Ethics statement

The study was designed in accordance with the Declaration of Helsinki and was approved by the institutional ethics committee of the Council of Scientific & Industrial Research-Institute of Genomics and Integrative Biology, Delhi, India (Ref No: CSIR-IGIB/IHEC/2023–24/04). The patients provided their written informed consent before participation in this study.

Study design and sample processing

The study was conducted by utilising the serum samples collected from the patients reporting to the Max Super Speciality hospital, New Delhi, India, during June to November 2023. The serum samples from symptomatic dengue-suspect patients were collected at the hospital and subjected to an NS1-Ag test. A total of 4045 NS1-Ag-positive serum samples were brought to our laboratory at CSIR-Institute of Genomic and Integrative Biology (CSIR-IGIB), New Delhi, India, in a cold storage box on a daily basis. The transportation time was ranging from 45 to 60 minutes. On the same day, the serum samples were aseptically transferred to the 2 ml tubes and processed for the RNA isolation. and remaining serum samples were stored at -80°C for long-term preservation (in case required). To each serum sample, a unique barcode ID was assigned. The clinical data for a subset of those patients who underwent a complete blood count (CBC) and/or liver function test (LFT) were also obtained from the hospital EHR (Electronic Health Record). Out of 4045 patients from which we have NS1-Ag positive serum samples, 1756 CBC reports and 790 LFT reports were obtained and utilized in this study for the analysis.

Further, for the purpose of analytical clarity, we categorized our patient cohort, based on the local clinical parameter data, into mild, moderate, and clinically severe as reported earlier [23–25]. Following these earlier studies, we have investigated our clinical data for the classification of dengue patients. Across the 1756 CBC reports, 473 patients demonstrated both Total Leukocyte Count (TLC) and platelet count within normal range. Similarly, 405 patients were observed with low TLC and normal platelet count; whereas, 416 patients had both low TLC and low platelet count compared to the normal range. Thus, based on CBC reports of the patients, we were able to categorize the 1294 dengue patients into three groups: dengue without warning signs (n = 473, patients with normal platelet and leukocyte counts), dengue with warning signs comprising two subgroups—n = 405, patients with normal platelet but decreased leukocyte counts (leukopenia), and n = 416, patients with both low platelet counts (thrombocytopenia) and decreased leukocyte counts. The symptoms associated with severe dengue, such as severe bleeding, organ failure, or abnormal liver parameters, were not exhibited by any patients in our study cohort. In this context, the term clinically severe is used solely to describe the concurrent presence of both thrombocytopenia and leukopenia. For analytical clarity, the groups are designated as mild (patients without thrombocytopenia and leukopenia), moderate (patients without thrombocytopenia but with leukopenia), and clinically severe (patients with both thrombocytopenia and leukopenia) dengue cases. These classification criteria have been summarized in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Criteria based on clinical parameters for defining mild, moderate, and clinically severe dengue patients.

https://doi.org/10.1371/journal.pntd.0013034.t001

RNA Isolation and determination of DENV serotype

The serum sample was thawed in ice. From the 150 µl serum samples, total RNA was isolated using QIAamp Viral RNA Mini Kit (Qiagen, Cat No. 52906) following the manufacturer’s protocol (and stored at -80°C until further use). NanoDrop (Thermo Fisher Scientific, USA) was used to assess the quality and quantity of the RNA. A fraction of RNA was utilized to determine the serotype of the DENV through RT-PCR. The DENV serotyping was performed using the DENV serotyping kit following the manufacturer’s instructions, with positive and negative controls included (TRUPCR; Cat. No. 3B237).

