Role of biomarkers in predicting disease severity in acute dengue and SARs-CoV-2-Infected patients

Purpose

The early stages of both dengue infection and COVID-19 can present similarly with acute febrile illness or influenza-like symptoms, and individuals with initially mild disease may progress to more severe symptoms. We performed biomarker analysis to determine if host immune responses can predict the disease severity of both diseases.

Methods

Differential immune response profiles in patient populations were compared during acute dengue or COVID-19 using a panel of 22 soluble biomarkers. Patient plasma biomarkers were measured by ELISA or the Meso Scale Discovery platform, and statistical analysis was performed using SAS software. Receiver operating characteristic (ROC) curves were created to identify the optimal cut-off values for differentially upregulated biomarkers in severe cases. Multiple logistic regression models were developed to predict disease severity using a combination of selected biomarkers, with or without demographic data, and were analyzed using GraphPad Prism software.

Results

Almost all of the biomarkers were higher in dengue compared to COVID-19 patients. Comparing severe to mild dengue illness, biomarkers related to monocyte activation (IL-1β, IL-12p70, soluble CD14) and Th2 cytokines (IL-4 and IL-13) were significantly elevated. Additionally, 1,3 β-D-glucan, a biomarker related to gut barrier disruption and microbial translocation, was elevated in patients with severe dengue and emerged as a key severity biomarker. In COVID-19 patients, the chemokine CXCL10 (IP-10) was the best predictive biomarker for severity. Moreover, biomarkers related to gut mucosal barrier disruption (lipopolysaccharide-binding protein, soluble CD14, and 1,3 β-D-glucan) and neutrophil extracellular trap (NET) markers were elevated in moderate to severe COVID-19. The multiple logistic regression models predicting severity for both diseases yielded ROC curves with an excellent area under the curve (AUC) greater than 0.95 and demonstrated sensitivity, specificity, positive predictive value, and negative predictive value greater than 90%.

Conclusions

Our analyses indicate that gut barrier disruption and subsequent microbial translocation are common phenomena in severe cases of dengue and COVID-19. Multiple logistic regression models using a combination of specific biomarkers have the potential to improve severity prediction.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-025-12196-4.

Plain Language Abstract

The early stages of both dengue infection and COVID-19 can present similarly with acute febrile illness or influenza-like symptoms, and individuals with initially mild disease may progress to more severe symptoms. Therefore, we analyzed host immune responses to determine whether specific biomarkers could predict disease severity in both infections. Using a panel of 22 soluble biomarkers measured by ELISA or the Meso Scale Discovery platform, we compared differential immune response profiles in acute dengue and COVID-19, generated receiver operating characteristic (ROC) curves to identify optimal severity-associated cut-off values, and constructed multiple logistic regression models—with and without demographic data—to evaluate predictive performance. We found that almost all of the biomarkers were higher in dengue compared to COVID-19 patients. Comparing dengue with warning signs to mild dengue illness, biomarkers related to monocyte activation (IL-1β, IL-12p70, soluble CD14) and Th2 cytokines (IL-4 and IL-13) were significantly elevated. Moreover, 1,3 β-D-glucan, a biomarker related to gut barrier disruption and microbial translocation, was elevated in dengue patients with warning signs and emerged as a key severity biomarker. In COVID-19, the chemokine CXCL10 (IP-10) was the strongest predictor of severity, while biomarkers linked to gut mucosal barrier damage (lipopolysaccharide-binding protein, soluble CD14, and 1,3 β-D-glucan) and neutrophil extracellular traps were elevated in moderate to severe cases. The multiple logistic regression models predicting severity for both diseases yielded ROC curves with an excellent area under the curve (AUC) greater than 0.95 and demonstrated sensitivity, specificity, positive predictive value, and negative predictive value greater than 90%. Overall, our analyses indicate that gut barrier disruption and subsequent microbial translocation are common phenomena in severe cases of dengue and COVID-19. Multiple logistic regression models using a combination of specific biomarkers have the potential to improve severity prediction.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12879-025-12196-4.

Introduction

Host immune response profiles based on protein or transcriptomic biomarkers help improve diagnoses, predict disease severity, and determine response to treatment for several diseases. These biomarkers typically relate to innate and adaptive immunity, coagulation pathways, and cell/tissue injury associated with organ-specific complications [1, 2]. Given that various viral infections exhibit unique cellular tropism and pathogenesis, we hypothesize that there might be differential biomarkers specific to each viral infection that correlate with disease severity.

