Trapping device. The trap used in this study was a modified BG-sentinel (Biogents AG, Regensburg, Germany, hereafter BG) equipped with a feeding system (Fig. 1) designed to keep mosquitoes alive longer. This system includes: i) a collection chamber that provides a more comfortable environment than the original mesh bag, reducing stress and mortality of mosquitoes during trap activity; ii) a pipe system that reduces airflow in the chamber, preventing rapid dehydration of mosquitoes; iii) an FTA card sugar delivery system (feeder) through which collected mosquitoes can release pathogens during the sugar meal. The feeder is a plastic tube filled with a honey-based solution (2% hydroxy-ethyl-cellulose in water and honey in 3:2 proportion) in which a Whatman FTA Classic card is partially soaked. The solution is dyed with 0.15% methylene blue (a dye with low toxicity for mosquitoes) to allow the identification of sugar-fed mosquitoes. The feeder was preliminarily laboratory tested on Aedes albopictus mosquitoes to assess the sugar feeding rates (Supplementary file 1: Table S1).
Collection sites and field sampling. The field samplings were carried out in eleven municipalities of Veneto Region (North-eastern Italy, Fig. 2): Badia Polesine, Ceneselli, Ficarolo, Minerbe, Villa Bartolomea, Nogarole Rocca, Oppeano, Erbè, Jesolo, Caorle. In 2021, a single site in Selvazzano Dentro was sampled. All sites were characterized by both high vector density and endemic circulation of WNV and USUV, as reported from historical data obtained between 2010 and 2020 by the Istituto Zooprofilattico Sperimentale delle Venezie (IZSVe)37.
During 2019 sampling, four collections were carried out on alternate weeks in ten sites from July to August, resulting in a total of 112 observations. Our BG trap prototype was compared to a CDC-like trap (Italian Mosquito Trap; PeP, Cantu, Italy; hereafter CDC), which was chosen as a comparator because it is considered a highly effective device for Culex pipiens sampling in Italy38 and is the most used trap in WNV surveillance39. Both traps were provided with 2 kg of dry ice as a source of CO2. In addition, the BG trap was baited with BG-Lure (Biogents). The traps were deployed approximately 50m from each other to avoid interference between them. The CDC trap was left active for about 24 hours, while the BG trap was left active for an additional day, replacing dry ice and battery at the time of mosquito collection on the first day of sampling. Because the CDC traps utilized for WNV surveillance are consistently placed in the same location each year, it was not feasible to rotate the positions of the CDC and BG traps.
To investigate the performance of the BG trap over multiple working days in relation to mosquito infection prevalence, in 2021, only one site (Selvazzano Dentro) was tested (July-August) following a different experimental design. Five BG traps were left operational for four consecutive days every week, while a single CDC trap was used as a control, working for one night only, following the setup of the 2019 sampling. During the four-day collections of the BG traps, the CO2 was continuously supplied by a gas cylinder, and the fan was powered through power line. The mosquitoes were collected at the end of the fourth day.
Mosquito identification and sample processing. The collected specimens were morphologically identified according to standard taxonomic keys40 and divided in pools of maximum 100 females, based on collection date, site, trapping method and species. In the first year of sampling (2019), the presence of blue dye was also assessed in all identified mosquitoes to determine the daily feeding rate for each species and predict the presence of saliva on the FTA card. All collected mosquitoes were identified while maintaining the cold chain and stored at -80°C. The FTA cards were individually placed into 2-ml Eppendorf tubes for at most 7 days at room temperature until following analysis. This time window was compatible with the detection of WNV according to preliminary tests performed under semi-field conditions (Supplementary file 1: Table S2, Fig. S1 and S2) and evidence from literature41. To avoid contamination, all cards were analysed separately from mosquitoes and on different days, sterilizing all the handling instruments after each sample manipulation.
Viral RNA extraction from FTA cards and mosquitoes. Viral RNA from FTA cards was extracted using the QIAamp Viral RNA Kit (QIAGEN, Valencia, CA, USA). AVL buffer (Viral Lysis Buffer with carrier RNA, QIAGEN), EtOH and RNA carrier amount were increased proportionally for a starting volume of 200µl (800µl AVL buffer, 8µl RNA carrier, 800µl EtOH). The samples were shacked for 2 hours at room temperature after rehydration with AVL buffer and then the extraction proceeded according to the manufacturer’s protocol. RNA from pooled mosquitoes was extracted with an automated nucleic acid extraction system, to decrease hands-on time, increase sample throughput and reduce the risk of contamination. Before extraction, two 5mm Tungsten Carbide Beads were added to each mosquito pool and the samples homogenized with the TissueLyser II (QIAGEN) at 30 Hz per 30’’ for two rounds. RNA was extracted from homogenate with an automatic extractor (Microlab STAR Hamilton, Americas, Australia & Pacific Rim) using MagMAX Pathogen RNA/DNA kit (Thermo Fisher Scientific, Waltham, MA, USA) following the high-volume manufacturer protocol.
