AD. Although MR method provides a reliable tool to estimate the causal link between an exposure and an outcome using genetic variants, some additional caveats and limitations should be listed. First, the varied data sources of SNP-exposure, SNP-disease should be non-negligible. As seen in Table 1, data used in the study were from population of a European ancestry. Our findings should be carefully when applied to other population. Second, the participants with AD may not be accordant in advancing stages and pathological indications. This might bias the MR estimates and cause the discrepancies between our findings and previous reports. Third, adiponectin and resistin perfumed their intrinsic abilities via the binding with their own receptors. Future efforts could be paid to explore the interactive mode between receptors and ligands. Fourth, not all factors in adipokine family were analyzed in the study, for the lack of adequate GWAS data sources. Further exploration for future datasets may help solve it.