SARS-CoV-2 causes a broad range of clinical outcomes. What determines whether an infected individual will progress to severe COVID-19 is a complex interaction between host and viral factors. Since the beginning of the COVID-19 pandemic, there has been significant interest in developing robust biomarkers to predict disease outcomes amongst COVID-19 patients. This facilitates patient triage and resource prioritisation, both of which become particularly important when healthcare centres are required to simultaneously treat many COVID-19 patients. Numerous prognostic indicators have been proposed for COVID-19. These include, but are not limited to, patient demographics, co-morbidities, lung computed tomography (CT) results, coagulation assays, alterations in white blood cell counts and inflammatory response biomarkers such as C-reactive protein and cytokines (1). Others have used machine learning to predict disease outcomes (2, 3). Despite these investigations, there is no consensus on the best approach for predicting COVID-19 outcomes. Moreover, as there is no indication as to whether such predictive tools would apply to other potentially lethal viruses, and therefore could become part of stockpile for future pandemic preparedness.

We have previously discovered that the transcription of the interferon alpha-inducible protein 27 (IFI27) is a signature marker of pandemic H1N1/09 influenza infection (4). IFI27 (also known as ISG12 or p27) is an interferon alpha (and to a lesser extent interferon gamma) inducible gene of unknown function, with the gene product residing in the nuclear membrane of the cell (5). IFI27 expression is associated with the outcomes of several different viral illnesses including respiratory syncytial virus (RSV) infection and Enterovirus 71 (EV71) hand foot and mouth disease (6, 7). Preliminary evidence suggests that IFI27 expression may be a useful biomarker for COVID-19 diagnosis (8–11). In the respiratory tract, IFI27 expression is significantly upregulated by SARS-CoV-2 infection, more so than by other respiratory viruses (8, 9, 12). IFI27 is also upregulated in the peripheral blood of COVID-19 patients (10, 11, 13) and may serve as a biomarker for pre-symptomatic SARS-CoV-2 infection (14). These data are consistent with observations that the induction of type I interferons in the lower respiratory tract and systemically is associated with immunopathology (15) whilst the induction of type I interferons in the upper respiratory tract is associated with virus control and not disease outcomes (12). However, others have suggested that IFI27 expression is low in patients with severe COVID (16, 17). Therefore, the usefulness of IFI27 expression as a prognostic biomarker for COVID-19 has yet to be determined.

Here, we use multiple patient cohorts to evaluate the role of IFI27 expression in predicting COVID-19 disease progression. We further examine the role of IFI27 expression for risk stratification in infections caused by other respiratory viruses, namely influenza virus, another virus of pandemic potential.

The findings of this study included expression data (gene/protein) generated across eight cohorts across six countries (Australia, U.S.A., Chile, Brazil, Iran, and Singapore). There was one case-control study (Cohort 1), two prospective studies (Cohort 7 and Cohort 8) and five cross-sectional studies (Cohorts 2-6). Patients included in the studies had different disease severities (mild, moderate, and severe). Cohort 8 (Iran) included COVID-19 patients recruited in 2021, who were infected during the Delta wave of the pandemic in Iran. Other COVID-19 cohorts recruited patients in early 2020 (Cohorts 1-7). Several platforms were used to measure IFI27 gene expression including PCR, microarray, and spatial transcriptomics. Table 1 provides full details of the IFI27 measurement methods and tissue sampling approaches in each cohort.

