Protocols for the collection and testing of samples were reviewed and approved by the institutional review board of the University of California, Berkeley (PDCS: 2010-09-2245; PDHS: 2010-06-1649; A2CARES: 2021-03-14191). They were also approved by the institutional review board of the Nicaraguan Ministry of Health (PDCS: CIRE 09/03/07/-008.Ver25; PDHS: CIRE 01/10/06-13. Ver. 18; A2CARES: CIRE 02/08/21-114 Ver. 4).

Samples were collected during June–December 2022 as part of 3 ongoing studies, the Nicaraguan Pediatric Dengue Cohort Study (PDCS), Pediatric Hospital-based Study (PDHS), and Asian-American Center for Arbovirus Research and Enhanced Surveillance (A2CARES) cohort study (part of the National Institutes of Health Centers for Research in Emerging Infectious Diseases Network), and through the National Dengue Surveillance Program of the Nicaraguan Ministry of Health (Appendix Table). Since 2004, the PDCS has been conducted within the Sócrates Flores Vivas Health Center catchment area in District 2 of Managua, with participants 2–17 years of age (3,13,15,16). The PDHS collects samples from patients with suspected dengue 6 months–14 years of age who receive medical attention in the Infectious Diseases Ward of the National Pediatric Reference Hospital (Hospital Infantil Manuel de Jesús Rivera) in Managua and who have received parent/guardian consent for the study (17). The A2CARES cohort study follows ≈2,000 persons 2–80 years of age in ≈1,000 households in District 3 of Managua.

We selected 3,171 suspected dengue cases for potential sequencing analysis, of which 1,353 were confirmed as DENV-positive with known serotype. To ensure a representative national distribution in the selection of samples for sequencing, we included 10 ±2% of the DENV-confirmed serotyped samples from each department that had adequate coverage for genomic analysis (n = 135), which included samples from the studies in Managua. We sequenced 49 DENV-1, 6 DENV-2, 38 DENV-3 and 42 DENV-4 samples that were positive by real-time reverse transcription PCR (rRT-PCR). To assess the relationships of the samples collected in 2022 to previously circulating strains from Nicaragua, we included 68 DENV-1, 62 DENV-2, and 51 DENV-3 previously published sequences (5,10,18).

Serum or plasma collected during the acute phase (days 1–5 since symptom onset) from patients with suspected dengue was processed at the National Virology Laboratory (Centro Nacional de Diagnóstico y Referencia, Ministerio de Salud, Managua, Nicaragua). We extracted viral RNA by using the QIAmp viral RNA mini kit (QIAGEN, https://www.qiagen.com) according to manufacturer instructions and used it to perform a single-reaction multiplex rRT-PCR for detection of Zika virus, chikungunya virus, and DENV (19). We further analyzed DENV-positive samples to identify the infecting serotype by using a multiplex DENV rRT-PCR (20,21). We prioritized cycle threshold values <28 for genomic sequencing to ensure the presence of ample genetic material, which ultimately enhanced sequence quality, minimized contamination risk, and improved detection sensitivity. In >90% of cases, sequences included in genomic analysis had >60% genome coverage.

We reverse transcribed sample RNA (8 μL) with 2 μL LunaScript RT SuperMix (New England Biolabs, Inc., https://www.neb.com) as follows: 25°C for 2 minutes, 55°C for 10 minutes, 95°C for 1 minute, and a 4°C hold. The PCR reaction mixture contained 13.75 μL of nuclease-free water, 1.5 μL of pool A and pool B primers, 5 μL of 5X Q5 reaction buffer, 2 μL of 2.5 mM dNTPs (Invitrogen, https://www.thermofisher.com), 0.25 μL of Q5 DNA polymerase, and 2.5 μL of sample cDNA (22,23). PCR amplification consisted of 98°C for 30 seconds, followed by 45 cycles of 98°C for 15 seconds and 65°C for 5 minutes, followed by a 12°C hold. We selected samples with a 900-bp band in the pool A and pool B mix for sequencing. Using 5 μL pooled PCR products (from pools A and B), we prepared the cDNA MinION library by using a native barcoding kit (NBD196; NgelabKampus, https://ngelabkampus.com) with a ligation sequencing kit (LSK-109; Interprise, https://interpriseusa.com) at 30°C for 2 minutes and 80°C for 2 minutes according to manufacturer instructions. We cleaned up PCR products by using AmpureXP purification beads (Beckman Coulter Diagnostics, https://www.beckmancoulter.com) and loaded the DNA library onto a primed MinION flow cell R9.4 (FLO-MIN 106; Oxford Nanopore Technologies, https//nanoporetech.com).

