Interferon gamma and inflammatory response genes by weighted gene co-expression network analysis reveal significant associations with severe dengue infection in India

Abstract

Objective

Dengue virus, with four serotypes (DENV 1-4), is transmitted by Aedes aegypti. India carries 34% of global cases, yet studies to understand how the immune system reacts to dengue infection remain limited.

Design

This study uncovers early transcriptional signatures from whole blood of 112 NS1 positive hospital reported dengue patients and dissect biological processes linked to varying disease severities “with (WS) and without (-WS) warning signs”.

Results

Differential gene expression analysis showed suppressed innate response in WS patients, with downregulated cytokines (CXCL8, CXCL16, CXCL2, IL1B) and disrupted platelet function. Upregulation was observed in genes related to glycosylation, nervous system, and metabolism. WGCNA module-trait analysis identified interferon response and inflammation as positively correlated with platelet count, highlighting key genes, IFI44L, IFI35, OAS1, HERC6 and SAMDL9 contributing to comparative severe outcomes. CTTN involved in maintenance of vascular permeability showed highest significant association with low platelet count. IFI44L emerged as a significant marker, showing highest transcript count difference between patients with platelet counts below 100 × 10^9/L and those between 100 and 140 × 10^9/L.

Conclusion

This study advances understanding of dengue infection and identifying probable clinical biomarkers at the initial phase of infection. These findings might hold potential for improving diagnostic and therapeutic strategies.

Highlights

• •Early transcriptomic profiling reveals distinct signatures across dengue disease severity.
• •WS patients exhibit suppressed cytokines/interleukins and impaired platelet pathways.
• •Interferon and inflammatory gene modules correlate strongly with thrombocytopenia and severity.
• •CTTN and IFI44L emerge as key candidate biomarkers linked to platelet decline and clinically severe outcomes.
• •Early-phase molecular insights highlight utility in risk stratification and clinical management.

1. Introduction

DENV infects 390 million people annually, with 96 million experiencing clinical or subclinical symptoms, roughly three times more than WHO's dengue burden estimate [1,2]. India accounts for about 30% of global dengue infections, with many regions being hyper-endemic with multiple serotypes co-circulating year-round [1,3].

DENV infection leads to a broad spectrum of clinical manifestations ranging from asymptomatic cases to severe and potentially life-threatening conditions. Revised classification of 2009 categorized dengue into dengue without or with warning signs (D-WS; DWS) and severe dengue (SD) [[4], [5], [6]], wherein ‘with warning signs’ included symptoms of coagulation abnormalities (thrombocytopenia), clinical fluid accumulation (hemoconcentration) and increased vascular fragility [7]. Therefore, an initial assessment of clinical symptoms in combination with laboratory clinical parameters is crucial for evaluating the severity of dengue infection.

Human host transcriptional dynamics capturing gene functions aid in comprehending the cellular processes leading to complexities of dengue pathogenesis Several host pathways significantly impacted by dengue virus infection include antiviral and pro-inflammatory cytokine, TLR signaling [8], NF-kappa B signaling [9], TNF signaling [10] and complement and coagulation cascades [11]. Endothelial cells infected with DENV secrete cytokines and interleukins, which exacerbate vascular permeability and trigger inflammatory response, leading to plasma leakage [12]. Thus, early gene expression patterns can provide clues about the progression and severity of the disease [13].

We have previously reported the pathogen load and host response from serum samples of a smaller set of 24 dengue patients [14].

In a country of 1.4 billion people, understanding the factors driving dengue disease severity remains a critical need. To address this, our study investigated early transcriptional signatures in whole blood from a robust cohort of 112 dengue-positive patients presenting with and without warning signs.’ Host mRNA expression pattern was comprehensively evaluated through Weighted Gene Co-expression Network Analysis (WGCNA) to identify biological processes altered and its association with severity, based on clinical parameters. We identified co-expression modules associated with clinical traits, providing novel insights into transcriptomic changes during dengue infection. Intriguingly, in patients demonstrating warning signs, the entire immune landscape especially interferon response and inflammation positively correlated with low platelet count, modulating the pathogenesis towards clinically severe outcome.

2. Methods

2.1. Patient recruitment, sample and data collection

This cross-sectional study included 112 hospital-reported symptomatic individuals, wherein informed consent was duly acquired from all participants, and blood samples collected using EDTA vials, by a proficient medical team at MAX Hospital, New Delhi, India, during the patients' initial hospital visit.

