Unraveling potential gene biomarkers for dengue infection through RNA sequencing

Abstract

Dengue virus hijacks host cell mechanisms and immune responses in order to replicate efficiently. The interaction between the host and the virus affects the host's gene expression, which remains largely unexplored. This pilot study aimed to profile the host transcriptome as a potential strategy for identifying specific biomarkers for dengue prediction and detection. High-throughput RNA sequencing (RNA-seq) was employed to generate host transcriptome profiles in 16 dengue patients and 10 healthy controls. Differentially expressed genes (DEGs) were identified in patients with severe dengue and those with dengue with warning signs compared to healthy individuals. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses were performed to elucidate the functions of upregulated and downregulated genes. Compared to healthy controls, 6466 genes were significantly differentially expressed (p < 0.05) in the dengue with warning signs group and 3082 genes in the severe dengue group, with over half being upregulated. The major KEGG pathways implicated included transport and catabolism (14.4%–16.3%), signal transduction (6.6%–7.3%), global and overview maps (6.7%–7.1%), viral diseases (4.6%–4.8%), and the immune system (4.4%–4.6%). Several genes exhibited consistent and significant upregulation across all dengue patients, regardless of severity: Interferon alpha inducible protein 27 (IFI27), Potassium Channel Tetramerization Domain Containing 14 (KCTD14), Syndecan 1 (SDC1), DCC netrin 1 receptor (DCC), Ubiquitin C-terminal hydrolase L1 (UCHL1), Marginal zone B and B1 cell-specific protein (MZB1), Nestin (NES), C–C motif chemokine ligand 2 (CCL2), TNF receptor superfamily member 17 (TNFSF17), and TNF receptor superfamily member 13B (TNFRSF13B). Further analysis revealed potential biomarkers for severe dengue prediction, including TNF superfamily member 15 (TNFSF15), Plasminogen Activator Inhibitor-2 (SERPINB2), motif chemokine ligand 7 (CCL7), aconitate decarboxylase 1 (ACOD1), Metallothionein 1G (MT1G), and Myosin Light Chain Kinase (MYLK2), which were expressed 3.5 times, 2.9 times, 2.3 times, 2.1 times, 1.7 times, and 1.4 times greater, respectively, than dengue patients exhibiting warning signs. The identification of these host biomarkers through RNA-sequencing holds promising implications and potential to augment existing dengue detection algorithms, contributing significantly to improved diagnostic and prognostic capabilities.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11262-024-02114-2.

Introduction

Dengue virus (DENV) infection is a significant public health concern, with an estimated 400 million cases occurring annually worldwide [2]. While most individuals experience mild, self-limiting dengue fever, approximately 15–20% develop severe dengue, a life-threatening complication characterized by plasma leakage, organ dysfunction, and hemorrhage [18, 40]. Despite the availability of supportive care and prompt hospitalization, severe dengue remains a major cause of morbidity and mortality, particularly in endemic regions [41].

Currently, there are no reliable prognostic markers to predict which individuals will progress to severe dengue. The World Health Organization (WHO) criteria for dengue classification, established in 2009, distinguish between dengue with or without warning signs and severe dengue. However, the clinical warning signs used to identify patients at risk of developing severe dengue often emerge late in the disease course, limiting their effectiveness [1, 34]. This delay in identification can lead to inefficient patient management, inadequate resource allocation, and continued morbidity and mortality.

The development of novel biomarkers for early detection and prediction of severe dengue is crucial to improve patient outcomes. Recent studies have investigated the potential of host biomarkers derived from publicly accessible datasets using microarray-based techniques and machine learning approaches [22, 29]. Furthermore, advancements in gene multiplexing technologies, such as RT-PCR [21] and semiconductor biochip-based nucleic acid amplification [10], offer promising avenues for cost-effective biomarker identification and validation.

