Longitudinal Cytokine Profile in Patients With Mild to Critical COVID-19

The cytokine release syndrome has been proposed as the driver of inflammation in coronavirus disease 2019 (COVID-19). However, studies on longitudinal cytokine profiles in patients across the whole severity spectrum of COVID-19 are lacking. In this prospective observational study on adult COVID-19 patients admitted to two Hong Kong public hospitals, cytokine profiling was performed on blood samples taken during early phase (within 7 days of symptom onset) and late phase (8 to 12 days of symptom onset). The primary objective was to evaluate the difference in early and late cytokine profiles among patient groups with different disease severity. The secondary objective was to assess the associations between cytokines and clinical endpoints in critically ill patients. A total of 40 adult patients (mild = 8, moderate = 15, severe/critical = 17) hospitalized with COVID-19 were included in this study. We found 22 cytokines which were correlated with disease severity, as proinflammatory Th1-related cytokines (interleukin (IL)-18, interferon-induced protein-10 (IP-10), monokine-induced by gamma interferon (MIG), and IL-10) and ARDS-associated cytokines (IL-6, monocyte chemoattractant protein-1 (MCP-1), interleukin-1 receptor antagonist (IL-1RA), and IL-8) were progressively elevated with increasing disease severity. Furthermore, 11 cytokines were consistently different in both early and late phases, including seven (growth-regulated oncogene-alpha (GRO-α), IL-1RA, IL-6, IL-8, IL-10, IP-10, and MIG) that increased and four (FGF-2, IL-5, macrophage-derived chemokine (MDC), and MIP-1α) that decreased from mild to severe/critical patients. IL-8, followed by IP-10 and MDC were the best performing early biomarkers to predict disease severity. Among critically ill patients, MCP-1 predicted the duration of mechanical ventilation, highest norepinephrine dose administered, and length of intensive care stay.


INTRODUCTION
Coronavirus disease 2019  is caused by the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) virus (1). Most patients infected with SARS-CoV-2 remain asymptomatic or only develop mild respiratory symptoms, but 5% develop critical illnesses (2,3). Age and comorbidities are important risk factors for mortality. However, the underlying reasons for why patients manifest a spectrum of disease severity despite infection with the same virus are unclear. The natural history of COVID-19 follows a distinct pattern starting with early mild respiratory illnesses with or without systemic symptoms shortly after infection. After 1 week from symptom onset, a small proportion of patients develop respiratory failure from pneumonia which may be complicated by multiorgan dysfunction with acute respiratory distress syndrome (ARDS), shock and renal failure (1).
The cytokine release syndrome has been proposed as the driver of inflammation that is thought to be central to the pathogenesis of severe COVID-19 (4)(5)(6). Systemic inflammatory cytokines such as interleukin-6 (IL-6) and tumor necrosis factor-alpha (TNF-a) and specific Th1 cytokines like interferon-induced protein-10 (IP-10) are associated with COVID-19 severity and mortality (4,7,8). Moreover, specific cytokines such as IL-6, interleukin-8 (IL-8), and interleukin-10 (IL-10) are higher in patients with COVID-19related ARDS (9). The host response and cytokine profile in COVID-19 is distinct from that of other b-coronaviruses and influenza A viruses (10). In particular, reduced type I interferon levels are associated with severe COVID-19, suggesting disease severity may be due to impaired viral clearance and uncontrolled viral replication (11)(12)(13).
Studies on longitudinal cytokine profiles in patients across the whole spectrum of COVID-19 severity are lacking (14)(15)(16)(17). Furthermore, correlation between cytokine and specific clinical end points such as severity of organ dysfunctions are underexplored. Clinical data supporting the use of specific biologics in COVID-19 to prevent severe disease and improve survival, although encouraging, are currently limited (18,19). Appropriate timing of immunotherapy may be important to optimize efficacy (20). Comprehensive understanding of temporal changes in cytokine profile in COVID-19 is needed to prioritize potential therapeutics to prevent and treat severe COVID-19. The primary objective of this study was to evaluate the difference in early and late cytokine profiles in patients with mild, moderate, and severe/critical COVID-19. The secondary objective was to measure the associations between cytokines and length of intensive care unit (ICU) stay, duration of mechanical ventilation, highest vasopressor dose, and worst P a O 2 /F i O 2 (PF) in patients with severe/critical COVID-19.

