Entomovirological Surveillance in Schools: Are They a Source for Arboviral Diseases Transmission?

Surveillance and control activities for virus-transmitting mosquitoes have primarily focused on dwellings. There is little information about viral circulation in heavily trafficked places such as schools. We collected and analyzed data to assess the presence and prevalence of dengue, chikungunya, and Zika viruses in mosquitoes, and measured Aedes indices in schools in Medellín (Colombia) between 2016–2018. In 43.27% of 2632 visits we collected Aedes adults, creating 883 pools analyzed by RT-PCR. 14.27% of pools yielded positive for dengue or Zika (infection rates of 1.75–296.29 for Aedes aegypti). Ae. aegypti was more abundant and had a higher infection rate for all studied diseases. Aedes indices varied over time. There was no association between Aedes abundance and mosquito infection rates, but the latter did correlate with cases of arboviral disease and climate. Results suggest schools are important sources of arbovirus and health agencies should include these sites in surveillance programs; it is essential to know the source for arboviral diseases transmission and the identification of the most population groups exposed to these diseases to research and developing new strategies.


Introduction
During the last few decades, the incidence of arboviral diseases has increased worldwide. A total of 390 million dengue cases are estimated every year, 96 million of which show symptoms and about 20,000 are fatal, mostly among children [1,2]. In 2016, more than 2.3 million cases were reported just in the Americas, 1032 of which were fatal [3], and 2019 had the highest number of dengue cases ever recorded in the Americas with more than 3 million reported cases and 1400 registered deaths [3]. Colombia is no stranger to this problem; in 2019, 127,553 cases and 87 confirmed deaths were reported. Cities such as Cali, Medellín, and Bucaramanga have been classified as hyperendemic because there is periodic circulation of all four dengue serotypes and these cities historically have had the highest burden of dengue in Colombia [4]. Although this disease can occur at any age, a study performed in Colombia found that 50.7% of cases were observed in individuals under 19 years old, whereas 49.5% of severe dengue cases were observed in children aged under 14 years old [5].
Recently, Zika and chikungunya have emerged as new arboviruses in the Americas with significant public health, social, and economic impact. After Brazil, Colombia had the highest number of reported Zika cases in South America during the outbreaks that occurred between 2015-2016 [6]. Also, chikungunya virus has caused considerable outbreaks worldwide. In 2015,~700,000 probable cases were reported in the Americas, most of them (356,079 cases) in Colombia. These arboviruses are primarily transmitted by the bite of Aedes aegypti and Ae. albopictus mosquitoes, both of which are present in Colombia. Moreover, in the case of Zika, natural infection and experimental transmission have been reported in Culex mosquitoes [7][8][9]. Because few vaccines are currently available to prevent the arboviral diseases of public health importance, the prevention and control efforts focus on reducing mosquito populations, and the main prevention strategy is vector control [10,11], which has traditionally focused on people's dwellings and residential areas [12,13]. However, there are non-residential areas that are highly trafficked and where immature and adult forms of these vectors are often identified [14][15][16][17]. Despite the potential importance of these other locations, there are few studies that have analyzed their role in arboviral transmission. Recently, new strategies for the prevention of disease caused by arboviruses have been developed, including the possible use of mosquito saliva proteins as a universal vaccine against arboviruses; it is very important to know the source for arboviral diseases transmission and the most exposed population to help specialists to research and develop on the relevant vaccines.
In Medellín, the Secretariat of Health (SSM, by its Spanish acronym) implemented a strategy called Healthy Schools, a tool aimed at performing health-promoting activities and preventing arboviral diseases at schools. Healthy Schools includes a vector control program where traditional Aedes indices are routinely measured to identify schools with the highest abundance of Aedes mosquitoes to then focus control efforts on those locations (SSM 2018). However, multiple studies have shown that there is a weak association between traditional Aedes indices and the risk of arboviral transmission [18], and detecting mosquito infection rates for arboviral disease may be a better measure of human risk [19,20]. In this study, we analyzed natural infection with dengue, Zika, and chikungunya viruses in Ae. aegypti and Ae. albopictus mosquitoes collected in schools in Medellín, as well as the association between infection rates in mosquitoes and traditional Aedes indices, some epidemiological (dengue and Zika cases) and climate (temperature, relative humidity, and precipitation) variables. The implications of arbovirus infection in mosquitoes collected in schools are then discussed in the context of disease mitigation in educational establishments.

