Typical epidemiology of respiratory virus infections in a Brazilian slum

Abstract Host population size, density, immune status, age structure, and contact rates are critical elements of virus epidemiology. Slum populations stand out from other settings and may present differences in the epidemiology of acute viral infections. We collected nasopharyngeal specimens from 282 children aged ≤5 years with acute respiratory tract infection (ARI) during 2005 to 2006 in one of the largest Brazilian slums. We conducted real‐time reverse transcription‐polymerase chain reaction (RT‐PCR) for 16 respiratory viruses, nested RT‐PCR‐based typing of rhinoviruses (HRVs), and collected clinical symptoms. Viruses were common causes of respiratory disease; with ≥1 virus being detected in 65.2% of patients. We detected 15 different viruses during 1 year with a predominance of HRV (33.0%) and human respiratory syncytial virus (hRSV, 12.1%) infections, and a high rate of viral coinfections (28.3%). We observed seasonality of hRSV, HRV and human coronavirus infections, more severe symptoms in hRSV and influenza virus (FLU) infections and prolonged circulation of seven HRV clusters likely representing distinct serotypes according to genomic sequence distances. Potentially unusual findings included the absence of human metapneumovirus detections and lack of typical FLU seasonal patterns, which may be linked to the population size and density of the slum. Nonetheless, most epidemiological patterns were similar to other studies globally, suggesting surprising similarities of virus‐associated ARI across highly diverse settings and a complex impact of population characteristics on respiratory virus epidemiology.


| INTRODUCTION
Acute respiratory tract infections (ARI) are the main cause of morbidity and mortality among children aged <5 years in the developing world. 1 Respiratory viruses cause up to 80% of ARI. 2 Respiratory viruses are spread via three different transmission routes: contact, droplet, and aerosol transmission. 3 Host population size, density, immune status, age structure, and contact rates affect the transmission patterns of viruses causing acute predominantly self-limiting infections, such as respiratory viruses. 4 Residents of resource-limited communities such as slums may be particularly vulnerable to virus-associated ARI. Hypothetically, virus transmission may be facilitated in these dense populations, characterized by frequent interindividual contact, crowded housing, improper sanitation systems, poor education, and poor nutritional status, exemplified by inversely correlated influenza virus prevalence and family income in a study from Bangladesh. 5 The United Nations define slums as human settlement areas that combine the following attributes: lack of basic services as sanitation and water sources, substandard housing or illegal and inadequate building structures, overcrowding and high density, unhealthy living conditions and hazardous locations, insecure tenure; characterized by irregular or informal settlements, poverty, and social exclusion. 6 Close to 880 million people worldwide reside in urban slums, and this number is expected to double by 2025. 7 Nonetheless, little is known about how and whether disease patterns in urban slums differ from affluent settings. 8 Pivotal epidemiological studies conducted in slum cohorts from Bangladesh and Kenya highlighted the importance of respiratory viruses in these communities. 5,9,10 In Brazil, 11.4 million people, nearly 6% of the country's population live in slums (https://ww2.ibge.gov.br/home/). Data on virus-associated ARI from slum communities, particularly from Brazil, are scarce. In a study based on clinical symptoms, ARI in children inhabiting a Brazilian slum were very frequent, representing 50% of child infections events through a 1-year period. 11 A study combining clinical and virological data reported ARI symptoms in 60% of children inhabiting another Brazilian slum and in 35% of these children, the virus was isolated by cell culture. 12 Moreover, surprisingly high virus or bacterial detection rates (up to 85%) were observed in a study conducted in northeastern Brazil in children under 5 years from low-income families presenting with ARI. 13 We previously found that virus-associated ARI showed similar epidemiological patterns between a rural African and an urban European setting, including an overall similar spectrum of viruses, age associations and seasonal fluctuation despite drastic differences of socioeconomic status (SES) and climatic patterns. 14 Hypothetically, lower SES in the African setting may have been equaled by higher population density in the European setting, both of which likely facilitate virus transmission.
Here, we investigated 16 respiratory viruses by real-time (reverse transcription)-polymerase chain reaction (RT-PCR)-based methods in one of the biggest slums of Latin America, namely Paraisópolis. This community is located within the urban area of São Paulo city, inhabited by 89.000 residents in a 1.5 km 2 area (approximately 59.300 inhabitants/km 2 ) and distributed in 21 thousand dwellings including shacks, masonry buildings, townhouses and older and solid buildings 15 ( Figure 1A).

