Inverse associations between serum levels of secreted frizzled-related protein-5 (SFRP5) and multiple cardiometabolic risk factors: KORA F4 study

The study is based on data from the Cooperative Health Research in the Region of Augsburg (KORA) F4 study (2006–2008), a follow-up examination of the population-based KORA S4 study (1999–2001) conducted in Augsburg (Germany) and two surrounding counties. The design of the KORA studies has been described before [16, 17]. Briefly, study participants were invited from the city of Augsburg and two adjacent counties based on the survey sampling method of the former World Health Organisation Monitoring of Trends and Determinants of Cardiovascular Disease (WHO MONICA) project [17]. Within each selected community in the study area, age and sex-stratified samples were drawn with four of these strata including men and women aged 55–74 years. This cross-sectional analysis uses data from all study participants between 62 and 81 years of age in the KORA F4 study (n = 1161). Additional file 1: Figure S1 provides an overview of the study design, and Additional file 1: Table S1 compares participants and non-participants in KORA F4. In our analysis of the KORA F4 study sample, we excluded persons with unclear glucose tolerance status due to missing values for fasting and/or 2-h glucose (n = 33), type 1 diabetes (n = 2), drug-induced diabetes (n = 1) or other missing covariables (n = 29), resulting in a sample size of 1096 individuals.

Height, body weight, waist circumference, systolic and diastolic blood pressure were measured as described [16, 17]. Body mass index (BMI) was calculated as body weight (kg)/[height (m)]². Actual hypertension was defined as blood pressure of 140/90 mmHg or higher, or use of antihypertensive medication given that the study participants were aware of being hypertensive.

Known type 2 diabetes was defined as self-report of a previous diagnosis of type 2 diabetes that could be validated by the responsible physician, or as current use of glucose-lowering agents. All study participants without known type 2 diabetes were assigned to a standard 75-g oral glucose tolerance test (OGTT). Glucose tolerance categories were defined using fasting and 2-h glucose concentrations according to criteria of the American Diabetes Association [18]. Prediabetes was defined as impaired fasting glucose (IFG), impaired glucose tolerance (IGT) or combined IFG/IGT.

Glucose and insulin levels and HbA1c were measured as previously described [19]. Insulin sensitivity was assessed by calculation of the homeostasis model assessment of insulin resistance (HOMA-IR) based on fasting glucose and insulin levels and the whole-body insulin sensitivity index (ISI [composite]) based on both fasting and 2-h levels of glucose and insulin [19‐21]. Fasting beta-cell function was estimated using HOMA-β [20].

Serum lipid levels were measured as described [19]. Kidney function was assessed from the estimated glomerular filtration rate (eGFR) using the chronic kidney disease epidemiology (CKD-EPI) creatinine equation [22, 23].

Trained medical interviewers collected information on medical history, physical activity, smoking, alcohol consumption and use of medication [16]. Study participants were classified as physically active if they reported >1 h of physical activity per week during leisure time in either summer or winter. Smoking status was defined as never, former or current smoking. Alcohol consumption was classified as none, moderate or high for an alcohol intake of 0, 0.1–39.9 or ≥40 g/day for men and 0, 0.1–19.9 or ≥20 g/day for women, respectively.

Plasma levels of high-sensitivity C-reactive protein (hsCRP) and serum levels of interleukin (IL)-6, IL-1 receptor antagonist (IL-1RA) and total adiponectin were measured as described [24].

SFRP5 levels were measured in serum samples between 08/2016 and 09/2016 that had been continuously stored at −80 °C between blood sampling and analysis. Measurements were carried out using the Enzyme-linked Immunosorbent Assay Kit for Secreted Frizzled Related protein 5 (SFRP5) from Cloud-Clone (Houston, TX; previously USCN, Wuhan, China). The limit of detection (LOD) was 1.56 ng/ml. All sera yielded SFRP5 levels above the LOD. Coefficients of variation (CV) were assessed using three control sera measured in duplicates on 16 plates. Mean intra- and inter-assay CV were 6.4 and 18.6%, respectively.

Cross-sectional associations between log₂(SFRP5) and cardiometabolic risk factors were estimated by multivariable linear regression analysis using the following predefined set of potential confounders from previous KORA analyses [19, 25]:

Age, alcohol consumption, BMI, HDL cholesterol, LDL cholesterol, triglycerides and eGFR entered the models as continuous variables. In a sensitivity analysis, we replaced BMI with waist circumference in models 3 and 4.

Associations of log₂ (SFRP5) with prediabetes/diabetes or separate glucose tolerance status groups were assessed by logistic and multinomial logistic regression analysis, respectively, with individuals with normal glucose tolerance as reference group adjusting for the aforementioned covariables.

