DEN is a serious cause of mortality and morbidity in the tropical and subtropical regions that infects fifty million people every year; approximately 500,000 of them are hospitalized and 5% to 15% of them die, which is a dramatic data 1. Positive-sense RNA viruses evolve rapidly, 234 allowing the virus population to quickly adapt to new environments and escape from host anti-viral responses. One of the principal causes of genetic diversity in DENV is the error-prone replication with RNA-dependent RNA polymerase (RdRp), 5 so that one genomic mutation occurs in nearly every cycle of virus replication. RNA virus, such as DENV populations at a particular region, may also rapidly change due to periodic selective sweeps6, by the introduction of foreign strains of virus 7892, and due to intra-serotypic recombination 1011121314. However, there has been considerable debate about whether recombination occurs in DENV 15 and the relevance of any recombination to the development of live-attenuated flavivirus vaccines 1617. Besides, there is a number of reasons for believing that recombination can occur in DENV and this process is being described with increasing frequency in DENV-1 1318 and other members of the family Flaviviridae 19202122. The recombination in DENV was reported in the structural genes region and particularly in E gene sequence through the use of the BOOTSCAN, DIVERSE PLOTS, and LARD software 14.

The co-circulation of multiple DENV populations increases the opportunities for a mosquito vector to ingest several variants by feeding on a number of diverse infected hosts, or for a host to be infected by vectors infected with distinct DENV variants. These conditions exist in Mexico, the Caribbean Area and South-East Asia 23. This is supported by the fact that there are many reports of multiple serotypes of DENV from single hosts 3232425. Furthermore, it is likely that mixed infections with different genotypes of the same serotype may also occur where they co-circulate 2627.

Oaxaca, Mexico is one of the states where DENV is endemic and serotypes -1, -2 and -3 of DENV are co-circulating 23. DENV-2 was reported as the serotype with higher frequency compared with DENV-1, -3 or -4. Six partial sequences of the genes encoding proteins: capsid (C), pre-membrane-membrane (prM-M), envelop (E), and non-structural 1 (NS1) represented as C₍₉₁₎-prM-E-NS1₍₂₄₀₀₎ from six different isolates of DENV-2 from the Oaxaca outbreaks 2005-2006 were obtained. In addition, the RT-PCR products of C₍₉₁₎-prM-E-NS1₍₂₄₀₀₎ and E genes obtained from the MEX_OAX_1656_05 isolate were cloned and sequenced.

MEX_OAX_1656_05 and MEX_OAX_1038_05 isolates displayed recombination in the prM-E and E-NS1 genes and the parental strains were the Asian/American and Cosmopolitan genotypes. In addition, the E gene sequences from the clone 7 (MEX_OAX_1656_05_C07) showed recombination between the nucleotides 906 to 1047 and the parental strains were Asian/American and American genotypes.

To determine recombinant sequences in DENV-2, the nucleotide sequences of the partial C₍₉₁₎-prM-E-NS1₍₂₄₀₀₎ genome from six isolates and 90 representative sequences of the different genotypes were aligned and analyzed by RDP3 and GARD. In addition, the RT-PCR product of the partial C₍₉₁₎-prM-E-NS1₍₂₄₀₀₎ genome from the MEX_OAX_1656_05 isolate was cloned in pGEM-3Z. The sequences of 9 clones were aligned with all of the above sequences (Figure 1A). We also sequenced 10 clones of the E structural gene from the isolate MEX_OAX_1656_05 and aligned with 180 representative sequences containing different genotypes by the programs mentioned above (Figure 1B).

The first task in this phylogenetic analysis was to determine the best model of nucleotide substitution for DENV-2 virus sequence evolution. This assignment was undertaken using the Model Selection test from DataMonkey online server 2829, which compares 201 models of DNA substitution. Our results demonstrated that the best model was TrN93 30. Accordingly, the most complex general time-reversible value was the best fit to the data (relative substitution rates of A↔C = 0.057, A↔G = 1, A↔T = 0.057, C↔G = 0.057, C↔T = 1, and G↔T = 0.057); the Ln likelihood = -4550.59; parameter count = 38; and AIC = 9177.19. Finally, the estimated base composition was A = 0.340, C = 0.278, G = 0.225, and T = 0.157.