Acute dengue infection is an arthropod-borne viral disease transmitted by the Aedes mosquito species in tropical and subtropical areas. It is caused by infection with any of the four dengue virus serotypes (DENV-1 to DENV-4) [3]. While the majority of cases are asymptomatic, dengue infection may present as a spectrum, ranging from undifferentiated viral fever to dengue fever (DF) or dengue hemorrhagic fever (DHF), with or without shock. The 2009 WHO guideline classifies dengue patients into those without warning signs, with warning signs, and severe dengue. Patients with severe disease can develop life-threatening complications, including acute liver failure, encephalopathy, and hemophagocytic lymphohistiocytosis [4, 5]. The pathogenesis of severe dengue infection involves several complex mechanisms, including antibody-dependent enhancement (ADE), antibody-dependent cellular cytotoxicity (ADCC), cytokine storm (hypercytokinemia), aberrant activation of the complement system, as well as endothelial dysfunction [6, 7].

Coronavirus disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS‑CoV‑2), emerged in Wuhan, China, in December 2019 and eventually led to the global pandemic. The spectrum of clinical manifestations of COVID-19 varies widely from asymptomatic to influenza-like syndrome, pneumonia, respiratory failure, and death [8]. Cytokine storm also plays a key role in the pathogenesis of severe COVID-19. Numerous studies have shown that COVID-19 patients have elevated levels of inflammatory cytokines and chemokines such as TNF-α, IL-1β, IL-6, IL-10, IL-18, IL-2, IFN-γ, CXCL8 (IL-8) and CXCL10 (IP-10), several of which correlate with disease severity [9, 10].

Gut mucosal barrier disruption can be a consequence of severe acute viral infections. It was first recognized in acute HIV infection and leads to microbial translocation and subsequent low-grade systemic inflammation, contributing to long-term consequences in chronic HIV infection [11, 12]. Although other viruses do not productively infect and directly kill T cells, gut mucosal barrier disruption can be observed in severe viral infections such as dengue [13] and COVID-19. Mechanisms of gut barrier disruption include direct cytopathic effects on intestinal epithelial cells [14, 15] and cytokine-mediated or activation-induced cell death of T cells [16, 17]. Elevation of plasma biomarkers such as β-glucan, LPS, LPS-binding protein (LBP) and soluble CD14 (sCD14) indicate gut mucosal barrier disruption. We hypothesize that these biomarkers, along with immune-related biomarkers, could be related to disease severity and might differ between dengue infection and COVID-19.

Here, we compared host biomarker profiles in dengue and SARS-CoV-2 infections, which are typically mild or asymptomatic but can lead to severe and fatal conditions. Since dengue and COVID-19 both present with similar fever or influenza-like symptoms during the early phase of infection and are associated with cytokine storm, identifying the specific biomarkers associated with each viral infection may be beneficial to diagnosis and prognosis prediction.

Material and methods

Patients

Patients with dengue viral infection: Patients were eligible if they were 18 years of age or older and presented with suspicion of dengue with fever ≥ 38 °C no later than 72 hours. Patients with fever onset more than 72 hours prior were still eligible if the fever persisted until the time of enrollment. The patients were diagnosed with dengue viral infection by a positive NS1 rapid test (SD BIOLINE) first, followed by a confirmed diagnosis and identification of dengue virus serotypes by Next-generation sequencing (NGS, Illumina). To differentiate between primary and secondary dengue infections, dengue IgG and IgM ELISA testing was performed at the Division of Dengue Hemorrhagic Fever Research, Siriraj Hospital. All patients were admitted between February 2014 and October 2017 at Siriraj Hospital, Bangkok, Thailand. The final diagnosis of dengue fever without warning sings (DF) and dengue with warning signs (DWS) was determined according to the WHO criteria. [5]. Patients with severe dengue—defined by severe plasma leakage, severe hemorrhage, or severe organ impairment—were not included in this study.

Patients with SARS-CoV-2 infection: Patients were eligible if they were 18 years of age or older and diagnosed with COVID-19 by a positive SARS-CoV-2 PCR test between March 2020 and August 2021 at Rush University Medical Center, Chicago, Illinois, US. All blood samples were collected no later than 7 days after the onset of symptoms and were grouped by disease severity based on the Infectious Diseases Society of America Guidelines on the Treatment and Management of Patients with Coronavirus Disease 2019 [18]. Mild COVID-19: patients presenting with COVID-19 symptoms and not requiring hospitalization; Moderate COVID-19: patients presenting with COVID-19 symptoms who required inpatient services due to evidence of lower respiratory tract infection but did not require oxygen supplementation; Severe COVID-19: patients presenting with COVID-19 symptoms requiring inpatient services, including oxygen supplementation and/or ICU admission.