Associations between HbA1c, time since diabetes diagnosis and glucose-lowering medication and log₂ (SFRP5) were also estimated using linear regression analyses.

We checked the degree of multicollinearity in all models by calculating the variance inflation factors (VIF). Multicollinearity is considered high for VIF > 10, but VIF were <1.7 in all analyses.

All statistical analyses were performed with R version 3.3.3 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria) and Python version 3.6.1 (Python Software Foundation, https://​www.​python.​org/​). A P value <0.05 was considered to indicate statistical significance.

This is the first study that investigated the association between SFRP5, prediabetes and type 2 diabetes in a large population-based study and controlled for a range of potentially confounding factors. We extend the current literature by using both fasting and dynamic measures of insulin sensitivity. Our data shed light on the currently controversial data on SFRP5 and type 2 diabetes by demonstrating that higher serum SFRP5 was significantly linked to lower odds of prediabetes/type 2 diabetes (as combined group of individuals with impaired glucose regulation) even after adjustment for age, sex, BMI, lifestyle factors and risk factors for cardiovascular diseases.

Our study is in line with previous reports suggesting a protective role of Sfrp5 in the development of type 2 diabetes in mice [9] and with human studies demonstrating an inverse association between circulating SFRP5 and HOMA-IR, a fasting measure of insulin resistance [11‐13, 15]. Additionally, two studies also reported lower Sfrp5 levels in patients with type 2 diabetes compared to individuals with normal glucose tolerance [11, 12]. However, it is in contrast to two human studies which found a positive [10] and no relationship [14] to HOMA-IR or higher Sfrp5 levels in patients with type 2 diabetes compared to prediabetic individuals, although no difference to the group with normal glucose tolerance was observed [14]. Different laboratory methods cannot explain the observed differences because this study and the aforementioned studies [10‐15] used the same method to measure circulating SFRP5 (purchased from Cloud-Clone, previously distributed by USCN). However, the previous studies were based on small sample sizes, they were not population-based and differed in their adjustment for potential confounders. In particular the study with the opposing result of higher Sfrp5 levels in patients with type 2 diabetes was based on highly selected individuals with optimal glycaemic control (median HbA1c 5.8%) who cannot be considered representative for people with type 2 diabetes in the general population [14].

In contrast to the majority of epidemiological studies, the treatment of primary human adipocytes with recombinant SFRP5 reduced the phosphorylation levels of members of the insulin signaling pathway, which would argue for a detrimental role of SFRP5 for insulin sensitivity in these cells [26]. It may be possible that systemic and local effects of SFRP5 on insulin-sensitive cells are different, because the in vitro study concentrated on the impact of SFRP5 on adipocytes and did not consider the interaction with other cells from the adipose tissue as well as potential locally occurring binding partners.

The occurrence of cardiovascular diseases can be an important consequence of type 2 diabetes. In this study, we also investigated the association of serum SFRP5 with multiple cardiovascular risk factors in addition to glycaemia and prediabetes/type 2 diabetes. We found that higher serum SFRP5 was associated with lower systolic and diastolic blood pressure, higher HDL cholesterol and lower triglycerides. All associations were independent of age, sex and lifestyle factors. BMI attenuated all investigated relationships apart from those with systolic blood pressure and HDL cholesterol.

Previous findings were based on small sample sizes from hospital-based studies and included only people with Asian ethnicity with known differences in body fat distribution and adipokine levels compared to other ethnic groups. A recent hospital-based study found that circulating SFRP5 was associated with a lower odds of the metabolic syndrome irrespective of age, sex and risk factors for cardiometabolic diseases [15]. In this study population [15], circulating SFRP5 correlated inversely with systolic and diastolic blood pressure, triglycerides, free fatty acids and atherosclerotic index and positively with HDL cholesterol in a model adjusted for age and weight-to-hip ratio. In obese children with hypertension plasma SFRP5 was reduced, while lifestyle intervention did not only improve blood pressure and BMI but also increased plasma SFRP5 [27]. Another study found that serum SFRP5 was lower in patients with coronary artery disease compared to controls and correlated inversely with the severity of the disease [13]. In mice, Sfrp5 reduced myocardial infarction size and myocyte apoptosis in the heart after ischaemia/reperfusion injury. This might be partially mediated by inhibition of pro-inflammatory effects of the Wnt5a/JNK signaling in macrophages [8] which also plays an important role in adipose tissue inflammation [28] and endothelial dysfunction [29] and thus underlines the protective role of Sfrp5 on cardiovascular diseases. Collectively, these data demonstrate robust inverse associations between circulating SFRP5 and a range of cardiovascular risk factors in human cohorts using different study designs and populations, which are corroborated by data from animal models. However, data on cardiovascular outcomes are still very limited.

