Hepcidin and ferritin levels in restless legs syndrome: a case–control study

The association between restless legs syndrome (RLS) and iron homeostasis remains unclear. We compared serum hepcidin and ferritin levels in patients with RLS and controls, and assessed their relationships with RLS phenotype, drug intake, and history of augmentation syndrome. 102 drug-free RLS patients (age 58.9 [24.5–77.2], 63 females) and 73 controls (age 56.8 [23.46–76.6], 45 females) underwent a polysomnography recording. Hepcidin levels were quantified by ELISA. 34 RLS patients had a second assessment after starting dopaminergic drugs. Ferritin level was low (< 50 µg/l) in 14.7% of patients and 25% of controls, with no between-group differences in the mean values. Hepcidin levels were higher in patients even after adjustment for confounding factors, and excluding participants with low ferritin levels. Ferritin and hepcidin levels were comparable before and after treatment, and between patients with (n = 17) and without history of augmentation. Ferritin and hepcidin levels correlated with age, body mass index, and periodic leg movements. Higher hepcidin levels were associated with older age, older age at RLS onset, less daytime sleepiness and familial RLS. In conclusion, serum hepcidin levels but not ferritin were higher in RLS patients regardless of treatment and history of augmentation. Serum hepcidin may be a more relevant biomarker of RLS than ferritin.


Ferritin and hepcidin quantification in drug-free patients with RLS and controls. Compared
with controls, serum ferritin levels were slightly higher in the RLS group in unadjusted associations, but not after adjustment for BMI (Table 2). Similar results were obtained when ferritin levels were divided into tertiles. Ferritin levels were low (< 50 µg/L) in 15 (14.71%) patients with RLS and in 18 (25%) controls, with no betweengroup difference. None of participants were taking iron replacement therapy for at least 2 years.
Conversely, hepcidin levels were significantly higher in patients than in controls, even after adjustment for BMI (median [ranges] 18.36 µg/L [0.00-77.17] vs 11.89 [0.00; 70.51], p < 0.01) ( Table 2). Five patients with RLS and four controls had undetectable hepcidin levels (i.e. below the detection limit 26 ). When hepcidin levels were divided into tertiles, we found that the number of patients in each tertile progressively increased in contrast to controls. ROC analyses were performed to determine the cutoff values for the hepcidin levels with the highest sensitivity and specificity to discriminate patients with RLS from controls. The area under the curve values was 0.62 (95% CI [0.55-0.71]) for hepcidin level. The best cut off hepcidin value to discriminate patients with RLS from controls was 18.1 µg/L, with a sensitivity of 52% and a specificity of 75%, and with significant betweengroup differences (p = 0.0004).

Determinants of high serum hepcidin levels in drug-free patients with RLS. To identify factors
that could be associated with high hepcidin levels in RLS, patients with RLS were divided in two groups based on the highest serum hepcidin tertile (≥ 20.99 ng/mL and < 20.99 ng/mL). High hepcidin level (≥ 20.99 ng/mL) was associated with older age, older age at RLS onset, older age at daily RLS onset, positive family history of RLS, lower daytime sleepiness (ESS score), lower total sleep time, and higher WASO, REM sleep latency, apnea-hypopnea index (AHI) and PLMW. RLS was very severe (score > 30) in 38.1% of patients with high hepcidin level and in 26.8% of patients with low hepcidin level (< 20.99 ng/mL). RLS was moderate in 40.5% of patients with high hepcidin level and in 64.3% of patients with low hepcidin level, suggesting a trend towards a non-linear relationship (U-shaped curve) between hepcidin level and RLS severity (p = 0.05) ( Table 3). After adjustment for age, the associations between hepcidin levels, daytime sleepiness, family history of RLS, total sleep time, and WASO remained significant; however given the multiple testing performed these results should be taken with caution. .52] during treatment with dopaminergic agonists (n = 29 with rotigotine alone; n = 2 with rotigotine and pregabalin; and n = 3 with pramipexole). After drug intake, the RLS severity, ESS, ISI and BDI scores improved as well as the PSG characteristics, including sleep efficiency, WASO, PLMS and PLMW (Table 4). Conversely, serum ferritin and hepcidin levels did not change after starting treatment. None of these 34 patients had augmentation at time of the second evaluation.