We first investigated whether IFI27 expression could be detected in COVID-19 patients by assessing gene expression in the lower respiratory tract of deceased COVID-19 patients (Cohort 1; Brazil; n=4). The virus load in the lung was assessed by RNAscope^®, which had the sensitivity to detect single molecules in a cell. The distribution of virus load was quantified by STRISH, a robust image processing pipeline (22). The average measurements of SARS-CoV-2 spike mRNA (nCoV2019) per grid (tissue region) of neighbouring cells were visualized by STRISH using a tissue heatmap ( Figure 1A ). Adjacent tissue sections from the same samples were then processed by the Visium^® spatial transcriptomic method using polyA-capture ( Figure 1B ). Based on comparison of RNAscope and Visium analyses, we discovered that areas of high viral load in the lung, as detected by RNAscope, also had high levels of IFI27 gene expression ( Figures 1C, D ). The correlation between high viral load and increased IFI27 gene expression was independently replicated by an additional Visium^® spatial dataset, here generated using a probe hybridization protocol (data not shown). Together, these data indicate that IFI27 expression can be detected in the lower respiratory tract during SARS-CoV-2 infection. This finding was validated by several recently published studies (8, 23, 24), which confirmed the upregulation of IFI27 expression in infected lung tissue of COVID-19 patients ( Figure 1E ).

The above data indicated that IFI27 gene expression may reflect local disease activity in infected lung tissue. However, lower respiratory tract is not readily accessible or available for prognostic testing. Accordingly, we sought to establish if IFI27 expression could be detected in the upper respiratory tract during SARS-CoV-2 infection ( Figure 2 ). To model the human upper respiratory tract primary nasal epithelial cells from healthy donors were differentiated at an air liquid interface, as described previously (25). Cells were infected with SARS-CoV-2 (QLD02) and IFI27 expression was assessed 72 hours post-infection. At 72 hours post-infection IFI27 expression was upregulated in infected nasal epithelial cells ( Figure 2A ). However, the difference in relative expression between infected and uninfected cells (i.e., the dynamic range of expression) was only moderate between infected and uninfected cells (a feature which is suboptimal for a prognostic biomarker). Nevertheless, we further assessed the prognostic potential of IFI27 expression in nasopharyngeal swabs of COVID-19 patients (Cohort 2; Chile). Here, no significant association was observed between nasopharyngeal IFI27 gene expression and virus RNA ( Figure 2B ). Further analysis in another cohort (Cohort 3; USA) showed a similar finding; there was no significant difference in IFI27 gene expression between mild, moderate, and severe disease ( Figure 2C ). Since these findings suggested that upper airway IFI27 expression levels do not have a strong prognostic value (and a limited dynamic range) we sought an alternative sampling route to measure IFI27 expression in COVID-19 patients.

The spread of SARS-CoV-2 infection to the lower airway is often associated with a systemic host response, and blood provides an easily accessible clinical sample to assess this response. Thus, we proceeded to measure IFI27 expression levels in peripheral blood of COVID-19 patients (Cohort 6; Iran). We found that blood IFI27 was higher in infected patients (compared to asymptomatic or uninfected individuals). Furthermore, there was a trend of increasing IFI27 gene expression in patients with a worsening disease, although this trend was not statistically significant ( Supplementary Figure 1A ). When protein expression of IFI27 was measured, there were no elevation in COVID-19 patients (Cohorts 4 and 5; Chile & Brazil) and it did not correlate with disease severity ( Supplementary Figure 1B ). These findings suggested that blood IFI27 gene expression (not protein expression) could act as a surrogate marker of disease outcome in COVID-19.

We then used an independent prospective cohort (Cohort 7; Australia & Singapore) to test the hypothesis that changes in blood IFI27 gene expression could predict outcomes in COVID-19 patients. In Cohort 7, blood sampling (for IFI27 gene expression) was performed upon initial presentation of each patient when the disease outcome was still unknown. Each patient was then followed up for 28 days and their clinical outcomes (e.g., acute respiratory distress syndrome) were recorded (see Methods). We found once again that blood IFI27 gene expression was significantly higher in patients who tested positive for SARS-CoV-2, a necessary pre-requisite for a prognostic biomarker. ( Figure 3A ) We also noted that IFI27 blood expression was high in COVID-19 patients who developed an adverse outcome ( Figure 3B ). It is important to emphasise that all patients in this cohort (Cohort 7) were prospectively recruited at the early phase of their infection, which was prior to their outcomes become known (e.g., discharge home or admission to ICU). The patient outcomes were ascertained later in the follow-up period. Therefore, Figure 3C demonstrates the predictive capability of blood IFI27 in the early phase of the infection when the patient’s subsequent trajectory was unknown to the clinician. Since all patients were recruited under the same protocol (i.e., prospective recruitment), the findings in Figure 3C reflects the forward performance of IFI27 of all patients in this cohort. In case studies of a more limited number of patients IFI27 gene expression increase could precede – by several days – clinical signs of deteriorations or abnormal changes in laboratory parameters such as high levels of C-reactive protein ( Supplementary Figure 2 ).

