Evolution of acoustic communication in fungus-growing ant societies

Do ants ‘talk’? If so, how important is talking in ant societies? Chemical communication, through pheromones for example, was thought to be the main communication tool in ant societies; however, a major question is whether chemical substances alone can control highly differentiated social behaviours. In this study, we focused on the ‘talk’ of agricultural ants, key organisms in the Neotropical ecosystem, and conducted a detailed acoustic analysis. Our results indicate that in fungus-growing ants, acoustic communication is a much more frequent and critical factor than previously believed. The frequency of stridulatory sound-production recordings from the ants were found to be signicantly correlated with social structure complexity. Structural analysis indicated that both the area and number of slits in the stridulatory acoustic organs were signicantly correlated with body size. The ability of leaf-cutting ants to maintain fungus gardens was signicantly lower in the sound-inhibited group than in the pheromone secretion-inhibited group. These results suggest that ants that have become ‘chattier’ may induce altruistic behaviours and create more complex societies. The ndings of this study may be applicable not only to social evolution studies but also for effectively controlling ant behaviours.

cultivate numerous mutual fungi, these ants play a critical role in the ecological succession of vegetation in Neotropical forests 7,8 . When their vigorous agricultural activities are directed towards human society, they cause enormous economic and agricultural damage. For instance, Atta texana in Texas can destroy >5,000 ha·yr -1 of pine seedlings, with annual control and management costs reaching ~US$ 2.3 million 9 .
To maintain such a complex social structure and elaborate agricultural system, leaf-cutters must perform >30 time-sensitive tasks with the most appropriate members 6,7,8 , and our preliminary research in which we recorded high-frequency acoustic signals in Atta led us to hypothesize that 'talking' (i.e. acoustic communication) is an important factor used to control such social behaviours.

