Ferritin modifies the relationship between inflammation and arterial stiffness in hypertensive patients with different glucose tolerance

The study group consisted of 462 Caucasian newly diagnosed hypertensive patients, 216 men and 246 women aged 49.6 ± 12.2 years, participating in the Catanzaro Metabolic Risk factors Study. All subjects underwent physical examination and review of their medical history. Causes of secondary hypertension were excluded by appropriate clinical and biochemical tests. Other exclusion criteria were history or clinical evidence of coronary and valvular heart disease, congestive heart failure, peripheral vascular disease, chronic inflammatory diseases, anemia, gastrointestinal diseases with malabsorption, history of any malignant disease, history of alcohol or drug abuse, liver or kidney failure and diabetes already diagnosed. No patient had ever been treated with antihypertensive drugs. All subjects underwent anthropometrical evaluation with measurements of weight, height, and body mass index (BMI). After 12-h fasting, a 75 g oral glucose tolerance test (OGTT) was performed with 0, 30, 60, 90 and 120 min sampling for plasma glucose and insulin. Glucose tolerance status was defined on the basis of OGTT using the World Health Organization (WHO) criteria.

Measurements of clinic BP were obtained in the left arm of the patients, in supine position, after 5 min of quiet rest, with a aneroid sphygmomanometer. A mean value of at least three BP readings were taken in three different visits at least 2 weeks apart. Systolic and diastolic BP were recorded at the first appearance (phase I) and the disappearance (phase V) of Korotkoff sounds, respectively. A value of clinic systolic BP (SBP) ≥ 140 mmHg and/or diastolic BP (DBP) ≥ 90 mmHg identified patients as hypertensive according to current guidelines [22]. In addition, pulse pressure (PP) was defined as the difference between SBP and DBP.

All blood samples were obtained after overnight fasting. Serum ferritin levels (10–290 ng/ml) were measured using an immunoturbidimetric assay (Roche Diagnostics, Indianapolis, IN, USA). The serum iron (50–150 µg/dl) was measured with a colorimetric method and transferrin (2–4 g/l) by a nephelometric assay, while the percentage of saturated transferrin iron binding capacity (TIBC) was measured using a colorimetric test (Roche Diagnostics, Cobas 8000, Switzerland). Hemoglobin was determined using an automated particle counter (Siemens Healthcare Diagnostics ADVIA^® 120/2120 Haematology System, Milan, Italy).

Plasma glucose was measured by the glucose oxidation method (Beckman Glucose Analyzer II; Beckman Instruments, Milan, Italy). Triglyceride, total, low- (LDL) and high-density lipoprotein (HDL) cholesterol concentrations were evaluated by enzymatic methods (Roche, Basel, Switzerland). Serum insulin levels were determined by a chemiluminescence-based assay ((Immulite ^®, Siemens, Italy).). Insulin sensitivity was evaluated using the Matsuda index [insulin sensitivity index (ISI)], calculated as follows: 10,000/square root of [fasting glucose (millimoles per liter) × fasting insulin (milliunits per liter)] * [mean glucose * mean insulin during OGTT]. The Matsuda index is strongly related to euglycemic hyperinsulinemic clamp that represents the gold standard test for measuring insulin sensitivity [23].

Alanine aminotransferase (ALT) and aspartate aminotransferase (AST) levels were determined using the alpha-ketoglutarate reaction, and γ-glutamyltransferase (γGT) levels with the l-γ-glutamyl-3-carboxy-4-nitroanilide rate method (Roche, Basel, Switzerland). High sensitivity CRP (hsCRP) were measured by automated instrument (CardioPhase ^® hsCRP, Milan, Italy). Finally, creatinine levels were measured by Jaffe methodology and estimated glomerular filtration rate (e-GFR) was computed by using the chronic kidney disease epidemiology (CKD-EPI) collaboration equation [24].

