This study included 8,951 serum samples received by the Rio Grande do Sul Central Laboratory of Public Health (LACEN RS) from 2015 to 2023 for DENV detection and serotyping. Ribonucleic acid (RNA) was isolated from samples using automated extraction on Extracta 96 equipment (Loccus), using the Fast Kit – Viral DNA and RNA (MVXA-P096 FAST) (Loccus, São Paulo, Brazil) according to manufacturer's instructions.

DENV detection and serotyping were performed through reverse transcriptionquantitative polymerase chain reaction (RT-qPCR) using the Kit Molecular ZDC (BioManguinhos, Brazil) according to the manufacturer's instructions. Samples showing amplification with RT-qPCR cycle threshold (Ct) ≤38 were considered DENV-positive; positive and negative controls were used in each analysis. The multiple detection of the four known DENV serotypes (DENV-1 to 4) was performed in a CFX Opus Real-Time PCR System (Bio-Rad, Hercules, CA, USA).

Extracted RNA from 667 samples showing amplification with RT-qPCR Ct ≤28 was used for genomic sequencing. The genomic libraries were prepared using an adaptation of the Illumina COVIDseq protocol (https://www.illumina.com/products/by-type/ivdproducts/covidseq.html), replacing the primers by previously published DENV primers sets (Brito et al., 2021; Gräf et al., 2023) . Sequencing was performed using Illumina Miseq (Illumina Inc.). A reference-based genome assembly was conducted using the ViralFlow 1.0 pipeline (Dezordi et al., 2022a) , using the closest and most well-annotated DENV-1 (GenBank: GU131863.1) and DENV-2 (GenBank: MW577818.1) reference genomes.

The intrahost variant analysis was also performed using the ViralFlow pipeline, as previously described (Dezordi et al., 2022b) . In brief, after consensus generation, the bamreadcount tool extracts the proportion of each base (A, C, T, G) at every position from the bam file. Subsequently, an in-house Python script (intrahost.py) is employed to identify intrahost single nucleotide variant (iSNV) sites based on specific criteria. The defined rules for iSNV identification include requiring the minor variant (MinV) to constitute at least 5% of the total position depth. Additionally, it should be present in both sense and antisense reads, with each contributing at least 5% to the overall variant frequency and a minimum depth of 100 reads. Following identifying iSNV sites, two consensus genomes are generated as output. The major variant (MajV) is derived from the nucleotide in the major allele frequency at each genomic position. On the other hand, the MinV is constructed based on the nucleotide found within the minor allele frequency at the same positions.

Our samples were analyzed using the Genome Detective Virus Tool Version 2.72 (https://www.genomedetective.com) to associate genotypes belonging to each sample. To characterize the transmission chains of the DENV clades, we downloaded all near complete DENV-1 and DENV-2 genome sequences with coverage breadth > 70% from NCBI (https://www.ncbi.nlm.nih.gov/), resulting in a data set of 3.475 DENV-1 and 3.090 DENV2 samples. Genomes were aligned with sequences generated in this study using MAFFT (Katoh et al., 2019) and manually curated to remove artifacts using AliView (Larsson, 2014) . Maximum likelihood (ML) phylogenetic trees were estimated using IQ-TREE2 (Minh et al., 2020) under the General Time Reversible (GTR) nucleotide substitution model, which was inferred as the best-fit model by the ModelFinder application implemented in IQ-TREE2 (Kalyaanamoorthy et al., 2017) . Statistical support for tree nodes and branches was estimated using aLRT and ultrafast bootstrap approximation approaches with 1000 replicates. To understand the phylogenetic placement of MajV and MinV genomes, we added both genomes to our datasets to reconstruct an ML tree, including the sequences collected from GenBank.

In the studied period (2015-2023) Rio Grande do Sul state registered 1,023,366 notified Dengue cases, totalizing 177,009 confirmed cases distributed in the following years: 2015 (1,299 cases), 2016 (2,442 cases), 2017 (24 cases), 2018 (28 cases), 2019 (1,346 cases), 2020 (3,638 cases), 2021 (10,601 cases), 2022 (67,326 cases) and 2023 (38,357 cases).

