COVID-19 and cytokine storms in adults

Initially described in the early 1990s, following the administration of the anti-T-cell antibody muromonab-CD3, the cytokine storm has been linked to different other conditions and therapeutic agents overtime, thereby broadening the spectrum of this disease.⁸ Nowadays, the term ‘cytokine storm’ applies to several fields of medicine being a potentially serious complication of immunotherapy and several infectious, neoplastic and autoimmune diseases.

Haemophagocytic lymphohistiocytosis (HLH) is a T-cell-driven hyperinflammatory, hyperferritinemic condition characterised by persistent IFNγ-dependent stimulation of Toll-like receptors (TLRs), antigen-presenting cells and T-cell uncontrolled activation, ultimately leading to a cytokine storm.⁹ According to the 2019 Histiocyte Society recommendations, HLH therapeutic strategies are based on the concept that it is a heterogeneous disease of variable aetiology and severity; hence, treatment should be tailored to control hyperinflammation, with glucocorticoids, intravenous immunoglobulin, cyclosporine and etoposide being the historical anchor drugs, and to treat identified disease triggers. Other immunomodulating agents like IL-1 or IL-6 inhibitors are only recommended in selected cases including the macrophage activation syndrome (MAS), a subtype of sHLH associated with systemic juvenile idiopathic arthritis (sJIA), adult-onset Still’s disease (AOSD) and other autoimmune disorders.¹⁰

A context of infection by a new viral agent adds another layer of complexity to the management of the cytokine storm, thus challenging the translation of the evidence from autoimmune diseases. The causal relation between SARS-CoV2 and sHLH relies in its ability to bind TLRs and induce the above-mentioned cascade via activation of the inflammasome and release of IL-1ß.¹¹ Importantly, lung tissue samples collected from patients with SARS-CoV during the SARS epidemic in the early 2000s have been reported to exhibit hemophagocytosis features.¹² However, not all patients with COVID-19 develop a fully blown sHLH, and those with worse prognosis may neither fulfil HLH-2004 diagnostic criteria (table 1)¹³ nor reach the HLH-probability calculator (HScore) cut-off values. The authors who developed the HScore pointed out that as each underlying disease is associated with peculiar biologic abnormalities, the cut-off values for laboratory criteria may need to be set accordingly.¹⁴ Ruan et al described significantly higher ferritin levels in severe COVID-19, putting forward the hypothesis that it can represent a prognostic marker in this disease. However, the mean ferritin values in patients with better prognosis were still over the HLH-2004 cut-off (≥500 µg/L), and it is unclear how many of the deceased patients reached the HScore ferritin cut-off (2000 ng/ml) and at which stage of the disease.¹⁵ Recently, it has been proposed that AOSD, MAS, catastrophic anti-phospholipid syndrome and septic shock, characterised by similar clinical picture and very elevated serum ferritin levels, are included in the so-called hyperferritinaemic syndrome.¹⁶ Interestingly, an association between pulmonary involvement and HLH was reported in patients with sJIA mediated by gene pathways related to the IFNγ response,¹⁷ and lung imaging pattern and inflammatory mediators appear to be similar to those of patients with COVID-19. On this basis, COVID-19 with pulmonary involvement could be considered itself a hyperferritinaemic syndrome, regardless of concomitant sHLH.¹² The identification of hemophagocytosis features in COVID-19 lung tissue, the assessment of ferritin and proinflammatory cytokines trends overtime, along with receiver operating characteristic curves computed with data of patients with COVID-19 would add significant knowledge on this matter. In particular, this would allow (1) to confirm whether COVID-19 is yet another hyperferritinemic syndrome, (2) to clarify whether current HScore cut-off values are reliable in these patients or should be redefined and (3) to understand the actual extent of the cytokine storm in all patients with COVID-19, regardless of the severity, and ultimately to support clinicians in the decision of when and in whom to start immunomodulating therapy and with which compound. Moreover, it is important to understand whether in severe forms of COVID-19-associated pneumonia, ferritin is a mere acute phase protein or rather an active pathogenic mediator. In addition, in light of the pro-inflammatory role of ferritin enriched in heavy subunits (heavy ferritin), it could be of interest to evaluate the heavy:light ferritin ratio in COVID-19.¹⁸ Finally, the potential usefulness of the ferritin:erythrocyte sedimentation rate ratio, which may be an immediate, inexpensive and easy-to-implement biomarker, needs to be considered.¹⁹