Library preparation and whole genome sequencing of DENV-2

A total of 3702 DENV-2 identified using serotype-specific RT-PCR was further processed for whole genome sequencing using Oxford Nanopore Technology (ONT). Briefly, the RNA was reverse transcribed into cDNA using the LunaScript RT Supermix (New England Biolabs, Cat. No. E3010L). The DENV-2 genome was amplified using cDNA as a template. For the amplification of the DENV-2 genome, an amplicon-based tiling PCR approach was employed. Briefly, for the amplification of DENV-2 genomes, 39 sets of overlapping PCR primers adopted from Samuel et al., 2020 [26] were used. Primers were synthesized by Integrated DNA Technologies and optimized for accuracy before use. The sequences of PCR primers are listed in the S1 Table. The 39 sets of PCR primers were divided into two pools: the odd set (pool 1) and the even set (pool 2). The PCR was performed in two sets separately using the primers from pool 1 and pool 2. The amplification is achieved by Q5 High Fidelity 2X Master Mix (New England Biolabs, Cat. No. M0494L). Followed by the PCR, pool 1 and 2 PCR products were pooled together for each reaction. The PCR product was purified using Ampure XP (AXP) beads (Beckman Coulter, Cat. No. A63881). The concentration of purified PCR product was determined using a Qubit 4.0 fluorimeter (Invitrogen, USA). The purified PCR products were then subjected to library preparation. End-repair and A-tailing were performed using the NEBNext Ultra II End Repair/dA-Tailing Module (New England Biolabs, Cat. No. E7546L), followed by barcode ligation (Native Barcoding Kit 24v14; SQK-NBD114.24). The barcoded library was pooled together and processed for purification using Ampure XP (AXP) beads. The concentration of purified barcoded libraries was determined using a Qubit 4.0 fluorimeter (Invitrogen, USA). To the purified barcoded library, the native adapter was ligated (Native Barcoding Kit 24v14; SQK-NBD114.24), and the final purification was done using Ampure XP (AXP) beads. The concentration of the adapter-ligated library was determined using a Qubit 4.0 fluorimeter (Invitrogen, USA). The final purified library was loaded onto a flow cell (R10.4.1) and sequenced using a GridION device (ONT), followed by high-quality DENV-2 genome analysis.

Analysis of sequencing reads, genome assembly and variant calling

Following the completion of the sequencing runs on the GridION device, the raw ONT reads were transferred to high-performance computing (HPC). Briefly, raw reads were basecalled and demultiplexed through dorado basecaller with a phred Q-score quality of > 9. The demultiplexed high-quality passed reads were mapped to the DENV-2 reference (NC_001474) genome using minimap2 [27]. Additionally, primer trimming was carried out using align trim [28]. Further, artic mask [28] was used to mask the low-quality reads and taken as ‘coverage mask’ files. Primer depleted high-quality DENV-2 reads were taken for variant calling through Clair3 [29]. Post-VCF filtering was carried out using bcftools [30] by removing the SNVs having genotype quality less than 3 and depth less than 20 as ‘fail vcf’ and taking genotype quality greater than 3 and depth greater than 20 as ‘pass vcf’. Lastly, consensus FASTA was created by masking the positions of ‘fail vcf’ and ‘coverage mask’ and implementing the SNVs from ‘pass vcf’ to the DENV-2 reference genome using bcftools [30]. After merging the multiple VCFs, the preprocessing steps involve ‘IndelGap’ correction and ‘multiallelic’ variant resolution using bcftools, allowing us to extract a single variant from a single position. To facilitate downstream analysis, we applied variant normalization using ‘norm’ and ‘sort’, ensuring proper merging of SNVs and maintaining a consistent positional order from start to end. The depth and breadth of genome coverage were determined for the reads mapping to the DENV-2 wild-type genome (S2 Table). Samples which are greater than 50% genome coverage are merges and annotations for the SNVs obtained using Snpeff are mentioned in S3 Table.

Assessment of coverage and depth uniformity

We have generated a significantly high number of passed sequencing reads (with a Q score of 9), achieving a highest read of 2,23,095, with average depth of 1081X and an average genome coverage of 87.5 per Dengue virus genome. The highest genome coverage obtained was 98.8, which is commendable for direct clinical sample sequencing at a large scale. For basecalling and barcode demultiplexing, we utilized the dorado (v7.2) with the “hac” (high-accuracy basecalling) algorithm. This algorithm enhances sequencing accuracy by using a deep-learning model trained to minimize basecalling errors, ensuring high-fidelity read conversion from raw nanopore signals. Furthermore, Variant calling was carried out using Clair3, a deep-learning-based variant caller specifically designed for long-read sequencing. To ensure high-quality variant identification with uniform depth of coverage, we applied a stringent read-depth threshold, masking variants with depth below 20X.