The Rio Grande do Sul Public Health Laboratory identified 3,984 DENV-positive samples using the RT-qPCR method over the nine years studied. The samples were sourced from individuals with suspected acute febrile illness from 2015 to 2023, collected across 239 cities in the state, and exhibited amplification with RT-qPCR Ct of ≤38. We obtained raw reads of 627 samples diagnosed with DENV-1 infection and 40 from DENV-2 infection, totaling 667 samples investigated through whole genome deep sequencing. The average coverage breadth across the genome from all samples ranged from 72.23% to 99.84%, and an average coverage depth of 1817.46 (SD=908.59), ranging from 194.0 to 6,625.5.

During the study period, we also identified the coinfection of serotypes DENV-1 and DENV-2 identified by RT-qPCR in a sample from a 52-year-old female patient without risk factors. The patient lives in a northwest municipality of the Rio Grande do Sul state, where there was the simultaneous transmission of DENV-1 and DENV-2 serotypes. We sequenced both patient genomes to confirm serotype coinfection and performed phylogenetic reconstruction using maximum likelihood. The genome coverage breadth was 99.22% for DENV-1 and 91.57% for DENV-2. Additionally, the average coverage depth was 286.0 for DENV-1 and 770.0 for DENV-2. The results confirmed the coinfection by allocating each genome to its corresponding serotype in the tree, with maximum bootstrap support values for the two main branches representing the two serotypes, as shown in Figure 1. More data on samples Denv1_16248_2023 and Denv2_16248_2023 are presented in the following discussion and Table 1.

Among the 667 genomes, 665 (99.70%) displayed at least one iSNV site, that is, at least one genomic position with more than 100 reads supporting a minimum of two alternative nucleotides. For DENV-1, an average of 20 iSNVs sites per sample was observed. For DENV-2, an average of 14 iSNVs was observed in the samples analyzed. Most samples (663) showed a range between 1 to 88 iSNVs. Interestingly, four DENV-1 samples showed a higher number of iSNVs between 212 and 295, suggesting possible coinfection, with a notable distance iSNV number gap between those and the remaining 663 samples. Notably, in the case of the serotype coinfection sample, both genomes Denv1_16248_2023 and Denv2_16248_2023 exhibited low iSNVs values (8 and 12, respectively), indicating a low intrahost diversity. The iSNVs values for all samples are available in Supplementary Table 1.

With few exceptions, most studies estimating DENV iSNV diversity have used cloning and Sanger sequencing (Descloux et al., 2009; Parameswaran et al., 2012; Wang et al., 2002) . By comparing our data, it is suggested that the intrahost genetic variability of DENV is lower than previously estimated, as determined by Sanger sequencing, probably due to the more stringent parameters to validate low-frequency variants obtained through deep sequencing. For DENV-2, the intra-host variability found in this study showed a range of 2 to 43 iSNVs, reinforcing previous findings from deep sequencing, where a range of 10 to 44 iSNV was found for the complete DENV-2 genome (Romano et al., 2013) . With the lack of comparable data from deep sequencing, we found a study evaluating intrahost variability for DENV-1 using cloning and Sanger sequencing, where they applied rigorous methods to validate low-frequency variants, obtaining an average of 15.61 iSNVs leading to a result consistent with the reports in this study for DENV-1 (Thai et al., 2012) .

The unusually high number of iSNVs found in four DENV-1 samples may be derived from de novos SNPs generated during viral replication or due to simultaneous infections of clades having enough differentiating SNPs. Both phenomena would result in high iSNV, but using the MajV and MinV genomes generated, we sought to distinguish between these hypotheses through phylogenetic analysis of minor and major reconstructed genomes. Coinfection is expected to generate two substantially different genomes that confidently cluster within different clades, while de novo-generated minor and major reconstructed genomes are not. These four outlier samples and serotype coinfection data are described in Table 1. To assess if codetection could result from sample contamination, we assessed the negative controls from extraction to sequencing runs present in the runs of these four outlier samples. All the negative control assemblies revealed a coverage breadth and depth of 5% and 5x, respectively.