Comparison of methods of quantification of hepcidin. We quantified serum hepcidin levels using two different methods by LC-MS/MS assay as previously reported 9 . The Bland-Altman plot showed that four individuals were out of the agreement limits ( Fig. 1). Hepcidin level was lower when measured with the LC-MS/MS assay than with the ELISA method (mean difference = − 6.05 95% CI [− 9.44; − 2.66] p = 0.0007). Concerning the intra-individual reproducibility, a significant difference in the variability between the two methods was found (Pitman's test of difference in variance, r = 0.552, p < 0.0001).

Discussion
This study using a validated ELISA assay showed that serum hepcidin levels are higher in patients with RLS compared with healthy controls, in both unadjusted and adjusted models, and also after excluding subjects with low ferritin levels. This higher serum hepcidin level was not associated with dopaminergic drug intake and with recent history of augmentation. Conversely, serum ferritin levels did not differ between patients with RLS and controls, neither after dopaminergic drug intake nor augmentation.
The 25-amino acid-long hepatic bioactive form of hepcidin is released after proteolytic cleavage of the 84-amino acid-long prohepcidin precursor and plays a major role in iron homeostasis. Hepcidin-25 synthesis is inhibited by erythropoietic demand, hypoxia, and low iron content, while inflammation and high iron content stimulate its expression 17,27 . Hepcidin-25 decrease mediates the internalization and subsequent degradation of the iron cell exporter ferroportin, leading to cellular iron retention. Although detectable in biological fluids, the role of the other hepcidin isoforms (20, 22, 24 amino acids in length) remains unknown.
In the last few years, different immunochemical and mass spectroscopy-based assays have been used to quantify serum hepcidin-25 level in humans, but its clinical relevance remains to be determined. We previously found using LC-MS/MS in the MRM mode (the reference method for peptide quantification 9,26 ) that serum hepcidin levels were higher in patients with RLS than in controls, and were correlated with RLS severity (nonlinear relationship), and with PLMS. By using an ELISA assay with a monoclonal antibody, another reliable but less expensive method for hepcidin quantification that could be easily implemented in clinical settings 28 , we confirmed the higher serum hepcidin levels in patients with RLS than in controls. These results remained significant even after exclusion of participants included in the previous hepcidin measurement RLS study tested by LC-MS/MS study 9 . Here, the population of consecutive patients with RLS was diverse, in terms of severity, range www.nature.com/scientificreports/ of ferritin levels, dopaminergic drug intake, and recent history of augmentation syndrome. The between-group difference remained significant even after adjustment for confounding factors, and after excluding participants with ferritin levels below 50 µg/L. Indeed, participants with low ferritin levels were not excluded, but those taking ongoing iron replacement therapy or in the last 2 years. Comparison of the two methods (LC-MS/MS and ELISA) showed that the ELISA-based test gave higher values, possibly due to immunoassay interference with the recognition of other hepcidin isoforms by the used antibody, in agreement with previous reports 29 . Despite the absence of standardized cut-offs for serum hepcidin levels, high serum hepcidin levels (≥ 20.99 ng/mL) were associated with older age, older age at RLS onset, positive family history of RLS, lower excessive daytime sleepiness; however these results should be taken with caution regarding the multiple testing performed and the sample size of the population. The best hepcidin cut Although serum ferritin is considered as the most informative RLS biomarker, in our study ferritin level was not different between patients with RLS and controls, as previously reported [9][10][11][12][13] . Moreover, the percentage of subjects with ferritin level < 50 ng/mL was similar in the RLS and control groups. However, identifying patients with low ferritin level remains important for RLS management, particularly because iron deficiency must be treated before considering treatment with a dopaminergic agonist. The value of measuring hepcidin before therapeutic management should be evaluated in future studies.
