Patients and setting
This research was conducted at Hospital Sultanah Aminah Johor Bahru (HSAJB), a 989-bedded hospital that serves as the main tertiary referral centre of Southern Malaysia, with its patient population reflecting the larger community in Malaysia. The period of study was from January to April 2014, coinciding with one of the peaks of DENV outbreaks. During the study period, all hospitalized patients with a positive non-structural protein 1 (NS1) antigen were identified from the microbiology laboratory database, HSAJB. The initial NS1 testing was done at the microbiology laboratory of HSAJB, using a commercially available rapid dengue diagnostic kit; SD BIOLINE Dengue Duo combo device (Standard Diagnostic Inc., Korea). Secondary DENV infections were detected using Panbio Dengue IgG Capture ELISA, which has incorporated a cut-off value of > 22 Panbio Units, equivalent to HAI level of 1:2560, indicative of secondary infections .
Clinical data was retrospectively collected by reviewing the medical case notes, microbiology, haematology and biochemical laboratory results. The clinical data retrieved on admission included demography, vital signs, underlying comorbidities, signs and symptoms, haematological, liver and renal function parameters. Warning signs and severe dengue manifestations were recorded throughout the hospital stay. In addition, nadir platelet counts and results of dengue serology were also noted.
Approval was obtained from the Medical Research Ethics Committee, Ministry of Health Malaysia (NMRR-14-617-21061). Informed consent was not obtained from the patients, as this was a retrospective study and data was analyzed anonymously.
In total 290 patients (non-duplicate) with NS1 antigen positive were identified. Their serum samples were stored at -80 °C for further testing. These tests were conducted at an infectious diseases research laboratory at Monash University Malaysia.
Patients’ serum samples and extraction of viral DNA
Viral RNA was extracted from 200 μl of the original serum using QIAamp viral RNA Mini Kit (Qiagen, Germany) according to the manufacturer’s instructions. Extracted RNA was stored either at -80 °C or used for RT-PCR immediately. Complementary DNA (cDNA) from viral RNA was synthesized by reverse transcription using AccessQuick RT-PCR System kit (Promega, USA). The RT mixture consisted of 10 μl (20–50 ng) of extracted RNA, 1 unit of reverse transcriptase enzyme, 12.5 μl of AccessQuick mastermix (2x), 1 μl of random primer and 20 U of RNase inhibitor (RNaseOUT, Invitrogen) in a final volume of 20 μl. The RT mixture was incubated at 65 °C for 5 min (min) followed by 37 °C for 1 h (h) and 72 °C for 5 min. The prepared cDNA was used for multiplex PCR.
DENV serotypes were determined using multiplex PCR , which amplified specific target regions using a forward conserved 5’UTR primer and four reverse primers targeting specific regions of the M and C genes of respective DENV-1, -2, -3 and -4 serotypes. To ensure the specificity of the primers to DENV and the absence of cross-reactivity with related flaviviruses, the primers were blasted through the National Centre for Biotechnology Information database . Amplifications were performed as described and the expected size of each of the amplicons was as follows: DENV-1:342 bps, DENV-2: 251 bps, DENV-3: 538 bps and DENV-4: 754 bps. To perform PCR, a primer mix was prepared by mixing 400 nM of forward conserved primer and 200 nM of each reversed primer with appropriate volume of DEPC-treated distilled water. The premix was added to PCR buffer containing 1.5 mM MgCl2, 0.2 mM of each of the dNTPs, 5U of Taq polymerase and 2 μl of viral cDNA. The thermal cycling profile of this assay consisted of 35 cycles of PCR at 95 °C denaturation for 30 s (s), 60 °C of annealing for 30 s and 72 °C extension for 1 min . PCR contamination was avoided by spatially separating the RNA extraction, cDNA preparation and amplification steps. In order to detect possible contamination, a no template negative control was incorporated in all the PCR reactions.
All samples were also subjected to RT-PCR for Chikungunya virus based on its non-structural protein 1 (nsP1) and glycoprotein E1 (E1) genes . Chikungunya virus infections are relatively common in Malaysia and can mimic DENV infections in clinical presentations. PCR was performed in a Mastercycler gradient machine (Eppendorf, Hamburg, Germany).
Gel elution and sequencing of amplicons
The detection and identification of DENV direct from serum samples by RT-PCR was accomplished based on the product size of the amplified-serotype specific amplicons by electrophoreses in a 1.5-2 % agarose gel stained with ethidium bromide. PCR products were cut from the gel, extracted using the QIAquick Gel Extraction kit (Qiagen, Germany) and were directly sequenced in both forward and reverse directions using the specific primers by a commercial sequencing services (First base, Singapore). Random amplicons of DENVs (DENV-1: 30, DENV-2: 30, DENV-3: 13 and DENV-4:1) were selected from the PCR reactions that showed both single and dual DENV infections. The identities of the sequences were confirmed by Basic Local Alignment Search Tool (BLAST). The sequences obtained in the present study and other sequences retrieved from the GenBank were aligned in ClustalW (2.1).
Virus propagation in C6/36 cells and total viral RNA extraction for next generation sequencing (NGS)
Confluent Aedes albopictus C6/36 monolayer cells were grown and maintained in minimum essential medium (MEM) supplemented with 2 % fetal bovine serum (FBS), HEPES buffer and 1 % penicillin/streptomycin (100 U/mL penicillin, 100 μg/mL streptomycin; Gibco®; USA). Virus isolation was performed by inoculating 50 μl of original serum onto C6/36 monolayer cells in Leighton tubes which were incubated at 30 °C for 7 to 10 days for growth of viruses. Viral RNA was extracted from 200 μl of the first blind passage of the serum-infected C6/36 culture supernatant and cDNA was synthesized using the method described above. The cDNA was used for multiplex PCR  and NGS. The amplicons derived from the multiplex PCR of supernatant of the first passage of the C6/36 infected cells were compared with those derived directly from serum. To confirm the DENV serotypes and to determine the heterogeneity of these viruses, random samples of five DENV-1 and six DENV-2 from mono-infected samples were subjected to NGS.
