Highly sensitive positron imaging reveals short-term food distribution patterns in ant groups

The idea that nutritional exchange has played an important role in the evolution of functional integration within groups of social insects such as ants, bees, and termites has a long history¹ and has recently received renewed attention^2,3. As an initial step toward understanding the dynamics of nutrient flow within groups, visualization experiments using radiolabeled food were actively conducted during the 1950s–1970s (reviewed in^4,5. Due to the technical limitations of that time, these studies could obtain only “snapshot” data, monitoring the radioactivity of groupmembers at fixed intervals after feeding. In ants, recent research has focused on trophallactic (mouth to mouth) nutrient exchange, and several approaches have been developed to monitor nutrient transfer among individuals in real time. Some studies have usedfluorescent liquid food to track food flow from when ants begin food collection until they become satiated^{6,7,8,9,10,11}. However, although some degree of quantification can be obtained, it is difficult to accurately quantify the food held by ants. Buffin et al. developed a method for visualizing food exchange using a radioisotope (^99mTc) and gamma camera¹². However, this method cannot visualize and quantify food exchange in individuals because gamma cameras have limited resolution.

Positron imaging also uses radioisotopes, and the most widely used technique is positron emission tomography (PET) in cancer screening. Positron imaging systems use simultaneous measurement of 511 keV annihilation gamma rays with a resolution of approximately 2 mm, which allows non-invasive and high-sensitive visualization of food exchange. However, most positron-emitting radionuclides used in medical research have short half-lives (e.g., ¹⁸F: 110 min, ¹¹C: 20 min) and must be produced using a cyclotron; consequently, there are few facilities where positron imaging is performed¹³.

We previously visualized the movement of elements in plants using a planar positron imaging system, PETIS (positron-emitting tracer imaging system)^14,15. We visualized sodium transport using ²²Na (T_1/2 = 2.6 years), a commercially available long half-life positron-emitting radionuclide¹⁶. Then we thought that this ²²Na positron imaging technique could be applied not only to the study of plants, but also to the study of ant groups. Here, we developed an experimental system to visualize real-time quantitative changes in ant group food distribution using ²²Na-labeled food (sucrose) and PETIS.

Temporal changes in food distribution in a large ant group

After feeding a single worker ant approximately 500 kBq of ²²Na-labeled food (30% sucrose solution) (Fig. 1A), this ant (the initially fed ant) was placed into a container with approximately 100 worker ants, which was then placed in the middle of the PETIS detector to visualize the exchange of ²²Na-labeled bait between individuals (Fig. 1B). We successfully visualized time course change of food distributions within an ant group in three different biological replication experiments (Exp1, Exp2, and Exp3). Throughout the experiments, the experimental groups consisted of worker ants only, without queens or brood (which differs from “stock colonies” that contained queens and broods, too). Top panels of Movie S1 show the animation of food distributions in Exp1, Exp2, and Exp3. Figure 1C shows the snapshot images from the top-left panel of Movie S1, which corresponds to the PETIS data of Exp1, captured at 0, 10, 40 and 160 min.

To quantify the disparity in the amount of food held by each ant, an index of dispersion was introduced. The index of dispersion for food was defined in Eq. 1 and calculated using the PETIS data shown in top panels of Movie S1. The index of dispersion increases when the disparity in the amount of food held by each ant is large. Conversely, as partitioning food progressed and the disparity in the amount of food among ants decreased, the index of dispersion also became smaller. In Exp1 and Exp2, partitioning food progressed and the index of dispersion converged to a low common value after approximately 20 min; however, in Exp3, the index of dispersion increased and decreased during the 3-h period and did not converge with the Exp1 and Exp2 values (Fig. 2). The bottom panels of Movie S1 show the animations of time-course changes of the index of dispersion in Exp1, 2 and 3.

To verify the consistency of PETIS measurements, the PETIS counts from all pixels within the group area were summed to calculate the total group radioactivity at each time point. The time course of total PETIS counts in Experiments 1–3 is shown in Supplementary Figure S1. The total signal remained stable throughout the 3-h imaging period. This stability indicates that the observed changes in the index of dispersion (Fig. 2) reflect redistribution of ²²Na-labeled food within the group, rather than measurement drift or fluctuations in total detected radioactivity.

²²Na autoradiography after imaging in PETIS and radioactivities of each individual quantified by a gamma counter showed that ²²Na was distributed throughout the group from the initially fed individuals in Exp1 and Exp2, whereas ²²Na was distributed among half of the individuals in the group in Exp3 (Fig. 3). These results were consistent with the index of dispersion results (Fig. 2).