Characterizing the host transcriptome through genome-wide gene expression analysis has become essential in understanding disease pathogenesis. Historically this was done using microarrays [20] but currently ‘next-generation RNA sequencing’ (RNA-seq) primarily utilizing the Illumina sequencing technology has become the method of choice for transcriptome studies [31, 45, 47]. Additionally, customized or existing ELISA gene panels are employed for the identification of specific gene subsets.

The combination of the NS1, IgM, and IgG biomarkers is currently the only reliable biomarker identified to be beneficial for the diagnosis of dengue. However, challenges remain, including false negatives and limited sensitivity of detection when employing these biomarkers [7, 9, 24]. This highlights the urgent need for continued research to identify novel and robust dengue biomarkers. Therefore, we initiated a pilot study that aimed to characterize the host transcriptome profile in dengue infection using RNA sequencing and identify differentially expressed genes that could serve as biomarkers for predicting both dengue infection and progression to severe dengue.

Methods

Study population and sample collection

This study included adult dengue patients admitted to Hospital Kuala Lumpur, Malaysia, between January 2020 and January 2022. A total of 30 participants were enrolled: 20 with dengue fever and warning signs and 10 with severe dengue. All patients were confirmed NS1-positive by rapid antigen combo test (SD Diagnostics) and fulfilled the case classification criteria for dengue with warning signs or severe dengue at enrollment. The exclusion criteria include (i) diagnosed co-infections (e.g., leptospirosis, HIV, COVID-19, and other arboviruses); (ii) dengue NS1 and IgM negativity; iii) underlying co-morbidities related to blood disorders (e.g., anemia, ITP, liver disease, cancer, thalassemia and; iv) lack of informed consent. Additionally, 10 healthy control volunteers from Universiti Kebangsaan Malaysia Medical Centre were included. These volunteers were confirmed negative for dengue infection and COVID-19 at the time of recruitment and had no history of prior dengue infection.

Seven milliliters (7 ml) of human whole blood were collected in sodium citrate vacutainer (5 ml) and serum separator vacutainer (2 ml) for dengue infected patients whereas 5 ml of whole blood were withdrawn from healthy controls in sodium citrate tubes only. Blood collected in the sodium citrate vacutainer was intended for RNA-seq whereas serum obtained from the serum separator tube was used for dengue virus serotyping.

Dengue virus serotyping

Viral RNA was extracted from patient serum using a QIAamp Mini Viral RNA Extraction kit (Qiagen, USA). Serotyping was performed using a multiplex Taqman Real-Time RT-PCR assay [16]. A positive result was defined as a Ct value ≤ 36, and serotype identification was based on sigmoidal amplification in the respective channels (FAM = DENV1, HEX = DENV2, TEXAS RED = DENV3, CY5 = DENV4).

PBMC isolation and total RNA extraction

Peripheral blood mononuclear cells (PBMCs) were isolated from blood samples using Ficoll-Paque density gradient centrifugation (GE Healthcare) as outlined by Boyum [3] with some modifications. A 6 ml of each patient blood was diluted with equal volume of Phosphate Buffered Saline (PBS). Next, 8 ml of Ficoll-Paque (GE Healthcare) which has been brought to room temperature was pipetted carefully to the diluted blood without mixing the solution and subjected to centrifugation for 20 min at 400 g using the “soft” mode and inactivated brake. The separated PBMC layer was transferred to a new centrifuge tube and resuspended with 6 ml of PBS. The mixture was centrifuged at 500 g for 15 min with the resulting pellet was washed with 8 ml of PBS and centrifuged at 100 g for 10 min. Total RNA was extracted using the innuPREP RNA mini kit 2.0 (Analytik Jena) according to the manufacturers’ instructions.

Transcriptome profiling based on RNA-seq

The purified RNA samples were quantified by RNA High-Sensitivity Assay kits (Thermo Fisher) with a Qubit 3.0 fluorometer (Invitrogen). Enrichment of mRNA was performed using Dynabeads mrRNA Purification Kit (Invitrogen), and library preparation was carried out using MGIEasy RNA Library Prep Set (MGI). We assessed the libraries for quantity and quality with Qubit™ 1X dsDNA High Sensitivity (HS) kit (Invitrogen), Bioanalyzer (Agilent Technologies), and KAPA Library Quantification Kits (Kapa Biosystems) and sequenced on the DNBSEQ-G400 platform (MGI).

