Comparing the Efficiency of Different Carbon Dioxide Sources in Collecting Mosquito Vector Species on the Tempe Campus of Arizona State University

Carbon dioxide (CO₂) plays an important role in the process through which mosquitoes find their human hosts (Smallegange et al. 2010, Jerry et al. 2017). Adding CO₂ sources to mosquito traps is known to enhance capture rates of Culex spp. (vectors of, e.g., West Nile virus) (Jerry et al. 2017), Aedes spp. (vectors of, e.g., Zika, dengue, and yellow fever viruses) (Saitoh et al. 2004), and Anopheles spp. (vectors of malaria pathogens) (Smallegange et al. 2010). The CO₂ sources commonly used in vector surveillance programs include sugar-fermented yeast (Smallegange et al. 2010, Jerry et al. 2017), dry ice (Oli et al. 2005, Benante et al. 2019), and pressurized CO₂ stored in cylinders (Gillies 1980), but it is often not studied if there are differences in the efficacy in attraction of mosquitoes between different CO₂ sources at a single site. In this operational note, we report on a comparison of the effectiveness of 4 different CO₂ sources in collecting mosquito vector species on the Tempe campus of Arizona State University, using the BG-Pro mosquito trap (Biogents AG 2022; EVS style). Traps were hung with the trap opening 66 cm aboveground in 4 different areas (at least 132 m apart, with the furthest distance being 644 m).

The following 4 CO₂ sources were tested in a 4 × 4 Latin square design: 1) Sugar-fermented BG-CO₂ (a formulation of different yeast strains) (Biogents AG 2021); 2) Sugar-fermented Fleischmann active dry yeast (ACH Food Companies Inc., Oakbrook Terrace, IL). A solution of each yeast product was prepared daily by adding 20 g of yeast and 500 g of white refined household sugar to 2 liters of warm water (38°C) in an inflatable BG-Pro mixing bag (Biogents AG 2021). The mixing bag was closed, and the solution was mixed thoroughly. The sugar-fermented yeast mixture was prepared 1 h before the start of the experiment. The bags were placed next to the trap on the ground, and CO₂ was released right above the trap opening through a silicone tube (Biogents AG 2021); 3) Dry ice: solid blocks of dry ice were placed inside the BG-Pro carrier bag (Biogents AG 2022), which was closed and hung above the trap (distance between trap opening and carrier bag was approximately 15 cm). During experiment 1, 3 lb (1,362 g) of dry ice were used; during experiment 2, 6 lb (2,742 g) of dry ice were used to account for the longer duration of the experiment; 4) Pressurized CO₂ cylinders: 5-lb (2,270-g) CO₂ cylinders were placed next to the trap. Carbon dioxide was released at a flow rate of 0.5 kg CO₂ per day (which equals approximately 175 ml/min) right above the trap opening through a silicone tube (Biogents AG 2022).

Two separate experiments were conducted consecutively. The 1st experiment consisted of 16-h mosquito surveillance periods. From 4 p.m. to 8 a.m. the next day, the BG-Pro traps ran simultaneously in all 4 locations, each with a different source of CO₂. As we observed Aedes aegypti (L.) in our traps, and because this species is known to be a late-afternoon and morning biter (Mutebi et al. 2022), we repeated the experiment but ran all traps and CO₂ sources for 24-h periods, starting daily at 2 p.m. Mosquitoes were collected daily and sorted by sex. Males were discarded and females were identified to species using a dichotomous key (APHC 2016).

The sample size for each treatment group in this study for both experiments was 20 (collection periods), based on the outcome of a power analysis (predicting 17 collection periods for each treatment group) using G*Power (Erdfelder et al. 1996) and ANOVA repeated measures for count data. All other analyses were performed using R version 4.0.2 (R Development Core Team 2020). Generalized linear mixed models were performed using packages “lme4” (Bates et al. 2015) and “MASS” (Venables and Ripley 2002). The fixed main effect was treatment. The model corrected for location, and day was added as a random factor. A negative binomial distribution was used to correct for overdispersion. Tukey HSD multiple comparisons tests were performed using the package “multcomp” (Hothorn et al. 2008) to identify differences in the specific pairs of CO₂ source treatments.

In the 16-h surveillance, a total of 958 female mosquitoes were collected over 20 days. Culex quinquefasciatus Say (n = 853) and Ae. aegypti (n = 90) were the most abundant mosquito species, followed by Culiseta inornata (Williston; n = 12) and Cx. tarsalis Coquillett (n = 3). Overall, dry ice collected the highest number of mosquitoes (n = 330), followed by the gas cylinder (n = 308), sugar-fermented BG-CO₂ (n = 184), and sugar-fermented Fleischmann's yeast (n = 136). Catches for Cs. inornata and Cx. tarsalis were too low for formal statistical analysis. Aedes aegypti catch rates were not significantly different between the different CO₂ sources; however, there were significant differences in the catch rates between treatments for Cx. quinquefasciatus. Tukey HSD multiple comparisons test showed that the CO₂ cylinder and dry ice caught 1.66 times and 1.75 times more Cx. quinquefasciatus than BG-CO₂ (P = 0.047 and 0.0205), respectively, and 2.2 times and 2.3 times more Cx. quinquefasciatus than the sugar-fermented Fleischmann yeast (P < 0.001), respectively (Fig. 1).