This observation was further confirmed by area under the curve of receiver-operator-characteristics curve (AUROC) analysis, which showed the IFI27 gene expression outperformed laboratory variables (e.g., C-reactive proteins), patient variables (e.g., age, comorbidity) and physiological parameters (e.g., respiratory rate) in predicting COVID-19 outcomes ( Table 2 and Figure 3D ). Notably, the IFI27 gene expression had a high sensitivity (0.95), high specificity (0.83) and an AUROC of 0.90, all of which were higher than known factors associated with COVID-19 outcomes (e.g., age, co-morbidity). To confirm the prognostic value of IFI27 expression in the blood of COVID-19 patients, a second prospective cohort was recruited (Cohort 8; Iran). Although this cohort was recruited during the peak of the Delta wave in Iran, the same trend was observed: increased IFI27 expression was observed in patients with more severe COVID-19 ( Figure 3E ).

Ideally, for pandemic preparedness, a prognostic biomarker would function for many viral infections and not be restricted to COVID-19 alone. To investigate this further we analysed blood gene expression data from a previously published microarray study of influenza patients (20). This analysis confirmed that IFI27 upregulation occurred in severe influenza infection ( Figure 4A ). Importantly, the IFI27 gene family- including interferon alpha-inducible protein 27-like protein 1 (IFI27L1), and not and interferon alpha-inducible protein 27-like protein 2 (IFI27L2) ( Figures 4B, C ) were best suited to discriminating disease outcomes. Together, these expanded analyses demonstrated that IFI27, or its gene family, could be a prognostic biomarker of the host response to respiratory viruses.

Globally, more than 500K new SARS-CoV-2 infections are recorded every day and the emergence of novel viral variants has raised concerns that these numbers will increase, despite the increasing availability of vaccines. The current global situation emphasises the ongoing need for COVID-19 prognostic biomarkers to facilitate both patient triage and resource prioritisation. Here, we have provided the first evidence of blood IFI27 expression as a potential biomarker for risk stratification of COVID-19 patients. When prospectively validated, blood IFI27 expression showed a high positive and negative predictive value, outperforming other known predictors of COVID-19 outcomes reported in the literature.

IFI27 is an interferon inducible gene. Accordingly, the strong association observed between IFI27 upregulation and severe COVID-19 could be indicative of an increased viral replication. However, this hypothesis would be inconsistent with the poor correlation observed between nasopharyngeal IFI27 expression and viral replication in the nose. Instead, we propose that IFI27 expression reflects increased immunopathology (either local or systemic) and could predict COVID-19 outcome. The correlation of interferon expression and the severity of COVID-19 shows significant anatomical variation. In the upper respiratory tract, rapid induction of type I interferons is typically associated with reduced COVID-19 severity, as their induction is associated with the ability to control viral replication with limited immunopathology (26, 27). In contrast, the relationship between type I interferons and disease pathology is more complex. Certainly, an intact interferon response is important to reduce the severity of SARS-CoV-2 infection. This has been clearly shown by the fact that patients with auto-antibodies to type I IFN (28) or inborn errors in type I IFN immunity (29) experience more severe COVID-19. However, at the same time it is known SARS-CoV-2–infected IFNAR1 KO mice, lacking IFN-α/β signalling, have reduced inflammation and lymphocyte activation, which would support the notion that IFN-α/β can drive pathogenic inflammation in COVID-19 (15). Similarly, patients with severe COVID-19 maintain an elevated interferon response throughout the course of infection, whilst those with moderate disease display a progressive reduction in type I IFN responses over time (30). These seemingly disparate results can be explained by the fact that interferons can play both a protective and pathogenic role during SARS-CoV-2 infection.

