Nutrient preference of Neotropical ant assemblages along a habitat strata gradient and across trophic levels

In different biome, we installed 10 transects as sampling units (6 × 10 = 60 transects in total) separated by at least 1 km on average. Each transect was 740 m long with 75 sampling points, each separated by 10 m. At each sampling point, we provided one of five liquid food resources into one of three habitat strata (subterranean, epigaeic and arboreal). Hence, within each transect, each resource and strata combination (5 × 3 = 15 combinations) were repeated five times in sequence along each transect. Consequently, there were five pseudoreplicates of the 15 resource/habitat strata combinations per transect (5 pseudoreplicates × 5 resource types × 3 habitat strata = 75 sampling points per transect), giving a total of 750 sampling points per biome (75 sampling points × 10 transects = 750) and 4500 sampling points across the study (750 points × 6 biomes = 4500 points in total). We placed the food resources in 50-mL Fisher Scientific polypropylene centrifuge tubes that had a 5 cm cotton ball containing 10 ml of solutions in distilled water for each resource type. The solutions were: 1% sodium (NaCl), 20% sugar (CHO, made with sucrose), 20% amino acids (AA, made with unflavoured whey protein isolate which contains L-glutamine and other branched-chain amino acids such as leucine, isoleucine, and valine), lipids (100% extra virgin olive oil, without water), and distilled water as a control. Analogous resource solutions have successfully been used in previous studies (e.g., Fowler et al., 2014; Peters et al., 2014; Guariento et al., 2021; Lasmar et al., 2021). The placement of the tubes was different for each habitat strata. In the subterranean strata, we placed the tubes in a plastic box (4.5 cm high, 8 cm wide and 15 cm long) with an access hole <em>c</em>. 1.5 cm in diameter on each side. We buried the box <em>c</em>. 20 cm below the soil surface. In the epigaeic strata, we placed the baited tubes horizontally on the ground. For the arboreal strata, we fixed the tubes horizontally <em>c</em>. 2 m above the ground on the tree trunk closest to the sampling point, using a piece of adhesive and a string to keep the tube in place. The placement of the baited tubes began at 7:00 am for each transect except in the Caatinga biome. We delayed the baiting arrangement in Caatinga until 1:00 pm because of morning rainfall that could potentially reduce ant activity. We restricted all of our sampling to take place during only sunny or partially cloudy days and never during rainfall events. We left tubes operating for three hours. After that, we closed the tubes to sample the ants inside. We considered a tube as visited when there was at least one ant individual inside. To understand whether our baits were representatively sampling from the wider ant community we also sampled with pitfall traps. The pitfall trap data act as a proxy of ant abundance for all 60 of the baiting transects across the six biomes. By comparing ant