Transcriptional landscapes of dengue patients reveal associations between HERV activity and disease severity

Abstract

Human endogenous retroviruses (HERVs), remnants of ancient viral infections, are increasingly recognized as potential regulatory elements influencing gene expression and immune responses. However, their role in infectious disease severity remains poorly understood. Here, we investigate the transcriptional impact of HERVs in dengue disease progression by leveraging RNA sequencing data from a large cohort of hospitalized patients across different severity subgroups. We identify differentially expressed HERVs and their co-expressed genes, revealing a severity-dependent transcriptional signature. Notably, genes co-expressed with HERVs in severe dengue were enriched in pathways related to immune modulation, phagocytosis, infectious disease response, and calcium signaling. We further demonstrate that HERVs may regulate genes through co-localization and chromatin interactions, with several HERV–gene pairs residing within same topologically associated domains (TADs). Using promoter interaction analysis, we identify potential regulatory mechanisms, exemplified by the interaction between HERV elements and MAGI3, an immune-modulatory gene. Validation in an independent cohort of 78 patients confirmed reproducibility of HERV expression patterns. Interestingly, genes encoding HERV envelope proteins, known for their immunomodulatory and antiviral properties, were consistently downregulated in severe dengue across both cohorts. These findings suggest a role for HERVs in shaping host transcriptional responses to dengue, with potential implications for disease progression and severity.

Graphical Abstract

Graphical Abstract.

Introduction

The human genome consists of ~3 billion base pairs, coordinated into 23 pairs of chromosomes, encoding both protein-coding genes and vast noncoding regions [1]. While protein-coding genes make up only ∼3% of the genome, the majority consists of intergenic DNA, introns, pseudogenes (∼0.5%), and transposable elements (TEs), such as long interspersed nuclear elements (LINEs), short interspersed nuclear elements (SINEs), and long terminal repeat (LTR) retrotransposons [2, 3]. These repetitive elements account for up to 69% of the genome [4].

Among these repetitive elements, human endogenous retroviruses (HERVs) are remains of ancient viral infections and constitute ∼8% of the human genome [5]. Although many HERV sequences have accumulated mutations and lost their ability to retrotranspose, some have retained functional roles, including transcriptional regulation, immune modulation, and potential involvement in diseases such as cancer and neurological disorders, autoimmune disease, and diabetes. For example, members of the HERV-W family, including MSRV and ERVWE1 (Syncytin-1), are implicated in multiple sclerosis, where Env proteins was hypothesized to foster neuroinflammation and demyelination, possibly through Epstein–Barr virus activation, and thus are emerging therapeutic targets [6]. This is similarly reflected in how HERV-W and HERV-K18 both potentially contribute to T1D by eliciting proinflammatory cytokines, such as interleukin 1 (IL-1), interleukin 6 (IL-6), and tumor necrosis factor-α (TNF-α), and impairing β-cell insulin secretion via TLR4 signaling, while acting as a T-cell superantigen that promotes autoimmunity [7]. Furthermore, in schizophrenia, increased brain and cerebrospinal fluid expression of HERV-W, together with the fusogenic activity of its Env protein and possible de novo genomic integrations, further point to its pathogenic involvement in neuropsychiatric disorders [8]. Recent studies suggest that HERV-derived regulatory elements can influence gene expression through mechanisms such as enhancer activity, alternative splicing, and noncoding RNA production [9]. RNA sequencing (RNA-seq) studies have increasingly highlighted the transcriptional activity of HERVs, with evidence suggesting that their expression is tightly regulated by host epigenetic mechanisms [10].

While previous studies have identified HERV-driven gene regulation, a comprehensive transcriptome-wide analysis linking specific HERV elements to host gene expression remains limited. The extent to which HERVs contribute to transcriptional networks, particularly in a cell type- or disease-specific manner, is not well understood [11]. Furthermore, the mechanisms by which HERV-derived sequences function as enhancers or regulatory elements require further investigation. Recent studies have reported HERV activation in COVID-19 [12–16, 17], and previous work from our lab has highlighted a significant association between HERV-derived LTR elements and COVID-19 disease severity [18–21, 22]. Despite this evidence, the role of retroviral elements in nonretroviral infections and their impact on disease severity remain largely unexplored. Notably, recent work has shown that the HERV-W ENV-U3R transcript, strongly expressed in B cells from patients with severe COVID-19 and post-acute sequelae, aligns best with the ERVWE2 locus on chromosome X and contains a premature stop codon [23]. However, translation of a full-length ENV protein was experimentally demonstrated via ribosomal readthrough, highlighting a non-canonical mechanism by which HERV elements with defective coding potential may still produce functional proteins. These findings underscore the need for deeper investigation into the transcriptional and translational regulation of HERV loci, beyond classical gene annotation frameworks.

In parallel, a growing body of evidence underscores the functional consequences of HERV protein expression in the context of infectious and neuro-inflammatory diseases. Several studies have demonstrated that HERV-encoded envelope proteins, once expressed, can actively participate in immune modulation and cellular dysfunction, often exacerbating disease pathology. For viral infections, including that of HHV-6A, MSRV-Env is induced rapidly and specifically in immune and brain-derived cells, at least in part by interaction with the CD46 receptor, especially its SCR3 and SCR4 motifs [24]. After induction, the surface unit of the Env protein (ENV-SU) is a ligand for pattern recognition receptors like TLR4/CD14 to trigger innate immunity. This stimulation enhances proinflammatory cytokine production, monocyte activation, and dendritic cell maturation in favor of a Th1-polarized immune response [25]. HERV-W Env is markedly upregulated in schizophrenia and serves as a plausible central molecular driver of neuro-inflammation, mitochondrial dysfunction, and synaptic degeneration. HERV-W Env activates innate immune signaling through both the TLR3–IL-6 and TLR4–MyD88 pathways, resulting in elevated production of inflammatory mediators such as TNF-α, IL-10, and C-reactive protein (CRP), thereby promoting chronic sterile neuro-inflammation [26, 27]. In addition to its immunomodulatory effects, HERV-W Env potentially disrupts neuronal energy metabolism by inhibiting mitochondrial complex I activity and perturbing calcium homeostasis, effects that are mediated via NDUFV2P1 dysregulation [28].

Beyond inflammation and bioenergetic impairment, HERV-W Env profoundly affects neuronal viability by activating the cGAS–STING–IFN-β axis, potentially promoting neuronal apoptosis, and repressing the neuroprotective lncRNA linc01930 [29]. It also drives ferroptosis through the downregulation of GPX4 and SLC3A2, leading to excessive lipid peroxidation and mitochondrial damage [30], while simultaneously triggering pyroptosis via the NLRP3–Caspase-1–Gasdermin D pathway, which enhances IL-1β release and neuroinflammatory injury [31]. These multifaceted mechanisms converge to amplify neuronal death through apoptosis, ferroptosis, and pyroptosis, establishing HERV-W Env as a key effector of neurodegeneration in schizophrenia.