What determines whether the interferon response is pathogenic or protective is largely dependent on the magnitude of the response, the kinetics of the response (i.e., if the period of interferon production overlaps with the period of viral replication) and the localisation of the response (i.e., interferon responses in the nose tend to be associated with a more protective response whilst interferon responses in the blood tend to be associated with a more pathogenic response). Importantly, we observed a stronger association between COVID-19 outcome and the IFI27 gene expression than that observed with the IFI27 protein expression in blood. This most likely reflects an increased dynamic range for measuring gene expression and thereby an increased ability to differentiate patient outcomes. Future attempts to translate these findings into routine clinical settings should therefore focus on PCR-based assays to measure IFI27 expression in blood.

Compared to other biomarkers reported in COVID-19 literature, IFI27 offers several clinical advantages. Firstly, IFI27 expression appears to be specific to viral illness. In contrast, most infection/inflammatory biomarkers (e.g., C-reactive protein, leukocytes, interlukin-6) are elevated in many non-viral illnesses (e.g., trauma, sepsis). Secondly, IFI27 expression is directly linked to intra-cellular recognition of respiratory viruses (15, 26, 27, 30). Given the control point of IFI27 expression lies in the disease causal pathway, it makes sense to track disease progression by using IFI27 expression, rather than via biomarkers that are unrelated to the underlying disease activity. Finally, blood IFI27 gene expression has a strikingly high dynamic range (e.g., up to thousands of fold changes in severe COVID-19, as shown in the present study) and a strong signal-to-noise ratio (which produced consistent findings across different measurement platforms, such as microarray or PCR, also evidenced in our findings). These favourable measurement characteristics makes IFI27 a preferred prognostic tool to other biomarkers.

The kinetics of IFI27 expression is poorly understood. As a result, the ideal sampling window was not well-defined in this study. It is also uncertain whether serial measurement of IFI27 expression would be more informative than a single time point measurement. Furthermore, the predictive performance of IFI27 could be confounded by other yet-to-be-defined variables, such as timing (e.g., early presentation versus late presentation), stages of disease (e.g., lung only infection versus multi-organ disease) and prior COVID-19 vaccination (only measured in Cohort 8 in the present study). Indeed, some studies have suggested IFI27 expression is lowest in the critical care or ICU COVID-19 patients (16, 31) which is as apparently incongruent with the data presented herein. These differences may reflect the fact some critical cases have known defects in their interferon response (28) or, age differences between cohorts (noting that the mean age for cohort 7 was relatively low). Alternatively, we noted that the two critical patients in the study of Huang et al. were first sampled approximately 14 days after diagnosis (which was markedly later than those in the present study) and these differences may reflect kinetic differences in IFI27 expression. Therefore, we are unable to extrapolate the prognostic value of blood IFI27 expression to a broader clinical context and additional studies are required. Such studies should consist of large-scale, prospective studies with sample sizes adequately powered to allow researchers to assess the independent confounding effect of each variable (e.g., timing, disease stages and prevalence) on IFI27 expression.

The COVID-19 pandemic has emphasised the need to have stockpiles of broad-spectrum diagnostic and therapeutic tools that can be rapidly deployed at the start of any future viral outbreak. As a component of the anti-viral interferon response, it is not surprising that elevated IFI27 expression was observed in the blood of patients infected with a broad range of different respiratory viruses but not in the blood of patients with a bacterial respiratory infection. These data, combined with previous suggestions that IFI27 expression is associated with the severity of RSV (6, 7), suggest that IFI27 may represent a pan-viral prognostic biomarker. Interestingly, when we investigated this further in the context of influenza virus, we found that expression of IFI27L1 in the blood, rather than IFI27 itself, was associated with disease severity. We therefore propose that future studies focus on developing PCR-based assays to measure a suite of genes associated with IFI27 expression including IFI27L1, IFI27L2 and IFI6. It is hoped that this broader, combinatorial approach, would lead to the development of pan-viral prognostic biomarker that could be incorporated into future pandemic preparedness planning.