Gene ontology and pathway enrichment analysis

The raw data from RNA-seq were filtered to remove low-quality reads. Clean reads stored in FASTQ format. HISAT was used to map clean reads to Homo sapiens reference genome (GCF_000001405.38_GRCh38.p1) and Bowtie2 [19] to reference gene. Genes expression level is quantified by a software package called RSEM and FPKM method. Genes with a |log2 fold change|≥ 1 and q value ≤ 0.05 were selected for analysis, and differential expression analysis of the transcripts was performed by employing the NOISeq method [38]. Clustering analysis of differentially expressed genes (DEGs) was performed with cluster [6, 8] and javaTreeview software [30] according to the cluster groupings for DEGs. For functional classification, the DEGs were analyzed using the Gene Ontology (GO, an international standardized system for a functional classification of genes) database (http://www.geneontology.org/) and WEGO software [46]. The genes were classified with the GO analysis for cellular components, biological processes, and molecular functions. Pathway enrichment analysis of DEGs was carried out using the Kyoto Encyclopedia of Genes and Genomes (KEGG) database (https://www.genome.jp/kegg). It is considered that there is a significant enrichment for GO and KEGG analysis when the P value is less than 0.05. Final analysis of DEGs was done in comparison of dengue with warning signs and severe dengue groups against healthy controls. The bioinformatics pipeline used in the study is summarized in Fig. 1.

Fig. 1.

Summary of pipeline for RNA sequencing

Results

Characteristics of study population

The study recruited 30 dengue patients and 10 healthy control volunteers. Table 1 presents the demographic and laboratory characteristics of the study participants.

Table 1.

Demographic and laboratory features of the recruited study population

	Healthy control n = 10	Dengue with warning sign n = 20	Severe dengue n = 10
Age, median years (IQR)	31 (23–35)	23 (18–66)	31 (21–52)
Days of fever, median days (IQR)	NA	4 (3–6)	6 (4–6)
Platelet count (× 109/L) Mean ± SD	232 ± 30.5	100.7 ± 33.5	97.9 ± 47.0
White blood cell (× 109/L) Mean ± SD	NT	3.3 ± 1.1	4.3 ± 1.7
Hematocrit level (%) Mean ± SD	NT	40.7 ± 3.6	42.0 ± 2.9
AST (IU/ml) Mean ± SD	NT	81.7 ± 52.9	82 ± 53.1
ALT (IU/ml) Mean ± SD	NT	71.0 ± 60.4	51.4 ± 30.1
Dengue serotype n (%)	NA	DENV1 = 5 (25.0) DENV2 = 4 (20.0) DENV3 = 3 (15.0) DENV4 = 8 (40.0)	DENV1 = 2 (20.0) DENV2 = 3 (30.0) DENV3 = 2 (20.0) Untypeable = 3 (30.0)

NA not applicable, NT not tested

Transcriptome sequence annotation

Transcriptome sequencing was successfully performed on 10 healthy controls, 13 dengue patients with warning signs, and 3 severe dengue cases. Table 2 summarizes the sequenced dengue cases. Sequencing was unsuccessful for the remaining samples due to inadequate RNA or library quality. An average of 45 million reads per sample was generated. The average mapping ratio to the reference genome and genes was 94.06% and 70.44%, respectively. Two samples, PTB4-21 and PTB12-21, exhibited notably lower mapping gene ratio (%) as compared to other samples. This discrepancy can occur for several reasons. First, RNA composition differences wherein samples may contain a higher proportion of non-coding RNA which could contribute more reads to map to genome but outside the annotated gene regions. Second, sample-specific biological factors and differences in transcriptional activity such as higher intronic read retention or different isoforms can lead to lower gene-mapping ratio. While the two samples had a lower total mapping gene ratio, they exhibited high overall read quality. The clean reads and total genome mapping are consistent indicating that the sequencing and alignment were performed correctly and efficiently. The sequencing depth for these two samples is sufficient for downstream analysis. Moreover, there was consistency with quality control metrics. If the standard quality control such as base quality, alignment rates, and duplication rates do not show significant problems with these samples, it suggests that they are technically sound. Table 3 provides a detailed overview of the mapped reads for each sample.