Moreover, HERV-W Env contributes to neuronal plasticity deficits and structural abnormalities by perturbing synaptic signaling networks. It enhances HTR1B expression through ALKBH5-mediated m⁶A demethylation, which activates the ERK–ELK1–Arc cascade, leading to reduced dendritic spine density in serotonergic neurons [32]. Concurrently, it disrupts the Wnt5a/JNK/Arp2 noncanonical pathway via miR-141-3p upregulation, resulting in altered dendritic morphology and hippocampal neuron abnormalities [33]. In addition to this, HERV-W Env also plays an indirect role in neuronal damage in neuro-inflammatory disorders by activating microglia and destabilizing NMDA receptor function, especially the GluN2B subunit, which is essential for synapse formation [34]. Likewise, in amyotrophic lateral sclerosis (ALS), the HERV-K (HML-2) Env protein contains neurotoxic amino acid motifs (e.g. Ser298 and Pro312) that trigger oxidative stress and play a direct causative role in neuronal degeneration [35]. Together, these findings highlight that HERV-encoded proteins are not merely passive remnants of ancient infections but active modulators of host–pathogen interactions with potential roles in shaping immune responses and disease outcomes.

Dengue virus is a persistent infection with recurrent annual outbreaks, particularly in tropical regions like India, placing a significant burden on the healthcare systems. As per the Dengue Worldwide Overview: ECDC Report, since early 2025, >1.4 million dengue infections have been reported worldwide. In India alone, over 233000 cases were recorded in 2024, and 49573 cases have been recorded till August 2025 [36]. Despite the increasing evidence of HERV involvement in RNA virus infections, its role in dengue remains poorly understood. Recent studies by Wang et al. and Ferriera et al. have characterized the HERV transcriptome in dengue infection [37, 38], while our group has reported elevated expression of HERV-derived LTR elements at splicing sites in severe dengue patients [39].

Given HERVs’ role in gene regulation and immune modulation, dengue severity—ranging from mild symptoms to severe cases marked by plasma leakage and immune dysregulation—provides a compelling context to examine these interactions. In this study, we aim to investigate the contribution of HERV elements to dengue disease severity. By integrating transcriptomic analysis with repeat annotation, we seek to identify transcriptionally active HERV elements, characterize their regulatory impact, and determine their potential influence on host gene expression and immune responses in dengue infection. These findings will offer novel insights into the functional significance of HERVs in shaping the human transcriptome and their role in disease progression, especially single-stranded RNA virus infections.

Materials and methods

Human subjects and clinical protocol

This cross-sectional study was conducted at the CSIR-Institute of Genomics and Integrative Biology (IGIB) in New Delhi, India. We retrospectively recruited 112 hospitalized symptomatic individuals who tested positive for dengue via NS1 antigen testing. Informed written consent was obtained from all participants, and blood samples were collected in EDTA vials by a trained medical team at the MAX Super Speciality Healthcare Hospital, New Delhi, during the patients’ initial hospital visit. Sample collection adhered to the principles of the Declaration of Helsinki and took place between August and November 2022, coinciding with the peak dengue transmission period in New Delhi. In an independent second cohort, 80 individuals were recruited from August to November 2023. Following informed consent, blood samples were collected and tested for dengue NS1 antigen, and were classified into dengue-negative and dengue-infected. Clinical data were subsequently retrieved from participants' electronic health records (EHR). The study was reviewed and approved by the CSIR-IGIB Human Ethics Committee (Ref. No.: CSIR-IGIB/IHEC/2020–21/01).

Collection and classification of clinical samples

According to the 2009 WHO Dengue Classification Guidelines, 45 of the 112 patients from the first cohort were categorized as “without warning signs” and classified as mild (with normal platelet and leukocyte counts). The remaining 67 patients fell into the “with warning signs” category and were further stratified into moderate (leukopenia, n = 46) and severe (both leukopenia and thrombocytopenia, n = 21) cases [39, 40]. In the second cohort, the participants were segregated into dengue negative (n = 20), and the dengue positive patients were further segregated into mild (n = 40) and severe dengue infection (n = 20) based on the WHO guidelines. Due to a smaller number of patients with intermediate severity, no moderate category was available in the second cohort.

RNA isolation and library preparation

Total RNA was extracted from 1 ml of blood per patient using the QIAamp RNA Blood Mini Kit (QIAGEN, Cat. No. 52304) with optimized modifications. These included reduced incubation and centrifugation times for erythrocyte lysis and an additional 3-min incubation during the washing steps. RNA purity was assessed using a NanoDrop spectrophotometer, and its integrity was confirmed by agarose gel electrophoresis. The extracted RNA was stored at −80°C until further processing.

For library preparation, 250 ng of total RNA was processed using the Illumina TruSeq Stranded Total RNA Library Prep Globin Kit (Illumina, Cat. No. 20020612), which included globin mRNA and ribosomal RNA depletion to enrich nonribosomal transcripts. First-strand complementary DNA (cDNA) synthesis was performed using random hexamer primers and reverse transcriptase in the presence of Actinomycin D to ensure strand specificity. The RNA template was then removed using RNase H, and second-strand cDNA synthesis was carried out using DNA polymerase I, yielding double-stranded cDNA. The cDNA was purified using AMPure XP beads (Beckman Coulter), followed by 3′-end adenylation to prepare for adapter ligation.

Libraries were indexed and amplified via PCR, with quality assessment performed on the Agilent 2100 Bioanalyzer. Final library concentrations were quantified using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific, Cat. No. Q32854). After dilution to 4 nM, libraries were pooled in equimolar proportions and sequenced on the Illumina NextSeq 2000 platform using paired-end 2 × 151 read lengths, with a final loading concentration of 650 pM.

Data analysis: quality control, mapping to human reference, and differential expression analyses

Raw sequencing reads were demultiplexed using bcl2fastq (v2.x). Read quality was assessed with FastQC to evaluate base quality scores and adapter content [41]. Low-quality bases (Phred score < 20) were filtered out, and adapter sequences were trimmed using Trimmomatic (v0.39) [42]. The quality-filtered reads were rechecked to ensure effective filtration before being mapped to the human reference genome (GRCh38.110 primary assembly, Gencode) using STAR (v2.7.11) [43]. –outFilterScoreMinOverLread and –outFilterMatchNminOverLread were set to 0.66 to ensure the alignment score must be at least 66% of the read length to be considered valid, and at least 66% of bases must match (not just align) for a read to be counted. Alignment quality was assessed using Samtools and Qualimap to ensure high mapping rates and proper genome coverage [44, 45].

HERV loci coordinates were obtained from HERVd, a curated knowledge base of HERV elements maintained by the Institute of Molecular Genetics, Academy of Sciences of the Czech Republic [46]. At first, multi-mapped reads were discarded to improve quantification accuracy. Aligned RNA-seq reads were intersected with annotated HERV regions using featureCounts to quantify reads mapping to HERV loci using the following command: featureCounts -p -O -T 40 -d 50 -a “hervd.gtf” -o “featureCounts_output.txt” “Aligned.sortedByCoord.out.bam”. A minimum fragment length of 50 bp was set for reads in order to be considered a valid read. Similarly, gene expression counts were generated using the Gencode transcript reference. A read filter cutoff of at least 10 raw reads in at least 50% of the samples of the smallest group was applied to remove the low-expressed HERVs.