| MATERIALS AND METHODS
Nasopharyngeal aspirates were collected longitudinally from human parechoviruses (hPeVs), adenoviruses (AdVs) and influenza A and B viruses (FLU) as described previously. 14,16 HRV typing was done using a nested RT-PCR assay targeting the VP4/VP2 domains. 17 The HRV evolutionary history was inferred using a Neighbor-Joining algorithm, a p-distance substitution model and a complete deletion option in MEGA6 (www.megasoftware.net/) on a data set comprising 434 nucleotides from the viral VP2/VP4 domains after deletion of 5′-untranslated region sequence portions. Recombination in the data set was discarded using RDP V4.95. Only HRV was typed due to frequent detection and rapid evolution allowing sufficient phylogenomic resolution.
Statistical analysis was performed by using SPSS 13.0 (IBM) with Χ 2 or Fisher's test when analyzing numbers less than 5 in any cell.
Confidence intervals were calculated using Open Epi (www.openepi. com) using the Wilson method.

| RESULTS
Many epidemiological features of respiratory viruses in our study were similar to those from other reports worldwide. The first similarity was a high overall virus detection rate of 65.2% (184 of 282 patients) together with the genetic diversity of the respiratory pathogens, including all viruses that were tested except hMPV (Table 1). Both attributes were previously reported in studies conducted in temperate and tropical regions of Brazil (~60.0%), 18,19 in a slum community from Kenya (71.0%), 20 and in a German cohort (56.6%). 14 The second similarity was the predominance of HRV (33.0%) and hRSV (12.1%) detections (Table 1), also observed in hospitalized patients in distinct cities of Brazil, in a slum community in Kenya [18][19][20] and in German children. 14 The third similarity was the high rate of viral coinfections (28.3%) ( Figure 1B) which was similar to rates observed in a study from Curitiba, southern Brazil (29.0%) 19 and in a Kenyan slum community (27.0%). 20 Our data also point to the predominance of HRV in cases of coinfections (30 of 52 coinfections, 58.0%), a feature similarly reported in a study conducted in Brazil (69.0% of all coinfections) 19 that is likely influenced by the overall high number of HRV infections. Albeit not statistically significant, HRV, hRSV, and FLU were commonly detected as monoinfections (67.7%, 73.5%, and 76.9% respectively; P > .05 for all three viruses), whereas HCoV, AdV, hPiV, EV, and hPeV were significantly more frequently detected as coinfections (53.8%, 61.1%, 66.7%, 80.8%, and 83.3%, respectively; P < .05 for all) ( Figure 1C).
These data were consistent with results of a study including data from eight tropical countries showing that HRV, hRSV, and FLU were more commonly detected as monoinfections, AdV as coinfections and HCoV and hPiV equally distributed between both coinfections and monoinfections. 21 Similarly, preliminary studies from Brazil showed that hRSV and FLU were most frequently detected as monoinfections and EV and AdV as coinfections. 13,19 The fourth similarity was the seasonal variation of hRSV and HCoV detections. Particularly, hRSVs were more frequently detected during autumn (Fisher's exact, P = .004), a pattern already observed in surveillance studies carried out in two distinct São Paulo city hospitals. 22,23 HCoV were more frequently detected during winter (Fisher's exact, P = .011; Figure 1D), as observed previously in a study conducted during 20 years in the United States. 24 The fifth similarity included statistically significant associations of FLU and hRSV infections with more severe symptoms such as fever (Fisher's exact, P = .007 and P = .02, respectively) and of hRSV infections with dyspnea (Fisher's exact, P = .038) ( Figure 1E)