We extracted raw sequence reads by using filters for quality and trimmed adaptors by using Porechop version 0.2.3_seqan 2.1.1 (https://github.com/rrwick/Porechop) and assessed resulting sequence quality by using Nanoplot (24). We mapped raw reads by using Minimap2 v2.17-R941 (https://github.com/lh3/minimap2) against reference genomes from GenBank for DENV-1 (accession no. NC_001474), DENV-2 (accession no. NC_001474.2), DENV-3 (accession no. NC_001475.2), and DENV-4 (accession no. NC_002640.1) by using the pipeline at https://github.com/gvaleman/CONSENSO_D. We generated consensus sequences by using SAMtools version 1.7 (https://www.htslib.org/doc/1.7/samtools.html) with the default minimum depth coverage of 1 and confirmed the serotype by using Genome Detective Arbovirus Typing software version 4.1 (https://www.genomedetective.com) and BLAST (https://blast.ncbi.nlm.nih.gov). Mean depth coverage by serotype per SAMtools-depth tool was 221 for DENV-1, 68 for DENV-2, 516 for DENV-3, and 393 for DENV-4.

To increase the resolution of the genetic relatedness of the 2022 viruses in terms of genotype, country, and time of sample origin, we constructed alignments with initial broad coverage by using the National Center for Biotechnology Information Virus database (https://www.ncbi.nlm.nih.gov/labs/virus) by first subjecting sequences to BLAST and selecting the top 100 hits, followed by performing a directed search across regions and timescales represented in the database. Thus, we cast a broad net of potential related sequences. To reduce redundancy while maintaining representation (genotype, country, year) for final analysis, we aligned the large sample sets and reviewed them as initial RAxML (25) trees (Appendix Figures 1–4) before downsampling. We conducted sequence alignments with MAFFT version 7.4 (26) and manually reviewed them by using AliView version 1.28 (https://ormbunkar.se). We constructed maximum-likelihood phylogenetic trees for each serotype by using RAxML-HPC Blackbox version 8.2.12 and using a general time-reversible 
substitution model with bootstrap replicates determined automatically. We visualized the maximum-likelihood trees by using FigTree version 1.4.4 (http://tree.bio.ed.ac.uk) (data not shown). We performed Bayesian phylogenetic analysis by using BEAST version 1.10.4 (https://beast.community) and a Hasegawa-Kishino-Yano substitution model, a less parameterized model (2 substitution rates, for transitions and transversions) often used in place of the general time-reversible model above (6 substitution rates, for every base combination) to reduce computational burden. We also used a strict molecular clock and a Bayesian constant size coalescent tree. We ran the Markov chain Monte Carlo analysis for 100 million steps, taking samples every 5,000 steps, discarding the first 10% as burn-in. We assessed convergence with Tracer version 1.7.1 (http://tree.bio.ed.ac.uk) by using an effective sample size threshold >200. We assessed parameter estimate uncertainty by using 95% high probability density (HPD) intervals. Last, we generated maximum clade credibility (MCC) trees by using TreeAnnotator version 1.10 (https://beast.community) and visualized them with Figtree version 1.4.4.

Dengue epidemics in Nicaragua typically occur every 2–3 years (Figure 1, panel A); a given serotype surges to epidemic proportions roughly every 10 years: a shift from DENV-3 in 1998 (14) to DENV-2 in 1999 (3,13,27) to DENV-1 in 2003 (3) and again to DENV-2 in 2007 (5), then another surge of DENV-3 in 2009–11 (28), DENV-1 in 2012–13 (10), DENV-2 in 2015 and 2019 (11,18), and the cocirculation of all 4 serotypes in 2022 (Figure 1, panels A, B). However, during 2015–2020, only DENV-2 was detected in patients, whereas large epidemics resulted from newly introduced arboviruses such as chikungunya and Zika, followed by SARS-CoV-2 (29–31). Although Nicaragua did not implement official lockdown measures in response to the COVID-19 pandemic, international and specifically airline travel was severely curtailed for at least 1 year, starting in May 2020 (32,33). Together, the prolonged dominance of DENV-2 and the limited movement of people during the pandemic created a window of opportunity for hyperendemic DENV transmission in 2022 because of the susceptibility of the population to other serotypes. Furthermore, the circulation of Zika in 2016 might have primed immunity to increase DENV-4 cases (12). The result was the cocirculation of all 4 serotypes in 2022 (Figure 1, panel C), associated with an ≈100-fold increase in reported dengue cases from the previous year.