2.2. RNA isolation and library preparation

RNA extraction from 1 ml of blood from each patient sample was conducted using QIAGEN QIAamp RNA Blood Mini Kit (Cat. No. 52304) with specific adjustments, including reduced incubation and centrifugation durations for erythrocyte lysis and an additional 3-min incubation during washing stages. RNA purity was evaluated using NanoDrop Spectrophotometer and confirmed through agarose gel electrophoresis. The RNA specimens were preserved at −80°C in preparation for library construction. Subsequently, 250 ng of total RNA underwent processing employing Illumina TruSeq Stranded Total RNA Library Prep Globin Kit (Illumina, Cat. No. 20020612). This process entailed depletion of globin mRNA and ribosomal RNA. cDNA was synthesized through a multi-step process, starting with the reverse transcription of RNA fragments into first-strand cDNA using random hexamer primers and reverse transcriptase in the presence of Actinomycin D, which ensured strand specificity. The RNA template was removed, and a complementary second strand DNA was synthesized using polymerase 1 and RNase H, to generate double-stranded cDNA. Magnetic bead-based purification of cDNA was done with AMPure XP (Beckman Coulter). The purified cDNA subsequently underwent adenylation at the 3′ termini of blunt end double stranded cDNA.

Sequencing libraries were uniquely indexed and amplified using PCR. Quality assessment of the libraries was done using Agilent 2100 Bioanalyzer, while concentrations were quantified using Qubit dsDNA HS assay kit (Thermo Fisher Scientific, Cat. No. Q32854). Following dilution to 4 nM, libraries were pooled in equimolar proportions and subjected to sequencing with paired-end 2×151 read lengths on Illumina NextSeq 2000 platform at a final loading concentration of 650 pM.

2.3. Differential gene expression analyses

Raw reads were demultiplexed using bcl2fastq and quality-checked with FastQC [15]. Adapter trimming and low-quality filtering (Phred <20) were performed using Trimmomatic [16]. Clean reads were quasi-mapped to the human transcriptome (GENCODE v44, GRCh38) using Salmon. Gene-level differential expression was computed with DESeq2 (genes with >10 reads in ≥10 samples). Significant differentially expressed genes DEGs were defined as adjusted p ≤ 0.05 and |log2FC| ≥1.5. Differential gene analysis was carried out between different severity patient groups, namely WS and -WS. Log fold change was plotted against p-adjusted value using the EnhancedVolcano R package [17].

2.4. Functional enrichment and WGCNA network analysis

Functional enrichment of DEGs was performed using Enrichr and Reactome database [18]. Weighted gene co-expression network analysis was conducted using WGCNA v1.72.5 [19] on 14,071 genes across 112 samples, with modules identified using dynamic tree cutting (using the cutreeDynamic function with a minimum module size of 30 genes (minModuleSize = 30) to ensure sufficient statistical power for module detection. Module-trait associations were assessed using Pearson correlation (|R|≥0.50, p < 0.05), and gene-trait relationships using Spearman correlation. To ensure the network adhered to a scale-free topology, we used the pickSoftThreshold function in WGCNA to determine the optimal soft threshold power (β). The results were visualized using heatmaps of module-trait associations and the plotDendroAndColors function from WGCNA, which displays the hierarchical clustering of genes within the modules.

2.5. Statistical analysis

Analyses were performed in R and licensed GraphPad Prism. For clinical parameters, non-normal continuous traits were analyzed via Mann-Whitney U tests and categorical variables via Chi-square tests. WGCNA module-trait associations used Pearson correlation for continuous linear dynamics, whereas individual gene-trait relationships used Spearman rank correlation to minimize outlier effects. Benjamini-Hochberg False Discovery Rate (FDR) corrections were systematically applied to all intergroup and module-trait multi-testing comparisons (p < 0.05). Visualization was conducted using ggplot 2.

3. Results

3.1. Study cohort and design of the dengue patients

The Dengue cohort included 112 patients with febrile illness. 53 were “In Patient” (hospitalized) whereas 59 showed at “Out Patient” Department. The clinical symptoms essentially included fever (with and without chills) ranging from 2 to 6 days, with varied patients reporting with vomiting, nausea, headache and generalized weakness (Supplementary Table S1). Blood parameters included CBC (Complete Blood Count) and LFT (Liver Function Test). Based on 2009 WHO Dengue classification guidelines, 91 fell in the category of “without warning signs” (-WS). Within this group, 46 patients had leukopenia (LP) in their CBC report. Hence, we segregated the “without warning sign” patients into -WS-LP (n = 45) and -WS + LP (n = 46). These groups were referred to as mild and moderate respectively. Further 21 patients fell in “with warning signs” category displaying thrombocytopenia (TC). Hence, they were designated as WS + TC. Incidentally, these patients also had leukopenia. None of the patients fell in Severe Dengue category based on the WHO guidelines, yet in our study, we considered WS + TC group to be demonstrating the propensity for clinically severe disease manifestation, due to low platelet count, hence they were referred as severe.