Table 2.

Demographic and laboratory features of the sequenced dengue cases in this study

	Dengue with warning sign n = 13	Severe dengue n = 3
Age, median years (IQR)	23 (18–33)	34
Days of fever, median days (IQR)	5 (4–5.5)	6
Platelet count (× 109/L) Mean ± SD	94.6 ± 36.8	52.5 ± 32.5
White blood cell (× 109/L) Mean ± SD	3.5 ± 1.1	4.7 ± 2.6
Hematocrit level (%) Mean ± SD	40.8 ± 8.8	39.5 ± 2.5
AST (IU/ml)	79.4 ± 64.1	82.5 ± 50.1
ALT (IU/ml)	59.9 ± 62.5	51.5 ± 19.5
Dengue serotype n (%)	DENV1 = 3 (23.0) DENV2 = 3 (23.0) DENV3 = 1 (7.7) DENV4 = 6 (46.2)	DENV2 = 3 (100.0)

Table 3.

Alignment results of clean reads mapped to reference genome and genes

No	Sample identity	Grouping	Mapping to reference genome		Mapping to reference genes
			Total clean reads (M)	Total mapping genome ratio (%)	Total clean reads (M)	Total mapping gene ratio (%)
1	HCB1-21	Healthy controls	44.75	96.15	44.75	79.71
2	HCB2-21		45.18	96.00	45.18	77.28
3	HCB3-21		43.91	95.87	43.91	78.37
4	HCB5-21		44.53	95.61	44.53	76.50
5	HCB6-21		45.58	95.35	45.58	72.14
6	HCB7-21		45.45	95.21	45.45	72.47
7	HCB8-21		44.35	95.82	44.35	76.75
8	HCB9-21		44.65	95.67	44.65	76.23
9	HCB10-21		44.47	95.58	44.47	74.42
10	HCB11-21		44.24	95.74	44.24	73.66
11	PTB1-21	Severe dengue	44.40	92.64	44.40	75.99
12	PTB9-21		45.57	92.95	45.57	65.07
13	PTB12-21		44.47	94.71	44.47	45.03
14 15	PTB2-21	Dengue fever with warning signs	44.24	95.66	44.24	80.97
15	PTB3-21		45.26	91.57	45.26	69.85
16	PTB4-21		45.59	87.00	45.59	35.16
17	PTB5-21		45.58	90.96	45.58	68.14
18	PTB6-21		45.61	95.76	45.61	79.30
19	PTB7-21		45.76	85.47	45.76	59.52
20	PTB8-21		45.71	95.58	45.71	81.65
21	PTB10-21		45.13	94.54	45.13	67.41
22	PTB11-21		44.35	88.32	44.35	58.20
23	PTB13-21		44.99	92.80	44.99	64.75
24	PTB1-22		49.63	95.81	49.63	83.96
25	PTB2-22		49.55	94.08	49.55	75.38
26	PTB3-22		49.51	90.53	49.51	70.45

Differentially expressed genes in dengue groups vs healthy control

Differential gene expression analysis (DEG) revealed significant changes in gene expression (p < 0.05) between dengue patients and healthy controls. The dengue with warning signs group displayed 6466 differentially expressed genes, while the severe dengue group showed 3082 differentially expressed genes (Table 4). Notably, the proportion of upregulated genes was higher than downregulated genes in both groups. The distribution of gene expression levels in each sample is illustrated in boxplot and scatter plot visualizations (Figs. 2, 3).