RS RS RS RS RS RS

City Porto Alegre

Ivoti Lindolfo Collor Lindolfo Collor Augusto Pestana Augusto Pestana 267 295 282 212 8 12

MajV and MinV consensus sequences were generated for all samples bearing wellsupported alternative nucleotides. Our analyses using Genome Detective showed that all 627 samples of DENV-1 belong to genotype V. Among the 40 samples of DENV-2, 33 were identified as genotype II - cosmopolitan, while seven were classified as genotype III Southeast Asian/American. We found no difference between the MajV and MinV genomic sequences for these DENV2 genotypes.Phylogenetic analyses unveiled the presence of six distinct clades within our DENV-1 samples (BR3, BR4, BR5, BR6, BR7, and BR8), as shown in Figure 2A. Additionally, our analyses revealed the presence of two distinct clades within the DENV-2 samples (BR4 and BR5), as shown in Figure 2B. Among the six identified DENV-1 clades in our study, three (BR3, BR4, and BR5) were previously documented by Brito et al., 2021. Similarly, one of the two DENV-2 clades identified in our analysis (BR4) was previously characterized in the same study. Meanwhile, the DENV-2 (BR5) belonging to genotype II - cosmopolitan was reported in the state by Gräf et al., 2023 .

Phylogenetic analysis revealed that MajV and MinV from the four outliers samples (Denv1_1985_2022, Denv1_2010_2022, Denv1_7052_2022 and Denv1_8625_2022) clustered with high branch support (using aLRT and ultrafast bootstrap approximation to assess branch supports) within different clades, confirming a coinfection by different clades for the four outlier samples. Coinfection occurred between two clades, BR5 and BR7. Denv1_2010_2022, Denv1_7052_2022, and Denv1_8625_2022, the MajV clustered within the BR7 clade, while their MinV genomic versions were clustered within the BR5 clade. The MajV genome of Denv1_1985_2022 clustered within the BR5 clade and MinV was placed within the BR7 clade (Table 1). The BR5 clade was described for the first time in 2021 by Brito et al. and is the most prevalent clade in our samples (51.38%). Its circulation in Rio Grande do Sul and Brazil was detected for the first time in 2015, being present and well dispersed in the state until the last sequenced samples of this study in 2023. The BR7 clade, also found in Peru in 2021, appears as an emerging clade in Rio Grande do Sul in 2022, present at an average frequency in our samples (19.61%).

The plausibility of the codetection events was further supported by the fact that the BR5 and BR7 clades identified by our study circulated simultaneously in the municipalities where the coinfection cases were found (Figure 2C). The four samples were sampled within a short time frame, with the first being collected on March 10, 2022, and the last on April 1st, 2022, with a distance of 21 days between them. All cases of clade coinfection were detected in 2022, where the two clades circulated in the metropolitan region of the state capital, Porto Alegre. The greatest geographical distances between the four samples are 60.9 kilometers from Porto Alegre to Lindolfo Collor. Notably, the occurrence of the four clade coinfection events coincided with a peak in observed cases in the region, emphasizing the plausibility of these detections. This context is further supported by the understanding that virus coinfection events are more likely to occur in endemic areas with cocirculation of different virus lineages (Figure 3).

Our current understanding of infection by different DENV clades and the occurrence of antibody-dependent enhancement (ADE) remains incomplete, with a lack of evidence for its occurrence and potential variation in the clinical outcome. Antibodies that developed during infection bind to DENV serotypes during subsequent infections. These antibodies activate the virus, forming a complex that binds to the Fcγ receptor (FcγR) on circulating monocytes. This interaction increases the ability of the virus to infect monocytes, leading to viral replication and an increased risk of severe dengue disease (Halstead, 2014; Shukla et al., 2020) . Addressing these knowledge gaps is needed to develop effective strategies to prevent and treat dengue. The symptoms of the patient coinfected by serotypes DENV-1 and DENV-2 were fever, headache, myalgia, nausea, back pain, retroorbital pain, arthritis, and leukopenia. The dual infection seems not to have caused a severe dengue clinical presentation in this patient since no hospitalization was required and no hemorrhagic manifestation or other more severe alterations occurred. Similarly, all four identified cases of coinfection by distinct clades presented only mild symptoms and did not require hospitalization.