All measurements were obtained using a validated system (Sphygmocor™; AtCor Medical, Sydney, Australia) that utilizes high-fidelity applanation tonometry (Millar) and an appropriate computer software for the evaluation of pressure wave (Sphygmocor™). At first, the pressure calibration was achieved by the non-invasive automatic recording of supine brachial artery BP in the dominant arm after a 30 min of rest (Dinamap Compact T; Johnson & Johnson Medical Ltd, Newport, UK). In particular, BP was measured five times over 10 min and the mean of the last three determinations was considered for calibration. On the radial artery of the dominant arm, the pressure wave was recorded as the average of single waves consecutively obtained for 8 s. Pressure wave determinations were considered reliable only if the variation of peak and bottom pressures of pressure waves was < 5%. The central pressure wave assessment was automatically derived from the radial measurements by a generalized transfer function [25]. Moreover, central waveforms were evaluated to identify the time to peak/shoulder of the first (T1) and second (T2) pressure wave elements during systole. The pressure at the peak/shoulder of T1 was classified as outgoing pressure wave height (P1), the pressure at the peak/shoulder of T2 was defined as the reflected pressure wave height (P2), either as an absolute value or as percent of ejection duration. Augmentation pressure (AP) was defined as difference between P2–P1, and augmentation index (AI) as [AP/pulse pressure (PP)] * 100. Aortic pulse wave velocity (PWV) was measured from carotid and femoral pressure waveforms. Carotid to femoral transit time (ΔT) was calculated from the foot-to-foot time delay between carotid and femoral waveforms. The distance between the landmarks of the sternal notch and femoral artery was considered to evaluate the path length between the carotid and femoral arteries (L), and PWV calculated as L/ΔT.

Data are reported as mean and standard deviation or as absolute and percent frequency, as appropriate. Comparisons among more than two groups were performed by One Way ANOVA, followed by a Bonferroni post hoc between-groups comparison. Chi-squared test was utilized for categorical variables. The association between continuous variables was assessed by Pearson product moment correlation coefficient (r) and p value. The relationship between gender and PWV was investigated by point biserial correlation coefficient and p value. As potential confounders for the relationship between ferritin and PWV we considered all variables listed in Table 1 and those associated with this biomarker with p ≤ 0.05 at univariate analyses (see last column in Table 1) were considered to be introduced into the multiple linear regression model. To avoid collinearity, we did not include iron and transferrin into the multiple liner regression model because these two variables lied into the same causal pathway between ferritin and PWV. Similarly, we did not adjust for insulin to avoid collinearity between this variable and the MATSUDA index. To assess the variance of PWV explained by each covariate into the model, we calculated the squared of the partial correlation coefficient. The effect modification by ferritin on the relationship between hs-CRP and PWV was investigated by introducing into the same linear regression model ferritin (the effect modifier), hs-CRP and their interaction term (ferritin * hsCRP) as well as a series of potential confounders. The estimated increase of PWV associated to a fixed increase in CRP (+ 1 mg/l) across quartiles of ferritin (36 ng/ml, 86 ng/ml, and 159 ng/ml) was investigated by the standard linear combination method [26]. In multiple linear regression models, data were expressed as standardised regression coefficients (beta) and p value. All statistical analyses were performed using SPSS version 22 for Windows (Chicago, Illinois, USA) and STATA statistical package (version 13, Texas, USA).

Out of 462 hypertensive (HT) patients examined by OGTT, 271 (58.7%) showed normal glucose tolerance (HT/NGT), 146 (31.6%) impaired glucose tolerance (HT/IGT) and 45 (9.7%) type 2 diabetes mellitus (HT/T2DM). The anthropometric, clinical and biochemical characteristics of the study population according to glucose tolerance status are shown in Table 1. Of note, ferritin, transferrin, TIBC, and hs-CRP increased from the first to the third group (Table 1) whereas serum iron and hemoglobin showed an opposite pattern (Table 1). It’s important to remark that hemoglobin and ferritin levels of all study population were within the normal range. Figure 1 reports two histograms plotting the distributions of ferritin levels separately in males and females.