Biomarker measurement

Plasma samples were analyzed using the Meso Scale Discovery (MSD) Platform. IFN-γ, IL-1β, IL-2, IL-4, IL-6, IL-8, IL-10, IL-12p70, IL-13, and TNF-α were assessed using the Proinflammatory Panel 1 (human) kit V-PLEX with a 1:2 dilution. CD163 and myeloperoxidase were assessed using the Immuno-Oncology Group 1 (human) Multiplex Assays U-PLEX with a 1:50 dilution. Matrix mettalloproteinase-9 (MMP-9) was assessed using the Immuno-Oncology Group 1 (human) Multiplex Assays U-PLEX with a 1:100 dilution. IP-10, IL-18, and TRAIL were assessed using the Biomarker Group 1 (human) Multiplex Assays U-PLEX with a 1:5 dilution. LBP and sCD14 were assessed using the Development Packs U-PLEX with a 1:2000 dilution. All MSD test kits were analyzed by the MESO QuickPlex SQ 120 instrument (MESO Scale Diagnostics, Rockville, MD, USA). CD25 and Neutrophil elastase were assessed using an ELISA kit (R&D Systems, MN, USA) with 1:4 and 1:200 dilutions, respectively. D-dimer was assessed using the Asserachrom D-Di ELISA test kit (Diagnostica Stago, NJ, USA). 1,3-β-D-glucan was assessed using the end-point assay Glucatell kit (Associates of Cape Cod, USA) with a 1:4 dilution. All ELISA test kits were analyzed by the BioTek Synergy 2 microplate reader at a 450 nm wavelength (Agilent, CA, USA). All ELISAs and MSD kits were performed according to the manufacturer’s instructions.

Sample size calculation

We used the following formula to calculate the sample size for two groups with an equal number of samples [19, 20]:

• n = 2(Z_α/2 + Z_β)² S²/D² where

• n = sample size required in each group

• Z_α/2 = 1.96 a significance level of 0.05

• Z_β = 0.84 for a power of 80% (β = 0.2)

• S = pooled standard deviation of the two groups

• D = difference in the mean between the two groups

We estimated the values of S and D from a previous study that compared cytokine and chemokine levels in patients with dengue and COVID-19 [21]. In each disease, cytokine levels in patients with mild illness differed markedly from those in patients with more severe disease, and a sample size of fewer than 20 was sufficient to detect a statistically significant difference. Therefore, we aimed to estimate the sample size required to compare biomarker levels between the two diseases of equivalent severity: mild disease (DF vs. mild COVID-19) and more severe disease (DWS vs. moderate–severe COVID-19). Based on the previous study [21], we estimated S/D values of 5.1/2.7 and 25/15 for mild and more severe disease, respectively. Based on the sample size calculated from the formula, we subsequently enrolled 57 patients with DF, 57 patients with mild COVID-19, 40 patients with DWS and 40 patients with moderate–severe COVID-19 to test our hypothesis.

Statistical analysis

Baseline demographic data, clinical, and laboratory results between the study groups were compared using the Chi-square test, Fisher’s exact test, Mann–Whitney U test, as appropriate. Differences in biomarkers in samples were analyzed using the two-tailed Mann–Whitney U test (for two-group comparisons) and the Kruskal–Wallis test with Dunn’s post-hoc test (for multiple comparisons). The p-values of each biomarker from the Mann-Whitney U test were adjusted to control for the false discovery rate (FDR) using the Benjamini-Hochberg approach. P-values or FDR-adjusted p-values < 0.05 were considered significant. Receiver operating characteristic (ROC) curves were generated to present the area under the curve (AUC) and to use the Youden index J to analyze cut-off values, sensitivity, and specificity to predict disease severity biomarkers. Multiple logistic regression was applied to assess the association between certain biomarkers and disease severity, and to calculate the probability of developing severe disease using a combination of these biomarkers. All statistical analyses were performed using SAS 9.4 (SAS, Cary, NC) and GraphPad Prism version 10.3.0 (GraphPad Software, Inc., Boston, MA).

Results

Description of patient cohorts and biomarker panels

Ninety-seven DENV-infected patients were enrolled, of whom 57 developed DF and 40 developed DWS according to WHO criteria. Approximately 10% of patients in both the DF and DWS groups had primary dengue infections, while 90% had secondary dengue infections. All four DENV serotypes were present in the cohort, with serotype 4 being the most prevalent in both the DF and DWS groups (43.90% and 62.50%, respectively). In contrast, serotype 3 was the predominant serotype among patients with primary dengue infections; six out of ten patients (60%) with primary infections were infected with serotype 3. The median age of patients was comparable between the DF and DWS groups (24.11 and 21.68 years, respectively). The median time of blood sample collection was day 2 after the onset of illness in both groups. The nonstructural protein 1 (NS1) levels were significantly higher in the DWS group. However, the hematocrit levels were significantly lower in the DWS group (Supplementary Table S1), likely due to blood samples being taken at the early stage before hemoconcentration. All patients in the dengue infection cohort were otherwise healthy, and all DWS patients survived after treatment.

For the COVID-19 cohort, 97 SARS-CoV-2-infected patients were enrolled, of whom 57 developed mild COVID-19 and 40 developed moderate to severe COVID-19. The patients with moderate to severe COVID-19 were older, predominantly male, had more underlying diseases, and required more intensive care and ventilator support (Supplementary Table S2).

