Of 1,656 fecal samples from remote communities analyzed for rotavirus, 136 (8.2%) were determined to be positive by the commercial immunochromatographic test. Of these positive samples, 96 were from the 411 patients with diarrhea and 40 were from the 1,245 asymptomatic controls. Diarrhea was significantly associated with being infected with rotavirus (odds ratio = 9.2; 95% confidence interval 6.1–13.9). Rotavirus RNA was detected at the highest rates from symptomatic infants and, surprisingly, persons >40 years of age (Table 1). No pronounced seasonality of rotavirus infection was determined, and incidence was not significantly associated with month of collection or with the observed 30-day rainfall for 15 days before the visit and during the visit (data not shown). This lack of seasonality in the tropics has been reported previously (19).

From the 136 rotavirus-positive community samples, a subset of 47 samples were genotyped for the VP4 and VP7 genes. This subset represented all samples collected from February 2005 through February 2006, from 14 rural communities and Borbón. Of these 47 samples, 35 (74%) yielded successful PCR typing results for the VP4 gene and 37 (79%) yielded successful PCR typing results for the VP7 gene. An additional 6 (13%) yielded successful PCR amplification at 1 of the 2 genes. The remaining 8 (17%) samples were not typeable. Six of these untypeable RNA samples, along with 6 typeable samples, were subjected to electrophoresis on an agarose gel and visualized by staining with ethidium bromide in an attempt to detect rotavirus RNA. None of the untypeable samples produced rotavirus RNA banding patterns, whereas 3 of 6 typeable samples could be visualized.

Among the successfully typed samples, genotypes P[8] and G9 predominated. A small proportion of the samples produced patterns corresponding to P[6]G1 and P[6]G9 (Table 2). Table 3 summarizes the percentage of patients infected with P[8], G9, and P[8]G9 based on 2 assumptions of the 13 samples in which 1 or both of the VP4 and VP7 genes were not typeable. The first assumption was that samples were nontypeable because they were degraded sometime between testing positive by immunochromatographic tests in the field and the sample’s arrival in the laboratory in Quito. In this case, we assumed that those samples were missing data. The second assumption was that the samples were nontypeable because they were novel strains, and we therefore included them in the dataset. G9 genotype was identified in 34 samples, resulting in a 72%–92% infection rate; P[8] genotype was identified in 31 samples, a 66%–89% infection rate; and the combination of P[8]G9 was found in 29 samples, a 62%–88% infection rate.

To determine whether the genotypes in remote communities corresponded to strains circulating elsewhere in Ecuador, we analyzed 29 rotavirus-positive samples from Hospital de Niños Baca Ortiz in Quito. Again, genotypes P[8] and G9 overwhelmingly predominated (Table 2). Electrophoretic evidence for 5 mixed infections was found among these urban samples, but this was not seen among the rural samples (Table 2).

Of the 29 P[8]G9 community samples, 22 were sent for sequencing to the University of Leuven in Belgium. Approximately 750 bp of high-quality nucleotide sequence data for the VP7 gene was obtained from each sample. The 22 P[8]G9 samples were remarkably homogenous at the sequence level, with only 3 single nucleotide polymorphisms found in the 22 sequences. Phylogenetic analysis of the sequences (Figure) showed that the Ecuadorian sequences grouped together monophyletically and were part of the large clade composed of most of the recently isolated G9 rotavirus sequences worldwide.

The present study reports a high rate of infection (72%–96%) with rotavirus G9 genotype among persons in 2 geographically distinct regions within Ecuador, a remote coastal rain forest and an urban Andean hospital. To our knowledge, it is the first description of rotavirus genotypes in Ecuador, and the results support the observation that the G9 genotype, particularly P[8]G9, is spreading throughout Latin America. Also, the present study appears to be one of the few community-based descriptions of rotavirus infection (20–22). Symptomatic persons were actively identified in the community, recruited into the study, and matched with 3 asymptomatic controls each. This approach presents a more complete picture of rotavirus infection in rural communities than would be possible with the clinical sampling used in most previous studies that presumably focused on more urban environments. The high rate of rotavirus infection among symptomatic persons >40 years of age may be due to this age group’s lack of exposure to the emerging rotavirus genotype and is an observation that might have been missed in a purely clinical study.