Quantified data were imported into the R environment using the tximport package. The expression was TPM normalized to account for the variable length of HERVs. Differential expression (DE) analysis was performed separately for genes and HERV repeats using DESeq2 [47]. DE analysis was conducted across severity subgroups: mild versus moderate, mild versus severe, and moderate versus severe for both genes and HERV repeats. Similarly, DE analyses were performed for the second cohort: mild versus negative, severe versus negative, and severe versus mild. A Wald test was applied for significance calculation, and Benjamini–Hochberg correction was used to account for the false discovery rate. Differential expression was considered statistically significant for |log₂ fold change| ≥ 1.5 and adjusted P-value ≤ 0.05. While we used genomic coordinates to get the length of HERV loci that are expressed in our data and used a density plot to compare the distribution of the HERV lengths between DE, non-DE, and global HERVs. The significance between the length distributions was tested using Wilcoxon.

Co-expression and co-localization analyses of HERV-mediated gene regulation

We systematically investigated the role of HERV elements in immune response regulation. First, HERV loci were identified using the HERVd database, a comprehensive catalog of annotated HERV elements across the human genome. RNA-seq analysis revealed 15 388 DE HERV repeats across the severity subgroups.

To explore their regulatory potential, we identified the nearest downstream gene for each DE-HERV using a Python-based algorithm, which assigned the closest gene based on the minimum distance to its transcription start site (TSS) (full implementation available on GitHub). We then performed Pearson Correlation Coefficient (PCC) analysis on DE-HERV elements and their downstream DEGs, identifying co-expressed HERV–gene pairs with P-value ≤ 0.05 and |r| ≥ 0.5. Given prior evidence of HERVs influencing genes up to 100 kb downstream [48, 49, 50], we extended our analysis to identify intervening HERVs between DE-HERVs and their nearest genes, capturing additional co-expressed and co-localized HERV–gene pairs within this region.

Functional enrichment analysis

To investigate the biological relevance of DEGs and DE-HERVs, we conducted functional enrichment analysis using Enrichr-KG and ShinyGO, leveraging Reactome pathway annotations [51, 52]. Pathways with P-value < 0.05 were considered statistically significant. The results were visualized as bubble plots, where log₁₀ (FDR) was color-coded, and gene counts were represented by bubble size, providing an intuitive interpretation of pathway significance.

Identification of HERV–gene distal interactions

To assess chromatin-level interactions, we utilized data from the Topologically Associating Domain Knowledge Base (TADKB), an integrated resource for TAD structures across seven cell lines [53]. Given that our samples were derived from blood specimens, we selected K562 cells, a hematopoietic cell line with a chromatin structure closely resembling that of blood cells.

Among all identified co-expressed and co-localized HERV–gene pairs, we found 47 pairs within shared TADs. To further predict HERV–gene interactions within TADs, we applied rgt-TDF promoter analysis [54], which identified binding target promoter sites for three significantly interacting co-expressed and co-localized HERV–gene pairs, reinforcing the potential role of HERV elements in gene regulation.

Statistical analysis and data visualization

The statistical analysis was conducted using licensed versions of GraphPad Prism and R (v4.0.2), available from CRAN. Wherever appropriate, differences between continuous data points were assessed using the two-tailed Mann–Whitney U test, whereas categorical data comparisons were conducted through chi-square testing. A significance threshold of P < 0.05 was applied unless otherwise specified. For data visualization, graphs and illustrations were generated using Matplotlib, Seaborn, and Plotly (Python), ggplot2 and ggbio (R), and licensed versions of GraphPad Prism and BioRender. Figures were further refined using Inkscape. Significance value is denoted as *, where * indicates P ≤ 0.05, ** indicates P ≤ 0.01, *** indicates P ≤ 0.001 and **** indicates P ≤ 0.0001.

Result

Clinical characteristics and patient segregation

In this study, 112 NS1 antigen-positive hospital-admitted patients with primary dengue infection were enrolled and stratified into three severity sub-phenotypes based on the 2009 WHO classification: mild (n = 45), moderate (n = 46), and severe (n = 21) (Fig. 1A). Detailed demographic and clinical information of the patients is recorded in Supplementary Table S1. The median age of patients across the three groups was comparable, with a slightly higher median in the severe group (35 years, IQR: 18–43) compared to the mild (27 years, IQR: 18–33), and moderate groups (26 years, IQR: 19–31.75). The principal coordinate analysis (PCA) highlighted clear segregation of samples across the groups (Fig. 1B and Supplementary Fig. S1A). Platelet counts progressively declined with increasing disease severity, with median values of 175 × 10⁹/L (IQR: 155–202) in the mild patients, 160 × 10⁹/L (IQR: 150–165) in the moderate, and 125 × 10⁹/L (IQR: 112–135) in the severe patients (Fig. 1C). Similarly, total leukocyte count (TLC) was significantly reduced in severe cases (2.7 × 10⁹/L, IQR: 2.3–3.1) compared to the moderate (3 × 10⁹/L, IQR: 2.5–3.575) and mild patients (5 × 10⁹/L, IQR: 4.6–5.9) (Fig. 1D). Lymphocyte percentages were elevated in the moderate (24.5%, IQR: 20.3–31.975) and severe patients (26%, IQR: 15.8–32.7) compared to the mild (14.1%, IQR: 9.4–21.3) (Fig. 1E). In contrast, neutrophil percentages were lower in the moderate (63.3%, IQR: 58.1–68.65) and severe patients (62.6%, IQR: 57.3–73) compared to the mild (75.9%, IQR: 66.7–79.5) (Fig. 1F). The increased lymphocyte frequency in moderate and severe cases likely reflects activation of adaptive immune responses during dengue infection, particularly T cells. Lymphocytosis is a common feature of viral infections and has been associated with recovery phases. Dengue progression is also characterized by a shift from neutrophil to lymphocyte predominance, captured by a decrease in the neutrophil-to-lymphocyte ratio (NLR), an established marker of immune dynamics and disease progression [55]. Notably, severe dengue cases also exhibited elevated direct bilirubin levels, a clinical marker indicative of liver dysfunction, which could reflect hepatic involvement in severe disease (Fig. 1G).

Figure 1.

Overview of study design, patient stratification, and key clinical parameters in dengue patients. (A) Schematic representation of the study design, including patient enrollment, dengue diagnosis using NS1 antigen testing, RNA isolation, sequencing, and subsequent bioinformatics analysis. Created in BioRender. Devi, P. (2026) https://BioRender.com/zablxvm. (B) PCA plot showing sample distribution in each group and segregation of groups. (C–G) Key clinical parameters associated with dengue severity and statistical significance between the groups were assessed using the Wilcoxon rank-sum test. (C) Platelet count (10⁹/L), showing a significant decrease in moderate and severe dengue patients compared to mild. (D) Total leukocyte count (10⁹/L), which decreases with increasing disease severity. (E) Lymphocyte percentage, showing a significant change with severity. (F) Neutrophil percentage, which decreases with disease severity. (G) Bilirubin levels (mg/dL), showing a significant elevation in severe dengue cases. Statistical significance is indicated by P-values in each panel.