Study Area
Medellín is in the department of Antioquia, at 75 • 34 05 W and 6 • 13 55 N and is approximately 376.2 km 2 . Medellín is relatively flat in the city center, which is at 1500 masl (meters above sea level), with a steep rise along the perimeter up to 1800 masl. The climate is subtropical sub-humid with mean temperatures ranging from 16 to 28 • C. The population is around 2.5 million, and both dengue and Zika are routinely reported. Chikungunya cases in Medellín occur on a sporadic basis [21].

Mosquito Collections and Traditional Aedes Indices
Between January 2016 and December 2018, 2362 inspections were performed to capture adult mosquitoes (Figure 1). Schools included elementary and high; these were chosen for Medellin Health Secretary criteria. Schools were distributed throughout the city. Each visit was performed by a single official, who spent between 45 and 60 min at each institution, depending on its size. An active search was used to collect mosquitoes with a mouth aspirator and an entomological net among common areas such as offices, classrooms, bathrooms, and different humid and dark places, which are the preferred resting places of these vectors [22,23]. Mosquitoes were stored in tubes and transported alive to the Medical Entomology Laboratory of the School of Medicine of Universidad de Antioquia for identification through published keys [24].  The presence and number of potential larval development sites (water-hold tainers), such as tires, flower vessels, bottles, trash cans, or gutters, within each sch recorded. All containers and schools were visually assessed for the presence of im forms and adults to calculate the following traditional Aedes indices: Adult Index percentage of schools infested with adults Aedes), house index (HI) (the perce schools infested with larvae and/or pupae), container index (CI) (the percentage o holding containers infested with larvae or pupae), and Breteau index (BI) (the nu positive containers per 100 schools inspected). Containers were counted as positiv had immature forms of the vector, whereas schools were considered positive if one container with immature forms of Aedes spp. During 2016, schools were in once per trimester, and in 2017 and 2018 schools were visited once a month. W gated ento-virological data to a three-month scale to enable comparison to entom risk indicators. The level of risk was based on three levels of risk for each index The presence and number of potential larval development sites (water-holding containers), such as tires, flower vessels, bottles, trash cans, or gutters, within each school was recorded. All containers and schools were visually assessed for the presence of immature forms and adults to calculate the following traditional Aedes indices: Adult Index (AI) (the percentage of schools infested with adults Aedes), house index (HI) (the percentage of schools infested with larvae and/or pupae), container index (CI) (the percentage of waterholding containers infested with larvae or pupae), and Breteau index (BI) (the number of positive containers per 100 schools inspected). Containers were counted as positive if they had immature forms of the vector, whereas schools were considered positive if at least one container with immature forms of Aedes spp. During 2016, schools were inspected once per trimester, and in 2017 and 2018 schools were visited once a month. We aggregated ento-virological data to a three-month scale to enable comparison to entomological risk indicators. The level of risk was based on three levels of risk for each index, as suggested by the Health Department of Antioquia: low risk (BI < 5%, HI < 4%, CI < 3%), medium risk (BI 5-50%, HI 4-35%, CI 3-20%), and high risk (BI > 50%, HI > 35%, CI > 20%).

