key: cord-0694061-d7gw1v9x
authors: Bell, Lucy CK; Noursadeghi, Mahdad; Pollara, Gabriele
title: Transcriptional response modules characterise IL-1 and IL-6 activity in COVID-19
date: 2020-07-22
journal: bioRxiv
DOI: 10.1101/2020.07.22.202275
sha: 5aa78502c830ea904edc5c1a540a3c035f17403f
doc_id: 694061
cord_uid: d7gw1v9x

Dysregulated IL-1 and IL-6 responses have been implicated in the pathogenesis of severe Coronavirus Disease 2019 (COVID-19). Innovative approaches for evaluating the biological activity of these cytokines in vivo are urgently needed to complement clinical trials of therapeutic targeting of IL-1 and IL-6 in COVID-19. We show that the expression of IL-1 or IL-6 inducible transcriptional signatures (modules) reflects the bioactivity of these cytokines in juvenile idiopathic arthritis (JIA) and rheumatoid arthritis, and discerns the effect of therapeutic cytokine blockade in JIA. In COVID-19, elevated expression of IL-1 and IL-6 response modules, but not these cytokines per se, is a feature of disease both in blood and in affected organs. We propose that IL-1 and IL-6 transcriptional response modules can provide a dynamic readout of the activity of these cytokine pathways in vivo, with potential applications for identifying COVID-19 patients who may benefit from IL-1 or IL-6 blocking therapy, and to aid quantification of the biological effects of these treatments.

Severe Coronavirus Disease 2019 (COVID-19) typically occurs over a week from symptom onset, when viral 34 titres have diminished, suggesting an unregulated host inflammatory response may be driving the 35 pathogenesis of severe disease (1-3). Elevated IL-1 and IL-6 responses have each been associated with disease 36 severity (1, (4) (5) (6) (7) (8) . In addition, the hyperinflammatory state in COVID-19 is reported to resemble some aspects 37 of haemophagocytic lymphohistiocytosis (HLH), a condition that may benefit from therapeutic IL-1 blockade 38 (9). These observations have generated hypotheses that IL-1 and / or IL-6 may be key drivers of pathology in 39 severe COVID-19, and led to clinical trials of IL-1 and IL-6 antagonists in this context (10, 11) . 40

The measurement of individual cytokines at the protein or RNA level may not reflect their biological activity 41 their cognate cytokines, and discriminate these from each other and from an alternative inflammatory 101 cytokine stimulus, TNF. 102 103 IL-1 and IL-6 module expression in chronic inflammation 104 To determine whether IL-1 and IL-6 response modules were able to detect elevated cytokine bioactivity in 105 vivo, we assessed the blood transcriptome of juvenile idiopathic arthritis (JIA) and rheumatoid arthritis (RA) 106 patients. These are conditions in which elevated IL-1 and IL-6 activity are considered to play a key role in 107 disease pathogenesis, evidenced by clinical improvement following therapeutic antagonism of these cytokines 108 (22) (23) (24) (25) . The blood transcriptome of untreated JIA patients displayed elevated IL-1 and IL-6 bioactivity (fig 2A) 109 (26), but this was not consistently evident in several RA blood transcriptome datasets (fig S1) (27) (28) (29) . 110 Discrepancies between molecular changes in blood and tissues have been previously described in RA (29) , and 111 therefore we tested the hypothesis that in contrast to blood, elevated IL-1 and IL-6 bioactivity was a feature 112 of the synovium in RA. Consistent with this hypothesis, a separate transcriptomic dataset of synovial 113 membrane biopsies from patients with RA (30) showed elevated levels of both IL-1 and IL-6 response module 114 expression compared to non-RA synovium ( fig 2C) . 115

We used the elevated cytokine activity in the blood of JIA patients to test the hypothesis that therapeutic 116 cytokine modulation would result in changes in cytokine bioactivity as determined by module expression. We 117 made use of the blood transcriptome of JIA patients 3 days following administration of canakinumab, a human 118 monoclonal antibody to . Patients who had a therapeutic response to canakinumab showed elevated 119 IL-1 module expression which reduced 3 days after canakinumab administration ( fig 3A) . In contrast, in those 120 who had no treatment response, IL-1 module expression was lower at baseline and was unaffected by 121 canakinumab (fig 3A) . Unlike the differences seen in the IL-1 module between responders and non-responders, 122 there were no differences between these groups in IL-6 module expression at baseline ( fig 3B) . This indicated 123 that these two cytokine response modules quantified two distinct biological processes. Interestingly, 124 expression of the IL-6 module was also diminished after canakinumab treatment in patients who responded 125 to treatment, suggesting that IL-6 activity may be downstream of IL-1 in this context. Of note in these IL-1 and IL-6 bioactivity in COVID-19 131 We tested the hypothesis that elevated IL-1 and IL-6 bioactivity is a feature of COVID-19 disease, using the 132 peripheral blood transcriptome of 3 patients with mild-moderate COVID-19 disease who were admitted to 133 hospital and recovered (18). This dataset was generated using the Nanostring system and consisted of 594 134 mRNA targets, which included only 7/57 (12.2%) and 7/41 (17.1%) constituent genes of the IL-1 and IL-6 135 response modules respectively. We demonstrated that IL-1 and IL-6 submodules, generated from these 136 shorter lists of constituent genes, were still able to recapitulate all the findings from fig 3 (fig S2) . Assessing 137 the expression of these submodules in the blood transcriptome of COVID-19 patients revealed that IL-1 and 138 IL-6 bioactivity more closely reflected the time-dependent recovery of the patients, compared to the 139 expression of IL1A, IL1B and IL6 genes over time ( fig 4A) . 140