We estimated the amount of ²²Na originally ingested by the initially fed ant using the quantitative gamma counter data shown in Fig. 3 (Bottom). The total ²²Na activity was calculated as the sum of radioactivity measured for all individual ants in each group, following autoradiographic image acquisition. Based on these measurements, the total redistributed activity derived from the initially fed ant was estimated to be 24 kBq in Exp1, 44 kBq in Exp2, and 32 kBq in Exp3. These values correspond to the overall amount of labeled food transferred within each group and were used to evaluate the quantitative relationship between the gamma counter-based activities and the PETIS signal intensities (summarized in Supplementary Table S1).

Partitioning food within a group is generally thought to increase over time. However, in Exp3, the food was restricted to a few ants (Fig. 2A). The normalized time-course change of the index of dispersion at t = 0 is also shown in Fig. 2B. Food distribution progressed for up to 15 min after foraging; however, between 15 and 20 min, the food was consolidated to a few individuals, and distribution and consolidation repeated several times thereafter. As there were no signs of food being dumped and few individuals retained ²²Na in Exp3 (Fig. 3), food might have been repeatedly exchanged among a few individuals, including the initially fed individual.

Visualization of food exchange within a small ant group

To visualize and quantify food exchange between individuals in a relatively small ant group, the experimental setup (Fig. 1A, B) included a single ant fed ²²Na-labeled food (30% sucrose solution) placed in a container with 11 ants. This container was placed between PETIS detectors to visualize ²²Na-labeled food exchange among individual ants, and static (Fig. 4) and dynamic (Movie S2) visualization results were obtained. We also visualized ant behavior with a webcam (Fig. 4A; Movie S2A), and then we inferred food exchange between individuals based on the distribution of ²²Na-labeled sucrose imaged by PETIS (Fig. 4B; Movie S2B).

To quantify the amount of food exchange, the XY coordinates of each individual ant were extracted for each second using video (Movie S2A), and its trajectory was tracked for 2000s. We tracked 12 ants discriminated by three-color labeling; however, data from 11 ants were used because tracking failed for one ant. The position information of each ant was fused with PETIS signal values and calculated the change over time of the sucrose amount possessed by each individual (Movie S2C, D). We successfully visualized that the ants were gathering and food exchange was taking place immediately after the start of the imaging (Fig. 4D; individual 2 was the initially fed ant). Figure 4 represents the snapshot image of Movie S2 at 600 s.

Movie S2A and Fig. 4(A) present a video and its snapshot in which multiple ants are simultaneously visible. In this video, visually tracking the temporal changes in individual positions is challenging. However, using the analytical method developed in this study, it became possible to assign unique identification numbers to each ant and track them over time, as shown in Panel C. Furthermore, by integrating the ²²Na data presented in Panel B, we were able to visualize the amount of food carried by each individual in conjunction with their spatial positions, as illustrated in Panel C. Additionally, as demonstrated in Panel D, we confirmed that it is also possible to display temporal changes in food quantity for each individual.

This experimental system successfully visualized and quantified food distribution changes within large ant groups. Buffin et al. quantified the increase in the amount of food entering a group⁸, whereas data obtained by the method we developed provide more detailed data on food distribution change within a group over time. Our data could be used to describe more detailed changes in food exchange under different environmental conditions and among different ant species.

The index of dispersion in this study is a useful value for quantifying temporal changes in food distribution within ant groups, but it has certain limitations. The index assumes that the total detected radioactivity within the group remains constant over time and that the entire group stays within the PETIS field of view. It is therefore most reliable under stable measurement conditions where the group remains confined within the region of interest. When individuals frequently overlap or when the group size exceeds the PETIS field of view, however, the index may underestimate disparities in food distribution. It should be noted that this image-based index of dispersion becomes less reliable in very small groups of ants or when individuals are tightly clustered, because the pixel-level variance may fluctuate substantially under such conditions. Furthermore, because it is based on the overall signal intensity of the group rather than the radioactivity of each individual ant, it cannot account for biological factors such as excretion, metabolic decay, or signal attenuation within dense clusters.

A notable advantage of methods utilizing fluorescent dye-labeled food to visualize trophallactic food dissemination^2,3,4,5,6,7 lies in their ability to accurately track, in real time, changes in the amount of food retained by individual workers. This is achieved through the combination of tagged individual tracking and image-based quantification.​ However, there are two primary limitations to this approach. First, its applicability is restricted to specific ant species. For instance, studies employing this method have utilized Camponotus sanctus, a species with pale or lightly pigmented gasters, which permits visualization of ingested food through the cuticle⁶. Consequently, this technique is not broadly applicable across a wide range of ant species, unlike radioisotope based experiments. Second, the method lacks the sensitivity required to detect minute quantities of food. The RI-based approach proposed in the present study not only facilitates approximate real-time tracking of changes in food retention by workers but also enables the detection of trace amounts of food through post-experimental quantification.