Molecular Arbovirus Detection in Mosquitoes
Mosquitoes were divided into pools according to species (Ae. aegypti and Ae. albopictus) and collection site, each pool contained from 1 to 10 mosquitoes. Mosquitoes captured in 2016 were only tested for dengue virus, and in 2017 and 2018 samples were tested for dengue, chikungunya, and Zika. RNA was obtained with the RNeasy Mini Kit (Qiagen, Germantown, USA commercial kit. Each pool was mechanically macerated in 100 µL of lysis buffer (buffer RLT) and RNA was obtained with the RNeasy Mini Kit (Qiagen) commercial kit according to manufacturer instructions. Viral detection was performed with RT-PCR. Copy DNA (cDNA) was synthesized using 10 µL extracted RNA, 1× specific buffer for reverse transcriptase, 1 mM dNTPs, 10 µM direct primer specific for the dengue virus (DV1-5 -GGRACKTCAGGWTCTCC-3 ) and 1 µL reverse transcriptase (Fermentas, Vilnius, Lithuania), as well as incubation at 42 • C for 1 h and enzyme inactivation at 94 • C for 10 min. The following was used for dengue virus amplification: 3 µL cDNA, 1× buffer, 2.0 mM MgCI 2 , 0.2 mM dNTPs, 0.2 µM of each reverse primer specific for the four virus serotypes (DSP1-5 -AGTTTCTTTTCCTAAACACCTCG-3 , DSP2-5 -CCGGTGTGCTCRGCYCTGAT-3 , DSP3-5 -TTAGAGTYCTTAAGCGTCTCTTG-3 and DSP4-5 -CCTGGTTGATGACAAAAGTCTTG-3), and 0.5 U taq polymerase. PCR amplification was performed at an initial temperature of 95 • C for 2 min, followed by 35 cycles at 95 • C for 30 s, 55 • C for 1 min, and 72 • C for 40 s, with a final extension at 72 • C for 3 min. The size of the NS3 gene fragments obtained was 169 pb for DENV-1, 362 pb for DENV-2, 265 pb for DENV-3 and 426 pb for DENV-4. The protocol used by [25] was applied to detect Zika virus. For chikungunya amplification, the protocol proposed by [26] was used.
Amplification products were analyzed in 2% agarose gels dyed with ethidium bromide and observed under ultraviolet light. For each reaction, extraction and amplification negative controls were included, as well as positive controls comprising RNA of Zika and chikungunya viruses, and the four serotypes of dengue virus obtained from supernatant of infected cells. To confirm pool positivity, 15% of positive samples were randomly selected for bidirectional sequencing at Macrogen (Seoul, Korea).

Epidemiological Data
The number of cases of dengue and Zika in the city of Medellín was obtained from the Colombian public health surveillance system (SIVIGILA), and the accumulated values per trimester were used for the analyses carried out in this study.

Data Analysis
Traditional Aedes indices of container (CI), housing (HI), Breteau (BI), and adult mosquito infestation (AI) were calculated following the estimations recommended by the World Health Organization (TDR/WHO 2009). Infection rates (IR) were estimated using the maximum likelihood estimator (MLE) method for differently sized pools [27] through the pooledBin function implemented in the binGroup statistical package of R.
The parameter used to estimate IRs was "bias-corrected MLE" with "skewnesscorrected confidence Interval", as has been described previously in a study testing arboviruses in Colombia [28]. IRs without virus distinction were individually estimated for each mosquito species (Ae. aegypti and Ae. albopictus), and for both pooled species. IRs were estimated for each dengue serotype (DENV-1, DENV-2, DENV-3, and DENV-4), and Zika virus.
Following Shapiro-Wilk tests for normality of the data, the relationships between IRs and some epidemiological (reported dengue and Zika cases), entomological (Aedes indices), and climate variables were explored using Spearman's correlation analysis. The climate variables included mean temperature (MEANT), maximum absolute tempera-ture (MAXABST), minimum absolute temperature (MINABST), average daily maximum temperatures (AVMAXT), average daily minimum temperatures (AVMINT), maximum daily difference in temperature (MAXDIFT), minimum daily difference in temperature (MINDIFT), average daily difference in temperature (AVDIFT), total precipitation (PRECI), maximum precipitation in 24 h (MAXPRECI), number of days with precipitation (PRECI-DAYS), and relative humidity (RH). Epidemiologic data were provided by SSM, whereas monthly climate data were obtained from the Hydrology, Meteorology and Environmental Studies Institute (IDEAM, by its Spanish acronym) of Colombia. Correlations between entomological, epidemiological, and climate variables were analyzed at lags of 0, 1, and 3 trimesters. All analyses were performed in R.