COVID-19 and cytokine storms in children

Despite earlier reports outlined that COVID-19 morbidity and mortality in children was very low,²⁰ very recently, several children with a novel multisystem inflammatory disease resembling toxic shock syndrome (TSS) and atypical Kawasaki disease (KD) with proven SARS-CoV2 infection have been observed in the UK. Likewise, an increased frequency of KD has been reported in Italy.^{21 22} TSS is a rare, life-threatening condition due in most cases to toxins produced by Staphylococcus aureus bacteria but also related to group A Streptococcus. TSS is characterised by fever, diffuse macular erythroderma, desquamation 1–2 weeks after onset of rash, hypotension and multisystem involvement of three or more of the following organ systems: gastrointestinal, muscular, mucous membrane, renal, hepatic, hematologic, central nervous system.²³ However, to be classified as probable or definite TSS, the clinical picture is necessary but not sufficient and either of the two above-mentioned bacteria need to be isolated from biologic samples. In the current circumstances, whether viral isolates may be also considered remains to be elucidated, as it is tempting to speculate that the SARS-CoV2, alone or in combination, could be triggering or facilitating the TSS. On the other hand, KD is an acute systemic vasculitis characterised by fever, non-suppurative conjunctival injection, rash, oral mucositis, extremity changes and cervical lymphadenopathy, and along with scarlet fever, meningococcaemia and other conditions represent a differential diagnosis of TSS. It comes with no surprise that elevated IL-1b and IL-6 levels in addition to a monotype profile shifted to intermediate monocytes are common hallmarks in KD.²⁴ MAS is a complication in about 2% of patients with KD; however, the incidence is likely underestimated as many clinical and laboratory features of both diseases overlap and the HLH 2009 criteria have low sensitivity and specificity for the diagnosis of MAS complicating KD.²⁵ Recently, classification criteria for MAS in the paediatric population have been released and will allow to provide more reliable epidemiological data.²⁶ On this basis, it is conceivable that what is defined as ‘a novel multisystem inflammatory disease’ may be yet another condition caused by the COVID-19-induced cytokine storm. The existing evidence, although limited, that unlike adults children with COVID-19 and a good prognosis display normal ferritin levels, underscores the importance to measure this molecule in those with severe COVID-19 and the above-mentioned multisystem inflammatory disease.^27–29 Furthermore, comparing the vascular involvement in the young and lung involvement in the adult/elderly will help not only to characterise the pathogenic mechanisms behind SARS-CoV2 infection but also to gain understanding towards differential responses across the lifespan. The latter may be useful to define personalised medicine algorithms or better adjust the existing ones.

Therapeutic perspectives and concluding remarks

Most of the aspects of the hyperinflammation during COVID-19 remain therefore to be elucidated, and it is not known who will develop the cytokine storm, when and at which extent. Nonetheless, several immunomodulating agents inhibiting the activity of IFNγ, IL-1β, TNFα, IL-6 and the Janus kinase 1 and 2–signal transducer and activator of transcription pathway are under investigation for the treatment of selected cases of COVID-19. Despite being the mainstay of treatment of hyperinflammation, immunomodulatory agents are double-edged swords in a viral infection. They must be managed with caution, and special attention needs to be considered when referring to non-viral hyperinflammation treatment guidelines. In this regard, the location, the timing and the viral phase deserve some considerations that should guide treatment with these compounds.

First, it must be noted that disease timing is associated with distinct organs involved. Although originated in the lungs, the ‘viral response phase’ leads to a ‘host inflammatory response phase’ that is responsible for the hyperinflammation and thus, cytokine storm.³⁰ The exact overlap between infection phase and organ involvement needs to be more precisely defined, but it is tempting to speculate that underlying immune and inflammatory mechanisms differ across these phases, which will have clear implications for the therapy. It is conceivable that local immune activation in the lungs precedes the general hyperinflammation observed at the systemic level. The multi-organ failure observed in the late stages of COVID-19 severe infection is aligned with this notion. Alternatively, it may be hypothesised that an uncontrolled replication in the lungs due to a poor innate or CD8⁺-mediated viral clearance may facilitate viral spreading via the vasculature to the systemic compartment. The fact that SARS-CoV2 binds the ACE2, which is expressed both in lungs and the vascular endothelium, may support this idea. An in-depth characterisation of the draining lymph node microenvironment should be pursued to better delineate infection kinetics and, hence, disease stages.

Second, during the acute inflammatory response to infective agents, we observe a rapid (within 30 min) increase of TNFα and IL-1β levels and a subsequent rise of IL-6. The high levels of IL-6 last for longer, while TNFα and IL-1β levels rapidly decrease (within 24–48 hours).³¹ Therefore, clinical trials need to clarify the outcome of treating patients with immunosuppressants at different disease stages and with different extent of pulmonary involvement. In this regard, interim results of the French CORIMUNO-TOCI open-label randomised controlled trial have been recently announced.³² A total of 129 patients being hospitalised for COVID-19 moderate or severe pneumonia not requiring intensive care upon admission were enrolled in the study and randomised 1:1 to receive standard of care with or without tocilizumab. A significantly lower proportion of patients needed ventilation (non-invasive or mechanical) or died by day 14 in the tocilizumab group confirming that early intervention may be a major determinant to ensure adequate tackling of the cytokine storm.

Furthermore, treatment with glucocorticoids may be beneficial for patients who develop ARDS, especially in the light of cost, availability, knowledge of the drug and previous use in hyperinflammation. However, they have been linked to exacerbation of COVID-19-associated lung injury.³³ Trials are needed to investigate the optimal dosage and timing of glucocorticoids for COVID-19 treatment.

Finally, as only some of the patients with COVID-19 are (yet) being treated with antiviral agents, clinicians should pay attention to viral load and find a balance between ablating the cytokine storm via the use of immunomodulating agents without significantly affecting the host defence against the virus, thereby preventing uncontrolled replication. In this regard, the ideal scenario would be to combine treatments for the cytokine storm with effective antiviral therapy, thus reinforcing the need for disease stratification in benefit of the treatment decision-making process.

In conclusion, although the cytokine storm we observe in COVID-19 is similar to that observed in other diseases, this scenario is more complex than initially thought due to the heterogeneity of hyperinflammation. In addition, the question remains on whether is it a general phenomenon or rather a reaction of the lung possibly in relation to a massive viral invasion secondary to a lack of antiviral response. Only a deeper understanding of the origin and the different phases of the cytokine storm in relation to the whole clinical picture and disease evolution will allow to draw a definitive conclusion on the nature and natural history of this process. A better characterisation of the host immune response across disease stages, systemic/lung phase transition and viral lung tropism is needed. The acquisition of such notions will allow to confirm whether COVID-19 can be considered a hyperferritinaemic syndrome by itself and to optimise the management of this condition based on disease stage stratification, viral load and cytokine signature to ultimately tailor the therapeutic approach and improve disease prognosis.