Table 4.

Number of significant DEGs identified between groups vs healthy control

Group	Number of DEG	Number of upregulated genes	Number of downregulated genes
Dengue with warning sign vs healthy control	6466	3538	2928
Severe dengue vs healthy control	3082	1888	1194

Fig. 2.

Boxplot showing distribution of gene expression levels of each sample sequenced in this study

Fig. 3.

Scatter plot showing DEG of each group in comparison with healthy controls

GO Enrichment analysis

To understand the functional implications of the differentially expressed genes, Gene Ontology (GO) enrichment analysis was performed. A lower enrichment Q value (corrected p value) indicates greater statistical significance. Figure 4 and Fig. 5 present the top 20 enriched GO terms categorized into cellular component, molecular function, and biological process. Host genes were highly expressed in cellular components such the cytoplasm, cytosol, and membrane in the dengue with warning sign and severe dengue groups. A high number of genes were involved in protein binding, cell cycle, cellular response to DNA damage, phosphorylation, protein transport, apoptosis, and viral process among patients with warning signs. In addition to these functions, a large number of genes were also associated with cell division and oxidation–reduction process in severe dengue.

Fig. 4.

Top 20 GO Enrichment data in relation to cellular component, molecular function and biological process for dengue with warning sign vs healthy control group

Fig. 5.

Top 20 GO Enrichment data in relation to cellular component, molecular function, and biological process for severe dengue vs healthy control group

KEGG pathway analysis

KEGG pathway analysis revealed a systematic gene function in terms of the networks of genes and molecules. In this study, KEGG analysis derived both pathway classification and enrichment. The dengue with warning sign and severe dengue groups demonstrated similar pattern of KEGG pathways whereby highly expressed DEGs are mainly focused in the pathways involving transport and catabolism (14.4%–16.3%), signal transduction (6.6%–7.3%), global and overview maps (6.7%–7.1), viral diseases (4.6%–4.8%), and immune system (4.4%–4.6%). In the KEGG enrichment analysis, the top 20 most significantly expressed genes in dengue infection were associated with metabolic pathways (Fig. 6).

Fig. 6.

Annotated classification and enrichment of KEGG Pathway. Q value was obtained by correction of p value

Host genes associated with dengue infection

The host transcriptome profile obtained from this study can be described in many ways. First, we selected 20 topmost upregulated gene sets in dengue patients with warning signs and severe dengue based on the fold change ranking, regardless of their functional groups (Supplementary Table 1). Next, we shortlisted top 20 upregulated genes that are associated with two crucial events in dengue infection, which are immune response (T cell-associated, B cell-associated, chemokines, interleukins, signaling receptors) and platelet activation pathways. The upregulated genes were the focus of the study as these have positive measurable values to be used as dengue biomarkers.

Among the top 20 upregulated genes, regardless of functional groups, Interferon-alpha inducible protein 27 (IFI27), Potassium Channel Tetramerization Domain Containing 14 (KCTD14), Syndecan 1 (SDC1), DCC netrin 1 receptor (DCC), Ubiquitin C-terminal hydrolase L1 (UCHL1), Marginal zone B and B1 cell-specific protein (MZB1), and Nestin (NES) were highly expressed at a constant fold change (ranging between > 5 to 8-fold) in both dengue with warning signs and severe dengue groups. Notably, the gene ATP5MF-PTCD1 readthrough exhibited a remarkable fold change of 23.87 and was specific to the severe dengue group.

Within the DEGs specifically associated with immune response, various chemokines and interleukins were found abundantly expressed in both groups (Supplementary Table 2). However, both dengue patients with warning signs and severe dengue showed high and nearly similar expression levels (ranging between > 2 to 8-fold) of IFI27, SDC1, C–C motif chemokine ligand 2 (CCL2), TNF receptor superfamily member 17 (TNFSF17), and TNF receptor superfamily member 13B (TNFRSF13B).

