This study presents an extended mathematical model, originally developed to explore biomass dynamics in Acromyrmex ants and their symbiotic fungus, which has been adapted to investigate biomass growth in Atta sexdens and its fungal partner. The model incorporates ant grooming as a defence against entomopathogenic fungi, building on experimental data where self-grooming and allogrooming were quantified across three groups: Metarhizium anisopliae, Escovopsis phialicopiosa, and a control. A second chamber was introduced to simulate biomass transfer between compartments. Inflection points in growth curves were identified to detect shifts in population dynamics, and bifurcation diagrams explored key parameters affecting system stability, namely worker allocation to fungal cultivation, ant mortality, and fungal mortality. Metarhizium anisopliae significantly reduced both ant and fungal biomass, even under optimal grooming conditions, owing to its direct virulence to workers. In contrast, E. phialicopiosa, an opportunistic pathogen, had minimal impact unless fungal mortality exceeded a critical threshold. Self-grooming proved more effective than allogrooming in mitigating M. anisopliae effects, likely owing to prioritisation of individual defences under high pathogen pressure. Spatial dynamics enhanced resilience: one-way transfer between chambers redistributed biomass, delaying inflection points and bolstering structural stability. Bifurcation analysis revealed that extreme proportions of workers cultivating the symbiotic fungus reduced the biomass for both partners, whilst ant and fungal mortality rates led to non-linear declines in most simulations. These findings underscore the role of multi-chamber architecture in mitigating pathogen impacts in Atta colonies and suggest potential applications for biological control strategies by identifying behavioural and structural factors that may limit or enhance the effectiveness of pathogenic fungi in field settings. The model provides a useful framework for understanding epidemiological dynamics in natural nests, integrating behavioural defences and spatial strategies to safeguard this mutualism.

Allogrooming, ant-fungus mutualism, interactions, leaf-cutting ants, self-grooming

The symbiosis between Attina ants and Basidiomycota fungi constitutes an obligate mutualism: ants provide substrates and protect the fungi from pathogens and parasites, while the fungi serve as the sole food source for larvae and a significant component of the workers’ diet (Weber 1966; Mehdiabadi and Schultz 2010; Hölldobler and Wilson 2011). This specialised relationship, approximately 66 million years old (Schultz et al. 2024), involves a complex network of microorganisms, including yeasts, viruses, and bacteria, whose roles remain largely unexplored (Currie et al. 1999). Dense, genetically uniform colonies render ants vulnerable to pathogenic microbes (Schmid-Hempel 1998; Rissanen et al. 2022). To counter this, they have evolved a multi-layered defence system, integrating physiological responses (Fernández-Marín et al. 2006) with behavioural mechanisms such as self-grooming, allogrooming, and cleaning of brood and fungal gardens (Currie et al. 1999; Goes et al. 2022).

Leaf-cutting ants of the genus Atta are among the most important agricultural pests in Latin America. Members of this genus occur from the southern United States to Argentina (Mariconi 1970; Delabie et al. 2011; Forti et al. 2020). Their ecological significance is reflected in their status as dominant herbivores in Neotropical biomes (Bucher 1982; Costa et al. 2008), with the potential to consume up to 17% of the available foliar biomass (Costa et al. 2008). From an economic perspective, Atta species cause substantial damage to forest plantations. In Eucalyptus spp., sustained defoliation reduces productivity (Zanetti et al. 2003), while in Pinus spp., it negatively affects tree growth (Scherf et al. 2022). In agricultural systems, severe infestations can lead to considerable yield losses in crops such as cassava (Manihot esculenta) and coffee (Coffea arabica) (Bertorelli et al. 2006; Varón et al. 2007).

In pastures, intensive removal of plant biomass limits forage availability (Fowler 1986). The combination of extensive distribution, highly efficient foraging, and broad host adaptability consolidates Atta species as a major challenge for agricultural and forestry management in the Neotropical region. These impacts underscore the economic importance of understanding the ecological dynamics of Atta colonies and the need for effective management strategies. In addition to physiological and behavioural defences, fungus-farming ants employ spatial strategies to minimise pathogen transmission.

