Evidence of a causal relationship between high serum adiponectin levels and increased cardiovascular mortality rate in patients with type 2 diabetes

Despite insulin-sensitizing, anti-inflammatory and endothelial protective effects [1‐4], adiponectin is an independent positive predictor of cardiovascular (CV) mortality in the general population [5] as well as several clinical sets [6‐15] including type 2 diabetes (T2D) [16]. Because of such paradox, it is logical to suspect that high adiponectin rather than being a true risk factor is simply a marker of other metabolic disturbances, which, in fact, are able to causally shape the risk of CV mortality.

Genetic variants, which are robustly associated with a given modifiable variable, are strong and easy-to-use tools for assessing if causality underlines the association between that variable, used as an exposure, and outcomes of interest [17‐19]. In our case, if genetic variants, firmly associated with circulating adiponectin levels, prove to be also associated with CV death, a strong case is made in favor of a causal role of adiponectin on CV mortality.

We and others have recently reported that single nucleotide polymorphism (SNP) rs822354, located in the locus of ADIPOQ, the gene which encodes for adiponectin [20], is associated with serum adiponectin at genome-wide level of statistical significance [21, 22].

Based on this background, we used rs822354 as an instrumental variable to investigate the causal nature of the counterintuitive independent association between adiponectin and CV mortality we have previously reported in patients with T2D [16].

Clinical features of patients from GHS are summarized in Table 1. During follow-up (5.4 ± 2.5 years), 58 CV deaths occurred, corresponding to an overall annual IR of 3.0 % (58 events/1922 py).

In these patients, the A allele of rs822354 was significantly associated with total adiponectin (β (SE) = 0.10 (0.042); p = 0.014). In addition, in a model comprising several variables relevant to the risk of death, including age, sex, smoking habits, BMI, HbA1c, total cholesterol, HDL-cholesterol, triglycerides, insulin therapy and hypertension, both the same allele at rs822354 (IRR = 1.94, 95 % CI 1.23–3.07; p = 0.005), as well as the genetic equivalent of total adiponectin change (IRR = 1.07, 95 % CI 1.02–1.12; p = 0.003), were significantly associated with CV mortality. Of note, the observed genetic effect was significantly greater than that exerted by the genetic equivalent change of serum adiponectin (i.e. change per one rs822354 allele) (p for IRR heterogeneity = 0.012) (Fig. 1a).

Similarly, rs822354 was associated with HMW adiponectin (β (SE) = 0.17 (0.06); p = 0.003), whose genetic equivalent change was, in turn, associated with CV mortality (adjusted IRR = 1.12, 95 % CI 1.04–1.21; p = 0.003). Also in this case, the association of CV mortality with either rs822354 or the genetic equivalent change of serum HMW adiponectin was significantly different (p for IRR heterogeneity = 0.021) (Fig. 1b).

Many cardiovascular risk factors recognize a genetic background [27‐29]. Beside unraveling new pathogenic pathways, such genetic markers may be used as a tool for addressing the intrinsic nature underlying the relationship between metabolic variables and a given outcome of interest, an approach known as Mendelian randomization [30, 31].

Our data clearly show that the A allele of SNP rs822354 (in the ADIPOQ locus, which encodes for adiponectin [20]), is associated not only with high serum adiponectin levels (both total and HMW isoform), but also and most importantly with high CV mortality rate.

This finding is perfectly in line with a previous study showing that the G allele of SNP rs4783244 (in CDH13, encoding for the HMW adiponectin receptor T-cadherin [32]) is associated with both high HMW adiponectin levels [33‐35] and increased mortality rate [35].

Although the exact mechanism through which SNP rs4783244 (or others in linkage disequilibrium, LD, with it) affects serum adiponectin is a matter of debate [33, 34], previous [35] and our present findings strongly suggest that the paradoxical association between increased mortality risk and high serum adiponectin, rather than recognizing the latter as only a marker of yet unknown biologically-unlinked abnormalities, is explained by a causal, deleterious effect of adiponectin action and/or metabolism [36, 37] on mortality rate.

Given the well-known beneficial effect of adiponectin on several metabolic and vascular phenotypes [4, 38‐40], the intrinsic biological nature underlying such deleterious effect(s) is difficult to understand or even only to speculate; otherwise, we would not be talking of “paradoxical effect/association”. Having acknowledge this, some deleterious actions of adiponectin on processes possibly involved in atherogenesis, including stimulatory effect on secretion of pro-atherogenic cytokines [41‐43] and angiogenesis [42, 44], have been recently reported in specific cell types. Whether such effects are observable also in human tissues involved in atherogenesis and whether they may play a role in increasing atherosclerotic processes is a possibility that deserves further, specifically designed studies, to be addressed. As a matter of fact, the available literature on the association with human traits does not unequivocally point to adiponectin as a metabolic and vascular protecting hormone. In fact, while low adiponectin has been associated with obesity, metabolic syndrome [39], some macro- and micro-vascular damages [38, 40] and T2D [39], so to make reasonable hypothesize that drugs activating adiponectin signaling pathways improve all the above mentioned metabolic and vascular disturbances [4], on the other hand, adiponectin levels are positively associated with subclinical atherosclerosis in patients with type 1 diabetes [45] and with all-cause mortality in patients with T2D [46, 47].

An additional, unexpected, finding of our study is that the association between SNP rs822354 and CV mortality is stronger than that expected if it would be entirely mediated by serum adiponectin changes. Such discrepancy might be due to the difference between a totally stable genetic effect, operating since birth on one hand and, conversely the role exerted by serum adiponectin, whose levels are quite variable [48, 49], especially in heavily treated patients as are those with T2D [50, 51], and whose effect has been evaluated only for a few years. As an alternative explanation, we cannot exclude that, beside the role on serum adiponectin, rs822354 exerts additional, distinct (i.e. pleiotropic) effects through other risk factors, which either we cannot account for or are still unknown. At this regard, it is of note that although neither rs822354, nor other SNPs in high LD with it, are associated with CV traits in humans [52, 53], it is located in the 3q27.3 region close to KNG1, a potential contributor in shaping CV risk [54]. In fact, KNG1 encodes for high molecular weight kininogen, the precursor of bradykinin and kallidin, belonging to the pathway activating coagulation factors XI and XII [55] and interrelated with the renin–angiotensin–aldosterone system [56].

We like to acknowledge some limitations of our study. Firstly, the sample we analyzed comprises only 356 individuals, so that the strength of the novel association we here report (i.e. between rs822354 and CV mortality) does not reach genome-wide level of statistical significance. In addition, we lack a second, independent sample on which to replicate our finding. In all, although we cannot exclude the possibility of a false positive result, our study plays the important role of hypothesis generating, thus representing a solid background on which further similar studies can be based on.

Conversely, strengths of our study are the deep clinical phenotyping and the genetic and environmental homogeneity of our sample recruited in a well-defined geographical region. In addition, also the determination of both total and HMW adiponectin, which allowed us to confirm our finding also with the biological active isoform of adiponectin [57], reinforces our study.

Finally, the observation that among community-dwelling Japanese individuals [35], a genetic marker of adiponectin levels is associated with all-cause mortality, suggests that our present finding is generalizable across different ethnicities and is not restricted to CV mortality in the specific subset of patients with T2D.

In conclusion, our present and previous [35] data suggest that adiponectin action/metabolism paradoxically causes increased CV mortality risk. Further studies are needed before this unexpected finding may be considered as established.