Both macro- and micro-scale observations were possible with our system, which represents a powerful research tool for building models of large- and small-scale food exchange within an ant group. In the present experiment, the time-course change of the index of dispersion for food was evaluated for 100 ants. Although typical ant groups can range from several thousand to many tens of thousands of individuals depending on the species, it is considered feasible to perform experiments to determine this index even with groups of such size. In contrast, the visualization of food exchange between individuals is currently challenging to apply to groups containing more than 12 ants, mainly due to the difficulty in extracting the XY coordinates of each individual from video recordings. This limitation is expected to be resolved through the application of more advanced coordinate extraction techniques.

Although this study aimed at technical development, let us interpret the obtained results biologically. In Exp1 and 2, the predicted pattern of food diffusion over time due to nutritional exchange was observed. In contrast, in Exp3, the individuals initially fed mostly retained the food throughout the experiment. This difference is likely because Exp3 resampled the same stock groupA soon after extracting 100 foragers for Exp1. There may have been very few foragers left in Exp3. Nutrient exchange is known to occur unevenly among honeybee workers, in which age polyethism, the pattern shared hymenopteran eusocial insects, i.e. young workers perform nurse and old workers forage characteristically, is considered a related factor^17,18,19. Also in Anoplolepis gracilipes it is reported that foragers are more likely to serve as donors in trophallaxis²⁰. It is possible that Exp3 sampled only nurse individuals, which typically do not serve as donors in nutrient exchange. Accordingly, because Exp3 was conducted using a stock colony with an insufficient number of foragers, the findings obtained from the small-group experiment are considered to be of limited usefulness for interpreting the results of Exp3.

Studied species and groups

The yellow crazy ant Anoplolepis gracilipes forms large groups with multiple nests and multiple queens, containing hundreds of thousands of workers. This ant is omnivorous, preying on diverse invertebrates in the wild, either alone or in groups, and may scavenge the carcasses of small vertebrates. It often tends honeydew-producing hemipterans. Mouth-to-mouth trophallaxis is frequently observed among workers²⁰.

Three A. gracilipes groups (A, B, C) were collected at three different locations (26.21°N,127.77°E; 26.25°N,127.76°E; 26.76°N,128.24°E) on the main island of Okinawa in March 2020. Each group was kept in a large plastic container (55 cm × 26 mm × 19 mm) with nest materials (sand and wood debilis) in a laboratory (25 °C, 14L:10D) for 1–5 months before the experiments. Groups were fed mealworms and 10–20% honey water twice a week ad libitum. A few days before the imaging experiment, ant groups were transported to the research facility of Takasaki Institute for Advanced Quantum Science in Gunma Prefecture. **1.

Imaging experiment

Figure 1A and B show schematics of the imaging experiment. At the start of the experiments, all three stock groups A, B, and C contained multiple queens and numerous brood (eggs, larvae, and cocoons). Groups A and B each contained approximately 500 workers. Group C contained approximately 1000 workers. The experiments were concentrated into two periods (each lasting 2–4 days). First, Experiments 1–3 (Exp.1–3, hereafter) were conducted. Exp.1 sampled workers from stock group A, while Exp.2 sampled workers from stock group B. Exp.3 was conducted using 100 individuals resampled from stock group A on the day following Exp.1. An additional experiment using small groups of 12 individuals were conducted ca. three months after Exp.1–3, using the stock group C.

The worker sampling method is as follows. In ants, division of labor generally exists among workers. In Anoplolepis gracilipes, it is known that donors of trophallaxis are more likely to be individuals responsible for foraging²⁰. Therefore, to select more foragers, we carefully picked up 100 individuals active outside the nest (rotting wood placed in a container) in the stock groupusing a brush. Since fewer than 100 individuals were active outside the nest at the start of sampling, we took some time to pick up all individuals emerging from the nest until we reached 100. We performed similar sampling for the 12-individual group experiment, but sufficient individuals were present outside the nest, so sampling did not take long.

Those 100 or 12 individuals were transferred to a tetrafluoroethylene-coated plastic imaging container (160 mm × 105 mm × 55 mm) to prevent the ants from escaping, and starved overnight. Just before the imaging experiment, 1 μL of ²²Na solution (radioactivity: approximately 500 kBq) was added to 19 μL of 30% sucrose solution to make a radio-labeled food. Parafilm (10 mm × 10 mm × 127 μm) was placed inside a feeding container (70 mm × 70 mm × 100 mm), 20 μL of radio-labeled food was added, and one ant from a starved group was selected from 100 or 12 individuals, marked with enamel paint on the dorsal abdomen and placed in the feeding container. A digital camera was used to confirm that this single ant had sucked on the radio-labeled food for at least 2 min, after which it was put into the imaging container where a group of 100 (Fig. 1C) or 12 (Fig. 4) ants were present. The imaging container was then placed between the PETIS detectors, and images of the ²²Na signal distribution were continuously measured for 3 h using PETIS. Simultaneously, the movement of each individual ant was imaged with a digital camera placed on the surface of the PETIS detector. In the experiment in Fig. 2, the dorsal abdomen of only ants that initially fed the labeled food was marked with enamel paint, and in the experiment in Fig. 4, the dorsal abdomen of all ants was marked with enamel paint in a combination of three colors.