Age- and caste-based spatial segregation exemplifies this approach (Wilson 1980), whilst controlled movement between functional groups and waste segregation further reduce disease spread (Farji-Brener and Tadey 2012), enhancing organisational immunity. In the genus Atta, an intricate network of underground chambers (Moreira et al. 2010) plays a pivotal role in collective immunity, bolstering colony resilience. The chamber expansion follows a unidirectional pattern during colony development. Studies show that, at the time of expansion, Atta workers relocate the brood and fungal biomass to locations where new chambers will be excavated and established (Römer and Roces 2014). This directional flow of resources reflects the colony’s developmental trajectory, in which older chambers act as sources of biomass for new compartments. This compartmentalised structure, dedicated to fungal cultivation, brood rearing, and waste disposal, limits pathogen dissemination (Hart and Ratnieks 2002), rendering these ants particularly resistant to biological control strategies.

A key challenge in biological control is the structural complexity of nests, since interconnected chambers hinder the uniform application of biocontrol agents and the empirical study of disease dynamics. Typically, control strategies involving entomopathogenic fungi, such as Metarhizium anisopliae, are successful when the fungi infect insects through spores that adhere to the insect’s surface, germinating and penetrating the cuticle to initiate an infectious process. Given the complexity of these social immunity mechanisms, analysing these systems poses a challenge, highlighting key considerations for using entomopathogenic fungi in biocontrol.

Mathematical models of organism interactions are vital in ecology for describing and predicting population dynamics within complex systems under diverse scenarios (Lima et al. 2009). Mutualistic or symbiotic relationships, such as those between ants and fungi in compartmentalised systems, are widely studied (Dejean et al. 2023). The relationship between fungus-cultivating ants and their fungi offers a unique opportunity to explore population dynamics and the factors influencing stability and growth under varying environmental conditions.

This study employed a mathematical model to investigate the dynamics between leaf-cutting ants and their symbiotic fungus within a two-chamber system, where one-way transfer between compartments is density-dependent and follows a logistic function. This configuration enables the examination of critical factors, including carrying capacity, the influence of external stressors (such as pathogens), and ant behavioural responses to population growth. Furthermore, it facilitates modelling the effects of self-care and social hygiene, specifically self-grooming and allogrooming, in response to pathogenic fungal spread, enhancing our understanding of how these behaviours mitigate disease impacts. The model extends the framework proposed by Kang et al. (2011), originally developed to describe biomass growth dynamics in Acromyrmex ants and their symbiotic fungus.

The present study addressed fundamental ecological questions regarding how hygienic behaviours, such as self- and allogrooming, contribute to the resilience of leaf-cutting ant colonies and their fungal symbionts when faced with pathogenic fungal infections. Specifically, we investigated how the addition of an extra nest chamber, a structural feature typical of the genus Atta, could mitigate the impact of pathogens, what critical mortality thresholds compromised mutualistic stability, and how worker allocation and behavioural strategies shaped the population dynamics of both the ants and their fungal gardens.

These questions are central to the development of effective biological control strategies, given that the structural complexity of subterranean nests and the behavioural defences of these ants present significant challenges to the application of pathogens under field conditions. By integrating experimental data with mathematical modelling, this study aimed not only to advance our ecological understanding of the ant-fungus mutualism but also to identify key parameters (e.g., optimal proportion of workers allocated to fungal cultivation, one-way transfer rates between chambers) that can inform more targeted interventions, minimising environmental impacts while optimising population control in regions where this genus is considered an agricultural pest.

Ant responses to the pathogens varied depending on the grooming type (Table 2). Table 2 presents laboratory estimates of the time spent on self-grooming and allogrooming for the control, M. anisopliae, and E. phialicopiosa treatments. The largest difference among treatments (T1, T2, T3) was between the maximum and mean allogrooming values (Table 2).

Fig. 1A depicts the ant and fungal populations according to the model of Kang et al. (2011), showing sigmoid growth. Fig. 1B includes a one-way transfer from chamber A to B, with biomasses decreasing in A and increasing in B. Fig. 2 illustrates maximum self-grooming, with inflection points (solid blue circles for chamber A and solid red circles for chamber B) in biomass curves for T1, T2, and T3, reflecting critical growth shifts. In T2 (M. anisopliae), the ant and fungal biomasses decreased compared to T1 (control), with inflection points delayed. T3 (E. phialicopiosa) closely resembled T1, with minor shifts in inflection values (Fig. 2). Figs 3, 4 show mean and minimum self-grooming, respectively, and curves exhibiting analogous trends with slight differences in their inflection points.