Peripheral and aortic hemodynamic parameters for the study population, according to glucose tolerance groups, are presented in Table 2. There were no significant differences among groups for peripheral BP values and heart rate. On the contrary, from the first to the third group of glucose tolerance, we noted a significant increase of central (c)-SBP and c-PP, without significant difference for c-DBP. Hypertensive T2DM patients showed a significantly higher c-SBP and c-PP values than HT/NGT, but similar to HT/IGT. Similarly, all indices of vascular stiffness progressively increased from the first to third group. In particular, PWV values were significantly higher in HT/IGT and HT/T2DM patients in comparison with HT/NGT, but without any significant differences between these two groups. Similarly, also AI and AP values were significantly higher in HT/IGT and HT/T2DM patients than HT/NGT but substantially comparable between them (Table 2).

A linear correlation analysis was performed to assess the association between PWV and all variables listed in Table 1. PWV was related directly with ferritin, hs-CRP, age, insulin, AST, γGT, transferrin, and hemoglobin and inversely with the MATSUDA index, serum iron, e-GFR, and HDL cholesterol. In multiple linear regression model (Table 3) testing a series of univariate correlates of PWV (according to the strategy described in “Methods” section—“Statistical analysis”), ferritin was the strongest correlate of PWV explaining a 14.9% of the variance of this biomarker, followed by MATSUDA index (9.0%), hs-CRP (5.6%), e-GFR (1.9%), and HDL cholesterol (0.9%) (Fig. 2). Age, AST, γGT, and hemoglobin did not significantly contribute to explain the variance of PWV (Fig. 2). Overall, the model was able to explain a 39.9% of PWV variability.

Circulating levels of ferritin were directly related to those of hs-CRP (r = 0.20, p < 0.001). Thus, in Table 3 the effect modification by ferritin levels on the hs-CRP-PWV relationship was investigated by crude and adjusted analyses (see “Methods” section, “Statistical analysis” for details). On crude terms, serum ferritin was a strong modifier of the relationship between hs-CRP and PWV (P for effect modification = 0.01 and Fig. 3). The effect modification by ferritin on the hs-CRP-PWV relationship remained significant also after data adjustment for potential confounders (P for effect modification = 0.004, Table 3 and Fig. 3). As shown in Fig. 3, the estimated increase in PWV value portended by a fixed increase in hs-CRP (+ 1 mg/l) was progressively higher across increasing values of serum ferritin. These results indicate that high serum ferritin levels are not only a risk factor of arterial stiffness, but also amplify the harmful effect of inflammation on PWV.

In recent years the interrelationship between ferritin, glucose metabolism, arterial stiffness and inflammation has been extensively investigated. Scientific evidence has shown that high glucose levels also in the prediabetic range and insulin resistance might lead to increased arterial stiffness and cardiac concentric remodeling [27], moreover in a sample of hypertensive patients a significant positive association between baseline brachial-ankle PWV and the risk of new-onset diabetes was found [28]. Also serum ferritin levels have been associated with new-onset T2DM risk in Chinese [29] and Korean [30] populations and with increased arterial stiffness in healthy Korean adults [20]. Moreover, besides classical confounders, 24-h ambulatory BP and hs-CRP levels were determinants of pharmacologically (empagliflozin) induced improvement of arterial stiffness parameters [31]. Furthermore, a strong association between serum ferritin levels and atherosclerotic injury has been reported in different settings of patients [32, 33]. In particular, increased common carotid artery intima-media thickness and presence of carotid plaques has been associated with serum ferritin in haemodialysis patients [32] and patients with non-alcoholic fatty liver disease [33]. However, there is growing evidence demonstrating a strong association between higher serum ferritin levels and functional vascular damage. According with this, Ha et al. reported, in a large cohort of healthy subjects, an association between ferritin and brachial-ankle PWV [20]. Moreover in haemodialysis patients, serum ferritin was associated with progressive arterial stiffness increase during a 3 years follow-up, but only for serum values > 500 ng/ml [21]. In hypertensive patients, only few data are reported in literature. In this regard, Valenti and co-authors [18] showed that hyperferritinemia (i.e. higher than 240 ng/ml in females and 320 ng/ml in males) was associated with high aortic stiffness measured as carotid-femoral PWV in patients with well-controlled hypertensives but with several comorbidities and already receiving drug treatment whereas data about the glycometabolic state were not available in this study. All these factors may explain why hs-CRP was not retained as an independent risk factor of PWV in the multiple regression analysis [18]. In respect to the study by Valenti et al. our study has the advantage to include younger naive hypertensive, without chronic conditions (in particular renal and liver diseases) and who received no pharmacological treatment thus avoiding/minimizing possible confounding factors. Remarkably, all patients were evaluated by OGTT to define glucose tolerance and the effect of insulin sensitivity on vascular damage was assessed by Matsuda index, a strong surrogate of euglycemic clamp. Finally, the most important observation was that, in our study, serum ferritin levels were within the normal range value.