Serum ferritin and hepcidin levels were not different between patients tested before and during treatment with dopaminergic agonists, and between patients with and without recent history of augmentation syndrome. Conversely, a previous report showed an association between low ferritin levels and augmentation syndrome 14 . Our RLS group with augmentation syndrome was rather small, with retrospective diagnosis, and blood was not www.nature.com/scientificreports/ collected at the exact time of the augmentation syndrome, but after an interval of 82.12 days [range  due to management of augmentation in out-patient clinic (reducing or eliminating dopaminergic therapy, switching to alpha-2-delta ligands, opiates, or combined therapy), with study inclusion requiring PSG recording in stable condition. However, none of these 17 patients were treated with iron intake in the past two years. Serum hepcidin and ferritin levels do not change quickly from one month to another in the absence of associated specific conditions (e.g., iron intake, specific diet, inflammation, surgery) that did not apply to the patients included in the present study. Based on this limited data and its retrospective nature of assessment, further studies are required to elucidate whether hepcidin levels may predict subsequent development of augmentation. Altogether, our results suggest that hepcidin may be a relevant RLS biomarker in terms of pathophysiology but less for diagnosis. The high serum hepcidin levels with normal ferritin concentrations in RLS may reflect a dysregulation of blood iron homeostasis rather than an absolute iron deficiency. This condition may mimic a functional peripheral iron deficiency, as we previously hypothesized 9 , thus limiting iron availability and its redistribution to different organs, including the brain, as part of RLS pathogenesis. However, hepcidin was assessed in serum samples, and may not reflect brain hepcidin concentrations 9 . High hepcidin levels have also been reported in other conditions (e.g., hemochromatosis 30 , metabolic syndrome 31 , atherosclerosis 32,33 , inflammation 34,35 , renal and cardiovascular diseases [36][37][38] ) that may be associated with RLS, but almost not in our current RLS population due to our exclusion criteria. Indeed, recent studies strongly suggest that RLS should be seen as a continuous spectrum with a major genetic contribution, and a major environmental or comorbid disease contribution.
The present study has several strengths. We studied here consecutive patients with RLS, with diversity in terms of severity, range of ferritin levels, dopaminergic drug intake, and history of augmentation syndrome. All participants were well-characterized (RLS phenotype and comorbidities) and all underwent one-night PSG recording. We included patients with RLS associated with different conditions: idiopathic form, iron deficiency, drug-free and with treatment intake. A subgroup of patients with RLS was evaluated twice in similar condition (before and during treatment with dopaminergic agonists). Venous blood samples were collected between 7 and 8 a.m. after overnight fasting to avoid errors potentially related to circadian hepcidin variability. No participant had history of iron intake in the past 2 years. Hepcidin-25 was assessed using a reliable commercial ELISA assay that can be performed in the routine practice.
The present study also has some limitations. Despite some adjustments, unmeasured confounders, such as inflammatory and gastrointestinal conditions, could explain part of the associations observed with serum hepcidin level. Neither marker of inflammation, liver or renal function, no other measures of iron homeostasis were evaluated concomitantly with serum hepcidin and ferritin quantification. Seventy-five subjects (42 RLS patients and 33 controls) participated in a previous study that quantified serum hepcidin levels by LC-MS/ MS, as comparison of methods of quantification of hepcidin was one secondary objective of the present study.
Our clinical RLS population sample was well characterized but relatively small that may not reflective of RLS To conclude, using a reliable commercial ELISA assay, our study confirms higher serum hepcidin levels in patients with RLS, a biomarker that is not influenced by the treatment status or by recent history of augmentation. Due to the easy-to-use ELISA assay, hepcidin quantification could be implemented in the routine clinical practice, in view of a possible personalized approach before therapeutic management and potentially with hepcidin antagonists in some patients with RLS.

Methods
Population. One hundred and two drug-free consecutive adult patients with moderate to very severe RLS (63 women; median age 58.91 years [24.50-77.22]) were included. All patients underwent a face-to-face semistructured evaluation and clinical examination by sleep experts to confirm RLS diagnosis and to exclude mimic conditions, using standard criteria as previously described 1,9 . Age at RLS onset, age at daily RLS onset, and family history of RLS were collected, as well as history of chronic diseases, such as hypertension, sleep apnea syndrome, cardiovascular, neurological, liver and renal diseases, diabetes and dyslipidemia.
None of patients had recent history (i.e. during the last 2 years) or ongoing iron intake, but iron deficiency (serum ferritin level < 50 ng/mL) discovered at time of study was not an exclusion criteria. All medications taken during the last month were recorded as well as their posology, and age at intake onset. No patient took drugs known to worsen (antidepressant, antihistaminic, or antipsychotic drugs) or alleviate (DAs, levodopa, alpha2-delta ligands, clonazepam and opioids) RLS symptoms for at least 2 weeks before study entry. All patients with RLS comorbid with clinically significant liver, renal, inflammatory, psychiatric or neurological diseases were excluded. RLS severity was evaluated according to the International Restless Legs Syndrome Study Group (IRLSSG) criteria (score ≤ 10: mild RLS; 11-20: moderate RLS; 21-30: severe RLS; and > 30: very severe RLS) as detailed elsewhere 9 . RLS augmentation during the last two years before the study was diagnosed retrospectively according to published criteria 16 .