Figure 1.

Biomass of leaf-cutting ants (a) and symbiotic fungus (b) modelled using Kang et al. (2011) A without one-way transfer between chambers, and B with one-way transfer D = 0.1 between two chambers, with identical initial biomass conditions for the ants (B_a = 0.05) and symbiotic fungus (B_a = 0.3), and independent population dynamics within each chamber prior to one-way transfer.

Figure 2.

Maximum self-grooming effects on biomasses of leaf-cutting ants and symbiotic fungus with inflection points. Biomass growth curves with identified inflection points (solid blue circles) for chamber A and (solid red circles) for chamber B across three treatments: T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa). D represents the one-way transfer of biomass from chamber A to chamber B (D = 0.1). The simulation uses maximum self-grooming durations measured in seconds: T1 (168 s), T2 (114 s), and T3 (185 s).

Figure 3.

Mean self-grooming effects on biomasses of leaf-cutting ants and symbiotic fungus with inflection points. Biomass growth curves with identified inflection points (solid blue circles) for chamber A and (solid red circles) for chamber B across three treatments: T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa). D represents the one-way transfer of biomass from chamber A to chamber B (D = 0.1). The model uses mean self-grooming durations measured in seconds: T1 (14.4 s), T2 (12 s), and T3 (14 s).

Metarhizium anisopliae (T2), even exposed to maximum allogrooming, significantly reduced the biomasses for both ants and the symbiotic fungus compared to T1 (Fig. 5), with inflection points shifting to higher values, resulting in a time-delayed effect, except in Chamber A. T3 mirrored T1, with slight differences in inflection point values. The mean and minimum times spent on allogrooming produced similar growth curves across the treatments, with minor variations in their inflection values (Figs 6, 7). Excluding grooming protection (Fig. 8) led to a decrease in the biomasses of both the ants and the fungus.

Figure 4.

Minimum self-grooming effects on biomasses of leaf-cutting ants and symbiotic fungus with inflection points. Biomass growth curves with identified inflection points (solid blue circles) for chamber A and (solid red circles) for chamber B across three treatments: T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa). D represents the one-way transfer of biomass from chamber A to chamber B (D = 0.1). The model uses minimum self-grooming durations measured in seconds: T1 (0.3 s), T2 (0.53 s), and T3 (0.73 s).

Figure 5.

Maximum allogrooming effects on biomasses of leaf-cutting ants and symbiotic fungus with inflection points. Biomass growth curves with identified inflection points (solid blue circles) for chamber A and (solid red circles) for chamber B across three treatments: T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa). D represents the one-way transfer of biomass from chamber A to chamber B (D = 0.1). The model uses maximum allogrooming durations measured in seconds: T1 (1701 s), T2 (169 s), and T3 (1639 s).

Figure 6.

Mean allogrooming effects on biomasses of leaf-cutting ants and symbiotic fungus with inflection points. Biomass growth curves with identified inflection points (solid blue circles) for chamber A and (solid red circles) for chamber B across three treatments: T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa). D represents the one-way transfer of biomass from chamber A to chamber B (D = 0.1). The model uses mean allogrooming durations measured in seconds: T1 (330 s), T2 (54 s), and T3 (82 s).

Figure 7.

Minimum allogrooming effects on biomasses of leaf-cutting ants and symbiotic fungus with inflection points. Biomass growth curves with identified inflection points (solid blue circles) for chamber A and (solid red circles) for chamber B across three treatments: T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa). D represents the one-way transfer of biomass from chamber A to chamber B (D = 0.1). The model uses minimum allogrooming durations measured in seconds: T1 (1.5 s), T2 (3.87 s), and T3 (1.23 s).

Figure 8.

Effects of the absence of self-grooming and allogrooming on biomasses of leaf-cutting ants and symbiotic fungus with inflection points. Biomass growth curves with identified inflection points (solid blue circles) for chamber A and (solid red circles) for chamber B under the condition where no grooming behaviour occurs. D represents the one-way transfer of biomass from chamber A to chamber B (D = 0.1).