Integrative analysis of SNVs across DENV-2 genome with clinical parameters across dengue patients

According to the disease sub-phenotype mentioned, the SNVs identified across DENV-2 genomes were categorized as mild, moderate, and clinically severe subgroups (S4 Table). In order to find the SNV burden in the genes, the presence and absence of SNVs in mild vs. moderate, moderate vs. clinically severe and mild vs. clinically severe were calculated (S5 Table). Further, chi-square test was applied on the obtained ‘presence-absence’ contingency table. Additionally, on the secured burden genes, the nonparametric fisher’s exact test was used to evaluate the independence between two categorical variables and to compare SNV profiles across the groups. Two-sided p-values were computed, with a significance threshold of 0.05. The direction and strength of the association between the SNVs and group classifications were assessed using the phi coefficient (φ). The chi-square and fisher exact tests were performed using the scipy.stats package in python. Phi correlation was performed using a psych library in R software (S6 Table).

Phylogenetic Analysis and Genotype detection

The DENV-2 reference sequence (NC_001474) was used to perform multiple sequence alignment with our cohort dataset above 90% coverage through Augur [31]. As these are direct sequencing of the clinical samples, we have kept the threshold at 90% coverage, so as not to miss important clinical implications. Prior to the phylogenetic analysis, we downloaded the DENV-2 genomes with a coverage of above 90% from GISAID on September 12, 2024. We obtained a total of 8085 such DENV-2 genomes. These genomes were submitted to GISAID from different countries/continents of the world (S7 Table). Specifically, in our global data set, the DENV-2 genome was represented from India (n = 1040), Africa (n = 556), Asia (n = 2921), the Caribbean (n = 29), Europe (n = 5), North America (n = 1025), Oceania (n = 75), and South America (n = 2934). Moving forward with the filter of continents having more than 100 genomes, which excludes Europe, the Caribbean, and Oceania. A total of 10164 genomes (Cohort Data = 2188, Global Data = 7976) were taken for phylogenetic analysis, using ‘iq-tree’ with the ‘maximum-likelihood’ algorithm. The tree was visualized using FigTree1.4.4 [32] as a cladogram with reference sequence as midpoints in increasing order nodes.

Evaluation of global frequency with significant SNVs

Further, to unravel the occurrence and frequency of the significant SNVs associated with disease severity across the globe, we conducted SNV analysis and frequency comparison across the genomes of DENV-2 reported worldwide. From the 7976 DENV-2 genomes, SNVs were called using the in-house Python code, and frequency was determined with respect to the continents. Additionally, comparisons of our statistically significant SNVs associated with mild and clinically severe phenotype were done using matrix plots.

DENV-2: A predominant serotype in circulation

RNA was isolated from 4045 serum samples collected from the NS1-Ag-positive patients in the Delhi-NCR (National Capital Region). To confirm the NS1-Ag positive results and identify the serotype of the DENV, serotype-specific RT-PCR was performed using RNA directly isolated from the clinical serum samples. The combined positives from NS1-Ag and RT-PCR have been used in this study. Out of 4045 RT-PCR performed, DENV-1 (n = 28), DENV-2 (n = 3702), DENV-3 (n = 45) and DENV-4 (n = 05) were identified. The remaining 265 were tested DENV-negative. Thus, the serotyping study indicates that the DENV-2 was the predominant serotype in circulation during the dengue outbreak in June to Nov 2023 in Delhi-NCR, India. As DENV-2 was identified as the predominant serotype (91.5%), with the objective of determining the genomic variability and potential genomic hotspots across the RT-PCR-confirmed DENV-2 population, whole genome sequencing of 3702 DENV-2 was performed using the ONT. Sequencing read analysis and mapping of the high-quality passed reads to a DENV-2 reference genome revealed differential genome coverage across the 3702 DENV-2 sequenced. Specifically, for a larger fraction (59.53%) of the DENV-2 population (n = 2189), 90–98.83% genome coverage was achieved. Moreover, for 434, the genome coverage was obtained in the range of 80–90%, whereas for 230, 167, and 235 DENV-2, 70–80%, 60–70%, and 50–60%, genome coverage was obtained, respectively (S2 Table). We were able to achieve genome coverage above 90 to 98.83% for a substantial number of the genomes. The differential genome coverage could be an attribute of the direct sequencing of the genomes from the clinical samples. It could also be a reflection of the virus copy number availability and integrity of the DENV’s RNA genome in the clinical serum samples. At the same time, it is important to note the necessity of genomic surveillance with relevant clinical implications. Overall, out of 3702, 3254 DENV-2 have greater than 50% genome coverage. Given the importance of capturing genomic variability across a larger set of DENV genomes as it could have a possible implication in disease outcome or clinical manifestation across the patients, further analyses were conducted across the DENV-2 population and demonstrated greater than 50% genome coverage (n = 3254) (Fig 1A). As GISAID provides a platform for the genomic sequence dissemination in the global community, we took forward this cut-off along with the clinical implications mentioned above.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Experimental workflow and SNVs spectrum across DENV-2 genomes.