Patients included in our study present iron levels within the normal range and for this reason the association between ferritin and PWV cannot be explained by iron overload. However, it’s known that ferritin is strictly regulated by cytokines at various levels (transcriptional, posttranscriptional and translational) during cellular differentiation, proliferation and inflammatory burden [34]. In particular, it has been demonstrated that tumor necrosis factor alpha (TNFα) and interleukin 1a (IL1α) are able to induce the expression of the H chain of ferritin at the transcriptional level in different cell types [35]. Moreover, cytokines have also posttranscriptional effects on ferritin expression. According with this, translation of ferritin is induced by IL-1α, IL-6 or TNFα in the HepG2 hepatic cell line [35, 36], and iron is required for this action in fact the ferritin upregulation may be inhibited by deferoxamine (an intracellular iron chelator) [35]. These studies along with cytokine-mediated regulation suggest that inflammation can affect ferritin regulation. Thus, it’s plausible to speculate that in chronic conditions characterized by subclinical inflammation, such as arterial hypertension and T2DM, dysregulation of iron metabolism may lead to increased oxidative stress promoting arterial stiffening. Our results and interpretation are in keeping with those reported in two previous population-based studies [37, 38]. In a paper by Wolff et al. [37], in both men and women a dose–response relationship was found between serum ferritin levels and carotid atherosclerosis, so that higher serum ferritin levels were associated with greater risk of carotid plaque development. In another paper by Ahluwalia et al. [38], a significant interaction was found between inflammation and ferritin levels to explain the severity of carotid atherosclerosis in the study sample. Given the fact that the effect modification is a mutualistic phenomenon, Ahluwalia et al. proposed the hypothesis that inflammation may act as an effect modifier of the relationship between ferritin and PWV [38].

It is well known that serum ferritin is an acute-phase reactant, reflecting the degree of acute and chronic inflammation in several systemic diseases. However, emerging evidence suggests a potential causative role for ferritin in the inflammatory process [4]. In this regard, in vitro studies showed that ferritin may work as a local cytokine, activating Mitogen-Activated Protein Kinases (MAPK)-induced NF-kB in an iron-independent manner. This activity induces the increased expression of multiple pro-inflammatory mediators and a near 100-fold increase in inducible NO synthase which amplify oxidative stress [39, 40]. In addition, ferritin is also able to directly modulate the lymphocyte function thus acting as pathogenetic player in the innate immune response [41]. Considering that our study population presented a normal inflammation and iron status, it is also clinically plausible that ferritin per se might affect the increase in oxidative stress promoting those mechanisms that lead to vascular remodeling and increased arterial stiffness. Clinical practice supports this hypothesis, because ferritin represents a biomarker of disease progress and an independent predictor of various clinical outcome in different settings of patients [42‐44]. Since the increase in carotid femoral PWV is associated with a significant risk increase for CV morbidity and mortality, we can speculate that the association between serum ferritin levels and PWV may be responsible for the negative prognostic effect of ferritin. Overall, the studies performed so far do not fully clarify which is the cause and which is the effect in the complex pathway involving ferritin, inflammation and atherosclerosis. For this reason, any hypothesis is purely speculative and the exact mechanisms underlying the relationship between ferritin and carotid atherosclerosis in humans with normal iron levels still remains to be fully elucitaded.