In the 24-h surveillance, a total of 529 female mosquitoes were collected over 20 days. Again, Cx. quinquefasciatus (n = 375) and Ae. aegypti (n = 121) were the most abundant mosquito species collected, followed by Cx. tarsalis (n = 33). Culiseta inornata was not collected. Overall, dry ice collected the highest number of mosquitoes (n = 160), followed by the gas cylinder (n = 151), sugar-fermented Fleischmann's yeast (n = 119), and sugar-fermented BG-CO₂ (n = 99). Catches for Cx. tarsalis were too low for formal statistical analysis. The Tukey HSD multiple comparisons test showed that there were no significant differences in total catch ratch rates between CO₂ sources for Ae. aegypti and Cx. quinquefasciatus (both P > 0.05) (Fig. 2).

Our results demonstrate that the CO₂ cylinder and dry ice were more effective in catching Cx. quinquefasciatus in the 16-h experiment. Although Ae. aegypti was the second-most abundant mosquito species collected in both experiments, we found no significant differences in catch rates using the different CO₂ sources. Monitoring Cx. quinquefasciatus is important, as this species vectors West Nile virus (WNV) (Crockett et al. 2012) and St. Louis encephalitis virus (SLEV) (Swetnam et al. 2020) in Arizona. In a large 2021 outbreak, there were 569 confirmed human cases of West Nile and 121 people who died from the disease (ADHS 2022). Monitoring Ae. aegypti is also a priority, as it is a potential vector of arboviruses like Zika and dengue (Godoy et al. 2021). Although local transmission of Zika is currently absent in Arizona, two autochtonous cases of dengue were reported in Maricopa County in November 2022 (Kretschmer et al. 2023). Also, these diseases are circulating in neighboring Mexico (Jones et al. 2016). While catches for Cx. tarsalis and Cs. inornata were too low for statistical analysis, these species are known vectors of SLEV and WNV (Goddard et al. 2002), and should therefore be monitored.

Decisions on the type of CO₂ to use will depend on the local context, and do not only depend on mosquito species abundance and diversity. For example, it remains difficult to obtain and/or generate a cheap and reliable CO₂ output in remote areas, as demonstrated in sub-Saharan Africa (Smallegange et al. 2010). Dry ice is not easy to obtain, and gas cylinders are heavy and bulky, and hard to safely transport from one location to the other. Currently, sugar-fermented yeast solutions are a more reliable, cheaper, and easier method to produce CO₂ for adult mosquito sampling in many geographical areas around the world (Saitoh et al. 2004). Studies show that yeast-generated CO₂-baited traps capture more mosquitoes than traps that are not baited (Smallegange et al. 2010). Alternatively, other mosquito trapping methods have been developed over recent years that use CO₂ mimics (e.g., 2-butanone, acetone, and cyclopentanone [Kessy et al. 2020]) and human baits (producing natural CO₂) for the surveillance of mosquitoes (e.g., Alafo et al. 2022).

The present study has several limitations. The CO₂ release rates were not quantified for the sugar-fermented yeast solutions and the dry ice. The climate (temperature) may affect the CO₂ release rate as the yeast ferments over time (Guadalupe-Daqui et al. 2023). The CO₂ release rate of dry ice is highly variable and diminishes over time as the ice sublimates (Smallegange et al. 2010). Also, we conducted our study during the early summer (May 2021–June 2021) and in only 4 selected locations on the Tempe campus of Arizona State University. Future studies should assess the effectiveness of CO₂ sources across seasons and in several locations, as confounding factors (such as climate, human population densities, and water availability) are known to affect mosquito abundance. Furthermore, although our sample size was larger than the estimated sample size in our power analysis, future studies would need a larger sample size considering the large variation in mosquito collections in this study. Additionally, future studies should also test different trap types, some of which are designed specifically for the capture of Culex species or Ae. aegypti. This information will allow us to design an effective mosquito surveillance network, and—combined with molecular analysis to detect viruses—to understand which vectors pose direct and indirect health threats in the area.

We thank Sarah Rydberg and Brook Jensen for their help with the study logistics, and Carolin Degener for assisting us with the statistical analysis. Biogents AG (Germany) provided in-kind support in the form of BG-CO₂ yeast and bags and the BG-Pro mosquito traps, but was not involved in the data collection, decision to publish, or preparation of the manuscript.