Our analysis with RDP3 showed that the sequences of isolate MEX_OAX_1038_05 and MEX_OAX_1656_05 present statistical evidence of recombinants for GENECOV (P-Val = 2.467 × 10^-2), BOOTSCAN (P-Val = 4.289 × 10^-5), MAXCHI (P-Val = 1.438 × 10^-5), CHIMERA (P-Val = 3.790 × 10^-3), SISCAN (P-Val = 1.108 × 10^-9), and 3SEQ (P-Val = 4.478 × 10^-4), in two regions (Figure 2): the first breakpoints were located in 499nt and 512nt respectively; the second breakpoints were located in 868nt and 826nt respectively, and the third breakpoint was located in 2239nt in both recombinants (Figure 2A, 2B respectively). In addition, the analysis with GARD confirmed the breakpoints and recombination data for maximum likelihood. This analysis displayed the same site for the three breakpoints in both isolates: the first, second and third breakpoints were located in the nucleotides 498, 828 and 2226, respectively (Figure 2C). The recombinant regions were the intersection of prM-M structural gene to intersection of M-E structural genes and the second recombinant region started in the intersection of E-NS1 genes (Figure 2D). Interestingly, we found that the parental major strain was the non-recombinant clone MEX_OAX_1656_05_ C241 (obtained from the MEX_OAX_1656_05 isolate) and the minor parental strain was the Cosmopolitan genotype strain INDI_GWL_102_01 (accession number DQ448235).

The ML tree constructed with our sequence of structural gene C-prM from nucleotide 1-497 from the MEX_OAX_1038_05 and MEX_OAX_1656_05 isolates clustered with the Asian/American genotype (Figure 3A); the analysis of the region from nucleotides 498-828 of the isolates MEX_OAX_1038_05 and MEX_OAX_1656_05 moved to the Cosmopolitan genotype (Figure 3B) and when the region from the nucleotides 828-2222 was analyzed the two strains clustered again with the Asian/American genotype (Figure 3C). Finally, when the region corresponding to nucleotides 2223-2310 was analyzed the isolates clustered with the Cosmopolitan genotype (Figure 3D).

To determine the nucleotides involved in these recombinants, the C₍₉₁₎-prM-E-NS1₍₂₄₀₀₎ sequences of the clone MEX_OAX_1656_05_C241, recombinants sequences MEX_OAX_1038_05, MEX_OAX_1656_05 and the Cosmopolitan strain INDI_GWL_102_01 were analyzed. The changes in the recombinant isolates are labeled with a black dot (Figure 4). This analysis showed no evidence of recombination in the recombinant strain MEX_OAX_1656_05.

Like other RNA viruses, DENV undergoes low fidelity replication 31, resulting in virus pools of mutants. Therefore, to determine if one of the parental strains and/or a recombinant sequence is present in these pools, the RT-PCR product of the E protein gene from the recombinant strain, MEX_OAX_1656_05 was cloned and analyzed (Figure 1B). We obtained 10 E protein gene clones that were studied using the RDP3 software and it was determined that the sequence of clone MEX_OAX_1656_05_C07 presents statistical evidence of recombination by GENECOV (P-Val = 7.356 × 10^-7), BOOTSCAN (P-Val = 1.378 × 10^-5), MAXCHI (P-Val = 1.764 × 10^-3), CHIMERA (P-Val = 1.392 × 10^-4) and 3SEQ (P-Val = 4.478 × 10^-4). The E protein gene of said clones contains two breakpoints. The first breakpoint was located in the nucleotide 906 of the coding region for protein E; the second breakpoint was located in the nucleotide 1047 of the same gene (Figure 5A, Figure 6). GARD analysis confirmed that this clone is recombinant displaying the first breakpoint in the nucleotide 906 and the second breakpoint in the nucleotide1047 (Figure 5B). The constructed ML trees showed that the MEX_OAX_1656_05_C07 clone clustered in the Asian/American genotype branch when the 1-905 E gene region was examined, and clustered in the American genotype when the E gene region from nucleotide 906 to1047 was analyzed (Figure 5C). Finally, when region 1048-1485 was analyzed, the clone clustered again with the Asian/American strains.

The nucleotides involved in this recombinant are displayed in the alignment of the E gene region sequences of the recombinant MEX_OAX_1656_05_C07 clone, the parental clone MEX_OAX_1656_05_C17 and the strain MEX_95 (Figure 6).