All 112 patient blood samples underwent total RNA isolation and sequencing to capture the transcriptomic profile arising due to Dengue infection. 2,076,080,122 reads were generated during sequencing, with an average of 18,536,420 reads per sample. Downstream differential gene expression (DESeq2) analysis involved mapping with protein coding genes from GENCODE GRCh38. p14. Moreover, to reduce potential bias of the data set, we incorporated batch correction criteria during DESeq2 analysis. Furthermore, confounder analysis for age and gender during DESeq2 did not reveal any significant differences between the clinical groups. WGCNA grouped thousands of genes into manageable modules correlating with biological annotations to study complex transcriptomic data. Fig. 1 illustrates the clinical cohort segregation, experimental overview and downstream analysis.

Fig. 1.

Study Design Workflow.

A schematic workflow illustrating the NS1 positive dengue sample clinical classification and categorization: –WS–LP (no warning signs and no leukopenia), –WS + LP (no warning signs with leukopenia), and WS + TC (presence of warning signs and thrombocytopenia), followed by RNA isolation and human host transcriptome sequencing. Data analysis, including identification of differentially expressed (DE) genes and WGCNA was performed for association studies with clinical parameters.

3.2. Demographic and clinical characteristics of the dengue cohort

All 112 patients in this cohort experienced primary DENV infection with no secondary infection as confirmed clinically as well as with the patients. Samples were collected during the acute phase at the patients' first hospital visit, and the day of fever at sampling was included. Using viral reads recovered from the RNA-sequencing data, dengue serotype analysis demonstrated that the majority of patients were infected with DENV-2, consistent with the circulating serotype in Delhi during the study period. The Dengue cohort showed uniform presence with regard to age and gender in -WS-LP and -WS + LP groups, (Supplementary Table S2), whereas the WS + TC group showed more of male patients (81%) with higher median age of 35yrs. The non-significant confounder analysis negated any biases arising due to difference in age and gender between the groups. Patients were in acute febrile phase (characterized by fever that lasts 2-7 days) wherein, the mean Days of fever (DOF) of patients was 3.69. The day of fever (DOF) onset in patients in each group is provided in the Supplementary Table S2.

The CBC profile provided an initial assessment of Dengue severity, as blood parameters are altered by Dengue virus. Along with TLC (Total Leukocyte Count) and platelets (p-value <.001), lymphocytes and neutrophils were also significantly different between the groups (p-value <.001) (Supplementary Fig. S1–A). The two groups which showed leukopenia, -WS + LP (moderate) and WS + TC (severe), displayed low neutrophil and higher lymphocyte values when compared to the -WS-LP (mild). Neutrophils are the first line of active defense during any bacterial or viral infection, killing pathogens and removing cellular debris [20], emphasizing a probable heightened infection response in -WS-LP; mild group. Further, NLR ratio is considered as predictive of disease severity [21]. One of the features of clinically severe dengue is plasma leakage. A 20% increase in hematocrit levels is widely used as an indicator of plasma leakage. Hence, normal Hematocrit level is expected in patients without warning signs which are mild and moderate groups. Incidentally, we did not observe rise in Hematocrit in WS + TC patients also, suggesting no capillary leak in patients during the time of initial investigation.

LFT report was available for a subset of patients, based on the clinicians’ evaluation, from each group (Supplementary Fig. S1–B). Hepatic dysfunction is a feature of DENV infection ranging from mildly elevated levels of transaminases, to jaundice to acute liver failure, dissociating cases on differential severity. Higher than normal range for SGOT and SGPT levels were observed in WS + TC, although significant difference was observed for SGOT only in comparison to -WS-LP. Total protein was significantly down in WS + TC whereas bilirubin levels were significantly elevated in WS + TC when compared to the -WS-LP and -WS + LP respectively, yet they all fell within the normal range, thereby suggesting concurrent derangement at later stage of infection (Supplementary Fig. S1–B).

3.3. Early whole blood transcriptional signature depicts altered states of dengue infection in WS and -WS patients

Extensive interrogation on early transcriptomic level variations in dengue was done through differential gene expression analysis, firstly, within no warning signs groups with and without leukopenia (moderate vs mild). Then, between no warning signs groups and with warning signs i.e. severe vs mild and severe vs moderate. Volcano plot depicts differentially regulated genes in the various comparison groups (Supplementary Fig. S2). Differential gene expression analysis between moderate vs mild yielded 15 significant downregulated genes in moderate patients. Majority of genes were associated with immune response (CXCL2, CCRL2, CD177, ORM1), regulated inflammation (PI3) and immune signaling (HES4, DUSP13). Functionally, CXCL2 is one of the early chemokines expressed via CCRL2 axis during DENV infection and responsible for attracting neutrophils, CD177 and ORM1 are markers of increased neutrophil production and acute inflammation. PI3 regulates inflammation by preventing excessive tissue injury. HES4 and DUSP13 regulate immune response via NOTCH and MAPK signaling respectively. These downregulated genes are suggestive of subdued immune response in moderate patients compared to mild.