Evaluation of additional clinical symptoms revealed that fever, dehydration, and body aches were more pronounced in severe cases, whereas symptoms such as diarrhea and vomiting did not significantly differ across severity subgroups. Given that these patients were NS1 antigen-positive, we also analyzed viral serotypes using bulk RNA-seq data. Our analysis revealed that the majority of patients were infected with DENV2, a serotype often associated with severe clinical manifestations. However, for some patients, the viral reads were insufficient to determine the serotype, and these cases were distributed across different severity subgroups. These clinical and virological findings reaffirm the severity-based classification of the dengue patients and highlight key physiological changes associated with disease progression.

HERV expression increases with dengue severity, showing distinct activation patterns, genomic, and functional shifts

To investigate the role of HERV elements in dengue pathogenesis, we performed RNA sequencing on blood samples from patients stratified into mild, moderate, and severe disease groups. Our experimental workflow involved isolating RNA from patient blood samples, followed by whole transcriptome sequencing; mapping the reads to the human genome; identifying HERV loci based on established genomic coordinates; filtering out lowly expressed loci; and finally, performing differential expression analysis using standard/customized bioinformatic pipelines. The characterization of HERV loci in the human genome is mentioned in Supplementary Table S2, while the sample-wise reads mapped to HERV loci are available as Supplementary Table S3, and the HERV coverage across groups is presented in Supplementary Fig. S1B.

In total, 565471 HERV loci were annotated in the human genome, with 25889 expressed in our dataset. Out of these, 15387 loci were differentially expressed (DE) across the patient groups (Supplementary Table S4A–C). While the median age and male-to-female ratio were numerically higher in the severe group, statistical testing did not show a significant difference in the age distribution across the severity groups (P-values = 0.14 and 0.34, respectively), and therefore these factors were not included as a covariate in the DE analysis. Importantly, HERV expression was lowest in mild patients, with a progressive increase observed in moderate and severe dengue infections (Fig. 2A–C). Figure 2D illustrates the minimal overlap of DE HERV loci across the different severity comparison groups, suggesting that unique subsets of HERVs are modulated when transitioning from mild-to-moderate versus mild-to-severe dengue. This distinct regulation highlights potentially different roles or activation thresholds of HERV loci at varying disease severity.

Figure 2.

Differential expression and genomic characteristics of HERVs across disease severity groups. (A–C) Volcano plots depicting differentially expressed HERV loci in pairwise comparisons: (A) moderate versus mild, (B) severe versus mild, and (C) severe versus moderate. Each dot represents an individual HERV locus, with significantly upregulated and downregulated loci highlighted in blue and red, respectively. (D) Upset plot representing the total number of DE HERV loci across severity subgroups, with statistical comparisons shown below to see the overlaps between severity subgroups. (E) Density plot showing the distribution of HERV repeat lengths across the genome, with comparisons between globally (all annotated HERV loci across the genome) distributed HERVs, non-DE HERVs, and DE HERVs. Statistical significance is indicated. (F) Ridge plot illustrating the distribution of HERV repeat lengths for different severity subgroup comparisons. (G and H) Bar plots showing the genomic distribution of DE and non-DE HERVs, categorized by their location within exonic, intronic, intergenic, or UTRs. (G) Comparison between DE, non-DE, and global HERVs. (H) Breakdown across disease severity subgroups. (I and J) Bar plots showing the proportion of HERVs overlapping with different gene biotypes, including protein-coding genes, lncRNAs, and intergenic regions. (I) Comparison of DE, non-DE, and global HERVs. (J) Breakdown across disease severity groups. (K) Distribution of HERV families across the three severity subgroup comparisons. Different colors represent the comparisons: moderate versus mild, severe versus mild, and severe versus moderate. Asterisks indicate statistical significance levels (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

We then went on to characterize the DE HERVs in each comparison group to better understand their potential functional implications (Supplementary Table S5). First, we compared the length of the HERV loci that were differentially expressed to that of the non-DE as well as the genome-wide length of HERV loci. The DE-HERV loci were significantly longer than both the non-DE and global (genome-wide) HERV loci (Fig. 2E). Figure 2F demonstrates that HERV loci upregulated in the moderate patients (versus mild) tend to be of longer length compared to those upregulated in the severe patients (versus mild). Furthermore, when comparing severe-to-moderate severity, the upregulated HERV loci exhibit the shortest length. These differences in locus length may reflect variations in regulatory potential or chromatin accessibility, which in turn could influence the magnitude or nature of host responses during dengue infection (Supplementary Table S2).

Further, we investigated the genomic distribution as well as the biotype of genes overlapping the HERV loci. While the majority of the HERV loci were within the intergenic or intronic region, Fig. 2G showed a significantly high number of DE-HERV loci to be present within the intergenic region as compared to the global distribution of HERV repeats, whereas the majority of expressed but non-DE HERVs were located within the intronic region (Supplementary Table S2). Figure 2H shows that HERVs upregulated in both moderate and severe patients (relative to mild) are primarily located in the intergenic regions. However, in the direct comparison between severe and moderate patients, a significant proportion of the upregulated HERVs (in the severe group) are situated within the intronic regions. The observed shift toward intronic localization in severe patients highlights a potential change in HERV genomic activity patterns that may coincide with altered host gene regulation during severe diseases. Analysis of overlapping gene biotypes (Fig. 2I and J) further underscores the differential impact of HERV activation. In line with the previous finding in Fig. 2G, genes overlapping the DE-HERV loci were primarily of intergenic origin, while genes overlapping the non-DE HERVs were of protein-coding origin (Fig. 2I). In comparisons of moderate and severe cases versus mild, the majority of upregulated HERV (in moderate and severe compared to mild) loci are associated with the intergenic regions. In contrast, HERVs that are specifically upregulated in the severe patients (when compared to moderate) predominantly overlap with protein-coding genes (Fig. 2J). This pattern suggests that in severe dengue, HERV activation may be more closely associated with the expression of protein-coding genes, potentially reflecting or contributing to altered cellular functions linked to disease pathology.

In addition, a similar comparison of the DE repeats within each HERV family revealed a varying pattern suggesting a potential role in disease progression per subtype (Fig. 2K, and Supplementary Fig. S1C andD). Notably, most of the HERVH repeats were elevated in the moderate patients (versus mild) but decreased in severe cases (versus moderate), indicating a possible early immune-modulatory role that diminishes with increasing disease severity. Conversely, the MLT and MES families showed increased expression in the severe (versus moderate) patients, suggesting their involvement in immune dysregulation and heightened inflammatory responses, potentially contributing to endothelial dysfunction and vascular leakage [56]. These findings suggest that specific HERV families may be differentially associated with immune responses in dengue, with some elements linked to milder disease states and others associated with immune over-activation in severe dengue.

Collectively, these results indicate that HERV expression is tightly correlated with dengue disease severity—starting at low levels in mild cases and increasing through moderate-to-severe cases. These findings provide a framework for understanding the mechanistic links between HERV activation and dengue disease progression.

HERV activation-associated immune, metabolic, and signaling pathways in dengue severity

To understand the functional implications of HERV activation across different severity subgroups, we performed pathway enrichment analysis of genes overlapping differentially expressed HERV loci. In the transition from mild-to-moderate dengue (Fig. 3A), enriched pathways included Ras signaling, phospholipase D signaling, and serotonergic and glutamatergic synapse pathways, indicating a potential link between HERV activity and immune signaling, neurotransmission, and metabolic dysregulation. Notably, metabolic pathways were significantly enriched in the moderate severity patient, suggesting HERV involvement in host metabolic reprogramming, which could influence immune cell function and viral pathogenesis.