(A) Study design includes NS1-Ag-positive serum sample collection, RNA isolation, and serotype detection, followed by whole genome sequencing, variant calling, and downstream data interpretation. This figure was created using licensed version of Biorender.com. (B) SNV spectrum with distribution of synonymous and non-synonymous SNVs across DENV-2 genomes whose coverage are > 50% (n = 3254). (C and D) Bar chart representing means of number of unique SNVs divided by gene length, a normalized count in the y-axis, with mean from bootstrap standard error estimation. (C) synonymous, and (D) non-synonymous SNV counts (x1000) across the DENV-2 genomes. (E) Horizontal stacked bar chart depicts the percentage proportion of synonymous and various types of non-synonymous SNVs—including missense, frameshift, and stop-gained—across different genes of DENV-2 genome.

https://doi.org/10.1371/journal.pntd.0013034.g001

Distinct Patterns of Synonymous and Nonsynonymous SNVs in DENV-2 genomes Linked with Viral Proliferation and Adaptation

Genetic variation may lead to a change in the properties and characteristics of the virus that could impact the epidemiology and associated virulence [33,34]. Therefore, we aim to investigate the variations in the genome of the DENV-2 population that caused an outbreak during June to November 2023 in Delhi-NCR, India. Variant analysis across the 3254 DENV-2 genomes yielded a total of 1618158 SNVs. All together, these SNVs were observed at a total of 3321 (out of 10723) genomic positions, representing 30.9% of the DENV-2 genome length. The total number of SNVs observed at a genomic position across our data set—the SNV spectrum—is represented by Fig 1B. As the variant analysis was conducted across the DENV-2 genomes with differential genome coverage, to ensure precise capturing of genomic variability, it is imperative to mention the SNV count across the DENV-2 genome vis-à-vis the genome coverage. Briefly, DENV-2 genomes with 90–98.83% coverage (n = 2189) exhibit SNV counts ranging from 437 to 741 per genome. Similarly, those with 80– < 90% coverage (n = 434) show SNV counts between 471 and 689 per genome. Genomes with 70– < 80% coverage (n = 230) display SNVs ranging from 345 to 575 per genome, while those with 60–70% coverage (n = 167) have SNV counts ranging from 310 to 531 per genome. Additionally, DENV-2 genomes with 50– < 60% coverage (n = 235) exhibit SNV counts between 203 and 478 per genome (S8 Table).

Furthermore, to understand the specific impact of SNVs on DENV virus proliferation and pathogenesis, it is crucial to investigate the distribution of the SNVs across various genic regions. Therefore, we assessed the gene-wise load of the SNVs in our data set. Overall, the highest numbers of SNVs were observed across the NS5 region, with the 2699 bp long NS5 region recording 882 SNVs. Further, 525 SNVs were observed across the NS3 region. Moreover, across the 1484 bp long E-gene, a total of 470 SNVs were observed. To capture the relative abundance of the variations with respect to the gene length, we have determined the gene length-normalized count of SNVs across the various genic regions of DENV-2 populations. The results indicate relative overrepresentation of synonymous SNVs in prM, NS2A, NS2B, and NS4A (Fig 1C). On the other hand, the overrepresentation of non-synonymous SNVs was identified across the ancC, prM, and M proteins (Fig 1D).