Severe vs Mild patients’ comparison showed 215 Up-/92 Down-regulated genes whereas severe vs moderate displayed 142 Up-/2 Down-regulated significant genes. Taking together, the genes captured in “with warning signs” (WS + TC; severe) spanned over vastly different gene functions responsible for major host processes like membrane trafficking, immune response & inflammation, cell signaling, transcription regulation, cell adhesion, and metabolism as depicted in Fig. 2A.

Fig. 2.

Functional Analysis of Differentially Expressed Genes.

(A) Chord diagram illustrating biological processes identified using Enrichr (Reactome database) with all significant DEGs. (B) Pathways with significant enrichment in Moderate vs Mild. (C) Associated pathways with the up regulated genes in Severe vs Mild and Severe vs Moderate. (D) Associated pathways for the down regulated genes in Severe vs Mild and Severe vs Moderate are visually represented.

To interpret molecular pathways associated with DE genes, we performed pathway enrichment analysis using Enrichr against Reactome database. Fig. 2B depicts significant pathways for moderate vs mild down regulated DE genes whilst heat map for gene expression is given in Supplementary Fig. S3. Since severe vs mild patients’ comparison had both up and down regulated genes in substantial numbers, we segregated the DE genes during pathway analysis featuring enhanced or depleted functions. Pathways were collated that were differentially altered in WS (severe) with respect to -WS (mild/moderate) due to the dengue infection. The upregulated pathways (Fig. 2C) majorly reflect processes of glycosylation and neuronal system, uniformly across both the groups. Interestingly, downregulated pathways between severe vs mild demonstrated immune system and inflammatory response pathways along with GPCR signaling involved in the generation of immune cascades. It also demonstrated pathways of platelet activation & aggregation along with prostaglandin synthesis in the platelets and endothelial cells, revealing altered platelet function and inflammatory mediators (Fig. 2D). Interestingly, although major immune response related genes were downregulated in WS + TC, yet a diverse repertoire of B cell receptors (BCR; immunoglobulin light and heavy chain variables) were found upregulated demonstrating acute antibody response associated with severity. Upregulation of KIR3DL3, LILRB5 and IRAK1BP1 diverts innate immune pathways toward resolution of inflammation. Together, the differentially expressed gene profile suggests an immune and inflammatory response attenuation in the severe patients, at the time when platelet count reaches below normal threshold due to platelet dysfunction during dengue infection. The heatmaps capturing up and down-regulated genes are given in Supplementary Fig. S3.

3.4. WGCNA analysis reveals modules with functional consequences conversely associated with WS and –WS groups

Next, we wanted to investigate the transcriptome of Dengue patients to identify clusters of co-expressed genes (modules) and their relationships with clinical traits, distinguishing clinical features of dengue -WS and WS. We performed WGCNA, which groups thousands of genes into manageable modules, making it easier to study complex transcriptomic data. A soft-threshold power value of 30 was selected with the assistance of “pickSoftThreshold” function (Supplementary Fig. S4–A), whereas Supplementary Fig. S4–B depicts mean connectivity of soft threshold powers. These plots ensured robustness of the network and are essential for constructing biologically meaningful modules. A total of 14,071 genes across 112 patients were screened for allocation of modules during the WGCNA analysis; sample dendrogram given as Supplementary Fig. S5–A. WGCNA organized the genes into 22 modules across the mild, moderate and relative severe patients’ subgroups. Each module represented a group of genes with highly similar expression patterns, with size of module reflecting number of genes it contained. The distribution of 14,071 genes across 22 modules is shown in Supplementary Fig. S4–C with grey, turquoise and blue modules carrying more than 1000 genes. Supplementary Fig. S4–D represents collated heat map capturing differences in modules across severity groups as a whole while a heatmap across individual patient samples is given as Supplementary Fig. S5–B.

Two major clades comprising several modules with stark differential co-expression patterns are demonstrated by WS + TC (severe) which have been represented as clade 1 (red; higher expression) and clade 2 (blue; lower expression) in Supplementary Fig. S4–D. We functionally annotated the modules through pathway analysis with gene sets using Reactome database. Importantly, amongst the modules with higher expression in WS + TC from MEgrey to MElightyellow (clade 1), significant pathways depicted metabolism as a major process for MEcyan, whereas MEblue was concordant with glycosylation & nervous system with MEblack & MEmagenta for DNA repair mechanism. The modules from MEdarkred to MEpink (clade 2) showed decreased expression in WS + TC. Significant enriched pathways revealed distinct biological functions for each module such as blood coagulation, platelet activation, cell adhesion - MEdarkred; immune response, regulation of virus replication, interferon response, cytokine signaling - MEgreen/greenyellow/brown/red; cellular respiration, protein degradation - salmon/tan; and catabolic processes - MElight green. We regard these modules mentioned above as key modules. The other modules showed mixed processes highlighted in Fig. 3.