After PETIS imaging was completed, ethyl acetate was added to the container with the groupand the individual ants were sacrificed. Individual ants were sampled and attached to paper mats using storage phosphor screens (BAS-IP-MS-2040E, GE Healthcare, Amersham, UK) and the enclosed samples were kept in contact for 21 h using exposure cassettes (GE Healthcare). Autoradiographic image data were acquired using a laser imaging scanner (Typhoon FLA-7000, GE Healthcare). Then, each individual ant was placed in a polyethylene container and measured with a well-type gamma counter (ARC-7001; Aloka Co., Ltd., Tokyo, Japan) for 10 min, and the ²²Na activity of each ant was quantified. The data shown in Fig. 3 is more than the background of this measurement.

Data analysis

Imaging data acquired with PETIS, representing changes in the distribution of ²²Na-labeled sucrose over time, were converted to TIFF images using ImageJ (National Institutes of Health, Bethesda, MD, USA; http://rsb.info.nih. gov/ij/) at 1 frame/s. The spatial resolution of the PETIS system is 1.1 mm per pixel. The TIFF images were subjected to the following two analyses using the Python SciPy library (http://www.scipy.org/).

Real-time visualization of temporal changes in food distribution in a large ant group

For Exp1, Exp2, and Exp3, which were real-time visualization experiments of the distribution of food in a large ant group over time, we quantified the time constant of the distribution over the entire group. Specifically, after converting the PETIS imaging data showing the distribution of ²²Na-labeled sucrose over time to 1 frame/s, the Python SciPy library was used to calculate the variance-to-mean ratio ({s}^{2}/overline{x }) for each frame as the index of dispersion²¹, defined by the following equation to determine the change over time:

where ({s}^{2}), (overline{x }), (N), and ({x}_{i}) are the variance, the mean value, the number of pixels in a flame, and the PETIS counts at the (i) ^th pixel of the frame, respectively. In this equation, we employed the approximations that the mean value (overline{x }) was almost zero. This was a quantification of the disparity in the amount of food held by ants, and the variance-to-mean ratio increased when the disparity in the amount of food held was large. Conversely, the variance-to-mean ratio decreased as partitioning food became more advanced and the disparity in the amount of food held by ants became smaller. Because plotting the variance-to-mean ratio for each frame makes it difficult to read the trend of change over time, the average of the data up to 500 frames ahead is plotted for each frame in Fig. 2.

Visualization of food exchange between individuals in a small ant group

Stereoscopic video images of 12 ants taken simultaneously with PETIS were analyzed using UMATracker²². The position of each ant, marked with a three-color ink pattern, was calculated over a 30-min period. When two or more ants were in close proximity or overlapped, tracking errors occasionally occurred in the automated UMATracker output. In such cases, the trajectories were manually reviewed and corrected by visual inspection of the original video recordings to ensure accurate individual identification.

The temporal variation data of each ant’s position obtained from the above analysis and the TIFF image data of PETIS for the corresponding experiment were used to determine the temporal variation of radioactivity for each individual ant. The radioactivity of each ant was defined as the sum of the pixels contained inside a circle with a radius of 5.5 mm (equivalent to 5 PETIS pixels) centered on the position of each ant. When one pixel was included in several circles, we assumed that the ant closest to the pixel possessed all the radioactivity of the pixel. This division method is known as Voronoi partitioning.

To integrate the PETIS data with the position data obtained from UMATracker, we first corrected a spatial mismatch between the two coordinate systems. The raw UMATracker coordinates extracted from the webcam video were offset and scaled to match the PETIS coordinate framework, and these transformation parameters were optimized so that the two datasets were spatially aligned. Temporal synchronization between the PETIS images and the UMATracker output required an additional corrective step. Because the frame rates of the two systems were not intrinsically synchronized, we applied a periodic modulation to the playback speed of the UMATracker data. The parameters necessary to achieve temporal and spatial consistency were initially determined manually by the researchers and subsequently refined using a least-squares optimization so that the estimated amount of radioactive tracer assigned to each ant was maximized.

While the corrected UMATracker output showed good agreement with the PETIS images (Fig. 4; Movie S2A), the webcam video itself was not temporally adjusted. UMATracker generates text-based coordinate data, allowing straightforward manipulation of playback speed, whereas the webcam recordings are stored as video files, for which precise temporal editing is technically challenging. For this reason, the raw webcam footage is presented without synchronization adjustments.