Mutation rate studies indicate that DENV genome averages 1 nucleotide change per cycle of virus replication 32 because of the lack of proofreading activity. Another means to generate genetic changes is through recombination that has been reported in different Flaviviruses, including hepatitis C virus (HCV), diarrhea bovine virus (DBV), DENV, Japanese encephalitis virus (JEV), and Saint Louis encephalitis virus (SLEV) 141621. Recently, some evidences have showed recombination in natural populations of DENV-1 -2, -3 and -4 13142533 and the incorporation of this recombination as a probable mechanism contributing to genetic variation in DENV. Furthermore, the circulation of different serotypes and genotypes of DENV in a particular geographical region has been documented 233435, as well as the coexistence of two different serotypes or genotypes in a given mosquito or patient 232627, which makes feasible the recombination in DENV. From the first identification of an intergenotypic DENV recombinant 12, several DENV-1, -2, -3 and -4 recombinant strains have been identified 14. More importantly, the identification of this recombinant strains demonstrates that DENV is capable of successfully completing all the simultaneous stages of the infection in the same cell: the simultaneous replication of both viral genomes and the template shift by the viral RNA polymerase, while keeping the correct reading frame, encapsidation and release of the recombinant genomes in the process. The products will be subjected to the population processes guiding the maintenance, expansion or disappearance of new variants in the heterogeneous viral population.

All these reports focused on DENV-1 131827 recombination, and to date, there are a few reports of DEN-2 recombinant strains detected by analysis of protein E sequences 142526. Besides, protein E gene of clones or C₍₉₁₎-prM-E-NS1₍₂₄₀₀₎ region from human serum isolates have not been reported. There is only one single report of putative DENV-2 recombinant clone isolated from mosquitoes in the coding region for protein E 26. In this report, the isolates MEX_OAX1656_05 and MEX_OAX1038_05 showed recombination within the C₍₉₁₎-prM-E-NS1₍₂₄₀₀₎ region. In addition, there was recombination clearly identified within the E protein gene of the clone MEX_OAX1656_05_C7. Furthermore, the parental strains from the recombinants were identified. These results are a strong evidence of the creation of new variants in a heterogeneous viral population. Furthermore, this is the first report of DENV-2 recombination in Mexico.

We detected two isolates containing recombination highly similar to the one obtained from different cities in the state of Oaxaca, which is an evidence of the maintenance and expansion of new variants. These two recombinants in the C₍₉₁₎-prM-E-NS1₍₂₄₀₀₎ region contained 3 breakpoints non-previously reported: one in the prM and two in the E protein (Figure 2, 3, 4, 5). We are showing DENV-2 recombination between different genotypes in the isolates and clones analyzed with high frequency of approximately 30% and 10%, respectively. The detection of the DENV recombinants supports a potentially significant role for recombination in the evolution of DENV by creating genetic variation. This result is very important since recombination may shift the virulence of DENV. One could speculate if this shift may increase or diminish the virulence, like has happened in other RNA viruses, such as poliovirus 3637 influenza virus 38 and the HIV virus 39. A dramatic example is the loss of the attenuated phenotype of the poliovirus vaccine by recombination, resulting in the generation of new phenotypes that produce the acute paralytic disease. Consequently, recombinants have the potential to generate strains with a higher or lower virulence. To test this issue for DENV recombinants will be necessary to have an animal model to study the virulence of these recombinants.

The two points in our experimental procedure that have been instrumental in obtaining the reported result and to build confidence are: First, we analyzed 6 isolates and one clone in the coding region C₍₉₁₎-prM-E-NS1₍₂₄₀₀₎ from Oaxaca and concentrated our efforts in sequencing the E gene of 10 clones from one isolate. These regions were chosen based on its biological relevance and on the location of breakpoints identified in previous reports of recombination in DENV 1213262733; secondly, we minimized the chance of detecting false, artifactual recombination by using long extension times 40 and a proofreading DNA polymerase (Platinum Taq Hi-Fi) 41.

Moreover, the breakpoints tested by RDP3 resulted significant by 7 statistical methods; besides, the GARD software displayed the same breakpoints as the RDP3 software package.

