Three-dimensional body reconstruction enables quantification of liquid consumption in small invertebrates

Abstract

Quantifying feeding patterns provides valuable insights into animal behaviour. However, small invertebrates often consume incredibly small amounts of food. This renders traditional methods, such as weighing individuals before and after food acquisition, either inaccurate or prohibitively expensive. Here, we present a non-invasive method to quantify food consumption of small invertebrates whose body expands during feeding. Using the markerless pose estimation software DeepLabCut, we three-dimensionally track the body of Argentine ants,Linepithema humile. Using these extracted markers, we developed an algorithm which computationally reconstructs the ant’s body, directly measuring volumetric change over time. Moreover, we provide measures of accuracy and quantify the ant’s feeding response to a range of sucrose concentrations, as well as a gradient of caffeine-laced sucrose solutions. Small invertebrates are often prolific invasive species and disease vectors, causing significant ecological and economical damage. Understanding their feeding behaviour could be an important step towards effective control strategies.

Invasive ants fed spinosad collectively recruit to known food faster yet individually abandon food earlier

“…Alternatively, spinosad could induce a state of hyperactivity, causing ants to move faster, be more motivated to find food or even forage more efficiently. In fact, Argentine ants fed moderate doses of caffeine, a neuroactive chemical, have been shown to have improved foraging (Galante et al ., 2024). Spinosad, predominantly acting on nicotinic acetylcholine receptors, has also been identified as an antagonist of GABA receptors (see Millar & Denholm, 2007 and Kirst, 2010 for reviews).…”

“…Interestingly, sub-lethal doses of the neonicotinoid imidacloprid, also an antagonist of nicotinic acetylcholine receptors, were shown to shift colony-level preference in the invasive ant Lasius neglectus towards toxicant-laced solutions (Frizzi et al ., 2022), even though imidacloprid has been shown to impair olfactory learning and memory in honeybees (Yang et al ., 2012; Williamson & Wright, 2013). However, were this the case, control attempts using spinosad as the toxicant would likely have been more successful, unless such effects are lost at high doses, as they are for caffeine and imidacloprid (Frizzi et al ., 2022; Galante et al ., 2024).…”

“…Following the final Y-maze test, around 30 unmarked workers from each colony were removed and their feeding patterns, both total volume ingested and consumption rate, were quantified using 3D body reconstruction following Galante, Czaczkes & De Agrò (2024). Half of the individuals tested were fed a sucrose apple solution and the other half a sucrose strawberry solution.…”

“…The setup consisted of a resin 3D-printed platform where the test solution was placed, attached to a Raspberry Pi HQ camera system and tracked using DeepLabCut version 2.3.8 (Mathis et al ., 2018; Nath et al ., 2019). The same DeepLabCut networks trained in Galante, Czaczkes & De Agrò (2024) were used, but the location of the cameras was adjusted and thus a new calibration performed (1px = 1.05 ± 0.20mm, N = 505). Each ant was recorded for its entire feeding event and removed from the platform once full, yet not returned to its colony of origin.…”