Our in-depth analysis revealed at least five codetection events, corroborated by epidemiological data from geographically and temporally cocirculating clades in Rio Grande do Sul. Our results suggest that DENV codetection/coinfection events are rare, occurring at a rate of 0.75% in the studied population. However, this is undoubtedly a lower-bound estimate due to the limitation of detecting coinfection events with poorly divergent viral lineages. These findings have important implications for understanding the epidemiology of DENV in our region. The detection of coinfection between different clades shows the complexity of viral dynamics and may have significant repercussions on disease control. For example, coinfection by different clades can lead to greater genetic diversity of the virus, potentially influencing the host immune response and disease severity; further studies may help to clarify the interactions between multiple clade infections and the role of ADEs in shaping the clinical outcome of infection. As observed in our study, identifying coinfections by clades that circulate simultaneously in a given region emphasizes the need for continuous monitoring and genomic surveillance to better understand virus spread and predict possible outbreaks. Our results showed the importance of deep sequencing and refined phylogenetic analysis to detect and characterize coinfection events by different DENV clades, providing valuable information to guide disease control and prevention strategies. These findings also highlight the continued need for research to improve our understanding of DENV viral dynamics and their public health impacts.

This work was supported by Fundação de Amparo à Pesquisa do Estado do Rio Grande do Sul (FAPERGS/FIOCRUZ 13/2022 – REDE SAÚDE-RS, grant process 23/2551-00005107 and FAPERGS 14/2022 - ARD/ARC, grant process 23/2551-0000852-1). Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPQ) and Fundação de Amparo à Pesquisa do Estado do Rio Grande do Sul has provided a fellowship to R.S.S (FAPERGS 14/2022 - ARD/ARC).

Thales Bermann: Conceptualization, Investigation, Writing – review & editing. Ludmila F. Baethgen: Conceptualization, Investigation, Writing – review & editing. Tatiana S. Gregianini: Conceptualization, Investigation, Writing – review & editing. Fernanda Godinho: Methodology, Investigation, Writing - Review & Editing. Regina B. Barcellos: Methodology, Investigation. Amanda P. Ruivo: Methodology, Investigation. Milena Bauerman: Methodology, Investigation. Taina M. Selayaran: Methodology, Investigation, Software. Franciellen M. Santos: Methodology, Investigation. Julio A. Schoerer: Data Curation, Formal Analysis. Érica B. Möllmann: Data curation, Methodology. Fernanda L. Martiny: Data curation, Methodology. Valeska L. Lagranha: Data Curation. Luiz F.V. Oliveira: Funding acquisition, Methodology, Resources. Ana G. Veiga: Funding acquisition, Writing – review & editing. Andreia C.T. Zolet: Supervision, Writing – review & editing. Gabriel L. Wallau: Supervision, Conceptualization, Investigation, Writing – review & editing. Richard S. Salvato: Supervision, Conceptualization, Investigation, Writing – review & editing.

None.

The authors would like to thank the Secretaria da Saúde do Estado do Rio Grande do Sul for infrastructure availability and the work of professionals in charge of arbovirus control policies and diagnosis.

Raw reads and consensus fasta sequences from coinfection cases are available in the GISAID database under accession codes: EPI_ISL_18907330, EPI_ISL_18907331, EPI_ISL_18907332, EPI_ISL_18907333, EPI_ISL_18907334, and EPI_ISL_18907335. Datasets used in phylogenetic analysis, major and minor consensus sequences, and genome assembly metrics are available on Github (https://github.com/salvatolab/coinfectionDengue-clades-Southern-Brazil).

Supplementary Figure 1: A. Maximum likelihood tree for complete dataset DENV-1. The six clades identified in our study are highlighted, as shown in the legend below the tree; B. DENV-2 maximum likelihood tree for the complete dataset. The two clades identified in our study are highlighted, as shown in the legend below the tree.

Supplementary Figure 2: Plot with the frequencies of the two well-supported alleles for the MajV and MinV genomes of coinfection cases: Denv1_1985_2022, Denv1_2010_2022, Denv1_7052_2022 and Denv1_8625_2022. On the X axis, we have the position in the genome; on the Y axis, we have the frequency in percentage over the depth of reads in a given allele. The part of the bar in blue represents the frequency of the MajV allele, and the part in red represents the frequency of the MinV allele in percentage. The nucleotide representing this frequency in each position is positioned in the center of the bar.

Supplementary Figure 3. Plot with iSNV values for complete genomes of all samples, highlighting the four outlier samples with a notable gap between the values. Supplementary Table 1. iSNV values for complete genomes of all samples. The first column contains all 627 DENV1 samples followed by their iSNV values and all 40 DENV2 samples followed by their iSNV values.