Delving deeper into the SNV spectrum, out of the 3321 unique SNVs across the DENV-2 genomes, the SNVs observed at 878 genomic positions potentially produce variation in the amino acid sequence of the polyproteins that were designated as non-synonymous SNVs. The non-synonymous SNVs identified in our study are further classified into frameshift, missense, and stop-gained variation. The gene-wise proportion of each type of SNV, such as synonymous, frameshift, missense, and stop-gained, is depicted in Fig 1E. As evident from Fig 1E, the proportion of non-synonymous SNVs was comparatively high in structural proteins (ancC, M), 3’UTR, and 5’ UTR, ensuing its role in viral entry, attachment, and replication of the viral genome. The proportion of synonymous SNVs appeared high in the E gene and non-structural genes, signifying its role in the adaptive evolution of the virus.

Significant Alterations in the TLC, Platelets, and Liver Enzymes across DENV-2 Infected-Patients

The complete blood count (CBC) profiles of patients with dengue were utilized to observe the deviations in clinical parameters upon the onset of dengue infection. Clinical parameters were categorized as low, normal, or high based on the WHO-recommended reference ranges. S9 Table summarizes the distribution of these parameters across the defined categories. As it is evident from the S9 Table, several key clinical parameters were low across the dengue patients. Specifically, 51.96% (n = 889) dengue patients exhibited low total leucocyte count (TLC), indicating the development of leukopenia across the dengue patients. A total of 42.65% (n = 749) dengue patients were observed to have a low platelet count, suggesting thrombocytopenia. Additionally, other clinical parameters, such as haemoglobin, mean corpuscular volume (MCV), and mean corpuscular hemoglobin (MCH), were low in approx. 25% of the dengue patients, pointing to the development of anaemia associated with the DENV infection.

Moreover, the red cell distribution width (RDW) appears high in 52.75% (n = 901) patients, further supporting the diagnosis of anemia. Liver function test (LFT) data, available for a subset of patients, revealed elevated levels of SGOT-Aspartate Transaminase (AST) in 71.54% (n = 538) and SGPT-Alanine Transaminase (ALT) in 55.44% (n = 438) of dengue patients, markers often associated with severe dengue. Kruskal-Wallis with Dunn’s comparison was performed to determine the significance of the deviations in clinical parameters from the normal range. Statistical analysis confirmed that deviations in clinical parameters across dengue patients were significant, underscoring the notable changes in blood parameters during dengue infection. At the same time, it highlights the need for evaluating whether differential clinical parameters are correlated with differential disease severities and potentially DENV-2 genomic SNVs within those patients.

Leukopenia, thrombocytopenia and elevated liver enzymes differentiates dengue severity

As reported earlier [23–25], for analytical clarity we have been able to segregate 1294 dengue patients into mild (n = 473), moderate (n = 405), and clinically severe (n = 416) subgroups (Fig 2A), based on the available clinical data of the patients. Importantly, we observed a significant reduction in the TLC and platelet count in the clinically severe subgroup as compared to the patients from mild and moderate phenotype, highlighting a strong association of leukopenia and thrombocytopenia with increased disease severity (Fig 2B and 2C). Furthermore, a significant elevation in the hemoglobin levels was observed in clinically severe compared to the moderate sub phenotype, potentially reflecting hemo-concentration due to plasma leakage, a hallmark of severe dengue [35] (Fig 2D). In line with the previous findings [23], clinically severe subgroup exhibited a marked decrease in the neutrophil levels accompanied by an increase in the lymphocyte levels compared to the mild phenotype, further underscoring the dysregulated immune response in clinically severe dengue sub phenotype (Fig 2E and 2F). Several CBC parameters, including mean MCH, mean corpuscular hemoglobin concentration (MCHC), MCV, and mean platelet volume (MPV), were significantly elevated in the clinically severe phenotype compared to both mild and moderate subgroups, suggesting alterations in the erythropoietic and platelet homeostasis during severe disease [36]. However, no significant differences were observed in packed cell volume (PCV), red blood cell count (RBC), or red cell distribution width (RDW) across the severity subgroups, indicating a degree of stability in these parameters despite disease progression (S1 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Dengue severity subgrouping and comparative account of clinical manifestations.