Comparison of the biomarkers during the early phase in dengue and COVID-19 patients

A panel of 22 biomarkers was selected to study the differential response elicited by DENV and SARS-CoV-2 infections. This panel includes innate and T cell-derived cytokines (TNF, IL-1β, IL-6, IL-8, IL-10, IL-12p70, IL-18, TRAIL, IFN-γ, IL-2, IL-4, and IL-13), markers of monocyte and macrophage activation (IP-10 or CXCL10, soluble CD163, MMP-9, and sCD14), markers related to NETs (myeloperoxidase and neutrophil elastase), markers of microbial translocation and intestinal epithelial cell injury (1,3-β-D-glucan and LBP), a coagulation marker (D-dimer), and a marker related to T cell activation and hemophagocytic lymphohistiocytosis (soluble CD25). All 22 biomarkers were analyzed in 97 patients of the DENV infection group and 97 patients of the SARS-CoV-2 infection group. Nineteen out of the 22 markers (excluding soluble CD163 and 1,3-β-D-glucan) were significantly higher in DENV-infected patients, whereas MMP-9 was significantly higher in COVID-19 patients (Table 1).

Table 1.

Comparison of biomarkers between patients with dengue infection and COVID-19

Markers, pg/mL	DENV infection	SARS-CoV-2 infection	FDR-adjusted p-values
Median (IQR)	(N = 97)	(N = 97)
TNF-α	20.86 (15.69–27.10)	5.87 (4.95–9.40)	< 0.001
IL-1β	13.91 (9.80–18.82)	9.12 (3.98–16.46)	< 0.001
IL-6	13.29 (8.05–18.61)	3.89 (2.19–9.39)	< 0.001
IL-8 (CXCL8)	62.23 (43.76–91.01)	14.90 (10.86–27.46)	< 0.001
IL-10	72.31 (20.51–288.41)	1.12 (0.77–3.15)	< 0.001
IL-12p70	5.03 (2.27–9.41)	1.26 (0–4.70)	< 0.001
IL-18	874.70 (672.71–1142.40)	600.09 (425.89–875.59)	< 0.001
IFN-γ	1805.77 (574.14–5823.99)	17.04 (12.71–39.01)	< 0.001
IL-2	42.71 (31.99–60.74)	8.26 (3.35–12.82)	< 0.001
IL-4	1.03 (0.72–1.51)	0.83 (0.41–1.27)	< 0.05
IL-13	4.38 (3.19–5.53)	2.16 (1.49–3.15)	< 0.001
TRAIL	193.25 (84.26–532.22)	77.22 (55.03–110.19)	< 0.001
IP-10 (CXCL10)	28634.06 (19780.98– 41,137.80)	546.94 (303.08–2051.04)	< 0.001
soluble CD163 (x105)	1.94 (1.52–2.73)	2.01 (1.39–2.81)	0.91
MMP-9 (ng/ml)	8.67 (5.56–13.78)	12.74 (6.09–26.15)	< 0.01
soluble CD14 (μg/mL)	2.32 (1.90–3.34)	1.93 (1.52–2.77)	< 0.001
LBP (μg/mL)	10.44 (8.29–12.60)	6.32 (4.12–13.95)	< 0.001
β-d-glucan	108.27 (84.92–146.58)	105.85 (76.18–126.30)	0.07
soluble CD25	2121.33 (1506.82–3623.74)	1257.35 (839.45–2240.85)	< 0.001
MPO (ng/mL)	27.88 (10.29–66.93)	2.36 (1.23–7.61)	< 0.001
Neutrophil elastase (ng/mL)	252.40 (226.46–289.09)	88.57 (49.75–140.95)	< 0.001
D-dimer (ng/mL)	1742.66 (1615.07–1853.90)	869.02 (628.85–1123.89)	< 0.001

TRAIL, TNF-related apoptosis-inducing ligand; MMP, matrix metalloproteinase

LBP, LPS binding protein; MPO, myeloperoxidase

Comparison of the biomarker response in early DENV-infected patients who developed DF or DWS

We stratified each cohort to identify predictors of disease severity. Comparing mild DF (n = 57) to DWS (n = 40), there were 14 biomarkers that were significantly higher in the patients who developed DWS compared to those who developed DF (Fig. 1A and Supplementary Table S3). To evaluate the usefulness of these 14 biomarkers as early predictive biomarkers for the development of DWS, we created receiver operating characteristic (ROC) curves and found that the area under the curve (AUC) for all 14 biomarkers was > 0.6 (Supplementary Figure S1). Five markers, including myeloid cell-derived innate cytokines (IL-1β and IL-12p70), type 2 immune response-related cytokines (IL-4 and IL-13), and MPO (myeloperoxidase), exhibited AUC values of 0.7–0.8. However, only 1,3-β-D-glucan had an AUC of 0.858 (95%CI: 0.784–0.934), which was significant (p-value < 0.0001). With levels of 1,3-β-D-glucan > 128.7 pg/mL, the sensitivity and specificity of developing DWS were 67.50% and 89.47%, respectively, with a likelihood ratio of 6.4 (Fig. 1B).