The analysis of 10 clones obtained from the isolate MEX_OAX_1656_05 showed one clone (MEX_OAX_1656_05_C07) containing recombination in the E gene (Figure 5, 6). Interestingly, the parental strains for this recombinant were the Asian/American and the American genotypes. This result is very important because the American genotype has the highest divergence among all the genotypes for DENV-2. Furthermore, this is the first report on recombination between the Asian/American (MEX_OAX_1656_05_C17) and American genotypes (MEX_95), which is supported by the analysis with RDP3 and GARD (Figure 5A-B). This recombinant displays the breakpoints between the nucleotides 906 and 1047. These results suggest that the frequency of recombination in DENV is higher than thought earlier, and the process will remain fundamentally hidden until more studies of clonal diversity to be undertaken. Nevertheless, the precise mechanism underlying the recombination events for DENV is unknown. To understand the mechanism of recombination the development of experimental models for co-infection to generate DENV recombinants is required.

The second breakpoint in the C₍₉₁₎-prM-E-NS1₍₂₄₀₀₎ region (nucleotide 868 and 826) for the MEX_OAX_1038_05 and MEX_OAX_1656_05 isolates was different for 40 nucleotides when determined by BOOTSCAN, but it was the same when GARD was used (Figure 4). This was not associated with a sequence that permits the inference of a hot-spot of recombination as previously reported 12132627 and does not permit the deduction of the mechanism of recombination as has been described for other flavivirus 3142.

Recombination may be a consequence of circulation of several genotypes at the same time in the same site and probably inside the same cell in the mosquito or human patient; in addition to the high density of different viruses circulating in the geographic area of Oaxaca. This is supported by a previous work that suggests that density of geographical and temporal sampling increases the probability for identifying recombinant sequences 25.

Phylogenetic studies have shown the circulation of the American 43, American/Asian 23, and Cosmopolitan 44 genotypes in Mexico, which makes feasible their recombination and explains the fact of the Cosmopolitan and American genotypes to recombine with the Asian/American genotype spread more broadly. Our results in combination with previous reports 26 on DENV-2 recombination suggest that the different genotypes of DENV-2 are circulating in the virus pool infecting the mosquitoes or the human cells around the world. Until now, it remains unclear whether the frequency of recombination seen in this and previous studies is driving an increasing virulence of DENV strains. However, the recombinant strains of this study were obtained from the outbreak 2005-2006 where the frequency of DHF cases was higher than the DF cases in comparing to previous epidemics 45. To elucidate the role of recombination in DENV virulence will be necessary to follow the generation of recombinants in outbreaks from other Mexican states.

It is unclear whether the recombination events took place in a human host or a mosquito vector co-infected by multiple DENV genotypes. In this study, we detected two recombinant isolates of DENV-2 from human hosts namely MEX_OAX_1038_05 and MEX_OAX_1656_05, which identify 3 breakpoints within the prM-E-NS1 genome. Particularly the recombination appeared to have involved two genotypes of DENV-2, the Asian/American clone (MEX_OAX_1656_05_C241) from the same strain and the Cosmopolitan strain (INDI_GWI_102_01).

It is remarkable that parental and recombinant viral sequences of protein E were observed in an isolate from a single patient, particularly when the recombination appeared to have involved two genotypes of DENV-2 (Asian/American and the American) from the same geographic area (Oaxaca, Mexico). This is only the second observation of one parental and recombinant of DENV-2 in a population within a single host 26. There are two more studies where both parental and recombinant viral genomes were observed in a DENV-1 isolate from a single patient. DENV recombination mechanism will be clarified by undertaking more studies of clonal diversity in both human and mosquito vector in Mexico.

GPR obtained the isolates and clones, carried out the RT-PCR assays using RNA from passages 3 to sequence the partial C91-prM-E-NS12400 genome and E gene to develop recombination and phylogenetic analysis. ADB determined serotype and helped in the phylogenetic analysis. MCN participated in obtaining the clones of E gene. AC, collected serum samples from patients from Oaxaca and helped to obtain the isolates and clinical data from Oaxaca, Mexico. GPR and MLM participated in the writing and discussion of results, helped to review the manuscript and assisted with the literature validation. MLM proof-read and assembled the manuscript. All authors participated in the discussion of results and read and approved the final manuscript.