Fig. 3.

Key Findings from WGCNA Analysis.

Heatmap showing the correlation between the identified modules and clinical parameters, including total leukocyte count (TLC), platelets, neutrophils, and SGOT. Horizontal bars to the left annotate each module with pathways most enriched for the genes within that module.

We performed correlation analysis of module eigengene (ME) values with clinical parameters, which were altered in WS + TC patients i.e. TLC, platelets, neutrophils, lymphocytes (CBC) and SGOT (LFT) (Fig. 3). We considered Pearson coefficient ≥ ± 0.4 with significant p value for correlation. TLC, neutrophils and lymphocytes showed no significant correlation with any module. Modules comprising clade 2 showed significant positive correlation with platelets whilst only turquoise module from clade 1 showed negative correlation, reiterating the association of immune response modules with platelet dysregulation. With SGOT, only MEcyan, MElight cyan and MElight yellow modules from clade 1 showed significant negative correlation, all modules signifying metabolism as their major process.

3.5. Identification of module associated key genes demonstrates a diversified innate immune landscape regulated by platelets

We delved deeper to acquire genes from these modules that might represent biomarkers for WS group due to its association with low platelet count and high SGOT levels, both acting as differential clinical markers for severity. Identification of relevant genes was facilitated by linear regression analysis that validated correlation with clinical measures. Fig. 4A illustrates regression plots for genes that showed significantly positive correlation with platelets. Noteworthy genes from MEgreenyellow include genes associated with Interferon pathway, IFI44L, IFI35, KAT2B, OAS1 and SAMD9L. IFI44L is linked to DENV infection through its role in suppressing type I interferon response in severe cases. IFI35 plays dual roles (pro- or anti-) in viral replication and immune modulation, whereas OAS1 and HERC6 are essential in interferon responses during viral infections. Additionally, SAMD9L acts as a crucial antiviral factor and tumor suppressor in immune defense. The other set of genes belonged to platelet function/coagulation and adhesion; MEdarkred. ARHGEF12, RhoGEF signaling molecule playing pivotal role in platelet function and thrombosis, whilst actin-binding protein, Cortactin (CTTN), regulates endothelial barrier function. SELP (Selectin P) expressed during platelet activation mediates adhesion of leukocytes and monocytes to endothelial cells and platelets. MFAP3L (microfibril associated glycoprotein) induces ERK signaling showing diverse roles of tumorigenesis (cancer) and cytokine production (bacterial infections). Interestingly, this protein reportedly interacts with DENV capsid protein. PPBP (Pro-Platelet Basic Protein) aids in immune cell recruitment and activation. PRKAR2B and F13A1 are associated with blood coagulation processes, potentially influencing disease severity.

Fig. 4.

Association of Genes within Modules of Clade 2 with Platelets.

(A) Scatter plot representing significant positive correlation between the relative expression of genes and platelet count (per 10^9/L). The Spearman correlation test was employed, and the corr and p-values are annotated for each observed correlation. Scale of x axis is the raw count read transcripts for each gene plotted against the platelet counts (CBC clinical parameter) recorded for the patients with warning signs. (B) Depiction of the functional consequences of genes within the clade 2 modules. (C) Comparison of ΔCPM values for genes that differentiate between two platelet count groups: samples with platelet counts <100 × 10^9/L and those with counts between 100 and 140 × 10^9/L. ΔCPM (Delta Counts Per Million) was calculated as the difference in mean CPM values between the two groups, quantifying changes in transcript abundance. Asterisks indicate the statistical significance of differences (*p < 0.05, **p < 0.01, ***p < 0.001).

A set of subcellular protein palmitoylation and trafficking genes of Golgi and ER (clade 2 modules), GOLGA7, RAB27B, RAB2A, RAB6A and ZDHHC2 are supposedly required for viral replication and virus-induced cell death (autophagy). MEbrown captures genes for protein modification and immune signaling, wherein RNA binding protein genes (HNRNPA2B1 and RBM8A) demonstrates antiviral function. DNAJB6 and DNAJA1 are molecular chaperones assisting in DENV virion biogenesis. IFIH1 encodes MDA5 and TRAF3IP3, regulator for RIG-I-MAVS signaling responsible for innate immune response. Interestingly, EIF4E severely impairs translation of cellular protein sparing DENV protein synthesis. Together, positive correlation of low platelet count associated genes highlights a complex interplay of anti- and pro-viral host responses culminating into severe dengue pathophysiology (Fig. 4B).