Among the DEGs associated with platelet activation pathway, platelet-related genes that are highly expressed in dengue with warning signs include fibronectin 1 (FN1) and serpin family G member 1 (SERPING 1) while coagulation factors such as F2RL2 and F3 were specific for severe dengue (Supplementary Table 3).

Certain genes showed significantly higher expression in severe dengue compared to patients with warning signs, potentially serving as biomarkers to predict progression to severe dengue. These genes include C–C motif chemokine ligand 7 (CCL7) (fold change 4.5 vs 10.5), Aconitate decarboxylase 1 (ACOD1) (fold change 3.9 vs 8.2), Metallothionein 1G (MT1G) (fold change 3.6 vs 6.0), TNF superfamily member 15 (TNFSF15) (fold change 1.3 vs 4.6), Myosin light chain kinase 2 (MYLK2) (fold change 3.7 vs 5.0), and Serpin family B member 2 (SERPINB2) (fold change 1.7 vs 4.9).

Discussion

The identification and analysis of dengue gene biomarkers have gained significant traction in recent years. Several studies have explored the potential of these biomarkers to predict disease severity and guide clinical management. One such study that produced notable results was Robinson et al. [29] that employed pre-existing gene expression datasets from five nations to create a predictive algorithm through an in silico technique. High-throughput microfluidic qRT-PCR was then used to confirm the 20 genes that were chosen in a Colombian cohort. These genes have strong prediction power for severe dengue and were found to be considerably downregulated in NK and NK-T cells. In a similar vein, another study analyzed pre-existing microarray dataset to deduce the validity of gene biomarkers [43] and found ten hub genes to be statistically significant in dengue infection, including MX1, IFI44L, IFI27, ISG15, IFI35, OAS3, OAS2, OAS1, IFI6, and USP18. A meta-analysis was carried out using dengue gene expression data obtained from Gene Expression Omnibus repository and discovered genes such as CCNB1, CCNB2, IL12A, CXCR3, TNFSF13B, IFI27, TNFRSF17, STAT, STAT2, and TLR7 to be signatures to dengue fever [17].

Our study, conducted in a Malaysian cohort, contributes to this growing body of research by generating original transcriptome data and identifying a unique set of genes significantly upregulated in dengue and severe dengue. This unique dataset provides valuable insights into the specific genetic responses associated with dengue infection in a Malaysian population. Through RNA sequencing, we identified a comprehensive set of host genes that characterize dengue infection. Due to the extensive data generated, we prioritized analysis of the top 20 gene sets, initially focusing on those consistently associated with dengue infection regardless of their functional category. We then delved deeper into genes involved in immune response and platelet activation pathways. Our analysis revealed seven genes – IFI27, KCTD14, SDC1, DCC, UCHL1, MZB, and NES – with potential utility as biomarkers for dengue detection across all disease phases. These genes were consistently upregulated in both groups analyzed, suggesting their potential for universal detection of dengue infection.

IFI27, exhibiting a significant increase in both the dengue warning sign and severe dengue groups, is a mitochondrial protein belonging to the interferon family. It plays a crucial role in apoptosis by disrupting mitochondrial function and is known to be secreted during innate immune responses, contributing to the regulation of viral life cycles. Multiple studies have consistently demonstrated the upregulation of IFI27 in dengue fever, across various measurement platforms. Furthermore, a negative correlation between IFI27 expression and disease severity suggests its potential as a prognostic marker. However, the reliability and specificity of IFI27 as a dengue-specific biomarker remain questionable. This gene has also been reported in a range of other diseases, including pancreatic cancers [12], hepatocellular carcinoma [4], Influenza A/(H1N1)pdm09 [37], COVID-19 [32], and Enterovirus-71 infected hand, foot, and mouth disease [25]. Interestingly, a recent study [15] demonstrated that IFI27 secretion can inhibit dengue infection, suggesting its potential as a therapeutic target rather than a diagnostic or prognostic marker.