Figure 3.

Pathway enrichment and differential expression of HERV loci across disease severity groups. (A–C) Dot plots representing pathway enrichment analysis of differentially expressed genes associated with HERV loci in pairwise comparisons: (A) moderate versus mild, (B) severe versus mild, and (C) severe versus moderate. Pathways are ranked based on fold enrichment, with dot size representing the number of genes involved and color intensity indicating statistical significance [−log10(FDR)]. (D–F) Volcano plots showing differentially expressed genes in pairwise comparisons: (D) moderate versus mild, (E) severe versus mild, and (F) severe versus moderate. Each dot represents an individual gene, with significantly upregulated genes highlighted in orange and downregulated genes in teal. Nonsignificant genes are shown in gray.

Interestingly, in the severe versus mild subgroup, the gene functional categories shifted toward more immune response functions, including ErbB signaling, T-cell receptor signaling, phosphatidylinositol signaling, and neurotrophin signaling, along with the inflammatory mediator regulation of TRP (Transient Receptor Potential) channels (Fig. 3B). These findings indicate that HERV-associated gene expression correlates with pathways involved in immune activation, neuronal signaling, and metabolic disruption in severe dengue [17, 24, 57], including mitogen-activated protein kinase (MAPK) and Ras signaling, which are characteristic of heightened inflammatory responses.

Comparing moderate to severe dengue (Fig. 3C), enriched pathways included platelet activation, chemokine signaling, ubiquitin-mediated proteolysis, and endocytosis. These results highlight a potential role of HERV-associated genes in dysregulating coagulation, immune cell trafficking, and protein degradation, processes that are hallmarks of severe dengue pathogenesis. The enrichment of O-glycan biosynthesis pathways further suggests potential alterations in glycosylation patterns, which could impact viral entry, immune evasion, or host inflammatory responses. Collectively, these findings indicate that HERV activation is linked to key molecular pathways that contribute to the progression of dengue severity, potentially influencing immune dysregulation, metabolic shifts, and endothelial dysfunction.

Co-expression and co-localization of metabolic, immune, and signaling pathway genes with HERV repeats in severe dengue

To investigate the potential regulatory influence of HERV repeats on gene expressions in dengue severity, we performed differential expression analysis across the severity subgroups and identified 56 DE genes in the mild versus moderate comparison, 4141 in mild versus severe, and 714 in moderate versus severe (Figs 3D–F, and Supplementary Tables S6A–C). Given the known transcriptional regulatory potential of HERV repeats on genes present downstream, we first identified differentially expressed genes located downstream of differentially expressed HERV repeats in each comparison group. We then assessed their potential regulatory associations through Pearson correlation analysis, allowing us to determine co-expression patterns between the HERV repeats and nearby genes (Fig. 4A). This analysis revealed significant correlations between three genes and five HERVs in the moderate versus mild group, 12 genes and 13 HERVs in the severe versus moderate group, and 873 genes with 1452 HERV repeats in the severe versus mild group. Notably, only significant positive correlations between HERV loci and genes were observed under the applied thresholds (Fig. 4B–D and Supplementary Table S7).

Figure 4.

Co-expression, correlation, and functional enrichment analysis of HERVs and host genes. (A) Schematic representation plot showing the correlation between HERV expression and host gene expression. Created in BioRender. Devi, P. (2026) https://BioRender.com/zablxvm. (B–D) Dot plot representing the correlation between differentially expressed HERV loci and downstream DE genes across conditions: (B) moderate versus mild, (C) severe versus mild (top co-expressed pairs of HERV loci and genes are visualized), and (D) severe versus moderate group. Dot size corresponds to correlation strength, and color indicates the correlation coefficient (red: positive, blue: negative). (E–G) Network analysis of significantly enriched pathways associated with genes co-expressed with HERVs. Nodes represent genes, with edges indicating interactions or shared pathways across (E) mild versus moderate, (F) mild versus severe, and (G) moderate versus severe group. (H) Genomic organization of a representative HERV locus and its proximity to neighboring genes. Green blocks represent HERV subunits (H1–H5), and the blue region represents a neighboring gene within 100 kb. Created in BioRender. Devi, P. (2026) https://BioRender.com/zablxvm. (I) Network analysis of significantly enriched pathways associated with genes co-expressed and co-localized with the HERV loci in the severe versus mild group.

Pathway enrichment analysis of genes significantly correlated with HERV repeats revealed distinct functional associations across the disease severity subgroups. In the moderate versus mild comparison, enriched pathways included bile acid metabolism, acetylcholine-mediated transmission, and membrane transport, potentially reflecting early metabolic and neuronal adaptations to infection (Fig. 4E). In the severe versus mild, enriched pathways included bile acid synthesis, pancreatic secretion, drug metabolism, calcium signaling, and neuroactive signaling, suggesting systemic metabolic dysregulation and neurological involvement in severe dengue (Fig. 4F). The severe versus moderate comparison showed enrichment in phagocytosis, infectious disease pathways, MAPK signaling, calcium signaling, and oxytocin signaling, indicative of heightened inflammatory responses, immune activation, and endothelial dysfunction in severe dengue cases (Fig. 4G). It is worth noting that the calcium signaling pathway is one of the key pathways altered in dengue infection, with several reports highlighting the role of calcium signaling in the replication of the dengue virus in the case of severe dengue [58, 59]. These findings suggest that HERV repeats may influence the expression of nearby genes and are associated with alterations in metabolic, immune, and signaling pathways linked to dengue disease severity.

To further investigate the potential regulatory impact of HERVs on their co-expressed genes, we assessed the spatial co-localization of HERV repeats with the DE genes. We first identified whether additional HERV loci were present between the previously identified HERV-downstream gene pairs. Next, we extended our search to 100 kb upstream from the start of the downstream gene to capture HERV elements within this range, given prior reports suggesting that HERVs can regulate gene expression up to 100 kb from their integration site (Fig. 4H). Finally, we evaluated potential co-expression between these HERVs and genes within the 100 kb region, identifying co-expressed and co-localized HERV–gene pairs. This analysis revealed one co-localized pair in mild versus moderate (Fig. 4B), 43 in mild versus severe (Fig. 4C), and two in moderate versus severe (Fig. 4D).

Next, to understand the function of the genes co-expressed and co-localized with the HERV loci, we performed pathway enrichment of the genes. The genes in the mild versus moderate, and moderate versus severe groups were pseudogenes and lncRNAs, and hence their function could not be established. Pathway enrichment analysis of genes co-localized and co-expressed with HERV repeats in the mild versus severe group revealed significant enrichment in neurological functions, O-glycan biosynthesis, glycosaminoglycan biosynthesis, glycolipid metabolism, and viral myocarditis pathways (Fig. 4I). The enrichment of biosynthetic pathways related to glycans and glycolipids, key components of cellular membranes, and immune recognition suggests a potential role of HERV-associated transcriptional regulation in modulating host–virus interactions, immune evasion, and endothelial function in severe dengue. The viral myocarditis pathway enrichment further highlights a possible link between HERV-mediated gene regulation and cardiovascular complications in severe dengue cases, which aligns with clinical observations of myocardial dysfunction in severe dengue patients. Interestingly, the genes identified in the mild versus moderate and moderate versus severe groups were predominantly pseudogenes, warranting further investigation into their potential functional roles in dengue pathogenesis.