Serum/plasma samples targeted for sequencing were collected through the PDCS (n = 57), PDHS (n = 13), A2CARES cohort study (n = 4), and the Nicaragua epidemiologic surveillance system (n = 61) (Appendix Table) (16,17). A total of 135 sequences were generated by using newly established in-country sequencing capability based on the MinION platform and in-house bioinformatics and phylogenetics pipelines, markedly accelerating the time from clinical sample collection to pathogen sequencing from months to days.

We used 49 DENV-1 sequences from our study along with 81 publicly available sequences to build phylogenetic trees by using maximum-likelihood and Bayesian Markov chain Monte Carlo methods to generate an MCC tree (Figure 2; Appendix Figure 1, showing the initial ML tree with 408 taxa). DENV-1 genomes recovered in 2022 all belong to genotype V (American/East African and Asian genotype), consistent with previous studies of DENV-1 in Nicaragua (10). However, the viruses from 2022 clustered into 2 distinct clades, DV1-NI-1 and DV1-NI-2. Most sequences clustered into DV1-NI-2, for which the closest relatives were Ecuador sequences from 2014, suggesting that DV1-NI-2 is derived from an introduction from South America into Nicaragua. The other clade, DV1-NI-1, is represented by only 4 sequences and is associated with strains previously circulating in Nicaragua in 2013 and 2016. The date estimate of the time to most recent common ancestor (tMRCA) for DV1-NI-2 was 2013.7 (95% HPD 2012.8–2014.6) and for DV1-NI-1 was 1986.9 (95% HPD 1985–1989) (Figure 2). The genetic distance between DV1-NI-2 (2022) and DV1-NI-1 (2013 and 2016) is substantial, based on branch lengths relative to the rest of the phylogeny; mean rate is 0.04 substitutions per site (Appendix Figure 1). The introduction of the new clade, DV1-NI-2, and its cocirculation with the longer-standing Nicaragua-endemic clade, DV1-NI-1, represent increased DENV-1 genetic diversity circulating in Nicaragua in 2022.

We paired 6 DENV-2 sequences with 130 previously published genome sequences for phylogenetic analysis and tree building as previously described (MCC tree, Figure 3; Appendix Figure 2, showing the initial ML tree with 436 taxa). DENV-2 viruses circulating in Nicaragua in 2022 descended from DENV-2 present in Nicaragua since 2013 (clade DV2-NI-3B) and associated with epidemics in 2016 (10) and 2019, the latter marked by elevated numbers of cases, increased levels of disease severity, and accelerated rates of adaptive evolution (18). The estimated tMRCA of DV2-NI-3 was 2011.7 (95% HPD 2010.8–2012.4) (Figure 3). Globally, this DENV-2 lineage is a member of the Southeast Asian–American genotype IIIb, the only genotype circulating in Nicaragua since 1999 and in the region since the 1980s (5,10,18).

We combined the 38 DENV-3 sequences generated in this study with 93 published sequences to produce an MCC tree (Figure 4; Appendix Figure 3, showing the initial ML tree with 253 taxa). DENV-3 sequences from 2022 all fell within the Indian-subcontinent genotype III, the only genotype to have been detected in Nicaragua, first recorded in 1994 (34). Our prior studies had captured DENV-3 sequence evolution until 2014 (10,28) with an estimated tMRCA of 2003.8 (95% HPD 2003.2–2004.5) (Figure 4), after which DENV-3 remained undetected until 2022. In 2013 and 2014, multiple lineages of DENV-3 were circulating (10), but none persisted; the DENV-3 viruses sequenced in 2022 cluster most closely with viruses reported from Puerto Rico (2022), Brazil (2023), Cuba (2023), the United States (Florida, 2023), and India (2018) (35) and have an estimated tMRCA of 2019.8 (95% HPD 2019.5–2020.2) (Figure 4). The basal sequences to this clade are from southern and Southeast Asia, suggesting that the clade was introduced recently (after 2018) into the region, including Nicaragua, replacing earlier strains of this serotype (11).

We combined 42 DENV-4 sequences with 73 publicly available sequences to generate an MCC tree (Figure 5; Appendix Figure 4, showing the initial ML tree with 168 taxa). DENV-4 has not been detected in dengue patients in Nicaragua since 2006. Our analysis places DENV-4 sequences within genotype II, Indonesia, with an estimated tMRCA of 2005.1 (95% HPD 2003.7–2006.5) (Figure 5). The sequences most closely related to the Nicaragua strains are from Mexico, El Salvador, and Florida (2021–2022). However, the clade includes basal sequences from Mexico (2010) and Nicaragua (1999), suggesting a regionally long-circulating lineage with not-infrequent exchange between countries since 1999 until 2022. From this information, we conclude that there is a regional reservoir and source of DENV-4 into Nicaragua.