Social evolution is based on communication
Eusocial insects, including fungus-growing ants, are conventionally referred to as 'superorganisms' 10 , and the study of communication within these species is expected to provide a major breakthrough in the theory of social evolution 11 .
Highly evolved social behaviours of insects are controlled by constant sharing of information through chemical 12,13 , acoustic 14,15 , visual 16,17 , and tactile 18 stimuli (Table 1). In ants, chemical communication is generally understood to be a foundational feature of their social evolution and cooperative behaviours 12,13 . Although visual, acoustic, and tactile communications are postulated as the next evolutionary steps, those have yet to be experimentally revealed 11 . Barbero et al. (2009) published a revolutionary study on acoustic communication in ants 19 . It had been established that larvae of the lycaenid butter y (genus Maculinea) used chemical mimicry to invade host ant (Myrmica) nests 20 ; additionally, the larvae of Maculinea (Phengaris) rebeli were found to mimic queen ant sounds of Myrmica schencki and received protection from the workers once inside the nest. The study revealed that queens sounds induce cooperative behaviours, bringing attention to the importance of acoustic communication among ants. Further examples of acoustic communication research are well known in leaf-cutting (genus Atta) 21 and re ants (Solenopsis invicta) 22 . Others are also known to produce drumming sounds by striking their abdomens against the ground 23,24 ; however, no quantitative studies have been conducted on these functions.
To establish the importance of acoustic communication in the evolution of sociality in fungus-growing ants, we proposed the following two hypotheses: (1) A signi cant correlation exists between social evolutionary levels and sound-production frequency and types; and (2) acoustic as well as chemical communication in agriculture-practicing ants require a high level of cooperative behaviour. To test these hypotheses, we obtained detailed stridulatory sound recordings with each independent behaviour of seven genera (eight species) of fungus-growing ants using a high-resolution recording device that we developed. Phonetic analyses of the recorded sounds and structural analysis of the sound-producing organs were carried out via SEM, and the effects of chemical and acoustic stimuli on fungus-garden maintenance were examined through their experimental inhibition.
'Chattier' ants formed complex societies Using a self-developed, high-resolution recording device, we recorded the acoustic signals of 3-10 workers and fragments of symbiotic fungi from their fungus garden for 15-30 min in attines and successfully obtained ant-derived stridulatory sounds for all species (Extended Data 1-3). The groups that retained ancestral traits produced acoustic signals <0.5 times per minute per individual (0.01 ± 0.002 in Myrmicocrypta, 0.07 ± 0.06 in Apterostigma auriculatum, 0.03 ± 0.03 in Ap. mayri, 0.26 ± 0.30 in Cyphomyrmex). In the moderately derived group, the average frequency was 6.37 ± 2.32 and 2.15 ± 2.24 times per minute per individual in Trachymyrmex and Sericomyrmex, respectively. The genera of the highly advanced group, Acromyrmex and Atta, were much 'chattier' (average frequency 19.15 ± 6.35 in Acromyrmex and 24.08 ± 24.21 in Atta) than those of the other groups. A Jonckheere-Terpstra test con rmed a signi cant positive correlation between the stridulatory sound frequency and estimated branch age of each lineage (Z-value, 2.97; P < 0.01).
We next recorded and phonetically analysed acoustic signals under independent stimuli and external conditions for the 'chattiest' leaf-cutting ant species, Atta colombica. The two independent external stimuli were (1) pinching with forceps and (2) burying with oats, whereas the recorded signals from independent conditions and locations were (3) cutting Leguminosae leaves, (4) cutting Clusiaceae leaves, Accordingly, acoustic communication in fungus-growing ant societies was found to be more important than previously assumed. The evolutionary nature of ants was further revealed by two ndings: the frequency of acoustic signal production increased with social evolutionary stage, and the complex society of leaf-cutting ants produced the highest number of eusocial insect sound types recorded thus far. Accordingly, acoustic forms of communication, much like chemical communication, are an important factor in the evolution and maintenance of social system in the ants, as 'chattier' communities appear capable of more complex societies than 'silent', small, and simple societies.
The honeybee is a eusocial insect that is capable of complex acoustic communication despite the absence of a speci c sound producing organ. They can produce vibroacoustic signals by shaking their thoraxes and wing vibrations 31 , and when workers combine gure-eight dances with acoustic signals, they can share information about both the distance and direction of nectar sources 31 . The tooting and quacking sounds of queen bees inhibit the hatching of new queens and inform nestmates of the presence and activity of enclosed queens 32 . In worker bees, a piping sound terminates the dance and helps recruit nectar receivers 33 . Alternatively, the leaf-cutting ants here demonstrated 10 signi cantly different acoustic signals, indicating a potentially more complex communication system.
A subset of ant species are the only eusocial insects with stridulatory organs to produce sound. Only groups with two petiole segments have these organs, comprising just six of the 20 subfamilies, including Myrmecinae, Paraponerinae, and Ponerinae 34 . Although there is some ant sound communication research in several ant species, no quantitative studies have been conducted on these functions. Accordingly, the high-resolution recording device we developed could be useful in facilitating discovery of more complex acoustic communication systems within eusocial insect societies in the future.  Fig. 2b). More speci cally, the area of the stridulatory organ was not correlated with head width in the ancestral and middle-advanced groups but highly correlated in the two genera of leaf-cutting ants. The slopes (α) were 12.66 for Atta colombica (S-value = 80.71; P < 0.001) and 1.01 for Acromyrmex octospinosus (S-value = 514; P < 0.001), whereas the intercepts (β) were -15.73 and 1.94, respectively. Similarly, no signi cant correlation was found between the allometry of observed slit number and the ancestral or middle-advanced groups, although a signi cant positive correlation with Atta (α = 3.18, β = -2.19; S-value = 138.12; P < 0.01) and Acromyrmex was observed (α = 1.09, β = 1.51; S-value = 503.58; P < 0.001). The data thus suggest that even among two advanced genera, Acromyrmex followed a relative growth curve, whereas the more advanced genus Atta had slopes much greater than 1, indicating a highly variable structure.
These observations prompted an assumption regarding the physical evolution of acoustic communication in ants. The allometry indicated that the area and slit numbers of the stridulatory organ increased with body size. Body size, in turn, was regulated by habitat, queen egg-laying ability, and colony size 6 , suggesting that such ecological and social changes had driven the evolution of communication ability in ants.
In Hymenoptera, allometric analyses have shown that the slope of their reproductive organs is signi cantly >1, whereas those for the brain and central nervous system are <1 35 . Reproductive organ size is correlated with egg size, and egg size depends on the presence of parasitic organisms; thus, changes in the allometric slope can indicate an adaptive pathway for external factors. This suggests that the stridulatory organs of Atta were subjected to positive selection pressure, signi cantly increasing their function as a critical means of communication. In turn, it may affect the ecological factors and account for the 100-fold greater colony size and more complex social structure of Atta than of Acromyrmex.