The bifurcation diagram for worker allocation (q) compares minimum, mean, and maximum self-grooming times (Fig. 9). The general shape of all curves trends downwards, indicating that the proportion of workers cultivating the symbiotic fungus (q) approaches 0 when low biomass is produced, due to a low number of workers growing the fungus, and approaches 1 when a high number of workers excessively prioritise fungal growth, neglecting other tasks essential to the colony. A comparison of control (T1) and pathogenic fungi treatments (T2, T3) suggested that ant and symbiotic fungal biomasses were linked to peak self-grooming decreases under M. anisopliae (T2) but not under E. phialicopiosa (T3). Downward curves persist in Fig. 10; however, the ant and fungus biomasses were markedly reduced with M. anisopliae in both chambers A and B. The mean and minimum self-grooming times showed the same pattern observed for maximum self-grooming, although they differed quantitatively. Ant biomass associated with self-grooming was significantly higher than the biomass associated with allogrooming; however, symbiotic fungal biomass did not follow this pattern (see Figs 9, 10).

Figure 9.

Bifurcation diagrams for q (the proportion of workers cultivating the symbiotic fungus) in T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa), considering minimum (red), mean (green), and maximum (blue) time spent on self-grooming.

Figure 10.

Bifurcation diagrams for q (the proportion of workers cultivating the symbiotic fungus) in T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa), considering minimum (red), mean (green), and maximum (blue) time spent on allogrooming.

A general decrease in the biomasses of ants and fungus in response to increasing mortality values occurred for both ant mortality (d_a) and fungus mortality (d_f), considering the minimum, mean, and maximum values of s_t and a_t (Figs 11–14). Although the two mortality patterns were generally similar, important differences appeared when analysing d_a and d_f separately, from a grooming perspective.

Figure 11.

Bifurcation diagrams for d_a (mortality rate of ants) in T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa), considering minimum (red), mean (green), and maximum (blue) time spent on self-grooming.

Figure 12.

Bifurcation diagrams for d_a (mortality rate of ants) in T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa), considering minimum (red), mean (green), and maximum (blue) time spent on allogrooming.

Figure 13.

Bifurcation diagrams for d_a (mortality rate of fungus) in T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa), considering minimum (red), mean (green), and maximum (blue) time spent on self-grooming.

Figure 14.

Bifurcation diagrams for d_f (total mortality of fungus) in T1 (control), T2 (Metarhizium anisopliae), and T3 (Escovopsis phialicopiosa), considering minimum (red), mean (green), and maximum (blue) time spent on allogrooming.

Ant biomass in chamber A decreased non-linearly under self-grooming in all three treatments, with quantitative differences among the three (Fig. 11). A pronounced decrease in symbiotic fungus biomass as a function of d_a occurred only above d_a = 0.5, indicating a delayed impact on fungus biomass. The minimum and mean self-grooming levels impacted both the ants and the fungus similarly. Maximum self-grooming provided slightly greater protection for the ant biomass compared to the other grooming levels but provided significantly more protection for the fungus than for the ants (Fig. 11). Results for allogrooming were similar to those for self-grooming, except in T2, where ant and fungal biomass curves overlapped (Fig. 12). Notably, the ant biomass was lower in the allogrooming treatment T2 (Fig. 12) compared to T2 in the self-grooming treatment (Fig. 11). The data in Fig. 13 indicate that there was no significant difference in the reduction of ant and fungal biomasses among the three treatments. The same overlap observed in Fig. 12 was also noted in allogrooming for T2 (Fig. 14).

The obligate symbiosis between leafcutter ants and basidiomycete fungi is based on a bidirectional dependency: the ants provide fresh plant substrate and protection to the symbiotic fungus, while the fungus serves as the exclusive nutritional source for the colony (Weber 1966; Mehdiabadi and Schultz 2010; Hölldobler and Wilson 2011). The stability of this mutualism is maintained through complex behavioural mechanisms, including dynamic worker allocation (Constantino et al. 2021) and individual (Currie and Stuart 2001) and collective hygiene strategies (Cremer et al. 2007). The introduction of pathogens disrupts this homeostasis, revealing critical resilience thresholds.

The simulations performed here were based on the model of Kang et al. (2011) but expanded to include the one-way chamber transfer parameter, demonstrating how variations in ant mortality (d_a) and fungal mortality (d_f) drastically alter the system biomass (Fig. 1B). The one-way transfer parameter (D) quantified biomass exchange of ants and the symbiotic fungus between nest chambers, reflecting ecological differences in nest architecture between Atta species with multi-chambered nests (Moreira et al. 2010) and Acromyrmex species, which typically have single-chambered nests (Forti et al. 2006).