KCTD14, a member of the KCTD family of proteins, stands out as a particularly intriguing finding in our study. Belonging to group H of this family, KCTD14 is known to play a role in protein homooligomerization. Unlike IFI27, KCTD14 has not been reported in other viral infections, making it a promising candidate for a dengue-specific biomarker. While other KCTD proteins have been linked to conditions like neurodegenerative and neuropsychiatric disorders (KCTD3 & 16), obesity and eating disorders (KCTD15), and oncogenesis (KCTD 12 & 16), KCTD14 remains relatively understudied (Angrisani et al., 2021). The specific role of KCTD14 in cancer development has not yet been documented. A previous study [39] analyzed five publicly available datasets generated by array-based transcription profiling, finding KCTD14 elevated in both non-severe and severe dengue cases. Our findings further corroborate these results, demonstrating an even greater fold change in KCTD14 expression in our study.

Syndecans are proteoglycans that interact with a range of growth factors, including VEGF, FGF, and TGF-β, activating these signaling pathways. They also play a critical role in cell adhesion, both cell-to-cell and cell-to-extracellular matrix. SDC1, specifically, is predominantly expressed in epithelial and plasma cells [27]. Previous research has demonstrated that SDC1 is released by activated endothelial cells and shows a steady increase in serum levels during the defervescence stage of dengue, but not significantly during the febrile stage (admission) [23]. A study by Suwarto et al. [36] linked SDC1 to plasma leakage in severe dengue. While our study identified SDC1 as one of the highly expressed genes in dengue infection, its specificity as a biomarker for severe dengue remains unclear. We observed consistent elevation of SDC1 expression even during the non-severe phase, suggesting a potential limitation in its ability to differentiate between disease severities.

DCC, NES, MZB1, and UCHL1 emerged as promising gene biomarkers with potential predictive value for dengue infection. DCC, a transmembrane protein with immunoglobulin-like and fibronectin type II-like repeats, is involved in netrin-1-mediated nervous system development. NES, encoding a member of the intermediate filament protein family, is primarily expressed in nerve cells. The elevated expression of these two genes in our study suggests a potential link to dengue encephalitis, a serious complication of dengue infection. Notably, to our knowledge, there is no prior evidence describing the expression of these genes in dengue infection, highlighting the need for further exploration of their utility in dengue diagnosis or prognosis.

MZB1, a gene involved in positive regulation of cell population proliferation and located in the cytoplasm and extracellular region, has been widely studied in other diseases, including multiple myeloma [5], systemic lupus erythematosus (SLE) [26], and acute pancreatitis [44]. UCHL1 encodes a peptidase C12 family member and is specifically expressed in neurons and cells of the diffuse neuroendocrine system. Its expression has been documented in viral infections such as Epstein-Barr and Hepatitis C.

A group of genes with potential predictive value for progression to severe dengue warrants further investigation. These include CCL7, ACOD1, MT1G, TNFSF15, MYLK2, and SERPINB2. While expressed in both dengue with warning signs and severe dengue, these genes exhibited an average fold change twice as high in severe dengue, suggesting their potential as indicators of disease severity. A quantitative assay measuring these genes longitudinally, starting from the early stages of dengue infection, could prove valuable in predicting progression to severe dengue. CCL7, encoding monocyte chemotactic protein 3, is a secreted chemokine that attracts macrophages during inflammation. The elevated levels of CCL7 observed in severe dengue may reflect an exacerbated inflammatory response, potentially contributing to disease progression.