Mosquito Collections
During the three years of this study, 2362 inspections were performed in schools, and Aedes spp. adult mosquitoes were captured in 43.3% of institutions. The percentage of institutions where adult Aedes spp. were found varied each trimester with no clear seasonal pattern between years: the highest rate of adult Aedes infestation in schools was trimester IV of 2017 (55.16%) and trimester I of 2016 was the lowest (14.29%) ( Table 1). whereas monthly climate data were obtained from the Hydrology, Meteorology and Environmental Studies Institute (IDEAM, by its Spanish acronym) of Colombia. Correlations between entomological, epidemiological, and climate variables were analyzed at lags of 0, 1, and 3 trimesters. All analyses were performed in R.

Mosquito Collections
During the three years of this study, 2362 inspections were performed in schools, and Aedes spp. adult mosquitoes were captured in 43.3% of institutions. The percentage of institutions where adult Aedes spp. were found varied each trimester with no clear seasonal pattern between years: the highest rate of adult Aedes infestation in schools was trimester IV of 2017 (55.16%) and trimester I of 2016 was the lowest (14.29%) ( Table 1). Overall, 2959 Aedes mosquitoes (1265 females, 1694 males) were captured, 91% of which were Ae. aegypti. Aedes aegypti were collected during all sampling periods, while Ae. albopictus were not (Figure 2).   I  II  III  IV  I  II  III  IV  I  II  III  IV   2016  2017  2018 Ae. aegypti

Mosquitoes/schools
Trimesters per year

Larval Habitats and Traditional Aedes Indices
Of the 2362 schools inspected, 15% had water-holding containers with immature forms of Aedes spp. A total of 8692 containers that were suitable for Aedes spp. growth were inspected (and removed in most cases), 8.3% of which tested positive for the presence of mosquito larvae or pupae. Similar to what was observed for adult mosquitoes, the percentage of institutions with containers positive for immature mosquitoes varied by trimester with no clear pattern by trimester between years (Table 2). According to the PAHO risk parameters (1995), all Aedes indices obtained were at a medium risk level. The highest BI and HI levels were recorded in trimester III of 2016, and the highest CI level was recorded in trimester IV of the same year. However, trimester I of 2016 had lower BI and CI values and trimester II of 2018 had a lower HI value (Table 3).

Molecular Arbovirus Detection in Mosquitoes
Captured mosquitos were divided into 883 pools; 20% of them were positive for dengue in 2016, 6.92% in 2017, and 2.60% in 2018. For ZIKV 9.49% were positive in 2017 and 7.80% in 2018. Chikungunya was not detected in any of the mosquitoes processed during the study period. Across the entire study period, a higher percent of Ae. aegypti pools were positive than Ae. albopictus pools. Some Ae. aegypti pools presented simultaneous DENV/ZIKV infection (Table 4).

Infection Rates
Cumulative IRs for Aedes spp. varied each trimester, ranging from 2.93 (trimester IV of 2018) to 282.03 (trimester I of 2018). For Ae. aegypti, IRs ranged from 1.75 (trimester IV of 2018) to 296.29 (trimester I of 2018). These rates were higher for Ae. albopictus because of the low number of mosquitoes (259) compared to Ae. aegypti (2692), which resulted in IRs ranging from 0 in most trimesters to 573.98 in trimester I of 2018 ( Figure 3).