Moreover, METurquoise was the only module showing significant negative correlation with platelet count. However, Spearman correlation analysis revealed different cellular processes like wound healing, energy metabolism, and antiviral regulation. Many genes (ATP5MJ, CASP4, COX7B, GPD2, ISCA1, and LPGAT1) had mitochondrial functions related to amelioration of DENV infection, implying an overall favorable association of genes of this module with protective host factors.

Regression analysis did not fetch significantly correlated genes with SGOT parameter, implying that minor increase in SGOT levels (∞ 100) does not substantially affect disease outcome. We further interrogated whether it was possible to dichotomize between WS + TC patients with severe thrombocytopenia (<100 platelet count 10^9/L) vs those with platelets falling between 100 and 140 counts 10^9/L. We took into account the baseline absolute transcript counts difference of the aforementioned genes and found significant CPM (counts per million) difference for 16 genes between the two groups, as depicted in Fig. 4C; Table 1. Notably, IFI44L gene had a remarkable ΔCPM of > 800 distinguishing the two groups.

Table 1.

Baseline expression (Δ CPM) for significant genes between the patients with <100 platelet counts 10^9/L and 100-140 counts 10^9/L.

Module	Gene	Spearman correlation	p-value	Δ CPM	p-value (Δ CPM)	Biological Pathway/Functional Category	Functional Role in Context of +WS + TC Disease Severity
Greenyellow	IFI44L	0.45	0.04	857	0.04	Interferon (IFN) Signaling	Suppresses type I interferon responses during acute DENV infection, serving as a prominent severity marker [22].
Greenyellow	OAS1	0.46	0.03	482	0.03	Interferon (IFN) Response	Key mediator of the host antiviral interferon response during infection [14].
Darkred	F13A1	0.48	0.03	447	0.03	Blood Coagulation	Coagulation factor subunit; influences vascular/hemostatic complications during severe outcome.
Brown	IFIH1	0.48	0.03	426	0.03	Innate Immunity/Signaling	Encodes MDA5; regulates RIG-I-MAVS signaling pathways for downstream innate antiviral defense.
Brown	HNRNPA2B1	0.52	0.02	415	0.05	Immune Signaling/RNA Binding	RNA-binding protein involved in host antiviral responses.
Darkred	PPBP	0.46	0.04	277	0.03	Platelet Function/Immunity	Drives immune cell recruitment and leukocyte activation.
Greenyellow	SAMD9L	0.48	0.03	238	0.02	Interferon (IFN) Pathway	Crucial antiviral factor in host immune defense.
Greenyellow	IFI35	0.50	0.02	193	0.02	Interferon (IFN) Pathway	Dual roles in viral replication control and immune modulation.
Brown	DNAJB6	0.50	0.02	182	0.04	Chaperone/Virion Biogenesis	Assists DENV virion biogenesis.
Green	GOLGA7	0.50	0.02	113	0.03	Membrane Trafficking	Golgi protein involved in vesicular transport and viral replication.
Greenyellow	KAT2B	0.48	0.03	93	0.05	Transcription Regulation	Histone acetyltransferase regulating inflammatory gene transcription.
Darkred	PRKAR2B	0.50	0.02	89	0.04	Blood Coagulation	PKA regulatory subunit involved in endothelial stability.
Darkred	CTTN	0.60	0.004	71	0.03	Cytoskeletal Structure/Actin	Regulates endothelial barrier stability and vascular permeability.
Darkred	MFAP3L	0.46	0.04	46	0.03	Cell Adhesion/Signaling	Interacts with DENV capsid and alters endothelial barrier integrity.
Darkred	SELP	0.53	0.01	44	0.03	Platelet Activation/Adhesion	Drives leukocyte-platelet adhesion during inflammation.
Darkred	RAB27B	0.51	0.02	41	0.05	Vesicular Trafficking	Regulates platelet exocytosis and viral release dynamics.

4. Discussion

Despite Dengue infection being a major concern across tropical and subtropical regions, multifactorial clinical presentation complicates its outcome. Dengue infection leads to a complex interplay of viral and host factors wherein the host immunity regulation involves a delicate balance between pro- and anti-viral responses, consequently tipping off the outcome towards recovery or severity. Although initial clinical parameters can provide a window to assess onset of severe manifestations, yet risk factors for severity are inconsistent amongst Dengue patients [23,24].

This study attempted to capture the subtle/elusive differences in host transcriptome features that sets in when the clinical parameters show signs of derangement early during infection (acute phase). Importantly, leading to identification of potential clinical biomarkers and ongoing biological processes that may contribute to efficient diagnosis and clinical management of severe cases. We have tabulated findings of genes and associated pathways of the present study in Supplementary Table 3.