Fig. 1.

Biomarkers for dengue infection severity prediction. Only 6 biomarkers whose AUCs were higher than 0.7 were shown. A. Comparison of these 6 biomarkers in 57 df and 40 DWS patients. *P-value < 0.01, **P-value < 0.001 by Mann-Whitney U test, B. ROC curves of these biomarkers

Prediction of DWS using a combination of biomarkers

We then selected six biomarkers—IL-1β, IL-12p70, IL-4, IL-13, 1,3-β-D-glucan (AUC > 0.7) and sCD14 (AUC = 0.68, supplementary Figure S1)—to perform multiple logistic regression. sCD14 was included because it reflects the activation of monocytes (a primary target cell of the dengue virus) and is associated with gut mucosal barrier disruption. In contrast, myeloperoxidase was not included because its specificity was lower than 80%. The results showed that five biomarkers—IL-1β, IL-12p70, IL-13, 1,3-β-D-glucan, and sCD14—were significantly associated with DWS. The sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) of this multiple logistic regression model were 90% (95% CI: 76.3–97.2%), 93% (95% CI: 83–98%), 90% (95% CI: 76.3–97.2%), and 93% (95% CI: 83–98%), respectively (Table 2). Multiple logistic regression using a combination of six biomarkers yielded an AUC of 0.9836 (95% CI: 0.9659–1) (Fig. 2A), which was better than using a single biomarker (Fig. 1B). Interestingly, two of the four DWS patients with primary dengue infections had predicted probabilities of 0.1 and 0.55, respectively (Fig. 2B). This suggests that primary dengue infections are associated with low biomarker levels.

Table 2.

Multiple logistic regression model using six biomarkers to predict the probability of developing DWS

A
	Regress. Coef. (Beta)	ORs (95% CI)	P-value
Intercept	−5.43		0.00007*
IL-1beta (High)	3.246	25.70 (3.58–358.9)	0.004*
IL-12p70 (High)	3.824	45.81 (3.92–1907)	0.011*
IL-4 (High)	−1.223	0.29 (0.02–3.34)	0.351
IL-13 (High)	3.595	36.40 (1.38–2186)	0.049*
Beta-glucan (High)	3.246	25.69 (3.62–364.2)	0.004*
sCD14 (High)	4.299	73.65 (9.47–1381)	0.0004*
B.			.
Classification table	Observed DWS	Observed DF	Total
Predicted DWS	36	4	40
Predicted DF	4	53	57
Total	40	57	97

A. Association of each biomarker with DWS. Regress. Coef., regression coefficients or beta ORs, adjusted odds ratios = e^beta; * statistically significant

High:IL-1beta > 16.75 pg/mL, IL-12p70 > 8.129 pg/mL, IL-4 > 1.462 pg/mL, IL-13 > 5.907 pg/mL, beta-glucan > 128.7 pg/mL, sCD14 > 3.506 μg/mL

B. Confusion matrix or classification table showing the number of predicted and observed DF and DWS patients, using a probability cut-off value of 0.5

Fig. 2.

A. ROC curve of the multiple logistic regression model predicting DWS using a combination of six biomarkers. B. Scatter plot showing the predicted probability of developing DWS. The red dot indicates DWS patients with primary dengue infections

Comparison of the biomarker response in acute SARS-CoV-2-infected patients who developed mild COVID-19 or moderate–severe COVID-19

We found 18 differentially expressed plasma biomarkers between patients with mild and moderate–severe COVID-19; 16 biomarkers were significantly increased, and 2 biomarkers (IL-1β and IL-4) were decreased in patients with moderate to severe disease compared to those with mild disease (Fig. 3A and Supplementary Table S4). We evaluated the usefulness of these biomarkers for their predictive value in developing moderate-severe COVID-19 disease by ROC curve analysis. Among the biomarkers elevated in moderate-severe COVID-19, the AUCs of 3, 6, 5, and 2 biomarkers were 0.6–0.7, 0.7–0.8, 0.8–0.9, and > 0.9, respectively (Supplementary Figure S2). Biomarkers with AUCs > 0.8 include inflammasome-related innate cytokine (IL-18), type 1 immune response-related chemokine (IP-10 or CXCL10), NET-related markers (myeloperoxidase and neutrophil elastase), and markers of monocyte activation and gut permeability disruption (sCD14, LBP, and 1,3-β-D-glucan) (Fig. 3B). The AUCs of IP-10 and LBP were both more than 0.9. The AUC of IP-10 (CXCL10) was 0.957 (95% CI: 0.921–0.992), which was significant (p < 0.0001). At IP-10 levels > 631.1 pg/mL, the sensitivity and specificity for developing moderate-severe COVID-19 were 90.00% and 89.47%, respectively, with a likelihood ratio of 8.55. This marker appears to be the most significant predictor for developing severe COVID-19. The AUC of LBP was 0.912 (95% CI: 0.845–0.979), which was also significant (p < 0.0001). At LBP levels greater than 7,494 ng/mL, the sensitivity and specificity for developing moderate–severe COVID-19 were 85.00% and 85.96%, respectively, with a likelihood ratio of 6.06 (Fig. 3B).