Seventy-three community-dwelling adult controls (45 women, median age 56.85 years [23.46-76.65]) were recruited through advertisements and local association networks, during the same period, from the general population. All controls were free of RLS, liver, renal, inflammatory, psychiatric, and neurological diseases.
None of participants (patients and controls) took drugs known to affect RLS (dopaminergic agonists, L-DOPA, alpha-2-delta ligands, benzodiazepines or opioids), iron drugs, or any other medication known to influence sleep or movement (antidepressant, antihistaminic, or antipsychotic drugs) for at least 2 weeks before study inclusion as detailed elsewhere 9 .
This study was approved by Montpellier University review board, France, and all participants signed the informed consent. The study was performed in accordance with the Declaration of Helsinki and the French Good Clinical Practices.
Polysomnography. All participants underwent a single one-night polysomnographic (PSG) recording in the sleep laboratory with six electroencephalogram leads, two electro-oculograms, one chin electromyogram, and electrocardiogram. Respiration was monitored with a nasal cannula/pressure transducer, mouth thermistor, chest and abdominal bands, and pulse oximeter. Sleep stages, microarousals and respiratory events were scored manually according to standard criteria as detailed elsewhere 9,24 . Surface electromyogram electrodes on the anterior tibialis muscles recorded limb movements (LM Blood analyses. Venous blood samples were collected between 7:00 and 8:00 a.m. after overnight fasting, following standardized procedures. Serum ferritin levels were quantified within few hours after blood collection by electrochemiluminescence immunoassay (Elecsys, ECLIA, Cobas; Roche Diagnostics International Ltd, Switzerland). Serum samples from patients and controls were prepared in the same way and frozen at − 80 °C for hepcidin quantification. Hepcidin-25 (the 25-amino acid active isoform) was measured using the validated HS ELISA kit (DRG Diagnostics GmbH, Marburg, Germany). This assay relies on competitive binding detection, and was performed according to the providers' instructions (DRG, EIA-5782).
Hepcidin-25 was also measured by LC-MS/MS, as previously described, in a subgroup of 75 subjects (42 patients with RLS and 33 controls) 9,26 . Statistical analysis. Categorical variables are presented as numbers with percentages, and quantitative variables as medians with ranges. The Shapiro-Wilk test was used to test the normality of continuous variables. Regression logistic models were performed to compare demographic, clinical, and PSG characteristics between groups (controls and patients with RLS). Demographic variables significantly different in univariate analyses (p < 0.10) were included in logistic models to analyze the adjusted relationships between serum ferritin and hepcidin levels and the two groups. Associations were quantified with odds ratio (OR) and the 95% confidence interval (CI). The relationships of demographic, clinical and polysomnographic variables with hepcidin levels, Scientific RepoRtS | (2020) 10:11914 | https://doi.org/10.1038/s41598-020-68851-0 www.nature.com/scientificreports/ categorized in two groups (≥ 20.99 µg/L that corresponds to the highest tertile in the RLS group versus < 20.99), were analyzed using logistic regression models. Hepcidin level was also dichotomized between patients with RLS and controls to compare the diagnostic performance using receiver operating characteristic (ROC) curves. The best possible threshold was defined by the highest Youden Index [(specificity (sp) + sensitivity (se)) − 1]. The Wilcoxon signed rank test was used to analyze differences between repeated and continuous measures (before and after treatment). The Cohen's Kappa coefficient and the Bowker's test of symmetry were used to test the agreement of categorical variables. Spearman's rank order correlations were used to determine associations between continuous variables. The agreement of the two methods to measure hepcidin (LC-MS/MS and ELISA assays) was described using Bland-Altman plots and the intra-individual reproducibility of the method was analyzed using Pitman's test. The significance level was set at p < 0.05. Analyses were performed using SAS (version 9.4; SAS Inc., Cary, NC, USA).