Study Design
This was a prospective observational study in COVID-19 patients admitted to two public hospitals in Hong Kong.
Patients were recruited at hospital admission and blood samples were taken for cytokine measurement during the "early phase" within 7 days of symptom onset and the "late phase" between 8 and 12 days from symptom onset. Clinical data on demographics, symptoms, and treatment outcomes were collected prospectively to correlate with cytokine profiles. This study was approved by the Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee (2020.076).

Severity of COVID-19
The severity of COVID-19 was classified as mild, moderate, severe, and critically ill as previously described (21). Mild cases were defined as light clinical symptoms only without signs of pneumonia on imaging. Moderate severity was defined as those with imaging evidence of pneumonia. Severe cases included any patient with respiratory distress, respiratory rate ≥30/min or oxygen saturation ≤93% in room air or PF ratio ≤300 mmHg. Critical severity was defined as patients who needed mechanical ventilation, developed shock, or had other organ failures requiring admission to critical care.

Viral Load
Viral load of each patient was determined by taking multiple upper respiratory tract specimens including nasopharyngeal swabs and deep throat saliva samples during hospitalization for real-time PCR as described (22). The peak viral load was defined as the one with the lowest cycle threshold (Ct).

Statistics
Continuous data were described with median and interquartile range (IQR) while categorical variables were presented as proportions. Kruskal-Wallis and Chi-squared tests with Bonferroni correction were used to assess the differences of continuous and categorical clinical variables across multiple severity groups, respectively. Jonckheere-Terpstra (JT) trend analysis using the R SAGx package was used to measure whether cytokines from early and later phases changed progressively as severity of disease increased (mild !moderate !critical). Univariable and multivariable regression analyses using a generalized linear model (glm) in the R Stats package was used with age as a confounding variable to perform pairwise comparisons of cytokines between critical and moderate, between critical and mild, or between moderate and mild groups. Wilcoxon signed rank (for matched samples) test was performed to compare the difference in cytokine levels between early and late phases within each severity group. Spearman's rank correlation was calculated to assess the associations between each cytokine and other cytokines and between cytokines and lowest PF ratio, highest norepinephrine dose, days on mechanical ventilation, or ICU length of stay. Receiver operating characteristic (ROC) curve analysis using the R pROC package was applied to assess the potential of early and late cytokine profiling as a biomarker of COVID-19 severity to discriminate severe/critical patients from the noncritical (mild and moderate) group. Age as a confounding factor was controlled using a glm algorithm to calculate the area under the ROC (AUC) value. In-house developed scripts and ggplot2 in R v3.6.2 were used for statistical analysis and visualization.

Patient Characteristics
A total of 40 adult patients (mild = 8, moderate = 15, severe/ critical = 17) hospitalized with COVID-19 were included in this study. Their baseline characteristics, treatment, and outcomes are shown in Table 1 and Supplementary Data Sheet 1. Patients with critical COVID-19 were older than patients with moderate (Mann-Whitney U test, p = 0.023) or mild disease (Mann-Whitney U test, p ≤ 0.001) (Supplementary Figure 1A). Patients with moderate disease had a much lower proportion of males (4/15 26.7%) compared with the critical and mild groups (Chi-squared test, p = 0.032).

Viral Load
The median peak viral load of critical patients was 22.   cytokines (IL-1b and IL-9) were lower ( Figure 2B) only during the early phase among patients with more severe disease. Whereas, five cytokines (eotaxin, IFN-g, IL-1a, IL-12p40, and IL-12p70) were lower ( Figure 2C) and two cytokines (EGF and IL-18) were higher ( Figure 2D) only during the late phase among patients with more severe disease. Interestingly, we also observed significant changes in the levels of several cytokines (IFN-a2, IL-1RA, IL-12p40, and MDC) between early and late phases in severe/critical patients ( Figure 3). Of the 30 patients who had early phase samples, eight were given interferon beta-1b (IFN b-1b) and three were given steroids prior to blood sampling. Out of 40 patients, 17 patients received IFN b-1b and 14 patients were treated with steroids prior to late phase sampling. There was no difference in proportion of patients given IFN b-1b across the three severity groups at both time points (p = 0.399 and p = 0.407 for early and late phases, respectively). However, at both time points, more patients in the critical group were given steroids prior to sampling (p = 0.048 and p < 0.001 for early and late phases, respectively).