(A) Schematic representation of the dengue patient categorization based on two primary CBC parameters: total leukocyte count (TLC) and platelet counts (thrombocytes) into mild, moderate, and clinically severe, and their association with fluctuations in CBC and LFT parameters during dengue virus infection. This figure was created using licensed version of Biorender.com. (B-F) Statistical comparison across the patient groups associated with mild, moderate, and clinically severe sub phenotypes for the major CBC parameters, including (B) total leukocyte count (10⁹/L); (C) platelets (10⁹/L); (D) hemoglobin (g/dl); (E) neutrophils (%); and (F) lymphocytes (%). (G-K) Statistical comparison across the patient groups associated with mild, moderate, and clinically severe dengue sub phenotypes for major LFT parameters, including (G) bilirubin (direct) (mg/dl); (H) bilirubin (total) (mg/dl); (I) ALT (U/L); (J) AST (U/L); and (K) GGTP (U/L). Kruskal-Wallis with Dunn’s comparison was performed for comparison of each clinical parameter across mild, moderate, and clinically severe. The significance value is denoted as *, where * indicates p ≤ 0.05, ** indicates p ≤ 0.01, *** indicates p ≤ 0.001, and **** indicates p ≤ 0.0001.

https://doi.org/10.1371/journal.pntd.0013034.g002

Previous studies have highlighted deviations in liver function parameters during dengue virus infection, which often correlate with the disease severity and systemic inflammation [9,37,38]. Consistent with this, the liver function test (LFT) revealed significant alterations in the liver profile of patients associated with clinically severe phenotype compared to the mild and moderate phenotypes. Clinically severe subgroup showed marked increase in bilirubin (both total and direct), SGOT-Aspartate Transaminase (AST), SGPT-Alanine Transaminase (ALT), and GGTP (Gamma GT) serum compared to mild and moderate sub phenotypes, indicating hepatic dysfunction potentially driven by the immune-mediated liver injury or direct viral effects (Fig 2G-2K). However, no significant differences were observed in the albumin-to-globulin (A/G) ratio or alkaline phosphatase levels across the severity subgroups, suggesting these parameters might remain unaffected by the disease progression (S1 Fig). Together, these findings reinforce the critical role of haematological and hepatic parameters in delineating the pathophysiology of the patients associated with clinically severe phenotype.

Monophyletic clustering of cosmopolitan DENV-2 genotypes from India

DENV-2 is classified into six genotypes: Asian I, Asian II, American, American-Asian, Cosmopolitan, and Sylvatic, with the Cosmopolitan genotype being the most prevalent in India [39]. To determine the genotype of the DENV-2 strains in our study, we conducted genotype identification and observed that DENV-2 genomes in our study cohort exclusively belonged to the Cosmopolitan genotype.

Further, a comprehensive phylogenetic analysis was performed to investigate the relationship of the in-house sequenced DENV-2 genomes with the global strains. For this, 7,976 DENV-2 genomes with >90% coverage were retrieved from GISAID, representing diverse geographical regions, including India, Asia, Africa, North America, and South America. The analysis revealed distinct phylogenetic clusters reflecting global diversity in the DENV-2 strains (Fig 3). Our in-house genomes clustered as a separate Clade indicating the independent evolutionary pattern of the DENV-2 population from India. Further, it appears that the DENV-2 genomes captured from the outbreak in India were monophyletic and uniform cosmopolitan genotypes, thus providing a more rational DENV-2 population for genetic variability analysis. We further investigated potential clustering of the DENV-2 genomes across the severity groupings, however no possible clusters were identified (Fig 4) These findings highlight the distinct genetic signatures of the in-house DENV-2 strains and their divergence from globally circulating genomes.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Phylogenetic analysis of 10164 DENV-2 genome sequences.

The distribution of genomes of different continents from various outbreaks, along with the in-house sequences from a single outbreak, is represented using distinct colors. The color codes are as follows: mild (green), moderate (orange), clinically severe (magenta), cohort data (sea green), India (salmon red), Asia (pink), Africa (cyan), North America (yellow), and South America (yellow-green).

https://doi.org/10.1371/journal.pntd.0013034.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Phylogenetic analysis for the DENV-2 genomes associated with different disease sub-phenotypes.