The G9 genotype has been documented since the early 1980s (23,24). Throughout much of the 1980s and 1990s, G9 was considered very rare; however, recent reports have described it as increasingly important (2,3,25). In the United States, the G9 genotype was detected in a 1995–1996 outbreak (8) and maintained its presence in the subsequent 2 years, with an average detection rate across 10 US cities of 7% (26). In Australia, the overall G9 detection rate, averaged across 3 population centers, increased from <1% in 1997 to 29% in 2001 (27). In Japan, G9 was essentially undetected throughout the mid-1980s and 1990s until it suddenly appeared in several cities in 1998–1999 (28). In India, G9 strains were detected for the first time in the late 1980s and throughout the early 1990s were usually found in combination with the P[11] or P[6] genotypes at a detection rate of about 20% (12). A study of 6 population centers across India during 1996–1998 found an overall G9 detection rate of 17% but found G9 as the major strain (and for the first time associated mainly with P[8] genotype) in New Delhi in late 1998 (29). At 17 sites throughout the African continent during 1996–1999, the G9 detection rate was generally ≤5% (30), with the exception of Ghana (1997–1999), where it comprised 28% of rotavirus positive samples, and Nigeria (1998–1999), where it comprised 37% (31). In Europe, many instances of G9 detection have been reported from the late 1990s through the early 2000s (16,32–34).

Latin America, in particular, has seen a surge in dominance of this genotype in recent years. In Rio de Janeiro, Brazil, during 1997–1999, the detection rate was ≈15% (6,7). In São Paulo, Brazil, during 1996–2003, the rate was 17% overall but in the last 2 years, G9 accounted for 30%–50% of rotavirus infections (35). To our knowledge, only 4 studies have reported G9 detection rates as high as those in our study: 75% in Paraguay, 2000 (36); 75%–90% in Salvador, Brazil, 1999–2002 (9); 92% in Chiang Mai, Thailand, 2000–2001 (37); and 73% in Alice Springs, Australia, 2001 (27).

A potential source of bias in this study comes from the incomplete typing of 13% of the putatively rotavirus-positive specimens and the inability to type an additional 17%. These incomplete or untypeable samples, which were positive by immunochromatographic tests, may be the result of inappropriate handling or storage of some fecal samples, which can be complicated in remote community studies such as this. Typing failure because of sample degradation or false-positive ELISA results is not likely to result in biased results, and the lack of visualizable rotavirus RNA bands among the untypeable samples suggests that degradation is a likely cause of typing failure. However, in any PCR-based typing scheme, typing failure may be caused by primer–template mismatch, which could bias results, especially with novel strains (16,38). A nonsystematic review of 16 recent studies suggests that failure of PCR-based G-typing is relatively widespread, although the failure rate varies. Four studies report <5% typing failure (16,33,35,37), 3 report 5%–10% failure (9,12,26), 2 report 10%–20% failure (29,36), 2 report 20%–30% failure (27,30), and 2 report >30% failure (31,34). An additional 3 studies do not explicitly state whether all typed samples yielded results (6,7,32). A more complete picture of global rotavirus diversity should facilitate efforts to improve molecular typing techniques. In this study, the maximum possible bias would affect the Esmeraldas community results by lowering the G9 infection rate from 92% to 72% and the P[8]G9 from 88% to 62%. In the Quito samples the effect is much less pronounced, potentially lowering the G9 infection rate from 96% to 90% and P[8]G9 rate from 93% to 86%. Further studies are required to narrow this uncertainty; however, even the lower end estimates indicate that P[8]G9 is the predominant strain in Ecuador.

Additional evidence that G9 rotavirus is spreading through Latin America comes from comparing our nucleotide sequences to other sequences reported in GenBank. The sequences from the current study cluster into a large clade, which includes most of the recently isolated G9 rotavirus reported in the literature. This “emerging clade” is relatively homogenous: most isolates within the clade have <1% sequence divergence, an observation about G9 that has been made previously (39). However, the more recent regional isolates do tend to cluster together, as is the case for subclades composed of strains from Australia, Paraguay, or Ecuador (Figure). The low bootstrap support for these subclades is due to the small number of single nucleotide polymorphisms differentiating them.

The increasing prevalence of G9 rotavirus is particularly relevant given that many countries, including Ecuador, have approved the use of 2 rotavirus vaccines (10,40), but despite the wide distribution of G9 during the past 9 years (2,3,6–9), neither vaccine formulation includes the serotype G9 antigen (10,40). Studies have shown that some vaccines that do not contain the G9 antigen may still be capable of eliciting protective immunity against the G9 serotype (10). This immunity is most likely attributable to the G9 genotype’s common association with P[8], which is included in both vaccines. However, cross-immunity may not be universal, as has been seen with type P[4]G2 (2,3,10). Continual surveillance of circulating types, therefore, should be carried out before to the introduction and during the implementation of rotavirus vaccination programs.