Collectively, these findings suggest that HERV repeats may contribute to dengue disease severity by modulating gene expression in key metabolic, immune, and signaling pathways, potentially influencing host–pathogen interactions and disease progression.

Topologically organized HERV-mediated gene regulation via cis-regulatory elements in promoter regions

Given the ability of HERV repeats to regulate gene expression over long genomic distances, we sought to determine whether these regulatory interactions occur within the same topologically associated domains (TADs)—self-contained chromatin structures that constrain gene regulation. Using TADKB (K562 cell line, 10 kb resolution Hi-C data), we assessed the spatial organization of the 47 co-expressed and co-localized HERV–gene pairs identified in our earlier analysis. We found that 36 of these pairs were located within the same TAD, suggesting that their transcriptional co-regulation could be influenced by shared three-dimensional chromatin architecture (Supplementary Table S8).

To further explore the mechanistic basis of HERV-mediated gene regulation, we next investigated whether there was a direct physical interaction between the HERVs and the promoters of their downstream co-expressed genes. Using the RGT Toolbox Triplex Domain Finder, we predicted potential triplex-forming interactions, revealing strong interactions between HERVs and the promoter regions of three genes: LINC01121, RP11-5P15, and MAGI3 (Supplementary Table S9). Given its relevance to immune function, we selected MAGI3 for further analysis. MAGI3 encodes a membrane-associated guanylate kinase involved in cell polarity, signaling, and immune modulation, making it a strong candidate for HERV-driven regulatory control. By mapping the interaction sites within the MAGI3 promoter, we identified overlaps with known cis-regulatory elements harboring HERV derived repeats, suggesting that HERV-derived sequences may influence gene expression by modulating promoter activity (Fig. 5).

Figure 5.

Schematic illustration of HERV-mediated gene regulation within TAD. An intergenic HERV (red) interacts with the MAGI3 promoter (blue) via a cis-regulatory element (CRE, yellow) within a TAD, suggesting a potential regulatory mechanism. Created in BioRender. Devi, P. (2026) https://BioRender.com/zablxvm The genome browser tracks show gene annotations, regulatory elements, and transposable elements.

Together, these findings provide compelling evidence that HERVs may regulate gene expression not only through long-range chromatin interactions but also by physically engaging with promoter regions within the same TAD. This mechanistic link underscores a potential role for HERV elements in modulating immune and cellular pathways in dengue pathogenesis, warranting further functional validation.

Validation of HERV dysregulation across dengue severity in an independent clinical cohort

In order to validate the HERV expression pattern observed across dengue mild, moderate, and severe groups, we performed validation in an independent clinical cohort. This cohort included 80 individuals who were stratified into dengue-negative (n = 20) and dengue-infected (n = 60) based on NS1 antigen test results. The infected individuals were further categorized into mild (n = 40) and severe (n = 20) groups according to the WHO guidelines of dengue severity (Fig. 6A). The demographic and clinical details of the cohort are provided in Supplementary Table S1.

Figure 6.

Validation of HERV dysregulation in an independent dengue patient cohort. (A) Graphical representation of the study design. Created in BioRender. Devi, P. (2026) https://BioRender.com/cn087m2 (B) PCA plot highlighting clear sample segregation across groups. (C–G) Key clinical parameters associated with dengue severity: (C) platelet counts (in 10⁹/L), and (D) total leukocyte counts (in 10⁹/L) showing a significant decrease with increasing dengue severity as compared to dengue negative and mild dengue cases. (E) Lymphocyte percentages showed an initial drop in mild cases and a rise in severe cases. (F) Neutrophil percentages were also found to be significantly decreased in severe dengue. (G) Direct bilirubin levels were elevated in severe dengue cases, indicating liver dysfunction. Statistical significance is indicated by P-values in each panel. (H–J) Differential expression of HERVs in (H) mild versus dengue-negative, (I) severe versus mild, and (J) severe versus dengue-negative groups. Log fold change corresponds to the first group in each comparison. (K) Venn diagram showing the overlap of HERVs expressed in both cohorts. (L) Venn diagram showing overlap of DE HERVs across both cohorts. (M) Venn diagram showing overlap of DE HERVs from cohort 1 and non-DE but similarly expressed HERVs in cohort 2.

Principal coordinate analysis revealed clear sample segregation across groups (Fig. 6B and Supplementary Fig. S1E). Similar to our first cohort, we observed a progressive reduction in platelet count and total leukocyte count with increasing severity (Fig. 6C and D). Lymphocyte counts initially dropped in mild cases compared to dengue-negative individuals but increased again in severe cases (Fig. 6E), while neutrophil counts showed a marked decrease in severe dengue (Fig. 6F). As in the discovery cohort, we also observed a significant increase in bilirubin—a marker of liver dysfunction—in severe dengue patients (Fig. 6G).

RNA sequencing and data processing were performed as described earlier, using the same workflow for HERV identification and differential expression analysis. Sample-wise HERV-mapped read counts are provided in Supplementary Table S3, and group-wise HERV coverage is shown in Supplementary Fig. S1F. Two samples were excluded from the analysis due to low read counts. Age and gender distributions were statistically non-significant and were therefore not included as covariates during DE analysis.

A total of 15106 HERV loci were expressed in this dataset, of which 4456 were differentially expressed across comparison groups (Fig. 6H–J). Consistent with our discovery cohort, we observed a global increase in HERV expression in severe dengue cases. Notably, 14926 (98.8%) of the expressed HERVs were also expressed in the first cohort (Fig. 6K), highlighting the reproducibility of HERV expression across independent datasets. Of the 4456 DE HERVs identified in this second cohort, 1758 (∼39.4%) overlapped with DE HERVs in the first cohort (Fig. 6L). Strikingly, an additional 7779 HERVs from the remaining expressed set (∼73%) followed a similar expression trend, albeit with a lower log fold-change that excluded them from DE classification (Fig. 6M).

Group-wise comparison of DE HERVs, as well as analyses of locus length, genomic location, and biotype distribution, followed patterns highly consistent with the first cohort and are summarized in Supplementary Table S10. These results collectively strengthen our findings, suggesting that HERV expression is robustly associated with dengue disease severity and may contribute to the underlying regulatory mechanisms.

Selective downregulation of HERV envelope-encoding genes despite overall HERV activation in severe dengue

While our global transcriptome analysis revealed a progressive increase in overall HERV expression from mild to severe dengue, this increase was largely driven by non-coding HERV elements, especially LTR sequences. To assess whether protein-coding HERV loci, particularly those encoding envelope (env) proteins, followed a similar trend, we analyzed the expression of a curated set of HERV env genes known to encode envelope proteins with potential immunomodulatory roles. Using reference sequences retrieved from NCBI, we performed transcript quantification with Salmon, followed by TPM normalization. Differential expression analysis across severity groups was conducted using the Wilcoxon rank-sum test in both datasets.