| 1319
Together with the low frequency of codetections of FLU and hRSV, this feature is consistent with higher pathogenicity of both viruses. 18 In contrast to hRSV, HRV infection was significantly less frequently associated with fever and dyspnea (Fisher's exact, P = .001, Χ 2 P = .005) ( Figure 1E). The lower proportion of HRV infections and cases of fever compared to other community-acquired respiratory virus infections was similar to a previous study conducted in patients hospitalized with ARI in Curitiba, southern Brazil. 19 The sixth similarity was a predominance of hRSV and hPeV detections in patients aged ≤1 year (Fisher's exact, P = .001 and P = .01, respectively) ( Figure 1F). Similarly, Annan et al 14 reported that pneumoviruses including hRSV were more frequently detected at younger ages in cohorts from Ghana and Germany, predominantly in patients less than 1 year of age.
The seventh similarity comprised two contrasting HRV epidemiological patterns, namely replacement of some HRV strains and maintenance of other HRV strains over time. HRV comprises three defined species termed HRV A-C, and >100 different types likely representing multiple distinct serotypes. 25 We successfully typed a total of 77 HRV strains representing all three HRV species (GenBank accession numbers MH824434-MH824510), whereas 16 HRV strains could not be typed. Proportions of individual HRV species in our study (53.2% HRV-A, 6.5% HRV-B, 40.3% HRV-C) were in agreement with other reports worldwide. 17,26 The majority of distinct HRV strains in our study were detected only in one season, a common HRV epidemiological pattern. 17,[25][26][27] In contrast, seven clusters termed I-VII, each composed of three to five patient-derived HRV strains presenting a very low mutual pairwise sequence distance (≤2%), were detected over more than two seasons. The seven clusters belonged to HRV species A and C and differed from one another by 15% to 39% mutual nucleotide sequence distance. Previous studies on HRV typing using the genomic fragment used in our study found that defined HRV serotypes differed by at least 10% mutual nucleotide sequence distance. 17 Following this criterion, strains belonging to one cluster in our study thus likely represent the same HRV serotype, whereas the seven clusters all represent distinct HRV serotypes. Community protective immunity can dramatically limit the circulation of defined viral serotypes, including HRV and other viruses. 17,28 This was apparently the case for some, but not all HRV serotypes in our study, the latter potentially facilitated by the population structure of the slum.
Prolonged circulation of HRV clusters was detected specifically during winter and spring ( Figure 1G, clusters I, III, V, VII), summer and spring (cluster II), winter and summer (cluster IV), and autumn and summer (cluster VI). These data were reminiscent of prolonged circulation of closely related HRV lineages for 4 to 12 months or over three consecutive seasons reported previously from Sweden and Finland 26,27 and are thus not unique to our study setting. Notably, we cannot exclude that HRV strains may have been re-introduced repeatedly from other areas of São Paulo, a large metropolis accumulating about 21 million inhabitants, potentially facilitated by commuting of slum inhabitants to nearby areas for labor. 15 We observed only two potentially unusual patterns including the total absence of hMPV detections and absence of FLU seasonality.
The first may be explained by widely documented local variation in annual hMPV circulation, 29  More specifically, usage of identical methodology enables direct comparisons between this study and our previous investigation of a rural area in Ghana and an urban environment in Germany. 14  approaches, yet studies focusing on respiratory virus epidemiology in slum populations are scarce compared to studies from affluent settings. 32,33 This knowledge will be crucial to inform potential public health interventions to reduce disease burden in the population.

ACKNOWLEDGMENTS
This study was supported by the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP; n o 02/08460-6) and by the European Union's Horizon 2020 program through the ZIKAlliance project (grant agreement no. 734548).