Extensive studies in Dengue disease have brought to table, a core set of immune response signatures highlighted by interferon-stimulated genes, cytokines, and chemokines, revealing conserved immune pathways that may be universal to severe viral infections [[25], [26], [27]]. In this context, we examined DENV-infected patients without warning signs, mild and moderate (-WS-LP; WS + LP) and with warning signs, severe (WS + TC). It was important to note that the feature that distinguished the groups without warning signs i.e. mild and moderate, was the presence of chemokines, activation of neutrophil degranulation and regulation of inflammation in mild patients compared to moderate. Unlike prior studies, we observed that -WS-LP patients exhibited robust chemokine activity, particularly with activated neutrophil degranulation and inflammation regulation, suggesting a strong but controlled immune response to curb infection. In moderate cases, lower CD177 and CXCL2 gene expression, as well as diminished neutrophil activation, implied a shift toward infection progression, wherein the inflammatory response plays a critical role in the outcome of dengue infection [28]. This aligns with recent findings showing that lower neutrophil activation is associated with poor immune response [29].

Our findings in WS + TC cases underscore a depressed innate immune and inflammatory state, with observed downregulation of cytokines and interleukins (CXCL8, CXCL16, CXCL2 and IL1B), the drivers of inflammatory cascade critical in endothelial and vascular dysfunction [30]. The non-inflammatory state of WS patients complicates the understanding of dengue pathogenesis associated with severe infections. Yet, studies have identified non-inflammatory pathways for endothelial disruption in dengue infections, with NS1 protein primarily disrupting the endothelial glycocalyx causing plasma leakage [31]. This was accompanied by dysregulated platelet function, a hallmark of thrombocytopenia observed in our WS + TC patients.

Additionally, we identified a set of upregulated genes in severe patients related to biological processes of glycosylation, nervous system and metabolism, marking a clear delineation from -WS patients. Global studies have indicated that endothelial glycocalyx disruption in severe dengue can occur through O-glycosylation alterations independent of inflammation, with altered glycosaminoglycan structures affecting vascular permeability [32,33]. Different glycosaminoglycan chains within the endothelial glycocalyx form proteoglycans via N- or O-linked protein glycosylation attachments [34]. Here, we observed upregulation of genes associated with O-linked glycosylation (GALNT13, B4GALT1, B3GNT4, and MUC12/4) in WS + TC patients. In concordance with these reports, our results suggest that these glycosylation pathways may impact endothelial integrity [31], warranting further investigation into their potential as biomarkers for vascular leakage.

WGCNA association with initial clinical parameters helped us to elusively segregate processes and identify key genes that might collectively alter disease state in WS patients. Specifically, low platelet counts were strongly associated with shifts in systemic immune response, particularly in pathways related to interferon signaling, innate immunity, and inflammation, indicating a profound role of thrombocytopenia in driving disease progression. In contrast, processes of DNA repair, metabolic activity, neuronal functions and glycosylation did not adhere to low platelets. Moreover, our study demonstrates absence of significant correlation of TLC (leukopenia) with any of WGCNA modules, although early decline and recovery of TLC is reportedly associated with onset of the recovery phase in Dengue [35].

Through Linear regression analysis, we identified a set of genes with varying biological effects that cumulatively defined the course of infection. Our study identified a set of genes ARHGEF12, MFAP3L, RUFY1, wherein these genes or their homologues have reported interaction with DENV proteins. Additionally, DENV infection promotes anticoagulant molecule expression on endothelial cells, contributing to a hemorrhagic tendency in patients by reducing vascular integrity [36,37]. Similarly, several factors CTTN, MFAP3L, SELP, PPBP, PRKAR2B and F13A1 might be implicated in altering endothelial integrity and barrier function in WS patients potentially driving progression towards severe disease.

WGCNA, similar to DEG analysis, identified glycosylation as a key process within Meblue module, independent of platelet count association. This emphasizes glycosylation as a critical and ongoing biological pathway in DENV pathogenesis. The glycocalyx of endothelium controls vascular permeability, prevents thrombosis and adhesion of leukocytes [34,38]. Further, disrupted glycocalyx due to DENV NS1 antigen can cause vascular leak independent of inflammatory cytokines [36]. To this effect, the non-inflammatory state observed in our dataset supports the disruption of endothelial glycocalyx functions. Since transcriptomic profiling was performed using whole-blood RNA, the observed differential expression patterns may partly reflect changes in circulating immune cell composition in addition to the transcriptional alterations occurring within the individual cell types. Although complete blood count parameters were incorporated during clinical interpretation and WGCNA identified biologically coherent modules associated with platelet counts, future studies employing single-cell transcriptomics or computational cell deconvolution approaches would further resolve cell type-specific contributions to the observed molecular signatures.