A magnified view of phylogenetic tree for 929 genomes comprising clinical sub-phenotypes mild (light green), moderate (orange), and clinically severe (magenta).

https://doi.org/10.1371/journal.pntd.0013034.g004

Distinct genomic hotspots in E, NS4B and NS5 regions underpin dengue disease severity

Aligning with the theme of our study, we investigated the possible association of SNVs observed across the DENV-2 genomes and the clinical parameters of the dengue patients. As mentioned in the study design, the SNVs association study was conducted across 1,294 patients with detailed clinical data. The analytical workflow for the SNVs association study is depicted in Fig 5A, whereas Fig 5B explains the genome coverage vis-à-vis the SNV count in 1,294 DENV-2 genomes. Across the 1,294 DENV-2 genomes, SNVs were identified at 2,496 genomic positions. Many of these SNVs were commonly distributed across the mild, moderate, and clinically severe subgroups, while several SNVs were uniquely distributed in each subgroup, underscoring potentially distinct genomic hotspots/signatures linked to disease severity (Fig 5C). Specifically, the mild subgroup exhibited 485 unique SNVs not observed in the moderate or clinically severe patients. Similarly, the moderate group had 333 unique SNVs, while the phenotype associated with clinically severe subgroup demonstrated 308 exclusive SNVs (Fig 5C). The frequency of the SNVs associated with mild, moderate, and clinically severe phenotypes are depicted in Fig 5D.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Association of SNVs with differential disease severity.

(A) The analytical workflow of the SNV association studies, showing a representation of clinico-genomic integration of mild, moderate, and clinically severe subgroups with differential genomic features, that is depicted through differential SNV burden in the genes for different categories. This figure was created using licensed version of Biorender.com. (B) Joint-grid plot depicting genome coverage and SNVs count across mild, moderate, and clinically severe sub phenotypes depicting a positive correlation for higher number of SNVs for higher coverage. (C) Venn diagram demonstrating the unique and common SNVs across the mild, moderate, and clinically severe patients. (D) Frequency of SNVs in mild, moderate, and clinically severe groups across the DENV-2 genomes has been smoothed by moving average in intervals of 10, with the average represented by square markers. (E) Bar plot depicting significant burden of SNVs across the gene of DENV-2, with the significance shown in the asterisk symbol obtained from the chi-square test of independence.

https://doi.org/10.1371/journal.pntd.0013034.g005

Further, we aim to explore the involvement of SNVs in the clinical manifestation of the DENV-2 infected patients. For this analysis, the presence and absence of SNVs in the DENV-2 genes across the phenotypes associated with mild, moderate, and clinically severe groups was tabulated, and a chi-square test was applied (S5 Table). We observed the significant SNV burden across the E, NS2A, NS4B, NS5, and 3’UTR for the mild vs. clinically severe comparison, indicating their role in differential severity, whereas for the mild vs. moderate group comparison, 5’UTR, M, E, NS1, NS2A, NS4B, and NS5 were observed having the significant burden of SNVs. However, only 3’UTR was noted for the presence of SNV burden in moderate vs. clinically severe sub phenotypes. To further investigate the significance and potential functional relevance of SNVs distributed across these significant genes, fisher’s exact test and phi-coefficient correlation analyses were performed. Our analysis identifies statistically significant SNVs associated with each severity category. Comparisons between mild and moderate subgroup of patients revealed 14 significant SNVs that were uniquely associated with the mild patients (Fig 5E). The mild vs. clinically severe group analysis yielded 83 significant SNVs, of which 56 were associated with mild cases and 27 with clinically severe patients (Fig 5E). However, no SNVs were found significant between moderate and clinically severe patients. Interestingly, the 14 SNVs significantly associated with mild group across mild vs. moderate, were also observed in the mild vs. clinically severe group, indicating exclusive association of these SNVs with mild phenotypes. These results provide evidence of distinct SNV patterns correlating with disease severity.