PCR protocols and sequencing for virus detection. RNA extracted from pooled mosquitoes and FTA cards was screened for the presence of flaviviruses using a RT-PCR, followed by hemi-nested PCR42 and sequencing. A SYBR Green-Based RT-PCR targeting 250bp of the conserved region of the non-structural NS5 gene was performed in a final volume of 20µl containing 10µl of 2X QuantiNova SYBR Green RT-PCR Master mix (QIAGEN) (final concentration 1X), 0.25µl of QuantiNova RT mix, 5.15µl of RNase free water, 1µl of 10µM of MAMD forward primer (final concentration, 0.5 µM), 0.6µl of 10µM of cFD2 reverse primer (final concentration, 0.3µM) and 3µl of RNA template. The PCR thermal cycling was performed with Applied Biosystems StepOnePlus Real-Time PCR System (Thermo Fisher Scientific) for samples collected in 2019 and with MIC (BMS, Resnova, RM, Italy) for samples of 2021, as follows: initial incubation of 10 min at 50°C and 2 min at 95°C, amplification of 45 cycles at 95°C for 5s and 30s at 60°C, dissociation melting from 60°C to 95°C with a ramping rate of 0.3°C/s. Analysis of the melting curve was carried out to determine the presence of viral RNA and the homogeneity of PCR products.
Positive results were confirmed with hemi-nested PCR and sequencing. PCR amplification was performed using FS788 e CFD2 primers42 targeting 220bp of NS5 gene in a final volume 50µl containing 5µl of 10X Buffer II (AmpliTaq Gold DNA Polymerase with Buffer II and MgCl2, Applied Biosystems, Thermo Fisher Scientific) (final concentration 1X), 4µl of 25mM MgCl2 (final concentration 2.0mM), 1µl of 10mM dNTP (final concentration 0.2mM), 2.5µl of 10µM of FS778 forward primer (final concentration 0.5µM), 2.5µl of 10µM of CFD2 reverse primer (final concentration 0.5µM), 5U of AmpliTaq Gold (final concentration 2.5U), 33.5µl of ultrapure DEPC-pre-treated H2O and 1µl of cDNA.
The PCR thermal cycling used for cFD2 and FS 778 primers were performed as follows: incubation of 10 min at 95°C followed by 25 cycles of denaturation for 30s at 94°C, annealing at 54°C for 30s, extension at 72°C for 30s and final extension at 72°C for 3 min.
The amplification products were identified by their molecular weights through electrophoresis in a 7% agarose gel stained with SYBR Gold Nucleic Acid Gel Stain 1X (Invitrogen, Thermo Fisher Scientific) and visualized under UV light using Gel Doc XR + Gel Documentation System (Bio-Rad, Hercules, California, USA). Positive PCR products were purified and sequenced in both directions using the same forward and reverse primers of heminested PCR, employing a 16-capillary ABI PRISM 3130xl Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). Sequence data were assembled and edited with SeqScape software v2.5 (Applied Biosystems). Sequences obtained were aligned and compared with representative sequences available on GenBank database using Basic Local Alignment Search Tool (BLAST; http://blast.ncbi.nlm.nih.gov/Blast.cgi).
Statistical analysis
Different generalised linear models (GLM) were used to investigate: 1) the relation between mosquito abundance observed with the BG with that observed with the CDC trap; 2) the variation of mosquito species diversity in relation to the trap; 3) the mosquito infection rate according to trap and working days. More specifically:
1) To define the abundance of mosquitoes collected by the BG in relation to the CDC trap, we tested a GLM model with negative binomial distribution (to overcome data overdispersion). The mosquito abundance of the first day of collection was used as the response variable. The trap type, the Julian day and the sampling sites were included as explanatory variables. In addition, we included the working time as an offset term, to account for differences in the sampling effort for each collection. This model was run for the total female mosquito abundance and separately by species.
2) To investigate the mosquito diversity, we calculated the Shannon diversity Index (SH) as:
$$\text{S}\text{H} = \sum _{i=1}^{s}{p}_{i}\text{ln}{p}_{i}$$
where pi is the proportion of individuals of the ith species divided by the total number of individuals found in each collection and S is the species number. We then developed a GLM with a normal distribution, with SH as response variable and as covariates the same covariates that we used in model 1.
3) To estimate the infection rate from pooled mosquitoes we calculated the maximum likelihood estimate (MLE) for each of detected virus, using the approach developed by CDC (https://github.com/CDCgov/PooledInfRate). We estimated the point and confidence interval of the MLE for each sampling year and for each detected virus, based on binary values samples of pooled specimens in relation to trap method, number of tested pools, size of tested mosquitoes per pool, and site (in case of 2019 sampling) or week (in case of 2021 sampling). We also estimated the MLE value for each year in relation to mosquito species and trap, including the number of tested pools and size of tested mosquitoes per pool.
All analyses were performed in the statistical environment R v.4.0.543, using the following packages: PooledInfRate44, mass45, performance46, visreg47, ggplot248, dplyr49 and reshape50.