The basis for ancestral differences in IFN expression is not well-explained. A study shows that in SLE patients a cluster of highly expressed ISGs including IFI27 is more associated with African ancestry than disease activity (32). Our study which combines cohorts from ethnically diverse populations (Middle Eastern, Asian, European & South American) shows robust nature of IFI27 as a prognostic biomarker.

The present study has several limitations. Firstly, the prospective validation cohorts were relatively limited in size. Secondly, most patient cohorts (Cohort 1-7) were recruited prior to the availability of vaccination. One cohort (Cohort 8) did include vaccinated individuals recruited during the Delta wave of COVID-19 in Iran. Accordingly, our findings do not necessarily apply to vaccinated individuals experiencing ‘breakthrough’ Omicron (or subsequent variant) infections. Finally, we could not correlate the changes in IFI27 expression levels between the peripheral blood and the infected lung. Future studies (e.g., in animal models) are needed to unravel the coupled dynamics of IFI27 expression between these tissue compartments.

The findings provided herein represent the first evidence that IFI27 expression has a potential as a blood biomarker for risk stratification in COVID-19 patients.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding authors.

The study was approved under the Human Research Ethics Committee at Western Sydney Local Health District (HREC Reference: 2020/ETH00886 (6439), Research Governance at Westmead Institute for Medical Research, and Nepean Blue Mountain Local Health District (HREC Reference: 2019/ETH01485. Autopsy and biopsy materials were obtained from the Pontificia Universidade Catolica do Parana PUCPR the National Commission for Research Ethics (CONEP) under ethics committee approval reference number 2020001792/30188020.7.1001.0020 and approval reference number 2020001934/30822820.8.000.0020. University of Queensland Human Research Ethics Committee (HREC) ratification. Human Research Ethics Committees of the University of Concepcion (Chile) (CEBB 676-2020) and ratified by the University of Queensland (2021/HE000319). Human Research Ethics Committee of the Pontifícia Universidade Católica do Paraná (PUCPR) (Number CAAE: 30833820.8.0000.0020) and by Comissão Nacional de Ética em Pesquisa (CONEP). Mazandaran University of Medical Sciences (approval number IR.MAZUMS.REC 1399.856). National Healthcare Group Domain Specific Review Board (DSRB 2014/00614). The University of Queensland’s Human Research Ethics Committee (2020001742), the Queensland Children’s Hospital and Health Service Human Research Ethics Committee (HREC/16/QRCH/215, HREC/10/QRCH/78) and the Queensland University of Technology Human Research Ethics Committee 17000000039). All studies with SARS-CoV-2 were performed under physical containment 3 (PC3) conditions and were approved by the University of Queensland Biosafety Committee (IBC/374B/SCMB/2020). Written informed consent was obtained from all study participants. All methods were performed in accordance with institutional guidelines and regulations.

Conceptualisation: AK, MS, KS, BT, AM. Investigation: AS, JM, LC, MT, CT, SMT, GR, TM, MN, MR, AR, RoG, MG, FR, PS, KS, KC, YZ,YX, TW, ACS, CK, CF, AFRS, LN, SM, RV, OA, RaG, EM, RA-N, LL, FZ, EN-L, GL, BK, VH, YW, AP, TC, TK, KK, ST, TP, TK, TJ, JI, KO’B, JF, MW, CG, FS-F-G. Formal analysis: MS, KS, BT, AK, CT, RoG, YW, KO’B, JF, MD, GB, QN. Writing - original draft: all authors. Writing - review & editing: all authors. All authors contributed to the article and approved the submitted version.