Serum H‐Ferritin‐To‐Ferritin Ratio as a Biomarker of Hemophagocytic Lymphohistiocytosis in Critically Ill Patients With Hyperferritinemia

A wide range of conditions, including malignancies, infections, autoimmune autoinflammatory diseases, and more recently described adverse effects of immunotherapies, can trigger a cytokine storm responsible for a devastating dysregulated immune response. Among cytokine storm syndromes, hemophagocytic lymphohistiocytosis (HLH) is a hyperinflammatory hyperferritinemic syndrome resulting from a highly stimulated but ineffective immune response, leading to potentially fatal multiorgan damage [1]. HLH can be suspected based on elevated serum ferritin levels [2, 3] but the very high levels of ferritin specific to hyperinflammatory states are reached when the disease progression is too advanced to be of any clinical use. Among the patients with hyperferritinemia admitted to intensive care units, HLH is found in 1.5% [4]. Other frequently encountered causes of hyperferritinemia in critically ill patients are sepsis, liver diseases, and hematological malignancies, conditions that may either mimic or trigger HLH [5]. As the clinical and laboratory features of HLH and sepsis frequently overlap, a reliable marker to differentiate HLH is needed. The human ferritin is a ubiquitous iron-storage protein, composed of 24 subunits with 2 types of peptide chains: light (FeL) or heavy (FeH). The ratio of FeL:FeH subunits varies according to the physiological status of the cell and tissue function [6, 7]. In normal conditions, the serum ferritin is predominantly composed of the L subunit and positively correlated with the size of the total body iron stores [8]. Immunoassays used in clinical laboratories recognize solely the L subunit of ferritin. However, inflammatory conditions have been shown to modulate the relative expression of the H and L subunits of ferritin, with most studied pro-inflammatory stimuli preferentially upregulating FeH synthesis [9]. Increased levels of FeH have been observed in ex vivo models, in bone marrow and liver tissue from patients with HLH [10], and its pro-inflammatory properties have been demonstrated on human macrophage cultures [11]. To our knowledge, quantification of circulating FeH in blood has not yet been reported, especially during acute inflammatory processes [11-13]. To improve the specificity of ferritin's assay, we used mass spectrometry (MS)-based proteomics to quantify both L-Ferritin (FeL) and H-Ferritin (FeH) in human sera. Starting from an untargeted liquid chromatography-tandem MS (LC–MS/MS) analysis, we developed a targeted LC–MS/MS method based on multiple reaction monitoring (MRM). First, we performed untargeted shotgun analysis of 6 sera of patients with hyperferritinemia (> 5000 μg/L), to maximize our chances of identifying tryptic signature peptides of FeL and FeH among peptides from other more abundant sera proteins. Among identified peptides, we selected those with optimal characteristics and highest signal intensity, IFLQDIK (from FeH) and ALFQDIK (from FeL) to further quantification by MRM LC–MS/MS analysis. MRM experiments were optimized using the obtained MS/MS spectra (shown in Figure 1). For MRM LC–MS/MS analysis, a calibration curve was drawn up on each series of samples to reduce the level of analytical variation. The analytical precision of the calibration standards ranged from 6% to 9%. The analytical precision of the internal quality controls with different values was lower than 15% (according to the CLSI C62 guidelines). To assess the diagnostic performances of FeL and FeH measurements, we studied a cohort of critically ill patients admitted to intensive care units with hyperferritinemia (n = 50). This cohort included 22 patients with HLH, classified as such if four or more HLH-2004 criteria were met [14] or the HScore was ≥ 169 [15], and 7 patients with macrophage activation-like syndrome (MALS), a particular endotype of sepsis, characterized by FeL above 4420 μg/L and associated with a rapid progression to death similar to the course of HLH [16]. The control group consisted of 21 patients with suspected but unconfirmed HLH and patients with sepsis, severe sepsis, and septic shock (criteria described in [17]), associated with hyperferritinemia (> 500 μg/L). Clinical characteristics of the patients are outlined in Table 1. In HLH patients, a significantly increased FeH-to-total-ferritin ratio was observed compared to patients with sepsis and unconfirmed HLH (p < 0.05), as shown in Figure 2A. Importantly, at equivalent FeL levels, the serum FeH-to-total-ferritin