For D > 0, one-way transfer redistributes resources (Fig. 1B), reducing biomass in the source chamber (A) whilst increasing it in chamber (B). This dynamic is particularly critical for Atta, where multi-chambered nests mitigate local stresses, such as pathogen outbreaks, by reallocating workers and fungal biomass. This is evidenced by delayed inflection points observed under D = 0.1 (Fig. 2). Future studies extending this model to interconnected multi-chamber networks could explore one-way transfer patterns, resilience, and epidemiological processes within nest topology, revealing how chamber complexity influences ecological stability.

The sigmoidal patterns observed in the biomasses align with classical models of obligate mutualism (Vandermeer and Boucher 1978). Introducing a second fungal chamber reveals critical specificities, with the reduced biomass in chamber A suggesting an energetic cost in resource reallocation. The maintenance of population stabilisation, even under one-way transfer, underscores the behavioural plasticity of ants in worker allocation (Gordon 1996), a key mechanism for mutualistic resilience. The experimental introduction of pathogens, such as M. anisopliae (T2) and E. phialicopiosa (T3), into the colonies had distinct results. Metarhizium anisopliae significantly reduced ant and symbiotic fungus biomasses, although the most marked differences among treatments arose from maximum self-grooming and allogrooming levels. Even at maximum levels of self-grooming and allogrooming, Metarhizium anisopliae infection was not eliminated.

The impact of M. anisopliae was evident even under moderate symbiotic fungus mortality (d_f ≈ 0.4–0.5, Figs 13, 14), directly compromising workers and the maintenance of the fungal garden. In contrast, E. phialicopiosa exhibited effects comparable to the control in most scenarios, with pronounced impacts only under high fungal mortality, at least examining d_f influenced by allogrooming (d_f > 0.5, Fig. 14). In treatment T2 (Metarhizium anisopliae), ant and fungal mortality curves overlapped among all allogrooming levels (max, mean, min; Figs 12, 14), indicating that social grooming fails to decouple the pathogen’s impact on both mutualistic partners. Although Escovopsis spp. are primarily pathogenic to the ants’ cultivated symbiotic fungus (typically Leucoagaricus or Leucocoprinus basidiomycetes) rather than to the ants themselves, their degradation of fungal gardens disrupts the colony’s food supply, threatening its survival. These simulations align with literature indicating that E. phialicopiosa acts opportunistically, exploiting the symbiotic fungus’ vulnerability under high mortality conditions (Jiménez-Gómez et al. 2021; Mendonça et al. 2021). Additionally, the limited impact of E. phialicopiosa may stem from the lower virulence of the strain (Montoya et al. 2023), which fails to trigger strong physiological or behavioural responses.

In contrast, the pronounced effects of M. anisopliae on both mutualistic components suggest a disruption to intra-colony labour allocation. As M. anisopliae infects insects via conidial adhesion to the cuticle and subsequent tegument penetration (Shah and Pell 2003), self-grooming alone may be insufficient to prevent infection despite reducing spore quantity and viability (Hughes et al. 2002). Workers may thus prioritise alternative defences, such as increasing the frequency of self-grooming over its duration, cleaning the fungal garden (Goes et al. 2022), or activating physiological defences integrated with behavioural responses (Fernández-Marín et al. 2006), which were not quantified in this study.

Infection may also impair worker activity, as evidenced by delayed ant and fungal biomass stabilisation under maximum self-grooming. Studies have shown that M. anisopliae can overwhelm workers and disrupt essential activities such as foraging (Jaccoud et al. 1999), suggesting that infection-induced destabilisation of social homeostasis is a key factor in colony decline. These results align with empirical studies reporting similar responses in leaf-cutting ants challenged with Beauveria bassiana and Trichoderma harzianum (Mota-Filho et al. 2021).

Whilst the factors regulating grooming intensity in leaf-cutting ants are not fully understood, they are generally considered multifaceted (Goes et al. 2022). Several hypotheses could be proposed to explain this variation, including behavioural effort linked to pathogen prevalence and infection risks, duration of exposure to pathogens or contaminants, age polyethism, colony-level genetic diversity, queen pheromone-mediated modulation, and environmental stressors such as temperature and humidity. However, it remains to be answered whether and how much such factors contribute to regulating the grooming intensity in leaf-cutting ants.