Potential of Cytokines in Discriminating Severity Patients
All 11 cytokines that achieved statistical significance across severity groups during JT trend analysis in both early and later phases (GRO-a, IL-1RA, IL-6, IL-8, IL-10, IP-10, MIG, FGF-2, IL-5, MDC, and MIP-1a) had satisfactory AUC values in both phases for discriminating severe/critical from mild/moderate infections when age was adjusted as a significant confounding factor ( Figure 4A and Supplementary Figure 2). During the early phase, IL-8 was the best performing biomarker in terms of sensitivity and specificity for severe/critical outcome, followed by IP-10 and MDC ( Figure 4A). Whereas, in the late phase, MDC became the best performing biomarker, followed by IP-10, IL-10, GRO-a, and IL-6. Correlations between individual cytokines at A B FIGURE 1 | Cytokines with same trends with increasing disease severity in both early and late phases. Cytokine levels (pg/ml) progressively increased (A) or decreased (B) in levels in severe/critical COVID-19 compared with moderate and mild patients in both early and late phases along the sequence of mild !moderate !critical using a Jonckheere-Terpstra (JT) trend analysis (p-values shown in grey box). Individual comparisons between groups are tested using a glm algorithm by controlling age as a confounding factor, with statistical significance shown as not significant (ns); *p ≤ 0.05; **p ≤ 0.01. GRO-a, growth-regulated oncogene-alpha; IL-1RA, interleukin-1 receptor antagonist; IL-5, interleukin-5; IL-6, interleukin-6; IL-8, interleukin-8; IL-10, interleukin-10; IP-10, interferon-induced protein-10; MIG, monokine induced by gamma interferon; FGF-2, fibroblast growth factor-2; MDC, macrophage-derived chemokine; MIP-1a, macrophage inflammatory protein-1alpha.
the two phases are shown in Figure 4B. For instance, IP-10 and IL-6 exhibited a strong positive correlation, whereas IP-10 and MDC exhibited a significant negative association ( Figure 4C).