Summing together, our findings show notable concordance with dengue patients transcriptomic studies conducted in Brazil, providing cross-regional support for the central role of interferon-stimulated genes (ISGs) in dengue severity. Usuda et al. (2023) identified top-ranking ISGs such as IFI27 and ISG15 as biomarkers capable of differentiating dengue hemorrhagic fever (DHF) from dengue fever (DF), particularly during late acute phase [22]. While our dataset did not capture IFI27, we observed consistent upregulation of ISG15 in patients with warning signs and low platelet counts, aligning with its role as a potential disease severity marker. Interestingly, our study also detected CTTN, gene associated with endothelial function, which complements the immune-endothelial axis that was not a focus in the Brazilian datasets, suggesting potential population-specific elements worthy of further investigation. Together, these comparisons reinforce the biological relevance of our findings and demonstrate convergence of dengue-associated immune responses between Indian and Brazilian patients while also highlighting need for broader multi-cohort validation to understand regional variations.

By integrating these distinct biological phenomena observed in our dataset, a clearer picture of the sequential pathogenic cascade in WS + TC patients emerges: immune suppression leading to thrombocytopenia and subsequent vascular leakage. Upstream, a sub-optimal or suppressed early interferon and innate immune response diminishes the host's early defenses, creating a permissive environment for viral replication and high antigenemia [39]. Downstream, this heightened viral stress drives platelet dysfunction and clearance, leading to the observed thrombocytopenia [40]. Concurrently, independent of traditional cytokine-mediated hyper-inflammation, this environment triggers the upregulation of O-glycosylation pathways and the downregulation of structural maintenance genes like CTTN, ultimately altering the endothelial glycocalyx matrix [41], and causing the vascular permeability issues that characterize the onset of clinical warning signs [42].

5. Conclusion

This study presents the early host transcriptomic response to dengue infection. The transcriptomic modules identified through WGCNA analysis reveal that low platelet counts in patients with warning sign are indicative of depressed INF state, which might support viral persistence, exacerbating infection. Consequently, a sub-optimal inflammatory and altered coagulation landscape builds up in the WS patients, which might change infection trajectory towards severity. Our findings also suggest that endothelial glycocalyx disruption is a central, non-inflammatory mechanism contributing to vascular leakage in dengue. The endothelial glycocalyx disruption, mediated through altered O-glycosylation pathways, may drive vascular leakage in dengue, independent of inflammatory cytokines. Monitoring glycocalyx-related gene signatures could enable early identification of patients at risk for severe disease, while therapeutic strategies aimed at preserving glycocalyx integrity may represent a novel approach to reduce vascular complications. Together, glycosylation pathways can act as key molecular biomarkers and therapeutic targets for predicting and mitigating severe dengue-associated vascular leakage. In conclusion, this study delineates a distinct, interconnected molecular cascade extending from early immune evasion to platelet depletion and structural vascular compromise in dengue patients presenting with warning signs.

6. Limitations of the study

This study has the inherent limitation of a cross-sectional design, as blood samples were collected at a single time point during the acute phase of dengue infection. Consequently, the temporal dynamics of gene expression throughout disease progression and recovery could not be assessed. Future longitudinal studies with serial sampling across different stages of infection are warranted to better understand the dynamic molecular changes associated with dengue pathogenesis.

CRediT authorship contribution statement

Uzma Shamim: Writing – original draft, Visualization, Supervision, Methodology, Investigation, Conceptualization. Smriti Arora: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis. Ranjeet Maurya: Writing – original draft, Visualization, Software, Methodology, Investigation, Formal analysis. Jyoti Soni: Writing – review & editing, Visualization, Resources, Investigation. Richa Shukla: Writing – review & editing, Visualization, Resources, Investigation. Priyanka Mehta: Writing – review & editing, Methodology, Investigation, Formal analysis. Bansidhar Tarai: Writing – review & editing, Resources. Sandeep Budhiraja: Writing – review & editing, Resources. Rajesh Pandey: Writing – review & editing, Supervision, Resources, Project administration, Investigation, Funding acquisition, Conceptualization.

Data availability statement

Sequence data used in this study are publicly available in NCBI SRA under the accession number with the BioProject number PRJNA1071729.

Ethics statement

This study was reviewed and approved by CSIR-IGIB's Human Ethics Committee Clearance with the approval number: CSIR-IGIB/IHEC/2020-21/01, dated 21-01-2020. The patients/participants provided their written informed consent to participate in this study.

Funding

Authors acknowledge the funding support from Bill and Melinda Gates Foundation (BMGF), Grant number - INV-033578 and Rockefeller Foundation, grant number-2021 HTH 018, to R Pandey.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

The following are the Supplementary data to this article.

figs1.

figs2.

figs3.

figs4.

figs5.