Infection Rates
Cumulative IRs for Aedes spp.   During 2016 and the beginning of 2017, all four DENV serotypes were found; during this same period, Medellín recorded the highest levels of dengue transmission [4]. DENV-1 was the only serotype detected after trimester III of 2017. A single specimen was found to be coinfected with both DENV-2 and DENV-4 (Figure 4). The other mixed infections pools were grouped by several specimens.  Changes in IRs for the different serotypes were also observed over time. DENV-1 IR was registered through the entire study and consistently was found at the highest values. The remaining serotypes had similar IR patterns over time ( Figure 5); 45% of pools were positive for Zika in trimester I of 2018; this period had the highest infection rate which is also reflected in the IRs over time ( Figure 5).

Relationship between Infection Rates and Epidemiological, Entomological, and Climate Variables
We found a significant relationship between Ae. aegypti and Aedes spp. IRs and number of human dengue cases (Table 5), but no significant relationship was observed between Ae. albopictus IRs and disease cases (data are not shown).
For Aedes container indices, no statistically significant correlation was found between any entomological indicator and arbovirus infection rates, and no statistical correlation was found for the Lag periods considered. No correlation was found between adult index and these parameters ( Table 6).

Relationship between Infection Rates and Epidemiological, Entomological, and Climate Va ables.
We found a significant relationship between Ae. aegypti and Aedes spp. IRs and nu ber of human dengue cases (Table 5), but no significant relationship was observed tween Ae. albopictus IRs and disease cases (data are not shown).    Table 7 shows the relationship between climate variables and IRs for Ae. aegypti and Ae. albopictus. Interestingly, results were inconsistent. A positive correlation between minimum absolute temperature, average daily maximum temperatures, minimum daily difference in temperature, average daily minimum temperatures and Ae. aegypti IRs was observed; however, it had a negative correlation with precipitation. On the other hand, the mean temperature, average daily maximum temperatures, minimum daily difference in temperature had a negative correlation with Ae. albopictus IRs, while relative humidity had a positive correlation.