Chemical vs. acoustic cues
To compare the e ciency of social behaviours and management among leaf-cutting ants, sound and pheromone inhibition experiments were conducted. Ten sub-colonies, each consisting of eight workers, were placed for 1 week near a small fungus garden, along with glued pheromone-producing organs, sound-producing organs, or upper part of the mesothorax. Daily garden weights and social behaviours were recorded. The mortality rates in each experimental treatment were negligible and did not signi cantly affect the experimental results. It was found that the fungus garden was signi cantly smaller in the sound-inhibition group (48.4% ± 9.2% of the initial garden size) than in the pheromoneinhibition group (71.6% ± 10.9% of the initial garden size; Extended Data 6); the control garden was 103% ± 9.19% of the initial garden size. Signi cant differences among the groups were observed using a multiple comparison test (Steel-Dwass method: control vs. sound, t = 3.81; P < 0.001; sound vs. pheromone, t = 3.66; P < 0.001; control vs. pheromone, t = 3.70; P < 0.001).

Acoustic recording device and recording conditions
For the stridulatory acoustic recordings of fungus-growing ants, it was essential to have a small recording device with high resolution and could be safely carried to the research sites. However, the commercially available devices were too large and expensive. Therefore, we developed a small, high-resolution recording device that we designed. Hoshiden's KUC3523-04-0245 lead condenser microphone (Hoshiden, Osaka, Japan) was used to detect stridulatory cues, and it was connected via a 3.5 mm jack. The sound was ampli ed by a portable multimixer AT-PMX5P (Audio Technica, Tokyo, Japan) and imported into a computer. The recording device was covered with three layers of acrylic casing (the size of the outermost acrylic case is 15 cm x 15 cm x 15 cm), with a cork mat at the bottom, and a rubber sheet underneath to absorb vibrations and eliminate noise. In total, 945 min of acoustic data were recorded for seven genera (eight species) of attines (Extended Data 2). For the basal group, 10 individuals were recorded at a time, whereas for all the other groups, three individuals were placed in the recording device for 10-30 min each in three conditions: with the fungus garden, larvae, and pupae; with the garden substrates (Leguminosae leaves, Clusiaceae leaves, and Heliconiaceae owers); and with the garbage dump.

Structural analysis of the acoustic organs
Images of the acoustic organs for 69 individuals from seven genera (seven species) of attines were obtained using a scanning electron microscope (JEOL JSM-6360LA). Image J (v.1.52) was used to measure the area and count the number of slits.

Sound and chemical inhibition experiments
Two plastic cases were prepared with 0.20 g of fungus garden fragments, and eight small-to-mediumsized leaf-cutting ants were introduced into one of the cases, whereas the other was lled with oats (0.50 g), a fungus garden substrate. Ten sub-colonies were prepared for each treatment, and three manipulation conditions were used: (1) xation of the postpetiole with glue to prevent it from producing sound (soundlessness of all individuals was con rmed using a microphone), (2) xation of the end of the abdomen with glue to prevent pheromone secretion (individuals were made to walk on lter paper for con rmation), and (3) the pronotum was glued as a control. Sub-colonies were kept at 25 °C and 70% humidity, and newly treated individuals were introduced daily to replace any mortalities (control, 0.1 ± 0.08 mortalities·day -1 ; pheromone inhibition, 0.51 ± 0.3; sound-inhibition group, 0.30 ± 0.32). Worker behaviours were observed for 5-min steps at regular intervals daily, and the weights of the fungus garden and oats were measured after 1 week.

Statistical analyses
A Cochran-Armitage analysis was used to analyse the correlation between the stridulatory sound frequency and evolutionary stage. Canonical discriminant analyses were used for statistical processing of the stridulatory acoustic types in each species, and Pearson's product-rate correlation tests were used for the allometric analyses of the stridulatory organ structure. Analyses of the manipulation experiments were performed using the multiple comparison of means method (Ryan method), and a generalized linear model analysis was used to assess the types of behaviour for each treatment. All analyses were performed in R (v.3.6.1). The raw data and the codes of statistical processes were saved in Kyushu University and is available if required to con rm reproducibility. Table 1 Due to technical limitations, table 1 is only available as a download in the supplementary les section. Figure 1 Discriminant analysis of 11 signi cantly different sound types from leaf-cutting ant (Atta colombica) mid-sized workers and queens. Discrimination rates ranged from 60% to 100%. Reanalysis of the seven types of acoustic signals that were unclear in (a) resulted in discrimination rates of 80-100% (b).

Figure 2
Allometric relationships between (a) stridulatory organ size and (b) slit number according to head width.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.