Fig. 3.

Biomarkers for COVID-19 severity prediction. Only 7 biomarkers whose AUCs were higher than 0.8 were shown. mpo, myeloperoxidase, A. Comparison of these 7 biomarkers in 57 mild and 40 moderate-severe COVID-19 patients. *P-value < 0.001 by mann-whitney U test, B. ROC curves of these biomarkers

Prediction of moderate to severe COVID-19 using a combination of biomarkers and demographic data

We then combined demographic data with five biomarkers (with AUC > 0.8, specificity > 85%, likelihood ratios > 5; Fig. 3B)—IP10 (the chemokine marker with highest AUC), LBP, sCD14, 1,3-β-D-glucan (three markers related to gut barrier disruption), and neutrophil elastase (a NET-related marker)—to perform multiple logistic regression. The results showed that three biomarkers—IP10 (CXCL10), LBP, sCD14—along with male gender were significantly associated with moderate to severe COVID-19. The sensitivity, specificity, PPV, and NPV of this multiple logistic regression model were 92.5% (95% CI: 79.6–98.4%), 96.5% (95% CI: 87.9–99.6%), 94.9% (95% CI: 82.7–99.4%), and 94.8% (95% CI: 85.6–98.9%), respectively (Table 3). The multiple logistic regression model, using a combination of five biomarkers and demographic data, yielded an excellent AUC of 0.9882 (95% CI: 0.9725–1) (Fig. 4). This model outperformed those using biomarkers or demographic data alone (Supplementary Figure S3).

Table 3.

Multiple logistic regression model using five biomarkers plus demographic data to predict the probability of developing moderate to severe COVID-19

A.
	Regress. Coef. (Beta)	ORs (95% CI)	P-value
Intercept	−7.075		0.021*
IP10 (High)	4.29	73 (6.47–3339)	0.004*
LBP (High)	3.455	31.66 (2.737–1279)	0.019*
sCD14 (High)	4.5	90.01 (4.59–7668)	0.013*
Beta-glucan (High)	0.6683	1.95 (0.15–26.27)	0.596
NE (High)	2.413	11.17 (0.41–725.2)	0.175
Gender (male)	3.59	36.23 (2.44–2558)	0.028*
Underlying disease (presence)	1.583	4.87 (0.33–125.4)	0.271
Age (in years)	−0.02251	0.98 (0.87–1.1)	0.681
B
Classification table COVID-19	Observed moderate to severe	Observed mild	Total
Predicted moderate to severe	37	2	39
Predicted mild	3	55	58
Total	40	57	97

A. Association of each biomarker and demographic data with moderate to severe COVID-19

*statistically significant; NE, neutrophil elastase; LBP, LPS-binding protein

High: IP10 > 631.1 pg/mL, LBP > 7.494 μg/mL, sCD14 > 2.178 μg/mL, Beta-glucan > 112.47 pg/mL, NE > 111.85 ng/Ml

B. Confusion matrix or classification table showing the number of predicted and observed mild and moderate-to-severe COVID-19 patients, using a probability cut-off value of 0.5

Fig. 4.

ROC curve of the multiple logistic regression model predicting moderate to severe COVID-19 using a combination of five biomarkers and demographic data

Comparison of 1,3-β-D-glucan levels in patients with dengue infection and COVID-19

As 1,3-β-D-glucan was the best predictive biomarker for severe dengue and COVID-19 disease, we next asked whether its levels differed significantly between patients with these two infectious diseases. Therefore, we compared 1,3-β-D-glucan levels in four groups of patients: DF, DWS, mild COVID-19, and moderate to severe COVID-19. Although 1,3-D-d-glucan levels were highly elevated in patients with more severe disease, no statistical differences were observed in patients with mild infections (DF vs mild COVID-19) or those with severe diseases (DWS vs moderate-severe COVID-19) (Fig. 5).

Fig. 5.