ratio increased significantly in patients with HLH and MALS. The FeH ratio allowed to discriminate adult HLH patients from patients with suspected but unconfirmed HLH, whereas no such differences were observed for FeL levels, as shown in Figure 2B. The FeH ratio allowed to distinguish the patients with more severe clinical conditions in our cohort. These findings reinforce the hypothesis that inflammatory cytokine-induced FeH becomes more abundant during times of active immunologic response, as reviewed by Kernan et al. [9]. At a certain level of FeH increase, its imbalance with FeL would turn into an injurious process, as suggested by Zarjou et al. [18]. To determine the potential diagnostic value of the FeH-to-total-ferritin ratio, we performed ROC analyses comparing the patients with HLH and MALS to those with unconfirmed HLH, sepsis, severe sepsis, and septic shock (Figure 2C). Sensitivity and specificity were determined for the cut-off with the highest Youden index. We defined a cut-off of FeH-to-total-ferritin ratio of 6 and a cut-off of FeL levels of 4465 μg/L, predictable of HLH and MALS. The FeL levels achieved an AUC of 0.85 (IC: 0.76–0.96), sensitivity of 90% and specificity of 71%. The FeH ratio achieved an AUC of 0.85 (IC: 0.72–0.98), sensitivity of 93% and specificity of 86%. With a fixed sensitivity set at 93%, the FeH ratio was the biomarker with the highest specificity (86%) compared to FeL (67%), demonstrating that the FeH-to-total-ferritin ratio was better able to identify HLH and MALS compared to FeL levels. The FeH ratio and FeL levels are shown in Figure 3 for the different patient groups. Of 22 patients with HLH, one yielded discrepant results presenting a FeH ratio of 17 and FeL levels of 1081 μg/L; this patient had a diagnostic H-score equal to 209. Of nine patients with suspected but unconfirmed HLH, 6 presented a FeH ratio below 6 and FeL levels above 4043 μg/L, including 5 cases with underlying hematologic malignancies, all had an H-score below 169. To better understand the cytokine response, we evaluated whether the FeH-to-total-ferritin ratio and the FeL levels correlated with laboratory parameters of HLH (ASAT, sCD25) and other pro-inflammatory cytokines (sCD163, CXCL9, INFɣ, IL-18). We showed that the FeH ratio was correlated to soluble CD163, released from activated macrophage CD163+ (r = 0.37, p = 0.01), and with circulating levels of soluble CD25 (r = 0.39, p = 0.007), related to an overactivation of cytotoxic T lymphocytes CD25+. By contrast to the FeH, the FeL levels correlated with ASAT (r = 0.43, p = 0.003), reflecting the degree of liver damage. These observations provide further evidence that in HLH, the circulating FeH is markedly released by activated macrophages, and its hepatic origin is not predominant. In parallel to this increased secretion, the high levels of FeH might also be the result of a decreased uptake and recycling of circulating ferritin by receptor-mediated mechanisms potentially affected by abnormal ratios of FeH and FeL or a disturbed expression of receptor levels on cell surface. Altogether, our finding suggests that the disruption of the FeH ratio is an earlier marker than FeL to identify patients at risk of developing cytokine storm syndromes. To take full advantages of new targeted therapeutics, notably the recombinant antagonist of the IL-1 receptor, it is essential to rapidly identify patients with HLH and MALS who are most likely to benefit from these therapies [19, 20]. Here, we propose that the FeH-to-total-ferritin ratio could be used to timely diagnose HLH. In this purpose, the implementation of an automated immunoassay in clinical practice would represent an ideal diagnostic tool for saving on the hand-on-time requirement of sample preparation and analysis required for mass spectrometry. In conclusion, FeL and FeH are both detected in human serum during acute inflammatory processes. FeH-to-total-ferritin ratio is a better diagnostic marker than FeL to distinguish HLH in critically ill patients and allows a better stratification of septic patients with valuable implications to optimize new targeted therapeutic strategies. The authors have nothing to report. Prospective observational trial NCT02143986, according to a protocol approved by the institutional ethics committees of the four participating hospitals (ce2014/33, AK/15-06 69/4524, 2411, ce45/15). The authors declare no conflicts of interest. The data that support the findings of this study are openly available in PRoteomics IDEntifications database at https://www.ebi.ac.uk/pride/, reference number PXD065102.