To assess the functionality and potential biological significance of statistically significant SNVs across the severity subgroups, we looked into the distribution of these SNVs across the DENV-2 genome. This analysis revealed intriguing clustering patterns of SNVs within the genic regions that differed between the severity subgroups. Fig 6 demonstrates the distribution of significant SNVs associated with each severity subgroup. For the mild vs. moderate group comparison, all significant SNVs associated were predominantly clustered in the NS4B regions in the mild patients. However, we did not observe any significant SNVs associated with the moderate group for the same comparison. For the moderate vs. clinically severe group analysis, none of the SNVs were observed to be significantly associated with any of the phenotypes. Across the comparison of mild vs. clinically severe phenotypes, 56 significant SNVs associated with mild group were distributed across E, NS4B, NS5, and 3’UTR regions, with the majority of the SNVs cluttered in the NS4B and NS5 regions. Interestingly, of these 56 SNVs associated with mild group across the mild vs. clinically severe comparison, 14 were common for the mild vs. moderate group and were exclusively present in the NS4B region. For the mild vs. clinically severe comparison, the clinically severe patients revealed unique associations of distinct SNVs distributed across the E, NS2A, NS4B, NS5, and 3’UTR regions, with the E and NS5 genes having their predominant contribution of SNVs towards the clinically severe phenotype.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Distribution of significant SNVs across disease severity.

Lollipop plot depicting significant SNVs across mild and clinically severe phenotypes. SNV is represented with different colors, signifying individual groups. Significance was calculated using fisher’s exact test (p < 0.05). The directionality and strength of association of significant SNVs with respective groups were confirmed with the phi coefficient correlation test.

https://doi.org/10.1371/journal.pntd.0013034.g006

Our domain-wise analysis of SNVs in the E-gene reveals a clear demarcation in the distribution of SNVs associated with clinically severe and mild phenotype, highlighting their potential functional significance in disease severity. SNVs associated with clinically severe phenotype predominantly occur in the DI-DIII region [A1335G, G1374A, G1503A, C1713T, A1830G, A2016G, and C2026T], which is critical for receptor binding, membrane fusion, and viral entry [40]. In contrast, SNVs associated with mild sub-phenotype are localized in the DIII and stem regions [C2073T, G2085A, A2105G, T2124C, G2157A, and G2160A], which are primarily involved in the virus assembly and host cell recognition [41]. The non-synonymous SNVs C2073T and A2105G result in amino acid changes in the viral proteins, while synonymous SNVs are likely to affect viral processes through codon usage bias, translation efficiency, transcript stability and interactions with host factors. Notably, a study by Luciana et al. [42] highlights the selection of synonymous mutations that favor DENV’s preferred nonoptimal codons, which enhance viral fitness. Further targeted studies in this direction would further strengthen these potential functional understanding of the synonymous mutations. Additionally, in the NS4B and NS5 regions, we observe a mixture of SNV associations, suggesting diverse functional impacts on viral replication, immune modulation, and overall disease progression. Notably, the non-synonymous SNVs modulate function through alteration in the amino acid sequence of the protein and could have detrimental effects for the host. Therefore, we specifically looked further into the non-synonymous SNVs and their possible association with the dengue severity across different disease comparison subgroups. We identified a total of 5 non-synonymous SNVs (E: Asn390Ser; NS4B: Leu13Phe, Ala19Thr, Val48Ile and 3’UTR: X3475Trp) associated with mild phenotype. Further, two non-synonymous SNVs (E: Pro364Ser and NS2A: Lys135Arg) were observed to be associated with clinically severe phenotype. Altogether, these findings underscore the distinct genomic signatures associated with dengue severity, highlighting the E, NS4B, and NS5 regions as critical hotspots for SNVs in dengue differential severity.

Statistically significant SNVs showing global frequency flip for mild and clinically severe patients

To understand the occurrence and frequency of the significant SNVs associated with the disease severity across the globe, we have conducted SNVs analysis and frequency comparison across the genomes of DENV-2 reported worldwide. Then we looked for the statistically significant SNVs across the global dataset. We observed the global occurrence of these SNVs with varying frequencies in the country cohort across the globe. The frequency comparison of the statistically significant SNVs with the global SNVs was represented through a matrix plot (Fig 7). Interestingly, the significant SNVs associated with mild phenotype are having uniform frequency in the minimum of two countries, while clinically severe phenotype have frequency flip in all the SNVs.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Comparison of statistically significant SNVs associated with mild and clinically severe phenotype with global frequency (A-B).

The matrix plot depicts the comparison of significant SNVs identified in the of DENV-2 genomes from the study with the global frequencies reported worldwide A) Mild, and B) Clinically severe.

https://doi.org/10.1371/journal.pntd.0013034.g007