Whole genome sequencing of DENV
Synthesized cDNA was converted into double stranded DNA using NEBNext® mRNA Second Strand Synthesis Module (New England Biolabs, Ipwich, MA) according to the manufacturer’s instructions. The reaction was purified using Ampure bead XP (0.8× vol. ratio), normalized to 0.2 ng/uL based on Qubit quantification (Invitrogen, Carlsbad, CA) and tagmented with Nextera XT (Illumina, San Diego, CA) according to the manufacturer’s instructions for small insert size library. The constructed libraries were quantified, normalized and sequenced on the MiSeq sequencer located at the Monash University Malaysia Genomics Facility (run configuration of 2 × 150 bps paired-end read). Reference mapping to the complete genome of DENV was performed using MITObim version 1.8 (default setting) . The assembled genomes of 6 DENV-2 and 5 DENV-1 (DENV-2: TM26, TM78, TM181, TM198, TM213, TM296; DENV-1: TM24, TM50, TM99, TM100, TM242) along with additional closely related genomes of DENV isolated from the South East Asia and Oceania regions were used to infer evolutionary relationship. Nucleotide alignment based whole genome sequence was performed using MAFFT v7.127b (default alignment setting) and a maximum likelihood phylogenetic tree was constructed using FastTree version 2.1.8 with the Jukes-Cantor + CAT model [22, 23]. Tree visualization and editing was performed using FigTree v1.4.1 (http://tree.bio.ed.ac.uk/software/figtree/). Further classification of genotypes in each serotype was determined using the Genotype Determination and Recombination Detection tool on Virus Pathogen Resource (http://www.viprbrc.org/brc/genotypeRecombination.spg?method=ShowCleanInputPage&decorator=flavi_dengue).
Warning signs (WS) assessed included abdominal pain or tenderness, persistent vomiting (≥2 consecutive days), clinical fluid accumulation, mucosal bleeding, hepatomegaly (>2 cm) and haematocrit rise concurrent with a rapid decrease in platelet counts . We chose to exclude lethargy as a WS due to ambiguity in patients’ perception of lethargy and lack of objective distinction from tiredness . The definition for severe dengue was obtained from the WHO 2009 criteria  with minor modifications and comprised at least one of the three criteria:
Severe plasma leakage leading to shock (narrowing of pulse pressure to ≤ 20 mmHg, systolic blood pressure < 90 mm Hg or the presence of signs of poor capillary perfusion such as cold extremities, poor capillary refill or tachycardia) [3, 4] or fluid accumulation with respiratory distress (respiratory rate ≥30/min with oxygen saturation ≤ 92 % on room air, or requiring mechanical ventilation) .
Severe bleeding was defined as bleeding with hemodynamic instability that requires fluid replacement for shock and/or whole blood or packed cell transfusion or any life threatening bleed, e.g. haematemesis, melaena or intracranial bleed .
Severe organ impairment comprised severe liver impairment (aspartate aminotransferase or alanine aminotransferase ≥1000 IU/L), encephalopathy, myocarditis  or acute renal impairment (Stage 2 Acute Kidney Injury) [26, 27].
Based on population background study conducted in Malaysia, the haematocrit parameters used to evaluate haemoconcentration were >40 % in female adults, > 46 % in male ≤ 60 years, > 42 % in male > 60 years and > 38 % in children . Leukopenia was defined as leukocyte count < 4,000/mm3 and thrombocytopenia as platelet count <150,000/mm3. Severe thrombocytopenia was referred to as platelet count < 50,000/mm3, a value shown to be associated with additional severe manifestations . Paediatric patients were defined as patients aged less than 18 years. Secondary DENV infections categorization was based on the results of Panbio dengue IgG capture ELISA . Pleural effusion or ascites was diagnosed based on conventional x-rays or ultrasound of the thorax and abdominal region. Diarrhoea was defined as the passage of three or more loose stools per day . The simultaneous detection of more than one DENV serotypes was classified as co-infection, in contrast to mono-infection where only one DENV serotype was identified.
Data was analyzed using the Statistical Package for Social Sciences (SPSS version 20.0); comparing patients with and without DENV co-infections. To further pinpoint the variances in clinical and laboratory findings attributable to a particular serotype and its co-infection, subgroup analysis (DENV-1 and DENV-2 with its respective co-infection) was performed. Similar analysis was not conducted for DENV-3 and DENV-4 as the numbers were too small for valid statistical comparison.
Categorical variables were expressed as numbers and percentages and comparison amongst variables was determined by the Fisher’s exact test or Chi-squared test. Continuous variables were expressed as median ± interquartile range (IQR) and comparison was made using the non-parametric Mann–Whitney test. The odds ratio (OR) and its 95 % confidence intervals (CI) were calculated. The p-value < 0.05 (two-tailed) was taken as the level of significance. We then performed a multivariate logistic regression analysis by including clinical manifestations and laboratory parameters which were significant in univariate analysis (P < 0.05), to evaluate the factors independently associated with co-infections. To obtain more reliable results, variables with more than 5 % of missing data were excluded from the final model.