Association With Clinical Endpoints in Critical Patients
At the time of ICU admission (median days from onset 9, IQR 8-10), the association of cytokines with clinical endpoints were analyzed for severe/clinical patients ( Figure 5).
The proinflammatory Th1 helper (IL-18, IP-10, MIG, and IL-10) and ARDS-associated cytokines (IL-6, MCP-1, IL-1RA, and IL-8) were increased progressively in patients with increasing severity of COVID-19. After adjusting for age, IL-8, IP-10, and MDC levels were useful early (within 7 days of illness onset) biomarkers to predict disease severity; whereas, MDC, IP-10, IL-10, GRO-a, and IL-6 also carried good performance at the late phase (between 8 and 12 days after illness onset). MCP-1 level at ICU admission predicted the days on mechanical ventilation, highest dose of vasopressor required, and length of ICU stay. Similar to SARS-CoV-1 in 2003, severe/critical COVID-19 was associated with higher levels of Th1 cytokines such as IL-18, IP-10, and MIG (23). IL-18 increases rat lung vascular permeability, neutrophil infiltration, and other cytokines (24). It also enhances IL-12-induced IFN-g production (25,26). We found that levels of IL-18 increased in the late phase coinciding with deterioration requiring ICU admission. Surprisingly, although IL-12 and IFN-g levels were initially elevated in all COVID-19 patients, no correlation with disease severity was observed. Furthermore, levels of IL-12 and IFN-g normalized around time of ICU admission, and as previously reported, were relatively lower in patients with severe/critical COVID-19 (13). Viral load also did not differ between patients with mild/moderate or severe/critical COVID-19. Overall, our observations suggest that the deterioration typically occurs around days 8-12 from symptom onset is not mediated by uncontrolled viral burden.
Elevated IP-10 levels are associated with COVID-19 severity and mortality (1,7,(27)(28)(29)(30). While previous studies focused on late sampling of IP-10, we showed that IP-10 was an excellent early biomarker to predict subsequent disease severity. Bronchial epithelium secretes IP-10 under IFN-g stimulation (31). Since the role of IP-10 is to attract effector T cells to sites of Th1 inflammation, it may be an important target in SARS-CoV-2induced lung injury. Encouragingly, specific blockade of IP-10 has been shown to reduce ARDS in rat sepsis models (32). Corticosteroids have also been shown to reduce IP-10 levels in preclinical studies and in vivo in patients with SARS-CoV-1 infection (23,33). Similarly, elevated MIG levels in SARS-CoV-1 were attenuated by corticosteroid administration (23). Taken together, these findings may explain why corticosteroids improve survival in COVID-19 patients who require oxygen (34).
In contrast, Th2 cytokines (IL-5 and MDC) and allergic inflammation-related cytokines (IL-5 and eotaxin) were reduced in patients with severe/critical COVID-19. We found that MDC was the best performing biomarker at late phase (8-12 days after illness onset) to predict severity. Of note, we found an anti-inflammatory cytokine, IL-10, consistently increased with disease severity both in the early and late phases. IL-10 is secreted by regulatory T cells and type 2 innate lymphoid cells (35,36). It blocks the synthesis of other cytokines such as IL-12 and IL-18 and provides negative feedback on proliferation and differentiation of Th1 cells (37,38). More recently, however, it has been shown that in sepsis, IL-10 may stimulate and oppose IFN-g and TNF-a production in mononuclear cells and T cells, respectively (39). The pathological role of IL-10 in COVID-19, as well as being a potential therapeutic target, deserves further investigations.
Th1 (IP-10 and IL-10) and ARDS cytokines (IL-6 and IL-8) measured at early and late phases were predictive of disease severity. While these cytokines could be useful biomarkers to stratify the risk of patients, whether these changes reflect the consequence or cause of disease severity is unclear. Nevertheless, tofacitinib, a Janus kinase inhibitor which suppresses Th1 response and IL-6 production has recently been shown to decrease mortality in patients hospitalized with COVID-19 even when the majority of patients were already given corticosteroids (19,58). This adds further evidence that selective immunomodulation is an important therapeutic approach in COVID-19.
Our study has several limitations. The sample size was small which limits the power to detect difference between the severity groups. However, we were able to uncover cytokines which were consistently different in the same direction from mild, moderate to severe/critical COVID-19. We were unable to analyze cytokine profiles against hospital length of stay since many patients in our cohort were admitted to hospital for isolation rather than severity of illness. Furthermore, we could not assess relationship between cytokine profile and mortality since the mortality rate was low in our cohort. Analysis on effect of antiviral treatment and immunomodulating agents was not feasible due to small sample size and changes in treatment protocols over the recruitment period. However, we were able to adjust for age as confounding factor in pairwise comparisons and severity predictive performance analysis. Lastly, measurement of plasma cytokine is only a surrogate for cytokine levels in the lung which may not be representative of the pulmonary inflammatory profile.
In conclusion, cytokine profile varied across different severity of COVID-19 over time. Th1 response and ARDS-associated cytokines were elevated in patients with increasing severity of COVID-19. IL-8, IP-10 and MDC were the best performing early biomarkers to predict severity. MCP-1 level at ICU admission was related to days on mechanical ventilation, highest dose of vasopressor, and length of ICU stay.

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

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee (2020.076). The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
LL, ZC, CW, and PC designed the study. LL, GL, ET, VC, KF, and WW recruited patients and collected clinical data. RN and AY processed cytokine samples. ZC performed the data analysis. LL, ZC, CW, and PC interpreted the results before LL drafted the first draft of the manuscript. All authors including DH provided feedback to the final version of the manuscript. All authors contributed to the article and approved the submitted version.

FUNDING
This study was funded by a grant from the Health and Medical Research Fund (COVID190107).

ACKNOWLEDGMENTS
We thank the healthcare workers who cared for COVID-19 patients at our institutions and Ms. Patricia Leung for helping with clinical data collection.

SUPPLEMENTARY MATERIAL
The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fimmu.2021.763292/ full#supplementary-material Supplementary Data Sheet 2 | Comparison of cytokine levels in this study across severity groups and in both early and late phases.
Supplementary Data Sheet 3 | Spearman's correlation analysis between paired cytokines in early and late phases.