Discussion
Traditionally, it has been assumed that dwellings are the main site of dengue transmission given the endophilic and anthropophilic behavior of Ae. aegypti. Therefore, most studies measuring vector abundance and routine control have focused on homes [20,[29][30][31]. However, because few studies have assessed the risk for transmission in areas other than dwellings that are also highly trafficked and known to have mosquitoes, it is not clear if non-residential locations such as schools are also important places of arboviral transmission [29]. Mosquito abundance in schools has already been recorded in some countries such as Mexico [32], Brazil [33], Thailand [34,35], Peru [14], and Colombia [17,36], but measuring mosquito abundance is inadequate to accurately assess arboviral disease transmission risk. As far as the authors are aware, this is the first study to explore the rate of arbovirus infection in mosquitoes in schools and analyze the relationship between abundance indices, infection rates, and cases.
Our results showed that both immature and adult forms of Aedes spp. mosquitoes were detected in a high percentage of schools (43.27%). This is not surprising given the anthropophilic behavior of these vectors. The student population spends a significant amount of time at school, and therefore they may be important transmission foci for arboviruses. A high dengue incidence in student populations has been extensively documented in countries such as Nicaragua [37], Venezuela [2], Thailand [35], Vietnam [38], and even already in Colombia [36,39]. Furthermore, the incorporation of schools into routine vector control programs has proved to be effective in decreasing the presence of both vectors and disease [40]. Thus, they should be key sites for interventions aimed at arbovirus prevention. The presence of Ae. albopictus inside schools was a relevant finding, which can be explained by the rapid and aggressive reproduction of this mosquito. Although Ae. aegypti is the main vector in Medellín, the continuous increase of Ae. albopictus abundancy, its presences across at least 70% of the city neighborhoods [41], and the possibility this species may not be well controlled contribute to increase its role as a vector, not only in Medellín, but also in other cities where it is distributed [42].
Containers with Aedes spp. were reported in 15.5% of inspections, although the entomological indices for the periods assessed had a medium risk level according to PAHO parameters (1995). Even during the 2016 epidemic, these values were not higher than the other years in the schools. Our study showed no correlation between Aedes indices and dengue infection rates in mosquitoes. The use of Aedes indices as risk indicators for dengue transmission has been discussed by different researchers because several studies have reported no association between them and the amount of disease cases [18,27,43]. Some of the weaknesses of these indices are that they do not consider adult mosquito productivity, and most searches are performed on visible containers, thus leaving cryptic breeding containers unattended. Consequently, these indices may not provide full and accurate information for a correct estimation of the transmission risk [43], leading to a defective representation of the actual number of vectors. However, these indices are still being used and required by government agencies in several countries.
Detection of arboviruses in field mosquitoes has been suggested as effective tool to identify outbreaks risk [19].Our results evidenced natural infection of Aedes aegypti and Ae. albopictus species with dengue virus in the schools of Medellín. Natural infection of Ae. albopictus with dengue and Zika viruses has already been reported [9,42,44]; however, its role as an arbovirus vector in the country should be further clarified, especially considering the results in our study. The low amount of Ae. albopictus as opposed to Ae. aegypti is attributed to the biology of this species, which is less anthropophilic than Ae. aegypti [45]. However, the high infection rates observed for Ae. albopictus can be explained by the low number of pools evaluated, which results in extensively varied infection rates and less solid statistical analysis. Another aspect that could contribute to this low number of captures was the method implemented; the efficiency of the sampling with oral aspirator is affected by the time spent for sampling and the ability of the person implementing the capture. These variables can cause the number of captures to be low, leading to an inaccurate estimate of infection rates. Consequently, further studies are required given the epidemiological relevance of this species in schools and the significant role it plays in arbovirus transmission in other countries [46,47].
A study performed in schools of Merida (Mexico) showed that 16.1% of the pools assessed were positive for dengue [32], whereas a higher percentage (20%) was observed in this study for 2016. However, this number decreased in 2017 and 2018. Our results may be explained by the fact that a large dengue epidemic was recorded in Medellín in 2016, in which all four serotypes circulated [20]. This is the first study the authors are aware of reporting mosquitoes infected with ZIKV in schools. Other studies [48] have shown natural infection in mosquitoes captured in dwellings, with a positivity percentage similar to what was observed in our study (10% in 2017 and 8% in 2018).
We analyzed which variables influence infection rates, and the results in our study suggested an inversely proportional relationship between precipitation and infection rates, suggesting that warmer and drier periods favor dengue transmission. However, there are divergent results regarding the relationship between precipitation and the disease, e.g., certain studies have shown a positive correlation explained by the existence of vector breeding sites, which could enhance transmission [49]. However, other studies have shown a negative correlation with dengue incidence [50]. We also found a high number of significant correlations with infection rates for Ae. aegypti. These associations were closely related with dengue and Zika cases reported in Medellín, which suggests that Ae. aegypti is mainly responsible for dengue and Zika viruses transmission in the city or, at least, in schools. Similar results were found by [32] in schools of Merida.
Additionally, we found that temperature variables have a positive effect in arbovirus infection rate, the association between temperature and arbovirus replication and transmission had previously been described [51][52][53], particularly in the municipality of Bello, which shares urban areas with Medellín city [27]. Moreover, it has been reported that an increase in environmental temperature raises the onset of dengue in cities with a mean temperature like the one registered for Medellín [28]. Accordingly, the case increase observed in 2016 could have been attributed to a general temperature increase observed during that year. During these periods, Ae. albopictus mosquitoes showed infection rates of zero, which could account for the negative coefficient correlations observed in our study for that species.

Conclusions
In conclusion, our results evidence the importance of incorporating schools into arbovirus surveillance programs to achieve a higher impact on arbovirus prevention and control activities. Additionally, detection of arbovirus-infected mosquitoes raises an early alarm about the potential risk of transmission in these sites, especially because students, professors and any people working in these places are exposed to infected mosquito bites most of the day. Community participation is essential to control this type of diseases, especially school participation, to eliminate breeding containers inside schools and reduce mosquito populations and consequently reduce the risk of being bitten by an infected female vector.