Our simulations indicate that hygiene behaviours differentially affect the system biomass. Self-grooming proved more effective against M. anisopliae than allogrooming, reflecting the prioritisation of individual defences under high pathogen pressure (Figs 2 T2, 5 T2). This aligns with observations that ants modulate allogrooming in the presence of virulent pathogens to minimise horizontal spore transmission (Theis et al. 2015; Konrad et al. 2018). The lack of impact of allogrooming on E. phialicopiosa likely stems from the pathogen’s lower virulence (Fig. 5 T1, T3). However, under high fungal mortality (d_f > 0.5), allogrooming may accelerate system collapse due to secondary spore transmission and opportunistic behaviour (Jiménez-Gómez et al. 2021).

The relationship between ant mortality (d_a) and population collapse revealed a critical threshold (d_a ≈ 0.5) beyond which the mutualism loses resilience. For M. anisopliae, high worker mortality directly impaired fungal care, accelerating the decline in fungal biomass (d_f > 0.4) even under maximum self-grooming, demonstrating a critical threshold for mutualistic collapse (Fig. 13). The fungal decline is not due solely to reduced care but also to the collapse of symbiotic interactions (Mueller et al. 2005; Dejean et al. 2023). In contrast, E. phialicopiosa (T3) exhibited a threshold-dependent response to fungal mortality, with pronounced effects only when the symbiont’s baseline mortality exceeded df > 0.5. Under self-grooming, T3 and T1 were indistinguishable at low-to-moderate mortality levels (Fig. 13).

Bifurcation diagrams revealed critical system transitions (Figs 9, 10). Under low biomass (q → 0 or q → 1), colonies had to choose between prioritising fungal care (q → 1) or abandoning it (q → 0). Intermediate values (q = 0.4 to 0.6) maximised stability by balancing essential functions. The parameter q represents the proportion of workers allocated to fungus cultivation. Therefore, when biomass is low and q approaches zero, few workers are available to cultivate the symbiotic fungus, resulting in reduced biomass levels. Conversely, when q approaches 1, workers are fully dedicated to fungus cultivation, focusing exclusively on this task. Consequently, other tasks may be neglected, to the detriment of overall colony function, in favour of maximising fungus growth.

Delayed inflection points in biomass curves influenced by M. anisopliae likely arise from high pathogen intensity (k_a), which increases worker mortality (d_a), impairs fungal care, and initially prioritises self-grooming over allogrooming (Figs 2 T2, 5 T2). This creates a lag between pathogen exposure and mutualistic collapse. The critical threshold (d_a ≈ 0.5) marks the point at which worker mortality exceeds the colony’s replacement capacity (r_a), accelerating fungal decline (Fig. 11). Reduced fungal care (q < 0.25) explains the loss of resilience beyond this threshold (Figs 10, 11).

These results provide critical insights into pathogen transmission dynamics within leaf-cutting colonies and the factors shaping the resilience of ant-fungus mutualism. Beyond ecological understanding, these findings could inform practical strategies for managing Atta spp., which are significant agricultural pests.

The results highlight the critical role of multi-chamber nest architecture and hygienic practices in mitigating pathogen impacts. Metarhizium anisopliae was identified as a more severe threat to colony stability compared to Escovopsis phialicopiosa, significantly reducing both ant and fungal biomasses. Self-grooming proved more effective than allogrooming in countering M. anisopliae, suggesting that individual defences are prioritised under high pathogen pressure. The study also revealed critical thresholds for ant and fungal mortality, beyond which the mutualistic system collapses. Worker allocation to fungus cultivation emerged as a key factor in maintaining stability, with intermediate proportions optimising resilience. These findings enhanced our understanding of the ecological and behavioural strategies that support the resilience of ant-fungus mutualisms and provided insights for developing targeted biological control strategies against leaf-cutting ants.

We express our sincere gratitude to the funding body CNPq and grant #2023/08286-2, São Paulo Research Foundation (FAPESP), for their support and contribution to this research. We also thank Dr Ricardo Toshio Fujihara for kindly providing the ant colonies used in this study. We are grateful to Dr Janet Reid for proofreading this article in English. We are also grateful to two anonymous reviewers who contributed significantly to improving the manuscript.