Interestingly, higher serum SFRP5 was associated with impaired renal function measured as estimated glomerular filtration rate. The link between SFRP5 and renal function has not been investigated in detail yet. One study reported that SFRP5 inhibited high phosphate-induced calcification of vascular smooth muscle cells which is a major cardiovascular risk factor in chronic kidney disease (CKD) [30]. However, another study showed that the Sfrp5 gene is hypermethylated and downregulated in a mouse model of CKD, whereas injection of recombinant Sfrp5 inhibited Wnt signaling and attenuated renal fibrosis [31]. Currently, data on the association between SFRP5 and eGFR or albuminuria as measures of kidney function from other study populations are not available, so that further studies are required to corroborate or refute our finding.

One important risk factor for type 2 diabetes and its comorbidities is subclinical inflammation [32]. In the current study, we found that higher circulating SFRP5 was associated with lower levels of the pro-inflammatory markers hsCRP and IL-6, and the anti-inflammatory protein IL-1Ra and with higher levels of adiponectin. BMI explained the associations between SFRP5, IL-6 and IL-1Ra, but not those with hsCRP and adiponectin. It might be surprising that high SFRP5 was associated with lower levels of the anti-inflammatory IL-1Ra which argues against an anti-inflammatory action of SFRP5 at a first glance. However, prospective studies clearly demonstrated that higher IL-1Ra levels were associated with a higher risk of type 2 diabetes and cardiovascular events [33‐35]. This association may at least partially reflect a response to proatherogenic factors such as subclinical inflammation (including IL-1β-mediated processes, which upregulate IL-1Ra) and oxidative stress.

Previous human cross-sectional studies supported the anti-inflammatory phenotype of SFRP5. They found that circulating SFRP5 was associated inversely with hsCRP, leptin, IL-6, TNFα and IL-1β and positively with adiponectin [12, 13, 15, 36, 37]. On the cellular level, the treatment with recombinant SFRP5 reduced the release of IL-6 by 49% in vitro in TNFα-induced primary human adipocytes. In that study, SFRP5 did not influence the activation of JNK, but decreased the activation of NFκB [26]. In obese mice Sfrp5 reduced the levels of the pro-inflammatory proteins TNFα, IL-6 and MCP-1 and the content of macrophages in adipose tissue [9]. Collectively, these data indicate that SFRP5 may have anti-inflammatory effects which could mediate the relationship between higher SFRP5 and lower cardiometabolic risk.

The most important confounder which influenced the associations between SFRP5 and risk factors for cardiometabolic diseases was BMI. In our study, high serum SFRP5 was associated with a lower BMI. This is in line with other human studies showing that circulating SFRP5 associated inversely with BMI [7, 13, 15, 36] and plasma levels of SFRP5 were decreased in obese people [11, 12]. Data from mouse studies reported that Sfrp5 levels depended on the duration of obesity. Mice fed a high-fat diet for a short time had increased levels of Sfrp5 in the adipose tissue, whereas a high-fat diet for a longer period of time decreased levels of Sfrp5 in the adipose tissue [9]. In a human study, caloric restriction led to a fast increase of serum levels of SFRP5 [5]. So far, it is not known whether SFRP5 plays a physiological role as mediator between BMI and obesity-associated comorbidities or between weight loss and an improved cardiometabolic risk profile.

The large sample size and the population-based design allowed a valid analysis of the association between circulating SFRP5 and cardiometabolic risk factors. An important limitation of our study is the cross-sectional design which precludes conclusions regarding the predictive value of circulating SFRP5 for the incidence of type 2 diabetes and cardiovascular events. Lower SFRP5 levels may be cause or consequence of prediabetes/type 2 diabetes. Our subanalyses in patients with type 2 diabetes showed no associations between metabolic control or medication and SFRP5 levels, which would argue against hyperglycaemia as cause of lower SFRP5. However, one study reported an increase of SFRP5 levels following glucose-lowering therapy [11], so that further studies are required to resolve the temporal relationship between changes in SFRP5 and changes in glucose metabolism. Additionally, participants and non-participants in the KORA F4 study showed some differences in their baseline characteristics in the population-based KORA S4 study, but these differences were small, so that we have no indication for a relevant selection bias due to loss to follow-up. Moreover, the study is based on older individuals of German descent which limits the generalisability of our observations to younger populations and people with different ethnic background.