Strikingly, in contrast to the global HERV upregulation trend, several env genes—including ERV3-1, ERVW-1, ERVV-1, ERVV-2, ERVK3-1, ERVMER34-1, and ERVFRD-1—were significantly downregulated in severe dengue patients compared to mild or dengue-negative cases (Fig. 7A–O). This trend was observed consistently across both cohorts, with most genes showing a clear and statistically significant decline in expression. This inverse expression pattern suggests that while the HERV transcriptome landscape, which is largely composed of noncoding LTR elements, is activated, HERV-derived protein-coding transcripts, particularly env genes, are selectively repressed in severe dengue. While the precise biological function of these HERV-derived envelope genes in dengue pathogenesis remains unclear, their consistent reduction in severe cases points toward a potential role in modulating host responses during infection.

Figure 7.

Differential expression of HERV envelope protein-coding genes across dengue severity in two independent cohorts. (A–G) Expression of HERV envelope genes ERV3-1, ERV3-3, ERVW-1, ERVV-2, ERVH48-1, ERVK3-1, and ERVMER34-1 across mild, moderate, and severe dengue patients in Cohort 1. (H–O) Expression of the same set of genes along with ERVFRD-1 and ERVV-1 across dengue-negative, mild, and severe dengue patients in Cohort 2. Expression values are represented as length-scaled TPM, and statistical significance was determined using the Wilcoxon rank-sum test. Across both cohorts, a consistent and significant downregulation of HERV env genes was observed in severe dengue patients compared to mild/dengue-negative individuals, suggesting a potential loss of HERV-mediated antiviral and anti-inflammatory regulation in severe disease states. Asterisks indicate levels of statistical significance: P < 0.05 (*), P < 0.01 (**), P < 0.001 (***), and P < 0.0001 (****).

Taken together, our comprehensive analysis reveals a dual landscape of HERV activation in dengue infection—marked by a global upregulation of HERV transcriptome elements alongside a selective suppression of protein-coding HERV env genes in severe cases. This paradoxical expression pattern, consistent across two independent cohorts, underscores the complexity of HERV-mediated regulation in viral pathogenesis. These findings not only reinforce the association between HERV dynamics and dengue severity but also set the stage for mechanistic investigations into their functional roles in modulating host immune responses.

Discussion

Endogenous retroviruses are remnants of ancient viral infections that have been co-opted into the human genome, and their regulatory roles in immune function and disease progression are increasingly being recognized. Our findings suggest that HERV elements may act as cis-regulatory sequences influencing gene expression, particularly in the context of immune and metabolic responses. By integrating differential expression analysis, co-expression studies, and chromatin conformation data, we demonstrate that HERVs are not merely passive genomic elements but potentially contribute to transcriptional programs associated with disease severity. While the current analysis focused on downstream neighboring genes, it is important to note that HERV elements can also influence gene expression in the antisense orientation. Future work incorporating strand orientation and upstream gene relationships will be essential to fully elucidate the bidirectional regulatory potential of HERV-derived sequences.

A key strength of our study is the use of total RNA-seq data from two large cohorts of hospital-admitted dengue patients spanning different dengue severity subgroups as well as dengue negative group, providing a robust and clinically relevant dataset for assessing HERV-associated transcriptional changes. Prior studies in this area have predominantly relied on in vitro models, which, while valuable, do not fully capture the complex immune and metabolic interactions occurring in vivo [37]. By leveraging patient-derived transcriptomic data, we offer a more physiologically relevant perspective on how HERVs may modulate host gene expression during disease progression, laying the groundwork for future functional investigations.

One of the key hypotheses emerging from our study is that HERVs located near immune-related genes may have been co-opted by the host to regulate transcriptional responses. Many genes central to innate and adaptive immunity are under strong selective pressure, and regulatory elements within these regions including HERV sequences may evolve to fine-tune immune gene expression. In line with this, our analysis revealed that HERVs and immune genes were frequently co-expressed and co-localized within the same topologically associated domains. This suggests a potential mechanism where the 3D chromatin structure facilitates HERV-mediated transcriptional regulation of immune genes.

Furthermore, HERV long terminal repeats have been implicated in interferon-inducible transcriptional programs, and our findings support the idea that HERV-derived sequences may contribute to immune activation. This is particularly relevant in the context of dengue infection, where innate immune signaling plays a pivotal role in disease progression. HERV sequences embedded within immune gene loci may function as latent regulatory elements that become activated under inflammatory conditions, driving the expression of genes involved in antiviral responses and immune modulation. Interestingly, previous work by Liu et al. [60] demonstrated that hepatitis B virus X protein can increase HERV expression via NF-κB activation in HepG2 liver cells [60]. While distinct in etiology, these findings suggest that virus-induced hepatic inflammation potentially involving shared pathways like NF-κB may also contribute to HERV activation during severe dengue infection.

The relationship between HERV activity and host gene expression appears to be multifaceted, reflecting both shared regulatory environments and potentially selective interactions. While co-expression of HERVs and adjacent genes could arise from coordinated chromatin accessibility or activation of common signaling pathways, our data indicate that these associations are not uniformly observed. For example, in our dataset, an HERV (ERV_ERVL-E_ERV_0307250) and its neighboring gene CR2 were both upregulated in severe compared with mild dengue, consistent with a potential co-activation event within a permissive chromatin domain. In contrast, another HERV (ERV_MLT2C1_ERV_1254104) showed increased expression without corresponding changes in its downstream gene (IL32), and ERV_HERVL_2109217 exhibited an inverse pattern, with HERV upregulation accompanying suppression of its nearby gene PI3. Moreover, ERV_MSTC_ERV_0770458 remained unchanged while its adjacent IGF1 was upregulated in severe dengue (versus mild), further supporting that not all gene–HERV pairs share regulatory behavior. These distinct patterns suggest that HERV activation does not merely reflect passive transcriptional leakage from open chromatin regions but may involve context-specific regulation influenced by local chromatin architecture, transcription factor availability, and cellular signaling states. Collectively, these observations highlight a nuanced regulatory landscape in which some HERV–gene relationships may represent coordinated transcriptional responses to immune or stress stimuli, while others reflect independent or even antagonistic regulatory mechanisms.

Beyond their role as regulatory elements, we hypothesize that HERV sequences within genic regions may act as molecular patterns recognizable by innate immune sensors. Certain classes of pattern recognition receptors (PRRs), such as Toll-like receptors (TLRs) and RIG-I-like receptors (RLRs), can recognize nucleic acid sequences derived from viral elements, potentially leading to HERV-driven immune activation. This may have significant implications for viral infections, including dengue, where excessive immune activation is linked to disease severity. The observed enrichment of calcium signaling, MAPK signaling, and phagocytosis-related pathways in HERV-associated genes further supports the notion that HERV activity could contribute to immune reprogramming in infected individuals.