1,3 β-d-glucan levels in 4 groups of patients: 57 DF, 40 DWS, 57 mild and 40 moderate-severe COVID-19. *P-value < 0.001 by Kruskal-Wallis with Dunn’s post-hoc test. ns, not significant

Discussion

Measurement of host immune responses early in infection yielded important insights into the trajectory of dengue and COVID-19 disease courses. Evaluating patients with mild and severe disease, 14 biomarkers were significantly increased in patients who developed DWS compared to DF, and 16 biomarkers were significantly increased in patients who developed moderate to severe COVID-19 compared to mild COVID-19. Based on the AUC of ROC analysis, 1,3-β-D-glucan was the best predictive biomarker for severe dengue infection. For COVID-19, CXCL10, NETs- and gut epithelial barrier disruption-related biomarkers were predictors of severity. Indeed, 1,3-β-D-glucan, the direct indicator of gut microbial translocation, was significantly elevated in both severe dengue and SARS-CoV-2 infections, suggesting that gut epithelial barrier disruption occurs in severe cases of both infections.

Our results demonstrated that 20 out of 22 biomarkers were significantly higher in patients with dengue infection compared to those with COVID-19. These findings can be explained by the tropism of the dengue virus for myeloid cells (monocytes/macrophages and dendritic cells) and the phenomenon known as “original antigenic sin” [7]. Severe dengue infection often occurs upon secondary infection. Since the four serotypes of the dengue virus are distinct from each other, the adaptive memory response mounted after the primary infection cannot prevent reinfection with another serotype. Instead, cross-reactive T cells and B cells trigger a cytokine storm and antibody-dependent enhancement (ADE), leading to increased viral load and disease severity [7, 22]. Consequently, several innate and T cell-derived inflammatory cytokines appear highly elevated in the early phase of infection, as observed in this study. In contrast, the SARS-CoV-2 virus has a tropism for non-immune cells rather than myeloid cells. Although re-infection by different SARS-CoV-2 variants is very common due to escape mutations in B cell epitopes of the RBD region, the T cell epitopes are mostly conserved [23, 24]. Therefore, T cell response against the original variant can reduce disease severity and prevent an excessive cytokine storm [25]. Taken together, this evidence may explain why most inflammatory biomarkers are much lower in SARS-CoV-2 infection than in dengue infection. On the other hand, MMP-9 was significantly increased in COVID-19 patients, particularly those with moderate to severe disease, compared to dengue-infected patients. MMPs are enzymes in a family of zinc-dependent endopeptidases that degrade extracellular matrix proteins and are involved in the tissue remodeling process. An increase in MMP-9 in moderate to severe COVID-19 patients may be associated with adult respiratory distress syndrome (ARDS)-related pulmonary fibrosis, which is a complication of severe COVID-19 disease [26].

When comparing DF to DWS patients, type 2 immunity cytokines, namely IL-4 and IL-13, were strikingly increased in DWS patients and may be used as predictive biomarkers for severe diseases. Apart from T helper 2 cells (Th2), certain innate immune cells, including type 2 innate lymphoid cells (ILC2) and mast cells/basophils, can secrete Th2 cytokines [27–29]. Human ILC2 cells are productively infected by DENV [30], and ILC2 from DHF patients produce higher levels of IL-4 than those from DF patients [31]. Human primary mast cells or mast cell/basophil cell lines can be infected by DENV, especially in presence of anti-DENV antibody [32, 33]. Indeed, mast cell degranulation has been implicated in pathogenesis of severe DHF [34]. Moreover, Th2 cytokines, particularly IL-13, can cause epithelial barrier dysfunction [35], resulting in gut microbial translocation, as evidenced by the increased levels of 1,3 β-d-glucan. Regarding innate cytokines, IL-1β and IL-12p70 are specifically secreted by myeloid mononuclear phagocytes (the target cells of dengue virus) [36, 37], whereas certain cytokines and chemokines (e.g. IL-6, IL-10, CXCL8 and CXCL10) can also be produced by non-immune cells [38, 39]. The myeloperoxidase enzyme is also found in monocytes [40], in addition to neutrophils. The increase in IL-1β, IL-12p70 and meloperoxidase in patients who would develop DWS may be explained by the effects of ADE triggered by viral entry into myeloid cells.

Several biomarkers were significantly elevated in patients with moderate to severe COVID-19. Based on the AUCs determined by ROC analysis, we suggest that IL-18, IP-10 (CXCL-10), myeloperoxidase, neutrophil elastase, 1,3 β-D-glucan, sCD14, and LBP could be used as predictive biomarkers for disease severity. CXCL10 attracts monocytes/macrophages and T cells during type 1 mediated immunity and has been previously implicated in COVID-19 pathology [39, 41]. This chemokine can be upregulated by cytokines such as IFN-γ and TNF which were also significantly increased in moderate-severe COVID-19 patients. NETs, evidenced by the increased levels of myeloperoxidase and enzymes in our study, reportedly play an important role in organ injury in COVID-19 and correlate with the cytokine storm [42]. Gut mucosal barrier disruption is well recognized in COVID-19 and can be caused by several mechanisms including direct viral infection causing cytopathic effects in intestinal epithelial cells, as well as cytokine-mediated mucosal damage [15]. Mucosal barrier dysfunction results in microbial translocation (increases in blood 1,3 β-D-glucan levels) and subsequent systemic inflammation (increase in sCD14 and LBP levels). Our results confirm that NETs and gut mucosal disruption markers are associated with and can be used as predictive biomarkers for COVID-19 severity.