Being a multifaceted regulator of infection and inflammation, especially in immunometabolism, ACOD1 has garnered a lot of interest. Changes in the interplay between immunological and metabolic pathways during stress are referred to as immunometabolism. ACOD1 expression has been reported in pathogens such as Escherichia coli, Mycobacterium tuberculosis, and Zika virus [42]. MT1G has been considered a metalloprotein that largely balances metals to alleviate heavy metal toxicity and lower stress reactions; however, its involvement in infectious diseases was also reported. Among these, there are Human coxsackievirus B type 3 (CB3) and Influenza A viral infections [14]. The TNFSF15 is a cytokine produced by endothelial cell in established blood vessels and function to maintain vasculature, a number of disorders such as tumors, Crohn’s disease, ulcerative colitis, and other autoimmune diseases, have been linked to TNFSF15 mutations [13]. In regard to dengue, elevated TNFSF15 could be attributed to its response to vascular leakage during severity.

The remaining two genes, MYLK2 and SERPINB2 that are predictive of progression to severe dengue belong to the platelet activation pathway. Recent studies demonstrated that dengue virus binds to lectin receptor and triggers platelet activation and apoptosis, which generates inflammatory responses in target monocytes [11, 35]. Another research reported the detection of dengue virus in platelets isolated from dengue patients [28]. This raised the possibility of dengue virus replicating in platelet cells and subsequently contributing to alteration of the platelet transcriptome profile. One study provides the first evidence that dengue virus attack platelets and utilize their translational machinery to replicate and produce infectious virus [33]. The MYLK2 is a myosin kinase exclusively expressed in adult skeletal muscle and prerequisite for ADP-induced platelet aggregation. Another gene, SERINB2, is a plasminogen activator inhibitor-2 and had direct link to coagulation factor and involvement in thrombosis. These two genes were collectively elevated in severe dengue indicating the role in bleeding manifestation.

A gene identified as ATP5MF-PTCD1 was found to be overexpressed in severe dengue with fold change of 23. This remarkable expression was not described or reported previously. It represents naturally occurring read-through transcription between the ATP5J2 (ATP synthase, H + transporting, mitochondrial Fo complex, subunit F2) and PTCD1 (pentatricopeptide repeat domain 1) genes on chromosome 7 and encodes a fusion protein. Further involvement in inflammation or infectious diseases remains unclear. However, the result needs to be interpreted with caution owing to the low number of severe dengue cases that were analyzed in this study.

While this study provides valuable insights, it has several limitations. The sample size was relatively small, necessitating validation in larger cohorts across different dengue serotypes. Additionally, the temporal gene expression dynamics during different stages of dengue infection require further exploration.

Conclusion

This study contributes to the growing body of research on dengue gene biomarkers by identifying several potential candidates, particularly IFI27 and KCTD14, which show promise for dengue detection, and CCL7, ACOD1, MT1G, TNFSF15, MYLK2, SERPINB2, and ATP5MF-PTCD1, which hold potential as predictive biomarkers for severe dengue. Further investigation is necessary to elucidate the specific roles of these genes in dengue pathogenesis, validate their diagnostic and prognostic utility, and ultimately translate these findings into clinically relevant applications. The transcriptome profile of dengue infected hosts that have been elaborated and summarized in this study at this stage could benefit researchers and prompt them to validate these genes in their own cohort settings.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

JS was involved in conceptualization, methodology, lab investigation, formal analysis, funding acquisition, supervision, and writing original manuscript draft. SSMS contributed to sample collection, supervision at sample collection site, consultation on clinical aspects, and review of manuscript. SSH and NIFN conducted lab investigation and data curation. NII contributed to healthy control sample provision and supervision at sample collection site. RMZ and RT were involved in supervision, providing intellectual input and review of the manuscript.

Funding

The study had obtained Ministry of Health Malaysia Research Grant (P42 00500 117 1002) under the communicable disease scope.

Data availability

The datasets analyzed in this study will be available in the Gene Expression Omnibus (GEO) public repository, https://www.ncbi.nlm.nih.gov/geo/ under accession number GSE279208. Other datasets used during the present study are available from the corresponding author upon reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

The study has obtained ethical approval from the Medical Research and Ethics Committee (MREC), Ministry of Health Malaysia (NMRR-19–1691-48343), dated 6.11.2019. All participants in the study have given signed consent. All methods were carried out in accordance with Declaration of Helsinki.