In addition to immune regulation, our results indicate a potential role for HERV elements in metabolic pathways. We observed strong co-expression between HERVs and genes involved in lipid metabolism, bile acid synthesis, and energy homeostasis, suggesting that HERVs may act as distal enhancers or insulators affecting metabolic gene transcription. This aligns with previous reports of HERV-mediated regulation of lipid metabolism genes [61], which could be particularly relevant in viral infections where host metabolic reprogramming is essential for viral replication and immune response. The correlation between HERV subfamilies with specific pathways can be explored further, for example, if HERVH has an influence on genes associated with metabolic pathways and whether they are closely associated with immune responses. However, our current analysis did not perform subfamily-specific analysis and may require further studies to explore such associations. Notably, in the most severe disease group, we observed enrichment of pathways related to O-glycan biosynthesis, glycosaminoglycan metabolism, and glycolipid metabolism, all of which play essential roles in cellular communication, immune signaling, and endothelial integrity. The neurological pathway enrichment, including neuroactive ligand–receptor interactions, suggests a possible connection between HERV activity and neuro-inflammation, which has been reported in severe dengue cases [62, 63]. This raises intriguing questions about whether HERV-derived regulatory elements contribute to the neurological complications observed in severe viral infections.

Our chromatin conformation analysis further provides mechanistic insights into how HERVs regulate gene expression. We chose K562 cells, as they are hematopoietic in origin and currently represent the closest available chromatin conformation dataset for our peripheral blood RNA-seq samples. We observed that many co-expressed HERV–gene pairs were within the same TADs, indicating that the spatial organization of chromatin may facilitate regulatory interactions. Notably, HERV family distribution within these TAD-associated pairs revealed that multiple classes of HERV families are represented, suggesting a broad involvement of endogenous retroelements in chromatin organization. However, a substantial proportion of these elements belong to long terminal repeat (LTR) retrotransposons, consistent with their known regulatory potential. Among these, members of the HERV-H family were also identified within TAD-associated regions, further supporting their potential role in chromatin domain formation [64]. Additionally, triplex-forming interactions between HERVs and gene promoters suggest that HERVs may directly influence transcription by engaging with promoter-bound cis-regulatory elements. The example of MAGI3, an immune-modulatory gene, highlights how HERV sequences may impact gene expression through physical promoter interactions (Fig. 6). Future work integrating immune cell–specific Hi-C data will be essential to accurately identify the TAD boundaries, which will further strengthen our hypothesis.

These findings support a broader model in which HERVs serve as functional regulatory elements within the genome, influencing transcriptional programs related to immune activation, metabolic adaptation, and disease progression. While our study provides strong evidence for the correlative relationship between HERV activity and host gene expression, further functional validation is needed to determine causal mechanisms. Importantly, we were able to replicate the major transcriptomic patterns of HERV activation in an independent clinical cohort, where over 98% of the expressed HERVs overlapped with our discovery dataset, and nearly 40% of the differentially expressed HERVs were also significantly altered in the validation cohort. Furthermore, a substantial proportion of non-significant HERVs still showed similar directional expression trends. This high degree of reproducibility not only strengthens the robustness of our findings but also reinforces the association between HERV activation and dengue disease severity.

The observed selective repression of HERV envelope-encoding genes in severe dengue adds a novel layer of complexity to the broader activation of the HERV transcriptome. While the bulk of differentially expressed HERV loci in severe dengue are noncoding LTR elements, our focused analysis reveals that env genes—many with known immunomodulatory and antiviral properties—are consistently downregulated. In both cohorts, we observed a general trend of decreasing HERV env gene expression with increasing disease severity. Although the two cohorts were stratified differently—by mild/moderate/severe disease categories in the first cohort and by negative/mild/severe groups in the second—the overall directionality of the trend was consistent. This suggests that reduced HERV env expression may be associated with disease progression. Given differences in cohort composition and patient heterogeneity, variability in the expression patterns of individual HERVs is expected. Nonetheless, the convergence of these trends across independent datasets strengthens the observation that HERV activity may reflect the host’s transcriptional adaptation to infection severity rather than being confined to cohort-specific effects.

Several HERV env proteins, such as Syncytin-1 (ERVW-1) and ERV3-1, have been implicated in immune tolerance, anti-inflammatory signaling, and even direct antiviral effects through membrane fusion interference or cytokine suppression [65–68, 69]. Downregulation of these genes in severe dengue may reflect a loss of an intrinsic immune regulatory mechanism, contributing to the exaggerated inflammatory response and cytokine storm observed in these patients. Notably, similar suppression of HERV env genes has been reported in severe COVID-19 [65], further supporting the idea that immune pathogenesis in severe viral infections may involve active silencing of HERV-derived regulators.

We also consider an alternative, testable hypothesis: dengue virus may actively suppress HERV env gene expression as a viral strategy to escape HERV-derived antiviral responses. This would align with their known role in restricting exogenous virus replication and dampening inflammation, potentially offering a selective advantage to the virus in promoting systemic infection and immune over-activation.

Additionally, as HERV env gene expression is enriched in leukocytes [70], the leukopenia commonly seen in severe dengue may also contribute to the observed reduction. However, the persistence of downregulation across samples suggests a more targeted regulatory mechanism beyond mere changes in cellular composition. Taken together, these findings highlight the importance of exploring HERV env gene regulation at single-cell resolution and in functional assays, to decipher whether their suppression is host-driven, virus-driven, or a combination of both. While our study strongly suggests HERV–gene co-expression, co-localization, and their association with dengue disease severity, functional validation remains necessary to determine whether HERVs causally influence transcriptional regulation. Although we strengthened our findings by examining HERV expression in an independent validation cohort, we acknowledge that, as with any observational clinical study, the findings remain correlative. Future mechanistic studies in human-relevant systems such as primary immune cell models will be essential to definitively establish causality. A key limitation is the lack of matched protein-level validation (e.g., cytokines or metabolic regulators) and also the vascular endothelial damage, such as D-dimer and vascular cell adhesion molecule (VCAM), which would further support the functional relevance of the observed transcriptomic changes. Future studies integrating transcriptomic and proteomic measurements, along with mechanistic investigations in human-relevant systems such as primary immune cell models, will be essential to establish causality. Additionally, investigating cell type-specific HERV expression using single-cell transcriptomics could provide more granular insights into their functional roles in the immune cells during disease progression. Another limitation of our study is the use of short-read RNA sequencing, which may not be optimal for accurately identifying and mapping long HERV loci, as these sequences can align to multiple genomic locations. Long-read sequencing approaches could provide higher resolution in distinguishing full-length HERV transcripts and their precise genomic context, further strengthening the mechanistic understanding of HERV function in disease.

Conclusion

Our findings suggest that HERVs may play an active role in immune and metabolic gene regulation, particularly in response to disease severity. By leveraging transcriptomic and chromatin interaction data, we provide a comprehensive framework for understanding how HERV elements may shape host gene expression vis-à-vis dengue disease severity through direct regulatory interactions. These insights pave the way for further mechanistic studies to explore how HERV activity contributes to immune modulation, metabolic adaptation, and disease progression.

Supplementary data

Supplementary data is available at NAR Molecular Medicine online.

Conflict of interest

All the authors affirm that there is no conflict of interest while conducting the study. We also confirm that the funding body did not have any role in planning, execution, and inferences drawn from the study.

Funding

This research was funded by the Bill and Melinda Gates Foundation, grant number INV-033578; and the Rockefeller Foundation, grant number 2021 HTH 018, to R.P.

Data availability

Sequence data used in this study are publicly available in NCBI SRA under the accession numbers with the BioProject numbers PRJNA1071729 and PRJNA1279769.