Finally, we determined the expression of IL-1 and IL-6 response modules at the site of COVID-19 disease, using 141 autopsy samples collected from pulmonary and non-pulmonary tissues (31). We tested the hypothesis that 142 cytokine activity would be highest in the lungs, the site of predominant disease. This approach revealed 143 elevated expression of IL-1 and IL-6 response modules in the lungs compared to other organs, using both full 144 length modules and the abbreviated submodules ( fig 4B & fig S3) . In contrast, IL1A, IL1B and IL6 expression 145 did not show such differences, further underlying the increased sensitivity of the transcriptional modules in 146 quantifying cytokine bioactivity in COVID-19 infection ( fig 4B) . The protracted clinical course, inverse relationship between viral load and symptom progression, and the 150 association between inflammation and worse clinical outcomes support a hypothesis whereby severe COVID-151 19 disease is predominantly driven by an exaggerated inflammatory response (1,2). Both IL-1 and IL-6 may 152 play a role in this process (1, (4) (5) (6) (7) (8) , and in this study we utilised transcriptional modules derived from cytokine 153 stimulated cells to demonstrate that their expression, but not that of their cognate cytokine genes, provided 154 a quantitative readout for cytokine bioactivity in vivo, both in the context of COVID-19 and other chronic 155 inflammatory conditions. 156

Studies modulating cytokine activity in COVID-19 have already been initiated, despite inconsistent 157 demonstration of increased activity of these pathways in vivo (10,11). We show that in COVID-19, IL-1 and IL-158 than the expression of the cytokine gene mRNA, the stability of which is subject to trans-regulatory factors 162 and feedback loops (32,33). Moreover, transcriptional modules are intrinsically composed of genes with co-163 correlated expression, minimising technical confounding of single gene measurements, demonstrated by the 164 strongly concordant expression between the full and Nanostring subset IL-1 and IL-6 response modules. 165 Therefore, we propose that transcriptional modules integrate the culmination of cytokine signalling events 166 and provide a functional measure of cytokine bioactivity in vivo. 167

Our findings have several implications. First, they support the rationale for investigating the therapeutic 168 benefit of neutralising IL-1 and IL-6 in COVID-19 (10,11), aiming to reduce cytokine induced pathology at the 169 site of disease. Second, IL-1 and IL-6 response modules may be used to measure cytokine bioactivity following 170 therapeutic immunomodulation with drugs that target these cytokines, such as anakinra, canakinumab and 171 tocilizumab (9-11), permitting correlation between clinical responses and levels of cytokine activity. This may 172

be key, as individual protein level measurements after cytokine blockade in vivo do not necessarily reflect 173 bioactivity, exemplified by the rise in IL-6 cytokine in blood following administration of tocilizumab, a 174 humanised monoclonal antibody against the IL-6 receptor (34). Finally, IL-1 and IL-6 response modules may be 175 sensitive biomarkers to stratify for disease severity in COVID-19. This hypothesis will need to be tested in larger 176 datasets with a greater range of clinical severity, alongside other gene signatures derived from unsupervised 177 analyses of the blood transcriptome in Our study has some clear limitations. Foremost, the sample sizes in the COVID-19 blood dataset utilised were 179 small, and as this cohort did not have severe COVID-19 disease, we were not able to assess how module 180 expression tracks more prolonged clinical syndromes. Thus, our findings will need to be validated in larger 181 datasets with a wider range of COVID-19 severity when these become available in the public domain. Equally, 182 assessing neutralisation of IL-1 and IL-6 with biologic agents was limited by the paucity of available datasets 183 that assessed the impact of these drugs in the days, not months, following administration, analogous to the 184 analyses planned in COVID-19 (10). Finally, determining the specificity of the IL-1 and IL-6 response modules 185 was limited to the available datasets and the range of cytokine stimulation conditions performed in those 186 experiments. Comparing the expression of these modules across a wider range of biologically paired cytokine 187 stimulations will allow refinement of their sensitivity and specificity. 188

In conclusion, our data support elevated activity of the inflammatory cytokines IL-1 and IL-6 in COVID-19, and 189 demonstrate the power of cytokine transcriptional response modules in providing a dynamic readout of the 190 activity of these pathways in vivo. We propose that use of these modules may enhance efforts to investigate Data and software availability 195 All transcriptional datasets used in this manuscript were derived from public repositories. Their source is 196 detailed in table 1 and software used to analyse the data described in the methods. in vitro with either IL-1β (15 ng/ml) or IL-6 (25 ng/ml) for 4 hours, B) PBMC stimulated with TNF (20 ng/ml) 299 or IL-1β (10 ng/ml) for 6 hours, C) human renal proximal tubular epithelial (HK-2) cells stimulated with IL-6 300 (200 ng/ml) or TNF (100 ng/ml) for 1.5 hours and D) human macrophage cell lines (THP-1) stimulated with 301 IL-6 (50 ng/ml) or TNF (10 ng/ml) for 2 hours. Transcriptomic datasets are designated adjacent to figure  302 panels. * = p < 0.05 by Mann-Whitney test. 303 

Multi-omics monitoring of drug response in rheumatoid arthritis.

Transcriptional Signature Associated with Early Rheumatoid Arthritis and Healthy Individuals at high risk to develop the disease. 

Clinical features of patients infected with 2019 novel 210 coronavirus in Wuhan, China. The Lancet

Predicting infectious SARS-CoV-2 from 212 diagnostic samples

Immune mechanisms of pulmonary 214 intravascular coagulopathy in COVID-19 pneumonia. The Lancet Rheumatology

Dysregulation of immune response in patients with 216 COVID-19 in Wuhan, China

Single-cell landscape of bronchoalveolar immune cells 218 in patients with COVID-19

Single-cell longitudinal analysis 220 of SARS-CoV-2 infection in human bronchial epithelial cells