Interestingly, we found that the 1,3 β-D-glucan levels were not statistically significant between patients with DWS and those with moderate to severe COVID-19, suggesting that gut mucosal disruption occurs in severe cases of both viral infections. Several biomarkers including 1,3 β-D-glucan, sCD14 and LBP are usually considered suggestive evidence of gut barrier disruption. However, only 1,3 β-D-glucan reflects fungal translocation from GI tract, whereas sCD14 and LBP indicate the subsequent activation of innate immunity. CD14, which normally functions as a receptor for bacterial LPS, is either shed from membrane-bound CD14 or secreted from intracellular compartments upon monocyte activation. Several inflammatory cytokines in addition to LPS simulate the release of sCD14 [43]. LBP, which facilitates the binding of LPS to CD14, is indeed an acute phase protein primarily produced by liver and its production is known to increase in several inflammatory conditions [44]. We believed that the greater increase in sCD14 and LBP in dengue infection compared to COVID-19 is the consequence of strong systemic inflammatory reactions, rather than gut microbial translocation alone. The mechanisms of gut mucosal barrier disruption in human dengue infection are largely unknown. Gut barrier dysfunction may be mediated by TNF, IFN-γ, or Th2 cytokines [45], all of which were highly elevated in DWS patients of our study. Inflammatory pathology of gut mucosa and lymphoid tissues was observed in mouse models of dengue infection [13]. Previous clinical studies have demonstrated the increased levels of LPS and 1,3 β-D-glucan in dengue-infected patients, which were prominent on day 7 compared to early stage of infection [46, 47]. Our study found increased levels of 1,3 β-D-glucan in DWS patients within 2–3 days of symptom onset, suggesting it may have utility as an early predictive biomarker for disease progression to DWS.

Our study has several limitations. First, the differences in demographic characteristics between dengue-infected and COVID-19 patients could influence their respective immune responses. Dengue-infected patients are from a Thai population, mostly young and healthy, whereas COVID-19 patients are from an American population, mostly older and with more underlying diseases. We selected the US population for the COVID-19 biomarker study because we wanted to exclude any past immune response to dengue in COVID-19 patients. Second, our dengue cohort demonstrated a difference in serotype distribution between patients with primary and secondary dengue infections: DENV-3 was predominant in primary infections, while DENV-4 was more common in secondary infections. Therefore, this finding may be less relevant to real-world situations where other serotypes are more prevalent. Third, these biomarkers were measured in single plasma specimens. Further studies in serial samples should be performed to confirm the kinetics of biomarker expression. Fourth, our study—particularly the multiple logistic regression models—is limited by a small sample size. Further studies with larger sample sizes should be conducted to confirm the model’s predictions. Additionally, these biomarkers should be validated in patients co-infected with dengue and SARS-CoV-2.

Conclusion

Several differences and similarities were observed in hosts responses to mild and severe forms of dengue and COVID-19. For severe dengue, we demonstrated that myeloid mononuclear cell-derived cytokines (IL-1β and IL-12p70) and type 2 immunity-related cytokines (IL-13) could be used as early predictive biomarkers. We further confirmed that CXCL10 and NETs are indicative of SARS-CoV-2 infected patients who go on to develop moderate-severe COVID-19. Furthermore, 1,3 β-D-glucan, a gut barrier defect marker, is a general predictor of disease severity expressed during early acute viral infection.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

The authors confirm the following contributions to this research study: study conception and design: EY, JK, MB, GC, JM, AL, YS, and WT; data collection: EY, JM, and YS; analysis and interpretation of results: EY, JK, AL, and WT; draft manuscript preparation: EY, JK, MB, GC, AL, YS, and WT. All authors reviewed the results and approved the final version of the manuscript to be submitted for journal publication.

Funding

Open access funding provided by Mahidol University. This study was supported by a grant from Abbott Diagnostics Division Research and Development.

Data availability

Additional data is available in the supplementary material. Requests for additional data should be directed to the corresponding author.

Declarations

Ethics approval and consent to participate

This study was approved by Siriraj Hospital Institutional Review Board (certificate of approval 409/2023) and Rush University Institutional Review Board, Office of Research Affairs# 21080502. The research was conducted in accordance with the Declaration of Helsinki. All participants were 18 years of age or older. Informed consent to participate was obtained from all of the participants before the study. The research team restricted the participants’ information and maintained confidentiality throughout the study period.

Consent for publication

Not applicable.

Competing interests

The authors declare the following financial interests/personal relationships: MB, and GC are employees and shareholders of Abbott Laboratories. This work was funded by Abbott Laboratories. AL is a consultant for Abbott Laboratories. All other authors report no potential conflicts of interest.