abundance in pitfall traps and the ant foraging activity recorded by the baited tubes, we would know: (i) whether ants did not visit a bait because they did not occur in this habitat or (ii) whether baits did not attract ants due to methodological issues; (e. g. high ant abundance in pitfall traps but low ant foraging in baiting tubes, possibly indicating a problem with our baiting experiment). Thus, we started the pitfall sampling transects after the baiting experiment to avoid possible ant population depletions caused by pitfall traps (Lasmar et al. 2017). Pitfall transects were placed 20 m away in parallel from the corresponding baiting transects, totalling 60 pitfall transects across the entire study. Pitfall transects were 200 m long and had five sampling points separated by 50 m. At each sampling point, we collected ants at three habitat strata using arboreal, epigaeic and subterranean pitfall traps. Arboreal traps were installed at 1.5 m above the ground, tied in the middle of a tree trunk. Epigaeic traps were installed at the ground level and subterranean traps were buried at 20 cm under the ground. Pitfall traps were 8 cm in diameter and 12 cm in depth, and contained a 200 ml solution of water, salt (0.4%) and liquid soap (0.6%). Arboreal and epigaeic traps had a lid to cover and protect them against rain and sunlight. Subterranean traps were closed with lids and had four lateral holes of <em>c</em>. 1.5.cm in diameter on the sides. All pitfall traps remained in the field operating for 48 h.在不同生物群区（biome）中，我们布设10条样带（transect）作为采样单元，各条样带平均间隔不低于1km，总计设置6×10=60条样带。单条样带总长740m，设有75个采样点（sampling point），相邻采样点间距为10m。在每个采样点，我们会向3类生境层（habitat strata，包括地下、地表与树栖）中的一处投放5种液态食物资源之一。因此，在每条样带内，每种食物资源与生境层的组合（5×3=15种组合）将沿样带依次重复5次。由此，每条样带内15种资源/生境层组合各设有5个伪重复（pseudoreplicate），即5个伪重复×5种资源类型×3类生境层=75个采样点/样带；单个生物群区总计750个采样点（75个采样点×10条样带=750），整个研究共获取4500个采样点（750个采样点×6个生物群区=4500）。我们采用50mL的Fisher Scientific聚丙烯离心管作为投放容器，每根离心管内放置一团直径5cm的脱脂棉，蘸取10mL溶于蒸馏水的对应食物溶液。所用溶液包括：1%氯化钠（NaCl）溶液、20%糖类溶液（CHO，以蔗糖配制）、20%氨基酸溶液（AA，以无风味乳清分离蛋白配制，内含L-谷氨酰胺及亮氨酸、异亮氨酸、缬氨酸等支链氨基酸）、纯脂质（100%特级初榨橄榄油，不含水分），以及作为空白对照的蒸馏水。类似的食物资源溶液已在既往研究中成功应用（如Fowler等，2014；Peters等，2014；Guariento等，2021；Lasmar等，2021）。不同生境层的离心管布设方式存在差异：地下生境层中，我们将离心管置于尺寸为4.5cm高、8cm宽、15cm长的塑料盒内，盒两侧各开一个直径约1.5cm的出入口，随后将该塑料盒埋于地表下约20cm处；地表生境层中，我们将装有诱饵的离心管水平放置于地面；树栖生境层中，我们将离心管水平固定于距采样点最近的树干上，高度约2m，使用胶带与细绳将其固定到位。每条样带的诱饵投放均于当日7:00启动，但卡廷加生物群区（Caatinga biome）除外。受晨间降雨可能降低蚂蚁活动率的影响，我们将该区域的诱饵投放时间推迟至13:00。所有采样仅在晴天或多云天气开展，绝对避开降雨时段。我们将离心管放置3小时后封闭管口，采集管内的蚂蚁。当离心管内至少存在1只蚂蚁个体时，即判定该管已被蚂蚁造访。为验证诱饵采样是否能代表性覆盖更广的蚂蚁群落，我们同步采用巴氏罐（pitfall trap）开展采样。巴氏罐采样数据可作为60条诱饵样带所属6个生物群区的蚂蚁丰度代理指标。通过对比巴氏罐捕获的蚂蚁丰度与诱饵管记录的蚂蚁觅食活动，我们可明确两类问题：(i) 蚂蚁未造访诱饵是否是因为该生境中本就不存在该类蚂蚁，或(ii) 诱饵未吸引蚂蚁是否源于方法学缺陷（例如巴氏罐捕获量高但诱饵管觅食活动低，可能提示诱饵实验存在问题）。因此，我们在诱饵实验结束后再启动巴氏罐采样，以避免巴氏罐导致的蚂蚁种群耗竭（Lasmar等，2017）。每条巴氏罐样带与对应的诱饵样带平行布设，间距为20m，整个研究共设置60条巴氏罐样带。单条巴氏罐样带总长200m，设有5个采样点，相邻采样点间距50m。在每个采样点，我们使用树栖、地表与地下三类巴氏罐分别采集对应生境层的蚂蚁：树栖巴氏罐安装于树干中部，高度距地面1.5m；地表巴氏罐安装于地面；地下巴氏罐埋于地表下20cm处。巴氏罐直径8cm，深度12cm，内部盛放200mL由水、0.4%盐与0.6%液体肥皂配制的混合溶液。树栖与地表巴氏罐配有盖子，以防雨水与阳光侵入；地下巴氏罐同样配有盖子，并在侧面开有4个直径约1.5cm的侧向出入口。所有巴氏罐均在野外放置48小时后回收。